Week 0 · Module 1
Where is your English actually weak?
An 8-question rapid triage that diagnoses the 4 core English habits costing you manuscript acceptances and interviews — and routes you directly to your study units
Your scenario
You send your CV and a writing sample to a medical communications agency in Hyderabad. Two weeks pass. No reply. Your science is sound, your therapeutic knowledge is better than most of the people who did get called — but your covering email opened with “Respected Sir, kindly do the needful and revert back at the earliest.” and your sample had eleven article errors in the first paragraph. The reviewer stopped reading at line three. This module finds out, in twenty minutes, exactly which errors you are making.
Read this before you start
This is not an English test. It does not produce a grade and it is not a judgement of your intelligence or your science. It is a diagnostic instrument — the same logic as a lab panel. It tells you which of twelve specific error classes you are producing, so that you spend your study time on the three or four that are actually holding you back instead of re-reading grammar rules you already follow correctly.
Almost every medical writer trained in India makes errors from a small, highly predictable set. There are twelve classes. Most people have serious problems in three or four of them and are fine in the rest. Studying all twelve is a waste of your time.
How to take it honestly
Do not look anything up. Do not ask anyone. A wrong answer here is worth more to you than a right one, because it is the only way the diagnostic can find the gap. If you cheat this test you will study the wrong units for six weeks.
—
—
Your twelve areas — study the red ones first
Before you move on
Write down the three weakest areas from your result, in order. Come back to this box after you finish those three units and note what changed.
People who write this down finish the module at roughly twice the rate of people who do not.
Next module
0.2 — Articles: a, an, the, and nothing
Week 0 · Module 2
Articles: a, an, the, and nothing
The single largest error class in Indian scientific English — and the one that never clears by itself, because your first language gave you nothing to transfer
Your scenario
A senior editor returns your first manuscript. The science is untouched. Every second line has a small blue mark. You count them: thirty-one corrections, twenty-two of them articles. She has not written a single comment, because there is nothing to explain — each one is a separate instance of the same missing instinct. You have made this error since school and no one has ever told you why.
Why you make this error
Telugu, Tamil, Kannada, Malayalam, Hindi, Bengali, Marathi and Gujarati have no article system. Definiteness is carried by word order, case marking, or simply by context. When you write English, there is nothing in your first language to transfer — you are not being careless, you are working without a template.
This is why article errors survive long after every other error in your writing has been fixed, and why reading more English does not repair them. Reading builds recognition. Articles are a production skill. Only drilling fixes them.
This is why article errors survive long after every other error in your writing has been fixed, and why reading more English does not repair them. Reading builds recognition. Articles are a production skill. Only drilling fixes them.
The three-question test
Forget the grammar-book rules. Every article decision in scientific writing can be resolved by asking three questions in order. Run them on the noun, not on the sentence.
| Ask | If yes | Worked example |
|---|---|---|
| 1. Can I count this noun? Can I say “two of them”? | If no → use no article. Stop here. | Evidence supports this approach. Not an evidence, not evidences. |
| 2. Am I talking about the whole class in general, using a plural? | If yes → use no article. Stop here. | Patients with diabetes have a higher risk. Diabetic patients everywhere, not one cohort. |
| 3. Does the reader already know exactly which one I mean? | Yes → the. No → a / an. | We enrolled a patient with EGFR mutation. The patient received osimertinib. |
The reader-already-knows rule, in detail
The reader knows which one you mean when: you have already mentioned it; you define it in the same phrase (the patients enrolled in this study); there is only one in the world (the World Health Organization, the liver); or it is a superlative or ordinal (the primary endpoint, the highest dose). Everything else on first mention takes a or an.
A or an — listen, do not read
The choice follows the sound of the next word, not its spelling. This matters constantly in medical writing because abbreviations are everywhere and their spelling lies about their sound.
Wrong
Right
a MRI scanM is a consonant letter
an MRI scanbut it is said “em” — a vowel sound
an one-year follow-upO is a vowel letter
a one-year follow-upbut it is said “wun” — a consonant sound
a hourH is a consonant letter
an hourthe h is silent
an SAE report / a IV line
an SAE report / an IV line“ess-ay-ee”, “eye-vee”
an unit of insulin
a unit of insulinsaid “yoo-nit”
The four traps in clinical text
1 — Generic plurals
2 — First vs later mention
3 — Uncountables
4 — Defined by the phrase
The trap
Adding the to a plural noun that means the whole class. This is the highest-frequency article error in Indian manuscripts.
Wrong
“The immune checkpoint inhibitors have transformed the treatment of the melanoma.”
Right
“Immune checkpoint inhibitors have transformed the treatment of melanoma.”
Checkpoint inhibitors as a drug class takes no article. Melanoma as a disease takes no article. But the treatment keeps its the — because “of melanoma” tells the reader exactly which treatment is meant.
The trap
Using the on first mention, before the reader has any way of knowing which one you mean.
Wrong
“We report the case of the 62-year-old man with the rare urothelial carcinoma. The patient received velparimab.”
Right
“We report the case of a 62-year-old man with a rare urothelial carcinoma. The patient received velparimab.”
First mention introduces — a man, a carcinoma. Second mention refers back — the patient. Note that “the case of” is correct, because “of a 62-year-old man” identifies it. Track this pattern across a whole paragraph and it becomes automatic.
The trap
Treating an uncountable noun as countable — the same instinct that produces informations and equipments.
Wrong
“A recent research provides an evidence that the therapy is effective. We reviewed the literatures.”
Right
“Recent research provides evidence that the therapy is effective. We reviewed the literature.”
Research, evidence, literature, information, equipment, advice, feedback, training, personnel and staff are all uncountable. If you need a countable version, add a counting word: a piece of evidence, a body of research, a line of therapy, two studies.
The trap
Dropping the when a following phrase has already pinned the noun down to one specific thing.
Wrong
“Primary endpoint of study was overall survival. Median duration of follow-up was 24 months.”
Right
“The primary endpoint of the study was overall survival. The median duration of follow-up was 24 months.”
“Of the study” identifies the endpoint. “Median” is effectively a superlative — there is only one. Whenever a noun is followed by of plus something specific, or preceded by primary, first, highest, median or overall, it almost always takes the.
Recognition drills
Drill 1 of 3
Which version is correct?
A“Biomarkers are increasingly used to select patients for immunotherapy in the CASTELLAN cohort.”
B“The biomarkers are increasingly used to select patients for immunotherapy in CASTELLAN cohort.”
C“Biomarkers are increasingly used to select patients for immunotherapy in CASTELLAN cohort.”
D“The biomarkers are increasingly used to select the patients for immunotherapy in the CASTELLAN cohort.”
Drill 2 of 3
Complete: “Each participant underwent ___ ECG at baseline and ___ 24-hour Holter recording at week 4.”
Aa / a
Ban / a
Can / an
Da / an
Drill 3 of 3
A Methods section reads: “Blood samples were collected in ___ EDTA tube and stored at −80 °C until ___ analysis.”
Aan / the
Ba / the
Can / no article
Da / no article
Production drill — this is where the learning happens
Rewrite the paragraph below, fixing every article error. There are nine.
“The lung cancer is leading cause of the cancer death worldwide. In present study, we evaluated efficacy of the pembrolizumab in patients with advanced disease. A total of 312 patients were enrolled. The primary endpoint was overall survival. We collected the informations on adverse events at each visit. A median follow-up was 18 months. The results suggest that pembrolizumab may provide the survival benefit in this population.”
Corrected: “Lung cancer is the leading cause of cancer death worldwide. In the present study, we evaluated the efficacy of pembrolizumab in patients with advanced disease. A total of 312 patients were enrolled. The primary endpoint was overall survival. We collected information on adverse events at each visit. The median follow-up was 18 months. The results suggest that pembrolizumab may provide a survival benefit in this population.”
The nine fixes:
1. The lung cancer → Lung cancer — a disease named in general takes no article.
2. is leading cause → is the leading cause — superlative, only one.
3. the cancer death → cancer death — generic, uncountable.
4. In present study → In the present study — present identifies it.
5. evaluated efficacy → evaluated the efficacy — “of pembrolizumab” pins it down.
6. the pembrolizumab → pembrolizumab — drug names never take an article.
7. the informations → information — uncountable, no article, no s.
8. A median follow-up → The median follow-up — median means there is only one.
9. the survival benefit → a survival benefit — first mention, unspecified, and hedged.
Score yourself: 8–9 correct means you have the rule. 5–7 means recognition without production — repeat this drill in three days. Below 5 means do not move on yet.
The nine fixes:
1. The lung cancer → Lung cancer — a disease named in general takes no article.
2. is leading cause → is the leading cause — superlative, only one.
3. the cancer death → cancer death — generic, uncountable.
4. In present study → In the present study — present identifies it.
5. evaluated efficacy → evaluated the efficacy — “of pembrolizumab” pins it down.
6. the pembrolizumab → pembrolizumab — drug names never take an article.
7. the informations → information — uncountable, no article, no s.
8. A median follow-up → The median follow-up — median means there is only one.
9. the survival benefit → a survival benefit — first mention, unspecified, and hedged.
Score yourself: 8–9 correct means you have the rule. 5–7 means recognition without production — repeat this drill in three days. Below 5 means do not move on yet.
Reflection
Open something you wrote in the last month — an email, a CV, a manuscript draft. Read only the first paragraph and mark every article decision. How many would you now change?
Auditing your own old text is more effective than any exercise, because these are the errors your hand actually produces.
Next module
0.3 — Tense: the IMRAD tense map
Week 0 · Module 3
Tense: the IMRAD tense map
Scientific English uses tense to signal how certain a statement is. Get it wrong and you accidentally claim more — or less — than your data allow
Your scenario
You submit a manuscript. The reviewer writes: “The authors appear to present their own preliminary finding as established fact.” You never claimed that. But in the Discussion you wrote “This biomarker predicts response to therapy” in the simple present — the tense English reserves for settled knowledge. You meant your study observed it once. The tense made the claim, not you.
Why you make this error
Indian languages mark tense reliably, so this is not a gap in your grammar — it is a gap in convention. English scientific writing loads extra meaning onto tense that everyday English does not: past means we observed this once, present means this is known to be true, present perfect means the field has established this. Nobody teaches this, because native speakers absorb it by reading hundreds of papers.
The second problem is transfer: Hindi and Dravidian languages allow the progressive with verbs of state, which is why “I am having a doubt” and “We are knowing the result” feel natural to you and mark you instantly to an international reader.
The second problem is transfer: Hindi and Dravidian languages allow the progressive with verbs of state, which is why “I am having a doubt” and “We are knowing the result” feel natural to you and mark you instantly to an international reader.
The map — nine decisions, in order
Print this table. Keep it beside you for the next ten documents you write. After that you will not need it.
| Where | What you are saying | Tense | Example |
|---|---|---|---|
| Introduction | Established background knowledge | present | Lung cancer is the leading cause of cancer death. |
| Introduction | What the field has found so far | present perfect | Several trials have shown a survival benefit. |
| Introduction | What one named study did | past | Paik et al. reported a response rate of 46%. |
| Introduction | Your objective | past | We aimed to assess efficacy in this subgroup. |
| Methods | What you did, once, in the past | past | Patients were randomised in a 1:1 ratio. |
| Results | What you found | past | Median progression-free survival was 8.2 months. |
| Results | What a table or figure displays | present | Table 2 shows baseline characteristics. |
| Discussion | Restating your own finding | past | In this study, response was higher in the biomarker-positive group. |
| Discussion | Interpreting it, or drawing an implication | present | These findings suggest that biomarker testing may guide selection. |
The one that gets manuscripts rejected
The last two rows are the whole game. Your finding stays in the past. Your interpretation moves to the present — but must be hedged. Writing “This biomarker predicts response” puts your single result into the tense of established fact, and a reviewer will read that as overclaiming. Write “In this cohort, biomarker positivity was associated with response” and then “These data suggest that the biomarker may predict response.”
The two habits to break
Wrong
Right
I am having a doubt about the endpoint.have is a state, not an action
I have a question about the endpoint.stative verbs never take the progressive
We are knowing the results already.
We already know the results.know, believe, understand, contain, consist, belong, need, prefer, seem
The sample is containing three analytes.
The sample contains three analytes.
We have performed the assay in March 2025.present perfect cannot take a finished time
We performed the assay in March 2025.a named past time forces the simple past
The study has been completed last year.
The study was completed last year.
Since 2019, we published four papers.
Since 2019, we have published four papers.since opens a window still running — present perfect
Recognition drills
Drill 1 of 2
You are writing the Discussion. Your trial found a numerically higher response rate in one subgroup, but it was not powered for that comparison. Which sentence is correct?
A“MET exon 14 alteration predicts a higher response rate to tepotinib.”
B“MET exon 14 alteration has predicted a higher response rate to tepotinib.”
C“MET exon 14 alteration is predicting a higher response rate to tepotinib.”
D“Response rate was numerically higher in this subgroup; these data suggest that alteration status may influence response.”
Drill 2 of 2
Which sentence correctly opens an Introduction paragraph?
A“Urothelial carcinoma is the second most common skin cancer. Several trials have evaluated PD-1 blockade in advanced disease.”
B“Urothelial carcinoma was the second most common skin cancer.”
C“Urothelial carcinoma has been the second most common skin cancer. Several trials evaluated PD-1 blockade.”
D“Urothelial carcinoma is being the second most common skin cancer.”
Production drill
Fix the tense in every sentence. There are seven errors.
“Diabetes was a major global health burden. Several studies demonstrated an association between glycaemic control and cardiovascular outcomes. In the present study, we are aiming to evaluate a novel GLP-1 agonist. Patients are randomised in a 2:1 ratio between March 2024 and August 2024. Median HbA1c reduction has been 1.4% at 24 weeks. Table 3 showed the full safety profile. These findings are suggesting that the agent improves glycaemic control.”
Corrected: “Diabetes is a major global health burden. Several studies have demonstrated an association between glycaemic control and cardiovascular outcomes. In the present study, we aimed to evaluate a novel GLP-1 agonist. Patients were randomised in a 2:1 ratio between March 2024 and August 2024. Median HbA1c reduction was 1.4% at 24 weeks. Table 3 shows the full safety profile. These findings suggest that the agent may improve glycaemic control.”
The seven fixes:
1. was → is — standing background fact, simple present.
2. demonstrated → have demonstrated — accumulated body of literature, present perfect.
3. are aiming → aimed — your objective is a past decision.
4. are randomised → were randomised — Methods, and a named past window forbids anything else.
5. has been → was — a Result you observed, finished, simple past.
6. showed → shows — the table displays it now, every time it is read.
7. are suggesting → suggest — stative verb, no progressive. And note the added may: interpretation in the present tense must be hedged, which is Unit 9.
Score yourself: 6–7 correct means the map is working. Below 5, print the table and write your next three documents with it open.
The seven fixes:
1. was → is — standing background fact, simple present.
2. demonstrated → have demonstrated — accumulated body of literature, present perfect.
3. are aiming → aimed — your objective is a past decision.
4. are randomised → were randomised — Methods, and a named past window forbids anything else.
5. has been → was — a Result you observed, finished, simple past.
6. showed → shows — the table displays it now, every time it is read.
7. are suggesting → suggest — stative verb, no progressive. And note the added may: interpretation in the present tense must be hedged, which is Unit 9.
Score yourself: 6–7 correct means the map is working. Below 5, print the table and write your next three documents with it open.
Reflection
Take any abstract from a journal in your therapy area. Label the tense of every sentence and match it against the nine rows of the map. Where does the published abstract deviate, and can you work out why?
Deviations are usually deliberate. Finding out why is how you move from following the rule to controlling it.
Next module
0.4 — Register: hedging, claims and Indianisms
Week 0 · Module 4
Register: hedging, claims and Indianisms
The unit that separates a competent writer from an employable one — calibrating how strongly you claim, and removing the administrative English that marks your drafts on sight
Your scenario
Two candidates apply for the same medical writer role. Both have an M.Pharm. Both know oncology. Candidate one writes: “Kindly find attached my CV. Please do the needful and revert back at the earliest.” Candidate two writes: “I have attached my CV and a short writing sample. I would be glad to discuss the role at your convenience.” Only one gets a call. Nothing in candidate one's message is grammatically wrong.
Part one — the claim ladder
In scientific and regulatory writing, the verb you choose is a statement about your study design. It is not decoration and it is not politeness. Choosing a verb your evidence cannot support is, in a promotional or regulatory document, a compliance failure — the kind that gets copy rejected in medical-legal-regulatory review or triggers a warning letter.
Learn the ladder in both directions: what each rung means, and what design licenses it.
| Verb | What it claims | What you need to say it |
|---|---|---|
proves / confirms | Settled. No further evidence needed. | Almost never appropriate for a single study. Reserve for mathematical proof or overwhelming consensus. |
demonstrates | This study establishes it. | An adequately powered, pre-specified primary endpoint that was met in a randomised controlled trial. |
shows | The data display it directly. | A directly observed result. Safe for describing figures and tables. |
indicates | The data point this way. | A consistent result, possibly secondary or from a well-conducted observational study. |
suggests | Compatible with this reading. | Exploratory analyses, subgroups, small samples, hypothesis-generating work. |
may suggest | One possible reading among several. | Underpowered, post hoc, or where confounding is plausible and unaddressed. |
Association is not causation, and the language must say so
Only a randomised design licenses causal verbs. Everything else takes association language.
Causal — needs randomisation: reduced, caused, led to, resulted in, improved, prevented.
Associational — safe for observational data: was associated with, correlated with, was accompanied by, occurred more frequently in, was observed alongside.
Writing “the drug reduced mortality” from a retrospective cohort is not a stylistic slip. It is a claim your data cannot support, and in promotional copy it is the single most common reason for rejection at review.
Causal — needs randomisation: reduced, caused, led to, resulted in, improved, prevented.
Associational — safe for observational data: was associated with, correlated with, was accompanied by, occurred more frequently in, was observed alongside.
Writing “the drug reduced mortality” from a retrospective cohort is not a stylistic slip. It is a claim your data cannot support, and in promotional copy it is the single most common reason for rejection at review.
Drill 1 of 2
A single-arm Phase II study of 42 patients met its primary endpoint of overall response rate. You are drafting the conclusion of the abstract. Which verb is defensible?
A“These results demonstrate that the agent is superior to standard care.”
B“These results show promising activity; randomised evaluation is warranted.”
C“These results prove the efficacy of the agent.”
D“These results may suggest possible activity of the agent.”
Part two — the kill list
The phrases below are correct English. Every one of them is understood in India and almost nowhere else in professional scientific communication. They appear in Indian office correspondence, and international editors, recruiters and clients read them as a signal about the writer — usually before they have read anything else.
You do not need to feel anything about this. Just delete them.
Delete
Write instead
Kindly do the needful
Please review and approvesay what the action actually is
Please revert back
Please replyrevert means return to a former state; back is redundant anyway
Please find attached the same
I have attached the protocolthe same as a pronoun is legal English from 1890
at the earliest
by Friday, 14 Marchgive a date, not a feeling
Respected Sir / Dear Sir
Dear Dr Anand / Dear Editoruse the name; Sir alone reads as junior and dated
I have a doubt
I have a questiona doubt means you suspect it is false
prepone the meeting
move the meeting earlierprepone does not exist outside South Asia
discuss about / emphasise on
discuss / emphasisethe verb already contains it
It is a matter of great pleasure to present this humble work
This study evaluated…ceremonial openings signal an unedited draft
The above said / the aforementioned study only
This studyonly as an emphasis particle is a direct calque from Indian languages
One thing to be clear about
Indian English is a legitimate variety with millions of fluent speakers, and there is nothing deficient about it. This unit is not about correcting who you are. It is a narrow, practical point: international scientific publishing and pharma communications run on a particular house register, and matching it is a professional skill, exactly like learning ICH formatting. Use the register the document requires, and use whatever you like everywhere else.
Drill 2 of 2
You are emailing a journal editor to ask about a manuscript submitted six weeks ago. Which version is appropriate?
A“Respected Sir, kindly update on the status of the same and revert at the earliest.”
B“Sorry for disturbing you. Kindly do the needful regarding my humble submission.”
C“Dear Dr Mehta, I am writing about manuscript EJC-2026-0412, submitted on 3 June. Could you let me know its current status?”
D“Status of my manuscript. Please respond immediately.”
Production drill — the one that gets you hired
Rewrite this Discussion paragraph. Fix every overclaim and every register error.
“It is a matter of great pleasure to present these findings. Our retrospective analysis of 118 patients proves that early initiation of the therapy reduces mortality and definitely improves quality of life. The above said results confirm that this approach should be adopted as standard of care. We are having a doubt only regarding the generalisability, and kindly request the readers to revert with their views on the same.”
Model rewrite: “In this retrospective analysis of 118 patients, earlier initiation of therapy was associated with lower mortality and with higher patient-reported quality-of-life scores. These findings suggest that the approach may warrant prospective evaluation. Because the analysis was retrospective and unadjusted for treatment-selection bias, its generalisability is uncertain, and confirmation in a randomised setting is required before any change in practice.”
What changed and why:
1. Ceremonial opening deleted — it carries no information and signals an unedited draft.
2. proves → was associated with. Retrospective design licenses association language only. This is the single most important fix in the paragraph.
3. reduces mortality → was associated with lower mortality. Causal verb removed.
4. definitely improves → deleted. Definitely has no place in a scientific claim, and quality of life became the measurable thing that was actually recorded.
5. confirm that this should be standard of care → suggest it may warrant prospective evaluation. A retrospective series cannot change standard of care and should not ask to.
6. The above said → These.
7. We are having a doubt only regarding → the limitation is now stated properly, with its reason. Naming the specific bias is what a reviewer is looking for.
8. kindly request the readers to revert on the same → deleted. A Discussion does not address readers directly.
Score yourself: the two that matter most are 2 and 5. If you removed the ceremony but left proves standing, you fixed the cosmetic problem and left the compliance problem — go back over the claim ladder before moving on.
What changed and why:
1. Ceremonial opening deleted — it carries no information and signals an unedited draft.
2. proves → was associated with. Retrospective design licenses association language only. This is the single most important fix in the paragraph.
3. reduces mortality → was associated with lower mortality. Causal verb removed.
4. definitely improves → deleted. Definitely has no place in a scientific claim, and quality of life became the measurable thing that was actually recorded.
5. confirm that this should be standard of care → suggest it may warrant prospective evaluation. A retrospective series cannot change standard of care and should not ask to.
6. The above said → These.
7. We are having a doubt only regarding → the limitation is now stated properly, with its reason. Naming the specific bias is what a reviewer is looking for.
8. kindly request the readers to revert on the same → deleted. A Discussion does not address readers directly.
Score yourself: the two that matter most are 2 and 5. If you removed the ceremony but left proves standing, you fixed the cosmetic problem and left the compliance problem — go back over the claim ladder before moving on.
Reflection
Open your sent-mail folder. Find the last five professional emails you wrote. Count the kill-list phrases. Then rewrite the one that matters most — the one to a client, editor or recruiter.
This is the single highest-return exercise in Week 0. Your email register is what people judge before they ever see your writing.
Week 0 status
You have completed the diagnostic and three of twelve units — articles, tense, and register. Units 3 to 8 and 10 to 12 (agreement, countables, prepositions, modifiers, punctuation, concision, parallelism and scientific style conventions) are in development. If your diagnostic flagged those areas red, note them and return when the units are released.
If your diagnostic put you in Band A or B, you are ready for Week 1 now.
If your diagnostic put you in Band A or B, you are ready for Week 1 now.
Next module
1.1 — The day a drug is born
Week 1 · Module 1
The day a drug is born
From a molecule in a laboratory to a medicine in a patient's hand — the journey that creates the industry you are entering
Your scenario
It is Monday morning in Hyderabad. You have just joined a CRO as a junior medical writer. Your manager drops a file on your desk — "Clinical Study Report — Phase III." She says: "Familiarise yourself with the study background. We present to the sponsor on Friday." You open it. Page 1 reads: "This study was conducted in accordance with ICH-GCP guidelines, following successful completion of Phase I and Phase II trials." You nod — but quietly wonder: what exactly happened in Phase I and Phase II? This module answers that question.
The journey of a single drug
Before a medicine reaches a patient, it travels through one of the most carefully regulated processes in human history. On average it takes 10 to 15 years and costs upward of Rs. 8,000 crore. Most candidates fail. The ones that survive carry the weight of every failure before them.
As a medical writer, you will work somewhere inside this journey — writing documents that describe what happened, how it happened, and what it means. To write those documents well, you must first understand the journey itself.
Why this matters for you as a writer
Every document you will ever write describes a specific stage of this pipeline. When you know which stage you are writing about, you understand what question the document is answering and who is reading it. That understanding is what separates a good medical writer from a great one.
Explore the pipeline — click each stage
Stage 1
Discovery
2–5 yrs
Stage 2
Preclinical
1–3 yrs
Stage 3
Phase I
1–2 yrs
Stage 4
Phase II
2–3 yrs
Stage 5
Phase III
3–5 yrs
Stage 6
Approval
1–2 yrs
Stage 7
Phase IV
Ongoing
Drug Discovery — finding the molecule
Scientists identify a disease target and screen thousands of chemical compounds. Most are discarded. A handful become "lead candidates." No humans are involved yet — this work is entirely laboratory-based.
Your role as a writer: Almost none. Occasionally internal scientific summaries or literature reviews for research teams.
Preclinical — testing in cells and animals
The lead candidate is tested in cell cultures then animals for basic safety and pharmacokinetics (ADME). If it passes, the sponsor files an IND application with CDSCO (India) or FDA (US) for permission to test in humans.
Your role as a writer: Preclinical summary reports, background sections for regulatory submissions, literature reviews supporting the IND application.
Phase I — first in human
20–80 healthy volunteers receive the drug for the first time. Primary question: safety. What dose is tolerable? How does the body absorb, distribute, metabolise and excrete it? These trials happen at specialised Phase I units in teaching hospitals.
Your role as a writer: Phase I CSRs, informed consent documents, investigator brochures, protocol summaries. This is where many medical writing careers begin.
Phase II — does it work?
100–300 patients with the target disease receive the drug. Now efficacy enters: Does it actually treat the condition? What is the optimal dose? Roughly 70% of drugs entering Phase II do not advance to Phase III — this is where most candidates fail.
Your role as a writer: Phase II CSRs, congress abstracts, interim analysis reports, protocol amendments.
Phase III — the pivotal trial
Thousands of patients across multiple countries receive the drug — usually compared against placebo or standard of care. This is the main event. The CASTELLAN study (velparimab, NCT0XXXXXXX) referenced throughout this course was a Phase III-equivalent study run by the sponsor.
Your role as a writer: Full CSRs, peer-reviewed manuscripts, congress presentations, MSL slide decks, regulatory dossiers. Senior writers at Sanofi and Novartis spend most of their careers here.
Regulatory Approval — the submission
The sponsor compiles the Common Technical Document (CTD) submitted to CDSCO, FDA or EMA. This enormous package summarises every study conducted. Regulators review it and decide: approve, reject, or request more data.
Your role as a writer: CTD Module 2 summaries — Clinical Overview, Clinical Summary, integrated summaries of efficacy and safety. Regulatory medical writing is the highest-paid subspecialty.
Phase IV — life after approval
The drug is on the market, but science continues. Post-marketing surveillance tracks long-term real-world safety. Pharmacovigilance teams monitor adverse events. New indications are explored. Comparative effectiveness studies compare against competitors.
Your role as a writer: PSUR/PBRER periodic safety reports, real-world evidence publications, post-marketing CSRs, medical education content. Phase IV work is often done by experienced freelancers.
Something to sit with
Of roughly 10,000 compounds entering drug discovery, only 1 reaches the market. Every approved medicine carries the invisible weight of 9,999 failures. When you write about a drug that made it — you are writing about something genuinely rare.
Key terms
Clinical Study Report (CSR)
Comprehensive document describing design, methods, results and conclusions of a clinical trial. Required for all regulatory submissions. Follows ICH E3 guideline.
You write your first CSR Methods section as Portfolio Piece 3 in Week 3.
IND — Investigational New Drug
Application to CDSCO or FDA requesting permission to test a new drug in humans for the first time. Must be approved before any Phase I trial begins.
Medical writers contribute literature reviews and background summaries to IND packages.
Common Technical Document (CTD)
Internationally standardised format for regulatory submissions. Five modules. Module 2 summaries are written by medical writers — clinical overview and clinical summaries.
Module 3.3 covers CTD structure in full detail.
Pharmacokinetics (PK)
How the body absorbs, distributes, metabolises and excretes a drug — summarised as ADME. Heavily reported in Phase I CSRs. Key parameters: Cmax, Tmax, AUC, t1/2.
"PK parameters were assessed following single and multiple oral doses."
CDSCO
Central Drugs Standard Control Organisation — India's national regulatory authority. The Indian equivalent of the US FDA. All clinical trials in India require CDSCO approval.
CDSCO Schedule Y governs clinical trial conduct in India.
Sponsor
Organisation responsible for initiating, managing and financing a clinical trial. Usually a pharmaceutical company that owns the resulting data and documents.
CASTELLAN study sponsor: Meridian Therapeutics.
Knowledge check
A Phase III CSR lands on your desk. The study compared Drug A against placebo in 2,400 patients. What is the most likely primary purpose of this trial?
ATo assess whether the drug is safe to give to humans for the first time
BTo determine the optimal dose of the drug
CTo monitor long-term safety after the drug reaches the market
DTo demonstrate efficacy in a large patient population for regulatory approval
Reflection
Think of any medicine you have dispensed or taken personally. Which stage of the pipeline do you think was most critical in determining whether it reached you — and why?
Your notes save automatically in your browser.
Coming next — Module 1.2
Now that you know what happens across the pipeline, Module 1.2 maps the people inside it. Who is the CRA? What does an MSL actually do all day? Where exactly does a medical writer sit and who do they report to?
Next module
1.2 — Who does what in pharma
Week 1 · Module 2
Who does what in pharma
The people inside the pipeline — their roles, their documents, and where you fit among them
Your scenario
Your manager introduces you to the team on your first day. Around the table: a CRA, an MSL, a regulatory affairs officer, a biostatistician, and a pharmacovigilance associate. Everyone uses acronyms you have never heard. By the end of this module, you will know exactly what each person does — and how your work connects to theirs every day.
The six key roles — explore each one
Medical Writer
CRA
MSL
Regulatory Affairs
Biostatistician
PV Associate
Full title
Medical Writer / Scientific Writer / Regulatory Medical Writer
What they do daily
Write, edit and review clinical documents — CSRs, abstracts, manuscripts, slide decks, regulatory submissions, patient leaflets. Translate complex clinical data into clear, structured, compliant documents for specific audiences.
Key skills
Scientific accuracy · Regulatory knowledge (ICH E3, CTD) · Plain language · Source data verification · Attention to detail
Salary range India 2025
Rs. 3.5–6 LPA junior · Rs. 8–14 LPA senior · Rs. 18–28 LPA principal or lead
This is your target role. Everything in this course is designed to make you hireable at junior level and credible at senior level.
Full title
Clinical Research Associate (CRA) / Clinical Monitor
What they do daily
Visit clinical trial sites to verify the trial is conducted correctly — patient rights protected, data accurate, protocol followed. They are the sponsor's eyes and ears at the investigator site.
Key skills
Site management · ICH-GCP compliance · Source data verification · Protocol knowledge · Communication with investigators
Salary range India 2025
Rs. 3–5 LPA entry · Rs. 7–12 LPA senior CRA · Frequent travel required
CRA to medical writer is one of the most common career transitions. CRAs understand the trial inside out — which makes them excellent writers. Many IMRAD Advanced Track students make this move.
Full title
Medical Science Liaison (MSL)
What they do daily
Meet Key Opinion Leaders (KOLs) — senior doctors and researchers — to discuss scientific evidence about the company's drugs. Not sales. Scientific relationship building. They use MSL slide decks which medical writers produce.
Key skills
Deep therapeutic area expertise · Scientific communication · KOL relationship management · Clinical trial knowledge
Salary range India 2025
Rs. 8–14 LPA · Usually requires clinical degree or PhD
In Week 3 you create an MSL slide deck. Understanding what an MSL does and who they present to is essential for writing a deck they can actually use in the field.
Full title
Regulatory Affairs Officer / Manager
What they do daily
Manage the drug's regulatory lifecycle — preparing and submitting dossiers to CDSCO, FDA or EMA. Interpret regulatory guidelines and ensure company compliance. Work closely with medical writers on CTD Module 2 submissions.
Key skills
ICH guideline knowledge · CTD structure · CDSCO/FDA/EMA requirements · Submission management · Regulatory strategy
Salary range India 2025
Rs. 4–7 LPA entry · Rs. 12–20 LPA senior · Highly specialised path
Regulatory medical writing is the highest-paid subspecialty of medical writing. Module 3.3 introduces you to this world in depth.
Full title
Biostatistician / Statistical Programmer
What they do daily
Design the statistical analysis plan for clinical trials, analyse the data after the trial ends, and produce statistical output tables — which the medical writer then interprets and writes about in the CSR. A constant collaboration partner.
Key skills
SAS/R programming · SAP design · ANCOVA, survival analysis · ITT/PP populations · Missing data handling
Salary range India 2025
Rs. 5–9 LPA entry · Rs. 14–22 LPA senior · Requires statistics background
You do not need to be a statistician. But you must read their output and describe it accurately. Module 2.4 builds this exact skill.
Full title
Pharmacovigilance (PV) Associate / Drug Safety Officer
What they do daily
Collect, assess and report adverse events from clinical trials and post-market surveillance. Write Individual Case Safety Reports (ICSRs), Periodic Benefit-Risk Evaluation Reports (PBRERs) and other safety documents.
Key skills
MedDRA coding · Causality assessment · ICH E2A/E2E guidelines · CDSCO reporting timelines · Safety narrative writing
Salary range India 2025
Rs. 3–5 LPA entry · Rs. 8–14 LPA senior · High demand, growing field
PV writing is a distinct and growing subspecialty. Many medical writers in India start here. Module 4.1 covers pharmacovigilance writing in depth.
The writer's unique position
A medical writer is the only person in the team who touches every other role's work. You receive data from the biostatistician, align with regulatory affairs, write content for the MSL, and produce documents the sponsor submits to CDSCO. You are the connective tissue of the entire operation — invisible but essential.
Knowledge check
A CRA visits a hospital site and finds that a patient's informed consent form was signed after the first dose was given — not before. What is the most appropriate next step?
ADocument it as a protocol deviation and report to the sponsor and ethics committee
BIgnore it — the patient received the drug and is fine
CNote it mentally and remind the investigator to be more careful next time
DImmediately withdraw the patient from the study
Reflection
Which of the six roles surprised you most — and why? Which role feels closest to what you want to be doing in 12 months?
This reflection will be useful in your Module 1.4 career planning exercise.
Therapeutic Area Endpoint Mapping Across Clinical Phases
Medical writers must be fluent across distinct therapeutic modalities and their regulatory endpoint expectations:
- Cardiology & Vascular (e.g. Heart Failure / CVOT): Phase II/III focus on composite 3-point MACE (CV death, non-fatal MI, non-fatal stroke), annualized heart failure hospitalisation rates, and KCCQ health status scores.
- Endocrinology & Metabolism (e.g. Type 2 Diabetes / Obesity): Primary efficacy centers on HbA1c reduction (LS Mean change from baseline at Weeks 24–52) and percentage body weight loss; key safety includes adjudicated pancreatitis, thyroid C-cell pathology, and continuous glucose monitoring (time-in-range).
- Immunology & Rheumatology (e.g. Rheumatoid Arthritis / Psoriasis): Categorical response thresholds like ACR20/50/70, PASI-75/90/100, and EASI-75. Safety monitors opportunistic infections, herpes zoster, and venous thromboembolism (VTE).
- Respiratory & Pulmonology (e.g. Severe Asthma / COPD): Annualized rate of moderate-to-severe exacerbations, pre- and post-bronchodilator trough FEV1, FeNO, and blood eosinophil stratification.
- Neurology & CNS (e.g. Multiple Sclerosis / Alzheimer's): Annualized relapse rates (ARR), EDSS disability progression, CDR-SB cognitive ratings, and volumetric brain MRI lesion burdens.
Next module
1.3 — ICH-GCP: the rulebook
Week 1 · Module 3
ICH-GCP — the rulebook
The guidelines that govern every clinical trial on the planet — and why you must know them before you write a single word
Your scenario
A junior writer submits a CSR section to the client. The client sends it back: "This does not align with ICH E6(R2) Section 4.8." The junior writer stares at the screen. They have heard of ICH-GCP. They have never actually read it. This module ensures that never happens to you.
ICH stands for the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use. GCP stands for Good Clinical Practice. Together, ICH-GCP E6(R2) is the international ethical and scientific quality standard for designing, conducting, recording and reporting clinical trials.
It exists to protect trial participants and ensure the data generated is credible and accurate. Every clinical trial conducted by a reputable sponsor anywhere in the world follows ICH-GCP. Every document you write as a medical writer describes work that was — or should have been — conducted under these guidelines.
The most important sentence in GCP
"The rights, safety, and well-being of the trial subjects are the most important considerations and should prevail over interests of science and society." — ICH E6(R2), Section 2.3. Every guideline, every SOP, every document in clinical research traces back to this single principle.
Five GCP principles every writer must know
Principle 2.1 — Ethics first
Clinical trials must be conducted in accordance with the ethical principles originating from the Declaration of Helsinki. As a medical writer: every document you produce must accurately reflect that these ethical standards were upheld. If they were not, the document must not obscure that fact.
Principle 2.3 — Subject rights prevail
The rights, safety, and well-being of trial subjects are the most important considerations — above science and commerce. As a medical writer: when writing about adverse events, never minimise or reframe safety signals to make results look better. Accurate, complete safety reporting is a non-negotiable ethical obligation.
Principle 2.5 — Scientific soundness
Trials must be scientifically sound and described in a clear, detailed protocol. As a medical writer: you will often write or edit the clinical trial protocol — the document that defines how the trial is conducted. Module 3.2 covers protocol writing in detail.
Principle 2.10 — Accurate records
All clinical trial information should be recorded, handled, and stored in a way that allows its accurate reporting, interpretation and verification. Good Documentation Practice (GDP) is the practical application of this principle: be accurate, be complete, be legible, be timely, be traceable.
Principle 2.13 — Quality assurance
Systems and procedures must assure the quality of every aspect of the trial. As a medical writer: quality review (QC) of every document is mandatory. In practice this means another writer or reviewer checks your work against source data before submission. You will experience this in every CRO and pharma company.
Key terms
Declaration of Helsinki
Foundational international ethical framework for medical research involving humans. First adopted 1964. ICH-GCP is built on its principles.
Every informed consent form references it, usually in the opening paragraph.
Informed Consent Form (ICF)
Legal and ethical document signed by each trial participant before any study procedure. Must explain purpose, risks, benefits, and the patient's right to withdraw at any time without penalty.
Medical writers draft ICFs — they must be in plain language a layperson can understand.
Protocol Deviation
Any departure from the approved clinical trial protocol. Must be documented and reported. Major deviations may require reporting to the ethics committee.
Protocol deviations are described in the CSR patient disposition section.
Good Documentation Practice (GDP)
The five principles of clinical documentation: accurate, complete, legible, timely, and traceable. Every document you produce as a medical writer must meet all five.
If it is not written down with a date and source — in regulatory terms it did not happen.
Knowledge check
A medical writer notices a CSR section describes an adverse event as 'mild and transient' but the source data shows the patient was hospitalised for 3 days. What should the writer do?
AKeep the 'mild and transient' description since the patient recovered
BFlag the discrepancy and correct the description to accurately reflect the hospitalisation
CAsk the biostatistician to re-categorise the event in the statistical tables
DRemove the adverse event from the section to avoid confusion
Reflection
When you read 'the rights, safety, and well-being of trial subjects prevail over interests of science and society' — what does that mean for you personally as someone who will write the documents that describe these trials?
Your notes save automatically.
Next module
1.4 — Career paths: your map
Week 1 · Module 4
Career paths — your map
Where you can go from here, how fast, and what each path actually looks like from the inside
Your scenario
You finish reading your first CSR on Day 2. Your manager asks: "Where do you see yourself in 3 years?" You need an honest, specific answer — not "I want to grow and learn." This module gives you the vocabulary and the map to answer that question with confidence.
Agency / CRO
In-house pharma
Freelance
Regulatory specialist
PhD bridge
Career progression
Junior Medical Writer → Medical Writer II → Senior Medical Writer → Principal Writer → Associate Director Medical Writing
Typical timeline
2 years junior → 2 years MW II → 3 years senior → leadership roles. Acceleration possible with strong portfolio and regulatory experience.
Best for
Freshers and early career professionals. CROs expose you to multiple therapeutic areas and multiple sponsors simultaneously — fastest way to build a versatile portfolio.
Top India CROs hiring writers
Siro Clinpharm (Mumbai) · Veeda Clinical Research (Ahmedabad) · Lambda Therapeutic (Ahmedabad) · Parexel India (Hyderabad) · ICON India (Bengaluru) · Syneos Health India (Hyderabad)
Most Foundation Track graduates start here. Your first placement will likely be at a Hyderabad or Bengaluru CRO — many accept hybrid or remote writers now.
Career progression
Scientific Writer → Senior Scientific Writer → Principal Scientific Writer → Medical Writing Lead → Head of Medical Writing
Entry requirement
Usually requires 3–5 years CRO or agency experience first. In-house pharma roles are fewer and more competitive — but offer deeper therapeutic area expertise and better work-life balance.
Best for
Writers who want to specialise deeply in one or two therapy areas (oncology, cardiology) and work closely with clinical development teams over multiple years.
Companies hiring in India
Sanofi India · Novartis India · AstraZeneca India · Dr Reddy's · Cipla · Biocon · Abbott India
These organisations are the main employers of medical writers in India, and between them account for most entry-level and mid-career openings.
The path
Build portfolio → first project on Upwork or direct outreach → regular clients → retainer contracts → full freelance practice
Realistic timeline
First freelance income: 2–4 months post-course with a strong portfolio. Full-time freelance income replacement: 12–24 months. Not a path for Day 1, but achievable faster than most think.
India pricing guide 2025
Abstract: Rs. 1,500–3,000 · MSL deck: Rs. 8,000–20,000 · CSR section: Rs. 15,000–40,000 · Regulatory synopsis: Rs. 30,000–60,000 · Editing: Rs. 500–1,000/hour
Best platforms
LinkedIn (most effective in India) · Upwork · Direct email to CROs · Your IMRAD Academy alumni network
Module 4.3 is entirely dedicated to building your freelance practice — including a cold email template to CROs and a complete pricing strategy.
Career progression
Medical Writer → Regulatory Medical Writer → Senior Regulatory Writer → Regulatory Writing Lead → Director Regulatory Affairs (writing track)
What makes it different
You specialise in CTD submissions — the enormous dossiers submitted to CDSCO, FDA and EMA for drug approval. Requires deep knowledge of ICH guidelines, CTD structure, and specific regulatory requirements per market.
Salary premium
Regulatory writers earn 30–50% more than clinical medical writers at equivalent experience levels. Global demand far exceeds supply — including in India.
How to enter
Start as clinical writer → take regulatory writing projects at your CRO → build a CTD portfolio piece → Module 3.3 gives you the foundation
If Module 3.3 excites you, this is your highest-value specialisation. Very few writers in India have deep regulatory writing experience — which means very strong demand for those who do.
Your advantage as a PhD
You already understand the scientific method, literature review, and data interpretation at a deeper level than most freshers. You need industry document knowledge — which this course provides — not more scientific training.
What changes
Academic writing is descriptive and exploratory. Industry medical writing is structured, constrained, audience-specific and regulatory-compliant. The gap is learnable in weeks, not years.
Career entry point
With the right portfolio you can potentially enter at Medical Writer II or Senior Medical Writer level — skipping the junior stage entirely. This saves 2–3 years of career progression.
Fastest bridge
IMRAD Advanced Track → 3 industry-format portfolio pieces → apply for Senior Medical Writer roles directly
PhD scholars who complete the Advanced Track and build strong portfolio pieces typically enter at Rs. 7–12 LPA — significantly above the fresher entry point.
Week 1 portfolio exercise — your 6-month plan
Write one page answering: (1) Which career track am I targeting and why? (2) What 3 specific skills do I currently lack for that role? (3) What exactly will I do in the next 6 months — concrete steps, not vague intentions? This document will be reviewed in your final module and used to build your personalised action plan.
Knowledge check
A fresher with an M.Pharm in Pharmacology wants to enter medical writing. They have no industry experience. Which is the most practical first step?
AApply directly for Senior Medical Writer roles at Sanofi or Novartis
BBegin freelancing immediately on Upwork without any portfolio or training
CBuild a 3-piece portfolio and apply to CROs for junior medical writer or internship roles
DComplete a PhD before attempting to enter medical writing
Reflection
If you could be doing any job in pharma in exactly 3 years from today — what does a Tuesday look like? Be specific: where are you, who are you talking to, what are you writing?
Your notes save automatically.
Next module
2.1 — IMRAD in depth
Week 2 · Module 1
IMRAD in depth
The universal structure of all scientific writing — and how to master each of its four sections
IMRAD — Introduction, Methods, Results, and Discussion — is the universal structure of every clinical paper, regulatory submission and scientific publication in the world. Every pharma professional, every journal editor, every CRO manager lives inside this framework daily.
Before the 1970s, scientific papers were written in whatever order the author preferred. In 1979, the ICMJE standardised the format. Today it is the default structure for CSRs, regulatory summaries, grant applications, conference abstracts and most scientific communication in the pharmaceutical industry. When you internalise this structure, you internalise the language of pharma.
Introduction
Methods
Results
Discussion
Purpose
Answer: Why was this study needed? What is already known? What gap does this study fill? What was the primary objective?
Structure — funnel shape (broad to specific)
1. Broad context — the disease and its burden
2. Current evidence — what is known from existing research
3. Gap in evidence — what is not known or not settled
4. This study's objective — to fill that specific gap
2. Current evidence — what is known from existing research
3. Gap in evidence — what is not known or not settled
4. This study's objective — to fill that specific gap
Example from CASTELLAN study
"Urothelial carcinoma (UC) is the second most common skin cancer worldwide, accounting for approximately 20% of all skin cancer-related deaths in patients with locally advanced or metastatic disease. Despite recent advances with PD-1 inhibitors, real-world comparative data between velparimab and historical systemic treatments remain limited."
The last sentence of the Introduction is almost always: "Therefore, this study aimed to [primary objective]." It closes the funnel and hands off to Methods.
Purpose
Answer: How was the study done? Could someone else reproduce it exactly using only this description? Does it match the approved protocol?
Required elements
Study design · Patient population (inclusion/exclusion criteria) · Interventions · Primary and secondary endpoints · Statistical analysis plan reference · Ethical approvals (IEC, ICF)
Voice and tense
Past tense, passive voice: "Patients were enrolled from three academic centres across France" — not "We enrolled patients." The convention makes methods sound objective and reproducible.
Every number in Methods must match the approved protocol. If the protocol says "patients aged 18–75" and you write "patients aged 18–70" — that is a source data discrepancy. This is why source data verification matters in CSR writing.
Purpose
Present findings exactly as they occurred — without interpretation. Data only. No opinion. No speculation. No comparison to other studies.
Standard order
1. Patient disposition (enrolled, completed, discontinued)
2. Baseline demographics
3. Primary endpoint results
4. Secondary endpoint results
5. Safety results (adverse events, SAEs)
2. Baseline demographics
3. Primary endpoint results
4. Secondary endpoint results
5. Safety results (adverse events, SAEs)
The golden rule
"Results reports findings. Discussion explains them." Never mix interpretation into Results. "The ORR was 46.1%" belongs in Results. "This represents a clinically meaningful improvement over historical rates" belongs in Discussion.
Every number in Results must trace back to a statistical table from the biostatistician. If the table says 46.1% and you write 46%, you have introduced a transcription error. Precision is non-negotiable — regulators count significant figures.
Purpose
Interpret the results: What do they mean? How do they compare with existing evidence? What are the limitations? What are the implications for clinical practice?
Structure — reverse funnel (specific to broad)
1. Main finding restatement (brief, not a Results repeat)
2. Context — how results compare to existing literature
3. Mechanistic interpretation — why this result makes scientific sense
4. Limitations — honest acknowledgement of study weaknesses
5. Conclusion and implications for practice
2. Context — how results compare to existing literature
3. Mechanistic interpretation — why this result makes scientific sense
4. Limitations — honest acknowledgement of study weaknesses
5. Conclusion and implications for practice
Common mistakes
Repeating Results verbatim · Overclaiming causation from observational data · Understating or omitting limitations · Making regulatory claims not supported by the study design
The Discussion is where your scientific judgement shows. A junior writer describes data. A senior writer contextualises it, acknowledges its limitations honestly, and frames its clinical significance with precision — without overstating what the evidence allows.
Portfolio piece 1 — your first assignment
Using the CASTELLAN study data from Module 1.1, write a 250-word structured abstract with five labelled sections: Background / Objective / Methods / Results / Conclusions. Word count must be 240–260 words. Every number must match the source data. This is Portfolio Piece 1. Keep it in a folder with your other portfolio pieces.
Knowledge check
A structured abstract is submitted to a congress. The Results section begins with: 'The primary endpoint was not statistically significant (p=0.12). However, notable improvements were seen in several secondary endpoints.' What is the most significant problem with this writing?
AThe word 'however' is informal and should not appear in scientific writing
Bp=0.12 should be reported as 'not significant' without providing the actual p-value
CSecondary endpoints should never be mentioned in a Results section abstract
DDescribing a non-significant result as a finding to be noted alongside 'notable improvements' in secondary endpoints creates a misleading impression of overall study success
Reflection
Think of the last scientific paper you read — a journal article, a thesis chapter, or even a research news summary. Can you identify which IMRAD section each paragraph belonged to? Which section was hardest to identify, and why?
Your notes save automatically.
Next module
2.2 — Literature search mastery
AI extension · Module 2·1
What reporting guidelines lose in an AI-assisted draft
Reporting guidelines exist because authors reliably omit the same things. Language models were trained on those same imperfect papers — so they omit the same things, faster, and in better prose.
A CONSORT, STROBE or PRISMA checklist is a list of items a reader needs in order to judge whether they can believe the study. Most of those items are unglamorous: how the allocation sequence was concealed, who was blinded, how missing data were handled. They interrupt the flow of a paragraph. A system optimising for readable prose has a structural incentive to drop precisely them.
This produces a specific and recognisable artefact: a manuscript that reads better than the underlying study can support. Reviewers rarely complain about it, because nothing on the page is wrong. The information required to spot the weakness has simply gone.
Frequently lost
Allocation concealment (CONSORT 10)What survives instead
"Patients were randomised to receive…" — randomisation is mentioned, the mechanism that made it credible is not.Frequently lost
Blinding detail — who, and how (CONSORT 11)What survives instead
"The study was double-blind." Which parties were blinded, and how the blind was maintained, disappears.Frequently lost
Sample size calculation (CONSORT 7a)What survives instead
The final n, with no assumptions, power, effect size or alpha. A number without a justification.Frequently lost
Harms, reported fully (CONSORT 19)What survives instead
Selected or summarised adverse events, with "generally well tolerated" doing work that data should be doing.Frequently lost
Participant flow numbers (CONSORT 13)What survives instead
Analysed n only. Screened, excluded, lost to follow-up and reasons for discontinuation quietly compress.Frequently lost
Registration and protocol access (CONSORT 23–24)What survives instead
Nothing. Models almost never generate a registration number unprompted — and will invent one if pushed.Frequently lost
Changes to outcomes after trial start (CONSORT 6b)What survives instead
A clean statement of the endpoint as analysed, with no record that it differs from the endpoint as planned. This one matters most.Frequently lost
PRISMA: search strategy, in fullWhat survives instead
"A systematic search of PubMed and Embase was conducted." Databases named, strings and dates omitted — the review becomes unreproducible.The one that should worry you most
Outcome switching (CONSORT 6b). If an AI-assisted draft presents the analysed endpoint cleanly and the
registered endpoint has been forgotten, the manuscript now silently misrepresents the trial. Nobody
intended it, no sentence is false, and it is exactly the pattern that trial-transparency researchers
look for. Always open the registry record alongside the draft.
Practical method
Run the checklist against the draft, not against your memory of the study. The failure mode is that you know the trial well, you read the sentence, your own knowledge fills the gap, and you record the item as present. Have someone who has not read the protocol do the pass if you can; if you cannot, check the item against the page number where the reader would find it, and write that number down. If you cannot write a page number, the item is missing.
- Trial registration number present, real, and checked against the registry record
- Analysed primary endpoint matches the registered primary endpoint — or the change is declared
- Randomisation: sequence generation, allocation concealment and implementation all described
- Blinding: who was blinded, and how it was maintained
- Sample size justification present with its assumptions
- Participant flow accounts for every screened patient
- Harms reported in full, not summarised into an adjective
- Missing data handling stated
- Every checklist item mapped to a page number, not to your memory
0 of 9 checked
Knowledge check
Week 2 · Module 2
Literature search mastery
How to find, filter and organise the evidence that every document you write depends on
A literature search is not typing a question into Google Scholar and clicking the first result. It is a documented, reproducible strategy that you can present to a client or include in a Methods section. Every medical writer conducts literature searches daily — for background sections, to find supporting evidence, to identify adverse event precedents.
The three essential databases every Indian medical writer must know: PubMed (free, comprehensive, global standard), Embase (broader European and pharmacological coverage, particularly strong for drug-related literature), and Cochrane Library (gold standard for systematic reviews and meta-analyses).
The Boolean operators — your search power tools
AND narrows: "velparimab AND urothelial carcinoma" — both terms must be present.
OR broadens: "UC OR urothelial carcinoma" — either term is sufficient.
NOT excludes: "skin cancer NOT melanoma" — removes melanoma papers.
Quotation marks = exact phrase: "overall survival" searches as a phrase.
MeSH terms = standardised vocabulary that catches all synonyms automatically.
OR broadens: "UC OR urothelial carcinoma" — either term is sufficient.
NOT excludes: "skin cancer NOT melanoma" — removes melanoma papers.
Quotation marks = exact phrase: "overall survival" searches as a phrase.
MeSH terms = standardised vocabulary that catches all synonyms automatically.
Basic search
Advanced search
Filters
MeSH terms
Basic keyword search
velparimab urothelial carcinoma
Results (approximate)
~340 results — too broad, too many irrelevant papers
Never submit a basic keyword search as your literature review. It is neither reproducible nor comprehensive enough for professional use.
Professional advanced search
("velparimab"[Title/Abstract] OR "MRD-2810"[Title/Abstract]) AND ("urothelial carcinoma"[MeSH Terms] OR "UC"[Title/Abstract]) AND ("clinical trial"[Publication Type])
Results (approximate)
~42 results — focused, relevant, reproducible
This search string belongs in your Methods section exactly as written. Anyone can reproduce it and get the same results. Always include the drug's development code (MRD-2810 for velparimab) to capture early-phase papers published before the commercial name was used.
Essential PubMed filters
Article type: Clinical Trial · Systematic Review · Meta-Analysis
Publication date: Custom range (e.g. January 2020 – December 2025)
Language: English
Full text: Free full text available
Publication date: Custom range (e.g. January 2020 – December 2025)
Language: English
Full text: Free full text available
When to apply date filters
Background sections: no strict limit — include landmark historical papers. Current evidence sections: last 5 years typically. Systematic reviews: document the exact search date — literature evolves, a search from March 2025 may return different results than the same search in December 2025.
Always record the date you ran the search. This is a GDP requirement — your search is part of the audit trail for the document.
What are MeSH terms?
Medical Subject Headings — standardised vocabulary used by PubMed to index papers. Even if a paper uses "bladder cancer," PubMed indexes it under "Urinary Bladder Neoplasms." Searching by MeSH captures all synonyms automatically.
How to find MeSH terms
Go to mesh.nlm.nih.gov → Type your concept → Find the official MeSH term → Use it in your PubMed search with the [MeSH Terms] tag.
A literature search without MeSH terms is incomplete. You will miss papers that use different terminology for the same concept — which is extremely common in clinical research.
PICOS framework
Population, Intervention, Comparison, Outcome, Study design — the framework for building a systematic search strategy. Define your PICOS before you start searching.
Grey literature
Scientific information outside peer-reviewed journals: conference proceedings, regulatory documents, clinical trial registries (ClinicalTrials.gov), WHO databases. Often missed but important for regulatory submissions.
PRISMA flow diagram
Standardised diagram showing how many papers were identified, screened, excluded and included in a systematic review. Required for all systematic reviews submitted to journals.
Reference management
Tools like Zotero (free), EndNote and Mendeley organise your references, format citations automatically, and prevent duplicate references in large documents.
Knowledge check
You run a PubMed search and find 847 results. What is the most appropriate next step?
ARefine the search with additional MeSH terms, Boolean operators and article type filters
BRead all 847 abstracts and manually select relevant papers
CTake the first 20 results and use those as your literature base
DAbandon PubMed and switch to Google Scholar
Reflection
Think of a clinical topic you find interesting — a drug, a disease, or a treatment approach. How would you build a PICOS question around it? What MeSH terms might you use?
Your notes save automatically.
Next module
2.3 — Abstract writing clinic
Week 2 · Module 3
Abstract writing clinic
The 250-word document that gets your paper published, your poster accepted, and your career started
A structured abstract compresses an entire study into 200–300 words. Each section has a label. Nothing from one section leaks into another. Every number is exact. And the whole thing reads in under 90 seconds. Writing a crisp, accurate structured abstract is one of the most valued and tested skills in medical writing — and one of the first things a CRO hiring manager will ask you to produce.
Word budget for a 250-word structured abstract
Background: 30–40 words — context and evidence gap
Objective: 15–20 words — one sentence, starts with "To..."
Methods: 60–70 words — design, population, intervention, endpoints, statistical approach
Results: 70–80 words — primary endpoint first, key secondary endpoints, key safety finding
Conclusions: 25–35 words — clinical significance, no new data
Total: 200–260 words. Not 261. Not 198.
Objective: 15–20 words — one sentence, starts with "To..."
Methods: 60–70 words — design, population, intervention, endpoints, statistical approach
Results: 70–80 words — primary endpoint first, key secondary endpoints, key safety finding
Conclusions: 25–35 words — clinical significance, no new data
Total: 200–260 words. Not 261. Not 198.
Original — broken
Rewritten — correct
What changed
Original draft submitted by investigator (287 words, unstructured)
"The results of our study were very promising. Patients showed good responses to the treatment in most cases. We studied a group of patients with advanced disease who received the new drug. In conclusion, this drug appears to work better than old treatments. Standard statistical procedures were used. Overall survival was longer in the treatment group. The drug was given orally once daily. Adverse events were manageable, with nausea being the most common side effect seen in around one third of patients approximately. Patients were recruited from three sites. The study was randomised. The primary endpoint was overall response rate. The drug showed clinical benefit in the majority of patients which was statistically significant with a p-value that was less than 0.05."
Problems: No section labels. Conclusion appears before results. Methods are vague ("standard statistical procedures"). AE frequency is approximate ("around one third... approximately"). 287 words — over the limit. "Our study" instead of passive voice.
Rewritten — structured, 247 words
Background: Patients with advanced disease have limited systemic treatment options, with poor long-term outcomes under conventional chemotherapy regimens.
Objective: To evaluate the efficacy and safety of Drug X versus standard of care in patients with advanced disease.
Methods: This randomised, open-label, Phase III trial enrolled patients with histologically confirmed advanced disease from three academic centres. Patients received Drug X 200 mg orally once daily or investigator's choice chemotherapy. The primary endpoint was objective response rate (ORR) per RECIST v1.1 criteria assessed by blinded independent central review. Secondary endpoints included overall survival (OS) and safety.
Results: In total, 248 patients were enrolled (Drug X: n=124; control: n=124). ORR was significantly higher with Drug X versus control (54.8% vs 18.5%; p<0.001). Median OS was 24.3 months (95% CI: 18.6–31.2) with Drug X versus 14.1 months (95% CI: 10.8–17.4) with control (HR 0.62; 95% CI: 0.48–0.79; p<0.001). The most common treatment-related adverse event was nausea (Drug X: 33.1%).
Conclusions: Drug X demonstrated statistically significant and clinically meaningful efficacy versus standard of care, with a manageable safety profile, supporting its potential use in this patient population.
Objective: To evaluate the efficacy and safety of Drug X versus standard of care in patients with advanced disease.
Methods: This randomised, open-label, Phase III trial enrolled patients with histologically confirmed advanced disease from three academic centres. Patients received Drug X 200 mg orally once daily or investigator's choice chemotherapy. The primary endpoint was objective response rate (ORR) per RECIST v1.1 criteria assessed by blinded independent central review. Secondary endpoints included overall survival (OS) and safety.
Results: In total, 248 patients were enrolled (Drug X: n=124; control: n=124). ORR was significantly higher with Drug X versus control (54.8% vs 18.5%; p<0.001). Median OS was 24.3 months (95% CI: 18.6–31.2) with Drug X versus 14.1 months (95% CI: 10.8–17.4) with control (HR 0.62; 95% CI: 0.48–0.79; p<0.001). The most common treatment-related adverse event was nausea (Drug X: 33.1%).
Conclusions: Drug X demonstrated statistically significant and clinically meaningful efficacy versus standard of care, with a manageable safety profile, supporting its potential use in this patient population.
Word count: 247. Structured. Each section in its correct place. Numbers exact (33% became 33.1%). Confidence intervals and HR added. Conclusion is interpretive, not repetitive. This abstract would be accepted by a major congress.
Key changes made
1. Added section labels — Background / Objective / Methods / Results / Conclusions
2. Moved conclusion to end where it belongs
3. Replaced "standard statistical procedures" with specific endpoint and assessment criteria (RECIST v1.1, blinded independent central review)
4. Changed "around one third approximately" to exact "33.1%"
5. Added confidence intervals and hazard ratio to OS result
6. Removed informal language ("our study," "very promising")
7. Cut 40 words by eliminating redundant phrasing
8. Objective begins with "To" per convention
2. Moved conclusion to end where it belongs
3. Replaced "standard statistical procedures" with specific endpoint and assessment criteria (RECIST v1.1, blinded independent central review)
4. Changed "around one third approximately" to exact "33.1%"
5. Added confidence intervals and hazard ratio to OS result
6. Removed informal language ("our study," "very promising")
7. Cut 40 words by eliminating redundant phrasing
8. Objective begins with "To" per convention
Every change serves precision, brevity or structure. Nothing was added that was not already in the data. This is the discipline of medical writing.
Portfolio piece 1 — submit this week
Using the CASTELLAN study data from Module 1.1 (NCT0XXXXXXX, velparimab vs historical treatments, primary endpoint ORR, multicentre non-interventional retrospective design), write a 250-word structured abstract with all five labelled sections. Keep it with your portfolio. This is Portfolio Piece 1 — the first document in your professional portfolio.
Knowledge check
In a structured abstract Results section, which piece of information should appear first?
AA summary of patient demographics and baseline characteristics
BThe primary endpoint result
CThe safety and adverse event summary
DThe clinical significance of the findings
Reflection
Find a published abstract on PubMed for any Phase III oncology trial. Count the words in each section. Does it follow the word budget guidelines? What would you change if you were the writer?
Your notes save automatically.
Next module
2.4 — Reading clinical data
Week 2 · Module 4
Reading clinical data
How to interpret forest plots, Kaplan-Meier curves, hazard ratios and p-values — without being a statistician
Your scenario
The biostatistician sends you a 60-page statistical output (TFL package). Page 12 has a forest plot. Page 23 has a Kaplan-Meier survival curve with a "Number at Risk" table. Page 31 has a summary table of hazard ratios with 95% confidence intervals. Your manager says: "Draft the Efficacy Summary for the manuscript by tomorrow afternoon." If you have an M.Sc. or M.Pharm, you know biology—but university rarely taught you time-to-event survival analysis. This module breaks down exactly how to read these figures and convert them into flawless, professional scientific prose.
Pedagogical Roadmap — From Outputs to Study Designs
In this module, you master reading and writing about clinical trial statistical outputs (survival curves, forest plots, hazard ratios). In Week 4 (Module 4.2), we expand on the underlying study design hierarchies (observational vs RCT, superiority vs non-inferiority) that govern what causal conclusions you are legally and ethically permitted to draw from these numbers.
1. Kaplan-Meier Curves — Survival Analysis Step-by-Step
In life sciences, we usually calculate arithmetic means (e.g., mean blood glucose). But in clinical trials, you cannot calculate an arithmetic mean survival time because when the study ends, many patients are still alive! We don't know their eventual survival time. This is called censoring.
The Kaplan-Meier (KM) method solves this by recalculating survival probability every time an event (death or disease progression) occurs.
Anatomy of the curve
What censoring means
Median survival explained
The writing formula
X-Axis & Y-Axis
Y-Axis: Probability of survival or remaining event-free (ranges from 1.0 / 100% at time zero down toward 0.0).
X-Axis: Time elapsed since randomisation (months, weeks, or years).
X-Axis: Time elapsed since randomisation (months, weeks, or years).
Steps & Separation
Every vertical drop represents one or more patients experiencing an event. A curve that stays higher for longer indicates superior survival. The point where the two curves visibly separate marks when the therapeutic benefit emerges.
Key visual rule: In oncology, the upper curve is almost always the experimental arm if the drug works. The greater the space between the curves, the greater the clinical benefit.
Tick Marks = Censored Patients
Tick marks (or crosses) along the curve indicate censored observations. A patient is censored if: (1) they were still alive and event-free at the study cut-off date, (2) they withdrew consent, or (3) they were lost to follow-up.
The Number at Risk Table
Look below the X-axis for the "Number at risk" table. It shows how many patients remain alive and under observation at each milestone (e.g., Month 0, 6, 12, 18, 24). As events occur and patients are censored, the number at risk drops. When the number at risk gets very small (e.g., < 10), the tail of the curve becomes statistically unstable.
Writer's rule: Never make grand claims about the far-right tail of a KM curve if the number at risk has dropped to 2 or 3 patients.
Median Overall Survival (mOS) / Median PFS
Draw a horizontal line across from 50% (0.50 probability) on the Y-axis until it intersects the curve, then drop straight down to the X-axis. That time is the Median Survival. It is the time by which 50% of patients have experienced the event.
What "NR" or "NE" Means
If the survival curve never drops to the 50% line before the study ends, the median cannot be calculated. Report this as "Not Reached (NR)" or "Not Estimable (NE)". Never guess!
Crucial distinction: Never call median survival an "average". Survival data are skewed and censored; mean survival is rarely estimable.
Standard Sentence Template for Medical Writers
"Median overall survival was [X.X] months (95% CI: [XX.X]–[XX.X]) in the [Drug Name] group compared with [Y.Y] months (95% CI: [YY.Y]–[YY.Y]) in the control group (hazard ratio [0.XX]; 95% CI: [0.XX]–[0.XX]; p = [0.0XX])."
Example in Published Literature
"Median progression-free survival was 18.4 months (95% CI: 15.2–21.6) with pembrolizumab versus 11.2 months (95% CI: 8.9–13.8) with chemotherapy (HR 0.65; 95% CI: 0.53–0.81; p < 0.001)."
Always report the point estimate, the 95% CI for both arms, the HR with its 95% CI, and the exact p-value in one cohesive sentence.
2. Hazard Ratio (HR) — Relative Risk Over Time
The Hazard Ratio is the most widely reported effect measure in modern clinical trials, yet it is the single most misunderstood number among junior writers.
Hazard Ratio Value
Clinical Meaning & Exact Phrasing
HR < 1.0 (e.g., HR = 0.65)
Event = Death / Disease Progression
Event = Death / Disease Progression
Treatment reduces event rate by 35% (1.0 − 0.65 = 0.35).
Correct phrasing: "Treatment was associated with a 35% reduction in the hazard of progression or death."
WRONG phrasing: "Patients lived 35% longer" (HR is a rate ratio, not a survival time multiplier).
Correct phrasing: "Treatment was associated with a 35% reduction in the hazard of progression or death."
WRONG phrasing: "Patients lived 35% longer" (HR is a rate ratio, not a survival time multiplier).
HR = 1.0
No difference in event rate between treatment and control arms.
The null hypothesis is true.
The null hypothesis is true.
HR > 1.0 (e.g., HR = 1.40)
Treatment increases event rate by 40% (harm / worse outcome).
Correct phrasing: "The experimental treatment was associated with a 40% higher hazard of adverse events."
Correct phrasing: "The experimental treatment was associated with a 40% higher hazard of adverse events."
The Confidence Interval (CI) Rule of 1.0
Always look at the 95% Confidence Interval accompanying the Hazard Ratio:
- Significant benefit: HR 0.72 (95% CI: 0.58–0.89) → Both bounds are below 1.0. Statistically significant!
- Not significant: HR 0.72 (95% CI: 0.49–1.05) → The interval crosses 1.0. Not statistically significant, even though the point estimate (0.72) looks favorable! You cannot claim a proven benefit.
3. Forest Plots — Subgroup Analysis at a Glance
A Forest Plot displays effect estimates across multiple patient subgroups (e.g., Age <65 vs ≥65, Male vs Female, Stage III vs Stage IV, Biomarker Positive vs Negative) on a single visual plane.
Vertical Line of No Effect (1.0)
A vertical solid line at 1.0 represents no difference between treatment and control. If a subgroup's horizontal whisker crosses this line, that subgroup alone did not reach statistical significance.
Left of 1.0 favours treatment; right of 1.0 favours control (in efficacy trials).
Square / Box (Point Estimate)
The square represents the point estimate (HR or OR) for that specific subgroup. The size of the square represents the sample size or statistical weight of that subgroup.
Larger subgroups have larger squares and narrower horizontal lines.
Horizontal Line (95% CI)
The horizontal line extending from the box represents the 95% Confidence Interval. A short line means high precision (large subgroup); a long line means low precision (small subgroup).
Always inspect whether the whiskers cross 1.0.
The Bottom Diamond (Overall Result)
The overall pooled treatment effect is shown as a diamond at the bottom. The center of the diamond is the overall HR; the width represents the overall 95% CI.
If the diamond does not touch 1.0, the overall primary trial result is statistically significant.
How to Write About a Forest Plot
When all subgroups benefit: "The survival benefit of Drug X was consistent across all pre-specified subgroups, including age, disease stage, and prior line of therapy (Figure 2)."
When benefit is restricted to a subgroup: "Progression-free survival benefit was observed predominantly in patients with PD-L1 positive tumours (HR 0.54; 95% CI: 0.40–0.72), whereas no clear benefit was seen in the PD-L1 negative subgroup (HR 0.94; 95% CI: 0.72–1.23; p for interaction = 0.008)."
When benefit is restricted to a subgroup: "Progression-free survival benefit was observed predominantly in patients with PD-L1 positive tumours (HR 0.54; 95% CI: 0.40–0.72), whereas no clear benefit was seen in the PD-L1 negative subgroup (HR 0.94; 95% CI: 0.72–1.23; p for interaction = 0.008)."
Objective Response Rate (ORR)
Proportion of patients achieving complete response (CR) or partial response (PR) per RECIST v1.1. Categorical endpoint reported as % with 95% CI.
Formula: (CR + PR) / Total Evaluable Patients.
Progression-Free Survival (PFS)
Time from randomisation to objective disease progression or death from any cause. Evaluated via KM survival curves and HR.
Key surrogate endpoint in oncology.
Intention-to-Treat (ITT)
All randomised patients analysed in their originally assigned group, regardless of adherence or protocol deviations.
Mandatory primary population for FDA/EMA submissions to prevent bias.
Per-Protocol (PP) Set
Subset of patients who completed the study without major protocol violations. Used as a sensitivity analysis.
Never use PP as primary efficacy in a regulatory submission without ITT.
Knowledge check
A clinical study reports HR = 0.71 (95% CI: 0.63–0.91; p=0.003) for overall survival. How would you correctly describe this result in a scientific manuscript?
APatients on treatment lived 29% longer than patients on control
BThe result is not statistically significant because the confidence interval is wide
CTreatment was associated with a statistically significant 29% reduction in the hazard of death (HR 0.71; 95% CI: 0.63–0.91; p=0.003)
Dp=0.003 proves the treatment is clinically significant
Practice Exercise
Look at this mock data: Treatment Arm mPFS = 14.2 mo (95% CI: 11.5–16.8); Control Arm mPFS = 8.1 mo (95% CI: 6.4–9.7); HR = 0.58 (95% CI: 0.44–0.76; p < 0.001). Write the complete two-sentence results summary for an abstract.
Remember: report medians with CIs for both arms first, then state the HR, its 95% CI, and the p-value.
Next module
3.1 — MSL slide decks
AI extension · Module 2·4
Hazard ratio described as a risk ratio, odds ratio or "chance of survival"
HR 0.68 rendered as an increase in risk, or a benefit reported for the wrong arm
"Patients survived on average 9.4 months"
Events described as though every patient was followed to the endpoint
An exploratory subgroup p-value reported as a finding
"Non-inferior" rewritten as "as effective as" or "equally effective"
ITT and per-protocol results interchanged, or population qualifier dropped
"There is a 95% probability the true HR lies between…"
When the model misreads the statistics
You have just learned to read a forest plot and a statistical analysis plan. That skill has acquired a second use: catching a language model that has read them wrong.
Large language models are trained on enormous quantities of published biomedical text, which makes them superb at producing sentences that sound like results reporting. They are markedly less reliable at the underlying reasoning. The result is a specific and predictable set of errors — statistically illiterate content in statistically fluent prose.
These are worth memorising, because unlike a typo they survive every automated check and most human ones.
The failure
Effect-measure conflationHazard ratio described as a risk ratio, odds ratio or "chance of survival"
How to catch it
An HR is a ratio of instantaneous rates over time, not a probability. If the sentence would still make sense with the time dimension removed, it is describing the wrong quantity.The failure
Direction inversionHR 0.68 rendered as an increase in risk, or a benefit reported for the wrong arm
How to catch it
Check the CI against the point estimate. An inverted HR usually leaves the original CI behind, producing an internally impossible statement. Always read the two together.The failure
Median treated as mean"Patients survived on average 9.4 months"
How to catch it
Survival data are censored and skewed; the median is reported precisely because the mean is not estimable. "Average" is never the right word for a median survival.The failure
Censoring ignoredEvents described as though every patient was followed to the endpoint
How to catch it
Ask what happened to patients still at risk at data cut-off. If the draft cannot answer it, the sentence is overstating what was observed.The failure
Subgroup promoted to confirmatoryAn exploratory subgroup p-value reported as a finding
How to catch it
Go to the SAP. If the analysis was not pre-specified and alpha-controlled, the only permissible language is hypothesis-generating.The failure
Non-inferiority read as superiority"Non-inferior" rewritten as "as effective as" or "equally effective"
How to catch it
Non-inferiority is a statement about a pre-defined margin, not equivalence. Locate the margin. If the draft never mentions it, the claim is unmoored.The failure
Analysis set substitutionITT and per-protocol results interchanged, or population qualifier dropped
How to catch it
Every efficacy and safety number carries a population. If a sentence names a number without naming its analysis set, treat it as unverified.The failure
Confidence interval misinterpreted"There is a 95% probability the true HR lies between…"
How to catch it
A frequentist CI is a statement about the procedure across repeated sampling, not a probability about this parameter. The correct phrasing is "consistent with".The asymmetry that matters
A model that misreads a statistic writes a clean sentence about it. Human statistical confusion
usually produces hedged, tangled prose that signals its own uncertainty. Machine confusion produces
confident, well-formed text. You lose the linguistic tell that used to warn you, which means the
numerical check is no longer optional.
Working checklist — statistics in an AI-assisted draft
- Every effect estimate reunited with its confidence interval and checked for internal consistency
- Every number carries an explicit analysis set (ITT / per-protocol / safety)
- Every "significant" traced to a pre-specified, alpha-controlled test in the SAP
- Every subgroup claim labelled exploratory unless the SAP says otherwise
- Survival language checked for median/mean and censoring assumptions
- Any non-inferiority claim states its margin
- Endpoints described in the draft match the endpoints defined in the protocol, by name
0 of 7 checked
Knowledge check
Week 3 · Module 1
MSL slide decks
The document that bridges scientific evidence and medical practice — and one of the first things you will be paid to write
Your scenario
An MSL is meeting a leading oncologist in Hyderabad next Tuesday. She needs a 10-slide deck covering the CASTELLAN study — concise, visually clear, no promotional language, built for a 20-minute scientific discussion. You have 3 days to deliver it. This module shows you exactly how.
What makes an MSL deck different from a presentation
An MSL slide deck is not a PowerPoint presentation. It is a scientific communication tool built for a specific conversation between two highly trained professionals. The MSL (who presents it) and the KOL (who receives it) both understand the science deeply. The deck's job is to structure that conversation — not to explain basic concepts.
The 5 cardinal rules of MSL deck writing
1. Maximum 15 words per bullet. If it cannot fit in 15 words, it is a paragraph — not a bullet.
2. Every data point has a source reference. Every number on every slide has a footnote citation.
3. No promotional language. "Breakthrough," "superior," "best-in-class" — never. These trigger MLR rejection instantly.
4. Efficacy before safety. Standard sequence: Study overview → Patient population → Primary endpoint → Secondary endpoints → Safety → Key takeaways.
5. Visuals carry the data, text carries the context. If a table can replace 5 bullets, use the table.
2. Every data point has a source reference. Every number on every slide has a footnote citation.
3. No promotional language. "Breakthrough," "superior," "best-in-class" — never. These trigger MLR rejection instantly.
4. Efficacy before safety. Standard sequence: Study overview → Patient population → Primary endpoint → Secondary endpoints → Safety → Key takeaways.
5. Visuals carry the data, text carries the context. If a table can replace 5 bullets, use the table.
Slide 1: Overview
Slide 2: Population
Slide 3: Primary endpoint
Slide 4: Safety
Slide 5: Takeaways
Slide title
CASTELLAN: Study design and objectives
Key elements
NCT number (NCT0XXXXXXX) · Study design (multicentre, non-interventional, retrospective) · Sponsor (a global pharma sponsor) · Publication reference with PMID · Simple design schematic
This slide answers: "What is this study?" in under 30 seconds. An MSL should be able to introduce the study from this slide alone without reading from it.
Slide title
Patient population — who was studied
Key elements
N enrolled (with treatment arms) · Key inclusion criteria (disease stage, prior treatment) · Key exclusion criteria · Baseline demographics table (age, sex, ECOG performance status)
KOLs always ask: "Is this population like my patients?" This slide answers that immediately. Baseline characteristics must come from statistical tables — never paraphrase numbers.
Slide title
Primary endpoint: Objective Response Rate (ORR)
Key elements
ORR with 95% CI for each arm · p-value · Visual (bar chart or figure from the publication) · Brief definition of ORR (per RECIST v1.1 or study-specific criteria)
The primary endpoint slide is the most important in the deck. Every number must be exact — pulled from the publication, not from memory. The visual must come from the paper, not be recreated from estimates.
Slide title
Safety summary — treatment-related adverse events
Key elements
Any-grade TRAE rate · Grade 3–4 TRAE rate · Most common TRAEs table (all events occurring in ≥10%) · Serious adverse events · Treatment discontinuation due to AEs · Deaths (treatment-related, if any)
Never present only favourable safety data. All serious adverse events and treatment-related deaths must be included. An MSL who cannot discuss safety honestly will lose the trust of any KOL permanently.
Slide title
Key clinical takeaways
Key elements
3–4 bullet points maximum · Each point = one clinical insight, not a data repeat · Starts with the primary result · Ends with a forward-looking scientific question or implication
Takeaways are interpretive — but every claim must be supported by data on previous slides. A non-interventional retrospective study cannot claim causation. "Was associated with higher ORR" not "demonstrated superior efficacy."
Portfolio piece 2 — your assignment
Create a 5-slide MSL deck for the CASTELLAN study using the data from Module 1.1. Use the slide sequence from this module. Maximum 15 words per bullet. Every number referenced. No promotional language. Keep it with your portfolio. This is Portfolio Piece 2.
Knowledge check
An MSL slide states: "Drug X is superior to chemotherapy and should be the first choice for all patients with advanced UC." What is wrong with this statement?
AThe font is too small for a presentation slide
BIt should not compare the drug to chemotherapy at all
CMentioning the disease indication constitutes off-label promotion
DIt uses promotional language and makes a prescriptive clinical recommendation not supported by the study design
Next module
3.2 — Clinical study reports
Week 3 · Module 2
Clinical study reports (ICH E3)
The backbone of regulatory submissions — demystifying the 340-page document and mastering Section 9 (Methods)
Your scenario
You open the Phase III Clinical Study Report (CSR) from Module 1.1. It is 340 pages long. Your manager says: "Write Section 9.1: Study Design and Plan. Here is the protocol and the SAP tables. I need the first draft by Thursday." You feel a sudden wave of panic—how can anyone write a 340-page document? Here is the secret: nobody writes a 340-page document from scratch. This module demystifies how a CSR is structured, where the data comes from, and guides you through writing Section 9.1 step by step.
Regulatory Context — The Single Study in the Global Dossier
A Clinical Study Report (CSR) is the comprehensive, definitive report of a single completed clinical trial. In the global Common Technical Document (CTD) hierarchy — which we explore across all five modules in Module 3.3 — every individual CSR lives inside CTD Module 5 (Clinical Study Reports). Mastering ICH E3 here prepares you to understand how multiple individual CSRs are synthesized into regulatory overviews.
1. Demystifying the 340-Page Document
When a fresher hears "340-page report," they imagine writing an entire textbook. In reality, a CSR is composed of three distinct parts:
CSR Component
Who Creates It & What It Contains
Core Narrative Body
(Sections 1–15)
~50–90 pages
(Sections 1–15)
~50–90 pages
Written by the Medical Writer in collaboration with the clinical team. Synthesizes the protocol, operations data, and statistical tables into formal regulatory prose following the ICH E3 guideline.
Tables, Figures, Listings
(TFLs / Section 14)
~150–400 pages
(TFLs / Section 14)
~150–400 pages
Generated by the Statistical Programming team (SAS programmers). Hundreds of pages of computer-generated tables (demographics, adverse event listings, laboratory shifts). You do not create these; you extract numbers from them!
Appendices & Attachments
(Section 16)
~100–500+ pages
(Section 16)
~100–500+ pages
Compiled by Clinical Operations & Data Management. Contains the final protocol, sample Case Report Forms (CRFs), investigator CVs, audit certificates, and randomisation lists.
2. The Source Document Triad — Where Your Content Comes From
A medical writer never invents facts. Every single paragraph in a CSR originates from one of three source documents:
1. The Clinical Protocol
Describes what was planned. Use this to write Section 9 (Investigational Plan / Methods), Section 9.3 (Eligibility), and Section 9.4 (Treatments). Every word must mirror the approved protocol.
Source for: Objectives, Design, Endpoints, Inclusion/Exclusion.
2. Statistical Analysis Plan (SAP) & Tables
Describes the mathematical rules and actual outputs. Use this to write Section 9.7 (Statistical Methods), Section 10 (Patient Disposition & Demographics), and Section 11 (Efficacy).
Source for: Sample size justification, ITT definitions, p-values, HRs.
3. Safety Database & Line Listings
Lists every single adverse event, death, and lab abnormality. Use this to write Section 12 (Safety Evaluation) and Section 12.3.2 (Patient Safety Narratives).
Source for: AEs, SAEs, discontinuations, laboratory shifts.
4. Protocol Amendments & Ops Logs
Records every change or deviation that occurred while the study was running. Use this to write Section 9.8 (Changes in Planned Conduct) and Section 10.2 (Deviations).
Source for: Protocol version tracking and major deviations.
3. The ICH E3 16-Section Structure
The International Council for Harmonisation (ICH) E3 guideline is the universal global template. Medical writers typically own Sections 6 through 14.
Section 9: Methods
Section 10: Patients
Section 11: Efficacy
Section 12: Safety
Section 16: Synopsis
Section 9 — Investigational Plan (Methods)
9.1 Study Design & Plan: Overview of trial structure, blinding, controls.
9.2 Selection of Population: Inclusion & exclusion criteria.
9.4 Treatments Administered: Drug doses, route, duration, blinding procedures.
9.5 Efficacy & Safety Variables: Exact timing and definitions of endpoints.
9.7 Statistical Methods: Analysis populations (ITT, Safety), handling of missing data.
9.2 Selection of Population: Inclusion & exclusion criteria.
9.4 Treatments Administered: Drug doses, route, duration, blinding procedures.
9.5 Efficacy & Safety Variables: Exact timing and definitions of endpoints.
9.7 Statistical Methods: Analysis populations (ITT, Safety), handling of missing data.
Written primarily in past tense, passive voice. Every statement must match the protocol exactly.
Section 10 — Study Patients
10.1 Disposition of Patients: Number screened, randomised, completed, and discontinued (broken down by reason: AE, consent withdrawal, disease progression).
10.2 Protocol Deviations: Major deviations (wrong dose, missed visit, ineligible patient enrolled).
10.3 Demographics & Baseline: Age, sex, race, baseline disease stage.
10.2 Protocol Deviations: Major deviations (wrong dose, missed visit, ineligible patient enrolled).
10.3 Demographics & Baseline: Age, sex, race, baseline disease stage.
Disposition numbers must reconcile perfectly across tables and text: Enrolled = Completed + Discontinued.
Section 11 — Efficacy Evaluation
11.1 Data Sets Analysed: ITT population, Per-Protocol population.
11.2 Demographic & Baseline Characteristics: Baseline comparability between arms.
11.4 Efficacy Results: Primary endpoint first, followed by secondary endpoints, subgroup analyses, and sensitivity analyses.
11.2 Demographic & Baseline Characteristics: Baseline comparability between arms.
11.4 Efficacy Results: Primary endpoint first, followed by secondary endpoints, subgroup analyses, and sensitivity analyses.
Strictly results only. No subjective speculation or marketing claims. Interpretation belongs in Section 13/14.
Section 12 — Safety Evaluation
12.1 Extent of Exposure: Total doses administered, median treatment duration.
12.2 Adverse Events (AEs): Common AEs, Grade 3/4 AEs, treatment-related AEs.
12.3 Deaths, Other SAEs & Dropouts: Detailed narratives for every death or life-threatening event.
12.4 Clinical Laboratory Evaluations: Liver enzymes, hematology shifts, renal function.
12.2 Adverse Events (AEs): Common AEs, Grade 3/4 AEs, treatment-related AEs.
12.3 Deaths, Other SAEs & Dropouts: Detailed narratives for every death or life-threatening event.
12.4 Clinical Laboratory Evaluations: Liver enzymes, hematology shifts, renal function.
Zero suppression: Every AE reported during the trial must be documented, whether drug-related or not.
Section 16.1.1 — The CSR Synopsis
A 3–5 page structured summary of the entire study (title, study period, objectives, methodology, number of patients, diagnosis, test product/reference, criteria for evaluation, statistical methods, summary of results, and conclusions).
Always written LAST. Regulators read the Synopsis first before opening the 300-page dossier.
4. Guided Clinic — How to Draft Section 9.1 (Study Design)
Let's walk through how a real junior medical writer transforms a protocol summary into ICH E3 Section 9.1 text:
Raw Protocol Excerpt (Protocol ABC-301)
Title: A Phase III, Randomised, Double-Blind, Placebo-Controlled, Multicenter Study of Compound X in Adult Patients with Moderate-to-Severe Atopic Dermatitis.
Design: 2-arm parallel group; 1:1 randomisation; 16-week double-blind treatment period followed by a 4-week safety follow-up. Stratified by baseline disease severity (moderate vs severe) and geographic region.
Intervention: Compound X 200 mg subcutaneous injection every 2 weeks vs matching placebo injection.
Primary Objective: Proportion of patients achieving IGA score of 0 or 1 at Week 16.
Design: 2-arm parallel group; 1:1 randomisation; 16-week double-blind treatment period followed by a 4-week safety follow-up. Stratified by baseline disease severity (moderate vs severe) and geographic region.
Intervention: Compound X 200 mg subcutaneous injection every 2 weeks vs matching placebo injection.
Primary Objective: Proportion of patients achieving IGA score of 0 or 1 at Week 16.
How the Medical Writer Transforms This into Section 9.1
"Study ABC-301 was a Phase III, randomised, double-blind, placebo-controlled, parallel-group, multicenter trial designed to evaluate the efficacy and safety of Compound X compared with placebo in adult patients with moderate-to-severe atopic dermatitis.
Eligible patients were randomised in a 1:1 ratio to receive either Compound X 200 mg administered subcutaneously every 2 weeks or matching placebo for a 16-week double-blind treatment period. Randomisation was stratified by baseline disease severity (moderate vs severe) and geographic region (North America, Europe, Asia-Pacific). Following the 16-week treatment period, all patients entered a 4-week safety follow-up period.
The primary objective of the study was to evaluate the proportion of patients achieving an Investigator's Global Assessment (IGA) score of 0 (clear) or 1 (almost clear) with at least a 2-grade reduction from baseline at Week 16."
Eligible patients were randomised in a 1:1 ratio to receive either Compound X 200 mg administered subcutaneously every 2 weeks or matching placebo for a 16-week double-blind treatment period. Randomisation was stratified by baseline disease severity (moderate vs severe) and geographic region (North America, Europe, Asia-Pacific). Following the 16-week treatment period, all patients entered a 4-week safety follow-up period.
The primary objective of the study was to evaluate the proportion of patients achieving an Investigator's Global Assessment (IGA) score of 0 (clear) or 1 (almost clear) with at least a 2-grade reduction from baseline at Week 16."
Portfolio Piece 3 — Your Assignment
Using the format demonstrated in the Guided Clinic above, write a 400-word Methods Section (ICH E3 Section 9) for a mock Phase III oncology study. Include: Study Design (9.1), Patient Eligibility (9.2), Treatments (9.4), and Primary Endpoint (9.5). Maintain formal passive voice, past tense, and exact technical precision. Save this draft in your portfolio as Portfolio Piece 3 (ICH E3 CSR Methods Section).
Knowledge check
A CSR Methods section states the primary endpoint was "objective response rate." The approved protocol specified "objective response rate per RECIST v1.1 assessed by blinded independent central review." Which version is correct for the CSR?
AThe full protocol wording — the CSR must match the approved protocol precisely
BThe shorter CSR description — it is cleaner and easier for reviewers to read
CFlag it as a protocol deviation in the patient disposition section
DUse both descriptions and let the regulatory reviewer decide which is accurate
Next module
3.3 — Regulatory writing
AI extension · Module 3·2
Signing off an AI-drafted CSR section
Regulatory drafting tools can now assemble CSR sections directly from study data. What they cannot do is sign the document. Understanding exactly where that line falls is the difference between a writer who is replaced by the tool and one who is required by it.
Purpose-built regulatory platforms will structure and populate large portions of an ICH E3 report from source data, and they do it quickly. That is a genuine efficiency gain and there is no professional dignity in pretending otherwise. But the output is a draft submitted to no one. A qualified regulatory writer must still verify every figure against source and accept responsibility for the result before it enters a submission — which means your value has moved from production to attestation.
What changes, concretely
Historically you built the document and reviewed it as you built it — verification was distributed invisibly through the act of writing. When the draft arrives fully formed, that verification has not happened anywhere. It now has to be performed deliberately, as a separate, planned, time-boxed task. Teams that do not schedule it simply skip it, and nobody notices until an inspection.
The four checks, in order
1 · Completeness against ICH E3. A model generates what it was prompted for. It does not know what a submission requires. Sections that are frequently thin or absent in generated drafts: protocol deviations and their handling, disposition accounting that reconciles across the whole report, the rationale for the statistical methods actually used versus those planned, and demographic subgroup presentation. Run the E3 table of contents as a physical checklist, not from memory.
2 · Number-to-TFL traceability. Every number in narrative text exists in a table, listing or figure. Take them one at a time, in order, and mark the source location. This is tedious and it is the job. Pay closest attention to numbers that appear in more than one place — a synopsis figure, a body figure and a discussion figure that have drifted apart is a classic finding.
3 · Qualifier integrity. The most common silent corruption. Analysis population, timepoint, censoring rule and denominator are precisely the words that a fluency-optimising system removes to make a sentence read better. A safety percentage without its safety-population denominator is not a rounding problem; it is a different claim.
4 · Safety narratives. Treat these as the highest-risk content in the document. Individual patient narratives contain dates, doses, timings, concomitant medications and causality assessments — every one of which is a specific, checkable fact, and every one of which a model will happily invent a plausible version of if the source is ambiguous. Verify narratives against the CRF and the safety database, never against the draft's own internal consistency.
Two things that are never delegated
- Causality and clinical judgement. A model can restate an investigator's causality assessment. It cannot make one, and text that reads as though it did is a serious problem.
- Confidential source data. Unblinded data, patient-level listings and unpublished protocols go only into systems with a governed data agreement in place. "It was just for a draft" is not a defence at inspection.
Verification pass — working checklist
- ICH E3 table of contents walked section by section; gaps listed before reading any prose
- Every narrative number matched to a specific TFL cell and marked
- Repeated figures (synopsis / body / discussion) reconciled against each other
- Every efficacy and safety figure carries its analysis population and denominator
- Disposition numbers reconcile: randomised = treated + not treated, and account for every discontinuation
- Safety narratives verified against source, not against the draft
- Causality statements traced to the investigator assessment, not generated
- Statistical methods text matches the SAP, including any amendments
- Tool used, stage of use, and reviewer name recorded in the project file
0 of 9 checked
Knowledge check
Week 3 · Module 3
Regulatory writing
The highest-paid specialisation in medical writing — and the field growing fastest in Indian pharma
Your scenario
A pharma company is submitting a new drug for CDSCO approval. The regulatory affairs manager emails you: "We need Module 2.5 — Clinical Overview — first draft by end of month. Thirty pages. You have the clinical dossier." In Module 3.2, you worked inside an individual trial report (CSR / Module 5). Now we zoom out to the entire submission dossier — the Common Technical Document (CTD) — where all clinical and non-clinical data come together. This module gives you the map to navigate it.
The Common Technical Document — the world's drug approval format
The CTD is the internationally agreed format for drug regulatory submissions. Used by CDSCO (India), FDA (USA), EMA (Europe), PMDA (Japan) and most global regulatory authorities. It has five modules. Medical writers own Module 2 — the written summaries that synthesise all clinical and non-clinical evidence into a scientific argument for approval.
CTD Module 2 — what medical writers write
2.1 Table of Contents
2.2 Introduction
2.3 Quality Overall Summary (CMC team)
2.4 Nonclinical Overview (preclinical scientists)
2.5 Clinical Overview — 30–50 pages. Expert scientific analysis of all clinical data. Written by medical writers.
2.6 Nonclinical Written Summaries
2.7 Clinical Summaries — structured summaries of biopharmaceutics, clinical pharmacology, efficacy, and safety. Written by medical writers.
Sections 2.5 and 2.7 are your primary regulatory writing deliverables.
2.2 Introduction
2.3 Quality Overall Summary (CMC team)
2.4 Nonclinical Overview (preclinical scientists)
2.5 Clinical Overview — 30–50 pages. Expert scientific analysis of all clinical data. Written by medical writers.
2.6 Nonclinical Written Summaries
2.7 Clinical Summaries — structured summaries of biopharmaceutics, clinical pharmacology, efficacy, and safety. Written by medical writers.
Sections 2.5 and 2.7 are your primary regulatory writing deliverables.
2.5.1 Background
2.5.2 PK overview
2.5.5 Efficacy
2.5.6 Safety
2.5.7 Benefit-risk
Section 2.5.1 — Background and overview
Disease epidemiology and unmet need · Current standard of care and its limitations · Drug's mechanism of action · Brief overview of the complete clinical development programme
Written like an Introduction — broad to specific. Answers: why does this drug exist and why does the world need it? Every epidemiology statistic requires a literature reference with year and source population.
Section 2.5.2 — Overview of biopharmaceutics and PK
Absorption profile · Distribution (volume of distribution, protein binding) · Metabolism pathway · Excretion route · PK in special populations (renal impairment, hepatic impairment, elderly, paediatric if applicable)
You summarise Phase I PK data here. You do not need to be a pharmacokineticist — but you must understand Cmax, Tmax, AUC, t1/2 and bioavailability and be able to describe them accurately from the statistical tables.
Section 2.5.5 — Overview of efficacy
Summary of all Phase II and III efficacy data · Primary and secondary endpoint results across all studies · Consistency of effect across subgroups and regions · Comparison to existing treatments where available
The most complex section. You synthesise data from multiple studies into one coherent scientific argument for why the drug works. Inconsistencies between studies must be acknowledged and explained — not hidden. Every claim needs a specific study reference.
Section 2.5.6 — Overview of safety
Integrated safety summary across all clinical studies · Adverse event profile by system organ class · Serious adverse events · Deaths · Special populations safety · Drug interactions · Abuse potential if applicable
Safety overview requires the most care. You are making an integrated argument about the drug's safety across all studies. Selecting only favourable safety data is a regulatory offence. All data — favourable and unfavourable — must be presented and discussed honestly.
Section 2.5.7 — Benefit-risk conclusions
The regulatory writer's highest-stakes section. Explicitly weighs the clinical benefits against the known risks for the proposed indication and patient population. Concludes whether the benefit-risk profile supports approval.
Regulators read this section most carefully. It must be honest, balanced, and evidence-based. A benefit-risk conclusion that ignores safety signals or overstates efficacy will be challenged — and could delay or prevent approval.
Salary premium in regulatory writing
Regulatory medical writers earn 30–50% more than clinical medical writers at equivalent experience levels. Global demand far exceeds supply — including in India. If Module 3.3 excites you, this is your highest-value specialisation path. Entry via clinical writing → regulatory projects at CRO → CTD portfolio piece → apply for regulatory writer roles.
Knowledge check
In the CTD Module 2.5 benefit-risk section, the writer notes Drug X showed superior efficacy but caused Grade 3-4 hepatotoxicity in 18% of patients. The sponsor wants this toxicity rate omitted from the benefit-risk section. What should the writer do?
AFollow the sponsor instruction — they own the data
BInclude the data accurately and document the sponsor instruction in writing
CMove it to an appendix where it is available but less prominent
DDescribe it as manageable hepatic effects to contextualise the finding
Next module
3.4 — Patient information leaflets
Week 3 · Module 4
Patient information leaflets
Writing for patients — the hardest audience, the most important document, and the one most medical writers underestimate
Your scenario
A patient with a standard Class 8 education opens her medicine box. Inside is a leaflet you wrote. It says: "Contraindicated in patients with hepatic insufficiency or concurrent administration of CYP3A4 inhibitors." She reads it three times. She understands nothing. She stops taking the medicine. This module teaches you how to prevent that outcome.
The plain language imperative
A Patient Information Leaflet (PIL) is a regulatory document. CDSCO and EMA both require that it be written in plain language that patients can actually understand. But most PILs are written by scientists, reviewed by lawyers, and cleared by regulatory affairs — and end up comprehensible only to the people who produced them.
As a medical writer, your job is to be the patient's advocate in this process. You write the leaflet that a real person — likely anxious about their diagnosis, possibly with limited formal education — can read and act on in 5 minutes.
The Flesch-Kincaid readability test
The Flesch Reading Ease formula measures how easy text is to read. A score of 60–70 is easily understood by 13–15 year olds — the appropriate target for most PILs. Check it in Microsoft Word (Review → Check Accessibility → Readability Statistics) or at readable.com.
Below 40: too complex for most patients. Above 80: may feel condescending for adult patients. Aim for 60–70.
Below 40: too complex for most patients. Above 80: may feel condescending for adult patients. Aim for 60–70.
Original — clinical language
Rewritten — plain language
Rules applied
Original PIL section — Side effects (clinical language)
"Adverse reactions: Hepatotoxicity (including fatal cases) reported with incidence of 1.2% in clinical trials. Immune-mediated adverse reactions including pneumonitis, colitis, hepatitis, endocrinopathies and nephritis have been observed. Contraindicated in patients with hepatic insufficiency (Child-Pugh B or C). Concurrent administration with CYP3A4 inhibitors may increase plasma concentrations."
Reading age: approximately 18+ (university level). Most patients will not understand hepatotoxicity, immune-mediated, endocrinopathies, CYP3A4 inhibitors or plasma concentrations.
Rewritten in plain language
Possible side effects:
Tell your doctor immediately if you notice any of these serious side effects:
- Yellowing of your skin or eyes, dark-coloured urine, stomach pain, or feeling very tired (signs of liver problems)
- New or worsening cough, shortness of breath (signs of lung problems)
- Diarrhoea, blood in your stools, stomach pain (signs of bowel problems)
- Changes in how much you urinate (signs of kidney problems)
Do not take this medicine if you have liver disease — your doctor will have checked this before prescribing it to you.
Tell your doctor about all other medicines you take, including vitamins and herbal supplements, as some medicines can affect how this medicine works.
Tell your doctor immediately if you notice any of these serious side effects:
- Yellowing of your skin or eyes, dark-coloured urine, stomach pain, or feeling very tired (signs of liver problems)
- New or worsening cough, shortness of breath (signs of lung problems)
- Diarrhoea, blood in your stools, stomach pain (signs of bowel problems)
- Changes in how much you urinate (signs of kidney problems)
Do not take this medicine if you have liver disease — your doctor will have checked this before prescribing it to you.
Tell your doctor about all other medicines you take, including vitamins and herbal supplements, as some medicines can affect how this medicine works.
Reading age: approximately 12–13 years. Every technical term translated to observable symptoms. Action steps are clear. Conditional statements removed. All safety information preserved — nothing removed.
Plain language rules applied
1. Hepatotoxicity → observable symptoms (yellowing skin, dark urine, stomach pain, fatigue)
2. Immune-mediated adverse reactions → split into individual organ symptoms patients can notice
3. Pneumonitis → lung problems + observable symptom (cough, breathlessness)
4. Contraindicated → Do not take if
5. CYP3A4 inhibitors → other medicines + action step (tell your doctor)
6. Plasma concentrations → how this medicine works (consequence for the patient)
7. Bullet points for scannability
8. Bold for action items the patient must act on
2. Immune-mediated adverse reactions → split into individual organ symptoms patients can notice
3. Pneumonitis → lung problems + observable symptom (cough, breathlessness)
4. Contraindicated → Do not take if
5. CYP3A4 inhibitors → other medicines + action step (tell your doctor)
6. Plasma concentrations → how this medicine works (consequence for the patient)
7. Bullet points for scannability
8. Bold for action items the patient must act on
Every technical concept is either translated into observable patient terms or replaced with a clear patient action. Nothing was removed — all safety information is preserved. This is patient-centred writing.
Knowledge check
A PIL section reads: "Patients should ensure adequate hydration status to minimise the risk of renal adverse events." How should this be rewritten in plain language?
A"Patients should increase H2O consumption to prevent kidney problems"
B"Hydrate sufficiently to avoid adverse kidney events"
C"Drink plenty of water every day. Not drinking enough water can affect your kidneys."
DRemove it — it is too technical to explain simply
Reflection
Think of a medicine you have dispensed from your pharmacy. Find its package insert. Identify the three most technically complex sentences in the side effects section. How would you rewrite each one in plain language?
This exercise directly prepares you for real PIL writing projects.
Next module
4.1 — Pharmacovigilance basics
Week 4 · Module 1
Pharmacovigilance basics
The science of drug safety after approval — and why medical writers are essential to the entire system
Your scenario
A doctor in Chennai reports that three patients on Drug X developed an unexpected severe liver reaction never observed in clinical trials. The report reaches the pharmacovigilance team on Monday. Within 15 days it must be sent to CDSCO. Within 30 days a full narrative must be written. That narrative is written by you.
What is pharmacovigilance?
Pharmacovigilance (PV) is the science and activities relating to the detection, assessment, understanding and prevention of adverse effects of medicines. Clinical trials see only a few thousand patients over a few years. The real world has millions of patients over decades. Safety signals invisible in trials often emerge only when a drug reaches the market at scale — which is why PV never stops after approval.
The four key PV documents medical writers produce
ICSR (Individual Case Safety Report): A single adverse event report. The narrative section — which describes what happened to the patient in clear, chronological, clinical prose — is written by PV medical writers.
PSUR / PBRER (Periodic Safety Update Report / Periodic Benefit-Risk Evaluation Report): Comprehensive safety review submitted to regulators at defined intervals. Large, complex documents. Senior PV writers specialise in these.
DSUR (Development Safety Update Report): Annual safety report for drugs still in clinical trials. Submitted to regulators and ethics committees.
RMP (Risk Management Plan): A strategy document outlining how known risks will be monitored, minimised and communicated post-approval. Submitted as part of the regulatory dossier.
PSUR / PBRER (Periodic Safety Update Report / Periodic Benefit-Risk Evaluation Report): Comprehensive safety review submitted to regulators at defined intervals. Large, complex documents. Senior PV writers specialise in these.
DSUR (Development Safety Update Report): Annual safety report for drugs still in clinical trials. Submitted to regulators and ethics committees.
RMP (Risk Management Plan): A strategy document outlining how known risks will be monitored, minimised and communicated post-approval. Submitted as part of the regulatory dossier.
Serious vs non-serious adverse events — the ICH E2A definition
A Serious Adverse Event (SAE) meets at least one of these criteria:
- Results in death
- Is life-threatening
- Requires hospitalisation or prolongs existing hospitalisation
- Results in persistent or significant disability or incapacity
- Is a congenital anomaly
- Is otherwise medically significant (requires regulatory or clinical judgement)
SAEs require expedited reporting to CDSCO: 15 calendar days for unexpected SAEs. Non-serious AEs are collected and reported in periodic reports (PSUR/PBRER).
Key principle: Recovery does not determine seriousness. A patient who recovered from 3 days of hospitalisation still experienced a Serious Adverse Event.
- Results in death
- Is life-threatening
- Requires hospitalisation or prolongs existing hospitalisation
- Results in persistent or significant disability or incapacity
- Is a congenital anomaly
- Is otherwise medically significant (requires regulatory or clinical judgement)
SAEs require expedited reporting to CDSCO: 15 calendar days for unexpected SAEs. Non-serious AEs are collected and reported in periodic reports (PSUR/PBRER).
Key principle: Recovery does not determine seriousness. A patient who recovered from 3 days of hospitalisation still experienced a Serious Adverse Event.
ADR vs adverse event
An adverse event (AE) is any undesirable medical occurrence regardless of causal relationship. An adverse drug reaction (ADR) implies a causal relationship with the drug has been established or suspected — a higher bar.
In PSUR writing you report all AEs but classify ADRs separately with causality assessment.
MedDRA coding
Medical Dictionary for Regulatory Activities — standardised vocabulary for coding adverse events globally. Every AE is assigned to a specific MedDRA preferred term (PT) and system organ class (SOC).
A writer who understands MedDRA coding is significantly more valuable to a PV team than one who does not.
Causality assessment
Evaluation of whether a drug caused an adverse event. WHO-UMC and Naranjo algorithm are common tools. Classifications: certain, probable, possible, unlikely, unclassifiable.
The medical writer documents the causality assessment determined by the physician — they do not make this clinical determination themselves.
Expectedness
Whether an adverse event is listed in the current approved product information (SmPC/PIL). Unexpected SAEs trigger 15-day expedited reporting. Expected SAEs are still serious but may follow standard periodic reporting timelines.
The CASTELLAN study's velparimab label defines which adverse events are expected — any hepatotoxicity beyond the labelled rate may be unexpected.
Knowledge check
A patient takes Drug X and is hospitalised for 2 days with severe diarrhoea. The hospitalisation resolves completely and the patient recovers. Is this a Serious Adverse Event?
ANo — the patient recovered completely, so it is non-serious
BIt depends on whether the diarrhoea was classified as moderate or severe
COnly if diarrhoea is not listed as a known side effect of Drug X
DYes — hospitalisation meets the ICH E2A definition of an SAE regardless of outcome
Reflection
From your pharmacy practice: have you ever had a patient report an unexpected reaction to a medicine? Using what you now know about PV classifications — was it an AE or an ADR? Was it serious or non-serious? What should have happened next under CDSCO regulations?
Your pharmacy background is directly relevant to PV writing — this is a genuine advantage you have over most candidates.
Next module
4.2 — Biostatistics for writers
Week 4 · Module 2
Biostatistics for writers
You do not need to run the statistics. You need to read them, understand them, and write about them without error
Your scenario
The biostatistician sends you a table. It shows: ORR 43% (95% CI: 34.7–51.6), p=0.003; mOS 18.4 months (95% CI: 14.2–NE); HR for PFS 0.67 (95% CI: 0.52–0.86). You need to write three sentences describing these results accurately. Every number must be correct. Every claim must be justified by the study design. In Module 2.4, you mastered reading survival curves and forest plots; this module gives you the foundational study design hierarchy and inference rules that keep your claims scientifically defensible.
Study designs — why they determine what you can claim
The study design determines what claims you can make about the data. An observational study cannot prove causation. A single-arm trial cannot claim superiority over standard of care. A post-hoc subgroup analysis cannot be presented as a primary finding. These distinctions are not pedantic — they are the boundary between accurate reporting and misleading scientific communication.
Study design hierarchy — from strongest to weakest evidence
Systematic review / Meta-analysis — combines data from multiple RCTs. Highest level of evidence. Cochrane reviews are the gold standard.
Randomised Controlled Trial (RCT) — gold standard for establishing causation. Random allocation eliminates selection bias.
Cohort study — follows two groups over time. Can show association but not causation.
Case-control study — compares people with a disease to those without. Good for rare diseases.
Cross-sectional study — a snapshot in time. Shows prevalence only.
Case report / case series — describes individual cases. Generates hypotheses only.
As a writer: the design determines your conclusion language. RCTs "demonstrated." Observational studies "showed an association." Case reports "suggested."
Randomised Controlled Trial (RCT) — gold standard for establishing causation. Random allocation eliminates selection bias.
Cohort study — follows two groups over time. Can show association but not causation.
Case-control study — compares people with a disease to those without. Good for rare diseases.
Cross-sectional study — a snapshot in time. Shows prevalence only.
Case report / case series — describes individual cases. Generates hypotheses only.
As a writer: the design determines your conclusion language. RCTs "demonstrated." Observational studies "showed an association." Case reports "suggested."
ITT vs PP — which analysis do you report first?
Intention-to-Treat (ITT): All randomised patients analysed in their assigned group regardless of whether they completed treatment or violated the protocol. The primary analysis for all regulatory submissions. More conservative estimate — less likely to overestimate treatment benefit.
Per-Protocol (PP): Only patients who completed treatment per protocol. Usually shows a larger treatment effect because it excludes dropouts. Used as a supportive analysis only.
Regulators always want ITT first. Reporting PP first and ITT second implies you are trying to inflate the effect estimate — which triggers regulatory scrutiny.
Per-Protocol (PP): Only patients who completed treatment per protocol. Usually shows a larger treatment effect because it excludes dropouts. Used as a supportive analysis only.
Regulators always want ITT first. Reporting PP first and ITT second implies you are trying to inflate the effect estimate — which triggers regulatory scrutiny.
Sample size and confidence intervals — what they mean for writing
A study with 50 patients finding 40% response rate has very wide CIs — the true rate could be 26–56%. A study with 500 patients finding the same 40% has narrow CIs — perhaps 35.6–44.5%. The second estimate is far more reliable and precise.
As a writer: always report CIs alongside point estimates. A response rate without a CI is an incomplete result. When CIs are wide, the study is underpowered and you must describe results with appropriate uncertainty — "numerically higher" not "statistically significantly better" when the p-value does not reach significance.
As a writer: always report CIs alongside point estimates. A response rate without a CI is an incomplete result. When CIs are wide, the study is underpowered and you must describe results with appropriate uncertainty — "numerically higher" not "statistically significantly better" when the p-value does not reach significance.
NE — Not Estimable
When median survival (or other time-to-event endpoint) cannot be calculated because too few events have occurred. Common in trials with immature survival data.
Write: "Median OS was 18.4 months (95% CI: 14.2–NE) — indicating that more than half of patients had not yet experienced the event at the time of analysis."
Post-hoc analysis
An analysis not pre-specified in the Statistical Analysis Plan (SAP), conducted after the data was unblinded. Considered hypothesis-generating only — cannot be presented as a confirmatory finding.
Write: "In a post-hoc exploratory analysis, patients with [characteristic] appeared to derive greater benefit (HR 0.54; 95% CI: 0.38–0.77)." Never present as a primary finding.
Clinical vs statistical significance
A statistically significant finding (p<0.05) may not be clinically meaningful. A 0.2 mmHg blood pressure reduction can achieve p<0.001 with 10,000 patients — but no clinician would treat for this. Always report both statistical and clinical context.
Multiplicity and alpha spending
When multiple endpoints or analyses are tested, the probability of a false positive increases. Alpha spending controls the overall Type I error rate. Unadjusted p-values for secondary endpoints require careful language — often "nominal p-value" is used.
Knowledge check
A non-randomised single-arm trial reports 65% of patients responded to treatment. The investigators want to conclude "Drug X is more effective than historical chemotherapy rates of 30%." As the medical writer, what is the most accurate way to frame this conclusion?
A"The ORR of 65% appeared higher than historical chemotherapy response rates of approximately 30%, an observation warranting further evaluation in a randomised controlled setting"
BCopy the investigators conclusion — they ran the study
COmit the comparison to historical controls and report only the 65% figure
D"Drug X demonstrated superior efficacy to chemotherapy with an ORR of 65% versus historical rates of 30%"
Next module
4.3 — AI in medical writing
Week 4 · Module 3
AI in Medical Writing: The Four Obligations
Every earlier module in this course taught you to produce a document. This one teaches you to be accountable for one you did not fully write.
Data rule: unpublished data stays out of public AI
Never paste unpublished or confidential material into a general-purpose AI tool. Public and free-tier AI chatbots may keep what you enter and, depending on the provider’s terms and your settings, use it to train future models. Treat anything you type into them as disclosed. That covers unpublished results, protocols, clinical study reports, embargoed abstracts, manuscripts under review, client documents and any patient-level information. Use only AI tools your employer or client has approved under a data-processing agreement, and follow their SOP.
Week 4 · Medical Affairs — Module 4·3
AI in Medical Writing: the four obligations
Every earlier module in this course taught you to produce a document. This one teaches you to be accountable for one you did not fully write.
The 11 pm draft
A co-author returns your manuscript at 11 pm with a note: "Tidied up the results, much clearer now." The prose is excellent. Better than yours, honestly. It flows.
Three of the references do not exist. The hazard ratio in the abstract has been inverted relative to Table 2 — the text now says the drug increased risk. One sentence about grade 3 events has quietly lost the phrase "in the safety population."
A proofreader catches none of this. A plagiarism checker catches none of this. Your journal's editorial system catches none of this. Your name is on the accountability line, and the submission deadline is 9 am.
§1
Why this module is not a tool tutorial
Tools change every quarter. The model you learn today will be deprecated before you finish this course. What does not change is the structure of the problem: a machine can now produce fluent scientific prose, fluency is no longer evidence of correctness, and a named human still has to answer for the result.
So this module teaches four obligations rather than four products. Each one answers a question that a regulator, a publisher, a compliance officer or a co-author may ask you, and each one will still be the question in five years.
The one sentence to remember
An AI produces a draft. A qualified human produces a document. The distance between those two things
is your job, and it is the part of your job that is getting more valuable, not less.
§2
Obligation one — disclosure
Who used what, and where is it declared?
AI cannot be an author. This is settled, and the reasoning matters more than the rule.
Authorship requires accountability: the ability to approve the final version, to answer for its integrity, and to declare competing interests. A model can do none of these. It cannot be sued, cannot retract, cannot hold an opinion about its own funding. That is why every major body reached the same answer independently — not because AI writes badly, but because authorship is a liability position and a model cannot occupy one.
The 2026 ICMJE Recommendations added a dedicated section covering AI use by authors, editors and reviewers. Read it directly rather than relying on summaries; it is short, and journals will hold you to the text, not to a blog post about the text.
Where declaration goes — the honest answer
There is no uniform position across publishers yet. Until there is, the defensible default is:
- Methods — if AI touched anything that affects interpretation of the data (analysis, literature selection, synthesis of results).
- Acknowledgements — if AI was used for language, structure or drafting assistance only.
- Cover letter — always, in one sentence, regardless of the above. It costs you nothing and it removes any later argument about concealment.
- Target journal's own instructions — checked at submission, not at kickoff. They change.
Note the shape of the risk here. Over-disclosure has never harmed anyone's submission. Under-disclosure is how retractions start.
§3
Obligation two — verification
Can every claim be traced back to a source?
Verification is not proofreading and it is not editing. Proofreading asks whether the sentence is well-formed. Verification asks whether the sentence is true given the source documents. An AI-assisted draft can be flawless by the first standard and catastrophic by the second, and the two failure modes look nothing alike on the page.
The four failure classes you are looking for
- Fabrication — a citation, author, DOI, trial registration number or quotation that does not exist. Plausible formatting is the giveaway that there is nothing to give away.
- Transposition — a real number, moved. Arms swapped, a hazard ratio inverted, a confidence interval attached to the wrong endpoint, a percentage borrowed from an adjacent row.
- Qualifier loss — the population, the timepoint, the analysis set, the censoring rule silently deleted in the interest of a smoother sentence. This is the most common and the hardest to see.
- Confidence inflation — hedged source language ("suggests a possible association") rendered as assertion ("demonstrates"). Nothing factual has changed and the meaning has changed completely.
Three of these four are invisible to a reader who does not have the source open beside the draft. That is the whole discipline: you cannot verify from the document alone. If you are reviewing an AI-assisted draft without the TFLs, the protocol or the source papers in front of you, you are not verifying — you are reading.
Verification drill — five errors, one paragraph
Click any span you believe is wrong. Then check your work. Source data is given below the draft.
Source of truth — extract from the results tables
Randomised: 412 (drug 206, placebo 206). Primary endpoint: progression-free survival, ITT population.
Median PFS 9.4 months (drug) vs 6.1 months (placebo). HR 0.68 (95% CI 0.54–0.86), p=0.001.
Grade ≥3 treatment-emergent adverse events: 34.0% (drug) vs 29.1% (placebo), safety population.
Overall survival: immature at data cut-off; not formally tested. Reference for prior work: Okonkwo et al., Lancet Oncol 2023;24:889–901.
Median PFS 9.4 months (drug) vs 6.1 months (placebo). HR 0.68 (95% CI 0.54–0.86), p=0.001.
Grade ≥3 treatment-emergent adverse events: 34.0% (drug) vs 29.1% (placebo), safety population.
Overall survival: immature at data cut-off; not formally tested. Reference for prior work: Okonkwo et al., Lancet Oncol 2023;24:889–901.
In this phase III study, 412 patients were randomised 1:1
to drug or placebo. The primary endpoint of progression-free survival was met, with a
median of 9.4 months versus 6.1 months and a
hazard ratio of 1.68 (95% CI 0.54–0.86), indicating a
statistically significant benefit for the treatment arm.
Grade 3 or higher treatment-emergent adverse events occurred in
34.0% of patients receiving drug, compared with 29.1% on placebo.
The treatment also conferred a significant overall survival advantage,
consistent with earlier findings from
Okonkwo et al. (Lancet Oncology, 2023) and with
Marchetti et al. (New England Journal of Medicine, 2024;390:1122–34).
0 flagged
§4
Obligation three — provenance
Whose AI touched this, at which handoff?
A single publication passes through a sponsor, an agency, several external authors, sometimes a freelance writer, and a publisher. Each of those parties may be running a different AI system under a different policy with a different data-retention posture. The industry's own assessment is blunt: most publication professionals use AI, and most do not know what their external authors are using.
This is the gap that governance work is currently rushing to close, and it is the one where a writer who thinks about it early looks conspicuously competent. You do not need a policy department to act on it. You need one question asked at kickoff.
The kickoff question, and why it works
"So that we can complete the AI disclosure accurately at submission — will any AI tools be used on this draft at your end, and if so which ones?"
Framed as an administrative necessity, it is answered honestly. Framed as a compliance challenge, it is answered defensively or not at all. Record the answer in the kickoff minutes. That record is your evidence of oversight, and evidence of oversight is what you will be asked for.
Note also the confidentiality dimension. Pasting an unpublished protocol, an embargoed dataset or patient-level information into a consumer AI tool is a data breach whether or not anything goes wrong. Enterprise agreements with defined data handling exist precisely so that this question has an answer.
§5
Obligation four — discoverability
Can a machine read what you published?
The first three obligations concern AI as a writer. The fourth concerns AI as a reader, and it is the one that changes publication strategy rather than publication craft.
Physicians increasingly reach clinical evidence through retrieval-based AI platforms rather than by searching databases directly. Those systems pull from curated evidence bases and surface a synthesised answer. If your trial is behind a paywall, published without structured metadata, or licensed in a way that excludes it from retrieval, it can be excellent work that the answering system never sees.
What determines whether a machine can use your paper
- Access — open access, or at minimum an accessible abstract carrying the actual findings rather than a teaser.
- Structure — structured abstract, explicit population, intervention, endpoint and effect size in machine-parsable form. A conclusion of "the drug was well tolerated" carries no retrievable content.
- Licensing — a licence that permits inclusion in curated evidence bases.
- Redundancy — a plain language summary, a registry record and a congress abstract each give the retrieval system another route to the same finding.
This is genuinely unsettled territory. Nobody has an agreed methodology for measuring it yet, which is exactly why understanding the mechanism now is worth more than memorising anyone's current answer.
Where the field stands
loading…Still unresolved
§6
Key terms
- Hallucination
- Output that is fluent, well-formatted and false. The formatting is what makes it dangerous — fabricated citations look more correct than real ones.
- Retrieval-augmented generation (RAG)
- A system that fetches documents from a defined evidence base and answers from them. Determines whether your publication is visible to AI-mediated readers.
- Human in the loop
- A named person who reviews and accepts responsibility for output before use. A checkbox is not a loop; a documented review is.
- Provenance
- The traceable record of who did what to a document, using which tools, at which stage.
- Author AI attestation
- A recorded statement from each contributor about AI tools used on their contribution. Currently voluntary; treat it as inevitable.
- Qualifier loss
- Silent deletion of the population, timepoint or analysis set that made a claim true. The most under-detected AI editing failure.
- Source-to-text traceability
- Every factual assertion in a document mapped to a specific location in a source. The operational definition of verification.
- Generative engine optimisation (GEO)
- Structuring content so retrieval systems can find, parse and cite it. The publication-strategy consequence of AI as reader.
§7
Knowledge check
Reflection
- Return to the 11 pm scenario. Your co-author is a senior investigator who will be irritated by the suggestion that their draft contains fabrications. Write the two sentences you would actually send at 11:15 pm.
- Take a paper you have read recently. Could a retrieval system extract its primary finding from the abstract alone — population, comparator, effect size, and direction? If not, name the single sentence you would rewrite.
- Of the four failure classes in §3, which would you personally be least likely to catch, and what would you have to have open on your desk to catch it?
- Your organisation has no AI policy. You are asked to draft one page. What is the first line?
Part of Module 4·4
Authorship, accountability and AI
You have just learned the four ICMJE authorship criteria. This section explains why a machine can satisfy the first one and none of the others — and why that single fact settles most of the questions people think are still open.
Currency note — read this before teaching
The 2026 ICMJE Recommendations added a dedicated section on the use of AI by authors, editors and
reviewers, alongside updates to trial data access, funding and affiliation statements, and registration
timing. Guidance in this area is moving. Verify against icmje.org before each cohort and update
the review date in the module header.
Why AI cannot be an author
Return to the four criteria. A model can plausibly contribute to drafting — criterion two. It cannot give final approval of the version to be published, and it cannot agree to be accountable for all aspects of the work, including the accuracy and integrity of parts it did not produce. Those are not capability gaps that a better model will close. They are statements about legal and moral standing.
This is why the answer arrived so quickly and so uniformly across bodies that disagree about most other things. Authorship has never been a reward for producing text. It is an assumption of liability, and a tool cannot assume one — the same reason your statistical software is not an author of your analysis.
The consequence people miss
Because AI cannot hold accountability, all of it stays with the human authors — including for
content they did not write and did not closely read. An author who accepts an AI-generated passage has
made an assertion about its accuracy. There is no partial ownership available here, and "the tool
produced that section" has never been accepted as mitigation for anything.
Three roles, three different obligations
Authors must disclose AI use and remain fully responsible for the content. The practical question is not whether to disclose but where — and since publishers have not yet aligned on a single location, declare in the acknowledgements for language and drafting assistance, in the methods where AI affected data handling or evidence selection, and in the cover letter in every case.
Reviewers face a distinct problem that is easy to overlook: a manuscript under review is confidential. Uploading it to a general AI tool to help draft a review is a breach of that confidentiality, independent of any question about review quality. If you review for journals, this is the rule that is most likely to catch you out, because the intent is entirely benign.
Editors carry the same confidentiality obligation and, additionally, must not delegate editorial judgement to a system that cannot account for it.
GPP and the medical writer's position
Good Publication Practice already required that professional medical writers be acknowledged, that their funding source be named, and that they work under author direction. AI does not create a new principle here; it stresses an existing one. The distinction the guidance rests on — who directed the work, who approved it, who answers for it — is exactly the distinction that makes AI-assisted drafting manageable. If a writer using an AI tool remains under author direction and the authors approve the final version, the existing framework holds. What breaks it is undisclosed use, not use.
Note the mirror image of the ghostwriting problem you studied earlier. Ghostwriting hides a human contributor who should have been named. Undisclosed AI use hides a non-human contributor that could never be named — and the remedy in both cases is the same: transparency about how the document came to exist.
Author AI attestation — a usable template
Send at kickoff, collect before submission
To complete the AI disclosure accurately at submission, please confirm for your contribution to this manuscript:
- Did you use any AI tool at any stage? (Yes / No)
- If yes, which tool, and for what — literature searching, drafting, language editing, figure generation, or other?
- Was any unpublished or confidential study material entered into it?
- You remain responsible for the accuracy of all content in your contribution, including any AI-assisted text. (Confirm)
Keep the completed responses with the project file. This record is your evidence of oversight — and question 3 is the one that occasionally surfaces a real problem in time to fix it.
Knowledge check
Next module
4.4 — Freelance and career launch
Week 4 · Module 4
Freelance and career launch
Everything you need to earn your first rupee from medical writing — and build a sustainable career
Your scenario
You have completed the core clinical research, regulatory writing, safety, and AI governance curriculum. You have three portfolio pieces. You have a working vocabulary of clinical research, regulatory writing and pharmacovigilance. You know what a hazard ratio means, how to write an MSL deck, and why GDP matters. Now what? This module answers that question with specific, actionable steps for the next 90 days.
Where to find your first opportunity
India CRO list
Freelance platforms
Cold email template
Pricing guide
90-day plan
Tier 1 CROs — actively hiring medical writers in India
Siro Clinpharm — Mumbai
Veeda Clinical Research — Ahmedabad
Lambda Therapeutic Research — Ahmedabad
Synchron Research Services — Ahmedabad
Parexel India — Hyderabad
ICON plc India — Bengaluru
Syneos Health India — Hyderabad
PRA Health Sciences India — Bengaluru
Covance India — Hyderabad
Veeda Clinical Research — Ahmedabad
Lambda Therapeutic Research — Ahmedabad
Synchron Research Services — Ahmedabad
Parexel India — Hyderabad
ICON plc India — Bengaluru
Syneos Health India — Hyderabad
PRA Health Sciences India — Bengaluru
Covance India — Hyderabad
Pharma companies with India medical writing teams
Sanofi India (Hyderabad) · Novartis India (Hyderabad) · AstraZeneca India (Bengaluru) · Dr Reddy's Laboratories (Hyderabad) · Cipla Medical Affairs (Mumbai) · Biocon Medical Affairs (Bengaluru)
Hyderabad has the highest concentration of CRO and pharma medical writing roles in India. Many are within a few hours of the major southern hubs. Many of these companies accept hybrid or fully remote arrangements for writers with proven portfolios.
LinkedIn — most effective platform in India
Update your profile with IMRAD Academy certificate. Connect with HR managers and medical writing leads at your target CROs. Post a 3-sentence open-to-work statement. Comment thoughtfully on posts from pharma professionals. Recruiters search LinkedIn daily for medical writers with pharma knowledge.
Upwork
Build your profile with your three portfolio pieces as attachments. Set your rate at the lower end of market for your first 3 projects to build reviews. Set up job alerts for medical writing, clinical research writing, regulatory writing, pharmaceutical writing.
Direct email — most underused and most effective approach
Email the medical writing department directly at your target CRO. This works better than job portals — portals have 200 applicants per role. A direct email has 2. Use the template in the next tab.
Your first freelance project will most likely come from LinkedIn or a direct email — not from a job portal. Job portals are a filtering exercise, not a discovery exercise.
Cold email template — fresher to CRO (adapt for your background)
Subject: Junior Medical Writer — [Your Name] — IMRAD Academy
Dear [First Name],
I am a [M.Pharm / M.Sc] graduate with a certificate in Medical Writing and Clinical Research from IMRAD Academy. I am writing to express my interest in any junior medical writer or medical writing intern roles at [CRO name].
I have completed structured training in ICH-GCP, clinical study report writing (ICH E3), MSL slide deck development, regulatory documentation (CTD Module 2), and pharmacovigilance. My portfolio includes three pieces: a structured abstract based on a published oncology study, a 5-slide MSL deck, and a 400-word CSR Methods section.
I am available immediately and open to remote, hybrid or Hyderabad-based arrangements. I have attached my CV and portfolio pieces.
Thank you for your time.
Regards,
[Your name] | [Phone] | [Email] | IMRAD Academy
Dear [First Name],
I am a [M.Pharm / M.Sc] graduate with a certificate in Medical Writing and Clinical Research from IMRAD Academy. I am writing to express my interest in any junior medical writer or medical writing intern roles at [CRO name].
I have completed structured training in ICH-GCP, clinical study report writing (ICH E3), MSL slide deck development, regulatory documentation (CTD Module 2), and pharmacovigilance. My portfolio includes three pieces: a structured abstract based on a published oncology study, a 5-slide MSL deck, and a 400-word CSR Methods section.
I am available immediately and open to remote, hybrid or Hyderabad-based arrangements. I have attached my CV and portfolio pieces.
Thank you for your time.
Regards,
[Your name] | [Phone] | [Email] | IMRAD Academy
This email is under 200 words. It has a specific subject line. It names three portfolio pieces. It signals ICH-GCP and regulatory knowledge. It removes barriers (remote, hybrid, immediate). Send it to 20 CROs. Expect 3–5 responses. Follow up once after 10 days.
Freelance pricing guide — India 2025
Structured abstract: Rs. 1,500–3,000
Congress poster abstract: Rs. 3,000–6,000
MSL slide deck (5–10 slides): Rs. 8,000–20,000
Clinical study report section: Rs. 15,000–40,000
Regulatory synopsis (CTD 2.5 section): Rs. 30,000–60,000
Patient information leaflet: Rs. 5,000–12,000
Editing and proofreading: Rs. 500–1,000 per hour
Full Phase I CSR: Rs. 80,000–1,50,000
Congress poster abstract: Rs. 3,000–6,000
MSL slide deck (5–10 slides): Rs. 8,000–20,000
Clinical study report section: Rs. 15,000–40,000
Regulatory synopsis (CTD 2.5 section): Rs. 30,000–60,000
Patient information leaflet: Rs. 5,000–12,000
Editing and proofreading: Rs. 500–1,000 per hour
Full Phase I CSR: Rs. 80,000–1,50,000
When to increase your rates
After 3 completed projects with positive feedback · After adding a regulatory writing sample to your portfolio · After your first PMID-level publication support credit · After 12 months of consistent freelance work
Start at the lower end of each range for your first 3 projects. Build reviews and references. After your first 5 completed projects, increase rates by 25–30% and do not apologise for it.
Days 1–30: Build visibility
Polish your 3 portfolio pieces → Update LinkedIn with IMRAD Academy certificate → Connect with 50 pharma professionals on LinkedIn → Send cold emails to 20 CROs → Set up Upwork profile → Apply for 5 junior medical writer roles
Days 31–60: Generate momentum
Follow up on all unanswered emails (once, politely) → Apply for 10 more roles → Accept your first freelance project even at a lower rate → Ask for a testimonial after completion → Post your first LinkedIn article on a clinical research topic from this course
Days 61–90: Consolidate and review
Review your 6-month career action plan from Module 1.4 — have you moved toward your 3-year vision? → Complete your second and third freelance projects → Review your three portfolio pieces against the standards in this course and identify the weakest one
Most students who complete this course and take these specific actions secure their first paid opportunity within 60–90 days. The portfolio is your proof. The cold email is your door. Consistency is what turns the first one into the second.
Reviewing your own portfolio
You have completed the core curriculum and produced three portfolio pieces. Now assess them yourself, the way an agency test would: does the abstract hold every number against the source? Does the deck survive a reviewer asking where each claim comes from? Does the Methods section let someone repeat the study?
Whichever piece fails hardest is the one to rewrite first — and the course covering it is where to go next.
Whichever piece fails hardest is the one to rewrite first — and the course covering it is where to go next.
Final reflection
Look back at what you wrote in Module 1.4 — your 3-year vision and your 6-month plan. How has your thinking changed after completing the core curriculum modules? What is the one specific action you will take in the next 48 hours?
You have earned your certificate. What you do next determines everything.
Next module
4.5 — Choose your professional pathway
Week 4 · Closing Module
Choose your professional pathway
You have surveyed the landscape of clinical research and medical writing. Now it is time to choose where you go deep — the seven professional domains, what each one demands, and which matches your background and ambitions.
Where you stand
The foundational tier gave you the shared vocabulary: IMRAD structure, statistical literacy, critical appraisal, regulatory awareness, ICH-GCP, source-document reading, and scientific English. Every professional domain needs all of these. What differs is the document types you will write, the stakeholders you will serve, and the depth of specialist knowledge required.
The seven professional domains
Each domain below represents a distinct career in medical writing. Some overlap — a pharmacovigilance writer uses regulatory knowledge, and an HEOR writer needs statistics — but the core deliverables and day-to-day work differ enough that specialising matters.
Domain 1 — Regulatory Medical Writing
Core documents: Clinical Study Reports (CSRs), Investigator's Brochures (IBs), CTD Module 2 summaries (2.5 and 2.7), protocols and protocol amendments, statistical analysis plans (SAPs).
Who hires: CROs (IQVIA, Parexel, Syneos, PPD), pharma medical writing departments, specialist regulatory writing agencies.
Key skills beyond foundation: ICH E3 structure, CTD formatting, amendment tracking, regulatory submission timelines, working with biostatisticians and clinical teams.
Best fit if you: are detail-oriented, enjoy structured formats, want the highest-volume and most stable demand in medical writing.
Foundation courses that matter most: ICH-GCP, Statistics, Reading Source Documents, Critical Appraisal.
Who hires: CROs (IQVIA, Parexel, Syneos, PPD), pharma medical writing departments, specialist regulatory writing agencies.
Key skills beyond foundation: ICH E3 structure, CTD formatting, amendment tracking, regulatory submission timelines, working with biostatisticians and clinical teams.
Best fit if you: are detail-oriented, enjoy structured formats, want the highest-volume and most stable demand in medical writing.
Foundation courses that matter most: ICH-GCP, Statistics, Reading Source Documents, Critical Appraisal.
Domain 2 — Medical Communications & Publications
Core documents: journal manuscripts, congress abstracts and posters, slide decks for advisory boards, publication plans, systematic literature reviews for publications.
Who hires: medical communications agencies (Ashfield, Caudex, Oxford PharmaGenesis, AMICULUM), publication planning teams within pharma.
Key skills beyond foundation: ICMJE guidelines, GPP4 (Good Publication Practice), journal submission workflows, reference management, working with KOLs and authors.
Best fit if you: enjoy writing narrative prose, like working with academics and clinicians, want variety in therapeutic areas.
Foundation courses that matter most: Scientific English, Statistics, Critical Appraisal, Reading Source Documents.
Who hires: medical communications agencies (Ashfield, Caudex, Oxford PharmaGenesis, AMICULUM), publication planning teams within pharma.
Key skills beyond foundation: ICMJE guidelines, GPP4 (Good Publication Practice), journal submission workflows, reference management, working with KOLs and authors.
Best fit if you: enjoy writing narrative prose, like working with academics and clinicians, want variety in therapeutic areas.
Foundation courses that matter most: Scientific English, Statistics, Critical Appraisal, Reading Source Documents.
Domain 3 — Systematic Reviews & Meta-Analyses
Core documents: systematic review protocols and reports (PRISMA), meta-analysis reports, network meta-analyses (NMAs), rapid evidence assessments, literature search strategies.
Who hires: evidence synthesis consultancies (RTI Health Solutions, Mapi Group, Evidera), HTA submission teams, Cochrane review groups, academic research centres.
Key skills beyond foundation: PRISMA 2020, Cochrane methodology, search strategy design, risk-of-bias assessment (the tools you previewed in Critical Appraisal), meta-analytic statistics, GRADE framework.
Best fit if you: enjoy methodological rigour, like research more than writing, have a background in epidemiology or public health.
Foundation courses that matter most: Critical Appraisal, Statistics, Reading Source Documents.
Who hires: evidence synthesis consultancies (RTI Health Solutions, Mapi Group, Evidera), HTA submission teams, Cochrane review groups, academic research centres.
Key skills beyond foundation: PRISMA 2020, Cochrane methodology, search strategy design, risk-of-bias assessment (the tools you previewed in Critical Appraisal), meta-analytic statistics, GRADE framework.
Best fit if you: enjoy methodological rigour, like research more than writing, have a background in epidemiology or public health.
Foundation courses that matter most: Critical Appraisal, Statistics, Reading Source Documents.
Domain 4 — Medical Information & Scientific Content
Core documents: standard response documents (SRDs), medical information letters, formulary dossiers, medical FAQs, internal training materials for medical affairs teams.
Who hires: pharma medical information departments, medical affairs outsourcing providers, drug information centres.
Key skills beyond foundation: off-label enquiry handling, AMCP dossier format, therapeutic area depth, rapid literature triage, compliance with promotional/non-promotional boundaries.
Best fit if you: are a pharmacist or have a clinical background, enjoy answering specific clinical questions, prefer shorter-form writing with fast turnaround.
Foundation courses that matter most: Reading Source Documents, Critical Appraisal, ICH-GCP, Scientific English.
Who hires: pharma medical information departments, medical affairs outsourcing providers, drug information centres.
Key skills beyond foundation: off-label enquiry handling, AMCP dossier format, therapeutic area depth, rapid literature triage, compliance with promotional/non-promotional boundaries.
Best fit if you: are a pharmacist or have a clinical background, enjoy answering specific clinical questions, prefer shorter-form writing with fast turnaround.
Foundation courses that matter most: Reading Source Documents, Critical Appraisal, ICH-GCP, Scientific English.
Domain 5 — Medical Affairs & Field Medical Writing
Core documents: MSL training decks, advisory board materials, medical education slide sets, congress booth content, medical strategy documents, field medical reports.
Who hires: pharma medical affairs departments, medical education agencies, MSL teams, medical affairs consulting firms.
Key skills beyond foundation: slide design principles, storytelling with data, working with MSLs and medical directors, understanding commercial–medical boundaries, congress planning.
Best fit if you: enjoy visual communication, like working at the interface of science and strategy, are comfortable presenting to clinical audiences.
Foundation courses that matter most: Scientific English, Statistics, Critical Appraisal, Reading Source Documents.
Who hires: pharma medical affairs departments, medical education agencies, MSL teams, medical affairs consulting firms.
Key skills beyond foundation: slide design principles, storytelling with data, working with MSLs and medical directors, understanding commercial–medical boundaries, congress planning.
Best fit if you: enjoy visual communication, like working at the interface of science and strategy, are comfortable presenting to clinical audiences.
Foundation courses that matter most: Scientific English, Statistics, Critical Appraisal, Reading Source Documents.
Domain 6 — Pharmacovigilance & Safety Writing
Core documents: Periodic Safety Update Reports (PSURs/PBRERs), Development Safety Update Reports (DSURs), Risk Management Plans (RMPs), aggregate safety reports, individual case safety reports (ICSRs), signal detection summaries.
Who hires: pharma safety/PV departments, PV outsourcing companies (Accenture, Cognizant, TCS in PV), regulatory safety writing agencies.
Key skills beyond foundation: MedDRA coding, ICH E2E/E2F, signal detection methodology, benefit-risk frameworks, CIOMS guidelines, EudraVigilance and FAERS databases.
Best fit if you: are methodical and patient, have a pharmacy or clinical safety background, want steady demand with regulatory deadlines.
Foundation courses that matter most: ICH-GCP, Statistics, Critical Appraisal, Regulatory Landscape.
Who hires: pharma safety/PV departments, PV outsourcing companies (Accenture, Cognizant, TCS in PV), regulatory safety writing agencies.
Key skills beyond foundation: MedDRA coding, ICH E2E/E2F, signal detection methodology, benefit-risk frameworks, CIOMS guidelines, EudraVigilance and FAERS databases.
Best fit if you: are methodical and patient, have a pharmacy or clinical safety background, want steady demand with regulatory deadlines.
Foundation courses that matter most: ICH-GCP, Statistics, Critical Appraisal, Regulatory Landscape.
Domain 7 — HEOR & Market Access Writing
Core documents: HTA submissions (NICE, G-BA, CADTH, PBAC), AMCP dossiers, cost-effectiveness analyses, budget impact models, value dossiers, payer slide decks, real-world evidence reports.
Who hires: HEOR consultancies (IQVIA HEOR, Evidera, RTI Health Solutions, BresMed), pharma market access teams, pricing and reimbursement departments.
Key skills beyond foundation: health economic modelling concepts (Markov, partitioned survival), utility measurement (EQ-5D), ICER/QALY interpretation, NMA for indirect comparisons, country-specific HTA requirements.
Best fit if you: have a health economics, pharmacy, or public health background, enjoy working with numbers and models, want to understand how drugs get priced and reimbursed.
Foundation courses that matter most: Statistics (especially the HEOR preview), Critical Appraisal, Reading Source Documents, Regulatory Landscape.
Who hires: HEOR consultancies (IQVIA HEOR, Evidera, RTI Health Solutions, BresMed), pharma market access teams, pricing and reimbursement departments.
Key skills beyond foundation: health economic modelling concepts (Markov, partitioned survival), utility measurement (EQ-5D), ICER/QALY interpretation, NMA for indirect comparisons, country-specific HTA requirements.
Best fit if you: have a health economics, pharmacy, or public health background, enjoy working with numbers and models, want to understand how drugs get priced and reimbursed.
Foundation courses that matter most: Statistics (especially the HEOR preview), Critical Appraisal, Reading Source Documents, Regulatory Landscape.
Self-assessment: which pathway fits?
Pathway selector
Answer these five questions honestly, then check which domain they point to.
1. What kind of writing do you prefer?
A) Highly structured, template-driven (CSRs, safety reports) → Domains 1, 6
B) Narrative, persuasive (manuscripts, dossiers) → Domains 2, 7
C) Methodological, analytical (SLRs, meta-analyses) → Domain 3
D) Short-form, fast-turnaround (SRDs, training decks) → Domains 4, 5
2. Which stakeholder do you want to serve?
A) Regulators (FDA, EMA) → Domains 1, 6
B) Journal editors and KOLs → Domain 2
C) HTA bodies and payers → Domain 7
D) MSLs and medical affairs teams → Domains 4, 5
E) Researchers and methodologists → Domain 3
3. What is your background?
A) Life sciences, no clinical experience → Domains 1, 2, 3
B) Pharmacy or clinical → Domains 4, 6
C) Public health, epidemiology → Domains 3, 7
D) Health economics → Domain 7
E) Communications, journalism → Domains 2, 5
4. How do you feel about numbers?
A) Comfortable — statistics is a strength → Domains 3, 7
B) Competent — can work with them → Domains 1, 2, 6
C) Prefer words to numbers → Domains 4, 5
5. What matters most to you?
A) Job stability and high demand → Domain 1
B) Intellectual variety → Domains 2, 3
C) Impact on patient access → Domains 6, 7
D) Working with people → Domains 4, 5
A) Highly structured, template-driven (CSRs, safety reports) → Domains 1, 6
B) Narrative, persuasive (manuscripts, dossiers) → Domains 2, 7
C) Methodological, analytical (SLRs, meta-analyses) → Domain 3
D) Short-form, fast-turnaround (SRDs, training decks) → Domains 4, 5
2. Which stakeholder do you want to serve?
A) Regulators (FDA, EMA) → Domains 1, 6
B) Journal editors and KOLs → Domain 2
C) HTA bodies and payers → Domain 7
D) MSLs and medical affairs teams → Domains 4, 5
E) Researchers and methodologists → Domain 3
3. What is your background?
A) Life sciences, no clinical experience → Domains 1, 2, 3
B) Pharmacy or clinical → Domains 4, 6
C) Public health, epidemiology → Domains 3, 7
D) Health economics → Domain 7
E) Communications, journalism → Domains 2, 5
4. How do you feel about numbers?
A) Comfortable — statistics is a strength → Domains 3, 7
B) Competent — can work with them → Domains 1, 2, 6
C) Prefer words to numbers → Domains 4, 5
5. What matters most to you?
A) Job stability and high demand → Domain 1
B) Intellectual variety → Domains 2, 3
C) Impact on patient access → Domains 6, 7
D) Working with people → Domains 4, 5
Count which domain numbers appeared most often in your answers. The domain with the highest count is your natural starting point. If two domains tie, choose the one whose core documents you find more interesting — you will be reading and writing those documents for years.
Important: you do not have to choose immediately, and you can always add a second domain later. Most experienced medical writers cover two or three domains. But starting with one lets you build depth faster and makes your first portfolio more focused.
Important: you do not have to choose immediately, and you can always add a second domain later. Most experienced medical writers cover two or three domains. But starting with one lets you build depth faster and makes your first portfolio more focused.
Your 90-day action plan
After this course
Step 1 (this week): Take the IMRAD Foundations Certificate exam. It covers the seven foundational courses and validates the shared knowledge every domain requires.
Step 2 (next week): Enrol in your chosen domain pathway. Each pathway is 12–17 modules and takes 4–6 weeks of focused study.
Step 3 (weeks 3–8): Complete the domain pathway. Each one includes document-writing exercises that become portfolio pieces.
Step 4 (weeks 8–12): Build three portfolio pieces in your chosen domain. Use the templates and rubrics from the pathway. Have them reviewed — by a colleague, a mentor, or the IMRAD Academy peer review system when available.
Step 5 (week 12): Apply for your first role or project in that domain. Your certification plus portfolio pieces are what agencies and pharma companies look for.
Step 2 (next week): Enrol in your chosen domain pathway. Each pathway is 12–17 modules and takes 4–6 weeks of focused study.
Step 3 (weeks 3–8): Complete the domain pathway. Each one includes document-writing exercises that become portfolio pieces.
Step 4 (weeks 8–12): Build three portfolio pieces in your chosen domain. Use the templates and rubrics from the pathway. Have them reviewed — by a colleague, a mentor, or the IMRAD Academy peer review system when available.
Step 5 (week 12): Apply for your first role or project in that domain. Your certification plus portfolio pieces are what agencies and pharma companies look for.
Final reflection
Which domain did you choose, and why? What is the single most important thing you need to learn in that domain that this foundational course did not cover?
Write this down. When you finish the domain pathway in six weeks, come back and read what you wrote. The gap you identified is the measure of what the pathway should have filled.
You have completed the course
Take the Foundation Certification → Choose your domain pathway
IMRAD Academy — Read like a scientist. Write like a professional.
—