How do you run a fair trial?

Definition: allocation of participants to intervention groups by a chance process, so that neither the participant nor the investigator can choose or predict the group.

Why it is done:

• Removes selection bias: the investigator cannot send milder cases to the new treatment.
• Balances known and unknown prognostic factors between the groups, on average. Observational studies can adjust only for factors that were measured.
• Allows allocation concealment and blinding, and gives a valid basis for statistical tests.

Clinical example: a physician gives a new antihypertensive to younger patients with milder hypertension. Their blood pressure falls more, partly because of who they are. Randomising patients to the new or the standard drug removes the physician's choice.

Process: generate the sequence (by someone not recruiting) → conceal it (central telephone or web system, or sequentially numbered, opaque, sealed containers) → enrol the patient, then reveal the allocation → report the method (CONSORT 2025, items 17a to 19).

Not randomisation: alternate patients, date of birth, hospital number, day of the week. The 1948 MRC streptomycin trial used random sampling numbers in sealed envelopes.

• Allocation: assigning each participant to an intervention group, in a stated ratio (1 : 1, 2 : 1).
• Allocation concealment: a process that “keeps clinicians and participants unaware of upcoming assignments” (Schulz and Grimes, 2002). Central web or telephone randomisation, or sequentially numbered, opaque, sealed containers. Inadequate concealment exaggerated odds ratios by 41% (Schulz et al., 1995).
• Randomisation: allocation by a chance process that cannot be predicted. Simple, block, stratified, minimisation, cluster.
• Stratification: randomising separately within subgroups defined by a strong prognostic factor (centre, severity), usually with blocks, so each subgroup is balanced. Most useful in small trials.
• External validity: whether results “can be reasonably applied to a definable group of patients in a particular clinical setting in routine practice” (Rothwell, 2005). Depends on setting, patient selection, intervention delivery, outcomes and follow-up. PRECIS-2 places a trial between explanatory and pragmatic.
• Sensitivity analysis: repeating the analysis under different assumptions (different handling of missing data, per-protocol set, excluding protocol violations) to see whether the conclusion changes. Prespecified and post hoc sensitivity analyses are reported separately (CONSORT 2025, item 21d).
• Fragility Index: in the group with fewer events, the number of patients whose status must change from non-event to event before p rises above 0.05. For binary outcomes. Walsh et al. (2014) found a median of 8 in 399 trials, with 25% at 3 or less, and in 53% of trials it was smaller than the number lost to follow-up.

Definition: keeping one or more groups of people unaware of the intervention a participant received, after allocation. Also called masking (ICH E9, section 2.3.1).

Purpose: prevents differences in care and co-interventions (performance bias), in outcome assessment (detection bias), in reporting of symptoms by patients, and in withdrawals between groups. Most important for subjective outcomes such as pain, nausea or quality of life.

• Open label: everyone knows.
• Single blind: the participant does not know, or the investigator does not (ICH E9).
• Double blind: neither the participant nor the investigators and sponsor staff involved in treatment or evaluation know (ICH E9).
• Triple blind: no fixed definition, usually adds outcome assessors or data analysts. Physicians give many different meanings to these labels (Devereaux et al., 2001), so state who was blinded (CONSORT 2025, item 20a).

Double-blind procedure:

1. An independent statistician generates the randomisation list. Pharmacy or the sponsor prepares identical packs (appearance, taste, smell, packaging), labelled only with trial numbers.
2. If the drugs differ in form, use the double-dummy technique.
3. Each enrolled patient receives the next numbered pack. The code is held centrally or in sealed individual code envelopes.
4. Investigators, patients and outcome assessors record data without knowing the allocation. Analysts work with groups labelled A and B.
5. The database is cleaned and locked before the code is broken for analysis.
6. Report who was blinded, how, and any unblinding.

Unblinding, and when it is allowed: for an individual patient when knowing the treatment is essential for care, such as a serious adverse event (ICH E9, ICMR 2017 section 7.4.1). For a planned interim analysis, by an independent data monitoring committee only. For all at the end of the trial. Every break of the blind, planned or accidental, is recorded and explained in the report.

Limitations: not possible for many surgical or device interventions, and distinctive adverse effects can reveal the drug. In those cases blind the outcome assessor and use objective outcomes.

Methods of randomisation are in the first outline above.

• Randomisation: chance allocation that removes selection bias and balances prognostic factors. Methods: simple, block, stratified (see the first outline).
• Placebo: an inert preparation identical in appearance to the active drug.
• Uses: separates the drug effect from spontaneous change, the placebo effect, concomitant treatment and observer expectations (ICMR 2017), and makes double blinding possible.
• Equipoise: randomising patients is ethical only when there is genuine uncertainty within the expert medical community about which treatment is better (clinical equipoise, Freedman, 1987).
• Ethics: the Declaration of Helsinki (2024, paragraph 33) requires comparison with the best proven intervention unless none exists, or unless there are compelling scientific reasons and no added risk of serious or irreversible harm.
• ICMR 2017 conditions: no established effective therapy, withholding therapy causes only temporary discomfort, self-limited disease, or an active comparator would not give reliable results.
• Safeguards: add-on design, 2 : 1 randomisation to reduce placebo exposure, rescue medication, clear withdrawal criteria, close monitoring, and ethics committee review.

Definition: an experimental study in which participants are allocated at random to an intervention or a control group and followed up for outcomes.

Types: by design (parallel, crossover, factorial, cluster), by comparison (superiority, non-inferiority, equivalence), by purpose (therapeutic, preventive such as vaccine trials, screening, diagnostic), by blinding (open, single, double), by setting (clinical trials in patients, field trials in healthy people, community trials in whole communities), by attitude (explanatory, pragmatic).

Steps, with an example comparing Drug A and Drug B in stage 1 hypertension:

1. Question: in adults with stage 1 hypertension, does Drug A lower 24-hour ambulatory systolic BP at 12 weeks more than Drug B?
2. Protocol: eligibility, interventions and doses, primary and secondary outcomes, superiority or non-inferiority, sample size from the SD of BP change and the smallest important difference.
3. Ethics committee approval, then CTRI registration before the first patient.
4. Consent and enrolment. Baseline BP measured with a validated device.
5. Randomisation stratified by baseline BP, in blocks of varying size, with central concealed allocation.
6. Double-dummy blinding if the tablets look different. BP measured by an observer unaware of allocation.
7. Follow-up at fixed visits, adherence and adverse events recorded.
8. Intention-to-treat analysis of the difference in BP change with its 95% confidence interval, per-protocol as a secondary analysis.
9. Report with the CONSORT 2025 checklist and flow diagram.

Control and outcome: use a concurrent randomised control (placebo, active treatment, usual care or sham), not historical controls, which favour the new treatment (Sacks et al., 1982). Choose one patient-important primary outcome, and a surrogate only when it is accepted as a reliable predictor of clinical benefit (ICH E9). In CAST (1989), drugs that suppressed ventricular premature beats increased mortality.

Advantages: controls selection bias and confounding, allows blinding, strongest single-study evidence for cause and effect. Limitations: cost and time, ethical limits on what can be randomised, selected patients may limit external validity, rare and late harms are missed.

1. RCT: definition, design diagram (eligible population → consent → randomise → intervention and control → follow-up → outcome), types and steps as in the outline above.
2. Randomisation techniques: simple, block, stratified, minimisation, cluster, with advantages and limitations as tabulated in the first outline. Add concealment and its methods.
3. CONSORT 2025: 30-item checklist and flow diagram. Sections: title and abstract, open science (registration, protocol, data sharing, funding), introduction, methods (design, eligibility, interventions, outcomes, sample size, randomisation items 17a to 19, blinding 20a and 20b, statistics 21a to 21d), results (flow 22a and 22b, baseline 25, outcomes 26, harms 27), discussion (interpretation 29, limitations 30).
4. Draw the flow diagram: enrolment, allocation, follow-up, analysis.

What: Consolidated Standards of Reporting Trials. First published 1996, updated 2001, 2010 and 2025. CONSORT 2025 is a 30-item checklist and a participant flow diagram (Hopewell et al., 2025).

Significance: readers can judge a trial only if methods and results are reported fully. CONSORT gives the minimum set of items, is endorsed by ICMJE and many journals, and helps editors, reviewers and guideline writers appraise reports. It is a reporting guideline, not a quality score.

Components:

• Title and abstract (1a, 1b).
• Open science: registration (2), protocol and statistical analysis plan (3), data sharing (4), funding and conflicts (5a, 5b).
• Introduction: background (6), objectives (7).
• Methods: patient and public involvement (8), design (9), changes (10), setting (11), eligibility (12a, 12b), interventions (13), outcomes (14), harms (15), sample size and interim analyses (16a, 16b), sequence generation (17a, 17b), concealment (18), implementation (19), blinding (20a, 20b), statistical methods (21a to 21d).
• Results: participant flow (22a, 22b), recruitment (23a, 23b), intervention delivery (24a, 24b), baseline data (25), outcomes and estimation (26), harms (27), ancillary analyses (28).
• Discussion: interpretation (29), limitations (30).

Flow diagram: enrolment (assessed for eligibility, excluded with reasons, randomised), allocation (allocated, received or did not receive intervention), follow-up (discontinued, lost to follow-up, with reasons), analysis (analysed, excluded with reasons), for each group.

Relevance to the protocol: the companion SPIRIT 2025 statement lists protocol items. Planning randomisation, concealment, blinding, outcomes and analysis sets in the protocol makes CONSORT reporting straightforward. Extensions exist for cluster, crossover, non-inferiority and equivalence, pragmatic and other designs.

• Phase 0 (exploratory IND): very small subtherapeutic doses in a few people to check pharmacokinetics and target effect before larger trials.
• Phase I (human pharmacology): first administration in humans, usually healthy volunteers, or patients for toxic drugs such as cytotoxics. Tolerance, safe dose range, pharmacokinetics and pharmacodynamics. Dose-escalation designs.
• Phase II (therapeutic exploratory): see below.
• Phase III (therapeutic confirmatory): larger randomised controlled trials, often multicentre, to confirm benefit and safety in the intended population. Basis for marketing approval.
• Phase IV (therapeutic use): after approval. Uncommon adverse effects, effectiveness in practice, new populations, drug interactions, cost. If none of n patients in earlier trials had an adverse event, its risk can still be as high as about 3 ÷ n (rule of three, Hanley and Lippman-Hand, 1983).
• Dose finding: up-and-down designs change each patient's dose by a fixed step according to the previous patient's response and estimate the ED50 with few patients (Dixon and Mood, 1948). Biased coin or continual reassessment designs are preferred for the ED90 or ED95 (Görges et al., 2017).

Phase II in detail:

• First studies whose main aim is to explore therapeutic efficacy in patients (ICH E8).
• Early phase II (often called IIa): does the drug show activity in the disease? Later phase II (IIb): which dose and regimen?
• Designs: early studies may use concurrent controls or comparison with baseline status. Later studies are usually randomised and concurrently controlled (ICH E8). Dose-ranging studies compare several doses.
• In cancer, single-arm trials measure response rate. Simon's two-stage design (1989) keeps the expected number of patients small when the drug has low activity. Randomised phase II designs are also used.
• Outcomes are often short-term or surrogate markers. Results decide whether phase III is justified, and supply the dose and effect size for its sample size.

Phases describe the usual order of development. ICH E8 notes that study types and phases are not the same thing.

Phases in cancer:

• Phase 0: target modulation in tumours at subtherapeutic doses, to decide early whether development should continue (Kummar et al., 2008).
• Phase I: done in patients with cancer, not healthy volunteers, because the drugs are toxic. Dose escalation in small cohorts to find dose-limiting toxicity and the recommended phase II dose.
• Phase II: response rate in a specific cancer, often single-arm (Simon two-stage) or randomised phase II.
• Phase III: randomised against standard treatment, with overall survival, progression-free survival or quality of life as outcomes.
• Phase IV: long-term and uncommon toxicity, use in wider populations.

Advantages of Indian participation: earlier access to new treatments for patients, data on how the drug works in Indian patients before approval here, training and infrastructure for Indian investigators and sites, faster recruitment.

Disadvantages and concerns: risk of enrolling poor or less literate patients without truly informed consent, a control arm that may be below the local best standard, uncertain access to the drug after the trial, disputes over injury compensation, and a drug that may be unaffordable in India once approved.

Safeguards in India: ethics committees registered with the Central Licencing Authority (NDCT Rules 2019, rules 7 to 9), CTRI registration before the first participant (rule 25), free medical management and compensation for trial-related injury or death (Chapter VI), audio-video recording of consent for vulnerable participants in trials of new chemical or molecular entities (Third Schedule), and post-trial provisions arranged in advance (Declaration of Helsinki 2024, paragraph 34). The Rules' preamble notes that the Supreme Court, in writ petition (civil) 33 of 2012, observed that new clinical trial rules should be finalised urgently.

Why randomised trials are needed in oncology:

• Survival depends strongly on stage, performance status and biology. Surgeons and oncologists choose fitter patients for new treatments, so non-randomised comparisons overstate benefit.
• Historical controls differ in staging methods, imaging and supportive care.
• Most new treatments give modest gains, which only large randomised trials can detect reliably.
• Response rates and other surrogate outcomes do not always translate into longer survival.
• New treatments add toxicity and cost, which must be weighed against a measured benefit.

Impact: randomised trials decide the standard of care and guideline recommendations, and stop ineffective or harmful treatments from spreading.

Intention to treat (ITT): every randomised participant is analysed in the allocated group, irrespective of adherence (ICH E9). Preserves randomisation, reflects the effect of offering the treatment, and is conservative in superiority trials. Needs complete follow-up and a stated method for missing data.

Per protocol (PP): only participants who met eligibility, received a minimum exposure and had no major protocol violations. Estimates the effect when taken as intended, but can be severely biased because adherence is related to treatment and prognosis.

Worked example (made-up numbers): Drug A event rate 24/100 by ITT and 16/80 by PP, control 30/100 and 27/90. Risk ratio 0.80 by ITT and 0.67 by PP. Excluding the sicker non-adherent patients from Drug A makes it look better.

Which to use: superiority trials, ITT primary with PP as secondary. Non-inferiority and equivalence trials, both, because ITT is not conservative there.

Modified ITT (mITT): ITT with some randomised patients excluded, for example those who took no dose, had no post-baseline assessment or did not have the target condition confirmed. There is no single definition. In 475 trials reporting mITT, 80% excluded patients after randomisation and descriptions were ambiguous (Abraha and Montedori, 2010). Report exactly who was excluded and why (CONSORT 2025, item 21b), and show whether results change.

1. When: an effective standard exists and the new treatment offers another advantage (less toxic, oral, cheaper). A placebo control would be unethical.
2. Active comparator: proven effective in earlier superiority trials. Keep their important design features: outcome, dose, eligibility (ICH E9).
3. Margin (Δ): the largest clinically acceptable difference, smaller than the effect seen for the comparator in superiority trials, justified clinically and fixed in the protocol (ICH E9).
4. Hypotheses: H₀, the new treatment is worse than the standard by Δ or more. H₁, it is worse by less than Δ (or better).
5. Analysis: one-sided confidence interval (commonly the lower limit of a two-sided 95% CI). Non-inferiority is shown when the lower limit lies above −Δ.
6. Analysis sets: full analysis set and per-protocol set both reported. ITT is not conservative in this design.
7. Sample size: based on the margin, the expected true difference, α and power. ICH E9 warns that assuming a zero true difference underestimates the size needed if the new drug is slightly less effective.
8. Reporting: CONSORT extension for non-inferiority and equivalence trials (Piaggio et al., 2012). CONSORT 2025 item 9 asks for the framework.

For a binary outcome, assuming equal true cure rates p in both groups:

n per group = (Z_1−α + Z_1−β)² × 2p(1 − p) ÷ Δ²

Z 1−α

1.96 for a one-sided α of 2.5%

Z 1−β

0.842 for 80% power

p

expected cure rate in both groups

Δ

non-inferiority margin, as a proportion

Teaching example: p = 0.85, Δ = 0.10 → (1.960 + 0.842)² × 2 × 0.85 × 0.15 ÷ 0.10² = 200.2 → 201 per group

Allowing for 10% dropout: 201 ÷ 0.9 = 223.3, so recruit 224 per group.

• Null hypothesis (H₀): the statement the trial tries to reject. The alternative (H₁) is what the trial hopes to show.
• Superiority trial: H₀, no difference between new treatment and control. H₁, a difference exists.
• Non-inferiority trial: H₀, the new treatment is worse by the margin Δ or more. H₁, it is worse by less than Δ.
• Equivalence trial: H₀, the difference is Δ or more in either direction. H₁, the difference lies within ±Δ. Tested as two one-sided tests (ICH E9).
• Type I error (α): rejecting a true H₀. In a superiority trial, declaring an ineffective drug effective. In a non-inferiority trial, accepting an inferior drug.
• Type II error (β): failing to reject a false H₀, such as missing a real benefit. Power = 1 − β.
• A non-significant superiority result does not prove H₀. Look at the confidence interval.
• Multiple outcomes, subgroups and interim looks increase the chance of a false positive unless planned for.
• Five analyses of accumulating data, each at p < 0.05, give a 14.2% chance of at least one false-positive result when the treatments are equal. Group sequential boundaries, such as O'Brien-Fleming, keep the overall rate at 5%.

See the hypothesis testing chapter.

Confounding: a factor linked to both the treatment received and the outcome. Randomisation distributes known and unknown confounders equally on average, which observational studies cannot do. In small trials chance imbalance can still occur. Reduce it with stratified randomisation or minimisation, and adjust for prespecified prognostic factors in the analysis. Per-protocol exclusions reintroduce confounding.

Historical controls: patients treated in an earlier period differ in selection and care. For the same six therapies, 79% of trials with historical controls but 20% of randomised trials found the new therapy better (Sacks et al., 1982).

See the bias and validity chapter.

Definition: public recording of a trial's key details (design, interventions, primary outcome, sample size, sponsor, ethics approval) in a registry before the first participant is enrolled.

Importance:

• Reduces publication bias: trials with unfavourable results cannot disappear without trace.
• Shows whether primary outcomes or analyses were changed after the results were known.
• Avoids duplicate trials and helps patients and doctors find ongoing trials.
• Required by journals (ICMJE, for trials starting on or after 1 July 2005), by the Declaration of Helsinki (2024, paragraph 35), by ICMR 2017 and by Indian law.

Registries: WHO International Clinical Trials Registry Platform and its primary registries (including CTRI), ClinicalTrials.gov.

CTRI in India:

• Launched by ICMR on 20 July 2007. Free online registry. WHO primary registry since December 2008.
• Mandatory for regulatory trials since 15 June 2009. The NDCT Rules 2019 (rule 25) require registration before the first subject is enrolled.
• Only prospective registration since 1 April 2018.
• Registers interventional, observational, BA/BE, post-marketing, AYUSH and postgraduate thesis studies.
• Needs ethics committee approval and DCGI approval where applicable. Records Indian sites, investigators, ethics committees, randomisation, concealment, blinding and outcomes. Updates remain publicly visible.

How CTRI improves transparency and accountability: named investigators, sites and ethics committees are public. CTRI checks trial details against the ethics approval, and this review has identified problems such as multicentre trials approved by one distant ethics committee and one investigator running many trials at once (Vardhana Rao et al., 2018). ICMR signed the WHO joint statement on results disclosure in May 2017. Limitations: analyses of CTRI records have found missing, inconsistent and non-standard entries (Pillamarapu et al., 2019).

• Importance: too small misses a real effect (type II error) and exposes patients without producing an answer. Too large wastes resources and gives more patients the inferior treatment. Ethics committees and CONSORT 2025 (item 16a) require the calculation and its assumptions.
• Inputs: primary outcome, design (superiority or non-inferiority margin), α, power, smallest important difference, SD or control event rate, allocation ratio, dropout rate.
• Two means: n per group = 2(Z_α/2 + Z_β)² × SD² ÷ d². Two proportions: the two-proportion formula. Non-inferiority: based on the margin (see the outline above).
• ITT dilution: ICH E9 notes that if the primary analysis uses the full analysis set, the expected effect may need to be reduced to allow for patients who stop treatment.
• Adjust: divide by (1 − dropout rate), multiply by the design effect for cluster trials, and round up.
• Changing the size during the trial: planned sample size re-estimation when the SD or event rate differs from the assumption (ICH E9, section 3.5), or a group sequential design, whose actual size depends on when a stopping boundary is crossed, up to a planned maximum.

Formulas and worked examples are in the sample size chapter.

• Define the target population and eligibility criteria.
• Recruitment is usually consecutive enrolment of eligible, consenting patients, a non-probability method.
• Random allocation (simple, block, stratified) then assigns treatment, which provides internal validity.
• External validity improves with broad eligibility, several centres and a record of screened and excluded patients (CONSORT flow diagram).
• Report settings and locations (CONSORT 2025, item 11) and eligibility (item 12a).

See the sampling chapter.

See the study designs and systematic review and meta-analysis chapters.

• Phases: safety and immunogenicity in small groups (I, II), then a large efficacy trial (III), then post-licensure surveillance (IV).
• Participants: usually healthy people, often children, so safety standards are high. Parental consent and assent of older children.
• Outcome: laboratory-confirmed disease with a clear case definition, found by active surveillance for an adequate period.
• Sample size: the disease is uncommon, so large numbers or high-incidence areas are needed.
• Unit of randomisation: individuals for direct protection. Clusters (villages, schools) when indirect protection of unvaccinated contacts is to be measured, with a design effect.
• Control: placebo or an unrelated vaccine in identical vials, keeping participants and assessors blinded. A placebo is unacceptable where an effective vaccine is already recommended.
• Vaccine efficacy: VE = (ARU − ARV) ÷ ARU × 100, where ARU and ARV are attack rates in unvaccinated and vaccinated groups. Teaching example: ARU 2%, ARV 0.5% → VE 75%.
• Safety: recording of adverse events following immunisation, an independent data and safety monitoring board, stopping rules.
• Implementation: ethics approval, regulatory permission, CTRI registration, cold chain, trained field staff, community engagement, ITT analysis, CONSORT reporting.

Conduct:

1. Question with a patient-important primary outcome (mortality, ventilator-free days). Sample size, often large for mortality.
2. Protocol and analysis plan. Consider a multicentre network to reach numbers.
3. Ethics committee approval. For a new indication, route, dose or form of an approved drug used only for research, it may be an academic clinical trial that needs no permission from the Central Licencing Authority (NDCT Rules 2019, rule 28). Otherwise regulatory permission is needed.
4. CTRI registration before the first patient.
5. Consent: most ICU patients cannot consent, so consent comes from the legally acceptable representative (ICMR 2017). For vulnerable participants in trials of new chemical or molecular entities, the consent process is audio-video recorded (NDCT Rules 2019, Third Schedule).
6. Central web randomisation, stratified by centre and severity, with concealment available at any hour.
7. Blinding where possible, otherwise blinded outcome assessment and objective outcomes.
8. Serious adverse event reporting, a data and safety monitoring board, ITT analysis, CONSORT reporting.
9. In investigator-initiated research the institution acts as sponsor and is responsible for compensation or insurance for research-related injury (ICMR 2017, section 2.6.4).

Issues and practical limitations: time-critical enrolment at night, heterogeneous patients and variable usual care, co-interventions, difficulty blinding procedures and devices, high mortality needing large samples, limited funding and research staff, clinical workload of investigators, loss to follow-up after discharge, and delays in ethics and registration processes.

1. Define the event: for example virological failure or death. Time origin is the date of randomisation.
2. Record for each patient: time to event, or time to censoring (lost to follow-up, withdrawn, or event-free at the end of the study).
3. Kaplan-Meier estimate: at each event time, the probability of remaining event-free = (number at risk − number of events) ÷ number at risk. Multiply these probabilities in sequence to obtain the cumulative survival.
4. Plot a step curve for each arm, with numbers at risk shown below the time axis.
5. Compare arms with the log-rank test. Estimate the hazard ratio with its 95% CI using a Cox model.
6. Report median time to failure and event-free proportions at fixed times (for example 48 weeks), by intention to treat.
7. Assumption: censoring is unrelated to prognosis. Heavy loss to follow-up in one arm biases the curves.

See the risk and survival chapter.

Constitution (NDCT Rules 2019, rule 7): at least seven members from medical, non-medical, scientific and non-scientific areas, with at least one lay person, one woman member, one legal expert and one independent member from a related field (social scientist, representative of a non-governmental voluntary agency, philosopher, ethicist or theologian). At least 50% of members are not affiliated with the institution.

Registration: the ethics committee applies to the Central Licencing Authority in Form CT-01 (rule 8). Registration is granted in Form CT-02 and is valid for five years unless suspended or cancelled (rule 9). Renewal is applied for in Form CT-01 ninety days before expiry (rule 10). Functioning follows the ICMR National Ethical Guidelines 2017.

Member Secretary (ICMR 2017): a staff member of the institution with knowledge of clinical research and ethics.

• Receives, screens, circulates and maintains proposals, and categorises them for exemption, expedited review or full committee review.
• Schedules meetings, prepares the agenda and minutes.
• Organises documentation, communication and archiving.
• Ensures training of members and the secretariat, and keeps SOPs up to date.

Legal expert (ICMR 2017): holds a basic degree in law, with experience. Reviews the informed consent document and translations, memoranda of understanding, clinical trial agreements, regulatory approvals and insurance documents, and informs members about new regulations.

Details are in the research ethics chapter.

Payment for participation (ICMR 2017, section 2.5): reimbursement of travel and incidental expenses, and payment for time, inconvenience and lost wages, in cash or kind. Participants do not pay for research-related tests or treatment. The ethics committee approves the amounts and checks that they are not an undue inducement.

Compensation for research-related harm (ICMR 2017, section 2.6): participants harmed by participation are entitled to financial or other assistance, and dependants to compensation in case of death. This applies to all arms, including control. In investigator-initiated and student research the institution acts as sponsor.

Regulatory trials (NDCT Rules 2019, Chapter VI):

• Free medical management for injury, as long as required or until the injury is shown to be unrelated to the trial (rule 40).
• Financial compensation for trial-related injury, permanent disability or death, paid by the sponsor. In death, to the legal heir (rule 39).
• Serious adverse events are reported by the investigator within 24 hours. The ethics committee gives its opinion, an expert committee examines deaths, and the Central Licencing Authority decides the amount (rule 42).
• Formula for death (Seventh Schedule): compensation = (B × F × R) ÷ 99.37, where B is the base amount of ₹8 lakh, F an age factor, and R a risk factor from 0.5 to 4.0 depending on the severity of the disease at enrolment. For non-permanent injury, the amount matches lost wages.

Principles: compensation depends on whether the injury is related to the trial. The informed consent document must describe medical management and compensation, and the sponsor gives an undertaking to pay before permission is granted. Full details are in the research ethics chapter.

Ethical issues:

• Scientific justification and adequate preclinical and earlier-phase data, ethics committee approval and regulatory permission.
• Informed consent in the local language, voluntary participation and the right to withdraw.
• Choice of control: active comparator where effective NSAIDs exist. Placebo only with rescue analgesia and short exposure (Declaration of Helsinki 2024, paragraph 33).
• Exclusion or careful monitoring of people at high risk of gastrointestinal bleeding, renal impairment or cardiovascular disease. Monitoring of blood pressure, renal function and gastrointestinal symptoms.
• Serious adverse event reporting, a data and safety monitoring board, stopping rules, free treatment and compensation for trial-related injury.
• Confidentiality, CTRI registration, and publication whatever the result.
• Clinical equipoise: genuine uncertainty in the expert medical community that the new NSAID is better or safer than the comparator (Freedman, 1987).

Double-blind crossover trial:

1. Suited to a stable chronic condition, such as osteoarthritis pain, and a drug with a short-lived effect.
2. Randomise each patient to a sequence: new NSAID then comparator, or comparator then new NSAID.
3. Treatment period 1, then a washout period long enough for the first drug's effect to disappear, then treatment period 2.
4. Identical capsules, or double dummy, so neither patient nor investigator knows the drug in either period.
5. Measure pain and function at the end of each period.
6. Analyse within-patient differences, and check for period effects and carry-over effects. In the 2 × 2 design carry-over cannot be statistically distinguished from the treatment-by-period interaction (ICH E9), so prevent it with an adequate washout.

Advantages: each patient is their own control, so fewer patients are needed. Limitations: carry-over, dropouts between periods, unsuitable for curable conditions or permanent outcomes.

• Peer review: evaluation of a manuscript by independent experts in the field before the editor decides to accept, revise or reject it. It checks originality, methods, analysis, interpretation and ethics.
• Single-blind (simple blind): reviewers know the authors' identity, but authors do not know the reviewers. Reviewers can write freely, but they may be influenced by the authors' reputation, institution or country.
• Double-blind: neither authors nor reviewers know each other. It reduces bias based on author identity, but anonymity is hard to maintain in small fields and when authors cite their own work.
• Other models: open peer review, where identities or reports are published.
• The word “blind” here means the same as in trials: unaware of identity.

See the publishing and IPR chapter.

Here “blinding” means causing blindness, not trial blinding.

1. Confirm the diagnosis: clinical examination, imaging and electrophysiology, and genetic testing where available.
2. Draw a pedigree of at least three generations, asking about consanguinity.
3. Establish the inheritance pattern: autosomal dominant, autosomal recessive, X-linked or mitochondrial.
4. Explain the risk to future children: autosomal dominant, 50% for each child of an affected parent. Autosomal recessive, 25% for each child of two carriers. X-linked recessive, 50% of sons of a carrier mother affected and 50% of daughters carriers. Mitochondrial, passed through the mother.
5. Discuss options: carrier testing of relatives, prenatal or preimplantation genetic testing, and screening of at-risk relatives for early treatment.
6. Counsel without directing the couple's decision, maintain confidentiality, and give psychological support.
7. Plan care: low-vision aids, rehabilitation and education support, and follow-up as new tests or treatments become available.