Critical appraisal of an antidepressant RCT abstract (MEQ)

(i) PICO. P: adults with moderate MDD (HAM-D 18–24). I: Drug A. C: placebo. O: depressive symptom change at 8 weeks (primary continuous HAM-D); response rates should also be sought as patient-interpretable outcomes.[1]

(ii) Validity threats. Open randomisation list → failed allocation concealment → selection/allocation bias; empirically linked to exaggerated effects.[3] Completer analysis with differential dropout (28% vs 18%) → attrition bias / missing outcome data domain; ITT not used.[4][5] Continuous scale primary outcome is acceptable if blinded, but abstract does not confirm rater blinding — if open, detection bias. Headline relative claim without absolute response rates is reporting/spin concern (CONSORT expects complete outcome reporting).[6] Map to RoB 2: randomisation process (concealment), missing outcome data, measurement of the outcome, selection of the reported result.[5]

(iii) Numbers for ARR/NNT. Need event rates (or counts) for a dichotomous patient-important outcome such as response or remission in each arm, analysed preferably by ITT. ARR = CER − EER; NNT = 1/ARR. RRR alone can make small absolute benefits look large and does not communicate effort required to help one patient.[2]

(iv) Worked numbers. CER 35%, EER 50% response → if response is the good event, absolute risk increase in response = 15% = 0.15; NNT = 1/0.15 ≈ 6.7 → 7. Relative increase in response = 0.15/0.35 ≈ 43% (similar rhetoric to the abstract's "40%" marketing line). Always state duration (8 weeks) and outcome definition.[2]

(v) Applicability. External validity is limited: a comorbid substance-using outpatient was excluded, so baseline risk, adherence, and drug–drug/interaction context differ; even if internal validity were repaired, one cannot assume the same ARR — shared decision-making with explicit uncertainty and monitoring is appropriate rather than automatic adoption.[1]