Research methods and study design — structured clinical viva

Drug X vs Y programme. As proposed, this is a non-randomised comparative study, not an RCT. Confounding by indication (sicker patients channelled to one drug), open-label performance/detection bias on PANSS, and completer analysis (attrition) are fatal threats for a causal therapy claim.^[1][4][7] Redesign: parallel-group RCT with random sequence, allocation concealment, blinding of outcome assessors at minimum, pre-specified primary outcome, and ITT primary analysis, reported with CONSORT.^[2][7][8] If randomisation is truly impossible, label it observational, use ROBINS-I thinking, and avoid causal marketing language.

Catchment depression and unemployment. Single-time household survey is a cross-sectional design: appropriate for prevalence if sampling and response are rigorous; association with unemployment cannot establish direction (unemployment → depression vs depression → job loss).^[1][4] Report with STROBE cross-sectional items; do not claim incidence or strong causation without longitudinal data.^[3]

Meta-analysis after systematic search. A systematic review can stop at structured narrative synthesis; meta-analysis is optional when studies are sufficiently similar clinically and statistically. PRISMA governs reporting of the review process, not a mandate to pool incompatible trials.^[5]

Follow-ups. Concealment ≠ blinding.^[8][2] Hill criteria help weigh observational associations (temporality, strength, consistency, etc.) without replacing experimental evidence when RCTs are feasible.^[6]