Advanced critical appraisal — structured clinical viva

Forest plot (panel 1). Open with design and outcome. Describe weights, CI whiskers, diamond, and — critically — the prediction interval if shown. Moderate–substantial I-squared means the average effect may not predict every setting; explore clinical heterogeneity (dose, chronicity, industry, scales) before formulary mandates.^[1][9] Map key included trials to RoB 2 domains, emphasising concealment and outcome measurement for subjective scales.^[2]

NMA rankings (panel 2). Cipriani-style antidepressant NMA integrates direct and indirect evidence for efficacy and acceptability.^[3] Users’ guide priorities: network geometry, transitivity, consistency/coherence, and absolute benefits/harms — not SUCRA league tables as clinical gospel.^[4][9] Small average differences can still be policy-relevant but should not force one drug for every patient.

GRADE language. Separate certainty (may be moderate/low after RoB, inconsistency, publication concerns) from recommendation strength (often conditional for agent choice among similarly effective options when values and side-effect profiles diverge).^[5][6]

Follow-ups. Observational metabolic or suicide signals → ROBINS-I and confounding by indication.^[7] 'Works only in severe depression' → Sun subgroup credibility criteria.^[8] Non-inferiority press claims → demand pre-specified delta and CI-vs-margin.

Formulary bottom line. Do not change first-line policy on a ranking figure alone. Prefer: low RoB evidence, absolute effects, harm profiles, local acquisition/cost, monitoring capacity, and shared decision-making — with honest certainty language.^[5][6][9]