The original trials underpinning SSRI approval were conducted under methodological standards that would not pass peer review today. This proposal seeks a structured reanalysis using modern meta-analytic techniques, accounting for publication bias and industry funding effects.
Kirsch's reanalysis and the Cipriani network meta-analysis point in somewhat different directions and I'd want to understand why. Publication bias correction methods have improved significantly - funnel plot asymmetry approaches have known limitations that newer Bayesian methods address. A properly specified modern meta-analysis would be genuinely useful.
The Cipriani paper is probably the best evidence we have and it still shows meaningful effects for moderate-to-severe depression. The controversy is more about whether the effects are large enough to justify first-line use for mild presentations. That's a different clinical question.
What I want to see addressed is the industry funding effect on effect size estimates. Several analyses have shown industry-funded trials report significantly larger effects than independent replications. That gap needs to be explicitly modelled.