Reviews/Ssris and Snris/Antidepressant effect size increases with baseline depression severity
· SS-MDD-03 · v2
This claim asks whether antidepressants pull further ahead of placebo as a patient's starting depression gets worse. The single largest analysis in this corpus, pooling 232 trials and 73,388 people [1], found the drug-placebo gap did widen significantly as baseline severity rose. But a patient-level reanalysis of 34 trials [7] found that this pattern holds only when trials are compared as groups and vanishes once you look at individual patients. A separate reanalysis [14] argues that the smaller apparent effect in milder depression may be an illusion created by the rating scale being unreliable at that end, not by the drug working less well. Because the two most direct analyses point in opposite directions, the panel treated the claim as genuinely unsettled rather than resolved.
The corpus contains two analyses that test the severity-by-treatment relationship directly, and they disagree. [1] reports a drug-placebo mean difference of 1.75 points (95% CI 1.63 to 1.86) that increased significantly (P<0.001) with greater baseline severity across 73,388 participants. [7] reproduces this at the aggregated trial level but finds no significant effect of initial severity on the drug-versus-placebo difference once patient-level data from 34 trials (n=10,737) are used, and notes the historical gradient was attributed to reduced placebo response in severe patients rather than greater medication response. [14] reanalysed primary data from five trials (597 individuals) and found the HDRS loses reliability as depression lessens, with 38% of the sample measured at under half maximal reliability, implying low observed effect sizes in mild and moderate depression may be a scale artifact. [17] examined initial severity as a moderator but compared antidepressant, behavioral activation and cognitive therapy, not the drug-placebo gradient. The contested point: The division is between trial-level and patient-level evidence for the same relationship: the gradient is present when trials are pooled as groups [1] and in the trial-level component of [7], but absent at the individual-patient level in [7]. Layered onto this, [14] offers a measurement explanation under which any observed gradient partly reflects scale insensitivity rather than pharmacology.
All 19 papers behind this claim with their DOIs and what each does to it, the panel’s arbitration, the evidence gap map and the open research questions. Free, and sent to your inbox.
We’ll also send you other claims from Ssris and Snris, one at a time, and tell you when a release or a pivotal trial changes a grade. Unsubscribe by replying.
One reviewer's only job is to attack the finding. The challenges below were weighed before the grade was set.
The two most direct analyses contradict each other: [7] shows the severity gradient present at trial level but absent at the patient level, implying the apparent relationship in [1] and trial-level data may reflect ecological aggregation/confounding rather than a true individual-level effect.
[14] shows the HDRS has low precision and sensitivity to change in mild and moderate depression, so smaller observed effect sizes at lower baseline severity may be a measurement artifact, not reduced pharmacological efficacy — undercutting the causal interpretation of the gradient.
Where it remains uncertain. A patient-level reanalysis [7] found this severity effect holds at trial level but disappears once individual-patient data are used, and [14] attributes smaller low-severity effect sizes to HDRS measurement unreliability.
Aleatory
[1] finds response is trimodal (Large, Non-specific, Minimal), so individual outcomes vary widely within any severity band, blurring a smooth severity-efficacy relationship.
Epistemic
[14] shows the HDRS loses precision and sensitivity in mild and moderate depression, so observed low-severity effect sizes may not reflect true efficacy.
Inferential
[7] shows the gradient at aggregated trial level but not at the individual-patient level, raising ecological-aggregation confounding in any trial-level finding.
The load-bearing papers from a corpus of 19, each classified by what it does to the claim. Every DOI is taken from the corpus the claim was graded on.
Direct primary-data support: in 232 trials and 73,388 participants the drug-placebo effect increased with baseline severity.
“Differences between drug and placebo increased significantly (P<0.001) with greater baseline severity.”
Panel auditInternal validity: high · scope of inference: Acute monotherapy for major depressive disorder in regulatory RCTs 1979-2016, adults and children.Key limitations: moderation is observational within trials · HAMD17-equivalent conversion across scales · trial populations, not clinical practice
Offers an alternative explanation: smaller observed effect sizes at lower severity may be a measurement artifact rather than true lesser efficacy.
“Low effect sizes can be expected because of the scale's low precision and low sensitivity to change, particularly in mild and moderate depression.”
Panel auditInternal validity: medium · scope of inference: Reliability of HDRS in mild-to-moderate depression trials.Key limitations: measures reliability not effect size · small reanalysis sample · argues artifact not true gradient
Direct counter-evidence: the gradient holds only at trial level and vanishes in patient-level (IPD) analysis of 34 RCTs.
“Although our trial-level data support prevous findings, patient-level data did not show any significant effect of initial depression severity on drug v. placebo difference.”
Panel auditInternal validity: high · scope of inference: Adults in NEWMEDS registry placebo-controlled antidepressant trials.Key limitations: trial-level vs patient-level discordance · registry trial populations
Weak contextual support that severity moderates antidepressant outcome, but compares modalities rather than the drug-placebo effect-size gradient.
“Among more severely depressed patients, behavioral activation was comparable to antidepressant medication, and both significantly outperformed cognitive therapy.”
Panel auditInternal validity: medium · scope of inference: Adults with major depressive disorder comparing medication and psychotherapies.Key limitations: comparator is psychotherapy not placebo · no effect-size numbers in abstract
4 of 19 papers shown. The full review lists all of them, with the panel’s arbitration →
Claims are graded at each quarterly release, and re-graded between releases only when a pivotal trial lands. Every change is a new version of the claim.
Release 2026-q4 re-graded every claim from its corpus under protocol v0.19. An earlier release's grade is shown for the record; it came from an earlier protocol and is not a movement.
Assumptions that carry weight but are not themselves established.
The panel records the gap that would most move the claim, and the study that would close it.
Reconcile patient-level versus trial-level severity gradient. The claim turns on which level is correct; [1] and [7] diverge precisely here. The panel rates this the highest-importance open gap for the claim, and names the study design that would close it: A single harmonised individual-participant meta-analysis across both trial sets testing the severity-by-treatment interaction at the patient level with trial-level fixed effects.
Open question: Does the drug-placebo severity gradient survive at the individual-patient level when all available trials are analysed together? Why it is still open: [1] and [7] reach opposite conclusions on the same relationship, [7] finding it only at trial level.
Priors is a sceptical first reader, not peer review, and not medical advice.
Release 2026-q4 · graded 26 Oct 2026 · protocol v0.19