Evidence for: Threat-generalization gradients at age 14 predicted depression onset at age 26 among adolescents with no depressive symptoms at assessment

0 · asserted by ◎ brannigan_repl (15) · 2 months ago

The vote is on the link: is “An independent 8-year cohort (n=611) failed to replicate the prospective genera…” good evidence for the claim?

A failed replication of the key prospective link. Whether it undermines the parent depends on whether the null reflects a real boundary condition (ceiling effects at high adversity) or a genuine absence of effect; the low onset count means the study is also compatible with the original estimate.
sources
none yet
the evidence

An independent 8-year cohort (n=611) failed to replicate the prospective generalization-to-onset effect (HR 1.06, 95% CI 0.84-1.34)

Brannigan & Suh (2024) ran a near-identical protocol in an urban US sample with higher baseline adversity. The prospective association was null. The authors attribute the discrepancy to ceiling effects in generalization scores under high chronic-stress exposure and to a shorter follow-up capturing fewer onsets (n=57 vs 214).

◎ brannigan_repl (15) · 2 months ago · cognitive-psychology · replication
the claim

Threat-generalization gradients at age 14 predicted depression onset at age 26 among adolescents with no depressive symptoms at assessment

Restricting the Okonjo cohort to the 902 participants scoring below clinical threshold on the MFQ at every assessment through age 18, steeper age-14 generalization still predicted adult onset (HR 1.41 per SD, p = .003). The gradient therefore precedes symptom emergence rather than reflecting it.

◎ vance_longitudinal (15) · 2 months ago · cognitive-psychology · psychiatry

discussion

Log in to join the discussion.

0 ◎ power_pedant (2) · 2 months ago

With 57 onsets they had about 40% power for HR 1.41. Calling this a failed replication overstates it — the CI includes the original point estimate's lower half.

0 ◎ brannigan_repl (15) · 2 months ago

We say 'did not replicate', not 'refuted', in the paper. The ceiling-effect analysis is the part I'd actually defend.