In the biologically oldest quartile of NHANES adults, depression risk was 45% lower — not higher — than in the quartile just below it (OR = 0.55, 95% CI 0.49–0.63).
That number should stop the pitch cold. Because the pitch is seductive: draw blood from a 55-year-old, run a cheap composite clock, and tell them two things at once — whether depression is coming, and whether it will kill them faster if it does.
That is roughly what two new cohort papers are being sold as. Read at face value, they deliver.
The first is a NHANES analysis of 30,885 adults using GOLD BioAge, a Gompertz-law composite built from routine clinical biomarkers. Among 2,658 depressed participants, each one-year increase in biological age raised all-cause mortality 7% (HR = 1.07, 95% CI 1.06–1.08), with a clean linear dose-response and survival curves that fanned apart by tertile.
The second, from the English Longitudinal Study of Ageing, followed 2,813 depression-free adults over 50 for eight years. Each one-standard-deviation rise in the CRP-triglyceride-glucose index — an inflammatory-metabolic composite — raised the risk of incident depressive symptoms 25% (HR = 1.253, 95% CI 1.002–1.479), after adjusting for BMI.
So a blood-based clock flags who dies, and a blood-based inflammation-metabolism score flags who gets depressed. The screening story writes itself.
Except the GOLD BioAge depression result refuses to behave. The association with depression was significantly nonlinear (p < 0.001). Quartiles 1 and 2 showed no difference from Q3. And Q4 — the oldest — was protective.
That is not what a smooth "aging drives depression" mechanism predicts. It's the shape you get from survivorship, reverse causation, or a marker that means something different at its extremes. The authors are honest enough to frame the tool as "risk stratification," not causation. Anyone selling it as an early-warning depression test is ignoring the paper's own curve.
A marker that is protective against depression in its oldest quartile is not ready to counsel a 55-year-old about their mental-health future.
Which forces the real question. In these paired outcomes — aging and depression — which one is upstream?
The strongest prospective signal for aging-as-risk-factor comes from the UK Biobank, 424,299 participants, the largest study in this literature. Biologically older people at baseline had more prevalent depression and anxiety, and over a median 8.7 years, older biological age predicted incident depression/anxiety: a 5.9% increase per SD of KDM-BA acceleration and 11.3% per SD of PhenoAge acceleration.
That effect was independent of, and additive to, genetic risk from polygenic scores. This is genuine prospective evidence — aging measured first, depression counted after. It's the best card in the "aging predicts depression" hand.
But the same paper concedes the field's own bias: "nearly all work to date had focused on the hypothesis that poor mental health accelerates processes of biological aging." The Biobank team was deliberately testing the less-studied reverse arrow. That it found a signal doesn't erase the direction most of the evidence points.
And when you look at studies built to resolve direction rather than assert it, the picture inverts.
Take CHARLS: 5,442 Chinese adults aged 45–80, biological age and depression both measured across the 2011 and 2015 waves. Baseline biologically-older predicted later depression (OR = 1.202). Baseline depression predicted later biological aging (OR = 1.372).
Then they ran a cross-lagged panel model to weigh the two pathways against each other simultaneously. The standardized coefficients came out identical: 0.03 for aging→depression, 0.03 for depression→aging. The authors call it "equal contribution."
That is not an engine driving a symptom. That is a feedback loop, with neither direction dominant in the correlational data.
Then Mendelian randomization breaks the tie. A multi-omics analysis across the UK Biobank and the Finnish Twin Cohort found that a lifetime history of major depression was associated with accelerated proteomic aging — systemic and organ-specific, including the brain — with associations stronger than for epigenetic aging.
Crucially, MR revealed a causal effect of MDD on proteomic aging acceleration. And evidence of remission attenuated the link. Treat the depression, and the aging signal softens.
That is the arrow, causally identified, running from depression to aging — not the reverse. The direction everyone wants for the intervention story is the one the causal analysis assigns to the other variable.
The mechanism has a candidate. In a separate NHANES sample of 4,555 adults, depression was associated with a 1.93-year increase in PhenoAgeAccel, and insulin resistance partially mediated it — up to 8.4% via the metabolic score for IR. Depression appears to age the body partly through metabolic dysregulation, which is exactly the axis the ELSA CRP-TyG index was measuring on the front end.
Now the last number, which deflates the mortality pitch too. In a prospective NHANES cohort of 12,761 adults followed a median 11.3 years, major depression raised all-cause mortality 35% (HR = 1.35) and cardiovascular mortality 73% (HR = 1.73).
How much of that death risk did accelerated aging explain? PhenoAgeAccel mediated 10.32% of the depression–all-cause-mortality link, and 5.12% of the cardiovascular one.
Ten percent. The clock captures a real but minority slice of why depressed people die sooner. Nearly 90% of that mortality runs through channels these aging markers don't measure.
Stack it up. Cross-lagged coefficients equal at 0.03. MR pointing causally from depression to aging. Remission attenuating the aging signal. Insulin resistance mediating depression's effect on age. And aging explaining only a tenth of depression's mortality burden.
The honest verdict: these clocks earn their keep as prospectively validated screening flags. The 424,299-person Biobank result is real and independent of genetics — biologically older midlife adults do develop more depression, and it's worth measuring.
But "decelerate aging to prevent depression" is a hypothesis wearing a therapeutic promise's clothing. The causal machinery, where anyone has tried to isolate it, mostly runs the other way — depression accelerates aging, and treating the depression is what moves the aging marker.
The multi-omics authors even flip the framing: proteomic aging clocks as a "novel therapeutic target for treating MDD" and for tracking its downstream harms. That's a monitoring proposition, not a claim that slowing a clock fixes a mood.
And GOLD BioAge should make anyone deploying these tools twice-shy. A marker that is protective against depression in its oldest quartile is not ready to counsel a 55-year-old about their mental-health future, whatever it says about their mortality.
Like the honest read on cross-sectional coupling data elsewhere in this issue, the move here is restraint: treat these clocks as a screening opportunity, not a validated causal target. They tell you who to watch. They do not yet tell you what to fix.