PCBs and heart risk: what a predicted score can and cannot show

Polychlorinated biphenyls were banned decades ago in most of the world. They remain measurable in almost everyone's blood, which is the defining feature of a persistent organic pollutant and the reason the class keeps generating epidemiology long after the factories closed. A study published on 30 August asks whether the residue that is left tracks with cardiovascular risk.

What was measured

The analysis draws on the prospective Wuhan-Zhuhai cohort in China [s1]. Seven serum PCB congeners — PCB-28, 52, 101, 118, 138, 153 and 180 — were measured by gas chromatography with tandem mass spectrometry [s1]. Predicted 10-year atherosclerotic cardiovascular disease risk was calculated with the China-PAR equation, a risk model validated for Chinese populations, and urinary 8-iso-prostaglandin F2α served as a marker of oxidative stress [s1].

That choice of outcome is the single most important thing to understand about this study, and it is addressed below.

The findings

Cross-sectionally, each one-unit increase in the natural-log-transformed concentration of PCB-118, PCB-138, PCB-153 and total PCBs was associated with 1.249, 1.253, 1.457 and 1.829 percentage points higher predicted 10-year risk respectively, all with P values below 0.05 [s1]. The associations for PCB-118, PCB-138 and PCB-153 remained significant after false discovery rate correction — the adjustment that accounts for having tested seven congeners at once [s1].

Longitudinally, persistent high PCB-118 exposure was associated with a 0.510 percentage-point greater increase in predicted risk (95% CI 0.023 to 0.996) and a higher risk of transitioning to predicted high-risk status (RR 1.432, 95% CI 1.012 to 2.027), each compared with persistent low exposure [s1]. Neither of those longitudinal associations remained significant after false discovery rate correction [s1].

An exploratory mediation analysis found urinary 8-iso-prostaglandin F2α statistically consistent with partial indirect pathways, with estimated indirect proportions ranging from 3.29% to 11.62% [s1]. Database-derived bioinformatics analyses pointed to TNF, NF-κB and IL-17 pathways and identified IL-6, TNF, IL-1β and ICAM1 as candidate hub genes [s1].

The outcome is a calculation, not an event

Nobody in this study had a heart attack for the purposes of the analysis. The outcome is a number produced by an equation from a person's age, sex, blood pressure, cholesterol and other inputs [s1]. A finding that PCB exposure is associated with a higher China-PAR score is, mechanically, a finding that PCB exposure is associated with the inputs to that equation.

That is not worthless — the risk equation is validated, and the inputs are real clinical measures. But it changes what the result means. If PCBs are associated with, say, higher blood lipids, a predicted-risk analysis will register that as elevated cardiovascular risk without adding independent information about cardiovascular disease. An analysis of actual incident events would.

A shift of 1.829 percentage points in predicted 10-year risk is also modest at the individual level, and its clinical meaning depends entirely on where a person sits in the distribution — a point the transition analysis gets at by asking who crosses into predicted high-risk status [s1].

What the authors claim, and what they do not

The paper's own summary is unusually restrained for this literature: several PCB measures were associated cross-sectionally with higher predicted risk, while longitudinal evidence was limited and strongest for persistent high PCB-118 exposure [s1]. The mediation and bioinformatics findings are labelled as highlighting possible pathways that require further validation [s1].

That restraint is warranted by the numbers. The cross-sectional associations held after correction; the longitudinal ones did not [s1]. When a study's prospective results are the ones that fall away, the prudent reading is that the cross-sectional signal may reflect shared determinants of both PCB burden and metabolic profile — age and body fat being the obvious candidates, since PCBs are stored in adipose tissue — rather than a forward-acting effect.

The bioinformatics component deserves a similar caution. Pathways were derived from databases rather than measured in these participants, and TNF, NF-κB and IL-17 signalling is implicated in a very wide range of conditions. Identifying them as candidates narrows nothing on its own.

Why legacy pollutants keep generating this literature

PCBs, like PFOS and PFOA, are covered by international controls that ended most production, and like them the environmental burden has not ended with the production. A life cycle emission inventory for mainland China covering 1985 to 2023 found PFOS and PFOA emissions following an increase-peak-decline pattern, peaking around 2010 and then falling, which the authors attribute to the Stockholm Convention and to domestic regulations [s2]. Their conclusion is that future emissions will increasingly come from legacy products and waste streams rather than from manufacture [s2].

The epidemiology has the same shape. Exposure that peaked a generation ago is still measurable in serum, still varies across a population, and still supplies enough contrast for a cohort study — which is why banned compounds continue to produce new findings, and why those findings describe a declining exposure rather than a current one.

What would settle it

An analysis of incident cardiovascular events rather than predicted risk, in a cohort large enough for that outcome, with repeated PCB measurement. The Wuhan-Zhuhai cohort is prospective, so that analysis is possible in principle [s1]. Until it exists, the defensible statement is that serum PCB concentrations track with the components of a cardiovascular risk score, and that the prospective version of the finding did not survive correction for multiple testing.

Sources

• [s1] Cross-sectional and longitudinal relationships between exposure to polychlorinated biphenyls and predicted 10-year atherosclerotic cardiovascular disease risk. Environment International, 30 August 2026. https://doi.org/10.1016/j.envint.2026.110493
• [s2] Highly Resolved Life Cycle Emissions of Per- and Polyfluoroalkyl Substances (PFAS) in China from 1985 to 2023. Environmental Science & Technology, 17 July 2026. https://doi.org/10.1021/acs.est.6c02449