Platelet-Derived Growth Factor D Emerges as a Surprising Driver of Heart Attack and Stroke Risk

A protein long known mainly as a growth signal released by platelets may be one of the hidden architects of the world’s deadliest disease. In a two-decade study of thousands of Chinese adults, researchers report that higher circulating levels of platelet-derived growth factor D, or PDGFD, are linked to a markedly elevated risk of atherosclerotic cardiovascular disease, the umbrella term for heart attacks and ischemic strokes caused by clogged arteries. More strikingly, when the team used advanced statistical methods to simulate what would happen if PDGFD levels were deliberately lowered, the predicted twenty-year risk of cardiovascular events fell in a clean, dose-dependent pattern, in some scenarios approaching the benefit of guideline-directed cholesterol lowering.

The study, published in the journal iScience, draws on the Chinese Multi-provincial Cohort Study-Proteomics Project, a community-based prospective investigation rooted in a population cohort initiated in 1992. For the proteomics arm, 2,621 individuals from the Beijing area with available blood samples were enrolled, and after excluding 71 people who had already experienced cardiovascular events before the 2002 baseline examination, 2,550 participants entered the final analysis. Their average age at baseline was 57.5 years, and 44 percent were men. Crucially, the researchers did not rely on a single blood draw: up to four repeated PDGFD measurements were available per participant across visits in 2002, 2007, 2012, and 2020 through 2023, yielding 5,308 protein measurements in total.

Measuring PDGFD required a high-throughput proteomic platform. The team used the SomaScan 11K assay, in which modified single-stranded DNA molecules called SOMAmers bind to target proteins in small aliquots of serum and are quantified on a DNA microarray, with concentrations reported in relative fluorescence units. Quality control figures pointed to solid reproducibility: the median intra-assay coefficient of variation was 3.67 percent in one subcohort and 3.42 percent in the other, while the inter-assay coefficient of variation between batches was 6.87 percent, considered acceptable for this type of measurement. Participants were followed for fatal and non-fatal acute coronary events and ischemic stroke until the end of 2023, with events adjudicated by physicians using the World Health Organization MONICA protocol and supplemented by linkage to national vital registration and hospital information systems.

Over a median follow-up of 20.3 years, 463 participants, or 18.2 percent of the cohort, developed atherosclerotic cardiovascular disease, while 264 participants experienced competing events such as hemorrhagic stroke or non-cardiovascular death. The central finding was consistent across every statistical model the researchers applied. Each standard deviation increase in circulating PDGFD was associated with a 19 percent higher risk of atherosclerotic cardiovascular disease, with a hazard ratio of 1.19. When the analysis was adjusted sequentially for medication use, for the full set of risk factors recommended by the 2020 Chinese Society of Cardiology guidelines for primary prevention, and even for a polygenic risk score capturing inherited cardiovascular susceptibility, the association did not budge, hovering between hazard ratios of 1.22 and 1.24, all statistically significant.

The robustness of the signal was tested from multiple angles. Accounting for competing risks, a statistical necessity when people die of other causes before ever developing heart disease, softened the estimate to a hazard ratio of 1.14 but left it significant. The association held when the researchers used only baseline PDGFD levels, when they used lagged measurements deliberately excluding the blood draw closest to a cardiovascular event to guard against reverse causation, and when they removed anyone who developed disease within the first five years of follow-up, a maneuver designed to weed out pre-existing, undiagnosed atherosclerosis. Only the analysis based on averaged PDGFD levels failed to reach statistical significance, though its effect estimate remained directionally consistent, likely because fewer participants and fewer events were available in that restricted sample.

What elevates the study beyond a conventional association analysis is its use of the target trial emulation framework, a technique that borrows the logic of a randomized controlled trial and applies it to observational data. Before emulating any intervention, the researchers verified a key biological premise: genetic variants associated with serum PDGFD levels mapped to the PDGFD locus itself and shared a consistent association signal, indicating that circulating PDGFD is predominantly regulated by cis-acting variation at its own gene region rather than by distant pleiotropic pathways. That genetic evidence, they argue, justifies treating hypothetical PDGFD-lowering interventions as genuine mimics of a therapeutic effect rather than statistical artifacts.

The emulation itself was technically ambitious. Using targeted maximum likelihood estimation, or TMLE, with a Super Learner algorithm combining generalized linear models, intercept-only models, and random survival forests, the team cloned the cohort into hypothetical treatment arms in which circulating PDGFD was reduced by sustained percentages ranging from 10 to 90 percent. Under the natural course of no intervention, the estimated twenty-year cumulative risk of atherosclerotic cardiovascular disease was 19.0 percent. Lowering PDGFD by 40 to 50 percent pushed that risk down to between 16.8 and 14.8 percent, an absolute risk reduction of 2.3 to 4.4 percentage points, a benefit statistically comparable to the treat-to-target LDL cholesterol lowering strategy used as a positive control, which achieved an absolute risk reduction of 2.8 percentage points. A parallel analysis using the parametric g-formula, a more conservative modeling approach, produced smaller estimates but the same dose-response pattern, with reductions ranging from 0.6 percentage points for a 10 percent PDGFD cut to 3.6 percentage points for a 90 percent cut.

The biology behind these numbers fits a coherent picture. PDGFD is a member of the platelet-derived growth factor family, and previous genetic studies had already flagged the PDGFD gene as perhaps the only growth factor pathway gene associated with coronary artery disease risk. Experimental work has shown that PDGFD promotes the recruitment of immune cells and the migration and proliferation of vascular smooth muscle cells, drives extracellular matrix deposition and vascular structural remodeling, and participates in the platelet-driven inflammatory cascade that creates the pro-inflammatory microenvironment in which atherosclerotic plaques take root. Elevated PDGFD expression has been observed in resident vascular cells and infiltrating inflammatory cells within human atherosclerotic lesions. Platelets, once viewed purely as clotting agents, are now understood to accelerate atherosclerosis by amplifying local inflammation and promoting vascular wall remodeling through crosstalk between endothelial cells and immune cells, and PDGFD appears to sit upstream of much of that signaling.

The subgroup analyses added a layer of personalization that could shape future trials. The estimated benefits of hypothetical PDGFD lowering were largest among people aged 60 and older, men, smokers, individuals with a body mass index of 24 or higher, and those classified as high risk by the ten-year cardiovascular prediction model, where absolute risk changes reached as much as 15.1 percentage points. The pattern echoes a Swedish cohort in which PDGFD gene polymorphism was linked to increased cardiovascular mortality specifically in elderly men, and preclinical findings that obesity upregulates adipose PDGFD and promotes its proteolytic activation, potentially explaining why the obesity subgroup showed outsized estimated benefits. Intriguingly, non-smokers displayed a clear dose-response relationship between the degree of PDGFD lowering and predicted risk reduction, while smokers showed a stronger overall effect but no dose-response gradient, possibly because smoking’s own vascular damage overwhelms the incremental gains from progressively lowering the protein.

The authors are careful to frame the findings as population-level evidence rather than proof of clinical efficacy. Residual confounding by factors such as socioeconomic status cannot be excluded, the causal assumptions underlying the emulation may not be fully satisfied, and the extreme intervention scenarios were explored to map dose-response patterns rather than to represent achievable therapies. The cohort’s geographic concentration in Beijing and its lower smoking prevalence than national estimates may limit generalizability, and the study did not evaluate other PDGF family members such as PDGFA, PDGFB, and PDGFC, leaving the specificity of PDGFD’s role unresolved. Still, the triangulation of longitudinal associations, genetic support, dual causal inference methods, and a positive control intervention gives the result unusual weight. With considerable residual cardiovascular risk persisting despite modern lipid, glucose, and blood pressure therapies, PDGFD now stands as a genuinely modifiable candidate, awaiting the randomized trials that will determine whether lowering it in the clinic can deliver the benefits the simulations promise.

Subject of Research: The association between circulating platelet-derived growth factor D levels and atherosclerotic cardiovascular disease risk

Article Title: Potential role of platelet-derived growth factor D in atherosclerotic cardiovascular disease risk

Article References: Deng, X., Yang, Z., Jia, P., Liu, Y., Zhang, Y., Deng, Q., Yang, N., Hao, Y., Han, L., Sun, J., Lu, W., Sun, X., Qi, Y., & Liu, J. (2026). Potential role of platelet-derived growth factor D in atherosclerotic cardiovascular disease risk. iScience, 29(11), Article 117864. https://doi.org/10.1016/j.isci.2026.117864

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117864

Keywords: PDGFD, atherosclerotic cardiovascular disease, platelets, target trial emulation, proteomics, SomaScan, coronary artery disease, LDL cholesterol, causal inference, polygenic risk score, vascular smooth muscle cells, inflammation