Simple Blood Ratio Predicts Diabetes Risk Years Before Diagnosis, Study Finds

A simple calculation involving two numbers that already appear on virtually every routine blood test report may offer an early warning signal for type 2 diabetes, according to a new prospective cohort study drawing on nearly a decade of health data from thousands of middle-aged and older adults in China. The measure, known as the platelet-to-high-density lipoprotein cholesterol ratio, or PHR, is obtained by dividing a person’s platelet count by the concentration of HDL cholesterol, the so-called good cholesterol, in their blood. Researchers who tracked 5,835 diabetes-free participants from the China Health and Retirement Longitudinal Study, better known as CHARLS, found that those with the highest PHR values at the study’s start were significantly more likely to develop diabetes over the following years than those with the lowest values, even after accounting for a wide range of other risk factors.

The findings, published in BMC Endocrine Disorders, arrive at a moment when diabetes has become one of the most pressing public health challenges of our time, and nowhere more so than in China, which carries one of the largest diabetes burdens of any nation on Earth. The appeal of the PHR lies in its radical simplicity. Platelet counts and HDL cholesterol levels are measured millions of times each day in hospitals and clinics around the world, meaning that the ratio costs nothing extra to calculate and requires no new technology, no specialized assay, and no additional blood draw. In an era when biomarker research increasingly leans on expensive multi-omics platforms and machine learning pipelines, there is something striking about a two-number quotient that appears to carry meaningful predictive information about a devastating chronic disease.

To understand why such a ratio might work, it helps to consider what each component represents. Platelets are best known for their role in blood clotting, but hematologists have long recognized that they are also active participants in inflammation. Elevated platelet counts can reflect a state of systemic low-grade inflammation, and activated platelets contribute to the endothelial dysfunction, atherosclerosis, and metabolic disturbances that characterize cardiometabolic disease. HDL cholesterol, by contrast, is generally protective: it participates in reverse cholesterol transport, ferrying cholesterol away from peripheral tissues and artery walls, and it also exhibits anti-inflammatory and antioxidant properties. Low HDL cholesterol is a well-established feature of the metabolic syndrome and a recognized predictor of type 2 diabetes. By placing an inflammatory marker in the numerator and a protective marker in the denominator, the PHR essentially combines two opposing physiological signals into a single index, one that rises when inflammation climbs or when protective lipid function falls.

The study’s design lent considerable statistical weight to its conclusions. The investigators used CHARLS, a nationally representative longitudinal survey of Chinese residents aged 45 and older, which has followed its participants with repeated interviews and physical examinations since 2011. From the baseline wave conducted in 2011 and 2012, the researchers identified 5,835 participants who were free of diabetes at enrollment. These individuals were then tracked across four subsequent follow-up waves in 2013, 2015, 2018, and 2020, yielding a median follow-up duration of 116 months, nearly ten years of observation per participant. Over that period, 669 participants developed incident diabetes, providing a substantial number of outcome events for the statistical models to work with.

The analytical approach followed the standard playbook for rigorous cohort research, with a few refinements that are worth highlighting. Participants were sorted into four groups according to the quartiles of their baseline PHR, allowing the researchers to compare diabetes incidence between the highest and lowest quarters of the distribution. Kaplan-Meier survival curves, which plot the cumulative probability of remaining diabetes-free over time, revealed a striking visual separation between the groups: the higher the PHR quartile, the steeper the accumulation of new diabetes cases, a difference that was highly statistically significant with a log-rank p value below 0.001. The researchers then turned to Cox proportional hazards regression, the workhorse of survival analysis, which models the hazard, or instantaneous risk, of an event as a function of predictor variables while appropriately handling censoring and differing follow-up times.

The headline result emerged from the fully adjusted Cox models, in which the association between PHR and diabetes was estimated while controlling for an extensive battery of potential confounders including age, sex, body mass index, blood pressure, lipid fractions, kidney function markers such as serum creatinine, and other clinical and demographic characteristics. Participants in the highest PHR quartile had a 41 percent higher risk of developing diabetes compared with those in the lowest quartile, with an adjusted hazard ratio of 1.41 and a 95 percent confidence interval of 1.11 to 1.78, and a p value of 0.004. Importantly, the confidence interval excludes the null value of 1.0, indicating that the association is unlikely to be a statistical fluke, and the width of the interval suggests reasonable precision given the sample size.

One of the more intriguing technical aspects of the study involved testing whether the relationship between PHR and diabetes risk was linear or followed a more complicated curve. The researchers employed restricted cubic spline regression, a flexible modeling technique that fits a smooth curve through the data using polynomial segments joined at designated points called knots, while constraining the tails of the curve to be linear. This approach can reveal threshold effects, U-shaped relationships, or other nonlinear patterns that a conventional linear model would miss. In this case, the spline analysis indicated a straightforward positive linear relationship: as PHR increases, diabetes risk increases proportionally, with no evidence of a plateau or threshold. A formal test for nonlinearity returned a p value of 0.339, well short of significance, supporting the interpretation of a linear dose-response pattern.

Robustness checks formed another pillar of the investigation. The researchers performed subgroup analyses, stratifying the sample by age, sex, obesity status, and other characteristics to determine whether the association held consistently across different segments of the population. By and large it did, suggesting that the PHR-diabetes link is not confined to a particular demographic niche. The team also ran sensitivity analyses using alternative definitions of the outcome. When they employed a composite diabetes definition, presumably incorporating self-reported physician diagnoses and medication use alongside laboratory measures, the association remained significant, with the highest quartile showing a hazard ratio of 1.30 and a confidence interval of 1.09 to 1.56. Additional analyses using multiple imputation, a statistical technique for handling missing data that generates several plausible completed datasets and pools the results, produced consistent findings, reassuring the authors that incomplete records were not driving the results.

Perhaps the most scientifically candid element of the paper concerns a question the authors themselves raise: does the PHR actually add predictive value beyond what its components provide individually? Component-specific analyses suggested that the association may be driven predominantly by the HDL cholesterol side of the ratio rather than the platelet count. This is an important nuance. If HDL cholesterol alone carries most of the signal, then the ratio’s incremental utility over simply measuring HDL is open to debate, and the authors explicitly note that whether PHR provides information beyond HDL-C alone requires further investigation. Such transparency is a hallmark of careful biomarker research, where enthusiasm for novel composite indices must always be tempered by head-to-head comparisons with their constituent parts.

The study does have limitations inherent to its design. As an observational cohort, it can demonstrate association but not causation, and residual confounding by unmeasured lifestyle or genetic factors cannot be fully excluded. The findings also derive from a Chinese population of middle-aged and older adults, and generalizability to other populations will need to be tested in independent cohorts. Nevertheless, the work adds to a growing body of literature linking integrated inflammatory-metabolic indices to incident cardiometabolic disease, and it identifies PHR as a candidate biomarker that is, in the authors’ words, simple and routinely available. For clinicians and researchers watching the diabetes epidemic, the prospect of squeezing additional predictive insight out of tests that are already being ordered every day is an attractive one, and future studies will determine whether this humble ratio earns a place in clinical risk assessment.

Subject of Research: Association between the platelet-to-HDL cholesterol ratio and incident diabetes risk in a national Chinese cohort

Article Title: Association between platelet-to-high-density lipoprotein cholesterol ratio and incident diabetes risk: a national representative cohort study based on CHARLS

Article References: Chen, L., He, C., & Chen, X. (2026). Association between platelet-to-high-density lipoprotein cholesterol ratio and incident diabetes risk: a national representative cohort study based on CHARLS. BMC Endocrine Disorders. https://doi.org/10.1186/s12902-026-02627-x

Image Credits: AI Generated

DOI: 10.1186/s12902-026-02627-x

Keywords: platelet-to-HDL cholesterol ratio, PHR, type 2 diabetes, diabetes risk prediction, CHARLS, HDL cholesterol, inflammation, biomarker, cohort study, Cox regression, cardiometabolic risk, endocrinology