Risk Estimates Split Sharply Across Three Cardiovascular Models
Cardiovascular prevention starts with a number, but the number depends on the model chosen. In a cohort of university employees in India, three widely used cardiovascular risk tools produced materially different 10-year estimates, with the WHO/ISH charts and ASCVD-PCE running close together while the Framingham equation pushed many more participants into higher-risk categories.
The divergence was not subtle. Mean estimated risk was 5.94% with WHO/ISH, 5.95% with ASCVD-PCE, and 12.86% with Framingham, a spread large enough to change how primary prevention would be framed at the bedside. In practical terms, one model suggests a modest risk burden while another assigns a far heavier future load of events, even though all three are intended to guide prevention rather than diagnose current disease.
The Occupational Cohort and How Risk Was Measured
The analysis drew on standardized health assessment data from 266 university employees, then narrowed to adults aged 40 to 74 years to create a common age range across the three prediction approaches. Among 127 employees in that age band, three with prior ischemic heart disease were excluded because the study focused on primary prevention. The final analytical cohort included 124 participants.
Mean age was 47.63 ± 6.45 years, and 86.3% were male. Diabetes affected 25.8% of the cohort, smoking or tobacco use was present in 19.4%, and 29.0% were already receiving antihypertensive treatment. Those baseline features describe a workforce that is not immunized against cardiometabolic risk simply because it is employed in an academic setting. The burden of diabetes and blood pressure treatment signals a population in which vascular risk may already be accumulating before the age of retirement.
Blood pressure assessment followed a standardized protocol with three readings taken at five-minute intervals, and repeat assessments were carried out during three subsequent weekly visits for employees with elevated initial readings. That design matters because blood pressure variability can distort risk estimation if a single office reading is allowed to stand in for the patient’s usual hemodynamic state. Risk was then calculated using the 2019 WHO/ISH laboratory-based charts for South Asia, the ACC/AHA Pooled Cohort Equation, and the 2008 Framingham general cardiovascular disease equation.
Framingham Classified More Employees as High Risk
The clearest signal came at the high-risk threshold. Using a 20% cutoff, WHO/ISH classified 2.4% of participants as high risk, ASCVD-PCE classified 4.8%, and Framingham classified 20.2%. A model that places one in five employees above a high-risk line will produce a very different clinical workflow than one that flags only a small minority.
Exact four-category agreement mirrored that split. WHO/ISH and ASCVD-PCE matched in 70.2% of cases, while WHO/ISH and Framingham agreed in only 31.5% of cases and ASCVD-PCE and Framingham in 26.6%. The categorical picture was reinforced by kappa statistics: 0.502 for WHO/ISH versus ASCVD-PCE, 0.079 for WHO/ISH versus Framingham, and 0.047 for ASCVD-PCE versus Framingham. Weighted kappa values were 0.639, 0.323, and 0.342, respectively.
A single sentence captures the clinical consequence: Framingham consistently pushed this Indian occupational cohort toward higher predicted risk.
Why Different Equations Produce Different Clinical Answers
These discrepancies are not random arithmetic errors. Risk models are built from different derivation populations, different endpoints, and different assumptions about how age, blood pressure, lipids, diabetes, and smoking interact over time. A chart calibrated for one population can overestimate or underestimate risk when moved into another, especially when baseline event rates and competing causes of disease differ.
The Framingham equation has long been influential, but it was developed in a U.S. community population and was designed around a broader cardiovascular endpoint structure. By contrast, WHO/ISH charts are intended for regional application and ASCVD-PCE was derived to estimate atherosclerotic cardiovascular outcomes. Those differences in endpoint definition matter at the decision threshold, where a patient may or may not qualify for more intensive lipid-lowering therapy, blood pressure optimization, or closer follow-up.
In a workforce like this one, the implications are immediate. If a model systematically assigns higher risk, clinicians may recommend more aggressive primary prevention, but they may also expose lower-risk individuals to unnecessary concern or treatment escalation. If a model estimates lower risk than is biologically plausible, opportunities for prevention can be missed.
What the Agreement Data Mean for Indian Primary Prevention
The close alignment between WHO/ISH and ASCVD-PCE suggests that these two tools may be more interchangeable than either is with Framingham in this setting, at least for broad risk categorization. Their mean estimates were nearly identical, and their categorical agreement was substantially stronger than the agreement seen with Framingham.
By contrast, Framingham’s higher mean risk and higher rate of 20% classification suggest a tendency to shift employees into more intensive management pathways. In a primary-prevention clinic, that could translate into more statin discussions, more frequent surveillance, and a lower threshold for intervention. Whether that improves outcomes depends on how well the score matches actual future events in Indian occupational populations, a question this study cannot answer because it compared models rather than tracking incident cardiovascular disease.
The cohort size also limits how firmly the findings can be generalized beyond this university workforce. Even so, the pattern is clinically coherent: models built on different populations do not behave identically when they are dropped into a distinct Indian setting with its own mix of diabetes, tobacco exposure, and treated hypertension.
A Reminder for Bedside Risk Discussion
Risk calculators are decision aids, not absolutes. In a middle-aged employee with diabetes and hypertension treatment already on board, the choice of equation may determine whether the conversation stays observational or shifts toward active pharmacologic prevention.
That makes model selection part of the clinical act itself. When a population has not been represented in the derivation set of a risk equation, the resulting estimate may be less a prediction than an approximation shaped by geography, endpoint definition, and calibration.
For Indian occupational health programs, this study argues for restraint before importing a foreign risk score without question. WHO/ISH and ASCVD-PCE produced closely similar estimates here, while Framingham gave a much larger risk burden and a far higher share of employees in the high-risk category. The unresolved issue is not whether one model can calculate a percentage, but which model best reflects future events in this specific population and what threshold should trigger treatment when the calculators disagree.