Arte SheetsLearn. Try. Understand.

Health & Fitness

Cox Survival Arithmetic: Framingham and Historical Pooled Cohort Context

Explore the survival transformation shared by some historical cardiovascular risk models, with model-specific parameters and no automatic clinical classification.

A survival-model calculation, not a complete risk calculator

Some cardiovascular risk models use a Cox proportional-hazards transformation of the form 1 − S_ref^EXP(L − L_ref), where S_ref is the event-free survival associated with a model’s reference predictor L_ref and L is an individual linear predictor from that same model. The interval and event definition belong to the fitted model.

The spreadsheet only illustrates this final transformation. It does not calculate L from patient data, choose a model, assess eligibility or generate a clinical recommendation. Parameters, coefficients and reference populations from different models cannot be freely mixed.

A documented historical Pooled Cohort example

The 2013 ACC/AHA Pooled Cohort Equations report gives S_ref = 0.9144 and L_ref = 61.18 for its historical White-men equation. Its worked example uses a rounded L = 60.69, producing approximately 5.3% for that model’s specified 10-year atherosclerotic endpoint. These are not universal Framingham constants or values for every population.

The full linear predictor includes more than a simple sum of logarithms: the 2013 equations use demographic-specific coefficients and some interaction terms. Framingham general cardiovascular disease and Pooled Cohort ASCVD models have different definitions and coefficients. The selected values below reproduce only the transformation, not the full model.

=IF(COUNT(B2:B4)<>3,"Enter all three numeric parameters",IF(OR(B2<=0,B2>1,ABS(B4-B3)>20),"Outside example domain",ROUND(1-B2^EXP(B4-B3),6)))

Calculate the transformation

B2 is reference event-free survival from a chosen model, B3 its reference predictor and B4 the compatible calculated predictor. All three must be numeric. This example requires survival greater than 0 and at most 1, and limits the predictor difference to ±20 to avoid extreme numerical overflow; those bounds are a spreadsheet exercise constraint, not model eligibility criteria.

With 0.9144, 61.18 and 60.69, the unrounded result is approximately 0.053347, or 5.3347%. When the predictors are equal, the result is 1 − S_ref; this is a reference-profile result, not necessarily the average risk across all individuals.

Current guidance and interpretation

The 2026 ACC/AHA multisociety dyslipidemia guideline recommends newer PREVENT-ASCVD equations for risk assessment in appropriate primary-prevention populations. Those equations are distinct from the historical Pooled Cohort example on this page.

An estimated risk depends on the model, inputs, population and endpoint and is not a diagnosis or automatic treatment decision. Use the appropriate validated calculator and clinician–patient discussion for clinical care.

Try the example

Use the original values or change them to test your understanding.

Further reading