Research · Apr 19, 2026

Control arms of 29 trials as the natural history of chronic CAD: what the Mahesh et al. meta-analysis shows

Mahesh R, Wong DHC, Anand VV, Soh DKE, Low GZ, Hao Y, et al. Natural history and outcomes of medically treated coronary artery disease: insights from reconstructed individual patient data of 29 randomized trials. Clin Res Cardiol. 2025 (published online). doi: 10.1007/s00392-025-02800-7

What the study did

The authors reconstructed individual patient data from the Kaplan-Meier curves of the control arms of 29 randomized trials in chronic CAD. They pooled them into a single curve of “medically treated chronic CAD.” The cohort has 51,275 patients and 185,455 patient-years, with a mean follow-up of 4.4 years. The headline is that cardiovascular risk “more than doubles at ≥5 years of follow-up.”

The idea

Reconstructed IPD is usually used to compare treatments: PCI versus CABG, TAVI versus SAVR, drug-eluting versus bare-metal stents. Here the technique was applied to a single arm, pooled across 29 trials, to describe the course of treated disease. Control arms are usually seen only as the reference for a treatment effect. The authors treated them as a data source of their own.

The quality argument holds. Control arms carry trial-grade data: prespecified inclusion, protocol-defined follow-up and blinded event adjudication with harmonized definitions. No registry delivers that measurement quality, at any sample size.

The limit: precise measurements of different populations

The internal consistency that gives trial data its value holds within each trial. It is lost when 29 trials published between 1979 and 2024 are pooled. The result is trial-grade measurement applied to populations and therapies that changed over four decades.

The guideline-directed therapy label

The paper reports aspirin in 91.2%, statins in 87.1% and P2Y12 inhibitors in 47.0%, as if these described the whole cohort. Each value was computed only among the trials that reported that drug class. The rest were dropped.

The statin figure comes from 14 trials, about 45% of the patients. The other 55% contribute nothing. The missing data are not random. The trials without statin data are mostly the oldest: CASS, VA, Kloster and the European Coronary Surgery Study. In those, statin use was close to zero. Excluding them pushes the estimate up.

4S, EUROPA and PEACE

The 4S placebo arm contributes 4,444 patients and 438 deaths, about 14% of all deaths in the meta-analysis. That arm received no statin by protocol. It is the counterfactual against which current therapy was defined, and it entered the curve as “treated CAD.”

The same applies to EUROPA (n = 12,218) and PEACE (n = 8,290). These are placebo arms of ACE-inhibitor trials, built to receive no renin-angiotensin inhibition. The three trials make up almost half of the cohort.

Even the more recent trials, such as COURAGE, BARI-2D, FAME-2 and ISCHEMIA, predate the use of SGLT2 inhibitors in atherosclerotic disease, GLP-1 agonists, PCSK9 inhibitors, colchicine, icosapent ethyl and LDL targets below 55 mg/dL. No trial in the cohort approximates the therapy of a well-treated patient in 2026. Most do not represent 2016 therapy either.

The “doubling” of risk after 5 years

The figure comes from the year-by-year estimates in Table 4. Each year is computed from a different subset of trials. Short-follow-up trials enrolled newer, lower-risk populations. Long-follow-up trials are mostly from the pre-statin era. The jump between years compares different subsets, not the course of the same patients.

The reconstructed Kaplan-Meier curve shows a linear risk of about 2% per year across the whole horizon, with no inflection. The acceleration exists between trials, not over time.

Revascularization

A description of one treated arm says nothing about revascularization. The paper’s Discussion frames revascularization as “a valuable adjunct rather than an alternative,” but its own data do not address this. That question is answered by trials that randomize strategies: ISCHEMIA, COURAGE, BARI-2D and FAME-2. Together they show equivalence for hard endpoints and benefit for symptoms.

What the paper shows

Chronic CAD patients on the therapy of their era had mortality of about 2% per year and MI of 2% to 3% per year over five years. The disease is not cured, and surveillance matters. The paper does not show acceleration of risk, does not support more revascularization and does not inform about current medical therapy.

My reading

Historical trial cohorts reflect a standard of care that today’s patients should improve on. The question I care about is how far that 2% per year falls with routine use of what none of these cohorts received. I expect the main gain from pharmacological, metabolic and anti-inflammatory therapy. More revascularization may help too, but that has yet to be shown.