Adding heart imaging lifted risk prediction from 0.61 to 0.76
In short
Among 27,254 UK Biobank participants with no prior cardiovascular disease, followed for a median of about five years, a model using only traditional risk factors predicted major adverse cardiovascular events at an AUC of 0.61; adding cardiac MRI measures raised it to 0.76, and to 0.90 for heart failure specifically. The features driving risk were left ventricular mass index and global wall thickness — but they formed a high-risk phenotype only alongside reduced myocardial strain and impaired atrial function.
"Your heart has thickened" is a sentence lifters hear. Whether a single wall-thickness number explains actual risk is a separate question, and a recent analysis of the UK Biobank imaging cohort measured the size of that contribution directly.
The cohort was 27,254 participants with no prior cardiovascular disease. Over a median follow-up of about five years, 785 (2.88%) had a major adverse cardiovascular event.
How much did imaging improve prediction?
- Traditional risk factors alone: AUC 0.61 (95% CI 0.58–0.64)
- With cardiac MRI features added: AUC 0.76 (95% CI 0.73–0.79)
- For heart failure specifically: AUC 0.90 (95% CI 0.86–0.94)
- Cross-modality check in an Asian cohort using echocardiography: AUC 0.69 (95% CI 0.56–0.83)
An AUC of 0.61 is barely better than a coin flip. Adding cardiac structure and function measures changed the model substantially, and 0.90 for heart failure is a clinically meaningful level of discrimination.
Which measures carried the risk?
Interpretable machine learning (SHAP) put the largest contributions on higher age, left ventricular mass index and global wall thickness, together with low HDL cholesterol, low left atrial ejection fraction and low right atrial stroke volume.
The composition matters more than the list. Clustering produced two phenotypes, and the higher-risk one bundled increased cardiac volumes, impaired function, a tendency toward myocardial hypertrophy, and reduced myocardial strain. It was not thickness on its own — it was thickening that arrived with falling function.
Does a trained heart belong in that group?
This study was not an athlete cohort, so it does not answer directly. It does make the dividing line visible. Cardiac adaptation to endurance and resistance training enlarges chambers or walls while contractile and relaxation function hold or improve. The high-risk phenotype here is the opposite arrangement: structural change accompanied by functional decline. How the two diverge at the molecular level is in physiological versus pathological hypertrophy.
This is a prediction study, not an intervention study. A higher AUC means risk is identified better, not that having an MRI improves outcomes. Whether asymptomatic imaging is worth adding is a clinical decision, not a self-directed one.
The usable part for a lifter
If a screening test flags something cardiac, the question is whether structure and function are being read together rather than one number in isolation. Why an ECG voltage finding cannot settle wall thickness is covered in the limits of ECG voltage criteria, and a chest X-ray finding has the same problem (cardiomegaly on chest X-ray).
Two items on that risk list are directly trainable: HDL cholesterol and blood pressure. What resistance training does to vessels and pressure is covered in resistance training and artery health and the exercise dose for blood pressure. None of that conflicts with pushing the Big 3 — check where your lifts currently sit with the calculator.
Frequently asked questions
How much did cardiac MRI improve cardiovascular risk prediction?
In 27,254 UK Biobank participants, a model using traditional risk factors alone reached an AUC of 0.61; adding cardiac MRI features raised it to 0.76, and to 0.90 for heart failure specifically.
Which cardiac measures were most linked to risk?
Higher age, left ventricular mass index and global wall thickness, along with low HDL cholesterol, low left atrial ejection fraction and low right atrial stroke volume were the largest contributors to elevated risk.
Is a thick heart wall always dangerous?
In this study the high-risk phenotype was not thickness alone but a bundle: increased cardiac volumes, impaired function and reduced myocardial strain occurring together. Risk rose when structural change came with functional decline.
Does a heart enlarged by training carry the same risk?
The study did not include athletes, so it cannot say directly. The distinction is that training-induced cardiac adaptation preserves contractile and relaxation function, whereas the high-risk phenotype here came with impaired function.
Source: PubMed