Thigh circumference and risk of heart disease and premature death

Several anthropomorphic indices have been devised to help clinicians predict cardiovascular risk, including body mass index, waist circumference, hip circumference, and waist-hip ratio. Because none has clearly been shown to be superior,1 investigators continue to look for better measures, and in the linked study (doi:10.1136/bmj.b3292) Heitmann and Frederiksen propose a new one—thigh circumference.2

In a cohort of 1436 men and 1380 women aged 35-65 years participating in the Danish MONICA (monitoring trends in and determinants of cardiovascular disease ) project, the authors examined the association between thigh circumference and the incidence of cardiovascular disease and coronary artery disease at 10 years and total mortality at 12.5 years. They fitted four separate proportional hazard regression models to the data for either sex to examine the association between thigh circumference, measured in centimetres directly below the gluteal fold of the right thigh, and hazard ratio of disease and death. The model adjusted for smoking, education, physical activity, menopause (in women), body fat percentage, height, body mass index, waist circumference, alcohol intake, systolic blood pressure, and concentrations of total cholesterol and triglycerides. They used centiles of thigh circumference and made the 50th centile (equal to 55 cm) the base reference (hazard ratio of 1). In most of the models they found an almost linear increase in risk as thigh circumference dropped below 55 cm, which was independent of age or sex; above that circumference, risk decreased but showed no consistent relation. They concluded that below a threshold of about 60 cm the risk of developing heart disease, or dying prematurely, was greatly increased.

The results raise several questions. Is this association real and independent, or a spurious or chance finding? The statistical modelling used in this study was rigorous—it removed the effects of known conventional risk factors and minimised (but did not, and cannot, totally eliminate) residual confounding. Unfortunately, with regard to coronary heart disease for both sexes and cardiovascular disease for women, the hazard ratios became non-significant as more variables were added to the models, which may not have happened if the sample had been larger.

Is this association biologically plausible? It would seem logical that having bigger thighs would be a reflection of greater adiposity, and that this would increase the risk of heart disease. However, the authors cite studies suggesting that too little muscle or subcutaneous fat (or both) in the lower limbs may predispose to adverse glucose and lipid metabolism. Has the association been replicated in other studies? They cite a single study of patients with chronic obstructive pulmonary disease in whom the mid-thigh muscle cross-sectional area was a better predictor of mortality than body mass index. Interestingly, other studies have shown that larger hip circumference (which might be a proxy for thigh circumference) significantly reduces the risk of incident diabetes and coronary heart disease.3

Will this association help clinicians predict risk in individual patients more accurately than they already do using readily accessible and validated risk calculators? The answer is—we do not know. To improve individual risk estimates beyond those that we can derive now, the hazard ratios would need to be much larger than those seen in this study. The highest risk equalled hazard ratios between 2.0 and 2.5, and these applied to only 2.5% of all patients—those with thigh circumferences between 46.0 cm and 46.5 cm. A risk predictor is potentially useful if people who go on to have an event have a higher predicted risk on the basis of the new predictor than those who do not have an event. The probability of seeing such an increased risk in every patient who demonstrates the risk predictor is called the c-statistic, and the higher it is (c-statistic ≥0.7), the more useful the predictor. To achieve such values, the strength of association between the risk predictor and the risk of disease needs to be high, close to an unadjusted hazard ratio of 10 or more.4 Given that the highest hazard ratio associated with models that were not fully adjusted in this study was 4.65, it seems unlikely that thigh circumference will be clinically useful.

If a risk prediction model that incorporates thigh circumference in addition to other known risk factors is to be incorporated into usual practice, we need to ensure several things—firstly, that the new model discriminates better (has a higher c-statistic) than existing models; secondly, that it is well calibrated—that the predicted and observed risk estimates for each stratum of risk are similar; and thirdly, that using the new model will lead to an appropriate change in intended management in more patients now correctly reclassified as having higher or lower risk than would be the case using existing risk prediction models.

More research is needed to see whether measuring the thigh circumference with a tape measure adds anything more to our clinical management than eliciting risk factors from the history, examining the cardiovascular system, and measuring serum lipids. Randomised trials are needed to test whether interventions that increase thigh muscle mass through increased physical activity—in addition to or separate from current primary prevention strategies—decrease cardiovascular risk more than current practice. If this approach is shown to be effective, the public health implications would be intriguing.