How body composition may confound effect estimates of cardiorespiratory fitness

In the present issue of the European Journal of Preventive Cardiology, Salokari and co-authors studied the prognostic value of the Duke Treadmill Score (DTS) in 3936 middle-aged Finns.1 Cardiorespiratory fitness (CRF) was determined by measurement of maximal oxygen uptake (VO2max) on a bicycle ergometer rather than treadmill exercise time of the Bruce protocol, which is used in the original DTS. The authors demonstrated that the DTS does not provide any prognostic information independent of its components and that CRF (the authors use the term ‘exercise capacity’) was the most important predictor of cardiovascular mortality. Converting VO2max into metabolic equivalents of tasks (METs) was mentioned as a limitation since the latter is prone to confounding by body composition.2,3 In this editorial, we will outline the challenges associated with converting individual measures of exercise capacity into (standardized) measures of CRF. Further, we will argue that the main findings of the current study are likely to remain valid, despite potential for confounding by body composition. Converting VO2max into a measure of fitness may seem straightforward: a bigger person has a larger body and must consume more oxygen than a smaller person while doing the same activity. Thus, on the surface, scaling VO2max as ml/min per kg seems appropriate. However, expressing fitness as VO2max in ml/min per kg implies that in the average person VO2max is proportional to total body mass. If that were true, then for every additional kilogram of body weight, VO2max would need to increase by a fixed volume for a person’s fitness to remain unchanged (ratio scaling: mass exponent = 1).4 However, observations suggest that the increase in VO2max is not a linear function of total body mass. In a study of >1100 elderly individuals we found that lean mass is a better basis for scaling VO2max than total-body mass3 and a meta-analysis involving >6500 participants demonstrated a pooled mass exponent for total body mass of only 0.7.5 In Figure 1, we outline four people: persons A and B have similar amounts of lean mass (i.e. size of heart, lungs, muscles and other organs) but differing body compositions (i.e. proportions of body fat) and persons C and D have similar body compositions but higher body weights than persons A and B, respectively. Converting VO2max into a measure of cardiorespiratory fitness: Bias introduced by ratio-scaling compared to emperical mass exponents. Figure 1(a) shows the VO2max calculations if scaled to total body mass. The fitness level of person A is high, according to both empirical scale and ratio scale, persons C and D have normal fitness according to ratio scale and high fitness according to empirical scale whereas person B is deemed to have low fitness. The red line is the demarcation between high and low fitness defined by the ratio scale, that is, ml/min per kg. It has to pass through the origin and the point where population means of total body mass and VO2max intersect. The green line is the empirical mass exponent of total body mass.5 Compared with ratio scaling, use of the empirical mass exponent (i.e. the rate of increase of fitness with increasing body weight that has been observed in real-life populations) implies a higher threshold for fitness in individuals with lower body mass. In other words, ratio scaling results in underestimation of fitness in heavier and overestimation in lighter individuals. Assuming that persons A and B consume the same amount of oxygen at peak performance (i.e. same ml/min), the fitness level of person B (VO2max expressed as ml/min per kg) will appear to be lower since the formula involves total body weight and person B weighs more due to larger inert fat mass. Figure 1(b) shows the same four-person scenario, but this time VO2max is scaled to fat-free mass. Person B, freed of the handicap of inert extra weight in the form of fat, has now the same fitness level as the leaner person A. Person C, who had normal fitness scaled to total body mass, suddenly finds him-/herself on the wrong side of the line of normal fitness whereas D rises to the top of our fitness scale. Note that ratio scale for fat-free mass is almost identical to the empirical mass exponent (1.0 vs. 0.9). Note further that use of the empirical mass exponent when scaling to total body mass could not prevent misclassification of person B as non-fit as the lower mass exponent can only reflect average reduction of additional VO2max with increasing body weight on a population level. Exceptionally obese or lean individuals may still be misclassified. By implication, just as ignoring body composition is a potential source of misclassification, scaling for fat-free mass may still be associated with a residual bias if there is another co-determinant of VO2max that is independent of fat-free mass. However, the empirical mass exponent of 0.9 for fat-free mass leaves considerably lower room for error, as it is much closer to the ideal mass exponent of 1.0 compared with the exponent of 0.7 associated with total body mass. Let us now consider how this bias may affect studies of CRF and overall health. When fitness is used to predict health outcomes two things may happen: (a) if the outcome is not caused by obesity then the lower level of CRF for a heavier person (i.e. person B) will falsely predict a worse prognosis than justified by his or her cardiorespiratory system;6 or (b) if the outcome is caused by obesity, part of the risk conveyed by person B’s obesity will be wrongly attributed to lack of fitness.7 A recent study by Haapala et al. is a good example of how body-composition bias can be ruled out by providing effect estimates of cardiorespiratory fitness scaled to both total body mass and fat-free mass.8 In the current study VO2max is converted into METs. As one MET is equal to 3.5 ml oxygen uptake per kg body weight this conversion introduces a potential for body-composition bias. However, the treadmill test that is used to assess CRF in the DTS has a similar potential for body composition bias: the weight bearing treadmill tests put heavier individuals at a disadvantage. Since validation against direct measurement of VO2max has been done in normal weight populations9,10 the resulting estimates of CRF are only valid for comparisons within strata of similar body composition. Thus, even if a body-composition independent measure of CRF had been available, it would have been of limited value to study the prognostic value of the DTS. In summary, CRF should be standardized for body composition if obesity is causally related to the outcome or if independent effects of fitness and obesity are studied. As the DTS is incapable of distinguishing between fitness and obesity, use of a body-composition confounded measure of fitness has no bearing on the findings of the current study. Body composition (two-compartment model): proportion of body fat and lean mass (=fat-free mass). Mass exponent: describes the association between body mass and VO2max. If the mass exponent is equal with 1, there is a linear association (as implied by the ratio standard). If the mass exponent is <1, each additional unit of body mass is associated with a decreasing (i.e. smaller than proportional) amount of additional VO2max. Empirical mass exponent: the rate of increasing VO2max with increasing weight that has been observed in real-life populations. It is 0.7 for total body mass and 0.9 for fat-free mass (compared with 1.0 in ratio scaling). Ratio scaling: obtaining a measure of fitness by dividing VO2max with a constant (e.g. weight). The line of normal fitness is defined by the origin (0/0) and population mean (e.g. mean VO2max/mean weight). Ratio scaling is based on the implication that there is a linear association between the base for scaling (e.g. total body mass) and VO2max. The potential error introduced by ratio-scaling is determined by the degree to which the empirical mass exponent (i.e. the real association observed in large populations) is different from 1.0. Scaling: to standardize VO2max to obtain a measure of fitness that allows valid comparisons between people of different body sizes and compositions. The author(s) declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article. The author(s) received no financial support for the research, authorship and/or publication of this article.