Reply to DA Schoeller

Dear Editor:

We respect that our former collaborator Dr. Schoeller prefers different definitions of some of the terms we used and defined in our energy flux article (1), favors the use of different covariates than we selected and were asked by reviewers to use, and has reservations about using air-displacement plethysmography to estimate total body fat percentage. These differences in opinion play a vital role in scientific progress and we are grateful that he decided to share his perspectives. We had originally invited him to collaborate on this article, but the differences in opinion precluded a successful collaboration.

Although we respect Schoeller’s opinions, the objective of our study was to investigate how best to combine the gold-standard estimates of habitual total energy intake and energy expenditure, on the basis of doubly labeled water, to predict future body fat change. Most researchers and clinicians believe that excessive weight gain results from consuming more calories than are expended, but we discovered that no prospective studies had tested whether estimated habitual energy intake minus estimated habitual energy expenditure predicts future body fat accumulation. We therefore provided what appeared to be the first prospective studies to test the energy balance model of excess body fat gain. In that context, we also tested an alternative theory proposed >20 y ago (2, 3), following the original definition of the construct of energy flux.

To our surprise, we found that energy intake minus energy expenditure did not predict body fat gain in 2 prospective studies, controlling for baseline body fat or body mass. More importantly, we also found that energy intake plus energy expenditure did predict future body fat gain in both studies, again controlling for initial body fat or body mass. For the latter analyses, we excluded participants who were markedly out of energy balance, because the original definition of energy flux outlines that energy intake and energy output must be roughly similar because individuals could have the same total energy flux for completely opposite reasons for those who are markedly out of energy balance. Our operationalization of energy balance did not predict future increases in body fat (mean r = −0.07), whereas our operationalization of energy flux did (mean r = −0.26).

Given that our primary objective was to identify the best way to combine estimates of habitual energy intake and expenditure to predict future body fat gain, we were happy to “let the data speak” with regard to which method of combining these 2 terms had greater predictive validity. After all, the only difference between the 2 predictors is that one is calculated by subtracting the gold-standard estimates of habitual energy intake and energy expenditure and the other by summing these 2 variables. Because it has been difficult to identify biological factors that reliably predict body fat gain, which we feel has limited the efficacy of prevention programs and treatments for obesity, the second leading cause of death worldwide, we were willing to think laterally and investigate the predictive effects of a unique way of combining energy intake and expenditure data. An important cornerstone of science is to try innovative approaches when accepted convention has not worked. We hope others appreciate this perspective.

Unconventional approaches have proven to be critical for addressing several public health problems, such as the initially controversial use of vaccines to eradicate small pox and polio, which saved millions of lives. The results of our study have very clear and testable implications for the prevention and treatment of obesity. Rather than instructing individuals to reduce caloric intake to reduce body fat, which has not produced lasting fat loss in hundreds of randomized trials because caloric restriction produces a range of energy-conserving changes that result in attenuated fat loss, the findings from the energy flux analyses have a different implication. Specifically, the results suggest that it might be more effective for people to increase their physical activity and not reduce caloric intake, because this may allow the natural processes that govern satiety and hunger to function properly. We plan to test the thesis that individuals who maintain high energy flux are more likely to eat healthier foods. This might add to the idea that long-term weight-loss maintenance is more achievable via upregulations in energy intake and expenditure rather than by reducing caloric intake alone.

Because another critical cornerstone to science is transparency and independent replication, we offer Schoeller the opportunity to model these data as he sees fit to determine whether his approach has greater predictive validity for forecasting future body fat gain. We are more interested in identifying reliable predictors of body fat change than in reifying accepted perspectives in the literature. We have already shared these data with him and hope he will communicate the results of the approach he feels is superior. Our expectation is that covarying for factors that naturally correlate with estimated habitual energy intake and expenditure, such as fat-free mass, will only reduce the amount of variance in future body fat change explained and not produce testable hypotheses about how better to prevent and treat obesity. But this is an empirical question, and one that is worth addressing given the pressing need to develop more effective weight-control interventions.

Because we were unable to assess the temporal stability of energy flux, future studies should assess energy intake and energy expenditure repeatedly over time with doubly labeled water to determine whether the estimates based on 2-wk observation periods show temporal reliability. It would also be useful to test whether energy flux, which is estimated over a longer observational period, shows stronger relations to future body fat gain. Perhaps more importantly, given that it is always possible that prospective effects identified in longitudinal analyses are driven by an omitted third variable, we think it would be important to use randomized trials to test whether increasing physical activity is more effective in producing long-term weight loss or preventing excessive body fat gain than is reducing caloric intake. This is a perfectly falsifiable hypothesis and one that we hope multiple independent teams will address in the future.