Choice of baseline for primary endpoint
We write to clarify selection of the baseline to determine the diet effect on total energy expenditure and other outcomes. As we discussed in the Supplement, we chose to measure baseline in the post-weight loss (not pre-weight loss) assessment period for several reasons:
1) As a general rule, anchor data should be collected as close to the time of randomization as possible, to decrease error introduced by any time-varying confounder. The pre-weight loss measurement was obtained 3 to 4 months prior to initiation of the test diets.
2) The pre-weight loss measurement is strongly confounded by weight loss, whereas our specific aim is to examine total energy expenditure during weight-loss maintenance (consistent with the registered title of the protocol, Dietary Composition and Energy Expenditure During Weight-Loss Maintenance).
3) The stated purpose of the Run-In phase is to produce 12% weight loss, changing biological state (i.e., creating a predisposition to weight regain) to test the study hypotheses. Thus, it would be inconsistent with study aims and methodologically inappropriate to use the pre-weight loss measurement as a precise and accurate anchor for evaluating change in total energy expenditure in response to the test diets. Doing so would necessitate a substantially larger number of participants (and increased cost) to account for the additional imprecision, with no scientific benefit.
Some have recently argued that there might be ongoing weight change during the post-weight loss period that could inflate our results (1). However, this measurement was done before randomization, so any individual variations in weight or other factors could not confound study findings. Although error in baseline measurement could reduce the precision of our effect estimates – a non-issue here, as we had abundant power to test the primary hypothesis – it would not create bias in favor of any one diet.
Furthermore, as can be determined from the publicly available database, weight change during the 15-day assessment period was very small, averaging 23 grams per day, with no significant difference between individuals who would be assigned to the 3 diet groups (p=0.2). This value compares favorably with Redman et al (2), who state that “an accepted criterion for weight stability in controlled feeding studies is fluctuations in weight <250 g/d.”
Finally, our fully adjusted model included “pre-weight loss values for BMI, percentage lean mass, and total energy expenditure; and weight loss from pre-weight loss to start of trial.” The results of this model were not materially different from those of the minimally adjusted model, indicating that these pre-randomization factors were unimportant.
1. Belluz J. Does cutting carbs really help keep weight off? The big new diet study, explained. Vox, 16 Nov 2018. https://www.vox.com/science-and-health/2018/11/16/18096633/keto-low-carb...
2. Redman LM, Kraus WE, Bhapkar M, Das SK, Racette SB, Martin CK, Fontana L, Wong WW, Roberts SB, Ravussin E; CALERIE Study Group. Energy requirements in nonobese men and women: results from CALERIE. Am J Clin Nutr 2014;99:71-8
Competing interests: As detailed in the manuscript