The primary finding from this large exercise intervention trial in postmenopausal women is that the difference between actual weight loss and predicted weight loss (compensation) increases with exercise dose. We confirmed the findings of previous studies that a relatively high dose of exercise (12 KKW or 194 minutes per week) results in compensatory mechanisms that attenuate weight loss in previously sedentary women. However, a dose of exercise consistent with the exercise prescription for general health (8 KKW or 136 minutes per week) did not result in compensation as the actual weight loss closely matched the predicted weight loss. An exercise dose of 4 KKW (72 minutes per week) also resulted in weight loss that slightly exceeded the predicted weight loss. Our findings are important because most exercise guidelines for weight loss recommend 200–300 minutes per week and we provide evidence that this amount of exercise induces compensation that results in significantly less weight loss than predicted. Our findings of different doses of exercise resulting in different amounts of compensation may in part explain the discrepant results from previous exercise and weight loss studies. If different doses of exercise result in different amounts of compensation it is not surprising that when exercise studies of varying doses and duration are examined collectively there is no relation between exercise dose and weight loss. Our findings provide an excellent example of the complex nature of the energy expenditure and energy intake relation, and the importance of considering both sides of the equation in creating weight loss programs.
Numerous well conducted clinical exercise trials of long duration have reported less weight loss than expected. 
For example, Donnelly et al observed that after 18 months of exercise training and achieving 2000 kcal per week of exercise, college-aged women had no weight loss, although women in the control group actually gained weight. 
In a follow-up report Donnelly et al suggested that the observed “compensation” was due to increased energy intake, not changes in metabolic rate or non-exercise activity. 
A recent report by King et al provides support that increased energy intake may be a major source of energy compensation as they noted that after 12 weeks of 2500 kcal per week in exercise training energy expenditure, in those individuals that did not lose the predicted amount of weight (compensators) there was a increase in caloric intake while in those that lost the predicted amount of weight (non-compensators) there was no change in caloric intake. 
Unlikely sources of the observed compensation in the DREW study include changes in resting metabolic rate and the thermogenic effect of food as previous studies have found these variables not to be influenced by exercise training. 
Our step counter data suggests that there was no change in outside activity in the exercise groups, nor were there any differences in average daily steps among the exercise groups. 
Though doubly-labeled water or accelerometer data would have presumably been more accurate than our step data, it is still unlikely that substantial changes in non-exercise activity explain our findings.
In the previously published DREW primary outcomes paper we reported that the mean follow-up waist circumference was lower in all exercise groups compared to the control group and there was no difference in mean follow-up weight across groups. 
The current analysis allowed us to further explore the change in weight data including examining individual level data as well as weight compensation. Further we were able to examine if exercise induced changes in waist circumference are independent of weight loss. Our observation that adjusting for changes in weight did not meaningfully affect the change in waist circumference in any of the exercise groups as well as the observation that even in the exercise sub-group that did not lose weight there was a reduction in waist circumference is in agreement to previous published reports. 
It is worth noting that the exercisers that did lose weight had approximately two times the loss of waist circumference compared to the exercisers that did not lose weight which is also in agreement with previous reports. 
Excess abdominal obesity is associated with increased risk of mortality, CVD, diabetes, insulin resistance and metabolic syndrome. 
These findings reinforce the positive health benefits of engaging in physical activity even in the absence of substantial weight loss.
The DREW study was not designed to examine differences in actual and predicted weight loss nor the mechanism responsible for these observations. As such our measures of food intake, a food frequency questionnaire, is not optimal to measure small changes in energy intake. While we cannot conclude that an increase in energy intake in response to increased energy expenditure is the source of the observed compensation, based on the work of others we conclude it is the most likely cause.
Our findings should not be interpreted as suggesting that lower doses of exercise are more effective in producing weight loss than higher doses. We emphasize that DREW was not a weight loss study and it was not designed to examine the nuances of exercise induced weight loss. As such we hope our findings stimulate more research into the variability and mechanisms of exercise induced compensation. In the 12 KKW group approximately 27% of individuals did not have significant compensation in response to exercise training suggesting there are individuals that can achieve higher doses of exercise without compensation. Though primarily focused on weight loss maintenance, not weight loss, data from the Weight Control Registry (WCR) demonstrates that individuals who have lost a substantial amount of weight and have maintained weight loss typically achieve 45 or more minutes a day of moderate intensity activity. 
These data from the WCR provides further data that there are individuals that either are not prone to compensation or who have developed strategies to combat compensation. Developing methods to identify individuals likely to compensate, and developing/testing strategies to combat this compensation are areas deserving of future work.
Our findings also offer important insight into the exercise dose required for the prevention of weight gain. Based largely on data from cross-sectional doubled labeled water studies, the 2005 USDA Dietary Guidelines and Institute of Medicine recommend obtaining at least 60 minutes of moderate intensity activity per day to prevent weight gain. 
Both the 4 and 8 KKW group in DREW not only did not gain weight but actually lost weight during the 6 month study period. These two groups performed 72 and 136 minutes per week of moderate intensity physical activity, which is considerably less than the recommended 60 minutes per day to prevent weight gain. Our findings highlight the need for more research, in particular large randomized controlled clinical trials, exploring the dose of physical activity required to prevent weight gain as 60 minutes a day may present itself as a daunting undertaking for most individuals but even more so for a habitually sedentary individuals. In addition, this dose may well lead to compensation by more individuals, and to less success in weight loss.
Strengths of the DREW study include that it is an efficacy study, using a tightly controlled exercise doses, with all exercise completed in the laboratory and extensive monitoring of exercise energy expenditure. We obtained excellent exercise adherence and had a low dropout rate. Monitoring of steps per day throughout the 6-month period indicates that outside physical activity remained constant throughout the trial for all exercise groups, thus ensuring that observed group differences were due to the prescribed exercise dose. Further, the exercise doses, including intensity, are easily obtainable and are well tolerated by sedentary women; and this has important public health implications for refining future physical activity recommendations.
The study has limitations because the sample is limited to sedentary, overweight or obese, postmenopausal women with elevated blood pressure. Thus, we do not know if the results will apply to other women or men. However, the study sample is a group that is likely to benefit from exercise training and represents a sizeable proportion, probably a majority, of U.S. women in the age range of 45 to 75 years. Further the study could have benefitted from a more rigorous and sensitive measure of energy intake. The food frequency questionnaire utilized in DREW prevented us from being able to definitively demonstrate that the observed compensation is the result of an increase in energy intake. However, it should be reiterated that examining compensation was not the a priori goal of DREW. The exercise training intensity was moderate and while this makes for good public health and clinical applicability, it is possible that higher levels of training intensity might produce different results in regard to compensation and changes in waist circumference.
In this study of previously sedentary, overweight or obese, postmenopausal women we observed no difference in the actual and predicted weight loss with 4 and 8 KKW of exercise (72 and 136 minutes respectively), while the 12 KKW (194 minutes) produced only about half of the predicted weight loss. We need to gain a better understanding of the mechanisms responsible for this exercise dose dependent phenomenon and develop strategies to identify and treat potential compensators.