Individual differences in inherited and acquired phenotypic characteristics may modify the response to a given exercise intervention resulting in substantial interindividual variability. Bouchard and colleagues (1–3) were among the first to demonstrate that for a given dose of exercise, the magnitude of response varies considerably, where some individuals experience substantial improvements for a given trait whereas others do not. A biological basis that partially explains the variation in response to exercise has been established in animal studies (4,5). However, in human trials, the veracity of the approach to determine the existence of individual variability has been questioned (6–9). Prior observations that purport to demonstrate exercise response variability are based on the assumption that observed variability is a consequence of exercise alone and ignore differences in response due to day-to-day biological fluctuation and measurement error. Critics assert that the variability in response that is not attributable to exercise must be isolated to adequately quantify the variability attributable to exercise alone (6–8).
Research to date has focused on the variability of cardiorespiratory fitness (CRF) in response to exercise (3,10–12), whereas anthropometric measures including waist circumference (WC) and body weight have received less attention. To our knowledge, no study has examined individual variability for WC response to exercise in adults and few have investigated the response for change in body weight (13–15). The adverse consequences of obesity, in particular abdominal obesity, are well documented and ubiquitous across virtually all body systems (16). Given the obesity epidemic globally and its associated medical costs (17), characterizing the heterogeneity in both body weight and WC response to exercise is clinically relevant and has important implications for exercise-induced improvements in cardiometabolic health.
Once variability in response to exercise is demonstrated, exploration of the determinants may help to explain why some individuals respond to exercise to a greater or lesser extent compared with others. In addition to phenotypic characteristics, it is possible that changes in behavior, including dietary consumption and/or physical activity performed beyond that prescribed (activities of daily living), explain some of the variability in WC and body weight normally attributed to the prescribed exercise dose.
The objectives of this secondary analysis are 1) to determine and quantify the existence of interindividual variability in WC and body weight reduction in response to standardized exercise, 2) to characterize the likelihood that an individual’s response is clinically meaningful, and 3) to identify potential determinants of the interindividual variability in response to exercise. Our findings will contribute to the ongoing debate concerning interindividual variability in response to exercise, challenge current approaches to dichotomous classification of individuals as “responders” and “nonresponders,” and address important considerations for exploring determinants of response.
Setting and participants
Details of the trial design and methods (18) and the primary findings have been published elsewhere (19). Briefly, we conducted a 24-wk, single-center randomized controlled trial with a parallel group design between September 1, 2009, and May 31, 2013. The primary objective of the original investigation was to determine the separate effects of exercise amount and intensity on WC and glucose tolerance among sedentary, abdominally obese men and women (n = 300). All participants provided written informed consent before participation. This study was approved by the Queen’s University Health Sciences Research Ethics Board.
Of those individuals originally randomized (n = 300), participants were excluded from this analysis if they did not complete the 24-wk trial (n = 83), did not have both WC and body weight measurements at week 24 (n = 2), had an exercise adherence (number of exercise sessions attended) of less than 80% (n = 26), or did not have any dietary intake data (n = 8). This resulted in a final sample of 181 participants.
Participants were randomized to one of four groups: no-exercise control (n = 44); low-amount, low-intensity exercise (LALI; n = 46) (180 and 300 kcal per session for women and men, respectively, at 50% of V˙O2 peak); high-amount, low-intensity exercise (HALI; n = 53) (360 and 600 kcal per session for women and men, respectively, at 50% of V˙O2peak); and high-amount, high-intensity exercise (HAHI; n = 38) (360 and 600 kcal per session for women and men, respectively, at 75% of V˙O2peak). All participants performed walking or jogging exercise on a treadmill for the time required to achieve the desired energy expenditure (kilocalories per session) five times per week at the prescribed intensity (relative to CRF [V˙O2 peak]) for 24 wk (19). Using heart rate and oxygen consumption data obtained from the baseline exercise test, the heart rate corresponding to approximately 50% (LALI and HALI) and 75% (HAHI) of maximum oxygen consumption was assigned for each participant. At these exercise intensities, the energy expenditure targets (exercise amount) for women and men were 180 and 300 kcal, respectively, for LALI, and 360 and 600 kcal, respectively, for both HALI and HAHI. Follow-up V˙O2 peak exercise tests were conducted at weeks 4, 8, and 16 to verify the heart rate–V˙O2 relationship with continual adjustments to time required to achieve the prescribed energy expenditure to account for improvement in CRF. Heart rate was continuously monitored at every session to ensure adherence to the prescribed exercise intensity. All exercise sessions were supervised by trained personnel, and all exercise participants were asked not to engage in any structured exercise outside of the supervised sessions.
Physical activity performed outside of the prescribed exercise regimen was monitored using ActiGraph GT3X accelerometers for 7-d periods at approximately weeks 0, 8, 16, and 25. Participants were required to wear the accelerometer for at least 4 d, 10 h·d−1 each monitoring period. Established accelerometer cut points were used to classify and estimate average duration of incidental physical activity (>100 counts per minute) (20).
During a 1-wk run-in period, all participants were instructed to maintain their body weight and record daily consumption of self-selected foods. During the intervention, participants were instructed to maintain the target daily energy intake estimated during the run-in period and were prescribed a balanced diet that is consistent with the general recommendations found in Canada’s Food Guide and which aimed to provide approximately 50% to 55% carbohydrate, 15% to 20% protein, and 30% fat.
All participants were asked to complete daily self-report diet records for the duration of the intervention as a strategy to help ensure compliance with the dietary recommendations. When necessary, the prescribed target energy intake was revisited and adjusted if change in body weight deviated from the predicted exercise-induced weight loss by greater than 1 kg. All dietary procedures were conducted and supervised by the intervention nutritionist.
Selection of diet records
Three-day diet records (two weekdays and one weekend day) were randomly selected from the run-in period (baseline) and at weeks 8, 16, and 25 of the intervention, which corresponded to the weeks wherein accelerometry data were obtained. In cases where 3-d diet records were unavailable at these time points, the next closest 7-d period was used. In cases where all three selected diet records were not complete, available data (i.e., two weekdays) were analyzed at baseline (n = 15), week 8 (n = 1), week 16 (n = 4), and week 24 (n = 8).
Dietary intake analysis
The selected diet records for each participant were entered into a recently developed web-based automated 24-h recall platform (R24W). Details of the R24W development (21) and validation (22) have been published elsewhere. Briefly, the R24W was developed for a French Canadian population and incorporated aspects of the Automated-Multiple Pass Method recommended by the U.S. Department of Agriculture for dietary self-monitoring (23). The R24W automatically calculates nutrient intakes and diet quality scores using the 2010 Canadian Nutrient File (24) or the Nutrient Database for Standard Reference of the U.S. Department of Agriculture when data from the Canadian Nutrient File were missing (25). Here, the R24W was used as a tool to compile the food intake data from the diet records.
Energy and macronutrient intake, and diet quality scores (Canadian Healthy Eating Index 2010 [C-HEI] and Mediterranean Score [MedScore]) from the 3-d diet records were averaged. The C-HEI was adapted from the 2005 American HEI (26) and reflects Canada’s current dietary recommendations. Briefly, the C-HEI consists of eight adequacy (total vegetables and fruit, whole fruit, dark green and orange vegetables, total grain products, whole grains, milk and alternatives, meat and alternatives, and unsaturated fats) and three moderation (saturated fats, sodium, and other food) components. Points are given for each category reflecting insufficient to excessive intake. The 11 individual components are summed to produce a single score between 0 and 100, with higher scores indicating greater adherence to Canada’s Food Guide (27).
The MedScore is based on the Traditional Healthy Mediterranean Diet Pyramid (28). The MedScore assigns adjusted partial scores (0–4 points) for 11 components of the Mediterranean pyramid: grains, fruits, vegetables, legumes, nuts and seeds, olive oil, dairy products, fish, poultry, eggs, sweets, and red meat/processed meat. The total score ranges from 0 to 44 points with higher scores reflecting a food pattern consistent with the traditional Mediterranean diet.
CRF was assessed using standard open-circuit spirometry techniques (SensorMedics Corporation, Yorba Linda, California) during a graded exercise test in which participants walked on a treadmill at a self-selected speed at zero elevation for 3 min, after which the incline was increased by 5% for 2 min, then by 2% every subsequent 2 min until volitional fatigue (19).
WC was measured using a spring-loaded tape measure (Gulick II) at the superior edge of the iliac crest at baseline, 8, 16, and 24 wk for all participants in an exercise group, and at baseline, 16, and 24 for participants in the control group. WC was measured twice at each time point. If the two measures differed by greater than 1 cm, a third measure was taken. The two closest measures were averaged for the final recorded value. Body weight was measured using a calibrated beam scale.
A one-way ANOVA was performed to compare continuous baseline variables between groups. A chi-square test was performed to compare the sex distribution between groups. To examine between-group differences for change in anthropometric, dietary, and physical activity variables, linear mixed-effects models for repeated measures over time were applied to each of the aforementioned dependent variables. The models were estimated by restricted maximum likelihood and used an unstructured covariance matrix to account for within-subject correlation between baseline, week 16, and week 24. The independent variables in the mixed model included intervention group, time, group–time interaction, sex, sex–time interaction, and age. Contrasts were constructed to estimate the between group differences in the changes from baseline to 24 wk.
To quantify interindividual variability in response to exercise, the variation due to random variability was separated from the variation because of the intervention alone by using the following equation described by Atkinson and Batterham (8):
. In this equation, SDR estimates the SD of the interindividual variation in response due to treatment by removing the observed variance of the control arm from the treatment arms. SDI and SDC represent the SD values of the change in the intervention and control group, respectively.
If the SDR value was greater than or equal to the predetermined minimal clinically important difference, then further exploration of potential determinants was performed. We considered that SD differences of 2 cm for WC and 2 kg for body weight compared with control were the minimally clinically important difference (MCID) based on evidence of the association between differences in WC and body weight with cardiometabolic risk factors and mortality (29).
To determine the probability that an individual’s response was greater than the MCID (2 cm for WC and 2 kg for body weight), the “Precision of the estimate of a subject’s true value” Excel spreadsheet, developed by Hopkins (30), was used. The spreadsheet estimates the probability that a subject’s true change is greater than the MCID, adjusting for the technical error of measurement (31,32).
The previously described mixed-effects model was extended to assess potential determinants of interindividual variability for change in WC and body weight. The model was coded to allow additional response variance due to the exercise groups to be isolated from the control group. Baseline value, sex, variation in energy intake, diet quality scores, change in physical activity, and an interaction term between each of these variables and group were added to the model separately to identify potential determinants of individual variability in response to exercise. If the variable accounted for some of the variability in response to exercise, then the extra variance for that exercise group, beyond that in the control, would be attenuated, and this attenuation provided an estimate of proportion of individual variability in response to exercise that could be attributed to the given variable. To increase statistical power, we repeated the analysis pooling the three exercise groups, with the assumption that the determinants of nonexercise variability are similar and have the same magnitude of influence across all exercise groups.
A two-sided α of 0.05 was used to determine statistical significance, and no adjustment was made for multiple comparisons. All analyses were performed using SAS, version 9.4 (SAS Institute, Cary, NC), and SPSS software, version 24.0 (SPSS Inc., Chicago, IL).
Participant characteristics are summarized in Table 1. With the exception of MedScore, where the HAHI group had a slightly worse score than the other groups, there were no significant between-group differences for any baseline characteristic (P > 0.05).
Table 2 shows the separate effects of exercise amount and intensity on change in anthropometric variables, as well as the concurrent changes of diet and physical activity variables at 24 wk. As shown previously, reductions in WC were greater in the LALI (adjusted mean difference, −4.7 cm; 95% confidence interval [CI], −6.4 to −3.0), HALI (−5.3 cm [−7.0 to −3.6]), and HAHI (−5.8 cm [−7.6 to −4.0) groups compared with control (all P < 0.05) but did not differ from each other (all P > 0.05). Similarly, reductions in body weight were greater in the LALI (adjusted mean difference, −4.0 kg; 95% CI, −5.7 to −2.3), HALI (−5.1 kg [−6.7 to −3.4), and HAHI (−5.2 kg [−7.0 to −3.4]) groups compared with control (all P < 0.05) but were not different from each other (all P > 0.05). Change in C-HEI was greater for the LALI (7.2; 0.1 to 14.0) group compared with controls (P = 0.048); however, changes in self-reported energy intake and MedScore did not differ between groups at 24 wk (both P > 0.05). With the exception of HALI (3.1 [0.2 to 5.9]), the change in amount of physical activity (% of day) performed outside of the exercise prescribed was not significant compared with control (all P > 0.05).
Figure 1 and Figure 2 illustrate the individual change values for WC and body weight at 24 wk. For change in WC at 24 wk, the SDR values were 3.1, −0.3, and 3.1 cm for LALI, HALI, and HAHI, respectively. For change in body weight at 24 wk, the SDR values were 3.8, 2.0, and 3.5 kg for LALI, HALI, and HAHI, respectively.
Figure 1 and Figure 2 also display the probability that the individual change values for WC and body weight are greater than the MCID after adjusting for the technical error of measurement. For body weight, the HAHI group had the highest proportion of individuals whose response was considered “likely” or “very likely” to be clinically meaningful (LALI, 49.9%; HALI, 66.1%; HAHI, 68.4%), whereas the LALI group had the highest proportion of individuals whose response was considered “unlikely” to be clinically meaningful (LALI, 8.7%; HALI, 0%; HAHI, 2.6%). For WC, the HALI group had the highest proportion of individuals whose response was considered “likely to be clinically meaningful (LALI, 50%; HALI, 69.8%; HAHI, 60.5%). Similar to body weight, the LALI group had the highest proportion of individuals whose response was deemed “unlikely” to be above the clinically meaningful threshold for WC (LALI, 6.5%; HALI, 3.8%; HAHI, 5.3%).
No dietary or physical activity variable was a significant determinant of interindividual variability for change in WC or body weight for any exercise group (all P > 0.05). Baseline values accounted for approximately 19% and 16% of the interindividual variability for change in WC (P = 0.05) and body weight, respectively, in response to exercise for LALI alone, but did not account for significant variability in response to the other exercise doses. Similarly, sex accounted for approximately 13% of the interindividual variability for WC within the LALI group (P = 0.01) and did not account for interindividual variability in any group for change in body weight (all P > 0.05).
We further examined the contribution of potential predictors of interindividual variability for change in WC and body weight in response to exercise collapsed across groups. No dietary or physical activity variable accounted for interindividual variability for change in WC or body weight in response to exercise (all P > 0.05). Sex accounted for ~13% of the interindividual variability for change in WC (P = 0.04) and approached significance for change in body weight (P = 0.06). Baseline values did not account for the variability in the 24-wk change in WC (P = 0.10) but approached significance for change in body weight (P = 0.06)
We observed substantial interindividual variability in response to exercise independent of amount or intensity for change in WC and body weight after accounting for the variability not attributable to the exercise interventions. Further, we observed that the high-amount exercise groups (HALI and HAHI) yielded the highest proportion of individuals with a clinically meaningful response and the lowest proportion of individuals deemed unlikely to respond to exercise. Our analytical approach responds to assertions that prior investigations failed to consider the individual response to exercise that was beyond the variability explained by the day-to-day biological variation and measurement error and also extends dichotomous approaches to classifying individuals as “responders” and “nonresponders.” Thus, our findings derived from randomized controlled trial evidence are novel and reinforce prior observations from human and animal genetic studies demonstrating that a substantial heterogeneity exists among adult men and women in response to a standardized exercise dose (4,5).
To our knowledge, this is the first study to quantify interindividual variability for change in WC and body weight in response to standardized exercise after accounting for the variability not attributable to exercise. Since the early observations of Bouchard and colleagues (1,33), a growing body of evidence has supported the notion that substantial variability exists for numerous traits in response to exercise (3,11,14,34). However, without exception, prior studies have assumed that the observed variability was a result of differences in response to exercise alone. A failure to consider the variability due to day-to-day biological fluctuations and measurement error (random variability) precludes the accurate quantification of interindividual variability for a given trait. To address this, our study applied the approach suggested by Atkinson and Batterham (8) that distinguishes the random variability from the intervention variability by subtracting the SD of the control group from that of the intervention groups. By adequately quantifying the magnitude of interindividual variability in response to exercise, our investigation overcomes the limitations of prior trials and extends the observations of years of research on individual variability due to exercise, suggesting that each individual responds uniquely to a standardized treatment; however, this method does not account for possible within-person variation to repeated treatments.
Although the SDR value proposed by Atkinson and Batterham represents a straightforward method to isolate the variability due to treatment from random variation, this approach is not without limitations. The underlying assumption of the SDR equation is that the biological variation and the random error present in the control group are similar to that of the intervention group, and thus the remaining variability in the intervention group can be attributed to exercise. However, the utility of this equation is questioned in instances wherein the variability in the exercise group is, in fact, homogenized due to treatment or where the control group is contaminated with other sources of variability, beyond that of which is random (35). For instance, the finding that the HALI group had less variability for change in WC than the control group highlights the limitations of the SDR value as it is unlikely that variability in individual response for WC existed for only two of the three exercise groups. Although this finding remains unexplained, that the HALI group experienced substantial variability in body weight and not WC, indicating that this observation may be due to chance. Without assuming that the variability in change not attributable to treatment is the same across all groups (including control), we would require more complex study designs, such as repeated crossover designs, that may not be practical in this setting.
To our knowledge, this is the first study to demonstrate the likelihood of a clinically meaningful response in WC and body weight to an exercise intervention. Traditional approaches broadly apply a dichotomous classification of individuals as “responders” or “nonresponders” based on a predefined threshold (MCID, technical error (TE), etc.) (10,34). Using this approach increases the likelihood of incorrectly classifying an individual’s response because any individual whose change is less than the chosen threshold is considered a “nonresponder” to exercise and vice versa, although substantial uncertainty exists. The use of the spreadsheet developed by Hopkins (2000) addresses limitations inherent in dichotomous classification and allows researchers and clinicians to estimate the probability that a response is clinically meaningful. Individuals whose response falls slightly below the threshold is classified as a possible responder, which represents a more accurate and optimistic description of the efficacy of an intervention for a given individual. Using this approach, clinicians can use their judgment to make adjustments to an individual’s exercise prescription based on the likelihood that their client is experiencing benefit. It is important to note that individuals’ observed responses using this approach cannot be attributed to exercise alone and is likely influenced by environmental, behavioral, and physiological factors. However, given that a practitioner is not likely to isolate changes in outcomes attributable to exercise alone in a clinical setting, the evaluation of observed responses to both exercise and external factors is more generalizable to a clinical context.
King et al. (36,37) have noted both behavioral and metabolic responses to exercise that may compensate for changes in energy balance to defend body weight loss and maintain homeostasis. These factors include alterations in energy intake, decreases in nonexercise activity thermogenesis, basal metabolic rate, energy cost of exercise, and substrate oxidation. However, in the present study, we were unable to demonstrate that any of the selected dietary or physical activity variables predicted the exercise-induced variability in WC and body weight change. For selected variables, despite no statistical significance determined by the CI except sex, the effect size was large, suggesting that we may be underpowered to detect determinants of variability in response. For example, in the HAHI group, women lost on average ~2.5 kg less than males relative to control, despite a P value >0.05 when comparing men and women. Such a large difference in body weight reduction is of clinical importance and should not be dismissed. Furthermore, despite collapsing across groups in an attempt to increase sample size, the results remained unchanged. That we were unable to identify predictors of variability despite our relatively large sample size suggests that either 1) there are no consistent predictors of individual response or 2) much larger samples are needed, which may not be practical. For instance, if we were to investigate whether the effect of treatment differed between sex in a simple 2 × 2 factorial design, it would require approximately 126 individuals per group to detect a moderate effect size of the treatment effect to achieve 80% power at a two-sided α = 0.05 (38). A total sample size (n = 504) of that magnitude is neither pragmatic nor feasible for most research groups. Therefore, we emphasize the need to develop pragmatic solutions, whether in initial study design or statistical approach, to elucidate potential determinants of interindividual variability in the future.
Our study has several limitations. Although the SDR approach accounts for random variability, it does not account for the within-subject variability (whether a participant responds similarly to repeated applications of the same intervention). Limitations inherent with self-reported diet records, including self-representation and reactivity bias, may prevent us from accurately assessing the influence of individual differences in both diet quality and change in energy intake on WC and body weight change. It has been previously reported that self-report diet records can underestimate energy intake and other macronutrients by as much as 4%–37% compared with recovery biomarker studies (39–41). However, this notion would be true across intervention groups. In addition, our sample is limited to sedentary, abdominally obese adult men and women who are primarily Caucasian. Thus, our findings may not apply to other populations differing in age and ethnicity. The principle strengths of this study include rigorously controlled exercise prescriptions with supervised exercise and frequent fitness tests. In addition, physical activity performed outside of the intervention was measured objectively using accelerometers at multiple times throughout the trial.
In summary, our study supports the existence of substantial interindividual variability in response to exercise for change in WC and body weight after accounting for random error and biological variability, although we were unable to identify any lifestyle-based determinants of the heterogeneity in response to exercise. These findings underscore the need for the appropriate and pragmatic quantification of individual variability for other health-related traits and support the continued exploration of determinants that may explain the substantial variability in response to exercise. However, we acknowledge that studies exploring individual response variability will need to be larger and more complex than studies simply comparing mean response between groups.
The authors thank all study participants who volunteered their time. This study was supported by Canadian Institutes of Health Research (grant no. OHN-63277).
The authors declare no conflict of interest. The results of the present study do not constitute endorsement by the American College of Sports Medicine. The results of the study are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation.
1. Prud’homme D, Bouchard C, Leblanc C, Landry F, Fontaine E. Sensitivity of maximal aerobic power to training is genotype-dependent. Med Sci Sports Exerc
2. Bouchard C. Individual differences in the response to regular exercise. Int J Obes Relat Metab Disord
3. Bouchard C, Rankinen T. Individual differences in response to regular physical activity. Med Sci Sports Exerc
. 2001;33(6 Suppl):S446–51.
4. Wisløff U, Najjar SM, Ellingsen Ø, et al. Cardiovascular risk factors emerge after artificial selection for low aerobic capacity. Science
5. Lessard SJ, Rivas DA, Alves-Wagner AB, et al. Resistance to aerobic exercise
training causes metabolic dysfunction and reveals novel exercise-regulated signaling networks. Diabetes
6. Hopkins WG. Individual responses made easy. J Appl Physiol
7. Hecksteden A, Kraushaar J, Scharhag-Rosenberger F, Theisen D, Senn S, Meyer T. Individual response to exercise training-a statistical perspective. J Appl Physiol
8. Atkinson G, Batterham AM. True and false interindividual differences in the physiological response to an intervention. Exp Physiol
9. Williamson PJ, Atkinson G, Batterham AM. Inter-individual responses of maximal oxygen uptake to exercise training: a critical review. Sports Med
10. Ross R, de Lannoy L, Stotz PJ. Separate effects of intensity and amount of exercise on interindividual cardiorespiratory fitness response. Mayo Clin Proc
11. Sisson SB, Katzmarzyk PT, Earnest CP, Bouchard C, Blair SN, Church TS. Volume of exercise and fitness nonresponse in sedentary, postmenopausal women. Med Sci Sports Exerc
12. Bonafiglia JT, Rotundo MP, Whittall JP, Scribbans TD, Graham RB, Gurd BJ. Inter-individual variability in the adaptive responses to endurance and sprint interval training: a randomized crossover study. PLoS One
13. King NA, Hopkins M, Caudwell P, Stubbs R, Blundell JE. Individual variability following 12 weeks of supervised exercise: identification and characterization of compensation for exercise-induced weight loss. Int J Obes (Lond)
14. Church TS, Martin CK, Thompson AM, Earnest CP, Mikus CR, Blair SN. Changes in weight, waist circumference
and compensatory responses with different doses of exercise among sedentary, overweight postmenopausal women. PLoS One
15. Donnelly JE, Hill JO, Jacobsen DJ, et al. Effects of a 16-month randomized controlled exercise trial on body weight and composition in young, overweight men and women: the Midwest Exercise Trial. Arch Intern Med
16. Despres JP, Lemieux I. Abdominal obesity
and metabolic syndrome. Nature
17. Anis AH, Zhang W, Bansback N, Guh DP, Amarsi Z, Birmingham C. Obesity and overweight in Canada: an updated cost-of-illness study. Obes Rev
18. Ross R, Hudson R, Day AG, Lam M. Dose–response effects of exercise on abdominal obesity
and risk factors for cardiovascular disease in adults: study rationale, design and methods. Contemp Clin Trials
19. Ross R, Hudson R, Stotz PJ, Lam M. Effects of exercise amount and intensity on abdominal obesity
and glucose tolerance in obese adults: a randomized trial. Ann Intern Med
20. Freedson PS, Melanson E, Sirard J. Calibration of the Computer Science and Applications, Inc. accelerometer. Med Sci Sports Exerc
21. Jacques S, Lemieux S, Lamarche B, et al. Development of a web-based 24-h dietary recall for a French-Canadian population. Nutrients
22. Lafrenière J, Lamarche B, Laramée C, Robitaille J, Lemieux S. Validation of a newly automated web-based 24-hour dietary recall using fully controlled feeding studies. BMC Nutrition
23. Moshfegh AJ, Rhodes DG, Baer DJ, et al. The US Department of Agriculture Automated Multiple-Pass Method reduces bias in the collection of energy intakes. Am J Clin Nutr
25. United States Department of Agriculture. National Nutrient Database for Standard Reference [Web page]. http://ndb.nal.usda.gov/ndb/search
. Accessed April 11, 2018.
26. Guenther PM, Reedy J, Krebs-Smith SM, Reeve BB, Basiotis PP. Development and Evaluation of the Healthy Eating Index-2005
. U.S. Department of Agriculture. Center for Nutrition Policy and Promotion; 2007.
28. Willett WC, Sacks F, Trichopoulou A, et al. Mediterranean diet pyramid: a cultural model for healthy eating. Am J Clin Nutr
29. Cerhan JR, Moore SC, Jacobs EJ, et al. A pooled analysis of waist circumference
and mortality in 650,000 adults. Mayo Clin Proc
30. Hopkins W. Precision of the estimate of a subject’s true value (Excel spreadsheet). A New View of Statistics sportsci org: Internet Society for Sport Science
31. Bouchard C, Blair SN, Church TS, et al. Adverse metabolic response to regular exercise: is it a rare or common occurrence? PLoS One
32. Centers for Disease Control and Prevention. National Health and Nutrition Examination Survey [Web page]. http://www.cdc.gov/nchs/nhanes.htm
. Accessed April 11, 2018.
33. Bouchard C, editor Human adaptability may have a genetic basis. Health risk estimation, risk reduction and health promotion. Proceedings of the 18th Annual Meeting of the Society of Prospective Medicine Ottawa: Canadian Public Health Association
34. de Lannoy L, Clarke J, Stotz PJ, Ross R. Effects of intensity and amount of exercise on measures of insulin and glucose: analysis of inter-individual variability. PLoS One
35. Waters L, Reeves M, Fjeldsoe B, Eakin E. Control group improvements in physical activity intervention trials and possible explanatory factors: a systematic review. J Phys Act Health
36. King NA, Caudwell P, Hopkins M, et al. Metabolic and behavioral compensatory responses to exercise interventions: barriers to weight loss. Obesity
37. King NA, Horner K, Hills AP, et al. Exercise, appetite and weight management: understanding the compensatory responses in eating behaviour and how they contribute to variability in exercise-induced weight loss. Br J Sports Med
38. Lachenbruch PA. A note on sample size computation for testing interactions. Stat Med
39. Trabulsi J, Schoeller DA. Evaluation of dietary assessment instruments against doubly labeled water, a biomarker of habitual energy intake. Am J Physiol Endocrinol Metab
40. Poslusna K, Ruprich J, de Vries JH, Jakubikova M, van’t Veer P. Misreporting of energy and micronutrient intake estimated by food records and 24 hour recalls, control and adjustment methods in practice. Br J Nutr
. 2009;101(2 Suppl):S73–85.
41. Prentice RL, Mossavar-Rahmani Y, Huang Y, et al. Evaluation and comparison of food records, recalls, and frequencies for energy and protein assessment by using recovery biomarkers. Am J Epidemiol