Study design and objectives
The E-MECHANIC study is a three-arm, 6-month randomized controlled trial (RCT). The selected exercise doses reflect current recommendations for (1) general health (8 kcal/kg body weight/wk (8 KKW), or approximately 800 to 1,000 kcal/wk) and (2) weight loss (20 KKW, or approximately 2,000 to 2,500 kcal/wk) [
14]. A nonexercise control group will also be included in the study.
The primary objectives of the E-MECHANIC trial are (1) to test whether energy intake increases in response to either or both of two doses of exercise (8 KKW vs. 20 KKW), (2) to test whether the discrepancy between observed and expected weight loss (Wt Lossdiff) differs between the two exercise groups (8 KKW vs. 20 KKW) and between the control and 20-KKW groups and (3) to determine whether change in energy intake mediates Wt Lossdiff.
The secondary objectives of the trial are (1) to determine whether change in resting metabolic rate (RMR; adjusted for change in body mass index (BMI)) and activity levels (excluding structured exercise) differ across treatment groups; (2) to determine whether change in body composition differs by exercise dose; and (3) to characterize participants who compensate (that is, fail to lose the amount of weight expected) and compare them with those who do not compensate (that is, lose the amount of weight expected or more).
Participants
The recruitment goal in the E-MECHANIC trial is a combined total of 198 men and women. The inclusion criteria are being sedentary (that is, not exercising for more than 20 minutes at least 3 days per week based on self-report and 1 week of accelerometer data) and being overweight or obese (BMI ranging from ≥25 kg/m2 to ≤45 kg/m2). The exclusion criteria are current consumption of more than 14 alcoholic drinks per week, smoking within the past 6 months, pregnancy, having been pregnant within the past 6 months, breastfeeding, history of weight loss surgery (if a participant had a gastric band that was removed, the participant may be eligible at the discretion of the Principal Investigator (PI)), current participation in a weight loss program, medical condition such as diabetes and cardiovascular disease (CVD; potential participants with a history of CVD who are under the care of a physician who is treating the CVD will be considered for enrollment in the study) and inability to complete the study within the designated time frame because of plans to move out of the study area.
Assessments and schedule of procedures
The E-MECHANIC trial will have three assessment periods during which procedures to collect data for the primary and secondary study outcomes will be carried out. The primary outcomes are (1) energy intake and (2) Wt Loss
diff. Energy intake will be measured using two methods: doubly labeled water (DLW) and laboratory-based food intake tests. The secondary outcomes are changes in RMR, objectively measured activity levels (excluding structured exercise) and body composition. The primary assessments will occur at baseline and at week 24, with identical procedures followed at both assessment periods, and a truncated assessment period at week 4 to allow for identification of potential short-term changes in the study outcomes. Table
1 outlines the schedule of the E-MECHANIC study procedures.
Table 1
Schedule of study procedures
a
Informed consent and HIPAA authorization | OR, SV1 | | | |
Randomization | | Day 0 | | |
Anthropometrics and body composition | | | | |
Height | RI1, SV1 | | | |
Waist and hip circumference | SV1 | | Day 28 | Days 160, 174 |
Weight | RI1, SV1 | Days −14, −7, 0 | Days 28, 35 | Days 160, 167, 174 |
Blood pressure and heart rate | RI1, SV1 | | Day 28 | Days 160, 174 |
ECG | SV1 | | | Day 160 |
DXA | | Day −14, 0 | | Days 160,174 |
Energy metabolism | | | | |
TEE based on DLW measurement | | Day −14 | | Day 160 |
Daily weight | | Day −14 to day 0 | | Day 160 to day 174 |
Urine collection | | Days −14, −7, 0 | | Days 160, 167, 174 |
RMR | | Day −14 | | Day 160 |
Physical activity | | | | |
Accelerometry | RI2 | Day −14 | Day 28 | Day 160 |
VO2max | SV2 | | | Day 163 |
Psychological questionnairesb
| | Day −14 | Day 28 | Day 160 |
Clinical chemistry panel (CBC, chem 15) | SV1 | | | Day 160 |
Food intake tests with VAS | | Day −4 | Day 28 | Day 170 |
Blood collection for satiety hormone archivec
| | Day −4 | | Day 170 |
The trial will include a multistage screening process prior to participant enrollment consisting of a 3- to 4-week period that includes 1 week of run-in. Following an initial web and/or telephone screen to determine eligibility and provide the potential participants with a brief description of the study, the potential participants will be scheduled for an orientation session where detailed information about the study will be provided. The detailed information will include the number and type of assessments, the length and nature of the exercise training and the time commitment required to complete the study. Next, potential participants will undergo three run-in visits meant to reduce attrition by allowing them to determine if they will be able to schedule the study into their weekly routine. All three run-in sessions must be completed within the 1-week period to satisfy the eligibility criteria. During screening, informed consent will be obtained from each eligible participant by study staff. Staff members will also verify that eligible participants are not currently meeting public health physical activity guidelines (that is, sedentary: <8,000 steps/day over the course of 1 week) [
15] by having them wear SenseWear Armband (BodyMedia, Pittsburgh, PA, USA) accelerometers.
After being screened into the trial, participants will complete the comprehensive baseline assessment and will be randomly assigned to one of the exercise arms (8 KKW or 20 KKW) or to the control group (nonexercise arm). Participants will be enrolled at a 2:1 female/male ratio and randomized separately to one of the three intervention groups at a 1:1:1 ratio to obtain an equal number of women and men in each treatment group. Participants who drop out (that is, fail to return for the 6-month examination) will not be replaced. Participants in the 8-KKW and 20-KKW intervention groups will continue to attend exercise sessions during week 4 and week 24 assessment periods. Participants who have not met the intervention energy expenditure goals by week 24 will be allowed to exercise up to 3 weeks after the week 24 assessment period.
Interventions
The intervention will commence within 48 hours after a participant’s randomization and span 24 weeks.
Control group
Participants assigned to the control group will receive multimedia health information twice weekly by text messaging or e-mail throughout the study period. The health information covers a variety of topics, including stress management and the benefits of eating fruits and vegetables [
16‐
19]. Monthly seminars will be available for control group participants and will cover topics related to a healthy lifestyle. In addition, they will be sent a quarterly newsletter that features information on healthy living. During the 6 months of study participation, control group participants will be instructed to maintain their baseline level of physical activity. After the 6-month study is complete, control group participants will have the option of exercising in the fitness center at Pennington Biomedical Research Center for eight sessions within 1 month after their last assessment visit (that is, day 174). The fitness center staff will be available to provide instruction and information on an appropriate exercise prescription. In addition, control group participants will be given an opportunity to meet with a dietitian following the intervention.
Exercise protocols for the 8-KKW and 20-KKW groups
In the 8-KKW and 20-KKW groups, the exercise intensity for each participant will be set at a target heart rate associated with 65% to 85% of peak oxygen uptake (VO
2peak). All exercise training will occur on a treadmill. The exercise prescription will initially be based on peak oxygen uptake determined by giving participants a graded treadmill exercise test. In brief, the treadmill speed will start at 2.4 mph on a level grade for 2 minutes, and then increases in speed and/or grade will be applied to produce about a 1 metabolic equivalent of task (MET) increase in workload every 2 minutes until volitional fatigue is reached. During exercise training sessions, the speed and grade of the treadmill will be altered to keep the participants within their target heart rate range (65% to 85% of VO
2peak), and the participants will be allowed to select their exercise intensity within that range. On the basis of speed, grade, participant’s weight and standard American College of Sports Medicine (ACSM) equations, energy expenditure will be calculated in real time and the length of each exercise session will be adjusted to meet the daily caloric goal [
12,
20]. The caloric goal of each session will be calculated by dividing the prescribed exercise dose (8 KKW or 20 KKW) by the participant-selected exercise frequency (that is, three, four or five times per week).
In order to precisely match the caloric exercise goals of the study, we will measure actual energy expenditure (AEE) while exercising and appropriately adjust exercise time. Within the first week of the exercise intervention and once every 2 to 4 weeks thereafter for the remainder of the intervention, trained personnel will measure caloric expenditure rate using a metabolic cart while the participant is walking on a treadmill at a predetermined speed and grade. The metabolic cart data will be used to (1) verify the accuracy of the ACSM equations on an individualized basis, (2) adjust the participant’s daily exercise time to account for changes in metabolic or biomechanical efficiency that may occur due to the exercise training and (3) ensure that calories carried over from previous weeks remain equivalent throughout the study. By correcting the daily exercise time, we can precisely increase or decrease the amount of calories expended per session to match the prescribed weekly exercise AEE.
During exercise training sessions, exercise intensity will be monitored by using a heart rate signal from a Polar transmitter placed around the participant’s chest (Polar Electro, Lake Success, NY, USA) and recorded every 5 minutes along with the participant’s subjective rating of perceived exertion using the Borg Rating of Perceived Exertion scale. Participants will be allowed to read, listen to music, watch television or socialize with study staff while they exercise. The training program consists of a 3-minute warm-up at a standard intensity level, and then the prescribed training intensity will be initiated. We expect that most participants randomized to the 8-KKW group will be capable of exercising at their required dose during the first week of exercise. Participants randomized to the 20-KKW group may not be able to complete their assigned dose of exercise immediately upon entry into the study. In the 20-KKW group, the exercise dose will progress from 8 KKW during the first week, to 14 KKW during the second week and ultimately to the full prescription (20 KKW) during the third week. By the end of the third week, participants will be expected to perform 100% of their weekly exercise dose for 3 or 4 days in the 8-KKW group and for 3 to 5 days in the 20-KKW group.
Table
2 shows the estimated number of calories and minutes per week required by participants to expend 20 KKW. The number of calories and minutes required per session for each exercise dose is shown for those assigned the weekly dose in 3, 4 or 5 days. The flexibility of allowing participants to choose the number of days per week that they would like to exercise allows for preferred differences in the length of exercise sessions. We have found that this flexibility helps with participant compliance and overall satisfaction.
Table 2
Estimates of energy expenditure and exercise duration
a
Treadmill (3.2 mph), 5% grade | 667 kcal | 92 | 500 kcal | 69 | 400 kcal | 55 |
Treadmill (4 mph), 5% grade | 667 kcal | 80 | 500 kcal | 62 | 400 kcal | 50 |
Secondary study outcome measures
Indirect calorimetry will be used to measure RMR on day −14 and day 160. Energy expenditure will be adjusted for change in BMI, as outlined by Martin
et al. [
37,
38]. Each participant’s resting metabolism will be measured over a 30-minute period after a 12-hour overnight fast using a MAX-II metabolic cart system (AEI Technologies, Pittsburgh, PA, USA). After the participant has quietly rested for 30 minutes, a transparent plastic hood connected to the cart will be placed over the participant’s head. For the duration of the test, the participant will be asked to remain motionless and awake. Prior to each measurement, the pneumotach flowmeter will be calibrated using a 3-L calibration syringe (Hans Rudolph Inc, Shawnee, KS, USA), and gas analyzers will be calibrated using standardized gas mixtures. The inspired gas samples will be diluted by flow adjustment to maintain a carbon dioxide concentration between 0.7% and 0.9%. Oxygen uptake and carbon dioxide production will be calculated using M-II software. The average of the last 20 minutes of the measurement will be used to calculate RMR using Weir equations [
39].
Activity level
SenseWear armbands will be used to measure physical activity and sedentary behavior at multiple time points throughout the study, including screening, the 2-week DLW period during baseline, week 4 and week 24. The armband is a small device that fits on the upper arm and has five types of sensors that continuously record data: a two-axis accelerometer, two galvanic skin response sensors, a heat flux sensor, a skin temperature sensor and a near-body temperature sensor. Body weight, height, handedness, age, sex and smoking status will be used in the calculation of energy expenditure, which will be carried out using proprietary software. The armband also records the length of time that it was worn on the body, which will allow us to detect noncompliance with data collection procedures. Activity data will be summated per day and displayed in 1-minute epochs, allowing data from days with poor compliance to be imputed or eliminated from the statistical analyses. Participants will be instructed to wear the armband 24 hours/day and remove it while showering, bathing or swimming. The participants randomized to one of the two exercise groups will be instructed to wear the armband during the exercise sessions. If the participant is <95% compliant with wearing the armband during screening or week 4 of testing, they will be asked to wear the armband for an additional 7 days. During the midpoint of baseline (day −7) and week 24 of testing (day 167), the participant’s armband will be checked for compliance. If found to be noncompliant, the participant will be counseled by study staff about the importance of wearing the armband as directed. Participants can be excluded prior to randomization for armband noncompliance.
In addition to quantifying duration on the body, the SenseWear armband records key aspects of sedentary and active behaviors, including components of posture allocation, such as (1) number of minutes lying down per day, (2) number of minutes of sleep per day, (3) number of steps taken per day and (4) minutes per day spent in activities of different intensities (sedentary, moderate, vigorous and very vigorous). The armband quantifies activity intensity on the basis of MET cut points; for example, ≤3.0 METs is classified as sedentary behavior. The armband also quantifies (1) TEE, (2) RMR, (3) activity energy expenditure (AEE) (4) mean MET cut point and (5) minutes per day spent in physical activity. Importantly, the armband quantifies aspects of active and sedentary behaviors not captured by other methods, namely, DLW and RMR. The armbands will be used to quantify important changes in active and sedentary behaviors in response to exercise (between the 8- and 20-KKW groups and between the control and 20-KKW groups). These armband data will also be used to determine if there have been different changes in active and sedentary behaviors in response to exercise between compensators and noncompensators.
Extensive validity data on the SenseWear armband have been published. These data show that the armband could be used to estimate TEE more accurately than DLW over the course of 10 days in free-living conditions [
40] and also estimated RMR more accurately than indirect calorimetry [
41]. The armband provides a good measure of the amount of time spent in activity [
42], and, of five accelerometers tested, it was the most accurate tool for estimating TEE across different intensities of activity [
43]. Other researchers found that the armband provides a valid and reliable measurement of energy expenditure during resting conditions and during different intensities of activity, though variability in measuring energy expenditure increased during activity [
44,
45].
Body composition
Measures of body composition, including fat and lean mass, will be assessed by dual-energy X-ray absorptiometry (DXA) (Lunar iDXA with Encore software version 13.60; GE Healthcare, Madison, WI, USA) at baseline (day −14 and day 0) and week 24 (day 160 and day 174). Lean mass measurements taken with DXA are used to adjust RMR for alterations in body composition [
37]. Measurement data from DXA recordings will also be used to explore whether fat or fat-free mass changes differ between treatment groups and between compensators and noncompensators. The use of DXA involves minimal X-ray exposure—about the same as 12 hours of background radiation from the sun.
Statistical power and sample size
The power analyses will focus on the primary outcomes, that is, energy intake (measured with DLW and laboratory-based food intake tests) and Wt Loss
diff. In our power analyses, we will make the following assumptions: (1) power ≥0.90 and above is ideal; (2) the significance level under the null hypothesis will be set at α = 0.05; (3) the power analyses will be based on the sample size expected at the end of the study (that is, after 10% attrition); and (4) null hypotheses will be tested against two-directional (that is, two-tailed) alternative hypotheses. We will rely on conservative estimates of effect size and liberal estimates of variance, as well as the assumption that 1 kg of weight represents 7,700 kcal energy, in our power analyses [
33]. We expect no less than 60 participants in each of the three groups to finish the study (8 KKW, 20 KKW and control). The estimated effect sizes are shown in Table
3. They indicate that statistical power is identical for the comparisons between the 8-KKW and 20-KKW groups and between the control and 20-KKW groups.
Table 3
Effect sizes and power calculations for the primary outcome measures
a
Energy intake (kcal), DLW | | | | | | | |
8 KKW vs. 20 KKW | 60 | 60 | 200 | 316 | 0.63 | 0.96 | 0.93 |
Energy intake (kcal), laboratory | | | | | | | |
8 KKW vs. 20 KKW, control vs. 20 KKW | 60 | 60 | 200 | 363 | 0.61 | 0.90 | 0.95 |
Wt Lossdiff
| | | | | | | |
8 KKW vs. 20 KKW, control vs. 20 KKW | 60 | 60 | 2.0 | 3.3 | 0.61 | 0.95 | 0.90 |
Energy intake
The power analyses for the two methods by which energy intake will be measured rely on the effect sizes given in Table
3. The power analyses for energy intake, measured on the basis of DLW, indicate that the study will have a power of 0.93 for us to detect a 200-kcal difference in changes in energy intake between the 8-KKW and 20-KKW groups and between the control and 20-KKW groups. The study will also have 0.95 power for us to detect a 200-kcal difference in changes in energy intake measured with food intake tests in the laboratory between the 8-KKW and 20-KKW groups and between the control and 20-KKW groups.
Discrepancy between expected weight loss and observed weight loss
The effect sizes for the calculations and the power analyses are summarized in Table
3. They indicate that the study will have a power of 0.90 for the Wt Loss
diff comparisons.
Data analysis
Analytic data will be archived in an electronic file. Prior to locking the file against further change, we will inspect all data for completeness, proper range and internal consistency. This data-cleaning process will include mock analyses by an analyst blinded to intervention group assignment. Any changes made will be documented in permanent records. Statistical analyses will be performed only after careful consideration is given to the assumptions underlying the statistical methods employed, using model diagnostics such as quantile plots of studentized residuals, component-plus-residual plots and examination of leverage points and outliers.
First, baseline participant characteristics (for example, age, sex, race, weight, BMI, waist circumference) will be summarized for each intervention group as counts and percentages for categorical variables and as means and 95% confidence intervals for continuous variables. The statistical significance of differential longitudinal changes in response to interventions will be assessed by employing repeated-measures mixed-effects models with maximum likelihood estimation and Kenward-Roger adjustment to the degrees of freedom of the relevant test statistic. The three interventions will be compared across assessment times (for example, baseline, week 4, week 24). Intervention, race, sex and assessment times will be taken as fixed effects. Participants within groups will be considered as having random effects. Covariates such as age and baseline assessments will be included in preliminary models and retained in the final analytic models if warranted. The model covariance structure (for example, unstructured, compound symmetric, autoregressive) across time will be investigated to enhance the efficiency of the statistical tests. The results for each outcome variable will be summarized as least-squares means and 95% confidence intervals for each intervention group across the assessment times. Model residuals will be tested to see if distributions are approximately Gaussian and data transformations (for example, logarithmic) will be performed if needed. Group differences in baseline characteristics will be tested, and corrective steps, such as poststratification and addition of model covariates, will be taken if necessary. Linear contrasts on least-squares means will be used to test pairwise differential mean changes in each outcome measure from baseline to subsequent assessment times between the three intervention groups. Null hypotheses will be tested against two-directional alternatives at the nominal 0.05 significance level using Tukey-Kramer–adjusted P-values when appropriate. All statistical analyses will be performed using SAS version 9.3 statistical software (SAS Institute, Cary, NC, USA).
Second, we will characterize compensators vs. noncompensators, where compensators will be defined as those who fail to lose the expected amount of weight during the exercise trial and noncompensators will be defined as those who meet or exceed the expected amount of weight loss during the trial. Differences between the groups on the following variables will be tested: (1) baseline values derived from self-report instruments, (2) baseline energy intake measures (for example, total kilocalories, macronutrient intake and fat and sugar intake) and (3) baseline activity levels (quantified by SenseWear armband readings). In addition, differences between compensators and noncompensators will be evaluated for sex and race disparities, and compensators and noncompensators will be compared with respect to changes in outcomes from baseline to week 24.
As an initial step in a mediator analysis, we will test whether changes in energy intake mediate or partially mediate the relationship between group assignment (exercise dose) and Wt Loss
diff. We will follow the methods of Baron and Kenny [
59] and determine whether Wt Loss
diff is predicted by group, with grouping variables coded as 8 KKW vs. 20 KKW and control vs. 20 KKW. Subsequently, we will test whether change in energy intake predicts Wt Loss
diff. Assuming that the results are significant, we will then determine whether group significantly predicts Wt Loss
diff independently of energy intake. If group does not independently predict Wt Loss
diff, then the relationship between group and Wt Loss
diff may be attributable to change in energy intake. That is, we will then conclude that change in energy intake mediated the relationship between group and Wt Loss
diff. If the relationship between group and Wt Loss
diff is reduced after accounting for change in energy intake, then partial mediation will have been demonstrated.