Changes in Pedometer-measured Physical Activity are associated with Weight Loss and Changes in Body Composition and Fat Distribution in Response to Reduced Energy Diet Interventions: the POUNDS Lost trial

Abstract

Aims:

To examine whether changes in objectively measured physical activity (PA) are associated with weight loss and changes in body composition and fat distribution in response to weight-loss diet interventions.

Methods:

This study included 535 participants with overweight/ obesity randomly assigned to 4 weight-loss diets varying in macronutrients. PA was measured objectively with pedometers, and body composition and fat distribution were measured using dual-energy X-ray absorptiometry (DEXA) scans and computed tomography (CT) scans at baseline, 6 months, and 24 months.

Results:

From baseline to 6 months, when the maximum weight loss was achieved, each 1000 daily steps increment in PA was associated with a greater reduction in body weight (β[SE]=−0.48[0.11]) and waist circumference (β[SE]=−0.49[0.12]). Similar inverse associations were found in changes in body composition and fat distribution (P<0.05 and FDR qvalue<0.1 for all). The trajectory of the above adiposity measures across the 24-months intervention period differed between the patterns of PA change. Participants with the largest increase in PA maintained their weight loss from 6 months to 24 months, while those with less increase in PA regained their weight. In addition, dietary fat or protein intake significantly modified the associations between changes in PA and changes in body weight and waist circumference over 24 months (P_ΔPA*diet <0.05).

Conclusions:

Changes in objectively measured PA were inversely related to changes in body weight, body composition, and fat distribution in response to weight-loss diets, and such relations are more evident in people with high-fat or average-protein diet compared with low-fat or high-protein diet.

Introduction

The prevalence of obesity has been increasing rapidly worldwide, leading to the growing risks of associated diseases, including cardiovascular disease, type 2 diabetes, and certain cancers^1–4. Reducing dietary intake has become a mainstream approach to promote weight loss and improve body composition and fat distribution^5–10. A body of weight-loss diet intervention trials has been conducted in the past decades^11–19. While the majority of these dietary interventions showed significant effects on weight loss^{12,13,15,16,19}, considerable individual variability in changes of body composition^12,14,15,17 and fat distribution^16,18 has been observed among the participants.

The American College of Sports Medicine has claimed less than 150 minutes of moderate-intensity physical activity (PA) per week results in no weight loss or minimal weight loss in the general population²⁰. Notably, in the behavioral weight-loss interventions with diet and PA, adequate amounts of PA combined with diet restriction increase weight loss modestly, and the additive effect of PA on weight loss is diminished as the level of diet restriction increases^7,20. Even though the participants in the randomized groups are generally instructed to comply with similar levels of PA, significant heterogeneity in changes of PA has been observed over time and associated with weight loss^21–24. Therefore, we hypothesized that variability of changes in PA might at least partly account for the individual differences in weight loss within the context of diet interventions. In addition, commonly used self-reported PA is subjective to measurement errors. To date, no studies have investigated the relationship between changes in objectively measured PA and variability in weight loss in response to reduced energy diet interventions.

Preventing Overweight Using Novel Dietary Strategies (POUNDS Lost) Trial¹¹ is one of the largest and longest weight-loss diet intervention trials among individuals with overweight or obesity. In the current study, we examined associations of changes in objectively measured PA with weight loss over 2 years. We particularly assessed the associations of PA changes and changes in various measures of body composition and fat distribution. In addition, we assessed whether the associations were modified by diet interventions varying in macronutrient intakes.

Methods

Study design

The POUNDS Lost study is a randomized dietary intervention trial conducted from October 2004 to December 2007 at two clinical research sites: the Harvard T.H. Chan School of Public Health in Boston, MA, and the Pennington Biomedical Research Center in Baton Rouge, LA¹¹. The study was approved by the human subjects committee at each site and by a data and safety monitoring board appointed by the National Heart, Lung, and Blood Institute. All participants provided written informed consent. Mass mailings to lists of registered voters or drivers were the primary means of recruitment. Detailed descriptions of study design, participants recruitment, inclusion, and exclusion criteria have been summarized¹¹. In brief, a total of 811 individuals who were obese or overweight (body mass index [BMI] ≥25 kg/M² and ≤40 kg/M) and aged 30 to 70 years were enrolled and randomized to one of four different calorie-restricted diets (Supplementary Figure S1). People with diabetes or unstable cardiovascular disease, using medications that affect body weight, or with insufficient motivation as assessed by interview and questionnaire were ineligible to participate¹¹. Eligible participants were randomly assigned to one of the four diets consisting of the following micronutrition compositions: 1) low fat (20% of energy), average protein (15% of energy); 2) low fat (20%), high protein (25%); 3) high fat (40%), average protein (15%); and 4) high fat (40%), high protein (25%). Thus, two diets were low fat (20%), and the other 2 diets were high fat (40%), 2 diets were average protein (15%), and the other 2 diets were high protein (25%), which constituted a 2-by-2 factorial design. A graded difference was thus produced in carbohydrate which ranged from 35% in high fat, high protein diet to 65% in low fat, average protein diet. Similar foods in different proportions were used for each diet prescription, representing a deficit of 750 kcal per day from baseline for each participant. Group and individual sessions combined with behavioral counseling were used to promote adherence to the assigned diets. A daily food diary and a web-based self-monitoring tool were used to provide feedback on how closely participants’ daily food intake met the nutrient and energy goals.

Study participants

In the current analysis, 535, 559, and 367 individuals with pedometer-measured PA were investigated at baseline, 6 and 24 months, respectively. Overall, PA increased similarly across the 4 diet groups²⁵. The study had 80% power to detect a 0.7-kg difference in total body fat loss and a 0.33-kg difference in visceral fat loss²⁶. A random sample of about 50% of the enrolled participants in the POUNDS Lost trial was selected to undergo dual-energy X-ray absorptiometry (DEXA) scans for body composition at baseline, 6 months, and 24 months of follow-up. A further 50% of the above individuals were randomly assigned to receive computed tomography (CT) scans, resulting in about 25% of the total participants. A total of 299 underwent a DEXA scan at the baseline examination, and 296 and 203 individuals had available data at 6 and 24 months. Of these individuals, 134, 132, and 96 received CT scans at baseline, 6 months, and 24 months.

Assessment of dietary intake and adiposity measures

Dietary intake was assessed by using a 5-days diet record at baseline to assess adherence to the diet. Biomarkers of nutrient intake¹¹, urinary nitrogen excretion for protein²⁷, and respiratory quotient for fat versus carbohydrates^28,29 were used to validate self-reported adherence to macronutrient targets. Demographic and socioeconomic factors including age, sex, race, and height were measured at baseline examination. Body weight and waist circumference were measured at baseline and at 6, 12, 18, and 24 months during the intervention. BMI was calculated as weight in kilograms divided by the square of height in meters (kg/m²). The DEXA scan was performed on a Hologic QDR 4500A bone densitometer for body composition after an overnight fast. Participants were measured in the supine position wearing a hospital gown. Total fat mass, total lean mass, whole-body total fat mass %, and trunk fat % were measured at baseline, 6 months, and 24 months. Total, visceral, deep subcutaneous, and superficial adipose tissue mass were measured by CT scanning and analyzed from eight contiguous cross-sectional images above and below the L-4/L-5 interspace at baseline, 6 months, and 24 months²⁶.

Assessment of physical activity

Besides the nutrition goals for the weight-loss intervention trial, the PA goal was to gradually increase moderate-intensity exercise from 30 minutes per week to 90 minutes per week during the first six months and to maintain constant thereafter over the trial for each diet group¹¹. PA was monitored by questionnaire³⁰ and by the online self-monitoring tool. Participants also received and wore pedometers that were attached at the beltline to measure average steps/day for consecutive 7 days. The average number of steps per day was recorded at baseline, 6 months, and 24 months in this sub-sample. We used the continuous average steps/day in the current study. Change in steps from baseline to 6 months and 24 months and from 6 months to 24 months were calculated and stratified into tertiles.

Statistical analysis

The primary outcome was the change in body weight and the secondary outcomes were changes in waist circumference, body compositions, and fat distribution among participants who remained in the 24-months trial. First, the association between PA (steps/day) and all adiposity measures over 24 months were evaluated using the linear mixed model (LMM) to account for the repeated measures of PA and adiposity outcomes, adjusting for age, sex, race, diet group, BMI at baseline, and baseline value of the corresponding outcomes (except for body weight). Moreover, general linear models (GLM) were performed to examine the associations between each concurrent change in PA and outcomes from baseline to 6 months and 24 months and from 6 months to 24 months. In the models evaluating the associations of changes in PA with outcomes from baseline to 6 months and 24 months, baseline PA (steps/day) was further included in the adjusted model. In the model that examined the changes in PA in relation to outcomes from 6 months and 24 months, PA and corresponding outcome values at 6 months were adjusted in the analysis. We also performed an analysis with the addition of concurrent changes in body weight in these multivariate-adjusted models to assess the associations of change in PA and adiposity outcomes beyond weight loss. To determine whether the association of changes in PA with changes in adiposity measures can be modified by dietary intake, potential interactions between changes in PA and diet groups (high/low-fat diet or high/average-protein diet) for the outcomes were tested. Finally, we performed additional analysis using the linear mixed model with time variable as a repeated measurement factor and examined associations of changes in PA with the trajectory of changes in these adiposity measures over the 24-months intervention by adding the interaction term of changes in PA and time variable. P <0.05 was considered statistically significant. To account for multiple comparisons, we considered a false discovery rate ((FDR; qvalue) correction as suggested by Benjamini-Hochberg^31,32. Since the adiposity outcomes are correlated and the cost of a false negative (missing a potentially important discovery) is high, we considered a fairly high false discovery rate < 0.1 (FDRqvalue < 0.1), as significant³³. Statistical analyses were performed with SAS version 9.4.

Results

The baseline characteristics of participants included in the study were shown in Table 1. Most participants were female (55.51%) and white (83.55%). Of the 535 participants, the mean ± SD age was 51.5 ± 9.2 years, with a mean BMI of 32.73 ± 3.86 kg/cm². At baseline, 299 and 134 individuals had complete body composition (DEXA measurements) and fat distribution (CT scan), respectively. The median value of PA was 6644 steps/day (interquartile range 3935). 382 and 256 participants had the complete measure of changes in PA from baseline to 6 and 24 months, respectively.

Table 1.

Characteristics of study participants who provided complete physical activity¹ information at the baseline examination

Variables	N2	Baseline value
Physical activity (steps/day)	535	6866 (3161)
median (25th, 75th)	535	6644 (4641, 8575)
Age, y		51.5 (9.2)
Female sex	535	297 [55.51%]
Race	535
Black	-	88 [16.45%]
White	-	447 [83.55%]
Height, cm	535	169.22 (9.07)
Body weight, kg	535	94.1 (15.72)
BMI, kg/m2	535	32.73 (3.86)
Waist circumference, cm	535	104.1 (13.18)
Diet group (fat/protein/carbohydrate)	535
0: High-fat, high-protein (40/25/35)	-	138 [25.79%]
1: High-fat, average-protein (40/15/45)	-	134 [25.05%]
2: Low-fat, high-protein (20/25/55)	-	127 [23.74%]
3: Low-fat, average-protein (20/15/65)	-	136 [23.42%]
Dietary intake per day
Energy (kcal)	259	2024.57 (587.06)
Protein (%)	259	18.24 (3.35)
Fat (%)	259	36.95 (6.11)
Carbohydrate (%)	259	44.45 (7.68)
Biomarkers of adherence
Urinary nitrogen, g	535	12.71 (4.5)
Respiratory Quotient	534	0.84 (0.04)
SBP, mmHg	535	119.54 (13.78)
DBP, mmHg	535	75.51 (9.68)

Data are mean (SD), median (25th, 75th), or N [% of the number of study participants].

¹Physical activity was measured by pedometers.

²N, number of study participants.

We then analyzed the associations between objectively measured PA (steps/day) and different adiposity measures over the course of the intervention using the mixed model. Repeated measures of PA by pedometers were significantly associated with decreases in body weight, waist circumference, and measures of body composition and fat distribution over 24 months (P<0.0001 for all) (Table 2).

Table 2.

Association between repeatedly measured physical activity and adiposity measures over the 24-months intervention period

	Physical activity (steps/day)1
	β (SE)	P	FDR (qvalue)
Body weight, kg	−0.69 (0.07)	<0.0001	<0.0001
Waist circumference, cm	−0.52 (0.05)	<0.0001	<0.0001
Body composition
Total fat, kg	−0.41 (0.05)	<0.0001	<0.0001
Total lean, kg	−0.14 (0.03)	<0.0001	<0.0001
Total fat mass %	−0.27 (0.03)	<0.0001	<0.0001
Trunk fat %	−0.33 (0.04)	<0.0001	<0.0001
Adipose tissue mass
Total	−0.28 (0.04)	<0.0001	<0.0001
Visceral	−0.12 (0.02)	<0.0001	<0.0001
Deep subcutaneous	−0.09 (0.02)	<0.0001	<0.0001
Superficial	−0.16 (0.03)	<0.0001	<0.0001

Beta represents the change in outcomes for the per 1000 increases in steps/day; FDR (qvalue) false discovery rate

¹Data with LMM model adjusting for age, sex, race, diet group, body mass index at the baseline examination, and variable of interest at the baseline examination (except for body weight).

We further analyzed the associations between changes in PA and changes in adiposity measures (Table 3, Figure 1). We observed significant associations of changes in PA with concurrent changes in body weight, waist circumference, total body fat, total lean mass, total fat mass %, and trunk fat % from baseline to 6 months and 24 months, after adjustment for age, sex, race, diet group, BMI at baseline, baseline PA (steps/day) and corresponding outcome at the baseline examination (P <0.05 for all outcomes, Table 3). For example, each 1000 daily steps increment in PA was associated with a greater reduction in body weight from baseline to both 6 months and 24 months (P<0.0001). From 6 to 24 months, the association between change in PA and change in total lean mass was attenuated to be non-significant. After false discovery rate correction, the significances above persisted (FDRqvalue<0.1 for all, Table 3).

Table 3.

Association between changes in physical activity and concurrent changes (Δ) in adiposity during weight loss intervention

	Changes of Physical Activity (steps/day)
		Baseline to 6 months1				Baseline to 24 months1				6 months to 24 months2
	N3	β (SE)	P	FDR (qvalue)	N3	β (SE)	P	FDR (qvalue)	N3	β (SE)	P	FDR (qvalue)
Δ Body weight, kg	382	−0.48 (0.11)	<0.0001	0.0002	256	−0.63 (0.14)	<0.0001	0.0002	340	−0.31 (0.09)	0.0003	0.0015
Δ Waist circumference, cm	382	−0.49 (0.12)	<0.0001	0.0002	256	−0.73 (0.14)	<0.0001	0.0002	340	−0.37 (0.09)	<0.0001	0.001
Δ Body composition
Δ Total fat, kg	218	−0.49 (0.11)	<0.0001	0.0002	145	−0.64 (0.15)	<0.0001	0.0002	185	−0.29 (0.09)	0.002	0.005
Δ Total lean, kg	218	−0.15 (0.06)	0.0096	0.012	145	−0.18 (0.08)	0.033	0.0471	185	−0.08 (0.06)	0.1624	0.1804
Δ Total fat mass %	218	−0.35 (0.07)	<0.0001	0.0002	145	−0.42 (0.09)	<0.0001	0.0002	185	−0.20 (0.06)	0.0015	0.005
Δ Trunk fat %	218	−0.46 (0.10)	<0.0001	0.0002	145	−0.50 (0.12)	<0.0001	0.0002	185	−0.23 (0.08)	0.0031	0.0052
Δ Adipose tissue mass
Δ Total	73	−0.22 (0.10)	0.0352	0.0391	53	−0.21 (0.13)	0.1122	0.1247	76	−0.20 (0.07)	0.0044	0.0063
Δ Visceral	90	−0.14 (0.04)	0.0011	0.0017	66	−0.14 (0.05)	0.0045	0.0075	90	−0.05 (0.03)	0.0784	0.098
Δ Deep subcutaneous	90	−0.13 (0.04)	0.0012	0.0017	66	−0.06 (0.04)	0.1274	0.1274	90	−0.04 (0.03)	0.1905	0.1905
Δ Superficial	73	−0.11 (0.06)	0.0832	0.0832	53	−0.13 (0.08)	0.1037	0.1247	76	−0.13 (0.04)	0.0028	0.0052

Beta represents the change in outcomes for the per 1000 increases in steps/day; FDR (qvalue) false discovery rate

¹Data with GLM model adjusting for age at the baseline examination, sex, race, diet group, body mass index at the baseline examination, baseline steps, and variable of interest at the baseline examination (except for body weight).

²Data with GLM model adjusting for age at the baseline examination, sex, race, diet group, body mass index at the baseline examination, steps at 6 months, and variable of interest at 6 months.

³N, Number of participants eligible for the analysis.

Figure 1.

Trajectories of changes in adiposity measures according to changes in physical activity over 24 months. (A) Mean change of body weight (kg); (B) Mean change of waist circumference (cm); (C) Mean change of total fat (kg); (D) Mean change of total lean (kg); (E) Mean change of total fat mass %; (F) Mean change of trunk fat %; Data are means± SE after adjustment for age, sex, race, diet group, body mass index at the baseline examination, baseline steps and variable of interest at the baseline examination (except for body weight); Circle filled symbol, smallest tertile (T1) PA; Square filled symbol, median tertile (T2) PA; triangle filled symbol, largest tertile (T3) PA; For changes of physical activity, median (25th, 75th) values were T1(light grey): −1630.9 (−2592.1, −903.6) steps/day, T2(dark grey): 1144.4 (535.6, 1572.7), and T3(grey): 4277.7 (3098.1, 5173.6), respectively, among the included participants, the largest tertile represents the largest increment of physical activity (step/day) from baseline to 24 months.

Changes in most of the abdominal fat measures (total, visceral, and deep subcutaneous adipose tissue mass) were also strongly correlated with changes in PA from baseline to 6 months. From baseline to 24 months, associations of changes in PA with changes in total and deep subcutaneous adipose tissue mass were attenuated and no longer significant. Conversely, the associations between changes in PA and changes in superficial adipose tissue mass were significant from 6 to 24 months with adjustment for steps and superficial adipose tissue mass value at 6 months. The significances persisted when accounting for the false discovery rate (FDRqvalue<0.1, Table 3).

After further adjustment for concurrent weight change in the model, the associations between changes in PA and changes in total adipose tissue mass from baseline to 6 months or changes in waist circumference, total fat mass %, and visceral adipose tissue mass from baseline to 24 months generally remained significant but the magnitude decreased (Supplementary Table S1). However, all other relations were no longer significant when controlling for concurrent weight change in the model. After false discovery rate correction, the significance of waist circumference, total fat mass %, and visceral adipose tissue mass from baseline to 24 months remained, but the significance of total adipose tissue mass from baseline to 6 months disappeared (FDRqvalue = 0.101 for total adipose tissue mass, Supplementary Table S1).

To examine how different patterns of change in PA from baseline to 24 months were related to adiposity changes (Figure 1), 256 participants were categorized into tertiles. There were significant interactions of tertiles of change in PA by time for changes in body weight, waist circumference, body composition measures, and visceral adipose tissue mass over time (P_ΔPA*time <0.05 and FDRqvalue<0.1 for all), indicating that the changes for these adiposity measures across the intervention period differed between patterns of change in PA. Participants engaging in the largest increase in PA (tertile 3) maintained their mean weight loss of −9.6 kg (SE=0.65) at 24 months, while those engaging in less increase in PA (tertile 1 and 2) regained their weight from 6 months to 24 months (Figure 1A). The associations of changes in PA and changes in adiposity were more pronounced at 24 months than at 6 months for body weight, waist circumference, body composition measures, and visceral adipose tissue mass (Table 3). The PA-time interaction effects for waist circumference, trunk fat %, and visceral adipose tissue mass remained significant after additional adjustment for concurrent weight change (P_ΔPA*time = 0.0265, 0.0188, and 0.0456, respectively).

We also analyzed the interactions between changes in PA and diet Interventions. During the 24-month intervention, we found that there were no significant differences in mean or median PA (steps/day) at all the time points across different diet intervention groups but there was a difference in mean change of PA (steps/day) from baseline to 24 months (Supplementary Table S2). We found significant interactions between dietary fat or protein intake and changes in PA on predicting changes in body weight and waist circumference at 24 months. Specifically, an increase in PA was associated with a greater decline in body weight and waist circumference in participants assigned to the high-fat diet or average-protein diet than those with the low-fat diet or high-protein diet (P_ΔPA*diet =0.035, 0.0077, 0.0077, and 0.0474, respectively) (Figure 2). However, after false discovery rate correction, the significance of the interaction between dietary fat and change in PA on body weight did not persist (FDRqvalue = 0.175, Supplementary Table S3). There was no interaction between dietary fat intake and changes in PA in predicting any changes in body composition and abdominal fat from baseline to 24 months. However, there were significant interactions between changes in PA and dietary protein intake on changes in body composition and visceral adipose tissue mass at 24 months relative to baseline, and the significances persisted after false discovery rate correction (FDRqvalue<0.1, Supplementary Table S3).

Figure 2.

Changes in body weight and waist circumference according to changes in physical activity in low/high-fat diet group and low/high-protein diet group at 24 months. (A) Changes in body weight for low-fat or high-fat diet group; (B) Changes in waist circumference for low-fat or high-fat diet group; (C) Changes in body weight for average-protein or high-protein diet group (D) Changes in waist circumference for average-protein or high-protein diet group; Data are predicted fit with 95% confidence limits for the mean (CLM) predicted value after adjustment for age, sex, race, body mass index at the baseline examination, baseline steps and variable of interest at the baseline examination (except for body weight)

Discussion

In this 24-month randomized, dietary weight-loss intervention trial, we found that changes in objectively measured PA were inversely associated with weight loss and changes in body composition and fat distribution. The trajectory of the adiposity measures across the intervention period differed between the patterns of change in PA. In addition, we found that the associations between changes in PA and adiposity changes were significantly modified by dietary fat or protein intakes.

This is the first study, to our knowledge, to investigate the relationships between changes in objectively measured PA by pedometers and changes in adiposity measures in a long-term diet intervention trial. In our study, while the PA goal was to gradually increase moderate-intensity exercise to 90 minutes per week among the participants, as recommended in the intervention method, significant variability in the changes of PA was observed at the individual level. We found that changes in PA were inversely related to change in body weight, body composition, and fat distribution in response to the dietary interventions over 24 months. This finding suggested that individual variability in weight loss in the context of diet interventions may be at least partly attributed to the differences in changes of PA among the participants. In contrast to our finding, as reported by Donnelly et al., a minimum of 150–250 min/week of moderate-intensity PA will improve weight loss in combination with moderate, but not severe, diet restriction²⁰. Our results differed to a certain extent from the guidelines of the American College of Sports Medicine in a sample of overweight or obese participants with less than 150 minutes per week of moderate-intensity PA²⁰. The discrepancy may be attributed to the significant individual variability of PA over diet interventions through the use of pedometers, versus self-report. Thus, future weight loss efforts among overweight or obese adults with moderate diet restrictions may target a lower level of PA compared to the recommendation from the American College of Sports Medicine. Further investigation is required to establish the validity of the results. Moreover, with the same degrees of changes in PA, the reduction of fat mass was greater than that of lean mass, and more reduction in visceral than deep subcutaneous or superficial adipose tissue mass. The consistent associations of changes in PA with changes in waist circumference, total fat mass %, and visceral adipose tissue mass support the idea that an increase in PA may be particularly beneficial for the reduction in waist circumference, visceral fat, or total fat, beyond total body weight loss.

In addition, we found that the overall trajectory of body weight, body composition, and fat distribution differed according to the patterns (tertiles) of PA change. In the Pounds Lost trial, overall, participants reached the maximum weight loss at 6 months and regained weight afterward¹¹. We found that participants with the largest increment in PA maintained weight loss, while the others regained body weight, from 6 to 24 months. A similar trajectory was also observed for the changes in body composition and fat distribution. Our results indicate that increasing PA may effectively limit regain of body weight and adiposity measures during long-term diet interventions. Nonetheless, we acknowledged that the changes in PA might be a marker of adherence to dietary intervention since adherence was associated with successful weight loss in Pounds Lost¹¹. Further studies are warranted to verify our findings.

Intriguingly, we observed that the relations of changes in PA with changes in adiposity measures were modified by diets varying in fat or protein. Greater reduction in adiposity measures was observed in participants assigned to the high-fat or average-protein diet than those in the low-fat or high-protein diet. A synergistic effect of PA and dietary fat on weight loss and body fat has been suggested in previous studies³⁴. The beneficial effect of a high-fat ketogenic diet was found to be facilitated by aerobic exercise through increased gene expression and enzymes involved in fatty acid oxidation³⁵. In addition, both high fat intake and exercise may improve lipid metabolic inflexibility and restore impairments in fatty acid oxidation capacity³⁶; therefore, we suggest that the synergistic effects may be partly via these pathways. Because high-fat intake is also characterized by low-carbohydrate intake, it is difficult to determine which macronutrient would best explain the interactions in our study. The evidence supporting the amplification of the beneficial effect of an increment of PA on weight loss by the average-protein diet is scarce, and more studies are needed to explore the potential mechanisms.

The major strengths of this study include the availability of objectively and repeatedly measured PA and long-term dietary interventions. Counting steps as a metric for quantifying PA is objective, intuitive, understandable to the layperson, and can be measured easily and accurately³⁷, therefore minimizing the errors in self-reported assessment. Our longitudinal change analyses of repeatedly measured PA and adiposity measures ensured robust, consistent, and biologically plausible relations³⁸. Multiple comparison issues were accounted for by Benjamini-Hochberg method to control the false discovery rate. The significant associations for most of the adiposity measures persisted and thus reduce the probability of chance findings. Nonetheless, we acknowledge several limitations. First, pedometer-measured PA only provides steps per day rather than more detailed information such as exercise volume³⁹, patterns⁴⁰, intensity^23,24,41, and types⁴² of PA. Second, some definitive and significant patterns and associations might have been obscured partly due to the relatively small sample size, investigating only the participants with complete PA and adiposity measures over 24 months, especially for the body composition and fat distribution by DEXA and CT scans. Therefore, results should be interpreted with caution owing to the low statistical power in these analyses.

In conclusion, the present evidence indicates that changes in PA may significantly account for the individual variability and trajectory of weight loss and changes in body composition and fat distribution in long-term dietary intervention studies. Our results also suggest that diets varying in fat or protein may modify the relationship between changes of PA and measures of adiposity.

Abbreviations used:

PA

physical activity

BMI

body mass index

DEXA

dual-energy X-ray absorptiometry

CT

computed tomography

Data availability:

Some or all datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.