Abstract
Objective:
Look AHEAD was a randomized trial comparing intensive lifestyle intervention (ILI) and diabetes support and education (DSE) on cardiovascular disease (CVD) among individuals with overweight/obesity and Type 2 Diabetes. We conducted a secondary analysis to evaluate the association between change in weight and waist circumference (WC) and CVD outcomes.
Methods:
We classified participants (n=5490) into four categories based on change in weight and WC between baseline and Year 1 (both increased, both decreased, etc.). We fit separate Cox proportional-hazards regression for ILI and DSE (using group that reduced weight/WC as reference) and compared time to first occurrence of primary and secondary CVD outcomes from year 1 through median of almost 10 years. Secondly, we evaluated time to first event among all four ILI groups relative to DSE.
Results:
Within DSE, CVD outcomes did not differ. ILI participants with increased WC had increased risk of primary outcomes (regardless of weight loss (HR[95%CI]: 1.55(1.11, 2.17) or gain HR[95%CI]: 1.76(1.07, 2.89)) and secondary outcomes (overall p<.01) relative to ILI participants who reduced weight and WC and relative to DSE participants.
Conclusions:
In this secondary analysis, increased WC during the first year of ILI, independent of weight change, was associated with higher risk for subsequent cardiovascular outcomes.
Keywords: Obesity, waist circumference, cardiovascular disease, central adiposity, weight loss
Introduction
Obesity is prevalent in the United States (1), contributing to high rates of cardiovascular and metabolic disease (2). Behavioral weight loss is an evidence-based treatment for obesity, with the primary goal of facilitating a weight reduction that will reduce the risk of or severity of adiposity-related morbidity and mortality. It is well-established that modest weight losses of 5-10% of initial body weight are associated with improvements in cardiovascular risk factors and glucose control (3-5). The potential cardiovascular benefits of weight loss may be especially pertinent for individuals with Type 2 Diabetes, as these individuals are at increased risk for cardiovascular morbidity and mortality (6-8).
The Look AHEAD trial was a randomized controlled trial designed to evaluate the effect of long term weight loss intervention on risk for cardiovascular morbidity and mortality among adults with overweight or obesity and Type 2 Diabetes. Despite individuals in the lifestyle intervention achieving significant weight losses (8.6%) and improvements in cardiovascular risk factors, there was no difference between groups in incidence of cardiovascular events over a median follow-up of 9.6 years (3). Secondary analyses of the trial have helped to clarify subgroups in which behavioral weight loss treatment might influence cardiovascular risk and outcomes, such as those who achieved a weight loss beyond a threshold of 10% and those who increased their level of physical fitness by at least two metabolic equivalents (METS) also observed reduced risk of secondary cardiovascular outcomes (9).
Another pathway through which weight loss treatment might confer cardiovascular benefits is through reductions in central adiposity. Central adiposity, i.e. body fat that is concentrated in the abdominal or mid-region of the body, is especially associated with risk for cardiometabolic disease (i.e., cardiovascular disease and diabetes mellitus), even after adjusting for BMI (10-11). In fact, recent findings suggest that waist circumference (a proxy measurement for central adiposity) varies considerably even among individuals with the same BMI (12-13) and can further delineate variability in cardiometabolic risk (14). Since changes in weight and waist circumference tend to be highly correlated in behavioral weight management (15-16), it remains unclear whether incorporating measurement of waist circumference change can provide additional information about the long-term effects of behavioral weight loss treatment on cardiovascular morbidity and mortality.
As the primary objective of weight management is to reduce risk of medical morbidity and mortality, it is essential to clarify the best indicators of adiposity-related risk and to identify the best tools for monitoring change in risk from intervention. The goal of the current observational study was to evaluate among participants in the Look AHEAD trial how direction of change in weight and waist circumference between baseline to year one is associated with cardiovascular outcomes. To test this question, we used two different approaches. In the first approach we evaluated how change in weight and waist circumference was associated with primary and secondary cardiovascular outcomes in models stratified by treatment condition. This approach allowed us to better understand variability in response to behavioral weight management within the treatment and control groups and to explore the cardiovascular implications related to variable changes in body weight and central adiposity. In the second approach we utilized the DSE group as the reference to evaluate the time to first event based on weight and waist circumference changes among ILI participants relative to DSE. This is more aligned with a per-protocol approach and allowed us to determine if there are subgroups within ILI at increased or reduced risk compared to individuals assigned to the control condition.
Methods
Participants.
Design.
The Look AHEAD trial was a multi-site randomized controlled trial designed to evaluate the effect of long term weight loss intervention on risk for cardiovascular morbidity and mortality among adults with overweight or obesity and Type 2 Diabetes. The design and primary results of the Look AHEAD trial have been published and discussed at length elsewhere (17-19). However, key aspects of the design are briefly reviewed for reference. Men and women between the ages of 45-76 years, with a BMI≥25 kg/m2, and who had a confirmed diagnosis of Type 2 Diabetes were randomly assigned to one of two treatment conditions: a behavioral weight loss Intensive Lifestyle Intervention (ILI) or a Diabetes Support and Education (DSE) control condition. Regardless of randomization group, all participants received a baseline class on diabetes management and were directed to follow their medical providers recommended course of medical treatment for managing Type 2 Diabetes.
Intensive Lifestyle Intervention.
Participants assigned to the ILI condition received a comprehensive behavioral weight management program designed to facilitate weight loss of ≤7% of initial body weight during the first year of the trial. To increase likelihood of achieving this goal, all participants were directed to aim for a 10% reduction. The program targeted dietary intake and physical activity with support for sustainable behavioral modification. Participants attended a combination of group and individual sessions with weekly meetings during months 1-6 and then weekly meetings three out of four weeks during months 7-12. During years 2-4, participants received two contacts per month (one in person and one via phone, email, or mail). After year 4, participants had at least two on-site contacts per year (17).
Diabetes Support and Education.
Participants assigned to the DSE condition received three group-based educational/social classes in years 1-4. Topics covered during these meetings included nutrition and exercise.
Measures
Participants were seen face-to-face at baseline and annually thereafter. The following measures were collected at these visits by research staff who were masked to the participants’ intervention condition.
Demographic information.
Participants self-reported basic demographic and health-related information including their age, sex, race/ethnicity, history of cardiovascular disease, and history of smoking.
Insulin use.
At baseline and annually thereafter, participants brought all medications they were currently taking including insulin for management of Type 2 Diabetes. Insulin was utilized in the current study as it reflects greater disease severity and diabetes-related cardiovascular risk.
Weight/height.
Weight was measured annually using a digital scale, with participants in light clothing and shoes removed. Height was measured at baseline using a wall-mounted stadiometer. BMI (kg/m2) was also calculated with these data.
Waist circumference.
Waist circumference was measured in centimeters while the participant was in light clothing with a nonmetallic, constant tension tape placed around the body at the midpoint between the highest point of the iliac crest and the lowest part of the costal margin in the mid-axillary line. This measurement was assessed twice with the average of the two values used for data analysis.
Cardiovascular Risk Factors at Baseline.
Blood was collected after a 12-hour fast and was analyzed by the Central Biochemistry Laboratory (Northwest Lipid Research Laboratories; University of Washington, Seattle WA) for total serum cholesterol, high-density lipoprotein, and triglycerides. Low-density lipoprotein was calculated using the Friedewald equation (20). Systolic and diastolic blood pressure was calculated by taking the average value of two measurements approximately 30 seconds apart. The assessment was completed after a 5-minute rest period using a calibrated device.
Cardiovascular outcomes.
Consistent with the pre-specified centrally adjudicated outcomes for the Look AHEAD trial, two composite scores were used representing primary and secondary cardiovascular outcomes. Study staff who were masked to treatment condition prompted participants to report medical events and hospitalizations every 6 months (in person or via telephone call). Hospital records were then reviewed for potential cardiovascular events which were adjudicated by a central review committee who were also unaware of study group assignment. Any of the following events that occurred after the year 1 assessment until the intervention was terminated (median of follow-up of approximately 10 years) were included.
Primary cardiovascular disease outcome included death from cardiovascular causes or first occurrence of non-fatal acute myocardial infarction, stroke, or hospital admission for angina.
Secondary cardiovascular disease outcomes included the events defined within the primary outcome in addition to coronary artery bypass grafting, carotid endarterectomy, percutaneous coronary intervention, admission to hospital for congestive heart failure, peripheral vascular disease, and total mortality.
Analysis
Changes in weight and waist circumference were calculated during the first year of the Look AHEAD trial as the most intensive phase of the lifestyle intervention was delivered between baseline and the end of the first year. Evaluating change within the first year as a predictor of cardiovascular outcomes in subsequent years allows us to draw more confident conclusions about the direction of the relationship between waist circumference change and cardiovascular events. Participants with missing weight or waist measures at baseline or year 1, who received bariatric surgery, or experienced an event included in the primary or secondary cardiovascular outcomes in the first year were excluded from these analyses. We assigned participants to one of four categories based on whether or not their weight and waist circumference increased or decreased from baseline to year 1 (any value greater than or equal to 0.0 was coded as an increase, anything less than 0.0 was coded as a decrease). Analysis of variance for continuous and chi-squared tests for categorical measures were utilized to compare demographic and study variables by treatment assignment and by weight and waist changes observed across the four groups.
To analyze the differences in the time to first primary or secondary cardiovascular outcomes among change and randomization groups, we implemented two different approaches. We conducted analyses to test for homogeneity of effect between weight and waist change groups across randomization assignment groups. Then we fit separate models for ILI and DSE using the group that lost weight and reduced weight circumference as the reference group and compared the time to first event in this group relative to the other three weight and waist change categories. In the second approach we utilized participants from both DSE and ILI. The DSE group was used as the reference to evaluate the time to first event among ILI participants in all four weight and waist change groups relative to DSE.
For each approach, we fit three sets of Cox proportional-hazards regression models with increasing levels of covariate adjustment. The first model was adjusted only for clinical site. The second set of models adjusted for sex, race, baseline age, and clinical site. The third set of models included all covariates from the second set in addition to potential cardiovascular disease-related confounders measured at baseline: cardiovascular disease history, insulin use, diabetes duration, weight, systolic blood pressure, diastolic blood pressure, LDL cholesterol, and smoking status. Type 3 tests of equal effects among weight and waist change categories were used to assess association between these categories and the outcomes. This means that the results for each weight and waist change category are fully adjusted for the other groups, regardless of the order that they are entered into the model. Kaplan-Meier plots of survival curves by randomization assignment and waist and weight change grouping were used to illustrate visual differences in event rates over follow-up. Alpha of 0.05 was used as the significance level for all tests. Analyses were performed using SAS version 9.4 (Cary, NC).
Results
Of the original sample of 5,145 randomized participants in the Look AHEAD Trial, 5 did not have baseline waist circumference, 167 did not complete a year 1 follow-up visit, 111 had missing weight or waist at the year 1 visit, 188 had gastric bypass during follow-up, 63 had a primary outcome during the first year, and an additional 21 participants had one of the secondary CVD outcomes during the first year. These individuals were excluded from the current study, resulting in a final sample of 4,590 individuals. Analysis of variance and chi-squared tests were conducted to ensure that this subsample of the total Look AHEAD sample retained the effects of randomization. As shown in Table 1, the two groups did not differ on any significant baseline characteristic including age, sex, race, or baseline weight and waist circumference.
Table 1.
Baseline Characteristics of the Subset of Look AHEAD Participants Included in this Secondary Analysis by Randomization Assignment
| Diabetes Support and Education (n=2260) |
Intensive Lifestyle Intervention (n=2330) |
P-value | |
|---|---|---|---|
| Age, mean ± SD, years | 59.1 ± 6.8 | 58.7 ± 6.8 | 0.0951 |
| Age Category (years), No. (%) | 0.1601 | ||
| 45-55 | 670 (29.6%) | 728 (31.2%) | |
| 55-65 | 1171 (51.8%) | 1217 (52.2%) | |
| 66-76 | 419 (18.5%) | 385 (16.5%) | |
| Sex, No. (%) | 0.7309 | ||
| Male | 916 (40.5%) | 956 (41.0%) | |
| Female | 1344 (59.5%) | 1374 (59.0%) | |
| Race, No. (%) | 0.9811 | ||
| African American / Black (not Hispanic) | 350 (15.5%) | 365 (15.7%) | |
| White | 1429 (63.2%) | 1461 (62.7%) | |
| Hispanic | 292 (12.9%) | 309 (13.3%) | |
| Other/Mixed | 189 (8.4%) | 194 (8.3%) | |
| Baseline Diabetes Duration, mean ± SD, years | 6.8 ± 6.5 | 6.8 ± 6.7 | 0.8617 |
| Baseline Smoking, No. (%) | 0.3584 | ||
| Never | 1168 (51.8%) | 1160 (49.9%) | |
| Past | 995 (44.1%) | 1059 (45.5%) | |
| Present | 92 (4.1%) | 107 (4.6%) | |
| Baseline CVD History, No. (%) | 0.3681 | ||
| No | 1970 (87.2%) | 2010 (86.3%) | |
| Yes | 290 (12.8%) | 320 (13.7%) | |
| Baseline Hypertension, No. (%) | 0.2727 | ||
| No | 402 (17.8%) | 386 (16.6%) | |
| Yes | 1858 (82.2%) | 1944 (83.4%) | |
| Baseline Insulin Use, No. (%) | 0.6088 | ||
| No | 1844 (84.5%) | 1919 (85.1%) | |
| Yes | 337 (15.5%) | 336 (14.9%) | |
| Body Mass Index, mean ± SD, kg/m2 | 35.8 ± 5.7 | 35.6 ± 5.9 | 0.3459 |
| LDL Cholesterol, mean ± SD, mg/dL | 112.1 ± 32.1 | 111.9 ± 32.3 | 0.8974 |
| Weight, mean ± SD, kg | 100.2 ± 18.6 | 99.8 ± 19.4 | 0.5245 |
| Waist Circumference, mean ± SD, cm | 113.5 ± 13.5 | 113.3 ± 14.3 | 0.5877 |
As shown in Table 2, the group defined by a decrease in weight and waist circumference from baseline to Year 1 emerged as the most common category with 2,840 individuals. Individuals with increased weight and waist circumference was the second most common category with 782 individuals followed by increased waist but decreased weight (n=568) and lastly the group with a decrease in waist circumference despite increased weight was least common (n=400). Significant differences in demographic variables emerged across the four change categories. Participants with a reduced weight and waist circumference were almost one year older (p=0.0046) and more likely to be male (p=0.0002), non-minority (p=0.0006), in Intensive Lifestyle Intervention (p<.0001), and have greater magnitude of weight loss (p<.0001) and waist circumference decrease (p<.0001) than participants in other change categories.
Table 2.
Baseline demographic characteristics and year 1 weight and waist change unadjusted means ± SE and N(%)for waist and weight category
| Baseline Demographic or Year 1 Change Measure |
Decreased Weight/Waist mean ± SE or N(%) |
Weight Increase/ Waist Decrease mean ± SE or N(%) |
Weight Decrease/ Waist Increase mean ± SE or N(%) |
Increased Weight/Waist mean ± SE or N(%) |
P- value |
|---|---|---|---|---|---|
| N | 2840 | 400 | 568 | 782 | |
| Age, years | 59.16±0.13 | 58.25±0.34 | 58.49±0.29 | 58.47±0.24 | 0.0046 |
| Sex | 0.0002 | ||||
| Female | 1628(57.3%) | 256(64.0%) | 376(66.2%) | 458(58.6%) | |
| Male | 1212(42.7%) | 144(36.0%) | 192(33.8%) | 324(41.4%) | |
| Race | 0.0006 | ||||
| African American | 425(15.0%) | 83(20.8%) | 95(16.7%) | 112(14.3%) | |
| White | 1846(65.0%) | 219(54.8%) | 325(57.2%) | 500(63.9%) | |
| Hispanic | 354(12.5%) | 57(14.3%) | 90(15.9%) | 100(12.8%) | |
| Other/Mixed | 214(7.5%) | 41(10.3%) | 58(10.2%) | 70(9.0%) | |
| Baseline CVD History | 0.5026 | ||||
| No | 2470(87.0%) | 349(87.3%) | 496(87.3%) | 665(85.0%) | |
| Yes | 370(13.0%) | 51(12.7%) | 72(12.7%) | 117(15.0%) | |
| Baseline Insulin Use | 0.0007 | ||||
| No | 2369(86.2%) | 304(78.6%) | 458(83.1%) | 632(84.4%) | |
| Yes | 380(13.8%) | 83(21.5%) | 93(16.9%) | 117(15.6%) | |
| Baseline Smoking Status | 0.7045 | ||||
| Never | 1421(50.2%) | 214(53.5%) | 297(52.3%) | 396(50.8%) | |
| Past | 1284(45.3%) | 170(42.5%) | 244(43.0%) | 356(45.6%) | |
| Present | 128(4.5%) | 16(4.0%) | 27(4.8%) | 28(3.6%) | |
| Randomization Assignment | <.0001 | ||||
| Diabetes Support and Education | 896(31.6%) | 333(83.3%) | 339(59.7%) | 692(88.5%) | |
| Intensive Lifestyle Intervention | 1944(68.5%) | 67(16.8%) | 229(40.3%) | 90(11.5%) | |
| Baseline BMI, kg/m2 | 35.58±0.11 | 35.52±0.29 | 35.83±0.24 | 36.04±0.21 | 0.1803 |
| Baseline Diabetes Duration, years | 6.87±0.12 | 6.95±0.33 | 6.77±0.28 | 6.52±0.24 | 0.5851 |
| Baseline Systolic Blood Pressure, mmHg | 128.57±0.32 | 128.75±0.85 | 130.08±0.71 | 128.04±0.61 | 0.1639 |
| Baseline Diastolic Blood Pressure, mmHg | 70.10±0.18 | 70.49±0.47 | 70.54±0.40 | 69.56±0.34 | 0.2187 |
| Baseline LDL Cholesterol, mg/dL | 111.57±0.61 | 111.84±1.61 | 113.05±1.35 | 112.88±1.15 | 0.6329 |
| Year 1 weight change (kg) | −8.34±0.11 | 1.91±0.28 | −3.12±0.24 | 3.37±0.20 | <.0001 |
| Year 1 weight change (kg) by Randomization Assignment | <.0001 | ||||
| Diabetes Support and Education | −4.23±0.17 | 1.96±0.29 | −2.10±0.28 | 3.42±0.20 | |
| Intensive Lifestyle Intervention | −10.24±0.12 | 1.74±0.64 | −4.64±0.34 | 2.88±0.54 | |
| Year 1 waist change (cm) | −8.25±0.12 | −3.87±0.32 | 3.34±0.27 | 4.07±0.23 | <.0001 |
| Year 1 waist change (cm) by Randomization Assignment | <.0001 | ||||
| Diabetes Support and Education | −5.20±0.21 | −4.67±0.77 | 3.29±0.34 | 4.10±0.24 | |
| Intensive Lifestyle Intervention | −9.66±0.14 | −3.71±0.34 | 3.42±0.41 | 3.74±0.66 |
For the analyses of the primary outcomes, 84 individuals were excluded for having a primary or secondary event during the first year (Primary: 63 participants (ILI=26, DSE=37; p=0.1373); Secondary: 84 participants which includes the 63 excluded for primary (ILI=33, DSE=51; p=0.0369)). During the median follow-up of 9.9 years (maximum = 11.4 years), 694 participants reported having a primary Outcome event while 1029 reported the occurrence of a Secondary Outcome event. Initial analyses examined the test of interaction between randomization group and weight and waist change category. There was no heterogeneity of effect of weight and waist change group for the primary CVD outcome, however, for secondary outcome, there was significant evidence for differences in effect due to randomization group (second level of covariate adjustment: weight and waist change group*randomization group p=0.0153; full adjustment p=0.498) leading to two approaches for examination of these differences. For the first approach, analyses were stratified by treatment condition (Tables 3&4), using the group that observed decreases on both measures as the reference. In analyses comparing the four weight/waist change categories within only DSE, there were no differences in occurrence of either primary or secondary outcomes regardless of level of adjustment (see Table 5). However, among ILI groups, significant differences in cardiovascular events emerged in models adjusted for sex, age, race, and clinical site (Table 4; primary outcome p=0.0096, Secondary outcome p=0.0002; see supplemental Figures S1&S2). Compared to individuals who lost weight and reduced waist circumference (REF group), both groups that had increase in waist (regardless of whether they lost weight or not) had increased risk of CVD outcomes. Specifically, relative to the group that decreased both weight and waist, the group that increased weight and waist circumference had significantly greater hazard of both primary outcome (HR[95%CI]: 1.76[1.07, 2.89]) and Secondary outcome (2.17[1.47, 3.20]). Likewise, the group that lost weight but increased waist circumference had approximately 50% increased risk of CVD events (primary outcome (1.55 [1.11, 2.17]) and Secondary outcome (1.42 [1.07, 1.89]). Results were similar yet attenuated for the model with full adjustment for CVD related confounders (primary outcome p=0.0609, Secondary outcome p=0.0056) as can be seen in table 4.
Table 3.
Cox Proportional Hazards Regression Estimates of differences in Primary and Secondary Outcome Hazards for year 1 weight and waist change categories for varying levels of covariate adjustment for Diabetes Support and Education.
| Level of Adjustment | Decreased Weight/Waist |
Weight Increase/ Waist Decrease HR(95%CI) |
Weight Decrease/ Waist Increase HR(95%CI) |
Increased Weight/Waist HR(95%CI) |
Overall Type 3 P-value |
|---|---|---|---|---|---|
| Primary CVD Outcome* | |||||
| Model 1 | REF | 0.92 (0.66, 1.29) | 1.15 (0.84, 1.58) | 0.99 (0.76, 1.28) | 0.6924 |
| Model 2 | REF | 0.99 (0.71, 1.40) | 1.18 (0.86, 1.62) | 1.02 (0.79, 1.32) | 0.7543 |
| Model 3 | REF | 0.98 (0.70, 1.38) | 1.18 (0.86, 1.63) | 0.92 (0.71, 1.20) | 0.5555 |
| Secondary CVD Outcome** | |||||
| Model 1 | REF | 0.90 (0.68, 1.18) | 1.08 (0.83, 1.40) | 0.98 (0.79, 1.21) | 0.7446 |
| Model 2 | REF | 0.97 (0.74, 1.28) | 1.10 (0.85, 1.43) | 1.03 (0.83, 1.27) | 0.8632 |
| Model 3 | REF | 1.01 (0.76, 1.34) | 1.13 (0.87, 1.48) | 0.98 (0.78, 1.22) | 0.7700 |
Model 1 is adjusted for clinical site only
Model 2 is adjusted for sex, age, race, and clinical site
Model 3 is adjusted for sex, age, race, clinical site, cardiovascular disease history, insulin use, diabetes duration, weight, systolic blood pressure, diastolic blood pressure, LDL cholesterol, and smoking status
Table 4.
Cox Proportional Hazards Regression Estimates of differences in Primary and Secondary Outcome Hazards for year 1 weight and waist change categories for varying levels of covariate adjustment for Intensive Lifestyle Intervention.
| Level of Adjustment | Decreased Weight/Waist |
Weight Increase/ Waist Decrease HR(95%CI) |
Weight Decrease/ Waist Increase HR(95%CI) |
Increased Weight/Waist HR(95%CI) |
Overall Type 3 P-value |
|---|---|---|---|---|---|
| Primary CVD Outcome* | |||||
| Model 1 | REF | 0.64 (0.28, 1.44) | 1.31 (0.94, 1.82) | 1.41 (0.86, 2.30) | 0.1386 |
| Model 2 | REF | 0.75 (0.33, 1.70) | 1.55 (1.11, 2.17)* | 1.76 (1.07, 2.89)* | 0.0096 |
| Model 3 | REF | 0.58 (0.24, 1.41) | 1.48 (1.05, 2.09)* | 1.32 (0.77, 2.25) | 0.0609 |
| Secondary CVD Outcome** | |||||
| Model 1 | REF | 0.95 (0.54, 1.66) | 1.20 (0.90, 1.59) | 1.70 (1.16, 2.50)* | 0.0370 |
| Model 2 | REF | 1.13 (0.64, 1.97) | 1.42 (1.07, 1.89)* | 2.17 (1.47, 3.20)* | 0.0002 |
| Model 3 | REF | 0.92 (0.50, 1.69) | 1.43 (1.07, 1.92)* | 1.80 (1.19, 2.74)* | 0.0056 |
Note: indicates HR is statistically significantly different from reference group; Model 1 is adjusted for clinical site only; Model 2 is adjusted for sex, age, race, and clinical site; Model 3 is adjusted for sex, age, race, clinical site, cardiovascular disease history, insulin use, diabetes duration, weight, systolic blood pressure, diastolic blood pressure, LDL cholesterol, and smoking status
Table 5.
Baseline demographic characteristics and year 1 weight and waist change unadjusted means ± SE and N(%)for waist and weight category
| Baseline Demographic or Year 1 Change Measure |
DSE mean ± SE or N(%) |
ILI Decreased Weight/Waist mean ± SE or N(%) |
ILI Weight Increase/ Waist Decrease mean ± SE or N(%) |
ILI Weight Decrease/ Waist Increase mean ± SE or N(%) |
ILI increased Weight/Waist mean ± SE or N(%) |
P-value |
|---|---|---|---|---|---|---|
| N | 2260 | 1944 | 67 | 229 | 90 | |
| Age, years | 59.05±0.14 | 59.04±0.15 | 56.64±0.83 | 57.34±0.45 | 56.72±0.14 | <.0001 |
| Sex | <.0001 | |||||
| Female | 1344(59.5%) | 1100(56.6%) | 45(67.2%) | 163(71.2%) | 66(73.3%) | |
| Male | 916(40.5%) | 844(43.4%) | 22(32.8%) | 66(28.8%) | 24(26.7%) | |
| Race | <.0001 | |||||
| African American | 350(15.5%) | 292(15.0%) | 15(22.4%) | 48(20.9%) | 10(11.1%) | |
| White | 1429(63.2%) | 1260(64.9%) | 27(40.3%) | 121(52.8%) | 53(58.9%) | |
| Hispanic | 292(12.9%) | 251(12.9%) | 13(19.4%) | 35(15.3%) | 10(11.1%) | |
| Other/Mixed | 189(8.4%) | 140(7.2%) | 12(17.9%) | 25(10.9%) | 17(18.9%) | |
| Baseline CVD History | 0.8661 | |||||
| No | 1970(87.2%) | 1678(86.3%) | 59(88.1%) | 197(86.0%) | 76(84.4%) | |
| Yes | 290(12.8%) | 266(13.7%) | 8(11.9%) | 32(14.0%) | 14(15.6%) | |
| Baseline BMI, kg/m2 | 35.77±0.12 | 35.59±0.13 | 35.46±0.71 | 35.94±0.38 | 35.32±0.61 | 0.7475 |
| Year 1 weight change (kg) | −0.65±0.12 | −10.23±0.13 | 1.74±0.70 | −4.64±0.38 | 2.88±0.60 | <.0001 |
| Year 1 waist change (cm) | −0.86±0.15 | −9.66±0.16 | −4.67±0.85 | 3.42±0.46 | 3.73±0.73 | <.0001 |
In the second approach to analyses where the DSE group was the reference (see Table 5 for demographics across groups), only the model adjusted for gender sex, age, race and clinical site (model 2) was significant for the primary cardiovascular outcomes (See Table 6 for results). Specifically, only individuals who lost weight and increased waist circumference had an increased risk of primary cardiovascular events compared to DSE (see supplemental Figure S3). No other groups emerged as significantly different in risk. For the secondary cardiovascular outcome, individuals who had increased waist circumference (regardless of whether they had increased or decreased weight) had significantly greater risk of a secondary cardiovascular event in models adjusted for sex, age, race, and clinical site, as well as models adjusted for demographics and CVD confounders (model 3; see supplemental Figure S4). Neither of the groups with decreased waist circumference (regardless of weight gain or weight loss) differed from individuals in the DSE.
Table 6.
Cox Proportional Hazards Regression Estimates of differences in Primary and Secondary Outcome Hazards for year 1 Intensive Lifestyle Intervention (ILI) weight and waist change categories compared to Diabetes Support and Education (DSE) for varying levels of covariate adjustment.
| Level of Adjustment |
DSE | ILI: Decreased Weight/Waist HR(95%CI) |
ILI: Weight Increase/ Waist Decrease HR(95%CI) |
ILI: Weight Decrease/ Waist Increase HR(95%CI) |
ILI: Increased Weight/Waist HR(95%CI) |
Overall Type 3 P-value |
|---|---|---|---|---|---|---|
| Primary CVD Outcome | ||||||
| Model 1 | REF | 0.98 (0.83, 1.14) | 0.60 (0.27, 1.34) | 1.25 (0.90, 1.73) | 1.34 (0.82, 2.18) | 0.2657 |
| Model 2 | REF | 0.96 (0.82, 1.12) | 0.71 (0.32, 1.60) | 1.46 (1.05, 2.02)* | 1.63 (0.99, 2.66) | 0.0292 |
| Model 3 | REF | 0.97 (0.83, 1.14) | 0.55 (0.23, 1.34) | 1.42 (1.01, 1.99) | 1.25 (0.74, 2.11) | 0.1191 |
| Secondary CVD Outcome | ||||||
| Model 1 | REF | 0.96 (0.84, 1.09) | 0.86 (0.49, 1.49) | 1.13 (0.85, 1.49) | 1.56 (1.07, 2.29) | 0.1195 |
| Model 2 | REF | 0.94 (0.83, 1.07) | 1.02 (0.59, 1.78) | 1.33 (1.01, 1.76)* | 1.95 (1.33, 2.86)* | 0.0012 |
| Model 3 | REF | 0.94 (0.82, 1.08) | 0.81 (0.44, 1.48) | 1.35 (1.02, 1.80)* | 1.65 (1.10, 2.47)* | 0.0135 |
Note: indicates HR is statistically significantly different from reference group; Model 1 is adjusted for clinical site only; Model 2 is adjusted for sex, age, race, and clinical site; Model 3 is adjusted for sex, age, race, clinical site, cardiovascular disease history, insulin use, diabetes duration, weight, systolic blood pressure, diastolic blood pressure, LDL cholesterol, and smoking status
Discussion
This post-hoc analysis of the Look AHEAD trial documents important long-term clinical outcomes related to increased waist circumference independent of weight change during one year of behavioral weight loss treatment. Across two different analytic strategies, and in models of varying degrees of adjustment, we see that individuals who had an increase in waist circumference (regardless of the direction of weight change) had increased risk of cardiovascular events. Equally interesting is the finding that individuals who reduced waist circumference despite weight gain did not differ in risk compared to those who lost weight and reduced their waist circumference. This finding is consistent with research on the ‘obesity paradox’, which suggests that the relationship between adiposity and cardiovascular outcomes is not straightforward (21). Findings are most consistent across the secondary cardiovascular outcomes. This may be because primary and secondary outcomes are not mutually exclusive and secondary outcomes also include events that are not exclusively cardiovascular (i.e., all-cause mortality), possibly resulting in greater power to detect effects. Although these findings are from a secondary analysis, they do present a consistent pattern of risk related to increased waist circumference independent of weight change.
This is the first report, to our knowledge, that weight changes and waist circumference changes do not always move in the same direction during weight loss treatment. This decoupling of body weight and waist circumference has been documented previously but in longitudinal epidemiological research. Indeed, secular trends indicate that increases in waist circumference and weight at the population level are not occurring at the same rate, with waist increases occurring independent of weight change (22). Although these two measures are usually highly correlated in weight loss treatment and the majority of participants in the current study demonstrated improvements in both, over 10% of participants in Intensive Lifestyle Intervention experienced discordant changes in waist circumference and body weight during treatment. The frequency of discordant changes in weight and waist was even higher within the Diabetes Support and Education (DSE) group where over 25% of individuals had changes in opposite directions. Yet among DSE, there was no difference in primary or secondary cardiovascular outcomes across the four weight and waist change groups. This may be because magnitude of changes in waist and weight were attenuated among DSE participants or although less likely, because more DSE participants were excluded from analyses due to occurrence of primary or secondary events during the first year of the trial (e.g., reduced time to first occurrence).
These findings suggest that monitoring waist circumference change during behavioral weight loss treatment may be important for tracking progress. This is consistent with recommendations from a National Heart, Lung, and Blood Institute (NHLBI) expert panel indicating that waist circumference should be measured and used to assess progress in weight management interventions in addition to BMI and body weight (2324). Even more so, the findings suggest that waist circumference change in relation to weight loss may be especially important as opposed to simply reporting on the two metrics independently. Indeed incorporating waist circumference (or other measures of body composition) hold promise for clarifying the heterogeneous effects of weight loss treatment on important health outcomes.
The conclusions that can be drawn from the current findings are limited as this is a secondary, post-hoc analysis coupled with the inherent difficulties disentangling the influence of weight and waist changes during treatment as they are highly correlated. It is unclear if reduction in waist circumference and weight loss represent two separate processes through which reduced cardiovascular risk can be achieved or if reduced waist circumference is a mechanism linking weight loss to cardiovascular benefits. Because individuals who decreased their waist circumference despite weight gain did not differ in cardiovascular outcomes from those who improved on both and because those who increased waist circumference despite weight loss had increased risk of cardiovascular outcomes compared to the reference group, it is tempting to speculate that reduction of central adiposity (i.e., as assessed by waist circumference) may be the primary driver of the cardiovascular benefits of weight loss. It may be the case that the changes in weight and waist circumference are not causal influences on cardiovascular outcomes but instead reflect a variable response to weight loss treatment that is a symptom or marker of differences in risk. Statistically controlling for factors that might make these four groups different such as medication use and cardiovascular history helps to address this concern but there may be other unmeasured and confounding factors that are not accounted for in this analysis.
A future study may aim to compare two different approaches to lifestyle intervention where traditional behavioral weight loss is implemented to facilitate clinically significant weight loss compared to an isocaloric lifestyle intervention designed to reduce waist circumference without changes in body weight. Observational data suggest that this could be accomplished by intervening on quality of dietary intake and level of physical activity or physical fitness, as both are associated with variability in the distribution of body adiposity (25-27). Improvements in either of these areas may result in changes in body composition (particularly reductions in central adiposity) that may contribute to reduced cardiometabolic risk even without changes in weight or BMI (28-29). Although it will continue to be difficult to disentangle the relative influence of weight and waist circumference changes, such a study design would provide more experimental control and therefore facilitate more confident causal conclusions about the influence of waist circumference on cardiovascular risk.
There are a number of limitations of the current study. The groups where weight and waist change occurred in different directions reflect a small proportion of individuals in weight loss treatment. However, the large sample size of the Look AHEAD trial enabled the study of these subsets. For the current approach, we relied on weight and waist changes as measured at baseline and year 1. This assessment of change does not account for variability in either of these metrics throughout the year. Our analytic approach also does not account for any changes in these metrics that occur beyond the first year. Further, the Look AHEAD trial was not designed to address this specific research question; therefore, the results should be interpreted conservatively until further verification is available. Indeed, the methodological decision to generate four groups defined by waist and weight changes during the first year of the Look AHEAD trial was driven by statistical limitations from multicollinearity. Other researchers studying the relationship of weight change and waist circumference change have also been challenged by this issue (30). The weight and waist change categories also disregard magnitude of changes in either weight or waist circumference and individuals with values near zero may represent measurement error rather than true changes in waist circumference or weight change. We believe the contributions of the current analytic approach outweigh the limitations but highlight them with the hope that future studies will be able to address these issues prospectively. Additionally, the use of anthropometry (i.e., waist circumference) to measure central adiposity is cost effective but less accurate than advanced methodology such as DXA (DEXA) or imaging. Alternative approaches would allow for more accurate and direct measurement of central adiposity and quantifying visceral versus subcutaneous central body fat. Lastly, the Look AHEAD trial included older adults with Type 2 Diabetes and therefore it is unclear if the current findings would generalize to persons without diabetes or a younger population.
The current study highlights a number of important implications of waist circumference change in the context of weight loss treatment. The relationship between weight change and waist circumference change during one year of treatment had long lasting implications for cardiovascular morbidity and mortality among individuals with Type 2 Diabetes. In the ongoing pursuit to elucidate variability in weight loss outcomes and in the benefits of weight loss treatment, these findings offer more nuance to how anthropometric changes are related to long-term cardiovascular health. Although it is widely accepted that central adiposity is especially important for cardiovascular risk, central adiposity remains underutilized as a clinical tool for monitoring progress in behavioral weight management and instead percent weight loss remains the predominant target and metric to monitor participant progress in weight management. By integrating these two measures we may be better able to understand how weight loss treatment confers (or fails to confer) long-term cardiovascular health benefits, to identify which participants are likely accrue these benefits, and to guide changes to treatment that increase clinical benefits.
Supplementary Material
What is already known about this subject?
Both Body Mass Index and waist circumference are used to define obesity and assess adiposity-related risk
Epidemiological work indicates that waist circumference may be especially important for understanding adiposity-related morbidity and mortality.
Behavioral weight loss trials often report waist circumference data but do not utilize it in evaluating the effects of weight loss treatment on important health outcomes.
What are the new findings in your manuscript?
Increased waist circumference, regardless of weight loss or gain in behavioral treatment, was associated with subsequent cardiovascular outcomes.
How might your results change the direction of research or the focus of clinical practice? Please remember to also include between the title page and structured abstract in your paper.
Variability in waist circumference change during weight loss can provide additional information about the impact of weight loss treatment on important health outcomes.
The findings are consistent with emerging calls to incorporate measures of central adiposity into obesity-related clinical practice and research.
Acknowledgements
Look AHEAD data may be requested from the National Institute of Diabetes and Digestive and Kidney Diseases Central Repository (https://Repository.niddk.nih.gov/studies/look-ahead). Protocols, forms, summary statistics, and data dictionary are also available.
Funding Support: The first author (K.O.) is funded on a NIH training grant (T32 HL076134).
Funding and Support
Funded by the National Institutes of Health through cooperative agreements with the National Institute of Diabetes and Digestive and Kidney Diseases: DK57136, DK57149, DK56990, DK57177, DK57171, DK57151, DK57182, DK57131, DK57002, DK57078, DK57154, DK57178, DK57219, DK57008, DK57135, and DK56992. Additional funding was provided by the National Heart, Lung, and Blood Institute; National Institute of Nursing Research; National Center on Minority Health and Health Disparities; NIH Office of Research on Women’s Health; and the Centers for Disease Control and Prevention. This research was supported in part by the Intramural Research Program of the National Institute of Diabetes and Digestive and Kidney Diseases. The Indian Health Service (I.H.S.) provided personnel, medical oversight, and use of facilities. The opinions expressed in this paper are those of the authors and do not necessarily reflect the views of the I.H.S. or other funding sources.
Additional support was received from The Johns Hopkins Medical Institutions Bayview General Clinical Research Center (M01RR02719); the Massachusetts General Hospital Mallinckrodt General Clinical Research Center and the Massachusetts Institute of Technology General Clinical Research Center (M01RR01066); the Harvard Clinical and Translational Science Center (RR025758-04); the University of Colorado Health Sciences Center General Clinical Research Center (M01RR00051) and Clinical Nutrition Research Unit (P30 DK48520); the University of Tennessee at Memphis General Clinical Research Center (M01RR0021140); the University of Pittsburgh General Clinical Research Center (GCRC) (M01RR000056), the Clinical Translational Research Center (CTRC) funded by the Clinical & Translational Science Award (UL1 RR 024153) and NIH grant (DK 046204); the VA Puget Sound Health Care System Medical Research Service, Department of Veterans Affairs; and the Frederic C. Bartter General Clinical Research Center (M01RR01346).
The following organizations have committed to make major contributions to Look AHEAD: FedEx Corporation; Health Management Resources; LifeScan, Inc., a Johnson & Johnson Company; OPTIFAST® of Nestle HealthCare Nutrition, Inc.; Hoffmann-La Roche Inc.; Abbott Nutrition; and Slim-Fast Brand of Unilever North America.
Some of the information contained herein was derived from data provided by the Bureau of Vital Statistics, New York City Department of Health and Mental Hygiene.
Footnotes
Principal Investigator
Program Coordinator
Co-Investigator
All other Look AHEAD staffs are listed alphabetically by site.
Disclosure: The authors declare no conflicts of interest.
Clinical Trial Registration: Clinicaltrials.gov identifier NCT00017953
References
- 1.Hales CM, Fryar CD, Carroll MD, Freedman DS, Ogden CL. Trends in obesity and severe obesity prevalence in US youth and adults by sex and age, 2007-2008 to 2015-2016. Jama. 2018. April 24;319(16):1723–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Abdelaal M, le Roux CW, Docherty NG. Morbidity and mortality associated with obesity. Annals of translational medicine. 2017. April;5(7). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Look AHEAD Research Group. Cardiovascular effects of intensive lifestyle intervention in type 2 diabetes. New England journal of medicine. 2013. July 11 ;369(2): 145–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Wing RR, Lang W, Wadden TA, Safford M, Knowler WC, Bertoni AG, Hill JO, Brancati FL, Peters A, Wagenknecht L, Look AHEAD Research Group. Benefits of modest weight loss in improving cardiovascular risk factors in overweight and obese individuals with type 2 diabetes. Diabetes care. 2011. July 1;34(7): 1481–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Zomer E, Gurusamy K, Leach R, Trimmer C, Lobstein T, Morris S, James WP, Finer N. Interventions that cause weight loss and the impact on cardiovascular risk factors: a systematic review and meta-analysis. Obesity reviews. 2016. October; 17(10): 1001–11. [DOI] [PubMed] [Google Scholar]
- 6.Haffner SM, Lehto S, Rönnemaa T, Pyörälä K, Laakso M. Mortality from coronary heart disease in subjects with type 2 diabetes and in nondiabetic subjects with and without prior myocardial infarction. New England journal of medicine. 1998. July 23;339(4):229–34. [DOI] [PubMed] [Google Scholar]
- 7.Rawshani A, Rawshani A, Franzén S, Sattar N, Eliasson B, Svensson AM, Zethelius B, Miftaraj M, McGuire DK, Rosengren A, Gudbjornsdottir S. Risk factors, mortality, and cardiovascular outcomes in patients with type 2 diabetes. New England Journal of Medicine. 2018. August 16;379(7):633–44. [DOI] [PubMed] [Google Scholar]
- 8.Shah AD, Langenberg C, Rapsomaniki E, Denaxas S, Pujades-Rodriguez M, Gale CP, Deanfield J, Smeeth L, Timmis A, Hemingway H. Type 2 diabetes and incidence of cardiovascular diseases: a cohort study in 1· 9 million people. The lancet Diabetes & endocrinology. 2015. February 1;3(2):105–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Gregg EW, Look AHEAD Study Group. Association of the Magnitude of Weight Loss and Physical Fitness Change on Long-term CVD outcomes: The Look AHEAD Study. The lancet. Diabetes & endocrinology 2016. November;4(11):913. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Balkau B, Picard P, Vol S, Fezeu L, Eschwège E. Consequences of change in waist circumference on cardiometabolic risk factors over 9 years: Data from an Epidemiological Study on the Insulin Resistance Syndrome (DESIR). Diabetes Care. 2007. July 1;30(7):1901–3. [DOI] [PubMed] [Google Scholar]
- 11.Yusuf S, Hawken S, Ounpuu S, Bautista L, Franzosi MG, Commerford P, Lang CC, Rumboldt Z, Onen CL, Lisheng L, Tanomsup S. Obesity and the risk of myocardial infarction in 27 000 participants from 52 countries: a case-control study. The Lancet. 2005. November 5;366(9497):1640–9. [DOI] [PubMed] [Google Scholar]
- 12.Janssen I, Katzmarzyk PT, Ross R. Waist circumference and not body mass index explains obesity-related health risk. The American journal of clinical nutrition. 2004. March 1;79(3):379–84. [DOI] [PubMed] [Google Scholar]
- 13.Nazare JA, Smith J, Borel AL, Aschner P, Barter P, Van Gaal L, Tan CE, Wittchen HU, Matsuzawa Y, Kadowaki T, Ross R. Usefulness of measuring both body mass index and waist circumference for the estimation of visceral adiposity and related cardiometabolic risk profile (from the INSPIRE ME IAA study). The American journal of cardiology. 2015. February 1;115(3):307–15. [DOI] [PubMed] [Google Scholar]
- 14.Menke A, Muntner P, Wildman RP, Reynolds K, He J. Measures of adiposity and cardiovascular disease risk factors. Obesity. 2007. March;15(3):785–95. [DOI] [PubMed] [Google Scholar]
- 15.Rothberg AE, McEwen LN, Kraftson AT, Ajluni N, Fowler CE, Nay CK, Miller NM, Burant CF, Herman WH. Impact of weight loss on waist circumference and the components of the metabolic syndrome. BMJ Open Diabetes Research and Care. 2017. February 1;5(1):e000341. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Wing RR, Jeffery RW, Burton LR, Thorson C, Kuller LH, Folsom AR. Change in waist-hip ratio with weight loss and its association with change in cardiovascular risk factors. The American journal of clinical nutrition. 1992. June 1;55(6):1086–92. [DOI] [PubMed] [Google Scholar]
- 17.Look AHEAD Research Group. The Look AHEAD study: a description of the lifestyle intervention and the evidence supporting it. Obesity. 2006. May;14(5):737–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Espeland M Reduction in weight and cardiovascular disease risk factors in individuals with type 2 diabetes: one-year results of the look AHEAD trial. Diabetes care. 2007. March 15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Look AHEAD Research Group. Long term effects of a lifestyle intervention on weight and cardiovascular risk factors in individuals with type 2 diabetes: four year results of the Look AHEAD trial. Archives of internal medicine. 2010. September 27;170(17):1566. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge. Clinical chemistry. 1972. June 1;18(6):499–502. [PubMed] [Google Scholar]
- 21.McAuley PA, Artero EG, Sui X, Lee DC, Church TS, Lavie CJ, … & Blair SN (2012, May). The obesity paradox, cardiorespiratory fitness, and coronary heart disease In Mayo Clinic Proceedings (Vol. 87, No. 5, pp. 443–451). Elsevier. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Gearon E, Tanamas SK, Stevenson C, Loh VH, Peeters A. Changes in waist circumference independent of weight: Implications for population level monitoring of obesity. Preventive medicine. 2018. June 1;111:378–83. [DOI] [PubMed] [Google Scholar]
- 23.Klein S, Allison DB, Heymsfield SB, Kelley DE, Leibel RL, Nonas C, Kahn R. Waist circumference and cardiometabolic risk: a consensus statement from shaping America's health: Association for Weight Management and Obesity Prevention; NAASO, the Obesity Society; the American Society for Nutrition; and the American Diabetes Association. Obesity. 2007. May;15(5):1061–7. [DOI] [PubMed] [Google Scholar]
- 24.Ross R, Neeland IJ, Yamashita S, Shai I, Seidell J, Magni P, Santos RD, Arsenault B, Cuevas A, Hu FB, Griffin BA. Waist circumference as a vital sign in clinical practice: a Consensus Statement from the IAS and ICCR Working Group on Visceral Obesity. Nature Reviews Endocrinology. 2020. February 4:1–3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Dekker MJ, Lee S, Hudson R, Kilpatrick K, Graham TE, Ross R, Robinson LE. An exercise intervention without weight loss decreases circulating interleukin-6 in lean and obese men with and without type 2 diabetes mellitus. Metabolism. 2007. March 1;56(3):332–8. [DOI] [PubMed] [Google Scholar]
- 26.Estruch R, Martínez-González MA, Corella D, Salas-Salvadó J, Fitó M, Chiva-Blanch G, Fiol M, Gómez-Gracia E, Arós F, Lapetra J, Serra-Majem L. Effect of a high-fat Mediterranean diet on bodyweight and waist circumference: a prespecified secondary outcomes analysis of the PREDIMED randomised controlled trial. The lancet Diabetes & endocrinology. 2016. August 1;4(8):666–76. [DOI] [PubMed] [Google Scholar]
- 27.Ross R, Janssen I, Dawson J, Kungl AM, Kuk JL, Wong SL, Nguyen-Duy TB, Lee S, Kilpatrick K, Hudson R. Exercise-induced reduction in obesity and insulin resistance in women: a randomized controlled trial. Obesity research. 2004. May;12(5):789–98. [DOI] [PubMed] [Google Scholar]
- 28.Boule NG, Haddad E, Kenny GP, Wells GA, Sigal RJ. Effects of exercise on glycemic control and body mass in type 2 diabetes mellitus: a meta-analysis of controlled clinical trials. Scandinavian Journal of Medicine & Science in Sports. 2002. February;12(1):60–1. [DOI] [PubMed] [Google Scholar]
- 29.Ross R, Bradshaw AJ. The future of obesity reduction: beyond weight loss. Nature Reviews Endocrinology. 2009. June;5(6):319. [DOI] [PubMed] [Google Scholar]
- 30.Vatcheva KP, Lee M, McCormick JB, Rahbar MH. Multicollinearity in regression analyses conducted in epidemiologic studies. Epidemiology (Sunnyvale, Calif.). 2016. April;6(2). [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.