Highlights
- •
Sustained weight loss significantly lowers incidence of cardiometabolic outcomes.
- •
Sustained weight loss significantly delays onset of cardiometabolic outcomes.
- •
Greater weight loss significantly delays onset of cardiometabolic outcomes.
Obesity increases the risk of developing type 2 diabetes, hypertension, and hyperlipidemia. We sought to determine the impact of obesity maintenance, weight regain, weight loss maintenance, and magnitudes of weight loss on future risk and time to developing these cardiometabolic conditions. This was a retrospective cohort study of adults receiving primary care at Geisinger Health System between 2001 and 2017. Using electronic health records, patients with ≥3-weight measurements over a 2-year index period were identified and categorized. Obesity maintainers (OM) had obesity (body mass index ≥30 kg/m²) and maintained their weight within ±3% from baseline (reference group). Both weight loss rebounders (WLR) and weight loss maintainers (WLM) had obesity at baseline and lost >5% body weight in year 1; WLR regained ≥20% of weight loss by end of year 2 and WLM maintained ≥80% of weight loss. Incident type 2 diabetes, hypertension, and hyperlipidemia, and time-to-outcome were determined for each study group and by weight loss category for WLM. Of the 63,567 patients included, 67% were OM, 19% were WLR, and 14% were WLM. The mean duration of follow-up was 6.6 years (SD, 3.9). Time until the development of electronic health record-documented type 2 diabetes, hypertension, and hyperlipidemia was longest for WLM and shortest for OM (log-rank test p <0.0001). WLM had the lowest incident type 2 diabetes (adjusted hazard ratio [HR] 0.676 [95% confidence interval [CI] 0.617 to 0.740]; p <0.0001), hypertension (adjusted HR 0.723 [95% CI 0.655 to 0.799]; p <0.0001), and hyperlipidemia (adjusted HR 0.864 [95% CI 0.803 to 0.929]; p <0.0001). WLM with the greatest weight loss (>15%) had a longer time to develop any of the outcomes compared with those with the least amount of weight loss (<7%) (p <0.0001). In an integrated delivery network population, sustained weight loss was associated with a delayed onset of cardiometabolic diseases, particularly with a greater magnitude of weight loss.
The prevalence of obesity has risen dramatically in the United States; per the 2017 to 2018 National Health and Nutrition Examination Survey, 42.4% of US adults have obesity. Managing obesity is a lifelong endeavor as there are many biological, social, psychological, and environmental factors contributing to weight gain and loss. , Modest weight loss of at least 5% is clinically beneficial and recommended by clinical treatment guidelines, which can be achieved with various clinical and behavioral treatment options. However, long-term weight loss maintenance remains challenging owing to the biology of obesity, hence weight regain is common; , about 80% of the weight loss is regained within 5 years. Clinical outcomes of lifestyle, behavioral, and clinical treatment interventions have been examined, but limitations include the relatively short duration of follow-up observations; additionally, there is scant literature describing clinical outcomes of sustained weight loss in real-world settings. This study aims to evaluate the long-term impact of obesity, weight loss with regain, and weight loss maintenance, with the latter explored across varying weight loss thresholds. This research seeks to understand the relation between long-term weight maintenance and clinical relevance to 3 cardiometabolic outcomes: type 2 diabetes, hypertension, and hyperlipidemia in a large integrated delivery network setting.
Methods
This is a retrospective observational study of patients receiving primary care between 2001 and 2017 at Geisinger Health System, a Pennsylvania-based integrated delivery network that includes a health plan, acute care hospitals, specialty hospitals, ambulatory surgery centers, and clinical services such as the Center for Nutrition and Weight Management. , Geisinger Health System is 1 of the largest healthcare organizations in the United States and serves over 3 million residents including employees, individuals/families, and adults aged 65 and older. The Geisinger Institutional Review Board reviewed the study and determined that the research does not involve human subjects, thus deeming it exempt from Institutional Board oversight (IRB #: 2019-0138). Data were extracted from the Epic electronic health record (EHR) system in several stages to efficiently capture eligible subjects and define and analyze study outcomes.
The study population was limited to adult patients (age ≥18 years) with 3 or more EHR-documented weight measurements within 2 years, denoted as the index period; these measurements included a baseline weight, a 1-year weight (within 6 to 18 months), and a 2-year weight (within 12 to 24 months). The index period was set at 2 years to allow adequate time to discern clinically relevant weight change patterns. Patients who underwent bariatric surgery before or during the index period and patients with prevalent cancer or a history of cancer during the same time window were excluded from the study; for pregnant women, the weight measurements within 6 months of the pregnancy indicators were also excluded.
The study sample was separated into 3 groups based on weight trends during each year of the index period: (1) obesity maintainers (OM), patients with a history of obesity who maintained weight within ±3% margin from baseline; (2) weight loss rebounders (WLRs), patients with a history of obesity who lost >5% weight via nonsurgical methods (i.e., pharmacotherapy and/or lifestyle intervention) and regained weight from baseline (defined as regaining ≥20% of 1-year weight loss ); and (3) weight loss maintainers (WLMs), patients with obesity at baseline who lost >5% weight via non-surgical methods and maintained weight loss from baseline (defined as maintaining ≥80% of 1-year weight loss). The WLM group was stratified by the amount of initial weight loss. Patients who did not meet the definition of the 3 groups were excluded from the analysis.
Outcomes analyzed include type 2 diabetes mellitus, hypertension, and hyperlipidemia. The status of these conditions was defined by EHR documentation of International Classification of Diseases 10th edition (ICD-10) codes (or at least 2 outpatient visits) or treatment for the condition; for diabetes, the presence of a hemoglobin A1c level of >6.5% with diabetes medication treatment was also included to identify the relatively few patients who had a strong indication of diabetes despite the lack of a diagnosis code (Online Appendix 1 ). Diabetes, hypertension, and hyperlipidemia were classified as prevalent or incident based on the timing of meeting the diagnostic criteria—before or during the index period was considered prevalent, and after the index period was considered incident (Online Appendix 1 ). Additionally, A1c and systolic blood pressure (SBP) were also examined as outcomes related to diabetes and hypertension, respectively.
EHR data on weight measurements, socio-demographics, vital signs, laboratory tests, encounters, procedures, diagnostic codes, orders (pharmacological, nutrition consults, diet, etc.) were extracted. The median height of a patient was calculated and used for all body mass index (BMI) measures (Online Appendix 1 ). Timing of weight measurements was determined to identify periods when patients had 3 EHR-recorded weight measurements over a 2 to 3-year period. All weight measurements during the first 15 months of participation in Geisinger primary care for each patient were excluded to provide a lead-in period to establish medical history. A baseline weight resulting in a BMI ≥30 kg/m² was used to define the beginning of the index period (considered time 0). A second weight measurement occurred approximately 12 months after baseline (within 6 to 18 months) and a third measurement occurred at least 12 months after the second (within 12 to 24 months of baseline). A follow-up visit occurred at least 6 months after the third weight measure. The median follow-up period between the baseline and 1-year weight measurement ranged from 365 to 366 days across the 3 study groups. Follow-up duration between baseline and 2-year weight was also similar for the 3 groups, ranging from a median of 787 to 798 days.
Analyses evaluated independent and joint associations of weight loss and weight maintenance on each clinical indicator: type 2 diabetes, hypertension, and hyperlipidemia, starting with descriptive statistics and unadjusted analysis, followed by adjusted regression modeling. The simple analyses evaluated the unadjusted association of each clinical indicator for the study groups (OM, WLR, and WLM) using Cox regression for dichotomous outcomes (i.e., time-to-event regression model). For each clinical indicator, time-to-outcome was calculated as the number of days between the initial baseline weight measurement until the outcome of interest happens. For patients who did not develop the outcome of interest, the time was censored at the last follow-up visit. Testing for proportional hazard assumptions allows for the examination of consistent effects in the short and long term. We used a repeatedmeasures linear model with a first-degree autoregressive covariance structure (SAS PROC MIXED). Model assumptions were validated using residual plots (e.g., QQ plots, residuals vs time, residuals vs predicted).
Following the unadjusted analyses were models adjusting for selected patient characteristics and testing whether these characteristics modify the effect of weight loss on the clinical outcomes using the OM group as the reference group. The final models were adjusted for age, gender, BMI, diabetes, hypertension treatment, hyperlipidemia treatment, depression/anxiety treatment, osteoarthritis, asthma, gastrointestinal reflux disease, and the Charlson Comorbidity index. The Charlson Comorbidity Index is a validated score combining multiple comorbidities into a 10-year survival predictor. Comorbidities in this research were based on EHR diagnosis codes, weighted higher for diseases with greater mortality risk. Subgroup analyses were conducted for A1c among patients with prevalent diabetes and uncontrolled A1c (A1c ≥6.5%) and for SBP among those with prevalent hypertension and uncontrolled SBP (SBP ≥140 mm Hg); analyses were not conducted for lipid levels owing to limited availability of laboratory test data.
The cumulative incidence of each outcome was estimated by the Kaplan-Meier method and plotted over 10 years of follow-up for each group. In addition, Kaplan-Meier curves were used to compare time until outcome within the WLM group stratifying by the amount of weight loss at the end of year 2 of the index period (<7%, 7% to 10%, >10% to 15%, and >15%) using <7% weight loss as the reference group. A minimum weight loss of 7% was examined based on research demonstrating the effect of weight loss of at least 7% on preventing or delaying the development of type 2 diabetes. , The analyses were conducted using SAS version 9.4 (SAS Institute Inc., Cary North Carolina).
Results
Of the study sample of 63,567 patients, the majority (67%) were classified as OM (reference group). The remaining sample was similarly distributed between WLR (19%) and WLM (14%). The mean follow-up was 6.6 years (SD = 3.9). Female and younger patients were significantly more likely than male and older patients, respectively, to lose weight at 1 year (p <0.0001 for both). Baseline descriptive statistics and disease status for the study population are listed in Table 1 . Median weight loss from baseline to the 2-year weight measurement was 2.5% for the WLR group and 10.7% for the WLM group.
| Variable | OM (n = 42,534) | WLR (n = 12,227) | WLM (n = 8,806) | p Value |
|---|---|---|---|---|
| Age (years) | 53.3 (14.5) | 47.9 (15.4) | 50.1 (17.2) | <0.0001 |
| Baseline BMI (kg/m 2 ) | 35.3 (5.4) | 35.9 (6.1) | 35.8 (6.2) | <0.0001 |
| Women | 22,136 (52.0%) | 7,600 (62.2%) | 5,838 (66.3%) | <0.0001 |
| White | 41,337 (97.2%) | 11,855 (97.0%) | 8,510 (96.7%) | 0.096 |
| Hispanic | 544 (1.3%) | 171 (1.4%) | 141 (1.6%) | |
| Black | 522 (1.2%) | 159 (1.3%) | 132 (1.5%) | |
| Asian | 59 (0.1%) | 21 (0.2%) | 9 (0.1%) | |
| Hawaiian | 26 (<0.1%) | 7 (<0.1%) | 1 (<0.1%) | |
| American Indian | 24 (<0.1%) | 9 (<0.1%) | 5 (<0.1%) | |
| Other/unknown | 22 (<0.1%) | 5 (<0.1%) | 8 (<0.1%) | |
| Type 2 diabetes mellitus | 7,175 (16.9%) | 1,709 (14.0%) | 1,648 (18.8%) | <0.0001 |
| Prediabetes | 5,981 (14.1%) | 1,445 (11.8%) | 1,059 (12.0%) | <0.0001 |
| Treatment for hyperlipidemia | 13,853 (32.6%) | 3,025 (24.7%) | 2,527 (28.7%) | <0.0001 |
| Treatment for hypertension | 20,166 (47.4%) | 4,775 (39.1%) | 3,899 (44.3%) | <0.0001 |
| Any cardiovascular disease | 10,007 (23.5%) | 2,516 (20.6%) | 2,252 (25.6%) | <0.0001 |
| Congestive heart failure | 903 (2.1%) | 258 (2.1%) | 307 (3.5%) | <0.0001 |
| Stroke | 57 (0.1%) | 20 (0.2%) | 22 (0.3%) | 0.042 |
| Myocardial infarction | 483 (1.1%) | 121 (1.0%) | 93 (1.1%) | 0.364 |
| Osteoarthritis | 8,107 (19.1%) | 2,141 (17.5%) | 1,813 (20.6%) | <0.0001 |
| Treatment for depression/anxiety | 12,489 (29.4%) | 4,488 (36.7%) | 3,437 (39.0%) | <0.0001 |
| Asthma | 4,118 (9.7%) | 1,484 (12.1%) | 1,089 (12.4%) | <0.0001 |
| Gastroesophageal reflux disease (GERD) | 10,147 (23.9%) | 2,942 (24.1%) | 2,254 (25.6%) | 0.0023 |
| Charlson Index = 0 | 28,753 (67.6%) | 8,442 (69.0%) | 5,488 (62.3%) | <0.0001 |
| Charlson Index = 1 | 10,177 (23.9%) | 2,757 (22.6%) | 2,229 (25.3%) | |
| Charlson Index = 2+ | 3,604 (8.5%) | 1,028 (8.4%) | 1,089 (12.4%) |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
