Abstract
Objective
To evaluate the heterogeneity of treatment effects (HTEs) of intensive lifestyle intervention (ILI) in adults with type 2 diabetes and overweight/obesity, and identify a subgroup with greater cardiovascular benefits from ILI.
Methods
In a post-hoc analysis of the Look AHEAD trial, causal forest modeling was used to identify HTEs of ILI. The study population was stratified into four subgroups, and the associations of ILI with cardiovascular outcomes were assessed using multivariable Cox modeling compared to diabetes support & education (DSE).
Results
Among 4710 participants (mean age 58.9 years, 58.5% women), 768 primary outcomes occurred over a median follow-up of 9.5 years. Key variables identified through causal forest modeling were SF-36 mental health, diabetes duration, and urine albumin-creatinine ratio (ACR). In Subgroup 4 (SF-36 mental health > 55.64 and ACR > 10 mg/g), which had more cardiovascular risk factors and comorbidities, ILI significantly reduced the primary outcome risk (HR: 0.65, 95% CI: 0.48–0.87, P = 0.004) and three secondary outcomes compared to DSE. No cardiovascular benefits were observed in participants with SF-36 mental health ≤ 55.64 or ACR < 10 mg/g.
Conclusion
This post-hoc analysis of the Look AHEAD trial showed HTEs of ILI in adults with type 2 diabetes and overweight/obesity. Participants with better mental health, poorer renal function, and more cardiovascular risk factors were more likely to benefit from ILI.
Subject terms: Risk factors, Cardiovascular diseases
Introduction
The substantial increase in the parallel occurrence of type 2 diabetes and obesity plays a crucial role in the surging prevalence of cardiovascular disease (CVD), making it one of the foremost global public health challenges [1]. In recent years, although the emergence of new drugs such as Glucagon-Like Peptide-1 Receptor Agonist (GLP-1 RA) have offered valuable insights into assisting adults with overweight/obesity and type 2 diabetes in achieving weight loss and obtaining cardiovascular benefits [2, 3], lifestyle interventions such as healthy eating and physical exercise remains the cornerstone of primary prevention for the treatment of individuals with type 2 diabetes and concurrent obesity [4]. However, the Look AHEAD (Action for Health in Diabetes) trial [5], the largest randomized controlled trial to date on an intensive lifestyle-based weight loss intervention for overweight or obese adults with type 2 diabetes, did not show a significant reduction in cardiovascular morbidity or mortality. A post hoc analysis of the Look AHEAD trial showed that the presence of heterogeneity of treatment effects (HTEs) may explain the lack of a significant effect on cardiovascular outcomes [6]. Thus, identifying individuals who can achieve cardiovascular benefits from intensive lifestyle intervention (ILI) is essential for personalized management of patients with obesity and diabetes in the prevention and treatment of CVD.
A series of secondary analyses identified populations within those receiving ILI who would benefit from cardiovascular outcomes, such as individuals who lost at least 10% of their bodyweight in the first year or those who were able to maintain weight loss [7, 8]. Actually, several patient characteristics in demographics, medical history, physical examination, and laboratory values were associated with the level of treatment effect [6]. Relying solely on weight as a single dimension may be insufficient to comprehensively assess the HTEs of ILI. A multidimensional approach that considers various variables could provide a more accurate evaluation of heterogeneity and help identify populations that derive cardiovascular benefits from ILI. Therefore, to better assess the HTEs of ILI, we conducted a secondary analysis of the Look AHEAD trial, utilizing data from 42 baseline characteristic variables to build a causal forest model aimed at identifying populations that derive cardiovascular benefits from ILI.
Methods
Study design and population
Look AHEAD trial (trial registration NCT00017953) was a multi-center, randomized controlled clinical trial that evaluated the effects of an ILI on the risk of cardiovascular events in comparison with diabetes support & education (DSE) done at 16 clinical sites in the USA between Aug 22, 2001, and April 30, 2004 [9, 10]. The trial included 5145 adults with overweight/obesity and type 2 diabetes, and randomly assigned them to the ILI (n = 2570) or DSE (n = 2575) groups, data for 4906 of whom are available in public access data sets. After excluding 25 participants with missing data on the primary outcome, the sample used to construct the machine learning causal forest model included 4881 participants, consisting of 2023 men and 2858 women, with a median age of 59 years. Since the optimal tree model within the causal forest included three key variables—short form (36) health survey (SF-36) mental health, diabetes duration, and albumin-creatinine ratio (ACR)—we excluded participants with missing data for any of these variables (n = 6 for SF-36 mental health, n = 30 for diabetes duration, and n = 135 for ACR). Finally, the study cohort included 4710 participants in the primary analysis (Fig. S1). Baseline clinical characteristics were similar between the included and excluded participants (Table S1).
Four subgroups identified by machine learning causal forest model
The causal forest is a modified machine learning method based on recursive partitioning, which was devised to evaluate HTEs and able to specify variables related to HTEs in observational or experimental studies [11]. This data-driven approach involves dividing the data into subgroups based on variations in the effectiveness of the treatment to identify specific subgroups that are more likely to benefit from the ILI. In the causal tree analysis, the criterion for splitting is to maximize the difference in treatment effect values between groups based on a particular feature, thereby estimating the treatment effects for each subgroup. At this juncture, the average treatment effect within a group represents the predicted treatment effect for individuals within that group. According to the “counterfactual” principle of causal inference, the average treatment effect of after stratification by feature X can be articulated by the following equation:
(Equation one: Average treatment effect for individuals)
Equation one assumes that each subject is eligible for the ILI or DSE, with indicating a specific characteristic within the population. represents the outcome for subject when subjected to the ILI, and represents the outcome for the same subject in the DSE group. The process involves calculating the disparity for each stratum post-segmentation and then selecting the characteristic that yields the greatest disparity among the strata as the optimal point for bifurcation within the model. In actual circumstances, only the observed outcome for individual , either or can be obtained, not both simultaneously. Therefore, within the context of Look AHEAD trial, we typically substitute the individual treatment effect of ILI with the average treatment effect across the population i.e., and to measure the disparity in outcomes between the two groups. This approach allows for the estimation of the overall impact of ILI by comparing the average outcomes of the ILI and DSE groups.
(Equation 2: Mean treatment effect of ILI across stratified population groups)
According to Equation 2, when constructing a causal tree, the algorithm automatically identifies the feature value that maximizes the difference in treatment effects between the ILI and DSE groups into which the current population is divided (the splitting criterion), and then stratifies the groups based on this feature value. Subsequently, it repeats the aforementioned steps within each subgroup until a causal tree that meets the preset requirements is generated. In this analysis, we incorporated a total of 42 baseline factors as characteristic variables, encompassing demographics, physiologic and biomarker, lifestyle factors, laboratory test indicators, medical history, and the SF-36 measurement scale (Table S2). After constructing the forest model, the importance of each feature variable is calculated based on the number of times it is used for splitting at each level of the forest, and the features are ranked according to their importance. Finally, the optimal tree model within the causal forest was determined based on the effectiveness and clinical significance of the model and the importance of the variables. Based on the optimal tree model within the causal forest, the population was stratified into four subgroups (Subgroup 1: SF-36 mental health ≤ 55.64 and Diabetes Duration ≤ 4 years; Subgroup 2: SF-36 mental health ≤ 55.64 and Diabetes Duration > 4 years; Subgroup 3: SF-36 mental health > 55.64 and ACR ≤ 10 mg/g; Subgroup 4: SF-36 mental health > 55.64 and ACR > 10 mg/g) (Fig. 1).
Fig. 1. Four subgroups with HTEs identified by the causal forest model.
SF-36: short form (36) health survey; ACR: albumin-creatinine ratio; HTEs heterogeneity of treatment effects.
Study intervention
Individuals assigned to the ILI group embarked on a structured behavioral weight management journey. This initiative featured a calorie-conscious diet, aiming for a daily intake of 1200–1800 kcal, complemented by a moderate-intensity physical activity regimen, striving for at least 150 min weekly. The objective was to facilitate a 7% reduction in baseline body weight. Throughout the initial half-year, these participants engaged in weekly sessions with a multidisciplinary team of behavioral psychologists, dietitians, and exercise professionals. In the subsequent six months, the frequency of these encounters was tailored to a Bi-weekly schedule. From the second to the fourth year, the engagement continued with Bi-monthly interactions, alternating between face-to-face meetings and remote check-ins via phone, email, or correspondence. In contrast, the DSE group was offered periodic educational and social gatherings, focusing on themes of nutrition and physical activity, to foster a supportive community and share knowledge. The details have been described previously [10, 12].
Primary and secondary outcomes
In adherence to the Look AHEAD trial protocol [9], we have constrained our current analyses to the outcomes that were pre-defined for the Look AHEAD trial, ensuring they were evaluated by an outcomes committee. The primary composite cardiovascular outcome was the first occurrence of death from cardiovascular causes, non-fatal myocardial infarction, non-fatal stroke, or hospitalization for angina. The following three secondary composite cardiovascular outcomes were also considered: death from cardiovascular causes, non-fatal myocardial infarction, or non-fatal stroke (Secondary outcome 1); death from any cause, non-fatal myocardial infarction, non-fatal stroke, or hospitalization for angina (Secondary outcome 2); and death from any cause, non-fatal myocardial infarction, non-fatal stroke, hospitalization for angina, coronary-artery bypass grafting, percutaneous coronary intervention, hospitalization for heart failure, carotid endarterectomy, or peripheral vascular disease (Secondary outcome 3).
Statistical analysis
Baseline characteristics were analyzed by descriptive statistics, which are presented as mean (standard deviation) for continuous data and number (%) for categorical data. The missing covariates were handled using multiple imputation with the “mice” package in R, which generates plausible values for the missing data through model-based prediction, creates a complete dataset, and incorporates these imputed values into the analysis results [13]. By using the “mice” package, we can more effectively handle missing data and reduce the resulting bias. The baseline characteristics were similar both before and after multiple imputation (Table S3). For the 4710 participants, the risks of cardiovascular outcomes associated with the ILI (vs. the DSE arm) were evaluated using multivariate-adjusted Cox proportional hazard models. All included participants were stratified into four subgroups based on the causal forest model. The risks of cardiovascular outcomes among these four subgroups were evaluated using multivariate-adjusted Cox proportional hazard models. Subsequently, cumulative incidences of the primary outcome were estimated for the ILI and DSE arms in the four subgroups using the Kaplan-Meier method and the log-rank test. Multivariate-adjusted Cox proportional hazards regression analysis was also used to assess the cardiovascular effects of ILI (vs. the DSE arm) in the four subgroups. Fully adjusted Cox regression models were generated for the primary and secondary outcomes, involving adjustment for age, gender, race, body mass index (BMI), glycosylated hemoglobin (HbA1c), low-density lipoprotein cholesterol (LDL-c), serum creatinine, smoking status, hypertension, and diabetes duration. In addition, to mitigate the influence of pre-existing CVD on the results, the analyses were repeated among participants with no previous history of CVD.
A two-sided P value < 0.05 was considered statistically significant. All analyses were performed using Stata version 14 (StataCorp, College Station, TX, USA) and R language (version 4.3.2).
Results
Cardiovascular effects of ILI in the included participants
A total of 4,710 overweight or obese participants with type 2 diabetes were assigned to the ILI (n = 2352) and DSE (n = 2358) groups, respectively, with both groups exhibiting similar baseline clinical characteristics (Table S4). During a median follow-up of 9.5 years (interquartile range: 8.6-10.3 years), the primary outcome occurred in 393 (16.6%) participants in the DSE group and 375 (15.9%) participants in the ILI group. As observed in the previous study [5], no significant differences were found in the risk of the primary outcome (hazard ratio [HR] 0.96; 95% confidence interval [CI] 0.83–1.11; P = 0.570) or any of the three secondary outcomes (all P > 0.05) between the ILI and DSE groups (Table S5).
Cardiovascular risks across four subgroups identified by the causal forest model
The 4710 enrolled participants were divided into four subgroups with HTEs identified by the causal forest model. Subgroup 1, consisting of 1286 participants (27.3%), was characterized by poor mental health and a short duration of diabetes; Subgroup 2, consisting of 1481 participants (31.4%), was characterized by poor mental health and a long duration of diabetes; Subgroup 3, consisting of 1095 participants (23.2%), was characterized by good mental health and good renal function; Subgroup 4, consisting of 848 participants (18.0%), was characterized by good mental health and poor renal function (Fig. 1). Compared to the other subgroups, participants in Subgroup 4 were older, more likely to be male, had a higher BMI and waist circumference, were more frequently smokers and drinkers, and had a higher prevalence of hypertension, CVD, and dyslipidemia (Table 1).
Table 1.
Baseline characteristics of participants in four subgroups identified by the casual forest model.
| Variables | Subgroup 1 (N = 1286) | Subgroup 2 (N = 1481) | Subgroup 3 (N = 1095) | Subgroup 4 (N = 848) | P value |
|---|---|---|---|---|---|
| Age, years | 57.3 (6.5) | 58.9 (6.7) | 59.6 (6.6) | 60.3 (6.9) | <0.001 |
| Female, n (%) | 835 (64.9) | 878 (59.3) | 622 (56.8) | 422 (49.8) | <0.001 |
| Race, n (%) | 0.08 | ||||
| White | 857 (66.6) | 976 (65.9) | 716 (65.4) | 561 (66.2) | |
| Black (not Hispanic) | 191 (14.9) | 231 (15.6) | 212 (19.4) | 136 (16.0) | |
| Hispanic | 190 (14.8) | 217 (14.7) | 130 (11.9) | 127 (15.0) | |
| Other/Mixed | 48 (3.7) | 57 (3.8) | 37 (3.4) | 24 (2.8) | |
| Education, n (%) | 0.123 | ||||
| <13 years | 239 (18.6) | 314 (21.2) | 193 (17.6) | 170 (20.0) | |
| 13–16 years | 477 (37.1) | 548 (37.0) | 400 (36.5) | 333 (39.3) | |
| ≥16years | 570 (44.3) | 619 (41.8) | 502 (45.8) | 345 (40.7) | |
| Income, n (%) | 0.018 | ||||
| < $20 K | 155 (12.1) | 183 (12.4) | 97 (8.9) | 88 (10.4) | |
| $20 K–$40 K | 271 (21.1) | 325 (21.9) | 206 (18.8) | 172 (20.3) | |
| $40 K–$60 K | 269 (20.9) | 318 (21.5) | 219 (20.0) | 169 (19.9) | |
| $60 K–$80 K | 194 (15.1) | 227 (15.3) | 191 (17.4) | 153 (18.0) | |
| > $80 K | 397 (30.9) | 428 (28.9) | 382 (34.9) | 266 (31.4) | |
| BMI, kg/m2 | 36.2(6.1) | 36.0(5.9) | 35.4 (5.6) | 36.3 (5.8) | 0.002 |
| Waist Circumference, cm | 113.3 (14.0) | 114.5 (13.9) | 112.4 (13.4) | 116.3 (14.5) | <0.001 |
| SBP, mmHg | 128.5 (16.8) | 129.4 (17.2) | 126.6 (16.7) | 132.3 (17.4) | <0.001 |
| DBP, mmHg | 71.0 (9.4) | 69.4 (9.5) | 69.7 (9.4) | 71.3 (10.0) | <0.001 |
| Smoking, n (%) | 0.016 | ||||
| Never | 643 (50.0) | 756 (51.0) | 560 (51.1) | 384 (45.3) | |
| Former smoker | 575 (44.7) | 665 (44.9) | 501 (45.8) | 427 (50.4) | |
| Current smoker | 68 (5.3) | 60 (4.1) | 34 (3.1) | 37 (4.4) | |
| Alcohol, n (%) | 0.076 | ||||
| Never | 841 (65.4) | 1025 (69.2) | 734 (67.0) | 548 (64.6) | |
| ≥1 g/week | 445 (34.6) | 456 (30.8) | 361 (33.0) | 300 (35.4) | |
| HbA1c, % | 7.0 (1.1) | 7.6 (1.2) | 7.1 (1.1) | 7.5 (1.2) | <0.001 |
| HDL-c, mg/dl | 43.9 (11.7) | 43.2 (12.0) | 44.4 (12.2) | 42.5 (11.7) | 0.001 |
| LDL-c, mg/dl | 117.3 (32.4) | 110.4 (31.8) | 112.8 (30.5) | 109.6 (33.2) | <0.001 |
| Creatinine, mg/dl | 0.8 (0.2) | 0.8 (0.2) | 0.8 (0.2) | 0.9 (0.2) | <0.001 |
| Hypertension, n (%) | 1024 (79.6) | 1299 (87.7) | 868 (79.3) | 752 (88.7) | <0.001 |
| CVD, n (%) | 120 (9.3) | 233 (15.7) | 144 (13.2) | 163 (19.2) | <0.001 |
| Dyslipidemia, n (%) | 864 (67.2) | 1062 (71.7) | 776 (70.9) | 623 (73.5) | 0.009 |
| ACR, mg/g | 2.7 (10.5) | 5.9 (27.4) | 0.6 (0.2) | 9.0 (27.4) | <0.001 |
| ABI | 1.2 (0.1) | 1.2 (0.1) | 1.2 (0.1) | 1.2 (0.2) | 0.22 |
| Diabetes Duration, years | 2.0 (1.3) | 10.8 (6.4) | 6.2 (5.9) | 7.5 (6.9) | <0.001 |
Continuous and categorical variables are presented as mean (SD) and number (%), respectively.
Subgroup 1: SF-36 Mental Health ≤ 55.64 and Diabetes Duration ≤ 4 years; Subgroup 2: SF-36 Mental Health ≤ 55.64 and Diabetes Duration > 4 years; Subgroup 3: SF-36 Mental Health > 55.64 and ACR ≤ 10 mg/g; Subgroup 4: SF-36 Mental Health > 55.64 and ACR > 10 mg/g.
BMI body mass index, SBP Systolic blood pressure, DBP Diastolic blood pressure, HbA1c glycosylated hemoglobin, HDL-c high-density lipoprotein cholesterol, LDL-c low-density lipoprotein cholesterol, CVD cardiovascular disease, ACR Albumin-Creatinine Ratio, ABI Ankle Brachial Index, SD standard deviation.
During a median follow-up of 9.5 years (interquartile range: 8.6–10.3 years), the primary outcome occurred in 158 (12.3%) participants in Subgroup 1, 266 (18.0%) participants in Subgroup 2, 160 (14.6%) participants in Subgroup 3, and 184 (21.7%) participants in Subgroup 4. Similarly, the highest event rates for the three secondary outcomes were observed in Subgroup 4 (Fig. S2). In the fully adjusted model, compared with the participants in Subgroup 4, those in Subgroup 3 were associated with a lower risk of the primary outcome (HR: 0.79, 95% CI: 0.63-0.98, P = 0.029), but no differences were observed in participants Subgroup 1 (HR: 0.87, 95% CI: 0.69–1.09, P = 0.228) and Subgroup 2 (HR: 0.87, 95% CI: 0.71–1.05, P = 0.143) (Table S6). Obviously, compared to the other subgroups, participants in Subgroup 4 had more cardiovascular risk factors, a higher number of comorbidities, and were associated with an increased cardiovascular risk.
Heterogeneous cardiovascular effects of ILI among the four subgroups
Participants in the four subgroups were assigned to either the ILI or DSE arm: Subgroup 1 (671 ILI vs. 615 DSE), Subgroup 2 (728 ILI vs. 753 DSE), Subgroup 3 (537 ILI vs. 558 DSE), and Subgroup 4 (416 ILI vs. 432 DSE), with similar baseline clinical characteristics between the ILI and DSE arms in each subgroup (Table 2 and Table S7). For the 848 participants in Subgroup 4, the mean age was 59.9 ± 6.7 years in the ILI group and 60.7 ± 7.1 years in the DSE group, with 50.5% and 49% females, respectively. Both arms exhibited a higher prevalence of hypertension, dyslipidemia, and CVD (Table 2). During a median follow-up of 9.3 years (interquartile range: 7.9–10.1 years) in Subgroup 4, 73 (17.5%) participants in the ILI arm and 111 (25.7%) participants in the DSE arm experienced the primary outcome. The Kaplan-Meier survival function curves in Subgroup 4 showed a lower cumulative incidence of the primary outcome in the ILI arm compared with the DSE arm (P = 0.004), and there was no difference between the ILI and DSE arms in Subgroup 1 (P = 0.934), Subgroup 2 (P = 0.187), and Subgroup 3 (P = 0.688) (Fig. 2).
Table 2.
Baseline characteristics of participants in the ILI and DSE arms within Subgroups 4.
| Variables | DSE arm (n = 432) | ILI arm (n = 416) | P value |
|---|---|---|---|
| Age, years | 60.7 (7.1) | 59.9 (6.7) | 0.099 |
| Female, n (%) | 218 (50.5) | 204 (49.0) | 0.729 |
| Race, n (%) | 0.802 | ||
| White | 280 (64.8) | 281 (67.5) | |
| Black (not Hispanic) | 74 (17.1) | 62 (14.9) | |
| Hispanic | 65 (15.0) | 62 (14.9) | |
| Other/Mixed | 13 (3.0) | 11 (2.6) | |
| Education, n (%) | 0.471 | ||
| <13 years | 90 (20.8) | 80 (19.2) | |
| 13–16 years | 175 (40.5) | 158 (38.0) | |
| ≥16years | 167 (38.7) | 178 (42.8) | |
| Income, n (%) | 0.986 | ||
| < $20 K | 45 (10.4) | 43 (10.3) | |
| $20 K–$40 K | 86 (19.9) | 86 (20.7) | |
| $40 K–$60 K | 88 (20.4) | 81 (19.5) | |
| $60 K–$80 K | 80 (18.5) | 73 (17.5) | |
| BMI, kg/m2 | 36.1 (5.6) | 36.5 (6.0) | 0.274 |
| Waist Circumference, cm | 115.8 (14.0) | 116.7 (15.2) | 0.376 |
| SBP, mmHg | 132.7 (17.7) | 131.9 (17.0) | 0.523 |
| DBP, mmHg | 71.4 (10.1) | 71.1 (9.8) | 0.641 |
| Smoking, n (%) | 0.262 | ||
| Never | 203 (47.0) | 181 (43.5) | |
| Former smoker | 207 (47.9) | 220 (52.9) | |
| Current smoker | 22 (5.1) | 15 (3.6) | |
| Alcohol, n (%) | 0.338 | ||
| Never | 272 (63.0) | 276 (66.3) | |
| ≥1 g/week | 160 (37.0) | 140 (33.7) | |
| HbA1c, % | 7.6 (1.3) | 7.4 (1.2) | 0.041 |
| HDL-c, mg/dl | 42.3 (11.7) | 42.7 (11.7) | 0.657 |
| LDL-c, mg/dl | 110.8 (34.5) | 108.3 (31.8) | 0.260 |
| Creatinine, mg/dl | 0.86 (0.21) | 0.84 (0.23) | 0.465 |
| Hypertension, n (%) | 381 (88.2) | 371 (89.2) | 0.730 |
| CVD, n (%) | 86 (19.9) | 77 (18.5) | 0.668 |
| Dyslipidemia, n (%) | 316 (73.1) | 307 (73.8) | 0.891 |
| ACR, mg/g | 8.2 (20.1) | 9.9 (33.3) | 0.375 |
| ABI | 1.2 (0.2) | 1.2 (0.1) | 0.853 |
| Diabetes Duration, years | 7.7 (7.0) | 7.2 (6.8) | 0.353 |
Continuous and categorical variables are presented as mean (SD) and number (%), respectively.
ILI intensive lifestyle intervention, DSE diabetes support and education, BMI body mass index, SBP systolic blood pressure, DBP diastolic blood pressure, HbA1c glycosylated hemoglobin, HDL-c high-density lipoprotein cholesterol, LDL-c low-density lipoprotein cholesterol, CVD cardiovascular disease, ACR albumin-creatinine ratio, ABI ankle brachial index, SD standard deviation.
Fig. 2. Cumulative incidence curves with 95% CIs for primary outcomes between the ILI and DSE arms across the four subgroups identified by the causal forest model.
CI: confidence interval; ILI: intensive lifestyle intervention; DSE: diabetes support and education.
In the fully adjusted model for the participants in Subgroup 4, participants in the ILI arm had a 35% lower risk of the primary outcome (HR: 0.65, 95% CI: 0.48–0.87, P = 0.004), a 29% lower risk of secondary outcome 1 (HR: 0.71, 95% CI: 0.50–1.02, P = 0.060), a 35% lower risk of secondary outcome 2 (HR: 0.65, 95% CI: 0.50–0.85, P = 0.002), and a 24% lower risk of secondary outcome 3 (HR: 0.76, 95% CI: 0.59–0.97, P = 0.025) compared with the participants in the DSE arm (Table 3). However, no differences were found in the risks of the primary outcome in Subgroup 1 (HR: 1.01, 95% CI: 0.73–1.38, P = 0.971), Subgroup 2 (HR: 1.27, 95% CI: 0.99–1.62, P = 0.056), or Subgroup 3 (HR: 0.88, 95% CI: 0.65–1.21, P = 0.436) between the ILI and DSE arms. The results for the three secondary outcomes were similar to those for the primary outcome in Subgroup 1 to 3 (Table 3).
Table 3.
Risk of primary and three secondary outcomes for ILI vs. DSE within each subgroup identified by the causal forest model.
| Subgroups | No. of events in ILI (%) | No. of events in DSE (%) | Unadjusted | Fully adjusted | ||
|---|---|---|---|---|---|---|
| HR (95% CI) | P value | HR (95%CI) | P value | |||
| Primary outcome: Death from cardiovascular causes, non-fatal myocardial infarction, non-fatal stroke, or hospitalization for angina. | ||||||
| Subgroup 1 | 82/671 (12.2) | 76/615 (12.4) | 1.01 (0.74–1.38) | 0.934 | 1.01 (0.73–1.38) | 0.971 |
| Subgroup 2 | 143/728 (19.6) | 123/753 (16.3) | 1.18 (0.92–1.50) | 0.187 | 1.27 (0.99–1.62) | 0.056 |
| Subgroup 3 | 77/537 (14.3) | 83/558 (14.9) | 0.94 (0.69–1.28) | 0.688 | 0.88 (0.65–1.21) | 0.436 |
| Subgroup 4 | 73/416 (17.5) | 111/432 (25.7) | 0.65 (0.48–0.87) | 0.004 | 0.65 (0.48–0.87) | 0.004 |
| Secondary outcome 1: Death from cardiovascular causes, non-fatal myocardial infarction, non-fatal stroke. | ||||||
| Subgroup 1 | 54/671 (8.0) | 54/615 (8.8) | 0.94 (0.64–1.37) | 0.733 | 0.93 (0.63–1.36) | 0.703 |
| Subgroup 2 | 95/728 (13.0) | 80/753 (10.6) | 1.20 (0.89–1.62) | 0.228 | 1.34 (1.00–1.81) | 0.054 |
| Subgroup 3 | 46/537 (8.6) | 52/558 (9.3) | 0.89 (0.60–1.33) | 0.573 | 0.84 (0.57–1.26) | 0.406 |
| Subgroup 4 | 54/416 (13.0) | 80/432 (18.5) | 0.68 (0.48–0.93) | 0.030 | 0.71 (0.50–1.02) | 0.060 |
| Secondary outcome 2: Death from any cause, non-fatal myocardial infarction, non-fatal stroke, or hospitalization for angina. | ||||||
| Subgroup 1 | 101/671 (15.1) | 93/615 (15.1) | 1.02 (0.77–1.35) | 0.895 | 1.01 (0.76–1.34) | 0.963 |
| Subgroup 2 | 95/728 (24.3) | 162/753 (21.5) | 1.11 (0.89–1.37) | 0.352 | 1.19 (0.96–1.47) | 0.116 |
| Subgroup 3 | 91/537 (16.9) | 104/558 (18.6) | 0.89 (0.67–1.17) | 0.393 | 0.83 (0.63–1.10) | 0.201 |
| Subgroup 4 | 91/416 (21.9) | 138/432 (31.9) | 0.65 (0.50–0.85) | 0.001 | 0.65 (0.50–0.85) | 0.002 |
| Secondary outcome 3: Death from any cause, non-fatal myocardial infarction, non-fatal stroke, or hospitalization for angina, CABG, PCI, hospital admission for heart failure, carotid endarterectomy, or peripheral vascular disease. | ||||||
| Subgroup 1 | 116/671 (17.3) | 106/615 (17.2) | 1.03 (0.79–1.34) | 0.823 | 1.02 (0.79–1.34) | 0.859 |
| Subgroup 2 | 197/728 (27.1) | 186/753 (24.7) | 1.07 (0.88–1.31) | 0.499 | 1.15 (0.94–1.40) | 0.181 |
| Subgroup 3 | 102/537 (19.0) | 118/558 (21.1) | 0.87 (0.67–1.14) | 0.316 | 0.83 (0.63–1.08) | 0.168 |
| Subgroup 4 | 122/416 (29.3) | 153/432 (35.4) | 0.78 (0.62–0.99) | 0.042 | 0.76 (0.59–0.97) | 0.025 |
The multivariable model was adjusted for age, gender, race, body mass index, glycosylated hemoglobin, low-density lipoprotein cholesterol, serum creatinine, smoking status, hypertension, and diabetes duration.
Subgroup 1: SF-36 Mental Health ≤ 55.64 and Diabetes Duration ≤ 4 years; Subgroup 2: SF-36 Mental Health ≤ 55.64 and Diabetes Duration > 4 years; Subgroup 3: SF-36 Mental Health > 55.64 and ACR ≤ 10 mg/g; Subgroup 4: SF-36 Mental Health > 55.64 and ACR > 10 mg/g.
ILI intensive lifestyle intervention, DSE diabetes support and education, HR hazard ratio, CI confidence interval.
Sensitivity analyses
To mitigate the influence of pre-existing CVD on the results, the sensitivity analyses were conducted after excluding participants with CVD history at baseline. In the fully adjusted model, participants in the ILI arm had a 34% lower risk of the primary outcome compared with the participants in the DSE arm in Subgroup 4 (HR: 0.66, 95% CI: 0.44–1.00, P = 0.048) (Table 4). However, no significant differences were observed in the risks of the primary outcome in Subgroup 1 to 3 (all P > 0.05) (Table 4). Similar results were also found in the analyses for the secondary outcome 1 and 2 (Table 4). Therefore, heterogeneous cardiovascular effects of ILI were also observed after excluding participants with CVD history at baseline.
Table 4.
Risk of primary and three secondary outcomes for ILI vs. DSE within each subgroup identified by the causal forest model after excluding those with a history of CVD at baseline (n = 4050).
| Subgroups | No. of events in ILI (%) | No. of events in DSE (%) | Unadjusted | Fully adjusted | ||
|---|---|---|---|---|---|---|
| HR (95% CI) | P value | HR (95%CI) | P value | |||
| Primary outcome: Death from cardiovascular causes, non-fatal myocardial infarction, non-fatal stroke, or hospitalization for angina. | ||||||
| Subgroup 1 | 56/560 (10.0) | 54/606 (8.9) | 0.90 (0.62–1.31) | 0.593 | 0.91 (0.62-1.33) | 0.613 |
| Subgroup 2 | 87/609 (14.3) | 84/639 (13.1) | 1.04 (0.77–1.41) | 0.777 | 1.09 (0.81–1.48) | 0.561 |
| Subgroup 3 | 42/460 (9.1) | 58/491 (11.8) | 0.74 (0.50–1.10) | 0.135 | 0.71 (0.48–1.06) | 0.098 |
| Subgroup 4 | 39/339 (11.5) | 61/346 (17.6) | 0.63 (0.42–0.94) | 0.025 | 0.66 (0.44–1.00) | 0.048 |
| Secondary outcome 1: Death from cardiovascular causes, non–fatal myocardial infarction, non-fatal stroke. | ||||||
| Subgroup 1 | 42/560 (7.5) | 35/606 (5.8) | 0.78 (0.50–1.22) | 0.271 | 0.77 (0.49–1.22) | 0.265 |
| Subgroup 2 | 55/609 (9.0) | 53/639 (8.3) | 1.05 (0.72–1.53) | 0.796 | 1.14 (0.78–1.67) | 0.493 |
| Subgroup 3 | 26/460 (5.7) | 37/491 (7.5) | 0.72 (0.44–1.19) | 0.204 | 0.69 (0.42–1.15) | 0.693 |
| Subgroup 4 | 25/339 (7.4) | 42/346 (12.1) | 0.59 (0.36–0.97) | 0.039 | 0.64 (0.39–1.06) | 0.080 |
| Secondary outcome 2: Death from any cause, non-fatal myocardial infarction, non-fatal stroke, or hospitalization for angina. | ||||||
| Subgroup 1 | 72/560 (12.9) | 71/606 (11.7) | 0.92 (0.66–1.28) | 0.628 | 0.94 (0.67–1.30) | 0.693 |
| Subgroup 2 | 115/609 (18.9) | 117/639 (18.3) | 0.99 (0.77–1.28) | 0.952 | 1.04 (0.80–1.35) | 0.771 |
| Subgroup 3 | 53/460 (11.5) | 75/491 (15.3) | 0.72 (0.51–1.02) | 0.068 | 0.69 (0.49–0.99) | 0.044 |
| Subgroup 4 | 54/339 (15.9) | 81/346 (23.4) | 0.66 (0.47–0.93) | 0.017 | 0.68 (0.48–0.97) | 0.031 |
| Secondary outcome 3: Death from any cause, non-fatal myocardial infarction, non–fatal stroke, or hospitalization for angina, CABG, PCI, hospital admission for heart failure, carotid endarterectomy, or peripheral vascular disease. | ||||||
| Subgroup 1 | 79/560 (14.1) | 82/606 (13.5) | 0.97 (0.71–1.33) | 0.860 | 0.99 (0.73–1.36) | 0.962 |
| Subgroup 2 | 129/609 (21.2) | 133/639 (20.8) | 0.98 (0.77–1.25) | 0.879 | 1.03 (0.81–1.31) | 0.811 |
| Subgroup 3 | 61/460 (13.3) | 83/491 (16.9) | 0.75 (0.54–1.04) | 0.086 | 0.73 (0.53–1.03) | 0.070 |
| Subgroup 4 | 79/339 (23.3) | 90/346 (26.0) | 0.88 (0.65–1.19) | 0.413 | 0.90 (0.66–1.22) | 0.496 |
The multivariable model was adjusted for age, gender, race, body mass index, glycosylated hemoglobin, low-density lipoprotein cholesterol, serum creatinine, smoking status, hypertension, and diabetes duration.
Subgroup 1: SF-36 Mental Health ≤ 55.64 and Diabetes Duration ≤ 4 years; Subgroup 2: SF-36 Mental Health ≤ 55.64 and Diabetes Duration > 4 years; Subgroup 3: SF-36 Mental Health > 55.64 and ACR ≤ 10 mg/g; Subgroup 4: SF-36 Mental Health > 55.64 and ACR > 10 mg/g.
ILI intensive lifestyle intervention, DSE diabetes support and education, HR hazard ratio, CI confidence interval.
Discussion
In this post-hoc analysis of Look AHEAD trial, we constructed causal forests based on 42 routinely available clinical variables that might impact therapeutic effects and clinical outcomes, with baseline SF-36 mental health, diabetes duration, and ACR identified as key factors distinguishing overweight or obese participants with type 2 diabetes who derive high from low benefits associated with ILI. For participants with SF-36 mental health > 55.64 and ACR > 10 mg/g, who had more cardiovascular risk factors and comorbidities, ILI was significantly associated with a lower risk of cardiovascular outcomes. In contrast, the cardiovascular effects of the ILI were not observed in other participants (Fig. 3). These findings suggest that the ILI exhibits HTEs on cardiovascular outcomes in participants with type 2 diabetes and overweight/obesity. Specifically, the ILI might help in lowering the risk of cardiovascular events in participants with better mental health, but poorer renal function, and more cardiovascular risk factors and comorbidities. Our study underscores the importance of personalized health management and precision medicine, highlighting that tailoring ILI to individuals based on specific baseline characteristics, such as mental health status, renal function, and cardiovascular risk factors, may enhance its effectiveness in reducing cardiovascular outcomes in adults with type 2 diabetes and overweight/obesity.
Fig. 3.
Heterogeneous cardiovascular effects of intensive lifestyle intervention in adults with type 2 diabetes and overweight/obesity identified by the causal forest model.
The Look AHEAD trial was the largest trial to date assessing an intensive lifestyle-based weight loss intervention. However, because the trial failed to achieve its primary objective of reducing cardiovascular risk through a lifestyle program focused on weight reduction, it was terminated early based on a futility analysis, with a median follow-up of 9.6 years [5]. Given the overall neutral results of the Look AHEAD trial, several post-hoc analyses were conducted. A secondary analysis revealed heterogeneous treatment effects of intensive lifestyle intervention in overweight or obese adults with type 2 diabetes, identifying that HbA1c levels and a short questionnaire on general health could help pinpoint individuals most likely to benefit from a weight loss-focused intensive lifestyle intervention [14]. While previous studies have examined the HTEs in the Look AHEAD trial, they did not account for the role of early kidney function in the relationship between diabetes and cardiovascular events. The unique contribution of our study is its pioneering use of the causal tree machine learning algorithm, highlighting the influence of early renal function on the heterogeneity of cardiovascular outcomes following ILI in individuals with type 2 diabetes and overweight/obesity.
In supplementary analyses, participants with better mental health status were found to have improved cardiovascular outcomes compared with those with poorer mental health status through ILI. Bąk-Sosnowska et al. demonstrated in a study of 768 patients with chronic diseases that higher levels of mindfulness and stronger internal health locus of control were significantly associated with better medication adherence [15]. This supports our observation in the Look AHEAD trial that participants with higher SF-36 mental health scores derived greater cardiovascular benefits from ILI, potentially attributable to enhanced adherence facilitated by favorable psychological states. In addition, weight management is also influenced by mental health status. Pacanowski et al. explored the relationship between psychological status and weight variability in the Look AHEAD trial, noting an intriguing observation that poorer psychological functioning often precedes greater weight instability [16]. Another study indicated that participants in the ILI group who maintained their body weight within the target range (50%-100%) had a reduced risk of cardiovascular events [8]. This suggests that within the Look AHEAD population, individuals with higher SF-36 scores may experience more stable weight changes, which could indirectly improve their cardiovascular outcomes. Furthermore, research by Zu and colleagues showed that lower body weight variability or more stable weight trajectories were associated with a reduced risk of renal outcomes in individuals with overweight/obesity and type 2 diabetes [17, 18]. These findings highlight the importance of weight reduction and sustained weight control, particularly in younger individuals, for improving both cardiovascular and renal outcomes in this patient population.
According to the KDIGO 2024 clinical practice guideline for the evaluation and management of chronic kidney disease (CKD), urinary ACR is emphasized as a sensitive and specific marker for the early detection of CKD [19]. The ACR threshold of >10 mg/g used in our causal tree was data-driven, identified by the model as the point that maximized heterogeneity in treatment response. Notably, this threshold aligns with growing evidence that even high-normal or mildly elevated albuminuria is associated with increased cardiovascular risk and may represent early renal damage. Recent studies have shown that the risks of hypertension, CVD, and all-cause mortality increase progressively with rising urinary ACR, even below the conventional threshold of 30 mg/g [20–23]. This suggests that what is traditionally considered ‘normal’ may, in fact, conceal a subtle yet significant health risk for certain individuals. Niu et al. further reported that patients with diabetes and albuminuria face a heightened risk of adverse clinical outcomes, underscoring the need for early intervention [24]. To the best of our knowledge, this study is the first to demonstrate that participants with an ACR > 10 mg/g and good mental health may derive cardiovascular benefits from ILI. This is crucial for the personalized management of patients with type 2 diabetes and overweight/obesity to reduce cardiovascular burden.
The cardiovascular-kidney-metabolic syndrome (CKM) is a systemic condition resulting from the complex pathophysiological interactions between obesity, diabetes, CKD, and CVD, driven by the cumulative effects of multiple risk factors [25]. In addressing such complex clinical scenarios, where various diseases intersect or coexist, there is a strong emphasis on a multidisciplinary, collaborative diagnostic and therapeutic approach, as well as shared decision-making between healthcare providers and patients. Current evidence supports the use of several drugs for managing CKM, including sodium-glucose cotransporter 2 inhibitors (SGLT2i) [26], GLP-1RA [27], finerenone [28], and the dual GIP/GLP-1 receptor agonist [29]. However, challenges related to medical insurance policies, reimbursement systems, patient education, and institutional regulations have limited the effective use of these medications in clinical practice, leading to delays in treatment for some patients. We look to the future with optimism, anticipating that high-quality clinical studies will provide further evidence of the cardiovascular benefits associated with a combination of ILI and emerging therapies, especially for individuals with early kidney damage who have type 2 diabetes and are overweight or obese.
This study has several notable strengths. It leveraged data from a large-scale, multi-center, randomized controlled trial, specifically targeting adults with overweight/obesity and type 2 diabetes. Additionally, the study was characterized by an extended follow-up period of nearly 10 years. In our study, we applied causal forest model to analyze the data. The model automatically identified the SF-36 mental health, diabetes duration, and ACR as key effect modifiers from 42 baseline variables. This machine learning approach demonstrates its unique advantage in extracting hidden information from high-dimensional data [30]. However, this study also had several limitations. First, the causal forest method requires careful tuning of hyperparameters and may be sensitive to the choice of input variables. Interpretability of the resulting subgroups can also be challenging, though the use of a single optimal tree helps to balance model performance with clinical clarity. Second, as with other observational studies, the potential influence of residual confounders-whether measured or unmeasured-cannot be entirely ruled out, despite adjusting for key confounders in our risk estimation models. However, the consistency of results across sensitivity analyses strengthens the robustness of our primary findings. Nevertheless, causality cannot be definitively inferred from these observations. Our findings should be regarded as hypothesis-generating and warrant further validation in ongoing and future research, including real-world studies. Third, in clinical practice, we have observed that the same person can exhibit significant variations in ACR measurements on the same day, which has also been reported in the relevant literature [31, 32]. A single ACR measurement from a spot urine sample is insufficient to accurately reflect the patient’s actual urinary albumin excretion. Therefore, it is necessary to collect 24-hour urine samples or conduct multiple tests to obtain accurate results in future studies. Fourth, the evaluation of SF-36 mental health relies on participants’ self-reports, which may be influenced by personal biases, emotional states, or their understanding of the questions. Additionally, participants may answer based on what they perceive as the ‘correct’ or ‘expected’ responses, potentially affecting the authenticity of the questionnaire. Fifth, the study cohort consisted of adults with overweight/obesity and type 2 diabetes who participated in a clinical trial focused on weight management. As such, the findings may not be generalizable to the broader population. Our results should be further validated in study populations derived from routinely collected clinical data or in future randomized controlled trials.
Conclusion
The current post-hoc analysis of the Look AHEAD trial demonstrated HTEs of ILI in adults with overweight/obesity and type 2 diabetes. The causal forest model has identified SF-36 mental health, diabetes duration, and ACR to assess the HTEs. Participants with better mental health, poorer renal function, and more cardiovascular risk factors and comorbidities were more likely to derive greater cardiovascular benefits from ILI. These hypothesis-generating findings highlight the cardiovascular benefits of ILI in specific populations and, although further validation is warranted, provide new evidence for the personalized management of adults with type 2 diabetes and overweight/obesity.
Study importance
What is already known?
The cardiovascular benefits of intensive lifestyle intervention (ILI) remain unclear in adults with type 2 diabetes and overweight/obesity.
A range of clinical characteristics may influence the cardiovascular benefits associated with ILI, and it is necessary to thoroughly explore the key factors affecting these benefits in order to identify populations who derive cardiovascular benefits from ILI.
What does this study add?
The heterogeneity of treatment effects (HTEs) of ILI was observed in adults with type 2 diabetes and overweight/obesity.
Participants with better mental health, poorer renal function, and more cardiovascular risk factors and comorbidities were more likely to derive greater cardiovascular benefits from ILI.
How might these results change the direction of research or the focus of clinical practice?
Individualized approaches based on mental health, renal function, and cardiovascular risk factors might optimize the cardiovascular benefits of ILI in adults with type 2 diabetes and overweight/obesity.
Future research should focus on identifying and targeting the key characteristics to enhance the efficacy of lifestyle interventions in high-risk populations.
Supplementary information
Acknowledgements
We thank all participants in this study.
Author contributions
XXF, LJJ, and HLX designed and conducted the research and wrote the manuscript. XXF, LJJ, and LMH conducted data management and statistical analysis. HLX, ZJT, WP, and GY provided constructive suggestions for further analysis of the data. All authors read and approved the final manuscript. LMH, ZXD, and LXX take responsibility for the integrity of the data and the accuracy of the data analysis.
Funding
This study was supported by the National Natural Science Foundation of China (82370358 to X.Liao; 82470361 to Y.Guo; 82400495 to M.Liu), Guangdong Basic and Applied Basic Research Foundation (2024A1515013234 to X.Liao; 2024A1515012356 to X.Zhuang; 2024A1515030241 to Y.Guo; 2022A1515111181 to M.Liu), Funding by Science and Technology Projects in Guangzhou (2023A04J2169 to Y.Guo) and Young Talent Support Project of Guangzhou Association for Science and Technology (QT2024-029 to Y.Guo). The Look AHEAD trial was conducted by the Look AHEAD Research Group and supported by the NIDDK, the National Institute of Nursing Research, the National Heart, Lung, and Blood Institute, the Office of Research on Women’s Health, the National Institute of Minority Health and Health Disparities, and the Centers for Disease Control and Prevention. The data from Look AHEAD were supplied by the NIDDK Central Repository. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Data availability
Data from the Look AHEAD study are available to all researchers upon application.
Competing interests
The authors declare no competing interests.
Ethics approval and consent to participate
The data used in this study were obtained from the LOOK AHEAD public database. These data have been previously de-identified and are available for research use under ethical approval from the Ottawa Health Science Network Research Ethics Board. Therefore, our study did not require additional ethical approval.
Footnotes
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
These authors contributed equally: Xingfeng Xu, Jiangjie Lei, Lixiang He.
Contributor Information
Menghui Liu, Email: liumh58@mail.sysu.edu.cn.
Xiaodong Zhuang, Email: zhuangxd3@mail.sysu.edu.cn.
Xinxue Liao, Email: liaoxinx@mail.sysu.edu.cn.
Supplementary information
The online version contains supplementary material available at https://doi.org/10.1038/s41387-026-00433-x.
References
- 1.Ng ACT, Delgado V, Borlaug BA, Bax JJ. Diabesity: the combined burden of obesity and diabetes on heart disease and the role of imaging. Nat Rev Cardiol. 2021;18:291–304. [DOI] [PubMed] [Google Scholar]
- 2.Wadden TA, Bailey TS, Billings LK, Davies M, Frias JP, Koroleva A, et al. Effect of subcutaneous semaglutide vs placebo as an adjunct to intensive behavioral therapy on body weight in adults with overweight or obesity: the step 3 randomized clinical trial. Jama. 2021;325:1403–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Solomon SD, Ostrominski JW, Wang X, Shah SJ, Borlaug BA, Butler J, et al. Effect of semaglutide on cardiac structure and function in patients with obesity-related heart failure. J Am Coll Cardiol. 2024;84:1587–602. [DOI] [PubMed] [Google Scholar]
- 4.Davies MJ, Aroda VR, Collins BS, Gabbay RA, Green J, Maruthur NM, et al. Management of hyperglycaemia in type 2 diabetes, 2022. A consensus report by the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD). Diabetologia. 2022;65:1925–66. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Wing RR, Bolin P, Brancati FL, Bray GA, Clark JM, Coday M, et al. Cardiovascular effects of intensive lifestyle intervention in type 2 diabetes. N Engl J Med. 2013;369:145–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.de Vries TI, Dorresteijn JAN, van der Graaf Y, Visseren FLJ, Westerink J. Heterogeneity of treatment effects from an intensive lifestyle weight loss intervention on cardiovascular events in patients with type 2 diabetes: data from the look AHEAD trial. Diab Care. 2019;42:1988–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Gregg EW, Jakicic JM, Blackburn G, Bloomquist P, Bray GA, Clark JM, et al. Association of the magnitude of weight loss and changes in physical fitness with long-term cardiovascular disease outcomes in overweight or obese people with type 2 diabetes: a post-hoc analysis of the Look AHEAD randomised clinical trial. Lancet Diab Endocrinol. 2016;4:913–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Liu M, Huang R, Xu L, Zhang S, Zhong X, Chen X, et al. Cardiovascular effects of intensive lifestyle intervention in adults with overweight/obesity and type 2 diabetes according to body weight time in range. EClinicalMedicine. 2022;49:101451. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Ryan DH, Espeland MA, Foster GD, Haffner SM, Hubbard VS, Johnson KC, et al. Look AHEAD (Action for Health in Diabetes): design and methods for a clinical trial of weight loss for the prevention of cardiovascular disease in type 2 diabetes. Control Clin Trials. 2003;24:610–28. [DOI] [PubMed] [Google Scholar]
- 10.Wadden TA, West DS, Delahanty L, Jakicic J, Rejeski J, Williamson D, et al. The Look AHEAD study: a description of the lifestyle intervention and the evidence supporting it. Obesity. 2006;14:737–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Athey S, Imbens G. Recursive partitioning for heterogeneous causal effects. Proc Natl Acad Sci USA. 2016;113:7353–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Wesche-Thobaben JA. The development and description of the comparison group in the Look AHEAD trial. Clin trials (Lond, Engl). 2011;8:320–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Buuren SV G-OK. MICE: Multivariate Imputation by Chained Equations in R. J Statistical Software. 2011;45.
- 14.Baum A, Scarpa J, Bruzelius E, Tamler R, Basu S, Faghmous J. Targeting weight loss interventions to reduce cardiovascular complications of type 2 diabetes: a machine learning-based post-hoc analysis of heterogeneous treatment effects in the Look AHEAD trial. Lancet Diab Endocrinol. 2017;5:808–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Bąk-Sosnowska M, Gruszczyńska M, Wyszomirska J, Daniel-Sielańczyk A. The influence of selected psychological factors on medication adherence in patients with chronic diseases. Healthcare. 2022;10. [DOI] [PMC free article] [PubMed]
- 16.Pacanowski CR, Linde JA, Faulconbridge LF, Coday M, Safford MM, Chen H, et al. Psychological status and weight variability over eight years: results from Look AHEAD. Health Psychol. 2018;37:238–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Zu C, Liu M, Su X, Wei Y, Meng Q, Liu C, et al. Association of body weight time in target range with the risk of kidney outcomes in patients with overweight/obesity and type 2 diabetes mellitus. Diab Care. 2024;47:371–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Wang Y, Li F, Chu C, Zhang X, Zhang XY, Liao YY, et al. Early life body mass index trajectories and albuminuria in midlife: A 30-year prospective cohort study. EClinicalMedicine. 2022;48:101420. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.KDIGO 2024 Clinical Practice Guideline for the Evaluation and Management of Chronic Kidney Disease. Kidney Int. 2024;105:S117-s314. [DOI] [PubMed]
- 20.Inoue K, Streja E, Tsujimoto T, Kobayashi H. Urinary albumin-to-creatinine ratio within normal range and all-cause or cardiovascular mortality among U.S. adults enrolled in the NHANES during 1999-2015. Ann Epidemiol. 2021;55:15–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Yang ZW, Fu YB, Wei XB, Fu BQ, Huang JL, Zhang GR, et al. Optimal threshold of urinary albumin-to-creatinine ratio (UACR) for predicting long-term cardiovascular and noncardiovascular mortality. Int Urol Nephrol. 2023;55:1811–9. [DOI] [PubMed] [Google Scholar]
- 22.Zeng C, Liu M, Zhang Y, Deng S, Xin Y, Hu X. Association of urine albumin to creatinine ratio with cardiovascular outcomes in patients with type 2 diabetes mellitus. J Clin Endocrinol Metab. 2024;109:1080–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Gao Z, Zhu Y, Sun X, Zhu H, Jiang W, Sun M, et al. Establishment and validation of the cut-off values of estimated glomerular filtration rate and urinary albumin-to-creatinine ratio for diabetic kidney disease: A multi-center, prospective cohort study. Front Endocrinol. 2022;13:1064665. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Niu J, Zhang X, Li M, Wu S, Zheng R, Chen L, et al. Risk of cardiovascular disease, death, and renal progression in diabetes according to albuminuria and estimated glomerular filtration rate. Diab Metab. 2023;49:101420. [DOI] [PubMed] [Google Scholar]
- 25.Ndumele CE, Neeland IJ, Tuttle KR, Chow SL, Mathew RO, Khan SS, et al. A synopsis of the evidence for the science and clinical management of cardiovascular-kidney-metabolic (CKM) syndrome: a scientific statement from the American Heart Association. Circulation. 2023;148:1636–64. [DOI] [PubMed] [Google Scholar]
- 26.Braunwald E. Gliflozins in the management of cardiovascular disease. N Engl J Med. 2022;386:2024–34. [DOI] [PubMed] [Google Scholar]
- 27.Gourdy P, Darmon P, Dievart F, Halimi JM, Guerci B. Combining glucagon-like peptide-1 receptor agonists (GLP-1RAs) and sodium-glucose cotransporter-2 inhibitors (SGLT2is) in patients with type 2 diabetes mellitus (T2DM). Cardiovasc Diabetol. 2023;22:79. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Fu EL, Kutz A, Desai RJ. Finerenone in chronic kidney disease and type 2 diabetes: the known and the unknown. Kidney Int. 2023;103:30–3. [DOI] [PubMed] [Google Scholar]
- 29.Garvey WT, Frias JP, Jastreboff AM, le Roux CW, Sattar N, Aizenberg D, et al. Tirzepatide once weekly for the treatment of obesity in people with type 2 diabetes (SURMOUNT-2): a double-blind, randomised, multicentre, placebo-controlled, phase 3 trial. Lancet. 2023;402:613–26. [DOI] [PubMed] [Google Scholar]
- 30.Wager S, Athey S. Estimation and inference of heterogeneous treatment effects using random forests. J Am Stat Assoc. 2018;113:1228–42. [Google Scholar]
- 31.Harrison TG, Tonelli M. Measuring albuminuria or proteinuria: does one answer fit all?. Kidney Int. 2023;104:904–9. [DOI] [PubMed] [Google Scholar]
- 32.Sallsten G, Barregard L. Variability of Urinary Creatinine in Healthy Individuals. Int J Environ Res Public Health. 2021;18. [DOI] [PMC free article] [PubMed]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data from the Look AHEAD study are available to all researchers upon application.