Longitudinal body mass index exposure-based threshold to prevent mortality for populations with type 2 diabetes mellitus

Summary

The long-term body mass index (BMI) threshold to prevent mortality among aging individuals with type 2 diabetes mellitus (T2DM) remains unclear. We quantify BMI exposure using two metrics: the percentage of time BMI above target range (TAR) and the percentage of time BMI within target range (TTR). In a cohort of 3,708 adults aged ≥40 years with T2DM and at least 5 BMI measurements across 4 years, 1,020 deaths occurred during a median 5-year follow-up. Sustained BMI ≥27 kg/m² (TAR) is positively associated with an increased risk of mortality. Furthermore, longer sustenance of BMI within the range of 18.5–26.9 kg/m² (TTR) is associated with dose-response survival benefits for all-cause, cardiovascular, and cancer mortality, respectively (all p < 0.05). The association is further validated in the China Health and Retirement Longitudinal Study cohort. These findings support that the threshold for increased mortality risk is sustained exposure to BMI ≥27 kg/m² in middle-aged and elderly adults with T2DM.

Graphical abstract

Highlights

• •Longitudinal BMI exposure is quantified by TAR and TTR
• •Sustained BMI ≥27 kg/m² increases mortality risk in Chinese individuals with T2DM
• •BMI range of 18.5–26.9 kg/m² is optimal for Chinese individuals with T2DM

Hu et al. show that the “healthy” body mass index for middle-aged and elderly individuals with type 2 diabetes needs to be refined. The threshold for increased mortality risk is sustained exposure to a body mass index ≥27 kg/m².

Introduction

The rising prevalence of type 2 diabetes mellitus (T2DM) in aging populations poses a critical challenge to global health systems, with adults ≥40 years accounting for more than 65% of cases worldwide.¹ The particular concern is the accelerating burden in Asia, where the aging rate is outpacing the capacity of the healthcare system to respond.² Consequently, the number of T2DM-related deaths in Asia has exceeded 2 million annually.^1,3 While glycemic control remains essential, international guidelines and consensus statements^4,5,6 have highly recommended “comprehensive metabolic management,” particularly emphasizing sustained weight management as a primary treatment target.

Obesity is a major driver of T2DM.⁷ However, the relationship between body mass index (BMI) and mortality in individuals with established diabetes is complex.^8,9 Recent studies have reported that, with increasing age, the upper threshold of BMI related to mortality risk shifts upward compared with traditional thresholds.¹⁰ In response, current national and international guidelines^5,11,12 recommend adjusting BMI thresholds based on individual characteristics for aging adults with T2DM, but critical gaps persist. First, no validated BMI thresholds have been established, and existing evidence regarding BMI thresholds is mainly derived from Western populations, which may not be applicable to Asian populations due to ethnic and physiological differences.¹³ For Asia, where 60% of global diabetes-related deaths occur,¹ defining an ethnic-specific BMI threshold is essential for ensuring equitable and effective care. Second, existing evidence has predominantly relied on single time point BMI assessment, failing to capture dynamic weight changes over time. This neglects clinically critical patterns such as weight gain, weight loss, and weight regain, potentially biasing risk estimates.

To address these gaps and align with the imperative of sustained weight control, we introduced BMI time above target range (TAR) and time in target range (TTR), adapted from glucose management, to quantify weight exposure over time.¹⁴ Using high-quality electronic medical record (EHR) data^15,16 containing 20,624 longitudinal BMI measurements (≥5 measurements per person) across the first 4 years, along with government-validated mortality data, we aimed to (1) determine the optimal mortality-related BMI threshold using BMI TAR in middle-aged and elderly individuals with T2DM and (2) validate whether maintaining BMI within the derived target range using TTR conferred survival benefits.

Results

Clinical characteristics of the study population

A total of 3,708 individuals aged 40–99 years (median age: 67 [60–76]) with T2DM were included (Figure 1). Among these individuals, 58.3% were men. The median duration of diabetes was 10 (5–17) years. The median baseline BMI was 24.5 (22.3–27.2) kg/m². These individuals had 20,624 BMI measurements during the first 4 years, with an average of 6 BMI measurements per person. As shown in Table 1, compared with individuals with lower BMI TTR (0%), those with BMI TTR 100% were more likely to be men, had higher estimated glomerular filtration rate (eGFR), greater use of lipid-lowering medications, and a higher prevalence of smoking (all p < 0.05). Conversely, they had lower triglyceride (TG) and hemoglobin A1c (HbA1c), showed a lower proportion of antihypertensive and glucose-lowering medication use, and exhibited a lower prevalence of chronic kidney disease and cancer (all p < 0.05). Baseline characteristics stratified by outcome status (all-cause, cardiovascular, or cancer mortality) are summarized in Tables S1–S3. In general, individuals who died were older and had a higher burden of comorbidities compared with those who survived.

Figure 1.

The study design and flow chart

Table 1.

Baseline clinical characteristics according to categories of BMI TTR during the first 4 years

Variables	BMI TTR, %				p value
	0 (N = 651)	0.1–62.7 (N = 579)	62.8–99.9 (N = 579)	100 (N = 1899)
Age, years	66 (59–75)	68 (59–78)	68 (61–77)	67 (59–75)	0.043
Men/women, n	332/319	329/250	346/233	1153/746	<0.001
Duration of diabetes, years	10.0 (5.00–17.0)	10.0 (5.00–18.0)	10.0 (5.00–16.0)	10.0 (5.00–16.5)	0.724
BMI, kg/m2	29.1 (27.7–31.0)	26.5 (22.5–28.4)	25.6 (22.6–27.5)	23.6 (22.0–24.8)	<0.001
Systolic blood pressure, mmHg	127 (120–137)	126 (120–136)	127 (120–135)	127 (120–136)	0.291
Diastolic blood pressure, mmHg	72 (67–80)	71 (67–78)	71 (66–77)	71 (66–77)	0.054
Total cholesterol, mmol/L	4.73 (4.27–5.44)	4.68 (4.26–5.39)	4.68 (4.18–5.44)	4.69 (4.19–5.54)	0.871
Triglyceride, mmol/L	1.84 (1.44–2.47)	1.74 (1.38–2.29)	1.80 (1.41–2.36)	1.74 (1.36–2.32)	0.005
Low-density lipoprotein cholesterol, mmol/L	2.97 (2.50–3.74)	3.07 (2.43–3.71)	3.06 (2.51–3.78)	3.01 (2.46–3.64)	0.690
High-density lipoprotein cholesterol, mmol/L	1.03 (0.88–1.25)	1.06 (0.89–1.27)	1.08 (0.91–1.26)	1.09 (0.92–1.28)	0.057
Fasting plasma glucose, mmol/L	7.70 (6.39–9.37)	7.62 (6.42–9.35)	7.53 (6.35–9.59)	7.41 (6.37–9.10)	0.201
HbA1c, %	8.5 (7.6–9.9)	8.3 (7.4–9.5)	8.3 (7.5–9.7)	8.3 (7.4–9.6)	0.004
Fasting C-peptide, ng/mL	1.88 (1.15–2.82)	1.94 (1.15–2.80)	1.86 (1.14–2.78)	1.70 (0.98–2.60)	<0.001
Fasting insulin, μU/mL	9.42 (3.70–19.0)	9.44 (3.46–18.5)	8.19 (3.00–17.5)	7.88 (2.86–16.4)	0.004
eGFR, mL/min/1.73 m2	87.4 (67.4–99.4)	84.0 (60.7–97.7)	88.1 (68.2–98.2)	88.7 (72.1–98.8)	0.002
Current smoker, n (%)	182 (28.0)	199 (34.4)	219 (37.8)	632 (33.3)	0.003
Current drinker, n (%)	119 (18.3)	106 (18.3)	110 (19.0)	345 (18.2)	0.976
Insurance type, n (%)	–	–	–	–	0.064
Insured	572 (87.9)	524 (90.5)	518 (89.5)	1648 (86.8)	–
Uninsured	79 (12.1)	55 (9.50)	61 (10.5)	251 (13.2)	–
Family history of diabetes, n (%)	139 (21.4)	104 (18.0)	116 (20.0)	385 (20.3)	0.509
Use of medications, n (%)	–	–	–	–	–
Lipid-lowering medication	286 (43.9)	261 (45.1)	266 (45.9)	961 (50.6)	0.006
Antihypertensive medication	565 (86.8)	497 (85.8)	489 (84.5)	1,566 (82.5)	0.034
Glucose-lowering medication	630 (96.8)	548 (94.6)	527 (91.0)	1,735 (91.4)	<0.001
Other comorbidities at baseline, n (%)	–	–	–	–	–
Ischemic heart diseases	150 (23.0)	134 (23.1)	104 (18.0)	437 (23.0)	0.063
Heart failure	95 (14.6)	90 (15.5)	76 (13.1)	224 (11.8)	0.067
Cerebrovascular disease	79 (12.1)	60 (10.4)	85 (14.7)	270 (14.2)	0.059
Chronic kidney disease	44 (6.76)	58 (10.0)	39 (6.74)	117 (6.16)	0.016
Cancer	112 (17.2)	91 (15.7)	78 (13.5)	249 (13.1)	0.047

BMI, body mass index; TTR, time in target range; HbA1c, hemoglobin A1c; eGFR, estimated glomerular filtration rate.

Association between different predefined BMI TARs and mortality risk

During a median follow-up of 5.0 years, 1,020 individuals died (57.6 per 1,000 person-years), among whom 260 died due to cardiovascular diseases (14.7 per 1,000 person-years) and 448 died due to cancer (25.3 per 1,000 person-years). As shown in Figure 2, multivariable Cox proportional hazards analysis demonstrated that TARs with BMI thresholds of 24–25 kg/m² (TAR_≥24 to TAR_≥25) were inversely associated with all-cause mortality risk, whereas TARs with BMI thresholds of 27 kg/m² or higher (TAR_≥27 to TAR_≥30) were positively associated with all-cause mortality risk. Furthermore, TAR with a BMI threshold of 26 kg/m² was not significantly associated with all-cause mortality risk (hazard ratio [HR], 1.01; 95% confidence interval [CI], 0.99–1.02). Similarly, TARs with a BMI threshold of 27 kg/m² or higher (TAR_≥27 to TAR_≥30) were positively associated with cardiovascular and cancer mortality, respectively (Figure 2). Subgroup analyses stratified by age, HbA1c levels, history of cardiovascular disease, and history of cancer consistently demonstrated that BMI TARs with an upper threshold of 27 kg/m² or above were significantly associated with an elevated risk of all-cause mortality (Figure S1; all p < 0.05).

Figure 2.

The association of TARs using different BMI thresholds with all-cause, cardiovascular, and cancer mortality

The model was adjusted for gender, age, smoking status, alcohol status, insurance type, diabetes duration, hemoglobin A1c, systolic blood pressure, low-density lipoprotein cholesterol, estimated glomerular filtration rate, history of cardiovascular diseases, history of cancer, use of antihypertensive medication, use of glucose-lowering medication, and use of lipid-lowering medication. Data are represented as HR (95% CI) for mortality by each 10% increase in different predefined TARs. HR, hazard ratio; CI, confidence interval.

Adding BMI TAR to the conventional model significantly improved the predictive value for all-cause, cardiovascular, and cancer mortality (Figure S2; all p < 0.001). Furthermore, BMI TAR significantly enhanced both integrated discrimination improvement (IDI) and net reclassification improvement (NRI) for all-cause, cardiovascular, and cancer mortality (Figure S2; all p < 0.001).

Spearman correlation analyses showed that BMI TAR_≥27 was positively correlated with diastolic blood pressure (Table S4; r = 0.040), HbA1c (Table S4; r = 0.038), and TG (Table S4; r = 0.100), while negatively correlated with high-density lipoprotein cholesterol (HDL-C) (Table S4; r = −0.080). BMI TAR_≥27 was significantly correlated with baseline BMI and the mean and variability of BMI during follow-up (Figure S3; all p < 0.001).

Association between BMI TTR and mortality risk

When BMI within 18.5–26.9 kg/m² was set as the target range, restricted cubic spline analysis revealed a nonlinear relationship between BMI TTR and all-cause mortality (Figure 3A; p_{non-linearity} = 0.011), with an inflection point at 31%. Beyond this inflection point, an increase in BMI TTR was associated with a sharper decline in mortality risk. As shown in Figure 4A, multivariable-adjusted HRs (age, gender, smoking status, alcohol status, and insurance type, model 2) for all-cause mortality associated with BMI TTR levels of 100% (reference group), 62.8%–99.9%, 0.1%–62.7%, and 0% were 1.00, 1.39 (95% CI 1.15–1.68), 2.06 (95% CI 1.74–2.43), and 2.43 (95% CI 2.08–2.85), respectively. After further adjustment for diabetes duration, HbA1c, systolic blood pressure, low-density lipoprotein cholesterol (LDL-C), eGFR, history of cardiovascular diseases, history of cancer, use of antihypertensive medication, use of glucose-lowering medication, and use of lipid-lowering medication (model 3), this inverse association did not change. The restricted cubic spline curve showed that BMI TTR was linearly associated with cardiovascular mortality (Figure 3A; p_{non-linearity} = 0.900), while non-linearly associated with cancer mortality (Figure 3A; p_{non-linearity} = 0.018). The inflection point of BMI TTR for cancer mortality risk was 54%. Cox proportional hazards regression analysis suggested that multivariable-adjusted (model 3) HRs associated with BMI TTR levels of 100% (reference group), 62.8%–99.9%, 0.1%–62.7%, and 0% were 1.00, 1.13 (95% CI 0.77–1.65), 1.67 (95% CI 1.19–2.33), and 2.38 (95% CI 1.74–3.27) for cardiovascular mortality, and 1.00, 1.09 (95% CI 0.81–1.47), 2.02 (95% CI 1.57–2.60), and 2.12 (95% CI 1.67–2.69) for cancer mortality, respectively (Figures 4B and 4C).

Figure 3.

The association of BMI TTR with the risks of all-cause, cardiovascular, and cancer mortality

(A) Dose-response associations of BMI TTR with all-cause, cardiovascular, and cancer mortality using restricted cubic spline analysis nested in time-dependent Cox models.

(B) Comparison of the predictive performance of BMI TTR for mortality with the conventional model using ROC curves.

(C) Comparison of the predictive performance of BMI TTR for mortality with the conventional model using IDI and NRI. Data are represented as IDI (95% CI) and NRI (95% CI). The conventional model included age, diastolic blood pressure, use of antihypertensive medication, insulin therapy, triglycerides, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, urine albumin-to-creatinine ratio, and BMI. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. ROC, receiver operating characteristic; IDI, integrated discrimination improvement; NRI, net reclassification improvement.

Figure 4.

The association of BMI TTR with the risks of all-cause, cardiovascular, and cancer mortality

The association of BMI TTR with the risks of all-cause (A), cardiovascular (B), and cancer mortality (C) using time-dependent Cox regression models. Model 1 was adjusted for gender and baseline age; model 2 was adjusted for gender, baseline age, smoking status, alcohol status, and insurance type; and model 3 was adjusted for gender, baseline age, smoking status, alcohol status, insurance type, diabetes duration, hemoglobin A1c, systolic blood pressure, low-density lipoprotein cholesterol, estimated glomerular filtration rate, history of cardiovascular diseases, history of cancer, use of antihypertensive medication, use of glucose-lowering medication, and use of lipid-lowering medication. Data are represented as HR (95% CI). HR, hazard ratio; CI, confidence interval.

The predictive values of different lower BMI thresholds in TTR for mortality were compared. The results indicated that the predictive performance was comparable for BMI target ranges of 18.5–26.9, 20.0–26.9, and 22.0–26.9 kg/m². The BMI target range of 18.5–26.9 kg/m² outperformed the narrower target range of 24.0–26.9 kg/m² in predicting all-cause and cancer mortality (Figure S4). BMI TTR of 18.5–26.9 kg/m² provided significantly better prediction than the current Chinese guideline-recommended BMI target of 18.5–23.9 kg/m² (Figure S4).

Subgroup analyses further demonstrated that (1) age and HbA1c interacted with the association between BMI TTR and all-cause mortality (Table 2; both p_interaction < 0.05), (2) eGFR modified the association between BMI TTR and cardiovascular mortality (Table 2; p_interaction = 0.004), and (3) HbA1c interacted with the association between BMI TTR and cancer mortality (Table 2; p_interaction = 0.013). To further validate the findings in different subcohorts, four sensitivity analyses were performed. When current and past smokers were excluded (n = 1,232), the findings were not materially changed (Table S5). Consistently, after excluding individuals who died within 1 year since baseline (n = 158), we still observed a significant association between BMI TTR and mortality (Table S5). When further adjusted for baseline BMI and the mean and coefficient of variation (CV) of BMI during follow-up, the association between BMI TTR and mortality persisted (Table S5). After excluding individuals with severe comorbidities (cancer, heart failure, end-stage renal disease, or cachexia) (n = 1,259), the association remained significant (Table S5).

Table 2.

Subgroup analysis of the association between BMI TTR and all-cause, cardiovascular, and cancer mortality according to different baseline characteristics

	BMI TTR, %				pinteraction
	100	62.8–99.9	0.1–62.7	0
All-cause mortality
Age, years	–	–	–	–	0.010
<65	1.00 (reference)	1.43 (0.98–2.09)	2.75 (2.01–3.76)	2.69 (1.99–3.63)	–
≥65	1.00 (reference)	1.28 (1.03–1.59)	1.62 (1.32–1.99)	2.20 (1.83–2.66)	–
Gender	–	–	–	–	0.202
Men	1.00 (reference)	1.42 (1.13–1.78)	1.82 (1.46–2.27)	2.11 (1.71–2.59)	–
Women	1.00 (reference)	1.22 (0.87–1.71)	2.04 (1.55–2.68)	2.71 (2.10–3.50)	–
HbA1c, %	–	–	–	–	0.008
<8.0	1.00 (reference)	0.97 (0.71–1.34)	1.73 (1.33–2.25)	2.21 (1.70–2.88)	–
≥8.0	1.00 (reference)	1.61 (1.27–2.04)	1.92 (1.53–2.40)	2.41 (1.97–2.94)	–
eGFR, mL/min/1.73 m2	–	–	–	–	0.693
<90	1.00 (reference)	1.31 (1.05–1.65)	1.71 (1.40–2.10)	2.15 (1.77–2.61)	–
≥90	1.00 (reference)	1.25 (0.88–1.77)	2.39 (1.75–3.27)	2.43 (1.84–3.19)	–
Smoking status	–	–	–	–	0.336
Current and past	1.00 (reference)	1.29 (0.96–1.75)	1.76 (1.31–2.35)	1.98 (1.48–2.65)	–
Never	1.00 (reference)	1.28 (1.00–1.63)	1.93 (1.57–2.38)	2.52 (2.08–3.06)	–
Use of lipid-lowering medication	–	–	–	–	0.122
Yes	1.00 (reference)	0.95 (0.68–1.33)	1.63 (1.22–2.16)	2.27 (1.76–2.93)	–
No	1.00 (reference)	1.58 (1.25–1.99)	2.03 (1.64–2.50)	2.28 (1.86–2.79)	–
Use of antihypertensive medication	–	–	–	–	0.355
Yes	1.00 (reference)	1.33 (1.09–1.63)	1.96 (1.64–2.34)	2.29 (1.94–2.72)	–
No	1.00 (reference)	1.20 (0.68–2.10)	1.25 (0.72–2.15)	2.34 (1.44–3.78)	–
Use of glucose-lowering medication	–	–	–	–	0.131
Yes	1.00 (reference)	1.22 (0.99–1.50)	1.87 (1.57–2.23)	2.27 (1.93–2.68)	–
No	1.00 (reference)	2.29 (1.37–3.84)	1.51 (0.82–2.78)	2.62 (1.28–5.37)	–
Cardiovascular mortality
Age, years	–	–	–	–	0.822
<65	1.00 (reference)	1.27 (0.48–3.37)	1.94 (0.81–4.65)	2.68 (1.29–5.55)	–
≥65	1.00 (reference)	1.09 (0.72–1.66)	1.60 (1.11–2.31)	2.30 (1.62–3.27)	–
Gender	–	–	–	–	0.607
Men	1.00 (reference)	1.16 (0.71–1.89)	1.81 (1.17–2.80)	2.04 (1.33–3.12)	–
Women	1.00 (reference)	1.15 (0.61–2.15)	1.55 (0.90–2.67)	2.85 (1.74–4.68)	–
HbA1c, %	–	–	–	–	0.446
<8.0	1.00 (reference)	1.01 (0.54–1.91)	2.01 (1.23–3.29)	2.61 (1.53–4.44)	–
≥8.0	1.00 (reference)	1.21 (0.75–1.95)	1.36 (0.86–2.18)	2.16 (1.45–3.22)	–
eGFR, mL/min/1.73 m2	–	–	–	–	0.003
<90	1.00 (reference)	0.99 (0.65–1.51)	1.62 (1.13–2.31)	2.35 (1.67–3.29)
≥90	1.00 (reference)	2.31 (0.92–5.82)	1.75 (0.60–5.11)	2.72 (1.09–6.81)	–
Smoking status	–	–	–	–	0.403
Current and past	1.00 (reference)	1.56 (0.86–2.83)	1.71 (0.94–3.12)	1.85 (1.01–3.36)	–
Never	1.00 (reference)	0.91 (0.55–1.50)	1.62 (1.08–2.44)	2.56 (1.76–3.74)	–
Use of lipid-lowering medication					0.314
Yes	1.00 (reference)	0.94 (0.53–1.65)	1.65 (1.04–2.64)	2.82 (1.86–4.27)	–
No	1.00 (reference)	1.27 (0.75–2.15)	1.56 (0.96–2.55)	1.91 (1.17–3.13)	–
Use of antihypertensive medication	–	–	–	–	0.105
Yes	1.00 (reference)	1.01 (0.67–1.52)	1.74 (1.23–2.46)	2.25 (1.61–3.13)	–
No	1.00 (reference)	2.62 (0.89–7.74)	0.77 (0.19–3.19)	4.14 (1.38–12.4)	–
Use of glucose-lowering medication	–	–	–	–	0.447
Yes	1.00 (reference)	0.97 (0.64–1.46)	1.69 (1.20–2.40)	2.34 (1.69–3.23)	–
No	1.00 (reference)	5.39 (1.52–19.2)	0.82 (0.18–3.73)	3.74 (0.57–24.5)	–
Cancer mortality
Age, years	–	–	–	–	0.057
<65	1.00 (reference)	1.24 (0.73–2.09)	3.20 (2.11–4.83)	2.63 (1.76–3.93)	–
≥65	1.00 (reference)	1.08 (0.75–1.56)	1.70 (1.23–2.35)	2.00 (1.48–2.70)	–
Gender	–	–	–	–	0.984
Men	1.00 (reference)	1.12 (0.78–1.61)	2.12 (1.54–2.92)	2.24 (1.66–3.02)	–
Women	1.00 (reference)	1.07 (0.63–1.83)	1.99 (1.29–3.04)	2.14 (1.44–3.17)	–
HbA1c, %	–	–	–	–	0.013
<8.0	1.00 (reference)	0.75 (0.44–1.28)	1.79 (1.18–2.74)	2.00 (1.34–2.98)	–
≥8.0	1.00 (reference)	1.38 (0.95–2.00)	2.18 (1.57–3.01)	2.18 (1.61–2.93)	–
eGFR, mL/min/1.73 m2	–	–	–	–	0.454
<90	1.00 (reference)	1.17 (0.77–1.79)	1.78 (1.25–2.54)	2.06 (1.46–2.91)	–
≥90	1.00 (reference)	1.09 (0.71–1.68)	2.43 (1.69–3.50)	2.18 (1.57–3.02)	–
Smoking status	–	–	–	–	0.696
Current and past	1.00 (reference)	1.09 (0.70–1.72)	1.99 (1.31–3.02)	1.85 (1.22–2.83)	–
Never	1.00 (reference)	1.03 (0.68–1.56)	2.14 (1.55–2.95)	2.32 (1.74–3.11)	–
Use of lipid-lowering medication	–	–	–	–	0.351
Yes	1.00 (reference)	0.74 (0.41–1.35)	1.61 (0.98–2.66)	1.89 (1.22–2.94)	–
No	1.00 (reference)	1.30 (0.92–1.85)	2.32 (1.72–3.13)	2.25 (1.70–3.00)	–
Use of antihypertensive medication	–	–	–	–	0.649
Yes	1.00 (reference)	1.17 (0.85–1.60)	2.10 (1.60–2.75)	2.13 (1.64–2.75)	–
No	1.00 (reference)	0.46 (0.15–1.35)	1.54 (0.75–3.16)	2.15 (1.13–4.10)	–
Use of glucose-lowering medication	–	–	–	–	0.583
Yes	1.00 (reference)	1.00 (0.72–1.40)	1.99 (1.52–2.60)	2.12 (1.65–2.71)	–
No	1.00 (reference)	1.30 (0.59–2.85)	1.92 (0.83–4.44)	1.49 (0.53–4.17)	–

Data are presented as HR (95% CI). Multivariable-adjusted models included gender, baseline age, smoking status, alcohol status, insurance type, duration of diabetes, hemoglobin A1c, systolic blood pressure, low-density lipoprotein cholesterol, estimated glomerular filtration rate, history of cardiovascular diseases, history of cancer, use of antihypertensive medication, use of lipid-lowering medication, and use of glucose-lowering medication. HR, hazard ratio; CI, confidence interval; BMI, body mass index; eGFR, estimated glomerular filtration rate.

Spearman correlation analyses revealed that BMI TTR_18.5–26.9 was positively correlated with HDL-C (Table S4; r = 0.044) and eGFR (Table S4; r = 0.042), while negatively correlated with diastolic blood pressure (Table S4; r = −0.042), fasting plasma glucose (Table S4; r = −0.034), HbA1c (Table S4; r = −0.046), and TG (Table S4; r = −0.043). Moreover, as shown in Figure S3, BMI TTR was inversely correlated with the mean and variability of BMI (all p < 0.001).

Mean BMI was inversely associated with the risk of all-cause and cancer mortality (Table S6). The CV of BMI was significantly and positively associated with all-cause mortality (Table S6). However, the association of BMI variability metrics with cardiovascular and cancer mortality did not reach statistical significance (Table S6).

The additional predictive performance of BMI TTR

As shown in Figure 3B, incorporating BMI TTR into the conventional model significantly enhanced the predictive performance for mortality. For all-cause mortality, the area under the curve (AUC) increased from 0.683 (95% CI 0.668–0.698) in the conventional model to 0.742 (95% CI 0.727–0.756) after adding BMI TTR (p < 0.001). Similar improvements were observed for cardiovascular mortality (AUC: 0.751 [95% CI 0.737–0.765] vs. 0.768 [95% CI 0.754–0.781], p < 0.001) and cancer mortality (AUC: 0.595 [95% CI 0.579–0.611] vs. 0.662 [95% CI 0.647–0.678], p < 0.001). Moreover, adding BMI TTR to the conventional model yielded significant increases in IDI and NRI for all-cause, cardiovascular, and cancer mortality, respectively (Figure 3C; all p < 0.001).

External validation

To further test the generalizability of our findings, we performed external validation using the China Health and Retirement Longitudinal Study (CHARLS) cohort. The baseline characteristics were shown in Table S7. During a mean follow-up of 9 years, 112 individuals (11.3%) with DM died. Consistent with our primary analysis, compared with those with BMI TTR_18.5–26.9 of 100%, individuals with BMI TTR_18.5–26.9 of 0% had a 59% higher risk of mortality (Figure 5B; model 3: risk ratio [RR] 1.59, 95% CI 1.02–2.47), after adjusting for demographic characteristics, lifestyle habits, and clinical risk factors.

Figure 5.

The association between BMI TTR and all-cause mortality in an external cohort

(A) General study design.

(B) The association between BMI TTR and all-cause mortality using modified Poisson regression models. Model 1 was adjusted for gender and baseline age; model 2 was adjusted for gender, baseline age, smoking status, alcohol status, socioeconomic status score, and physical activity level; and model 3 was adjusted for gender, baseline age, smoking status, alcohol status, socioeconomic status score, physical activity level, diabetes duration, hemoglobin A1c, systolic blood pressure, low-density lipoprotein cholesterol, estimated glomerular filtration rate, history of cardiovascular diseases, history of cancer, use of antihypertensive medication, use of glucose-lowering medication, and use of lipid-lowering medication. Data are represented as RR (95% CI). RR, risk ratio; CI, confidence interval.

Discussion

Using EHRs, including 20,624 BMI measurements within the first 4 years, and government-validated mortality data from 3,708 middle-aged and elderly individuals with T2DM, we established a mortality-related BMI threshold using TAR and TTR. Our study demonstrated that cumulative exposure to BMI ≥27 kg/m² (TAR) was positively associated with an increased risk of all-cause, cardiovascular, and cancer mortality among individuals with T2DM. Furthermore, maintaining a BMI within 18.5–26.9 kg/m² (TTR) conferred dose-response survival benefits. Our findings address the ongoing debate around BMI thresholds for long-term weight control among middle-aged and elderly individuals with T2DM and provide evidence to support the concept of “sustained weight management” as a potential strategy in diabetes care.¹⁷

National and international guidelines have increasingly emphasized the importance of long-term weight management in diabetes care.^1,4,5 However, the appropriateness of traditional BMI thresholds for long-term weight control in middle-aged and elderly Asian individuals with T2DM is controversial and remains a subject of debate owing to important gaps in evidence. Prior studies that attempted to identify BMI thresholds for this population have largely relied on single time point BMI measurement,^{9,18,19,20,21} self-reported BMI data,¹⁹ and cohorts predominantly comprising Western populations,^{9,18,19,20,21} limiting their application to the Asian population, which accounts for 65% of global cases.¹ To address this unmet need, we introduced two dynamic metrics, BMI TAR and TTR, to estimate long-term weight management thresholds for mortality in middle-aged and elderly Asian individuals with T2DM. TAR quantifies exposure to an elevated BMI beyond a threshold and helps determine when to initiate intervention, while TTR evaluates the efficacy of maintaining target range. Given the epidemic of obesity in Asia and the substantial mortality burden associated with excess adiposity,^22,23 defining an actionable upper threshold is a pressing clinical and public health priority, whereas the lower threshold of BMI is well established and less debated.²⁴ Therefore, we specifically focused on identifying an upper BMI threshold for clinical intervention.

Previous studies using single time point BMI demonstrated that overweight was inversely associated, while obesity was positively associated with mortality risk.^10,25 However, this approach does not account for dynamic weight fluctuations or cumulative exposure over time, potentially masking critical dose-response relationships and leading to oversimplified or misleading conclusions about the long-term impact of BMI on mortality. Using longitudinal weight records, our results extend these findings and show that cumulative exposure to BMI ≥24 kg/m² was inversely associated with mortality risk, whereas sustained exposure to 27 kg/m² or higher was positively associated with mortality risk. Of note, maintaining BMI within 18.5–26.9 kg/m² was associated with a reduced risk of mortality, providing essential context for interpreting the results of the LOOK AHEAD trial.²⁶ Although the trial reported that initial weight loss failed to confer survival benefits after a median 9.6-year follow-up in individuals with T2DM, its post-hoc analysis demonstrated that TTR, defined as sustained weight loss ≥7% of baseline, was significantly associated with a reduced risk of cardiovascular events and composite kidney outcomes.^27,28 This discrepancy may reflect challenges in the long-term maintenance of weight loss rather than the ineffectiveness of weight reduction. Subsequent studies have focused on the association between body weight variability (e.g., standard deviation and CV) and mortality,^29,30 but lacked actionable thresholds for clinical practice. Our study directly addressed this gap by defining BMI TTR as a dynamic monitoring tool for long-term weight management, shifting the focus from a static BMI threshold to sustained maintenance within a target range. Moreover, incorporating TTR into the conventional model enhanced the predictive performance for mortality. This improvement was observed beyond conventional models that already incorporated precise laboratory examinations. This advantage is particularly important in resource-limited settings, where advanced testing may not be feasible. Therefore, the enhancement in AUC is not only statistically significant but also clinically actionable and valuable. Taken together, our study underscores the importance of long-term weight maintenance within a healthy range. Modest fluctuations are acceptable, as long as patients avoid sustained deviations outside the recommended range.

Subgroup analyses showed no significant interactions with the use of glucose-lowering, lipid-lowering, and anti-hypertensive medications, suggesting that the survival benefits of maintaining BMI within target range were consistent and significant regardless of medication use. Notably, we observed a significant interaction between age and BMI TTR for all-cause mortality. The association was stronger among individuals aged <65 years, although it remained statistically significant in both subgroups. This pattern suggests that sustained weight management confers benefits at any age, but earlier and longer maintenance within the BMI target range may yield greater survival advantages. Moreover, the associations between BMI TTR and both all-cause and cancer mortality were more pronounced among those with higher HbA1c. Previous studies have demonstrated that chronic hyperglycemia accelerates oxidative stress, advanced glycation end-product formation, and systemic inflammation, all of which contribute to tissue catabolism, frailty, and increased mortality risk.³¹ Poor glycemic control is also strongly linked to cancer development and progression, in part through local immunosuppression, pro-inflammatory cytokines, and enhanced tumor cell proliferation.³² In addition, the association between BMI TTR and cardiovascular mortality was more evident in individuals with higher eGFR but remained significant across subgroups. In advanced chronic kidney disease (CKD), CKD-specific processes (e.g., inflammation, vascular calcification, anemia, toxin burden, and uremic toxins) can increase the risk of cardiovascular events, potentially overshadowing BMI-related pathways.³³ This highlights the interplay between cardiovascular, kidney, and metabolic systems and the importance of co-management. Collectively, these findings indicate that the weight control target for diabetes care is not “one size fits all.” The upper threshold of BMI may vary across subpopulations defined by different clinical characteristics or outcomes, warranting further investigation.

There are several potential explanations for the protective role of BMI within 24.0–26.9 kg/m². It is found that modest excess adiposity may provide a metabolic reserve for middle-aged and elderly individuals. It upregulates protective adipokine profiles, enhances endotoxin-lipoprotein interaction, and sequesters toxins, collectively counteracting frailty, malnutrition, and osteoporosis.^34,35,36 Consequently, long-term weight targets should be set with greater caution in the aging population. Future studies on the association between body composition and mortality risk may help to elucidate this phenomenon. Nevertheless, once BMI reaches the upper threshold, the adverse consequences of excess adiposity may surpass these protective factors. First, higher BMI, particularly in Asian populations, is accompanied by visceral and ectopic fat, which aggravates insulin resistance, promotes hepatic steatosis, and accelerates atherogenic dyslipidemia, collectively elevating cardiovascular and cancer risks.³⁷ Second, higher adiposity may promote the activation of inflammatory pathways and adipokines, including tumor necrosis factor-α, interleukin-6, and interleukin-1, which upregulate inhibitor of kappa B kinase (IKK)/nuclear factor kappa-B (NF-κB) and mitogen-activated protein kinase (MAPK) pathways, inducing cell apoptosis and ultimately contributing to metabolic disorders.³⁸ The activation of oxidative stress is also involved in this progression.³⁹

Improving and optimizing access to weight management services and therapies for Asian populations could mitigate the diabetes epidemic, particularly given the heavy burden of diabetes-related mortality in this region.¹ However, access to weight management services in clinical practice is largely determined by BMI cutoffs, which, as our study shows, are unsuitable for mortality risk stratification in middle-aged and elderly individuals with T2DM. Moreover, our study indicated that maintaining BMI within 18.5–26.9 kg/m² over time could confer survival benefits, based on longitudinal weight measurements. Collectively, these findings underscore that, while BMI remains a simple, valuable, and cost-effective tool for population-level screening, its thresholds demand recalibration to align with ethnic-specific pathophysiology and aging-related metabolic shifts. Meanwhile, BMI thresholds must evolve from rigid cutoffs to dynamic monitoring tools. Without these changes, health systems, especially in low- and middle-income countries, will continue to misallocate resources by overtreating low-risk individuals. It is therefore crucial to revise weight thresholds for the Asian population.

The primary strength of our study is the high-quality database, which includes internationally standardized laboratory parameters, government-validated mortality data, and ≥5 BMI measurements per person within 4 years. To the best of our knowledge, few large-scale databases contain such high-frequency longitudinal BMI measurements per person as those used in our study. These high-quality data enabled us to quantify cumulative adiposity exposure and estimate BMI thresholds, addressing the limitations of single time point assessments that overlook weight fluctuation patterns. The second strength is the special focus on mortality-associated BMI thresholds. By directly linking longitudinal BMI exposure to all-cause mortality rather than surrogate metabolic markers, we enhance clinical validity, particularly for aging populations in whom survival equity should be prioritized. Third, we applied the metrics BMI TAR and TTR to evaluate mortality-associated BMI thresholds, providing clinically actionable insights for long-term weight management strategies. Fourth, we conducted external validation, which increased the generalizability of our findings.

Limitations of the study

Our study has several limitations. First, it focused on the Chinese population, and multi-ethnic validations remain essential. Second, this was an observational study. Whether maintaining a long-term BMI target range reduces mortality risk remains to be validated in future randomized controlled trials. Third, although our analyses adjusted for an extensive set of confounding factors, residual confounding from unmeasured factors such as physical activity, education, and dietary factors cannot be excluded. Future studies should incorporate these factors for a more comprehensive understanding. Fourth, BMI TAR and TTR reduced longitudinal BMI trajectories into percentages and therefore cannot fully capture heterogeneity in trajectory patterns. More frequent or shorter-interval BMI measurements may further reduce interpolation errors.

Conclusions

To conclude, our results suggest that the “healthy” BMI for middle-aged and elderly individuals with T2DM needs to be refined. The threshold associated with increased mortality risk was sustained exposure to BMI ≥27 kg/m². By moving beyond static, single time point BMI assessment toward dynamic monitoring, our study provides evidence to support the concept of “sustained weight management” as a potential strategy in diabetes care to promote healthy longevity.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Yuqian Bao (yqbao@sjtu.edu.cn).

Materials availability

This study did not generate new, unique reagents.

Data and code availability

• •The dataset analyzed during the current study is available from the lead contact upon reasonable request.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

We thank Prof. Weiping Jia from Shanghai Sixth People’s Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Prof. Gang Hu from Pennington Biomedical Research Center, the research staff, and all participants for supporting this study. The study was supported by funding from the Noncommunicable Chronic Diseases-National Science and Technology Major Project (2023ZD0509200 and 2023ZD0509201), Shanghai Research Center for Endocrine and Metabolic Diseases (2022ZZ01002), National Key Clinical Specialty (Z155080000004), Three-year Action Program of Shanghai Municipality for Strengthening the Construction of the Public Health System (2023) (GWVI-11-36 and GWVI-11.2-YQ07), Shanghai Action Plan for Science, Technology and Innovation (23JS1401000), Three-year Action Program of Shanghai Municipality for Strengthening the Construction of the Public Health System (2023–2025) (GWVI-11.1-49), and the Shanghai Oriental Talent Program (Outstanding Project).

Author contributions

T.H. and R.C. analyzed the data and wrote the manuscript. Y.X., C.W., L.W., and L.C. collected the data. Y.S. evaluated data analysis. Y.B. and T.X. conceived the study. All authors contributed to the conception and design of the work, interpretation of the data, reviewed and provided edits and comments on the manuscript, approved the final version of the manuscript, and agreed to be accountable for all aspects of the work. Y.B. and T.X. are the guarantors of this work and, as such, have full access to all the data in the study.

Declaration of interests

We declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Biological samples
Blood	Participants in this study	N/A
Software and algorithms
R version 4.4.2	R Project	https://www.r-project.org

Experimental model and study participant details

The PREMIRE cohort

The PREMIRE cohort (PREdictors of MortalIty and Risk factors in patiEnts with T2DM) was established using EHRs from Shanghai Sixth People’s Hospital affiliated to Shanghai Jiao Tong University School of Medicine to investigate risk factors of health-related outcomes among individuals with T2DM, which has been clearly described elsewhere.^15,16 In brief, this cohort contained data from over 100,000 individuals with DM, including demographic characteristics, anthropometric measurements, laboratory results, diagnostic records, and medication prescriptions, systematically collected in a high-quality electronic database that continues to be regularly updated. The study protocol was approved by the Institutional Review Boards of Shanghai Sixth People’s Hospital affiliated to Shanghai Jiao Tong University School of Medicine, and all participants provided written informed consents.

Individuals who were on regular visits according to the guidance, with at least 5 measurements of BMI across the first 4 years after enrollment in the database were eligible for this study, with follow-up ended when they experienced death or until 31 December 2021. Furthermore, individuals with the following conditions were excluded from this analysis: (1) aged <40 years (2) missing data, (3) mortality-related data cannot be retrieved (Figure 1).

The CHARLS cohort

The CHARLS project was approved by the Biomedical Ethics Committee of Peking University, and all participants signed informed consent. This project was aimed to evaluate the health, socioeconomic, and demographic characteristics of Chinese adults aged over 45 years. The national baseline survey of CHARLS was carried out in 2011, adopting the multi-stage probability proportional to size (PPS) sampling approach. The sampling covered 450 villages, 150 counties, and 28 provinces, covering over 17,000 individuals from approximately 10,000 households. The CHARLS conducted follow-up assessments every 2 to 3 years. Participants underwent face-to-face interviews in their homes via computer-assisted personal interviewing (CAPI) technology. The detailed information about the CHARLS project was clearly described in previous studies.⁴⁰ The CHARLS datasets can be downloaded at http://charls.pku.edu.cn/en.

This study used data from four waves collected in 2011 (wave 1), 2013 (wave 2), 2015 (wave 3) and 2020 (wave 5), respectively. During the first three waves, BMIs were measured (Figure 5A). The sample size in wave 1 was 17,705. After excluding individuals aged <45 years, without three BMI measurements, missing value of covariates, and lost to follow-up, 990 eligible individuals with diabetes were included in the analysis.

Method details

BMI and its longitudinal metrics

Trained healthcare professionals measured height and body weight using calibrated stadiometers and digital scales according to standardized protocols, with individuals wearing light clothing and no shoes to ensure accuracy. BMI = weight (kg)/height² (m²).

We evaluated longitudinal weight changes primarily using TAR, TTR, and additional metrics including mean BMI and BMI variability (standard deviation and CV). TAR and TTR were calculated by linear interpolation using the Rosendaal method.²⁸ TAR was defined as the percentage of time during which BMIs were above target range. TTR was defined as the percentage of time during which BMIs were within target range. Owing to the relatively large proportion of individuals who had a TTR of 0% or 100%, those with a TTR of 0% or 100% were assigned to two subgroups, respectively, and the remaining individuals were divided by the median BMI TTR value. The detailed calculation process for these metrics was illustrated in Figure S5.

Assessment of covariates

The standardized electronic database was constructed by linking patient records through government-issued unique personal identifiers. Demographic and clinical data including birthdate, gender, age of diabetes diagnosis, smoking status, alcohol status, and medication details (antihypertensive, glucose-lowering, and lipid-lowering medications) were systematically extracted from well-standardized and structured Medical Record Audit Form. Current smokers were defined as those who smoked at least one cigarette per day for over six months.⁴¹ Current drinkers were defined as those who consumed at least 20 g/day alcohol for at least six months.⁴² Anthropometric measurements (height, body weight, blood pressure) and biochemical parameters (fasting plasma glucose, HbA1c, fasting C-peptide, fasting insulin, total cholesterol [TC], TG, HDL-C, LDL-C, and serum creatinine) were uniformly assessed. eGFR was calculated according to the Chronic Kidney Disease Epidemiology Collaboration equation.⁴³ The International Classification of Diseases 10th Revision (ICD-10) codes were used to identify comorbidities. All data underwent rigorous quality assurance through independent verification by two chief physicians. Education level was coded as a quantitative score: primary school or below: 0 points, secondary school or vocational education: 1 point, and higher education: 2 points. Household wealth was categorized into quartiles and assigned values from 0 to 3, with higher scores indicating greater wealth. A composite socioeconomic score was then calculated as the sum of the education and wealth scores. Physical activity level was defined according to the International Physical Activity Questionnaire 2010, and classified as low, moderate and high.⁴⁴

Clinical outcome

Causes and date of death were ascertained from the Shanghai Municipal Center for Disease Control and Prevention database, and linked with study data via personal identification number, as detailed in prior studies.¹⁶ Causes of death were recorded using the ICD-10 codes. The rate of missing death events in Shanghai was proven to be 0.7‰ (administrative data). Structured chart review was used to evaluate the confirmation of death (COD) through the Shanghai adaptation of the Medical Record Audit Form. Healthcare professionals have reviewed the medical records of a death event and reassigned the COD, which provided a gold standard to measure the quality of routine COD data. The death events identified by Shanghai Civil Registration and Vital Statistics routine monitoring were thus reported with 85.7% sensitivity and 90.0% specificity, respectively. The primary outcome in this study was all-cause mortality. The secondary outcomes were cardiovascular mortality and cancer mortality, respectively. ICD-10 codes for these outcomes were listed in Table S8. The follow-up ended at death or on 31 December 2021, whichever came first.

Quantification and statistical analysis

Continuous variables with normal distribution were expressed as mean ± standard deviation, whereas those with skewed distributions were presented as median (interquartile range). Categorical variables were expressed as number (proportion). ANOVA tests, Kruskal-Wallis test, and Chi-square test were applied to compare baseline characteristics for normally distributed, skewed, and categorical variables, respectively.

To handle multiple BMI observations per individual, we used a time-dependent Cox model, incorporating BMI threshold values from 24 kg/m² to 30 kg/m² to identify the optimal upper threshold for mortality risk. Each TAR threshold was evaluated separately by entering TAR as a continuous variable into the model, and HRs with corresponding 95% CIs were calculated for every 10% increase in BMI TAR. Stratified analyses according to age (<65 years and ≥65 years), HbA1c (<8% and ≥8%), history of cardiovascular diseases, and history of cancer were used to evaluate the association between different predefined BMI TARs and the risk of all-cause, cardiovascular, and cancer mortality, respectively.

To further validate the upper threshold, the association between BMI TTR_18.5-26.9 and mortality risk was explored. First, the restricted cubic spline analysis nested in time-dependent Cox models was used to test whether there were dose-response associations of BMI TTR with all-cause, cardiovascular, and cancer mortality, and the inflection point was estimated by “segmented” package in R. Second, three models were applied in the time-dependent Cox regression analysis. Model 1 was adjusted for gender and age. Model 2 was adjusted for gender, age, smoking status, alcohol status, and insurance type. Model 3 was adjusted for gender, age, smoking status, alcohol status, insurance type, duration of diabetes, HbA1c, systolic blood pressure, LDL-C, eGFR, history of cardiovascular diseases, history of cancer, use of antihypertensive medication, use of glucose-lowering medication, and use of lipid-lowering medication. Third, the predictive values for mortality using different lower BMI thresholds in TTR and the guideline-recommended range were compared. Fourth, subgroup analyses according to baseline characteristics (age, gender, HbA1c, eGFR, smoking status, use of lipid-lowering medication, use of antihypertensive medication, and use of glucose-lowering medication) were applied to test the robustness of their relationship and evaluate potential effect modifications. Fifth, to avoid the potential bias, sensitivity analyses were conducted by 1) excluding current smokers and past smokers, 2) excluding individuals who died within 1 year, 3) further adjusting for baseline BMI, mean and CV of BMI during the follow-up, 4) excluding those with severe comorbidities (cancer, heart failure, end-stage renal disease, or cachexia). Sixth, Spearman analysis was used to evaluate the correlations between BMI TAR, BMI TTR and key clinical parameters (systolic blood pressure, diastolic blood pressure, fasting plasma glucose, HbA1c, TC, TG, HDL-C, LDL-C, and eGFR). We also assessed the correlation between BMI TAR, TTR, baseline BMI and other longitudinal BMI metrics (mean and variability of BMI). Multivariate-adjusted time-dependent Cox models were also applied to investigate the association of longitudinal BMI metrics including mean, and variability of BMI with the risk of mortality. Finally, C-statistics, NRI, and IDI were used to evaluate the enhanced predictive performance of BMI TAR and TTR for mortality compared with the conventional model including age, diastolic blood pressure, use of antihypertensive medication, insulin therapy, TG, LDL-C, HDL-C, urine albumin-to-creatinine ratio, and BMI.⁴⁵

To further test the generalizability of our findings, we performed the external validation using the CHARLS cohort. Information on deaths was obtained from the CHARLS Exit Interview, which was completed by a family member or other proxy respondent for participants who had died before the next survey wave. Considering the incidence rate of death was >10%, a modified Poisson regression model with robust error variance was used to estimate the RR and 95%CI. Model 1 was adjusted for gender and age. Model 2 was adjusted for gender, age, smoking status, alcohol status, socioeconomic score, and physical activity level. Model 3 was adjusted for gender, age, smoking status, alcohol status, socioeconomic score, physical activity level, diabetes duration, HbA1c, systolic blood pressure, LDL-C, eGFR, history of cardiovascular diseases, history of cancer, use of antihypertensive medication, use of glucose-lowering medication, and use of lipid-lowering medication.

Data were analyzed using R 4.4.2 (R Foundation for Statistical Computing, Vienna, Austria), and a two-tailed p < 0.05 was considered statistically significant.

Published: January 20, 2026