Abstract
A high body mass index (BMI) is associated with the onset of diabetes mellitus (DM). However, evidence regarding the association between changes in BMI and DM onset is limited, and the effect of annual BMI (kg/m2/year) change on DM onset is unknown. Therefore, we assessed the effects of changes in BMI and annual BMI on DM onset. We enrolled 13,949 participants aged 21–81 years who underwent an annual health checkup at least twice between April 2003 and March 2021 and examined the effect of BMI change and annual BMI change on DM onset. In total, 462 individuals newly developed DM. Compared with a BMI change of –0.25–<0.25, univariate and multivariate analyses—adjusted for age, sex, BMI, systolic blood pressure, creatinine, total cholesterol, triglycerides, alanine aminotransferase, hemoglobin A1c, and family history of DM—showed that a BMI change <–2 was associated with a lower risk, while 2 ≤ BMI change < 4 and 4 ≤ BMI change were associated with a higher risk of DM onset. In contrast, compared with –0.05 ≤ BMI change per year < 0.05, univariate and multivariate analyses showed a significant association between DM onset and 0.3 ≤ BMI change per year. In the age-stratified analysis, these associations were significant among younger and middle-aged participants but not in older adults. In conclusion, changes in BMI affect DM onset. Therefore, clinicians can prevent DM onset by providing guidance based on BMI and focusing on a ≥2 increase in BMI and a ≥0.3 increase per year of BMI in young and middle-aged individuals.
Keywords: Body mass index, Obesity, Glycated hemoglobin
Graphical Abstract
Introduction
Unhealthy lifestyle changes, including excessive and imbalanced dietary intake and physical inactivity, have resulted in a chronic energy surplus, which has been a significant factor in the recent rise in the prevalence of diabetes mellitus (DM) [1, 2]. DM is a risk factor for serious diseases such as cardiovascular diseases (CVD), renal failure, decompensated liver diseases, and cancers, and it can pose a threat to healthy life expectancy [3-10]. Therefore, appropriate management of glycemic control and prevention of DM onset is essential.
Body mass index (BMI) is the simplest indicator of body composition and is used in health checkups and outpatient settings as an important surrogate marker for assessing health status [11-14]. A high BMI is a risk factor for mortality in coronary heart disease, stroke, respiratory disease, and cancer [11, 12]. In contrast, a low BMI is also associated with the incidence of death caused by respiratory disease, coronary heart disease, stroke, and cancer [11]. Additionally, BMI is a predictive factor for the onset of metabolic dysfunction associated with steatotic liver and CVDs [13-15].
Concerning the association between BMI and diabetes, prior studies have focused on individuals with established diabetes, showing that BMI level and BMI change were related to the risk of microvascular and macrovascular complications of diabetes [16]. In contrast, studies investigating the risk of incident diabetes have demonstrated that higher baseline BMI confers a greater lifetime risk [17], and that BMI change, expressed mainly as percentage change or ±5% categories, is associated with diabetes onset in Western and Middle Eastern cohorts [18-20]. Additionally, a BMI of 25–29.9 kg/m2 or ≥30 kg/m2 has consistently been identified as a risk factor for diabetes onset compared with a BMI of <25 kg/m2 [21]. Although BMI change (%) has been linked to diabetes risk [22], evidence remains limited, and the roles of absolute (kg/m2) and annual (kg/m2/year) BMI change in diabetes development are not well defined.
Few studies have simultaneously assessed absolute BMI change, annual change, and percentage change for incident diabetes in a large Japanese cohort. Moreover, age- and baseline BMI–stratified analyses are particularly relevant in Asian populations with a high prevalence of lean diabetes, but evidence remains scarce. To address these gaps, in this community-based longitudinal cohort study, we assessed the effects of absolute (kg/m2), annual (kg/m2/year), and percentage BMI changes on the development of DM, and derived ROC-based thresholds to provide clinically interpretable guidance.
Materials and Methods
Approval of the research protocol
This study adhered to the ethical guidelines set forth in the Declaration of Helsinki and its later revisions and received approval from the Research Ethics Committee of Ehime University Hospital (approval number: 1709007). The requirement for informed consent was waived because the study used data from routine health checkup. To maintain participant anonymity, individuals were assigned a unique numerical identifier, and all data were securely stored in a protected database.
Patient characteristics
In this study, we enrolled 14,934 participants aged 21–81 years who underwent an annual health checkup at least twice between April 2003 and March 2021 at the Ehime General Health Care Association and reviewed their data. We excluded 985 participants who had DM at baseline (n = 758) and missing data for parameters, such as BMI (n = 1), hemoglobin A1c (HbA1c) (n = 8), and total cholesterol (TC) (n = 250) (Fig. 1). Finally, 13,949 participants were included in the analysis. In the final analytic cohort, the median number of health checkups per participant was 4 (interquartile range 2–6), with a mean of 4.65 visits (range 2–26). Accordingly, the follow-up duration from the first to the last checkup was not uniform (mean 4.72 years; median 3.18 years [interquartile range 2.07–7.00]; range 0.34–16.91 years).
Fig. 1. Flowchart of participant selection in this study.
The assessment items included age; sex; basic physical data, such as height and weight for calculating BMI while wearing light clothes; blood pressure in a seated position using an automatic sphygmomanometer; blood chemistry tests for >10 h, such as creatinine, TC, triglycerides (TG), alanine aminotransferase (ALT), fasting plasma glucose (FPG), and HbA1c levels; a questionnaire about past medical history; and family history of DM within second-degree relatives.
DM diagnosis and BMI change
The following criteria were used to diagnose and determine the onset of DM: 1) HbA1c ≥6.5%, 2) FPG level ≥126 mg/dL, or 3) history of diabetes treatment [23, 24]. Baseline BMI was defined as the value at the first health checkup during the study period, and endpoint BMI as the value at the final health checkup. For participants who developed DM during follow-up, the endpoint was the time of DM diagnosis, and subsequent data were not used. Participants who did not develop DM were followed until their last checkup. For those with three or more visits, BMI change was calculated from the first to the last available examination. The change in BMI was calculated using BMI at baseline and endpoint. The annual change in BMI was calculated by dividing the change in BMI by the observation period. In addition, percentage change in BMI was calculated as ΔBMI% = (BMIendpoint – BMIbaseline)/BMIbaseline × 100. The participants were categorized into six groups according to the degree of BMI change: BMI change < –2 (n = 523), –2 ≤ BMI change < –0.25 (n = 3,876), –0.25 ≤ BMI change < 0.25 (n = 2,832), 0.25 ≤ BMI change < 2 (n = 5,647), 2 ≤ BMI change < 4 (n = 910), and 4 ≤ BMI change (n = 161). Additionally, the BMI change per year was divided into six groups: BMI change per year <–0.2 (n = 2,553), –0.2 ≤ BMI change per year < –0.05 (n = 2,159), –0.05 ≤ BMI change per year < 0.05 (n = 2,197), 0.05 ≤ BMI change per year < 0.2 (n = 2,970), 0.2 ≤ BMI change per year < 0.3 (n = 1,272), and 0.3 ≤ BMI change per year (n = 2,798).
Analysis
All statistical analyses were conducted using JMP software, version 14.2.0 (SAS Institute Japan, Tokyo, Japan). A p-value <0.05 was considered statistically significant. We used the χ2 test to compare categorical variables such as sex, family history of DM, and onset of DM. Continuous variables, such as age, BMI, systolic blood pressure, creatinine, TC, TG, ALT, FPG, and HbA1c, were analyzed using the Steel test. Univariate and multivariate analyses, which were adjusted for other risk factors such as age, sex, BMI, systolic blood pressure, creatinine, TC, TG, ALT, HbA1c, and family history of DM [25-27], were performed using a logistic regression model. The observation period was not uniform. Since BMI change was calculated as the difference between baseline and endpoint rather than time-to-event, results were expressed as odds ratios (OR). Additionally, we modeled BMI changes as continuous variables (ΔBMI, ΔBMI per year, and ΔBMI%) and performed receiver-operating-characteristic (ROC) analyses to derive Youden-index cut-offs.
Results
Characteristics at baseline
Table 1 presents the baseline characteristics of all participants across the groups. The median age was 44 years, and 6,687 participants (47.9%) were male. The median BMI, FPG, and HbA1c were 22.1 kg/m2, 91 mg/dL, and 5.3%, respectively. A total of 462 individuals developed DM.
Table 1. Characteristics at baseline.
| BMI < –2 (n = 523) |
–2 ≤ BMI < –0.25 (n = 3,876) |
–0.25 ≤ BMI < 0.25 Control (n = 2,832) |
0.25 ≤ BMI < 2 (n = 5,647) |
2 ≤ BMI < 4 (n = 910) |
BMI ≥ 4 (n = 161) |
|
|---|---|---|---|---|---|---|
| Age, years | 45 (37–51)* |
45 (38–52) |
45 (38–52) |
43 (37–51)* |
39 (35–45.25)* |
36 (32–41)* |
| Sex, n (Men/Women)** | 298/225 | 1,992/1,884 | 1,343/1,489 | 2,625/3,022 | 391/519 | 38/123 |
| BMI, kg/m2 | 25.2 (23.1–27.8)* |
22.5 (20.6–24.7)* |
21.6 (19.7–23.8) |
21.8 (19.9–24.1)* |
22.3 (20.4–24.6)* |
22.7 (20.5–25.4)* |
| SBP, mmHg | 119 (109–132)* |
113 (103–125)* |
111 (101–123) |
110 (100–123) |
110 (100–121) |
109 (100–120) |
| Cre, μM | 70.7 (56.5–79.5)* |
67.1 (55.0–79.5) |
64.5 (53.9–78.6) |
63.6 (53.9–77.7) |
61.8 (53.0–71.6)* |
57.4 (53.0–64.9)* |
| TC, mM | 5.35 (4.78–5.97)* |
5.19 (4.68–5.84) |
5.19 (4.65–5.79) |
5.12 (4.57–5.71)* |
4.99 (4.52–5.53)* |
4.96 (4.43–5.58)* |
| TG, mM | 1.20 (0.80–1.77)* |
0.96 (0.66–1.45)* |
0.88 (0.64–1.32) |
0.88 (0.62–1.31) |
0.85 (0.59–1.23) |
0.74 (0.59–1.11)* |
| ALT, U/L | 22 (15–36)* |
18 (13–27)* |
17 (13–24) |
17 (13–24) |
16 (12–23) |
14 (11–19)* |
| FPG, mM | 5.21 (4.93–5.55)* |
5.10 (4.82–5.43)* |
5.05 (4.77–5.38) |
5.05 (4.77–5.38) |
5.05 (4.77–5.32) |
4.93 (4.66–5.16)* |
| HbA1c, % | 5.7 (5.4–5.9)* |
5.6 (5.4–5.8) |
5.6 (5.4–5.8) |
5.6 (5.4–5.8) |
5.5 (5.2–5.8)* |
5.4 (5.2–5.6)* |
| DM onset, n (%)** | 6 (1.1%) |
106 (2.7%) |
83 (2.9%) |
211 (3.7%) |
44 (4.8%) |
12 (7.5%) |
| DM family history, Yes, n (%) | 113 (21.6%) |
804 (20.7%) |
589 (20.8%) |
1,261 (22.3%) |
198 (21.7%) |
32 (19.9%) |
Data are presented as median (interquartile range) or number (percentage). The Steel test was used to analyze continuous variables compared with the control (–0.25 ≤ BMI change < 0.25), and the χ2-test was used to analyze categorical variables. Differences were considered statistically significant at p < 0.05. * indicates a significant difference compared with the control. ** indicates a significant difference among the categories.
ALT, alanine aminotransferase; BMI, body mass index; Cre, creatinine; FPG, fasting plasma glucose; HbA1c, hemoglobin A1c; SBP, systolic blood pressure; TC, total cholesterol; TG, triglycerides.
The other groups had significantly higher BMI than the control group (–0.25 ≤ BMI change < 0.25). Furthermore, participants in BMI < –2, 0.25 ≤ BMI < 2, 2 ≤ BMI < 4, and BMI ≥ 4 groups were significantly younger than those in the control group. Regarding TG, ALT, and FPG levels, BMI < –2 and –2 ≤ BMI < –0.25 groups showed significantly higher values, while the BMI ≥ 4 group showed significantly lower values than the control group. Systolic blood pressure was elevated in BMI < –2 and –2 ≤ BMI < –0.25 groups, while TC levels were higher in the BMI < –2 group and lower in the 0.25 ≤ BMI < 2, 2 ≤ BMI < 4, and BMI ≥ 4 groups.
Association between total BMI change and DM onset
In the univariate analysis, BMI change < –2 indicated a lower risk and 2 ≤ BMI change < 4 and BMI ≥ 4 change indicated higher risks for onset of DM than the control group (–0.25 ≤ BMI change < 0.25; Table 2). Additionally, in the multivariate analysis, adjusted for age, sex, BMI, systolic blood pressure, creatinine, TC, TG, ALT, HbA1c, and family history of DM, these associations remained significant (Table 2). When expressed as a percentage change from baseline, greater BMI increase was significantly associated with incident diabetes. In the multivariable logistic regression analysis, the adjusted odds ratio (OR) was 1.11 per 1% increase in BMI (95% CI 1.09–1.13, p < 0.001). For interpretability, the OR per 5% increase was 1.70 (95% CI 1.54–1.87, p < 0.001) (Supplementary Table 1). In the age-stratified multivariate analysis, for participants aged 21–40 years, BMI change ≥ 4 indicated a higher risk for onset of DM than that in the control group, and in those aged 40–61 years a 2 ≤ BMI change < 4 indicated a higher risk for DM onset. BMI change < –2 and –2 ≤ BMI change < –0.25 indicated a lower risk for DM onset than that in the control group (Table 3). However, an analysis for participants aged 61–80 years could not be performed owing to insufficient statistical power (Table 3). In a sensitivity analysis combining BMI changes into three categories (decrease, stable, increase), results were consistent with the six-category model. When analyzed as a continuous variable, each 1 kg/m2 increase in BMI was significantly associated with an increase in diabetes risk in the overall cohort (adjusted OR 1.57, 95% CI 1.45–1.70, p < 0.001). Similar associations were observed across different age groups. In the three-category analysis, statistical significance was not observed in the young and older adult subgroups, likely due to limited sample size. In contrast, when analyzed as a continuous variable, BMI increase was significantly associated with an increase in diabetes risk in all age groups (Supplementary Tables 2, 3). When treated as continuous variables, the difference between ΔBMI, ΔBMI/year, and ΔBMI% was modest (AUCs 0.553–0.590). The Youden-derived cut-offs were approximately 0.20 kg/m2 for ΔBMI, 0.31 kg/m2/year for ΔBMI/year, and –0.89% for ΔBMI%, with limited incremental AUC gains over the multivariable base model (increased from 0.902 to 0.905–0.911) (Supplementary Table 4). In the stratified analyses by baseline BMI (<22, 22–<25, ≥25 kg/m2), patterns differed across strata (Supplementary Table 5). In participants with BMI ≥ 25 kg/m2, BMI decrease was significantly associated with a reduced risk of diabetes (adjusted OR 0.37, 95% CI 0.23–0.58), whereas BMI increase was not significant. In the 22–<25 kg/m2 stratum, BMI decrease was not significant, but BMI increase was significantly associated with a higher risk of diabetes (OR 2.22, 95% CI 1.32–3.73). In the <22 kg/m2 stratum, BMI decrease did not confer protection and tended to be associated with an increased risk of diabetes (OR 2.07, 95% CI 0.95–4.54).
Table 2. Analysis of the association between total body mass index change and onset of diabetes mellitus.
| BMI change | BMI < –2 | –2 ≤ BMI < –0.25 | –0.25 ≤ BMI < 0.25 Control |
0.25 ≤ BMI < 2 | 2 ≤ BMI < 4 | BMI ≥ 4 |
|---|---|---|---|---|---|---|
| DM onset rate | 1.15 | 2.73 | 2.93 | 3.74 | 4.84 | 7.45 |
| Univariate analysis OR (95%CI) | 0.38 (0.16–0.88) |
0.93 (0.69–1.24) |
1 | 1.28 (0.99–1.66) |
1.68 (1.15–2.44) |
2.66 (1.42–4.99) |
| p value | 0.024 | 0.630 | — | 0.056 | 0.006 | 0.002 |
| Multivariate analysis OR (95%CI) | 0.14 (0.06–0.36) |
0.71 (0.51–1.00) |
1 | 1.36 (1.01–1.83) |
3.08 (1.97–4.80) |
15.3 (7.11–33.0) |
| p value | <0.001 | 0.051 | — | 0.037 | <0.001 | <0.001 |
Differences were considered statistically significant at p < 0.05.
The multivariate analysis was adjusted for age, sex, BMI, SBP, Cre, TC, TG, ALT, HbA1c, and family history of DM.
ALT, alanine aminotransferase; BMI, body mass index; CI, confidence interval; Cre, creatinine; DM, diabetes mellitus; HbA1c, hemoglobin A1c; OR, odds ratio; SBP, systolic blood pressure; TC, total cholesterol; TG, triglycerides.
Table 3. Analysis of the association between total body mass index change and onset of diabetes mellitus stratified by age.
| BMI change | BMI < –2 | –2 ≤ BMI < –0.25 | –0.25 ≤ BMI < 0.25 Control |
0.25 ≤ BMI < 2 | 2 ≤ BMI < 4 | BMI ≥ 4 |
|---|---|---|---|---|---|---|
| 21–40 years | ||||||
| DM onset rate | 0.55% (1/181) |
1.03% (13/1,265) |
0.76% (7/919) |
1.36% (29/2,139) |
2.15% (11/511) |
8.47% (10/118) |
| Univariate analysis OR (95%CI) | 0.73 (0.09–5.92) |
1.35 (0.54–3.40) |
1 | 1.79 (0.78–4.10) |
2.87 (1.10–7.44) |
12.06 (4.50–32.34) |
| p value | 0.76 | 0.521 | — | 0.169 | 0.031 | <0.001 |
| Multivariate analysis OR (95%CI) | N/A | 1.28 (0.31–5.20) |
1 | 2.25 (0.62–8.19) |
2.77 (0.56–13.75) |
32.45 (5.97–176.54) |
| p value | N/A | 0.733 | — | 0.219 | 0.214 | <0.001 |
| 41–60 years | ||||||
| DM onset rate | 1.52% (5/330) |
3.34% (82/2,454) |
3.91% (69/1,766) |
5.04% (166/3,296) |
8.25% (32/388) |
4.65% (2/43) |
| Univariate analysis OR (95%CI) | 0.38 (0.15–0.95) |
0.85 (0.61–1.18) |
1 | 1.30 (0.98–1.74) |
2.21 (1.43–3.41) |
1.20 (0.28–5.06) |
| p value | 0.038 | 0.330 | — | 0.070 | <0.001 | 0.804 |
| Multivariate analysis OR (95%CI) | 0.20 (0.06–0.65) |
0.77 (0.50–1.21) |
1 | 1.25 (0.84–1.86) |
4.89 (2.63–9.09) |
6.46 (1.29–32.26) |
| p value | <0.001 | 0.257 | — | 0.279 | <0.001 | 0.023 |
| 61–81 years | ||||||
| DM onset rate | 0% (0/12) |
7.01% (11/157) |
4.76% (7/147) |
7.55% (16/212) |
9.09% (1/11) |
N/A |
| Univariate analysis OR (95%CI) | N/A | 1.51 (0.58–4.20) |
1 | 1.63 (0.68–4.20) |
2.00 (0.10–12.94) |
N/A |
| p value | N/A | 0.405 | — | 0.28 | 0.565 | N/A |
| Multivariate analysis OR (95%CI) | N/A | 0.26 (0.05–1.21) |
1 | 1.08 (0.29–3.98) |
N/A | N/A |
| p value | N/A | 0.09 | — | 0.91 | N/A | N/A |
Differences were considered statistically significant at p < 0.05.
The multivariate analysis was adjusted for age, sex, BMI, SBP, Cre, TC, TG, ALT, HbA1c, and family history of DM.
ALT, alanine aminotransferase; BMI, body mass index; CI, confidence interval; Cre, creatinine; DM, diabetes mellitus; HbA1c, hemoglobin A1c; OR, odds ratio; SBP, systolic blood pressure; TC, total cholesterol; TG, triglycerides.
Association between BMI change per year and onset of DM
In the univariate analysis, BMI change per year < –0.2 and BMI change per year ≥ 0.3 indicated higher risks for onset of DM than that in the control group (–0.05 ≤ BMI change per year < 0.05; Table 4). Additionally, in the multivariate analysis adjusted for age, sex, BMI, systolic blood pressure, creatinine, TC, TG, ALT, HbA1c, and family history of DM, the association between DM onset and BMI change per year ≥ 0.3 remained significant (Table 4). In the age-stratified multivariate analysis, for participants aged 21–40 and 41–60 years, a BMI change per year ≥ 0.3 indicated a higher risk for onset of DM than the control group, and for those aged 40–61 years, a BMI change per year < –0.2 indicated a lower risk for onset of DM (Table 5). However, in the multivariate analysis for subjects aged 61–80 years, there was no significant association between BMI change per year and the onset of DM (Table 5).
Table 4. Analysis of the association between body mass index change per year and the onset of diabetes mellitus.
| BMI change | BMI change per year < –0.2 |
–0.2≤ BMI change per year < –0.05 |
–0.05 ≤ BMI change per year < 0.05 Control |
0.05 ≤ BMI change per year < 0.2 |
0.2 ≤ BMI change per year < 0.3 |
BMI change per year ≥ 0.3 |
|---|---|---|---|---|---|---|
| DM onset rate | 3.1 | 2.6 | 2.1 | 2.2 | 2.7 | 6.5 |
| Univariate analysis OR (95%CI) | 1.53 (1.06–2.22) |
1.18 (0.80–1.76) |
1 | 1.02 (0.70–1.50) |
1.38 (0.88–2.16) |
3.00 (2.16–4.17) |
| p value | 0.06 | 0.41 | — | 0.90 | 0.16 | <0.001 |
| Multivariate analysis OR (95%CI) | 0.67 (0.44–1.01) |
0.90 (0.58–1.40) |
1 | 0.98 (0.64–1.50) |
1.33 (0.81–2.19) |
2.27 (1.56–3.30) |
| p value | 0.059 | 0.63 | — | 0.94 | 0.26 | <0.0001 |
Differences were considered statistically significant at p < 0.05.
The multivariate analysis was adjusted for age, sex, BMI, SBP, Cre, TC, TG, ALT, HbA1c, and family history of DM.
ALT, alanine aminotransferase; BMI, body mass index; CI, confidence interval; Cre, creatinine; DM, diabetes mellitus; HbA1c, hemoglobin A1c; OR, odds ratio; SBP, systolic blood pressure; TC, total cholesterol; TG, triglycerides.
Table 5. Analysis of the association between body mass index change per year and the onset of diabetes mellitus stratified by age.
| BMI change | BMI change per year < –0.2 |
–0.2 ≤ BMI change per year < –0.05 |
–0.05 ≤ BMI change per year < 0.05 Control |
0.05 ≤ BMI change per year < 0.2 |
0.2 ≤ BMI change per year < 0.3 |
BMI change per year ≥ 0.3 |
|---|---|---|---|---|---|---|
| 21–40 years | ||||||
| DM onset rate | 0.94% (7/745) |
1.10% (8/730) |
0.57% (5/875) |
0.72% (9/1,242) |
0.98% (5/509) |
3.59% (37/1,032) |
| Univariate analysis OR (95%CI) | 1.65 (0.52–5.22) |
1.93 (0.63–5.92) |
1 | 1.27 (0.42–3.80) |
1.73 (0.50–5.99) |
6.47 (2.53–16.54) |
| p value | 0.394 | 0.251 | — | 0.669 | 0.390 | <0.001 |
| Multivariate analysis OR (95%CI) | 1.04 (0.11–9.95) |
3.03 (0.33–28.28) |
1 | 0.99 (0.09–11.49) |
2.75 (0.26–28.81) |
7.91 (1.02–61.62) |
| p value | 0.976 | 0.330 | — | 0.995 | 0.398 | 0.048 |
| 41–60 years | ||||||
| DM onset rate | 3.97% (67/1,689) |
3.04% (41/1,347) |
3.20% (40/1,249) |
3.22% (52/1,617) |
3.87% (28/724) |
7.75% (128/1,651) |
| Univariate analysis OR (95%CI) | 1.25 (0.84–1.86) |
0.95 (0.61–1.48) |
1 | 1.00 (0.66–1.53) |
1.22 (0.74–1.99) |
2.54 (1.77–3.65) |
| p value | 0.275 | 0.816 | — | 0.984 | 0.283 | <0.001 |
| Multivariate analysis OR (95%CI) | 0.69 (0.39–1.22) |
0.89 (0.47–1.66) |
1 | 0.72 (0.38–1.35) |
1.01 (0.50–2.02) |
1.97 (1.17–3.34) |
| p value | 0.208 | 0.707 | — | 0.309 | 0.987 | 0.011 |
| 61–81 years | ||||||
| DM onset rate | 7.56% (9/119) |
3.66% (3/82) |
2.74% (2/73) |
6.31% (7/104) |
7.69% (3/39) |
9.57% (11/115) |
| Univariate analysis OR (95%CI) | 2.90 (0.61–13.84) |
1.35 (0.22–8.30) |
1 | 2.39 (0.48–11.84) |
2.96 (0.47–18.51) |
3.75 (0.81–17.45) |
| p value | 0.15 | 0.80 | — | 0.29 | 0.25 | 0.09 |
| Multivariate analysis OR (95%CI) | 0.43 (0.05–3.59) |
1.29 (0.12–13.40) |
1 | 1.30 (0.16–10.36) |
1.54 (0.15–16.42) |
3.06 (0.39–23.80) |
| p value | 0.44 | 0.83 | — | 0.65 | 0.719 | 0.285 |
Differences were considered statistically significant at p < 0.05.
The multivariate analysis was adjusted for age, sex, BMI, SBP, Cre, TC, TG, ALT, HbA1c, and family history of DM.
ALT, alanine aminotransferase; BMI, body mass index; CI, confidence interval; Cre, creatinine; DM, diabetes mellitus; HbA1c, hemoglobin A1c; OR, odds ratio; SBP, systolic blood pressure; TC, total cholesterol; TG, triglycerides.
Discussion
In this study, we examined the association between BMI change and the onset of DM and showed the correlation between risk of DM onset and BMI change. A ≥2 increase in BMI elevated the risk of developing DM, and a ≤–2 reduction in BMI decreased the risk of DM onset regardless of the BMI and HbA1c at baseline. Additionally, regarding the BMI change per year, the ≥0.3 increase in BMI change per year was associated with an increase in risk for developing DM. In older participants, the associations were weaker and did not reach statistical significance. However, sensitivity analyses combining the BMI-change categories into three groups or treating BMI change as a continuous variable showed consistent trends with the overall cohort. These findings suggest that the lack of significance in the older adult subgroup is more likely due to limited statistical power rather than an absence of association. Taken together, to prevent the development of DM in younger and middle-aged individuals, it is important to control BMI and ensure that it does not increase by ≥2 or ≥0.3 per year.
The association between BMI and onset of DM have been examined [17-20, 22, 28-30]. Bae’s study included 8,900 participants with a mean age of 52.3 years from the KoGES Ansan and Ansung studies who had no history of CVD and were followed up. This study showed an association between BMI and the development of type 2 DM (T2DM), adjusted for age, smoking, alcohol intake, and regular exercise. The analysis revealed an elevated risk of developing DM at BMI levels above 23.7 kg/m2, with the lowest risk observed around a BMI of 21 kg/m2 in men (p-value for nonlinearity = 0.029). In contrast, women showed a gradual increase in hazard ratio (HR) for developing DM with increasing BMI (p-value for nonlinearity = 0.108) [28]. Other studies aimed to explore the link between BMI and the incidence of diabetes over a median follow-up period of 12.5 years. The analysis evaluated the relationship between various BMI categories [underweight (<18.5 kg/m2), normal (18.5–<23 kg/m2), overweight (23–<25 kg/m2), grade 1 obesity (25–<30 kg/m2), grade 2 obesity (30–<35 kg/m2), grade 3 obesity (≥35 kg/m2)] and DM onset, adjusting for variables such as age, systolic blood pressure, glucose levels, TC, smoking status, alcohol consumption, physical activity, and socioeconomic factors. Compared to individuals with normal BMI, the HRs and 95% CIs for DM in men were 0.84 (0.72–0.99) for underweight, 1.60 (1.52–1.69) for overweight, 2.43 (2.32–2.55) for grade 1 obesity, 4.78 (4.37–5.24) for grade 2 obesity, and 5.59 (3.85–6.62) for grade 3 obesity. In women, the corresponding HRs (95% CIs) were 0.74 (0.61–0.91), 1.62 (1.53–1.73), 2.57 (2.43–2.71), 4.37 (3.99–4.79), and 4.81 (3.62–6.38), respectively [29]. Furthermore, a meta-analysis of prospective cohort studies indicated that the pooled risk ratios for T2DM, based on BMI categories adjusted for ethnic differences, were 0.93 (95% CI 0.75–1.15, I2 = 55.5%, n = 12) for underweight, 2.24 (95% CI 1.95–2.56, I2 = 92.0%, n = 47) for overweight, 4.56 (95% CI 3.69–5.64, I2 = 96%, n = 43) for obesity, and 22.97 (95% CI 13.58–38.86, I2 = 92.1%, n = 6) for severe obesity, compared to individuals with normal BMI [30]. Nonetheless, these studies did not explore the relationship between changes in BMI and the development of diabetes, which was the primary focus of our research. Regarding the association between BMI change and DM onset, Ohno et al. reviewed the Jinnah Medical and Dental College Claims Database between 2005 and 2021. They examined 3,400,303 individuals with no previous history of DM or use of glucose-lowering medications to explore the association between value or change and the risk of DM onset [22]. The risk of DM onset increased significantly when the BMI surpassed 20–21 kg/m2. Additionally, compared with participants with a stable BMI (–5.0% to +5.0%), those whose BMI increased by 5.0% or more had a high risk for DM onset after adjustment for age, sex, baseline BMI, FPG, hypertension, dyslipidemia, cigarette smoking, alcohol consumption, and physical inactivity [adjusted hazard ratios (aHR):1.33, 95% CI 1.31–1.36]. In contrast, the aHR for DM onset among individuals whose BMI decreased by 5.0% or more, was 0.82 (95% CI 0.80–0.84). In the normal weight (18.5 ≤ BMI at baseline ≤ 24.9 kg/m2), overweight (25.9 ≤ BMI at baseline 29.9 kg/m2), and obese groups (30 kg/m2 ≤ BMI at baseline), a decrease in BMI was associated with a low risk of developing DM. However, in the underweight group (18.5 > BMI at baseline), a decrease in BMI was associated with an increased risk of DM onset [22]. This study defined the incidence of DM based on the use of DM medications along with a diagnosis of DM by ICD-10 code; therefore, the number of patients with newly onset diabetes may be underestimated. Moreover, this study examined BMI by percentage change, not by the amount of change, and did not examine the risk associated with annual BMI change. Additionally, several studies have provided important but incomplete insights into the relationship between BMI and DM [17-20]. Narayan et al. reported BMI-specific lifetime risk of diabetes in the U.S. population but did not evaluate longitudinal changes in BMI [17]. Ye et al. reported that percentage changes in BMI were associated with DM risk in a Canadian cohort [18]. Asgari et al. demonstrated that weight change categories of ±5% predicted diabetes risk in Iranian adults, but did not evaluate absolute or annual BMI changes [20]. In contrast, our study simultaneously assessed absolute BMI change (kg/m2), annual change (kg/m2/year), and percentage change in a large Japanese health-checkup cohort and further performed stratified analyses by age and baseline BMI. We also derived ROC-based cut-offs, which corresponded well with our categorical definitions, thereby providing interpretable thresholds for preventive guidance. These analyses extend the existing literature by demonstrating the clinical utility of BMI change metrics in a large Asian cohort with a high prevalence of lean DM. In contrast, another study indicated that changes in BMI had no effect on the onset of DM [19]. Salesi et al. conducted a study involving 1,557 participants with an average age of 47.1 years as part of the Isfahan Cohort Study—a large population-based research project in central Iran [19]. Individuals without a history of DM at the start were monitored for 13 years. DM was defined by FPG levels exceeding 126 mg/dL (7.0 mmol/L), a 2-h post-challenge glucose reading above 200 mg/dL (11.1 mmol/L), or the use of antidiabetic medications. The study found a link between BMI change and DM onset after adjusting for factors such as age, gender, family history of DM, smoking, dietary habits (healthy eating index), and physical activity. Compared to the control group (±5% change in BMI), no significant difference in diabetes onset was observed among individuals who experienced a BMI decrease of ≥10%, a 5%–10% decrease, a 5%–10% increase, or a ≥10% increase, regardless of their baseline BMI classification [normal/underweight (<25 kg/m2), overweight (25–29.9 kg/m2), or obese (≥30 kg/m2)] [19]. Therefore, the association between diabetes onset and changes in BMI remains controversial and requires further examination.
In the additional analyses of BMI change as continuous variables, ROC-based cut-offs were derived using the Youden index. The optimal thresholds were approximately 0.20 kg/m2 for ΔBMI and 0.31 kg/m2/year for annual change, which were consistent with our pre-specified categorical definitions (e.g., ≥0.3 kg/m2/year). The cut-off for ΔBMI% (–0.9%) indicated that failure to achieve at least 1% BMI reduction was associated with higher diabetes risk, whereas greater reductions had protective effects. Although the overall discriminatory ability of single BMI-change metrics was modest, these findings highlight that BMI change is an important factor influencing diabetes risk and a meaningful target for prevention. Our categorical approach is therefore both clinically interpretable and aligned with ROC-derived thresholds.
Baseline BMI–stratified analyses further clarified that weight reduction was beneficial among individuals who are overweight or obese (≥25 kg/m2) but did not reduce the risk in lean individuals (<22 kg/m2). Excessive BMI reduction in lean participants tended to be associated with higher diabetes risk, suggesting that preventive strategies should prioritize weight control in overweight/obese groups while avoiding undue weight loss in lean individuals.
This research has some limitations. First, the observation period was not consistent across participants. Therefore, the short observation period may have led to missed cases of DM onset. Second, as the participants were all from a single region in Japan, regional bias may have occurred. Third, although treatment intervention after DM onset may reduce BMI, in our cohort only 20 of the 462 participants with newly diagnosed DM (4.3%) were receiving treatment at the time of diagnosis. Thus, the influence of treatment-related BMI reduction is considered limited. Finally, a significant relationship was observed between changes in BMI and DM onset regardless of BMI or HbA1c level at baseline; however, the limited number of participants prevented analysis stratified by BMI or HbA1c level at baseline. Despite these limitations, this study has important implications for the prevention of DM onset.
In conclusion, changes in BMI affect the onset of DM, and a ≥2 increase in BMI and ≥0.3 increase in annual BMI are the risk factors for DM onset, and ≤–2 reduction in BMI decreases the risk of DM onset in young and middle-aged participants (Graphical Abstract). Therefore, clinicians can prevent the onset of DM by providing guidance based on BMI.
Graphical Abstract.
Acknowledgments
We thank Editage (www.editage.com) for English language editing.
Disclosure
None of the authors have any potential conflicts of interest associated with this research.
Shinya Furukawa is a member of Endocrine Journal’s Editorial Board.
Funding
The work submitted did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Supplementary Material
Supplementary Tables
References
- 1.Kolb H, Martin S (2017) Environmental/lifestyle factors in the pathogenesis and prevention of type 2 diabetes. BMC Med 15: 131. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Volaco A, Cavalcanti AM, Filho RP, Précoma DB (2018) Socioeconomic status: the missing link between obesity and diabetes mellitus? Curr Diabetes Rev 14: 321–326. [DOI] [PubMed] [Google Scholar]
- 3.Miyake T, Furukawa S, Matsuura B, Yoshida O, Miyazaki M, et al. (2024) Glycemic control is associated with histological findings of nonalcoholic fatty liver disease. Diabetes Metab J 48: 440–448. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.En Li Cho E, Ang CZ, Quek J, Fu CE, Lim LKE, et al. (2023) Global prevalence of non-alcoholic fatty liver disease in type 2 diabetes mellitus: an updated systematic review and meta-analysis. Gut 72: 2138–2148. [DOI] [PubMed] [Google Scholar]
- 5.Miyake T, Kumagi T, Furukawa S, Tokumoto Y, Hirooka M, et al. (2013) Non-alcoholic fatty liver disease: factors associated with its presence and onset. J Gastroenterol Hepatol 28 Suppl 4: 71–78. [DOI] [PubMed] [Google Scholar]
- 6.Wang M, Sperrin M, Rutter MK, Renehan AG (2023) Cancer is becoming the leading cause of death in diabetes. Lancet 401: 1849. [DOI] [PubMed] [Google Scholar]
- 7.Yun JS, Ko SH (2021) Current trends in epidemiology of cardiovascular disease and cardiovascular risk management in type 2 diabetes. Metabolism 123: 154838. [DOI] [PubMed] [Google Scholar]
- 8.Joseph JJ, Deedwania P, Acharya T, Aguilar D, Bhatt DL, et al. (2022) Comprehensive management of cardiovascular risk factors for adults with type 2 diabetes: a scientific statement from the American Heart Association. Circulation 145: e722–e759. [DOI] [PubMed] [Google Scholar]
- 9.Deng Y, Li N, Wu Y, Wang M, Yang S, et al. (2021) Global, regional, and national burden of diabetes-related chronic kidney disease from 1990 to 2019. Front Endocrinol (Lausanne) 12: 672350. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Xie Y, Bowe B, Mokdad AH, Xian H, Yan Y, et al. (2018) Analysis of the global burden of disease study highlights the global, regional, and national trends of chronic kidney disease epidemiology from 1990 to 2016. Kidney Int 94: 567–581. [DOI] [PubMed] [Google Scholar]
- 11.Global BMI Mortality Collaboration, Di Angelantonio E, Bhupathiraju Sh N, Wormser D, Gao P, et al. (2016) Body-mass index and all-cause mortality: individual-participant-data meta-analysis of 239 prospective studies in four continents. Lancet 388: 776–786. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Bhaskaran K, Douglas I, Forbes H, dos-Santos-Silva I, Leon DA, et al. (2014) Body-mass index and risk of 22 specific cancers: a population-based cohort study of 5·24 million UK adults. Lancet 384: 755–765. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Miyake T, Kumagi T, Hirooka M, Furukawa S, Koizumi M, et al. (2013) Body mass index is the most useful predictive factor for the onset of nonalcoholic fatty liver disease: a community-based retrospective longitudinal cohort study. J Gastroenterol 48: 413–422. [DOI] [PubMed] [Google Scholar]
- 14.Khan SS, Ning H, Wilkins JT, Allen N, Carnethon M, et al. (2018) Association of body mass index with lifetime risk of cardiovascular disease and compression of morbidity. JAMA Cardiol 3: 280–287. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Dwivedi AK, Dubey P, Cistola DP, Reddy SY (2020) Association between obesity and cardiovascular outcomes: updated evidence from meta-analysis studies. Curr Cardiol Rep 22: 25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Polemiti E, Baudry J, Kuxhaus O, Jäger S, Bergmann MM, et al. (2021) BMI and BMI change following incident type 2 diabetes and risk of microvascular and macrovascular complications: the EPIC-Potsdam study. Diabetologia 64: 814–825. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Narayan KM, Boyle JP, Thompson TJ, Gregg EW, Williamson DF (2007) Effect of BMI on lifetime risk for diabetes in the U.S. Diabetes Care 30: 1562–1566. [DOI] [PubMed] [Google Scholar]
- 18.Ye M, Robson PJ, Eurich DT, Vena JE, Xu JY, et al. (2018) Changes in body mass index and incidence of diabetes: a longitudinal study of Alberta’s Tomorrow Project Cohort. Prev Med 106: 157–163. [DOI] [PubMed] [Google Scholar]
- 19.Salesi R, Kermani-Alghoraishi M, Sadeghi A, Roohafza H, Talaei M, et al. (2023) Body weight changes and diabetes mellitus incident: a cohort study from the Middle East. J Prev Med Hyg 64: E345–E351. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Asgari S, Khalili D, Azizi F, Hadaegh F (2024) Weight change and risk of incident type 2 diabetes: short, medium and long-term follow-up in tehran lipid and glucose study. Cardiovasc Diabetol 23: 207. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Meigs JB, Wilson PW, Fox CS, Vasan RS, Nathan DM, et al. (2006) Body mass index, metabolic syndrome, and risk of type 2 diabetes or cardiovascular disease. J Clin Endocrinol Metab 91: 2906–2912. [DOI] [PubMed] [Google Scholar]
- 22.Ohno R, Kaneko H, Ueno K, Aoki H, Okada A, et al. (2023) Association of body mass index and its change with incident diabetes mellitus. J Clin Endocrinol Metab 108: 3145–3153. [DOI] [PubMed] [Google Scholar]
- 23.American Diabetes Association (2021) 2. Classification and diagnosis of diabetes: standards of medical care in diabetes-2021. Diabetes Care 44: S15–S33. [DOI] [PubMed] [Google Scholar]
- 24.Todo Y, Miyake T, Furukawa S, Matsuura B, Ishihara T, et al. (2022) Combined evaluation of Fibrosis-4 index and fatty liver for stratifying the risk for diabetes mellitus. J Diabetes Investig 13: 1577–1584. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Heianza Y, Kato K, Kodama S, Ohara N, Suzuki A, et al. (2015) Risk of the development of Type 2 diabetes in relation to overall obesity, abdominal obesity and the clustering of metabolic abnormalities in Japanese individuals: does metabolically healthy overweight really exist? The Niigata Wellness Study. Diabet Med 32: 665–672. [DOI] [PubMed] [Google Scholar]
- 26.Zhou X, Ji L, Luo Y, Han X, Zhang X, et al. (2009) Risk factors associated with the presence of diabetes in Chinese communities in Beijing. Diabetes Res Clin Pract 86: 233–238. [DOI] [PubMed] [Google Scholar]
- 27.Garcia-Compean D, Jaquez-Quintana JO, Gonzalez-Gonzalez JA, Maldonado-Garza H (2009) Liver cirrhosis and diabetes: risk factors, pathophysiology, clinical implications and management. World J Gastroenterol 15: 280–288. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Bae JC, Cho NH, Kim JH, Hur KY, Jin SM, et al. (2020) Association of body mass index with the risk of incident type 2 diabetes, cardiovascular disease, and all-cause mortality: a community-based prospective study. Endocrinol Metab (Seoul) 35: 416–424. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Bae YJ, Shin SJ, Kang HT (2022) Body mass index at baseline directly predicts new-onset diabetes and to a lesser extent incident cardio-cerebrovascular events, but has a J-shaped relationship to all-cause mortality. BMC Endocr Disord 22: 123. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Yu HJ, Ho M, Liu X, Yang J, Chau PH, et al. (2022) Association of weight status and the risks of diabetes in adults: a systematic review and meta-analysis of prospective cohort studies. Int J Obes (Lond) 46: 1101–1113. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary Tables