Association between body roundness index and the prevalence of diffuse idiopathic skeletal hyperostosis in elderly patients: the Bunkyo health study

Abstract

Diffuse idiopathic skeletal hyperostosis (DISH) is characterized by calcification and ossification of soft tissues. Previous studies have explored its links with diabetes, obesity, and visceral fat accumulation; however, no study has specifically evaluated its relationship with body roundness index (BRI). This cross-sectional analysis used baseline data from a prospective cohort of Japanese residents to investigate the association between DISH and the BRI, visceral fat volume, insulin resistance, and bone metabolic markers in an elderly urban Japanese population. We included 1,258 participants (535 men and 723 women, aged 65–84 years) from Bunkyo-ku, Tokyo. All participants underwent anthropometric measurements, InBody assessment, lateral thoracolumbar radiography, and blood sampling for bone metabolic markers. BRI was calculated using height and waist circumference. Participants with and without DISH were compared for BRI, visceral fat volume, and bone metabolic markers. Multivariate analyses were conducted to assess independent associations with DISH. DISH was identified in 194 participants (15.4%). Multivariate analysis demonstrated that higher BRI and visceral fat volume were independently associated with DISH (odds ratios: 1.5 and 1.01, respectively). We concluded that higher BRI and visceral fat volume were significantly associated with DISH in this elderly Japanese population, whereas bone metabolic markers showed no significant differences.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-47398-6.

Introduction

Diffuse Idiopathic Skeletal Hyperostosis (DISH) is characterized by continuous bony bridges spanning at least three vertebral levels at ligamentous and tendinous attachment sites, primarily along the anterior longitudinal ligament of the spine^1,2. Clinically, cervical lesions may cause dysphagia, whereas thoracic and lumbar lesions often lead to chronic back pain, restricted mobility, and reduced quality of life^3,4. DISH is also associated with an increased risk of post-fracture motor decline, secondary neurological injury, and postoperative implant failure⁴.

Patients with DISH exhibit significantly elevated fasting insulin levels and homeostasis model assessment of insulin resistance (HOMA-IR) compared with healthy controls^5,6, suggesting that adipose tissue–derived cytokines may contribute to DISH pathogenesis by modulating bone metabolism. Moreover, DISH has been linked to body mass index (BMI), waist circumference, and particularly visceral fat accumulation⁷. However, individuals with similar BMI values can have markedly different fat distributions and body compositions.

To address this limitation, Thomas et al. proposed the Body Roundness Index (BRI) as a more accurate measure of fat distribution⁸. BRI models the human body as an ellipse, using its eccentricity to estimate visceral and total body fat percentages. Calculated from height and waist circumference, BRI provides a more comprehensive assessment of visceral fat than BMI. Despite growing evidence linking visceral adiposity to DISH, no large-scale studies have investigated the relationship between BRI and DISH. Therefore, in this study, we aimed to elucidate factors associated with DISH prevalence, with a focus on BRI, using integrated blood tests and imaging evaluations in an elderly Japanese population.

Materials and methods

Study design

This study utilized data from the Bunkyo Health Study⁹, a 10-year prospective cohort of community-dwelling elderly residents in Bunkyo Ward, Tokyo, Japan. The cohort included 1,629 individuals aged 65–84 years, randomly selected from the local resident registry. After excluding 371 participants with missing radiographs, 1,258 participants were included in the final analysis. The dataset comprised general blood test results, muscle strength, lumbar spine and femoral neck bone mineral density, cognitive function, visceral fat volume, and lateral thoracolumbar radiographs. The Hosmer–Lemeshow test indicated adequate model fit (p = 0.12).

The study was approved by the Ethics Committee of Juntendo University (approval number: E24-0156-M01, July 2024) and conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent and were informed of their right to withdraw at any time (Fig. 1).

Fig. 1.

ROC curves for BRI, BMI, and VFA. Abbreviations: ROC, Receiver operating characteristic; AUC, Area under the curve.

Evaluation of DISH

In total, 1,258 individuals (mean age 73.1 ± 5.4 years; age range 65–84, 535 men and 723 women) underwent lateral thoracolumbar radiography to assess the presence of DISH. Diagnosis followed the Resnick/Niwayama classification¹⁰, which defines DISH as the presence of bony bridging across three or more contiguous vertebral levels.

Two spine specialists (D.F. and Y.S.) independently evaluated all 1,258 participants under blinded conditions to assess diagnostic reliability. Cohen’s κ coefficient (κ = 0.82) indicated strong inter-observer agreement. Discrepant cases were resolved through a consensus discussion between the two spine surgeons.

BRI measurement

BRI was calculated as 364.2 − 365.5 × √(1 − [waist circumference in centimeters/2π]²/[0.5 × height in centimeters]²), according to the formula developed by Thomas et al.⁸. BRI was categorized into four groups based on the 25th, 50th, and 75th percentiles to investigate its association with DISH prevalence.

Evaluation of biomarker and insulin sensitivity

Fasting blood and urine samples were collected in the morning after an overnight fast. In addition to standard biochemical tests, 25-hydroxyvitamin D (25(OH)D), bone-specific alkaline phosphatase (BAP), intact parathyroid hormone (iPTH), tartrate-resistant acid phosphatase 5b (TRACP-5b), adiponectin, and saturated fatty acids were measured.

Insulin resistance was evaluated using the HOMA-IR and Matsuda index, calculated as previously described¹¹. The Matsuda index was classified into three categories: < 2.5, 2.5–10, and 10 <. While a value < 3.0 generally indicates normal insulin sensitivity, interpretation should consider differences in calculation methods.

Measurement of visceral and subcutaneous fat areas

Visceral and subcutaneous fat areas were measured using a 0.3 Tesla MRI scanner (AIRIS Vento, Hitachi, Japan) following established protocols⁹. T1-weighted axial scans were obtained at the L4–L5 vertebrae level, and fat areas were quantified using AZE Virtual Place software (Canon Medical Systems Corporation, Japan).

Other measurement

Skeletal muscle mass (SMM) and femoral neck T-scores were measured using dual-energy X-ray absorptiometry (DXA) (Discovery DXA System, Hologic, Tokyo, Japan). Waist circumference (WC) was measured with participants standing upright, feet evenly spaced, and abdomen relaxed. The tape measure was positioned horizontally at the umbilical level.

Statistical analysis

Statistical analyses were performed using GraphPad Prism version 10.2.2 (341). Continuous variables were compared between groups using the independent samples t-test, whereas categorical variables were analyzed using the chi-squared test. The prevalence of DISH was evaluated relative to age, sex, and comorbidities.

Multivariate logistic regression was used to identify independent predictors of DISH. The explanatory variables in multivariate models included risk factors identified in previous studies¹². Variables with P < 0.05 in univariate analyses were included in the multivariate models, with age and sex used as baseline covariates. The use of antidiabetic and antilipid medications was also included as an adjustment factor. Subgroup analyses were conducted according to sex.

Results

Participants were classified according to their DISH status. The overall prevalence of DISH was 27.5% in men and 6.5% in women, with men accounting for 74.6% of all DISH cases. The BRI was significantly higher in participants with DISH. In addition, the prevalence of smoking history, hypertension, diabetes, and cardiovascular disease was significantly greater in the DISH group (Table 1).

Table 1.

Comparison of clinical characteristics between patients with and without diffuse idiopathic skeletal hyperostosis.

	Over all			Male			Female
	DISH n = 194	Non DISH n = 1064	p value	DISH n = 147	Non DISH n = 388	p value	DISH n = 47	Non DISH n = 676	p value
Sex, men, n(%)	147 (74.6%)	388 (36.4%)	< 0.001
Age(years)	74.0 ± 5.5	72.8 ± 5.4	0.004	74.0 ± 5.7	72.5 ± 5.2	0.004	74.3 ± 5.2	73.0 ± 5.5	0.120
65–70	64 (14.6%)	435 (85.4%)	< 0.001
71–75	55 (15.6%)	294 (84.4%)	< 0.001
76–80	42 (16.1%)	219 (83.9%)	< 0.001
81-	33 (22.1%)	116 (77.9%)	< 0.001
BRI	4.6 ± 1.3	4.4 ± 1.3	0.021	5.4 ± 1.8	4.6 ± 1.4	0.002	4.4 ± 1.0	4.0 ± 0.9	< 0.001
BMI	24.3 ± 3.0	22.6 ± 3.0	< 0.001	24.5 ± 2.9	23.2 ± 2.6	< 0.001	23.8 ± 3.2	22.3 ± 3.1	0.001
VFA	99.3 ± 44.0	75.2 ± 36.3	< 0.001	102.8 ± 45.3	85.4 ± 36.1	< 0.001	88.4 ± 38.3	69.4 ± 35.1	0.001
WC (cm)	83.8 ± 8.7	77.1 ± 8.0	< 0.001	85.4 ± 8.6	81.0 ± 7.4	< 0.001	78.7 ± 7.2	74.9 ± 7.4	0.001
Comorbidity (%)
smoking	113 (58.2)	410 (38.5)	< 0.001	106 (72.1)	295 (76.0)	0.322	7(14.9)	114 (16.9)	0.723
hypertension	113 (58.2)	469 (44.1)	< 0.001	87 (59.2)	195 (50.3)	0.064	26(55.3)	274 (40.5)	0.049
Diabetes	3 6(18.6)	110 (10.3)	0.002	28 (19.0)	58 (14.9)	0.256	8(17.0)	52 (7.7)	0.045
Dyslipidemia	66 (34.0)	438 (41.2)	0.060	44 (30.0)	138 (35.6)	0.216	22(46.8)	300 (44.4)	0.746
Stroke	6 (3.1)	38 (3.6)	0.735	5 (3.4)	15 (3.9)	0.799	1(2.1)	23(3.4)	0.616
cardiovascular	16 (8.2)	42 (3.9)	0.015	14 (9.5)	23 (5.9)	0.156	2(4.2)	19(2.8)	0.592
Cancer	27 (13.9)	137 (12.9)	0.694	22 (15.0)	59(15.2)	0.945	5(10.6)	78(11.5)	0.850

Data are expressed as mean ± SD or N (%).

The prevalence of DISH showed a significant difference across BRI quartiles. The prevalence across the three Matsuda Index groups showed a significant difference in the overall population and in men, but not in women (supplemental Table 1).

Significant differences were observed for BMI and Visceral fat areas (VFA), whereas subcutaneous fat areas (SFA) showed no significant difference. Among biochemical parameters, participants with DISH exhibited significant differences in serum calcium, phosphorus, adiponectin, low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), triglycerides (TG), HOMA-IR, Matsuda index, and insulin-like growth factor-1 (IGF-1). However, no significant differences were observed for creatinine, albumin, 25(OH)D, TRACP-5b, BAP, saturated fatty acid, HbA1c, or high-sensitivity C-reactive protein (hs-CRP) levels (Table 2).

Table 2.

Univariate analysis between diffuse idiopathic skeletal hyperostosis and various factors.

Variable	Odds	95% CI	p value
BRI	1.215	1.058 to 1.398	0.006
BMI	1.197	1.138 to 1.260	< 0.001
VFA	1.014	1.011 to 1.018	< 0.001
SFA	1.000	0.998 to 1.003	0.759
Femoral neck T score	1.528	1.369 to 1.708	< 0.001
Lab data
Cre	0.766	0.358 to 1.375	0.451
Alb	1.000	0.999 to 1.000	0.806
Ca	0.552	0.340 to 0.888	0.015
P	0.506	0.356 to 0.716	< 0.001
25(OH)D	0.994	0.964 to 1.025	0.720
BAP	0.987	0.952 to 1.020	0.433
iPTH	1.002	0.993 to 1.009	0.709
TRACP-5b	1.001	1.000 to 1.000	0.239
Adiponectin	0.911	0.881 to 0.940	< 0.001
Saturated fatty acids	0.999	0.998 to 1.001	0.566
LDL-C	0.992	0.987 to 0.997	0.002
HDL-C	0.975	0.964 to 0.985	< 0.001
TG	1.003	1.000 to 1.005	0.043
HbA1c(%)	1.074	0.827 to 1.373	0.581
HOMA-IR	1.403	1.229 to 1.604	< 0.001
Matsuda Index	0.672	0.553 to 0.815	< 0.001
IGF-1	1.006	1.001 to 1.010	0.018
hsCRP	1.000	1.000 to 1.000	0.221

Abbreviation: CI, confidence interval.

Multivariate logistic regression analysis, adjusted for sex, age and medication use, identified significant associations between DISH and several variables, including BRI, BMI, VFA, Matsuda index, HOMA-IR, HDL-C, adiponectin, IGF-1, femoral neck T-score, and SMM (Table 3).

Table 3.

Multivariate logistic regression analysis of each parameter for the prevalence of diffuse idiopathic skeletal hyperostosis, adjusted for age, sex, and the use of antidiabetic and lipid-lowering medications. (Crude OR (adjustment for sex, age) and Adjusted OR (adjustment for sex, age, antidiabetic and lipid-lowering medications)).

	Crude OR	95% CI	p value	Adjusted OR	95% CI	p value
BRI	1.504	1.275 to 1.782	< 0.001	1.501	1.273 to 1.778	< 0.001
BMI	1.176	1.113 to 1.244	< 0.001	1.169	1.107 to 1.236	< 0.001
VFA	1.011	1.007 to 1.015	< 0.001	1.011	1.007 to 1.015	< 0.001
HbA1c	1.047	0.801 to 1.350	0.733	0.877	0.624 to 1.216	0.437
Matsuda-Index	0.955	0.915 to 0.993	0.018	0.957	0.918 to 0.994	0.029
HOMA-IR	1.444	1.175 to 1.780	< 0.001	1.357	1.179 to 1.566	< 0.001
HDL-C	0.989	0.979 to 1.000	0.048	0.989	0.978 to 0.999	0.041
LDL-C	0.998	0.991 to 1.003	0.395	0.998	0.993 to 1.003	0.451
TG	1.002	0.999 to 1.004	0.180	1.002	0.999 to 1.004	0.154
Adiponectin	0.946	0.913 to 0.978	< 0.001	0.949	0.917 to 0.980	0.002
hsCRP	1.000	1.000 to 1.000	0.521	1.000	1.000 to 1.000	0.482
IGF-1	1.006	1.001 to 1.010	0.024	1.005	1.000 to 1.010	0.036
Femoral neck T score	1.209	1.055 to 1.387	0.006	1.215	1.060 to 1.393	0.005
Skeletal muscle mass	1.069	1.008 to 1.134	0.026	1.068	1.007 to 1.132	0.028

Abbreviations: OR, Odds Ratio; CI = confidence interval; T score, Femoral neck T score

In sex-stratified analyses (Table 4), significant associations in men were observed for BRI, VFA, Matsuda index, HOMA-IR, adiponectin, femoral neck T-score, and SMM. Among women, significant differences were found for BRI, BMI, VFA, Matsuda index, HOMA-IR, HDL-C, adiponectin, IGF-1, femoral neck T-score, and SMM.

Table 4.

Subgroup analyses of each parameter for men and women, adjusted for age and sex.

	Men			Women
	OR	95%CI	p value	OR	95% CI	p value
BRI	1.562	1.278 to 1.918	< 0.001	1.394	1.052 to 1.885	0.020
BMI	1.195	1.112 to 1.288	< 0.001	1.140	1.044 to 1.243	0.003
VFA	1.582	1.218 to 2.075	0.001	1.886	1.293 to 2.795	0.001
Matsuda-Index	0.695	0.546 to 0.882	0.003	0.633	0.425 to 0.928	0.021
HOMA-IR	1.407	1.109 to 1.793	0.005	1.540	1.056 to 2.282	0.027
HDL-C	0.995	0.982 to 1.007	0.381	0.973	0.952 to 0.993	0.012
LDL-C	0.995	0.9891 to 1.002	0.164	1.004	0.994 to 1.014	0.401
TG	1.002	0.999 to 1.005	0.265	1.003	0.997 to 1.007	0.329
Adiponectin	0.956	0.914 to 0.997	< 0.001	0.938	0.885 to 0.987	0.021
hsCRP	1.103	0.877 to 1.390	0.405	1.193	0.816 to 1.747	0.362
IGF-1	0.997	0.990 to 1.004	0.412	1.825	1.253 to 2.698	0.002
Femoral neck T score	1.168	1.001 to 1.366	0.049	1.378	1.039 to 1.811	0.023
Skeletal muscle mass	1.077	1.010 to 1.149	0.023	1.027	0.885 to 1.188	0.722

Abbreviations: OR, Odds Ratio; CI = confidence interval.

Discussion

This study is the first to demonstrate a clear association between the BRI, an easily obtained surrogate for visceral adiposity, and the prevalence of DISH. While the relationship between DISH and visceral fat volume or BMI is well established, no study has evaluated the combined association of visceral fat volume and WC with DISH. In this study, we identified associations between BMI and BRI, as well as insulin resistance, lipid profiles, bone mineral density, and muscle mass. Similar to that of previous reports, the prevalence was significantly higher in men than in women and increased with age.

BRI is a reliable indicator of visceral adiposity. The significant correlation between BRI and DISH observed in this study suggests that visceral fat accumulation may contribute to DISH pathogenesis. High BMI is a risk factor for DISH^5,13. Our multivariate logistic regression analysis further demonstrated that a higher BRI was more strongly associated with DISH than BMI, as reflected by the higher 95% confidence interval for BRI. The discriminative ability of BRI, BMI, and VFA was moderate, with no significant differences observed among the individual indices (Figure 1). Due to the characteristics of their calculation formulas, BMI increases when either subcutaneous or visceral fat is elevated, whereas BRI tends to increase particularly when visceral fat is high. Therefore, BRI may be more suitable for the simple assessment of visceral fat accumulation, and its inherent characteristics suggest that it may have superior latent discriminative performance compared with other indices. Furthermore, BRI offers a practical advantage in clinical settings, as it enables estimation of visceral fat without the need for MRI, thereby serving as a useful tool in routine practice. This finding supports the hypothesis that BRI may better capture obesity-related metabolic risk relevant to DISH development.

DISH is highly comorbid with metabolic syndrome conditions such as hypertension, diabetes, and dyslipidaemia^14,15. Visceral fat volume is strongly linked to these metabolic abnormalities¹⁶, suggesting that visceral adiposity may also play a key role in DISH pathogenesis.

The prevalence of DISH appears higher in Japan than in Western countries^17,18. Japanese individuals are known to exhibit a lower tolerance for visceral fat accumulation and a greater susceptibility to metabolic disorders than Western populations^18,19. Even at comparable BMI and WC, Asians tend to accumulate more visceral fat, which may explain their higher susceptibility to DISH²⁰. The BRI has also been reported as a useful predictor of diabetes and prediabetes, suggesting its superiority over BMI as a measure of central obesity²¹.

In our study, participants with DISH exhibited significantly higher HOMA-IR and Matsuda index values and lower adiponectin levels. According to Naito et al., visceral fat volume and adiponectin are key determinants of insulin resistance²². Our findings align with this, as the DISH group showed lower adiponectin and higher VFA. In contrast, SFA was not associated with DISH, likely due to metabolic and functional differences between visceral and subcutaneous adipose tissue²³. The strong link between visceral fat and insulin resistance²² supports the hypothesis that DISH development is closely related to visceral fat accumulation.

Mader et al. reported that patients with low HDL and high TG levels were more likely to develop DISH^15,24. However, our multivariate and subgroup analyses did not confirm significant associations between lipid parameters and DISH, except for lower HDL levels among women. Similar findings were reported by Okada et al., who found that insulin resistance (HOMA-IR), rather than lipid profiles, was significantly associated with DISH⁵.

IGF-1 is a key regulator of bone metabolism, coordinating the coupling of osteoblast and osteoclast activity and thereby influencing ossification²⁵. The biological activity of IGF-1 is regulated through complex interactions among the IGF-1 receptor/insulin receptor signaling pathway²⁶, the Wnt pathway²⁷, and phosphate metabolism mediated by FGF-23²⁸. Reflecting this complexity, although many studies have reported elevated IGF-1 levels in DISH, the findings are not entirely consistent. In the present study, a significant difference was observed only in women with DISH; however, the underlying reasons for this sex-specific association remain to be elucidated.

Previous studies have not reached a consistent conclusion regarding the association between SMM and the prevalence of DISH. SMM promotes ossification through multiple mechanisms, including enhanced ossification at the enthesis via muscle–tendon attachments², the formation of a bone-forming environment within muscle tissue (such as trauma-²⁹ or postoperative-related heterotopic ossification³⁰), and secondary involvement mediated by qualitative deterioration of the paraspinal extensor muscles³¹. Despite these mechanistic differences, these pathways do not fully explain the observed sex-specific differences in the present study, and further analyses are needed.

This study is the first to concurrently demonstrate associations between BRI and visceral fat volume with DISH. The strong correlation between these indices highlights BRI as a practical and informative tool for evaluating DISH risk and for identifying individuals at high risk for disease development.

Regarding bone parameters, significant differences were observed in bone mineral density between the lumbar spine and hip. Lumbar BMD is often overestimated owing to ossified lesions³²; therefore, we analyzed hip BMD, which was also elevated in the DISH group. Similar results were reported by Uehara et al.³³. Given that DISH involves ossification at spinal entheses and peripheral tendons and ligaments, increased BMD at the hip is a plausible finding.

Adiponectin contributes to bone formation by promoting osteoblast differentiation and inhibiting osteoclast formation. When insulin resistance increases, adiponectin levels decrease and insulin secretion increases. In this study, patients with DISH had lower adiponectin levels. The mechanisms by which low adiponectin, high insulin, or reduced insulin sensitivity enhance bone formation and lead to DISH remain complex and unclear.

Previous studies have reported inconsistent findings regarding bone turnover markers in DISH. While Ca and P levels differed significantly between groups, we found no significant differences in 25(OH)D, BAP, iPTH, and TRACP-5b. These discrepancies suggest that systemic bone turnover markers may not adequately reflect local ossification activity. Adiponectin’s known role in promoting osteoblastogenesis and inhibiting osteoclastogenesis supports our observation of altered metabolic patterns in DISH^34,35.

Recently, epidemiologic data on DISH have been reported mainly from Asian countries with wide variations in prevalence rates, such as 30.8% in Pakistan³⁶, 24.4% in Korea³⁷, 10% in Oman³⁸, 3.8% in China³⁹, 13.2% in the United States¹⁷ and 7.8% in Iceland⁴⁰. The prevalence of DISH in this study (15.4%) was slightly higher than that in Europe and the United States.

In a similar study in Japan, Ohshima et al. reported a prevalence of approximately 18% (men, 27.7%; women, 12.2%), with DISH being more common in men⁴¹. Our findings that men had higher BRI values likely explain the higher prevalence of DISH among them. Differences in ethnicity, culture, lifestyle, and visceral fat tolerance may underlie the global variation in DISH prevalence.

Age remains a major determinant of DISH. Our study revealed that the prevalence of DISH increased with age. Oshima et al. reported that the prevalence of DISH increases with age and that it may develop as early as in the 40s³⁴. Kim et al. and Ciaffi et al. also reported a strong positive correlation between age and prevalence of DISH^37,42. In all studies, the prevalence of DISH increased with age. The higher prevalence in older populations may be influenced by age-related worsening of insulin resistance and reduced spinal mobility⁴³.

Study limitations

This study has some limitations. First, it was conducted in a single urban district (Bunkyo Ward, Tokyo), and only 13 of 68 communities were sampled⁹, which may limit generalizability and result in a different obesity profile than in rural populations⁴⁴. The exclusion of incomplete questionnaires may also have introduced selection bias⁹. 371 excluded patients (22%) has the risk of introducing selection bias and should be taken into account. Moreover, this study did not include participants younger than 65 years, potentially overestimating the prevalence of DISH compared with studies that included younger age groups. 371 excluded patients (22%) has the risk of introducing selection bias and should be taken into account.

Second, the cross-sectional design limits causal inference, as DISH develops gradually over time, whereas blood tests represent a single time point. While this study adjusted for medication history, the duration and continuity of medication use were not assessed, potentially influencing the observed associations. Furthermore, in the present study, medication use history obtained based on questionnaire data was incorporated as covariates into each model, and multivariate analyses were performed. This approach may explain why the level of statistical significance in Table 3 did not change substantially before and after adjustment for medication use. Considering these findings, further longitudinal studies are needed to clarify causal relationships.

Conclusion

This study highlights the multifactorial etiology of DISH and its associations with age, sex, obesity, and insulin resistance. We demonstrated a strong relationship between DISH prevalence and BRI, a surrogate for visceral adiposity. Our findings suggest that BRI is a useful indicator for visceral fat accumulation in the context of DISH. Because the impact of visceral fat on ossification likely varies by ethnicity, sex, and age, multiple determinants contribute to DISH development. Further longitudinal and multiethnic studies are warranted to elucidate causal mechanisms and improve the generalizability of these findings.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Y.S. wrote the main manuscript text. H.N. and S.T. provided guidance on the structure and content of the manuscript. Y.T. provided data from the Bunkyo Health Study. M.N. contributed to the statistical analyses. M.I. critically reviewed the entire manuscript. All authors reviewed the manuscript.

Data availability

The data that support the findings of this study were obtained from the Bunkyo Health Study. Due to ethical and privacy restrictions, these data are not publicly available. The data are available from the corresponding author upon reasonable request and with permission from the Bunkyo Health Study.

Declarations

Competing interests

The authors declare no competing interests.