Japan’s Diabetes Divide: Men Face Rising Burden While Women Buck the Trend Through 2040

Japan is staring down a quietly diverging diabetes epidemic, according to a new study published in The Lancet Regional Health – Western Pacific. Using two decades of nationally representative health data and a sophisticated Bayesian age-period-cohort (BAPC) modeling framework, researchers projected that diabetes mellitus prevalence in Japanese men will climb from 16.4 percent in 2020 to 18.1 percent by 2040, while prevalence among women is expected to fall from 9.6 percent to 7.7 percent over the same period. The study, led by Rajib Mondal, Soshiro Ogata, and colleagues, is the first in the Western Pacific region to produce comprehensive sex-specific diabetes projections while explicitly accounting for the intertwined effects of age, calendar time, and birth cohort. The findings suggest that Japan’s total diabetes burden will grow to roughly 10.4 million cases by 2040, driven overwhelmingly by men and by the relentless demographic aging of one of the world’s oldest populations.

The modeling effort drew on the National Health and Nutrition Survey (NHNS), an annual cross-sectional household examination survey that has tracked the diet, physical activity, lifestyle, anthropometrics, and laboratory profiles of approximately 8,000 to 11,000 Japanese residents aged one year and older since its restructuring in 2003. The team focused on glycemic data from adults aged 40 and above, defining diabetes as glycated hemoglobin A1c at or above 6.5 percent under the National Glycohemoglobin Standardization Program, or self-reported use of glucose-lowering medication. This definition corresponds to the official NHNS category of ‘strongly suspected diabetes’ rather than strictly diagnosed disease, a choice the authors argue is well suited to capturing the full population-level burden for surveillance and long-term projection. HbA1c values measured under earlier Japanese Diabetes Society guidelines were converted to NGSP equivalents using a standard linear formula, ensuring consistency across the entire 2003 to 2019 observation window.

What sets this study apart from earlier forecasting efforts is its treatment of three distinct temporal dimensions that conventional models typically ignore. Age effects capture how diabetes risk changes across the lifespan as glucose metabolism deteriorates with biological aging. Period effects reflect influences that operate across the entire population at a given calendar moment, such as expanded healthcare access, mass screening programs, and rising consultation rates. Cohort effects, the most subtle of the three, capture how people born in particular eras carry distinct lifetime risks shaped by early-life nutrition, wartime deprivation, and generational lifestyle patterns. Because age, period, and cohort are mathematically linearly dependent, the researchers relied on identifiable non-linear components, or curvatures, and fitted smoothing priors using first- and second-order random walks within a hierarchical Bayesian framework implemented in the R bamp package.

To test whether this added complexity actually buys predictive power, the team pitted the BAPC model against two conventional alternatives: one assuming constant prevalence frozen at 2019 levels, and another assuming monotonic change based on mean percentage shifts observed from 2003 to 2019. Splitting the data into a training period from 2003 to 2014 and a validation window from 2015 to 2019, they measured performance using root mean square error, or RMSE, which quantifies the average annual discrepancy between projected and observed case counts. The BAPC model achieved the lowest error for the total population and for both sexes, and its projections for 2022 through 2024 aligned more closely with the officially reported NHNS prevalence than either conventional model, providing a rare external validation of a long-range health forecast against real-world data.

The headline projections reveal a nation moving in two directions at once. Overall crude prevalence is projected to rise from 12.6 percent in 2020 to 14.0 percent in 2040, with total cases climbing from roughly 9.9 million to 10.4 million. Among men, the burden is heaviest in older age brackets: prevalence among those aged 70 to 74 is projected to reach 26.7 percent by 2040, and among those 75 to 79 it may hit 29.1 percent. Women show the opposite trajectory, with prevalence declining across nearly every age group, dropping most steeply among those aged 55 to 64. Notably, when the researchers age-standardized their estimates using the 2015 Japan Standard Population, the apparent rise in men largely flattened out, indicating that population aging, rather than worsening age-specific risk, is the dominant engine of the male increase.

A formal decomposition analysis confirmed this interpretation, attributing the growing prevalence in men to both rising age-specific prevalence and demographic aging, with the latter contributing more. The age-effect patterns estimated by the model showed an upward curvature in men that intensified until ages 65 to 69 and continued climbing thereafter, consistent with aging-related deterioration in glucose metabolism. The cohort effects told an equally revealing story. In men, prevalence rose gradually across successive generations, with modest upward curvature in older and intermediate birth cohorts. In women, the model detected downward curvature in intermediate and younger cohorts, suggesting that diabetes risk has been falling steeply among women born after the mid-twentieth century, a divergence the authors link partly to sex differences in body mass index trends and lifestyle factors.

The shadow of history looms over these cohort patterns. The intermediate birth cohort in the analysis, spanning those born between roughly 1940 and 1955, came of age amid the social upheaval of World War II and its aftermath, conditions that likely produced poor early-life nutrition. Prior research, including studies of the Dutch famine, has shown that nutritional deficiencies in early life can trigger elevated diabetes incidence decades later in adulthood. The researchers also point to more recent pressures: declining dietary fiber intake, reduced physical activity, and rising body mass index over the last decade, trends that have been notably more pronounced among Japanese men than women. Meanwhile, increasing healthcare access and consultation rates for diabetes may be raising detected and treated case counts, contributing to the estimated period effects even as serum glucose levels in the broader population have reportedly stabilized.

The projections come with important caveats. The 95 percent credible intervals widen dramatically as the forecast horizon extends, reaching as high as 40.8 percent for men by 2040, reflecting the genuine uncertainty inherent in predicting two decades into the future. The NHNS was suspended in 2020 and 2021 due to the COVID-19 pandemic, and the definition based on ‘strongly suspected diabetes’ may underestimate true prevalence by excluding the ‘diabetes cannot be ruled out’ category. The models also could not adjust for socioeconomic factors, lifestyle variables, or body mass index directly, and the 17-year observation window, constrained by the availability of standardized HbA1c data since 2003, is relatively short for long-horizon projection. Sensitivity analyses using stronger smoothing priors and down-sampled data for the expanded 2012 and 2016 survey cycles, however, produced results consistent with the main findings.

For policymakers, the message is unambiguous: Japan’s diabetes burden is set to grow, and it will fall disproportionately on men and the elderly. The authors urge a reform of public health policy encompassing strengthened primary care screening and early detection, targeted attention to the health needs of older adults and men, sustained monitoring of the disease burden, and preventive lifestyle interventions to head off complications. Previous Japanese modeling work has shown that rising diabetes prevalence partly offset reductions in coronary heart disease deaths achieved between 1980 and 2012 through medical advances and favorable risk factor changes, underscoring how easily gains in cardiovascular health can be eroded by a swelling diabetic population. As other nations in the Western Pacific grapple with similar demographic transitions, Japan’s experience, and the Bayesian tools used to forecast it, may offer a preview of the region’s metabolic future.

Subject of Research: Projection of diabetes mellitus prevalence and case numbers in Japan through 2040 using Bayesian age-period-cohort modeling of national survey data

Article Title: Projections of diabetes mellitus in Japan through 2040: a bayesian age-period-cohort analysis of the national health and nutrition survey population

Article References: Mondal, R., Ogata, S., Kiyoshige, E., Nakaoku, Y., Nakatsuka, K., Watanabe, T., Teramoto, K., Kadota, A., Miura, K., Kypridemos, C., Yoshikawa, Y., & Nishimura, K. (2026). Projections of diabetes mellitus in Japan through 2040: a bayesian age-period-cohort analysis of the national health and nutrition survey population. The Lancet Regional Health – Western Pacific, 74, Article 101988. https://doi.org/10.1016/j.lanwpc.2026.101988

Image Credits: AI Generated

DOI: 10.1016/j.lanwpc.2026.101988

Keywords: diabetes mellitus, Japan, Bayesian age-period-cohort model, prevalence projection, National Health and Nutrition Survey, population aging, cohort effects, sex differences, public health policy, HbA1c, epidemiology, The Lancet Regional Health Western Pacific