Mortality trends in diabetes with proxy-defined steatotic liver disease in the United States, 1999-2023

To characterize U.S. mortality trends for diabetes mellitus (DM), proxy-defined metabolic dysfunction-associated steatotic liver disease (MASLD), and deaths listing both conditions, and to present exploratory conditional overlap extrapolations through 2040.

We analyzed CDC WONDER multiple-cause-of-death data from 1999-2023 among adults aged ≥25 years. DM was identified by ICD-10 codes E10-E14 and steatotic liver disease by K75.8 or K76.0. Because clinical MASLD cannot be directly ascertained in CDC WONDER, MASLD was defined using a death-certificate proxy combining steatotic liver disease codes with recorded cardiometabolic conditions. Age-adjusted mortality rates were assessed using Joinpoint regression. Prophet models generated conditional extrapolations through 2040. Sensitivity analyses used K76.0-only coding, K75.8-only coding, restriction to 1999-2019, and an underlying-cause-of-death-only analysis requiring K76.0 to be listed as the underlying cause of death.

DM-involved mortality changed little from 1999 to 2023 (118.47 to 122.61 per 100,000; average annual percent change [AAPC], -0.03%). In contrast, overlap mortality increased from 0.08 to 0.84 per 100,000 (AAPC, 11.20%), and proxy-defined MASLD mortality increased from 0.13 to 1.29 per 100,000 (AAPC, 9.83%), although absolute rates remained low. Sensitivity analyses using K76.0-only coding, restriction to 1999-2019, and an underlying-cause-of-death-only definition showed directionally consistent increases, whereas K75.8-only overlap deaths were sparse, zero, or suppressed. The underlying-cause-only analysis also showed a directionally consistent increase from 2000 to 2023 (AAPC, 14.46%; 95% CI, 13.46 to 21.51). Conditional extrapolations suggested continued growth through 2040 but were inherently uncertain.

Proxy-based mortality involving co-recorded DM and steatotic liver disease increased substantially in U.S. death-certificate data. These findings were directionally consistent across alternative coding, period-restricted, and underlying-cause-only sensitivity analyses but remain limited by under-ascertainment, misclassification, and evolving death-certificate documentation. They should be interpreted as descriptive surveillance patterns rather than evidence of biological expansion or causal mechanisms.