Non-alcoholic fatty liver disease (NAFLD), defined as the accumulation of hepatic steatosis in ≥5% of hepatocytes without heavy alcohol consumption, is currently the most prevalent chronic liver disease and is one of the leading causes of advanced liver disease worldwide (1, 2). It is currently estimated to affect approximately one-quarter of the global population in the USA (3). NAFLD is a hepatic manifestation of metabolic disease. Among metabolic diseases, diabetes mellitus (DM) has become a major medical problem in the 21st century. Diabetes is a metabolic disorder characterized by increased blood glucose levels following absolute or relative insulin deficiency, β-cell dysfunction, and insulin resistance (4). Such highly prevalent chronic diseases with chronic hyperglycemia are associated with long-term damage to blood vessels and multiple organs. Accumulating evidence has demonstrated that NAFLD is a multi-organ disease, suggesting a strong association between NAFLD and diabetes (5, 6). Diabetes and NAFLD are prevalent diseases that often coexist and can function synergistically to exacerbate the risk of hepatic and extrahepatic clinical outcomes. Recent epidemiological studies have revealed that NAFLD diagnosed using ultrasonography (US) could be used to estimate the development of incident diabetes, and patients with NAFLD have a higher risk of developing diabetes than those without NAFLD (5, 7–9). A recent meta-analysis including 20 studies revealed that NAFLD was related to an increased probability of incident diabetes during a median 5-year follow-up (10). In fact, factors involved in liver fat accumulation generally overlap with risk factors for diabetes. However, there is limited evidence regarding whether diabetes in NAFLD patients over a long period of time enhances the probability of all-cause and cardiovascular mortality.

The Third National Health and Nutrition Examination Survey (NHANES III), conducted between 1988 and 1994, including US data for NAFLD, is a well-designed population-based cohort with a large sample of adults in the USA. In the present study, we aimed to evaluate whether a history of diabetes could affect the risk of all-cause and cardiovascular mortality in patients with NAFLD compared with that in patients with NAFLD but without diabetes by analyzing 27-year follow-up outcome data.

In this nationally representative cohort study, we aimed to evaluate the impact of diabetes on mortality risk in individuals with NAFLD. We found that a history of diabetes was significantly associated with a higher risk of all-cause and cardiovascular mortality in individuals with NAFLD. These associations were independent of sociodemographic factors, medical history, laboratory investigation indicators, and degree of liver fibrosis. Furthermore, these results remained unchanged in stratification analyses. Stratified analysis according to sex, race, marital status, chronic bronchitis, congestive heart failure, hypertension, stroke, cancer, and degree of liver fibrosis confirmed that NAFLD with diabetes significantly increased the risk of all-cause and cardiovascular mortality in the age, BMI, waist circumference, poverty income ratio, albumin, CRP, HbA1c, cholesterol, triglycerides, HDL-c, glucose, blood urea nitrogen, creatinine, FIB-4 (continuous variable), AST, ALT, and urinary creatinine-adjusted model as compared with individuals without diabetes. An important potential source of bias in the observational epidemiological analysis is residual confounding. However, calculating the E-value supported the robustness of the main findings.

DM and NAFLD share common pathophysiological mechanisms, including insulin resistance and chronic inflammation (16–18). The link between diabetes and NAFLD is reflected by a series of metabolic changes, such as insulin resistance, triglyceride metabolism, and defective hepatic lipid profile, which results in fat accumulation, immune responses, and hyperinsulinemia induced by β-cell dysfunction in individuals with diabetes (19). Insulin resistance has been reported in 66%–83% of patients with NAFLD (20). NAFLD and insulin resistance share some common predisposing genetic variations, such as Apo C3 (16, 17). Hepatic insulin clearance is suppressed in individuals with diabetes and is related to the severity of metabolic syndrome (21). Insulin has anti- and pro-inflammatory properties (22). Moreover, a recent study revealed that serum insulin levels were strongly correlated with ballooning and hepatic lobular inflammation (23).

Several epidemiological studies have reported an association between diabetes and NAFLD. Recent evidence supports a more complex link between diabetes and NAFLD than was previously observed, exhibiting a bidirectional and mutual association between them (24). NAFLD has been reported to be a risk factor for type 2 DM (T2DM), and T2DM is a risk factor for liver disease in patients with NAFLD (6, 25, 26). A survey performed in Scotland revealed that, if a diabetic individual had ever been discharged with a hospital diagnosis of NAFLD, there was a greater risk of cardiovascular (HR 1.70 [95% CI 1.52–1.90]) and all-cause (HR 1.60 [95% CI 1.40–1.83]) mortality compared with those free of NAFLD (27). A recent study revealed that advanced liver fibrosis, measured using hepatic transient elastography, is a risk biomarker for cardiovascular mortality in individuals with T2DM and NAFLD. The prevalence of diabetes in individuals with NAFLD depends on the degree of NAFLD, ranging from 9.8% in mild NAFLD to 17.8% in moderate-to-severe NAFLD (28–30). NAFLD is reported as an independent risk factor for the development of diabetes, with a two-fold increase in incidence in these patients, and participants with non-alcoholic steatohepatitis have a more than three-fold higher risk of developing incident diabetes than those with simple steatosis (10, 28). A previous study included 132 patients with liver biopsy-confirmed NAFLD, and 44 (33%) had a history of diabetes (30). After multiple potential confounders (age, BMI, and cirrhosis) were adjusted, all-cause mortality was greater in patients with NAFLD and diabetes (relative risk 3.30 [95% CI 1.76–6.18]). The risk of liver-related mortality was higher in patients with NAFLD and diabetes. However, the main shortcoming was the selection bias introduced by the inclusion of individuals from a tertiary care center. In fact, patients undergoing liver biopsy demonstrated a higher HR for mortality, despite the fact that this was unaffected by age, sex, or date of diabetes diagnosis (31). This highlights the selection bias of non-community-based studies that depend on biopsy and the challenge in generalizing findings from participants who have undergone liver biopsy or liver US among the general population in the community. Furthermore, participants who undergo biopsy are more likely to have serious liver disease and, thus, exhibit a greater risk of liver-related mortality. Therefore, these studies overestimate mortality rates when used in the general community. Our study was based on a nationally representative cohort that can be generalized to the general population.

This mechanism between diabetes and NAFLD is complex and involves not only the liver but also the immune system, skeletal and cardiac muscle, brown and white adipose tissue, pancreas, vascular system, gut, and brain (32). The link between diabetes and NAFLD can be described by a spectrum of metabolic changes represented by insulin resistance, defective hepatic lipid profile, and triglyceride metabolism, which result in fat accumulation, immune responses, and hyperinsulinemia, as determined by β-cell dysfunction in T2DM (19). The link between diabetes and NAFLD further confirmed that diabetes aggravates the course of NAFLD and that, conversely, NAFLD will result in metabolic decompensation of diabetes (33). Another possible mechanism may be that ectopic liver fat is perhaps part of the pathogenic process in diabetes, resulting in hepatic insulin resistance, excess gluconeogenesis, and higher fasting glucose levels (34). Moreover, hepatic steatosis due to NAFLD leads to increased aminotransferase levels, with ALT levels exceeding those of AST (35). Increased ALT levels are prevalent in patients with T2DM, and for a specific serum ALT, those with T2DM exhibit more liver fat than those without diabetes (36).

The present study has several limitations, the first of which was that the identification of NAFLD status relied on a retrospective chart review of liver US results without biopsy evaluation/confirmation. However, liver biopsy is unethical for epidemiology studies of asymptomatic individuals in the NHANES III. It is the most comprehensive population-based database with ultrasound-controlled attenuation parameters to confirm hepatic steatosis using well-defined criteria (37). US is presently the most commonly available and non-invasive technique to diagnose hepatic steatosis in clinical practice. The sensitivity of US may vary depending on the hepatic fat content; when implemented properly, US has been uncovered to identify as little as ≥5% hepatic fat content (38). A meta-analysis that included 49 studies has displayed high accuracy of US (area under the curve (AUC) value of 0.93) for the diagnosis of moderate-to-severe steatosis (39). Second, due to the observational nature of the study, causation cannot be determined. Third, several variables were self-reported, which may have resulted in recall and reporting biases. Third, individuals were only queried about diabetes at baseline at a single time point; as such, any time-varying effect could not be investigated. Future in vitro and in vivo experiments should aim to elucidate the molecular mechanisms by which diabetes is associated with the prognosis of patients with NAFLD.

Notwithstanding these limitations, this study had several strengths. First, we used nationwide survey data representing the U.S. general population, and mortality data were linked to the NDI of the National Center for Health Statistics with a long follow-up period to observe mortality events, and ICD-10 codes were used to identify the cause of death. Second, the prospective design with a relatively large sample size and the use of a nationally representative sample of USA individuals with NAFLD facilitated generalization of the conclusions. Furthermore, several stratification analyses and calculations of E-values support the robustness of the findings.

In a large cohort of adults with NAFLD, we found that diabetes was significantly associated with higher all-cause and cardiovascular mortality. This link could be further characterized in future studies assessing the degree of glycemic control and its relationship with mortality in patients with T2DM and NAFLD.

Publicly available datasets were analyzed in this study. The raw data used in the article are available from National Health and Nutrition Examination Survey program (https://www.cdc.gov/nchs/nhanes/index.htm).

WW and XC are the principal investigators. WW and JX conducted the formal statistical analysis and data management. WW and JX wrote and XC edited and revised the manuscript. All authors read and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.