The current study is a secondary analysis of a large longitudinal cohort study conducted in China. Briefly, the longitudinal study aimed to assess chronic diseases and their risk factors. The original cohort consisted of 685,277 participants who underwent health examinations at the Rich Healthcare Group in 11 cities (Nantong, Wuhan, Hefei, Guangzhou, Chengdu, Changzhou, Shenzhen, Suzhou, Nanjing, Beijing, Shanghai) in China between 2010 and 2016. These participants underwent various measurements, including general physical parameter assessments, lifestyle evaluations, and blood tests as part of the medical examinations. The available data from the longitudinal cohort have been uploaded to the Dryad database for sharing by Chen et al. (21), and we utilized these data for our secondary analysis while respecting the authors’ rights. In a previous study, Chen et al. analyzed the relationship between body mass index (BMI) and diabetes, and the detailed research design and methods have been published elsewhere (22), in which subjects with the following characteristics were excluded from their study: (1) diagnosed with diabetes at baseline, (2) unknown glycemic status during follow-up, (3) follow-up duration less than 2 years, (4) missing or extreme values of sex, height, weight, BMI, and fasting plasma glucose (FPG), and (6) participants who withdrew from the study for unknown reasons. Ultimately, they included 211,833 participants for analysis (22). In the current study, we extracted the clinical data uploaded by Chen et al. from the Dryad database and further excluded participants with missing baseline lipid parameters (n=95,172) and ALT parameter data (n=410) based on the research objectives. Finally, we included 116,251 participants. Figure 1 illustrates the flowchart of this study. The current study is a secondary analysis, and the new research protocol has been authorized by the ethics committee of the authors’ institution (Jiangxi Provincial People’s Hospital Ethical Approval Number: 2021-067), which oversaw the entire research process to ensure compliance with the Helsinki Declaration. Additionally, the Ethics Committee of Jiangxi Provincial People’s Hospital waived the requirement to obtain informed consent, given that the current study data did not contain subject’ identity information.

All baseline data included in the database were collected through standardized questionnaires during the initial assessment, which encompassed age, sex, height, weight, systolic blood pressure (SBP), diastolic blood pressure (DBP), family history of diabetes, smoking, and drinking status. General physical parameters were measured by trained medical personnel in a quiet indoor environment, with weight measured to an accuracy of 0.1 kilograms and height measured to an accuracy of 0.1 centimeters.

Prior to each participant’s visit, they were required to fast for at least 10 hours. Venous blood samples were collected at the examination center, and professional healthcare personnel measured the levels of serum triglycerides (TG), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), HDL-C, FPG, ALT, aspartate transaminase (AST), blood urea nitrogen (BUN), and creatinine (Cr) using the Beckman 5800 Automatic Biochemical Analyzer

The start of follow-up was defined as the time of the initial health assessment, and the endpoint of follow-up was the diagnosis of incident diabetes or the end of the study (whichever occurred first). According to the diagnostic criteria of the American Diabetes Association, diabetes was defined as an FPG measurement ≥ 7.00 mmol/L during the follow-up period or self-reported diagnosis of diabetes (23).

Smoking status: At baseline data collection, participants were categorized into four groups based on their smoking history: non-smoker, former smoker, current smoker, and not recorded.

Drinking status: At baseline data collection, participants were categorized into four groups based on their alcohol consumption history: non-drinker, former drinker, current drinker, and not recorded.

BMI was calculated as weight divided by height squared, and participants were classified according to the BMI categories for the Chinese population: < 24 kg/m² (non-obese), ≥ 24 kg/m², and < 28 kg/m² (overweight), ≥ 28 kg/m² (obese) (24).

Data analysis was performed using R software version 3.4.3 and Empower(R) software version 2.20. A two-sided P-value < 0.05 was considered statistically significant.

Baseline characteristics were compared among groups defined by quintiles of the ALT/HDL-C ratio, including proportions of study participants, means, or medians. Differences in continuous variables among the groups were compared using one-way analysis of variance or the Kruskal-Wallis H test, while differences in categorical variables were compared using the chi-square test. Cumulative incidence curves for the primary endpoint in each quintile of the ALT/HDL-C ratio were estimated using the Kaplan-Meier method, and differences between groups were compared using the log-rank test.

To further explore the relationship between the ALT/HDL-C ratio and diabetes risk, the ALT/HDL-C ratio was included in Cox regression models both as a continuous variable and a categorical variable, and Hazard ratios (HRs) with 95% confidence intervals (CI) were recorded. Collinearity among covariates was assessed using multiple linear regression analysis, with the variance inflation factor used as an assessment tool for collinearity between covariates ( Supplementary Table 1 ) (25). Initially, an unadjusted Crude Model was evaluated and served as the reference for subsequent models. Model 1 was adjusted for sex, BMI, and age based on the Crude Model; to account for potential influences of family history and lifestyle habits factors on diabetes risk, they were further adjusted in Model 2. Finally, to assess the independent relationship between the ALT/HDL-C ratio and diabetes, potential influences of blood pressure (SBP, DBP), blood glucose (FPG), blood lipids (TG, LDL-C), BUN, and renal function (Cr) were further considered in Model 3. Furthermore, to mitigate the potential influence of undiagnosed liver diseases, we continued to exclude participants with abnormal liver function (baseline ALT or AST higher than 40 IU/L) in the current study and followed the aforementioned analytical procedures.

The predictive performance of the ALT/HDL-C ratio and its components (ALT and HDL-C) for diabetes was assessed using receiver operator characteristic (ROC) curves, and the corresponding area under the curves (AUCs) was calculated. Their performance was compared using the Delong test. Finally, considering the important influence of factors such as age, sex, BMI, and family history of diabetes on the risk of diabetes, we further performed subgroup analyses based on these factors in Model 3, and examined the differences between groups by likelihood ratio test.

After excluding participants who did not meet the criteria, a total of 116,251 baseline normoglycemic participants were included, with an average age of 44 years, including 62,622 males and 53,629 females. Table 1 presents the baseline characteristics of participants stratified by quintiles of the ALT/HDL-C ratio. From the table, it can be observed that as the quintiles of the ALT/HDL-C ratio increase, the proportion of male participants gradually increased, while the proportion of female participants gradually decreased. In addition, we found that except for HDL-C, which showed a decreasing trend, other indicators such as height, weight, BMI, SBP, DBP, FPG, TC, TG, LDL-C, ALT, AST, BUN, and Cr levels increased with increasing quintiles of the ALT/HDL-C ratio.

During a median follow-up period of 3.10 ± 0.95 years, a total of 2,674 (2.3%) participants were diagnosed with incident diabetes. Figure 2 displays the Kaplan-Meier curves of cumulative incidence of diabetes according to quintiles of the ALT/HDL-C ratio, showing an increasing trend in cumulative incidence of diabetes with increasing quintiles of the ALT/HDL-C ratio (log-rank P < 0.001).

Table 2 summarizes the relationship between the ALT/HDL-C ratio and diabetes incidence in the multivariable Cox regression analysis. Four multivariable-adjusted models were established, and a positive trend was observed between the ALT/HDL-C ratio as a continuous variable and the risk of incident diabetes in the crude model and Models 1-3. When the ALT/HDL-C ratio was treated as a categorical variable, compared to the lowest quintile (Q1), the highest risk was consistently observed in the fifth quintile (Q5) in all models (P-trend < 0.0001). Although the HR values slightly decreased in the adjusted Models 1, 2, and 3 compared to the crude model, the overall positive trend remained unchanged.

As a sensitivity analysis, we further examined the association among participants with normal liver function. After excluding 13,079 participants with ALT or AST levels higher than 40 IU/L, we conducted the same analytical procedures on the remaining 103,172 participants, and the new analysis results aligned with the results of the whole population analysis, indicating a positive correlation between the ALT/HDL-C ratio and diabetes risk, with the correlation progressively intensifying across the ALT/HDL-C ratio quintiles ( Table 3 ).

Subgroup analysis was conducted based on age, sex, family history of diabetes, and BMI to explore the relationship between the ALT/HDL-C ratio and diabetes risk in different commonly categorized populations. The study revealed significant statistical differences (P-interaction < 0.05) in the association between the ALT/HDL-C ratio and diabetes risk among different age groups, sexes, and BMI groups. Table 4 demonstrates that the risk of diabetes incidence varied among different age groups, with a higher risk observed in participants aged ≤ 60 years compared to those aged > 60 years (P-interaction < 0.0001). In the sex subgroup analysis, males had a higher risk of developing diabetes (HR: 1.08 for males, 1.04 for females; P-interaction = 0.0204). In the BMI subgroup analysis, compared to non-obese and overweight individuals, the highest diabetes risk related to the ALT/HDL-C ratio was observed in obese individuals (HR: non-obese 1.05 VS overweight 1.06 VS obese 1.17; P-interaction = 0.0001). Additionally, there was no statistically significant difference in the influence of family history of diabetes on the relationship between the ALT/HDL-C ratio and diabetes risk (P = 0.8113).

To compare the predictive value of the ALT/HDL-C ratio and its components for incident diabetes, ROC curve analysis was performed, and the corresponding AUC values were calculated ( Figure 3 ). The results indicated that the ALT/HDL-C ratio had the highest predictive value for incident diabetes compared to ALT and HDL-C alone (AUC: ALT/HDL-C ratio: 0.6716, ALT: 0.6653, HDL-C: 0.5817; all Delong P < 0.05). The optimal threshold point for the ALT/HDL-C ratio was calculated as 14.9248 ( Table 5 ).

Strengths: (1) This study involved a large sample size of 116,251 participants, which makes the results of the current study relatively reliable. (2) The current study revealed, for the first time, a positive correlation between the ALT/HDL-C ratio and diabetes risk in the Chinese population. (3) Several subgroup analyses were performed to compare differences in the correlation between ALT/HDL-C ratio and diabetes risk in populations with different characteristics, and these findings inform the clinical application of the ALT/HDL-C ratio.

Limitations: (1) Although the current research population hailed from various cities across China, the majority (10 out of 11 cities) are situated in the southern region. Therefore, the evidence of the current research may be more suitable for the population in the south of China, and the relevance of these results to the northern Chinese demographic requires further investigation. (2) The current study dataset is sourced from a public database and lacks liver disease information, detailed alcohol consumption information, and drug use information. considering that these factors may affect liver function, subsequently impacting the ALT/HDL-C ratio and the study’s outcomes, the absence of these information introduce certain limitations to our research. However, as an alternative solution, we further excluded the subjects whose baseline ALT and AST exceeded the normal reference level and conducted the same analysis steps; the results showed that the relationship between ALT/HDL-C ratio and diabetes remained positive in the population with normal liver function. This result largely reduces the potential impact of changes in liver function due to other causes and supports the conclusions of the main analysis. (3) Only baseline measurements of ALT, HDL-C, and other parameter indicators were considered, without accounting for numerical variations in ALT/HDL-C ratio over time. (4) The study did not evaluate data on postprandial blood glucose levels, which may have resulted in the underdiagnosis of some diabetes cases. (5) The current study excluded a large number of people who did not meet the inclusion criteria, which would bring about a certain selection bias.