This study focused on longitudinal data and large population cohorts. The study population was gathered from a multi-center physical examination database located in Shijiazhuang, Hebei Province, China, during the period from October 2011 to October 2024. Follow-up continued until the date of the endpoint of observation on October 31, 2024. Each participant’s follow-up began at enrollment and concluded upon an occurrence of NAFLD/MASLD, loss to follow-up, or at the endpoint of observation in October 31, 2024. Participants who engaged in at least twelve sessions during the 14-year follow-up period were eligible for inclusion in the study; notably, 46% of the population participated in all fourteen rounds of surveys. All participants were required to complete all physical examinations before leaving each session. Further diagnosis and treatment would be necessary if any medical conditions were identified during these examinations.

The data collected through lifestyle survey questionnaires (e.g., sociodemographic characteristics, dietary preferences, family and personal medical history), physical measurements (e.g., height, body weight, and blood pressure), clinical laboratory tests (e.g., blood, urine, and feces), and abdominal ultrasound examinations (e.g., liver, gallbladder, pancreas, spleen, and kidney). Standardized questionnaire surveys were administered prior to the participants’ physical examinations. During this period, a total of 23,666 Chinese individuals aged 25–60 years who completed both the questionnaire survey and physical examinations were recruited for this study. All participants were free from NAFLD/MASLD and advanced liver fibrosis as identified by abdominal ultrasound examination and clinical laboratory tests at baseline. Their institutions or companies organized the physical examinations, respectively, per year.

This research was granted ethical approval by the Institutional Review Board (IRB) of the Second Hospital of Hebei Medical University. Informed consent was obtained from all participants involved in the study. The procedures adhered to the Declaration of Helsinki and other applicable regulations.

We assessed the frequency of spicy food consumption over the course of 1 week utilizing a structured questionnaire. Two questions were employed to assess spicy food intake. The first question was: “How often do you consume spicy foods in past week over the past months?” Participants were instructed to respond by selecting one of the following options: never, less than 1 day per week, 1–2 days per week, 3–5 days per week, or 6–7 days per week. Spicy food intake was defined as follows: direct consumption of fresh chili peppers; fresh/fermented/dried chili peppers; sweet peppers; chili oil; chili sauce/paste; curry; or other hot red chili peppers used in cooking. A subsequent question was directed to participants who reported consuming spicy foods at least once per week: “What is your preferred level of spiciness?” Participants were presented with three options for spiciness levels: heavy, moderate, and mild.

In this study, data were collected from three dimensions to explore potential associations between spicy food consumption and NAFLD/MASLD or liver fibrosis. Sociodemographic characteristics included age, gender, smoking and drinking status, educational condition, work intensity, household income, and marital status. The smoking status of participants was assessed through questions to determine if they currently smoke. Excessive alcohol consumption was assessed via a questionnaire, defined as exceeding 30 grams per day for male participants and exceeding 20 grams per day for female participants (4). The educational level was categorized as either below a university degree or at the university degree level and above. The intensity of work was self-reported by participants, categorized as either light or heavy. A high household income was defined as an annual household income exceeding 150 thousand yuan. Marital status was categorized into two groups: currently married or not. Dietary preferences across all survey waves included weekly consumption of vegetables, fruits, meat, and eggs; responses were categorized into two groups: frequent and seldom.

Anthropometric variables, including waist circumference (WC), height, and body weight, were measured following a standardized protocol. Body mass index (BMI) was calculated as body weight in kilograms divided by height in meters squared. Blood pressure readings were obtained as the average of three distinct measurements taken from the upper right arm at the brachial artery, utilizing an automatic device following a 5-min rest period in a seated position. In this study, we collected blood samples for various indices related to blood glucose and lipid metabolism, including total cholesterol (CHOL), triglycerides (TG), high-density lipoprotein (HDL), low-density lipoprotein (LDL), platelet count (PLT), albumin (Alb), alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin (TBil), glycosylated hemoglobin (HbA1c), fasting blood glucose (FBG), and direct bilirubin (DBil).

Family medical history was systematically gathered, with a particular emphasis on three prevalent health conditions: diabetes, hypertension, hepatitis, and cancer. Data concerning the age of onset, as well as the duration for which family members had been diagnosed and treated, were meticulously documented.

Abdominal ultrasonography was utilized to investigate SLD. A skilled technician conducted the liver ultrasound, and all images were prospectively assessed by experienced hepatologists who were blinded to the clinical data. SLD was defined by positive ultrasound findings, characterized by two or more abnormal results in liver ultrasonography. Specifically, these criteria included: (1) diffusely increased echogenicity of the liver; (2) heightened echogenicity of the liver compared to that of the kidney or spleen; and (3) vascular blurring accompanied by gradual attenuation of the ultrasound signal. The ultrasonographic assessment of the SLD was classified as normal, mild, moderate, or severe hepatic steatosis.

NAFLD was defined as the presence of steatosis in individuals without excessive alcohol consumption or evidence of active hepatitis viral infection (33). Participants who tested positive for hepatitis B surface antigen or exhibited positive antibodies for hepatitis C virus were excluded from this study based on their blood laboratory test results.

MASLD referred to steatosis that is associated with the presence of at least one indicator of cardio-metabolic dysregulation. This condition was characterized by the occurrence of at least one of the following abnormalities related to cardio-metabolic risk: (1) a BMI ≥ 24 kg/m² or WC ≥ 94 cm for male participants or ≥80 cm for female participants; (2) systolic blood pressure (SBP) ≥ 130 mmHg, diastolic blood pressure (DBP) ≥ 85 mmHg, or receipt of specific anti-hypertensive medication; (3) plasma TG ≥ 1.70 mmol/L, or treatment with lipid-lowering medications; (4) plasma HDL < 1.0 mmol/L for male participants and <1.3 mmol/L for female participants, or treatment with lipid-lowering medications; and (5) pre-diabetes indicated by FBG levels between 5.6–6.9 mmol/L, glycated hemoglobin levels ≥5.7%, diagnosis of type 2 DM, or treatment with specific anti-diabetic medications (34).

The NAFLD Fibrosis Score (NFS) and the Fibrosis-4 index (FIB-4) were utilized to assess advanced liver fibrosis, which was widely employed in clinical practice for non-invasive evaluation of the risk of advanced liver fibrosis. Specifically, NFS was calculated using the following formula: NFS = −1.675 + 0.037 × age (years) + 0.094 × BMI (kg/m²) + 1.13 × diabetes (yes = 1, no = 0) + 0.99 × AST (U/L)/ALT (U/L) − 0.013 × PLT (×10⁹/L) − 0.66 × Alb (g/dL). A NFS value greater than −1.455 indicated advanced fibrosis (35). FIB-4 was developed using the formula: FIB-4 = (age [years] × AST [U/L])/((PLT [×10⁹/L]) × (ALT [U/L]) ^ (1/2)). A FIB-4 score exceeding 1.3 suggested advanced fibrosis (36).

The multiple imputation approach was employed to handle missing values. We utilized a binary logistic regression model to investigate the mechanism underlying the missing data and subsequently generated imputed datasets in accordance with this mechanism. With less than 5% of the data missing, we handled them using the Multivariate Imputation by Chained Equations (MICE) approach. The imputation model included all variables intended for subsequent analyses. Continuous variables were imputed using predictive mean matching, and categorical variables were imputed using logistic regression. We created 20 imputed datasets, running 10 iterations per dataset to ensure model stability and convergence. Subsequently, the results from analyses on these 20 datasets were pooled following Rubin’s rules (37–39). All imputations were performed using the mice package in R software.

The distribution of baseline variables was presented according to the reported frequency of consumption of spicy foods. Continuous variables were expressed as means with 95% confidence intervals (CIs), while categorical variables were displayed as frequencies and percentages. The relationship between spicy food consumption/chili level preference and the risk of NAFLD/MASLD, as well as advanced liver fibrosis, was assessed using Cox proportional hazards regression models across different variables. These adjusted models examined the association between the frequency of spicy food consumption and the risks associated with NAFLD/MASLD and liver fibrosis. Model I provided crude hazard ratios (HRs) along with 95% CIs; Model II further adjusted for demographic factors; Model III included additional adjustments for dietary preferences, and family medical history. Given that markers of obesity, blood pressure, blood glucose, or liver function may serve as intermediate factors in the potential causal pathway linking spicy food consumption to NAFLD/MASLD and advanced liver fibrosis, we excluded them from the multivariate analyses. In all analyses, participants who never consumed spicy foods were the reference category. For further analyses on chili level preference, those who never consumed spicy foods and those who consumed them less than once a week served as the reference category. HRs along with 95% CIs were utilized to compare risks among various groups. To evaluate trends in HRs across increasing categories of spicy food intake, we employed a Mantel–Haenszel extension chi-square test.

Cox proportional hazards regression models with adjusted RCS were used to explore the nonlinear relationship between spicy food consumption frequency and the risk of incident NAFLD/MASLD and liver fibrosis. Participants who never consumed spicy foods served as the reference category for all analyses of nonlinear associations. Nonlinear curve fitting was optimized by including three knots in the models, reducing accuracy loss from over-fitting (40).

Additionally, we conducted subgroup analyses by categorizing several demographic covariates for further study. These analyses included subgroups based on age group, sex, smoking status, educational level, work intensity, household income, and marital status. Furthermore, sensitivity analyses were performed to ensure the robustness of our findings across three distinct aspects. First, we compared results before and after including participants with missing data. Second, we assessed results prior to and following multiple imputation procedures. Lastly, we excluded participants with a reported family medical history to evaluate the stability of our conclusions.

As presented in Supplementary Table 1, we observed that the proportion of missing data across all variables was less than 5%. Furthermore, our logistic regression analysis investigating the relationship between all variables with missing data and NAFLD/MASLD as well as advanced liver fibrosis revealed that all p-values were non-significant, as shown in Supplementary Table 2. These findings indicate that the missing data for the independent variables are independent of the dependent variables in this study. Consequently, we infer that the missing data can be classified as missing at random.

After imputing the missing data using multiple imputation techniques, we found that the distributions of the imputed values closely mirrored those of the observed values. As illustrated in Supplementary Figure 1, gray represented observed data while red depicts imputed data.

Detailed baseline information on participants categorized by the frequency of weekly spicy food consumption in this study was shown in Table 1. A total of 23,666 participants aged 25–60 years, free of NAFLD/MASLD and advanced liver fibrosis, were enrolled between 2011 and 2024. The average age of the participants was 40.0 years (95% CI = 39.8–40.1) in 2011. Among the participants, 9,990 (42.2%) reported consuming spicy food at least once per week. Specifically, there were 5,930 participants who reported never consuming spicy food; 7,746 participants consumed spicy food less than once per week; 2,587 participants consumed spicy food one to two days per week; 2,431 participants consumed spicy food three to five days per week; and 4,972 participants consumed spicy food six to seven days per week. A total of 7,965 patients with NAFLD and 7,311 patients with MASLD were identified after a 12.6-year median follow-up in this cohort study. Among these individuals, there were 342 diagnosed with advanced liver fibrosis based on the Nonalcoholic Fatty Liver Disease Fibrosis Score (NFS), and an additional 380 diagnosed as having advanced liver fibrosis based on the FIB-4 index.

As shown in Table 2, after adjusting for covariates, participants consuming spicy food 1–2 days per week had an 18.9% lower risk of NAFLD (HR = 0.811, 95%CI = 0.750–0.878) and a 19.7% lower risk of MASLD (HR = 0.803, 95%CI = 0.741–0.871). Those consuming spicy food 3–5 days per week experienced a 33.2% lower risk of NAFLD (HR = 0.668, 95%CI = 0.612–0.729) and a 42.0% lower risk of MASLD (HR = 0.580, 95%CI = 0.528–0.637). Participants consuming spicy food 6–7 days per week had a 36.4% lower risk of NAFLD (HR = 0.636, 95%CI = 0.593–0.681) and a 44.6% lower risk of MASLD (HR = 0.554, 95%CI = 0.515–0.596), compared to those who never consumed spicy food. However, we did not observe a significant difference in the risk of advanced liver fibrosis between participants who consumed spicy foods and those who did not.

A significant non-linear relationship between the frequency of weekly consumption of spicy foods and the risk of NAFLD/MASLD (p < 0.001) was illustrated in Figure 1. However, this relationship was not observed in advanced liver fibrosis for either diagnostic standard.

The results of subgroup analyses stratified by age, gender, smoking status, and educational level, working intensity, household income, and marital status are presented in Table 3. Compared to the overall analysis, the consistent findings from subgroup analyses suggested that individuals who consumed spicy foods exhibited a lower risk of NAFLD/MASLD than those who did not engage in spicy food consumption. Furthermore, we did not observe significant changes in the subgroup analyses related to advanced liver fibrosis (Supplementary Table 3).

The robustness of the results was evaluated through sensitivity analyses conducted from three distinct perspectives, as outlined in Table 4. Firstly, excluding participants with missing data did not compromise the integrity of the primary findings. Secondly, consistency was observed between the dataset without missing values and the dataset utilizing multiple imputation techniques. Finally, after removing participants with a family medical history at baseline, it was evident that consumption of spicy foods remained associated with a lower risk of NAFLD/MASLD compared to those who never consumed spicy foods.

In further studies, we investigated the relationship between chili level preference and NAFLD/MASLD, as well as advanced liver fibrosis. As presented in Table 5, participants who never consumed or consumed spicy foods less than once per week had a 22.9% lower risk of NAFLD (HR = 0.771, 95%CI = 0.719–0.825) and a 24.6% lower risk of MASLD (HR = 0.754, 95%CI = 0.701–0.810) compared to those preferring mild pungency in their diet. Participants consuming spicy food with moderate pungency exhibited a 33.9% reduction in the risk of NAFLD (HR = 0.661, 95%CI = 0.570–0.761) and a remarkable 43.3% decrease in the risk of MASLD (HR = 0.567, 95%CI = 0.478–0.666). However, no significant association was observed between heavy pungency preference and either NAFLD or MASLD outcomes; furthermore, we did not find any significant correlation between chili level preference and advanced liver fibrosis.