In the training set, we gathered comprehensive data on 54,563 individuals who underwent H. pylori testing (C13/C14) at the Physical Examination Center of Taizhou Central Hospital (Taizhou University Hospital) between May 1, 2018, and December 31, 2022. This compilation encompassed information on patients’ age, gender, height, weight, blood pressure, as well as laboratory indicators, such as triglycerides (TG), total cholesterol (TC), glucose (GLU), glycosylated hemoglobin (HbA1c), albumin (ALB), creatinine (Cr), uric acid (UA), alanine aminotransferase (ALT), aspartate aminotransferase (AST), white blood cells (WBC), mean corpuscular volume (MCV), hemoglobin (Hb), platelets (PLT), and erythrocyte sedimentation rate (ESR). During screening, 233,816 patients with extensive missing data were excluded from the analysis. Furthermore, 84 patients with chronic liver disease, 30 patients with cirrhosis, and 21 patients with a history of malignant tumors were also excluded, ultimately leading to the inclusion of 30,612 patients in our analysis ( Figure 1 ). It is noteworthy that all research procedures adhered strictly to the ethical principles outlined in the Declaration of Helsinki, and the study received approval from the ethics committees of all participating institutions.

Obtained the patient’s age and gender, and measured their diastolic blood pressure (DBP), systolic blood pressure (SBP), height, and weight while they were in a calm and relaxed state. After an overnight fasting period of at least 8 hours, ultrasonography of the hepatobiliary system and arteries was performed, and venous blood samples were collected for laboratory parameter analysis. Additionally, a glycated hemoglobin analyzer was used to measure HbA1c levels.

H. pylori is detected through the C13 or C14 urea breath test (Fischbach and Malfertheiner, 2018). Initially, a breath sample is collected from an individual in a fasting state. Subsequently, the individual orally administers the C13 capsule and waits for 30 minutes before collecting another breath sample to complete the C13 test. Alternatively, the individual orally ingests the C14 urea capsule along with warm water, waits for 15 minutes, and then blows air evenly through the tube for 1-3 minutes. Finally, the gas collection card is inserted into the detector to finalize the C14 test.

The NHANES database (https://wwwn.cdc.gov/nchs/nhanes/) served as the validation set. NHANES, a public database, employs a cross-sectional, stratified, multistage probability design to capture a representative sample of the civilian, non-hospitalized population in the United States. Specifically, we chose the population from the 1999-2000 cycle, as it was the only cycle that included IgG results for H. pylori. Out of the total 9,965 individuals, we excluded 2,472 individuals who lacked H. pylori data and another 1,308 individuals with incomplete data. The ELISA optical density values for all subjects ranged from 0 to 5.73. Drawing from previous research criteria (Meier et al., 2020), we classified individuals with H. pylori antibody values below 0.9 as negative for H. pylori infection and those with values exceeding 1.1 as positive. Additionally, we excluded 176 individuals with the values falling within the range of 0.9 to 1.1, ultimately including a total of 6,009 individuals in our study ( Figure 1 ).

Continuous data are reported as mean ± standard deviation, with inter-group differences evaluated using t-tests. For data that deviate from normal distribution, they are presented as median (interquartile range), and inter-group disparities are assessed via the Mann-Whitney U test. Categorical variables are reported in terms of frequency (%), and chi-square tests are employed to assess inter-group differences. Restricted cubic spline (RCS) curves are utilized to explore the nonlinear relationship. Additionally, univariate and multivariate logistic regression analyses are conducted to identify independent risk factors, and the regression equation is constructed utilizing the forward likelihood ratio. The model’s fitting effect and clinical utility are evaluated through the receiver operating characteristic curve (ROC), area under curve (AUC), calibration curve, and decision curve analysis (DCA). A two-tailed p-value <0.05 is deemed statistically significant. Statistical analysis is performed using SPSS Statistics 22.0 (IBM Corp., Armonk, New York, USA) and R (version 4.3, Statistical Computing Foundation, Vienna, Austria, https://www.R-project.org). The rms, mice, rcssci, and nhanesR packages are utilized for comprehensive statistical analysis and graphical representation.

Based on the results of the C13 or C14 tests, patients were categorized into two groups: the H. pylori infection group, consisting of 22,296 individuals (72.8%), and the H. pylori non-infection group, encompassing 8,316 individuals (27.2%). Upon examination of the compiled data, it was observed that a minimal portion of the information was missing ( Supplementary Figure S1 ). Then the mice package was utilized to perform multiple imputation for the incomplete data points. The comparative analysis revealed that the H. pylori infection group exhibited significantly higher levels of age, weight, and BMI compared to the non-infection group (average p < 0.001). There were no significant differences in gender and height between the two groups ( Table 1 ).

The blood tests revealed significantly higher levels of TG, TC, Cr, Glu, WBC, and PLT in the H. pylori infection group compared to the non-infection group (average p < 0.01). On the contrary, the ALB level in the infection group was significantly lower (p < 0.01). The ESR levels did not adhere to a normal distribution, with the Mann-Whitney U test revealing a significantly elevated ESR level in the infection group (p < 0.001). However, no significant differences were observed in other laboratory parameters, including UA, ALT, and AST, between the two groups ( Table 1 ).

The ultrasound features of coarse or dense liver echo were utilized to assess the intrahepatic inflammation, while gallbladder wall thickening or roughness were utilized to assess the cholecystitis. Further comparison revealed significantly higher occurrences of fatty liver (42.4%), gallbladder cholesterol crystals (6.0%), gallbladder polyps (20.2%), and atherosclerosis (25.6%) in the H. pylori infection group compared to the non-infection group, with respective rates of 41.1%, 5.1%, 19.1%, 21.4% (average p < 0.01). However, no significant differences were observed between the two groups in terms of other features, including intrahepatic inflammation, gallstones, cholecystitis ( Table 2 ).

Logistic regression and RCS were employed to assess the association between the aforementioned parameters and H. pylori. The analysis revealed that individuals aged below 56 years exhibited an increasing risk of H. pylori infection with advancing age, whereas those over 56 years displayed a decreasing risk. Within the BMI range of 19.8-27.8, a significant elevation in the risk of H. pylori infection was observed with a rise in BMI index, whereas the risk decreased in individuals with a BMI exceeding 27.8. Notably, a significant negative correlation was observed between the ALB level and H. pylori infection, whereas the positive correlation was evident in TC. Furthermore, within the GLU range of >4.18mmol/L, a marked increase in the risk of H. pylori infection was observed with elevated glucose levels ( Figure 2 ). These findings suggest a potential association between H. pylori infection and metabolic abnormalities such as glucose, cholesterol, and albumin. Additionally, a significant increase in the risk of H. pylori infection was observed with an elevation in ESR levels within the ESR range of <21.1 mm/H ( Figure 2 ), indicating a potential link between H. pylori and the inflammatory state.

The association between H. pylori and metabolism may be influenced, to some extent, by abnormal glucose metabolism. Consequently, we conducted a focused analysis on the correlation between the aforementioned parameters and H. pylori in both diabetic and non-diabetic individuals. We further narrowed down our analysis to 18,887 individuals who had undergone HbA1c testing. Based on their medical histories and HbA1c tests, we identified 15,525 non-diabetic individuals and 3,362 diabetic individuals. Compared to the diabetic population, the trend of increasing risk of H. pylori infection with advancing age was more pronounced in the non-diabetic population. Similarly, the trend of elevated risk associated with decreasing ALB levels was also more evident in the non-diabetic group. More noteworthy, the significant associations between the risk of H. pylori infection and BMI, TC, and ESR were exclusively observed in non-diabetic individuals, excluding those with diabetes ( Figure 3 ).

Consistent with the findings from the overall population, significant differences were observed in age, BMI, TC, ALB, WBC, ESR, and HbA1c levels between the H. pylori-infection and non-infection groups among non-diabetic individuals ( Table 3 ). Similarly, the prevalence of cholesterol crystals (6.9%), gallbladder polyps (22.5%), and atherosclerosis (28.5%) was significantly higher in the infection group compared to the non-infection group, with respective rates of 5.9%, 20.5%, and 24.6% (average p<0.05) ( Table 4 ). However, among diabetic individuals, due to the limited sample size, only ALB, WBC, HbA1c, and gallbladder polyps exhibited significant differences between the two groups ( Supplementary Tables S1 , S2 ).

Similarly, the RCS curve plotted for the non-diabetic population revealed that the risk of H. pylori infection increased with age among individuals younger than 55 years, whereas it decreased among those older than 55 years. Within the BMI range of 20.0-27.4, a significant elevation in the risk of H. pylori infection was observed with an increase in BMI. Notably, ALB levels exhibited a significant negative correlation with the risk of H. pylori infection, whereas TC and ESR levels demonstrated a significant positive correlation. Furthermore, within the HbA1c range of >5.1%, a marked increase in the risk of H. pylori infection was observed with a rise in HbA1c levels ( Figure 4 ). These findings further underscore the association between H. pylori and abnormal metabolism of glucose, cholesterol, and albumin.

Multivariate regression analysis revealed that H. pylori infection, age, BMI, ALB, and TC were independent risk factors for cholesterol crystals (average p < 0.1, Table 5 ). Utilizing these parameters, a predictive model for cholesterol crystals centered on H. pylori was developed, formulated as follows: Y=0.112*H. pylori (yes=1, no=0) + 0.032*Age (years) + 0.049*BMI - 0.024*ALB (g/L) + 0.049*TC (mmol/L). This model exhibited an AUC value of 0.650 (95% CI: 0.644-0.655), with a statistical Z-score of 23.519 (p < 0.001). It demonstrated good sensitivity of 75.1% but relatively poor specificity of 48.6% ( Figure 5A ). Calibration curve analysis revealed a moderate alignment with the ideal state, yielding a Brier score of 0.05 ( Figure 5B ). DCA indicated moderate clinical utility for the model ( Figure 5C ). Additionally, H. pylori emerged as an independent risk factor for gallbladder polyps, albeit with a low predictive value ( Supplementary Table S3 , Supplementary Figure S2 ).

Subsequently, multivariate regression analysis revealed that H. pylori infection, age, BMI, ALB, TC, GLU, and ESR are independent risk factors for atherosclerosis (average p < 0.01, Table 6 ). Leveraging these parameters, a predictive model for atherosclerosis centered on H. pylori was formulated: Y=0.094* H. pylori (yes=1, no=0) + 0.096*Age (years) + 0.064*BMI - 0.034ALB (g/L) + 0.16*TC (mmol/L) + 0.205*GLU (mmol/L) + 0.013*ESR (mm/H). This model exhibited an AUC value of 0.828 (95% CI: 0.824-0.832), a statistical Z-score of 127.825 (p < 0.001), demonstrating high sensitivity (82.6%) and specificity (69.3%) ( Figure 5D ). Calibration curve analysis revealed a strong alignment with the ideal state, scoring a Brier of 0.14 ( Figure 5E ). Furthermore, DCA underscored the model’s significant clinical application value ( Figure 5F ).

The NHANES dataset comprised 3,718 individuals in the H. pylori non-infection group and 2,291 in the infection group. Consistent with the findings from the training set, the H. pylori infection group exhibited significantly higher ages and BMI indices compared to the non-infection group (average p < 0.001). Additionally, the levels of TC, Cr, GLU, and HbA1c were notably elevated, and the levels of ALB and HDL were significantly reduced in the infection group (average p < 0.01) ( Supplementary Table S3 ). Given the absence of ESR data, we resorted to the non-normally distributed C reactive protein (CRP). The Mann-Whitney U test revealed that the ESR level in the infection group was significantly elevated compared to the non-infection group (p < 0.001) ( Supplementary Table S3 ).

The RCS curve demonstrates that the risk of H. pylori infection rises with age among individuals younger than 55 years old. Within the BMI range of less than 28, a significant elevation in the risk of H. pylori infection is observed as the BMI index increases. Furthermore, in the ALB range exceeding 41g/L, there is a notable negative correlation between ALB and the risk of H. pylori infection. Additionally, TC and GLU levels exhibit a significant positive correlation with the risk of H. pylori infection. Within the CRP range below 1.2mmol/L, a marked increase in the risk of H. pylori infection is observed with rising CRP levels (average p<0.01) ( Figure 6 ). These findings align closely with the results of the training set, further corroborating the association between H. pylori infection and metabolic abnormalities in glucose, cholesterol, and albumin.