Study design and population
All the data for our study were sourced from the NHANES database, a comprehensive national survey conducted in the USA to assess the nutritional and health profiles of the population. This database is refreshed biennially with new sets of data, encompassing demographic, dietary, examination, laboratory, questionnaire, and limited access information. The NHANES data are publicly available at: https://www.cdc.gov/nchs/nhanes/index.htm. The research protocols for NHANES are reviewed and approved by the Ethics Review Board (ERB) of the National Center for Health Statistics (NCHS), and all participants provide informed consent.
For the purposes of our analysis, we extracted data from 24,814 individuals who participated in NHANES during the cycles spanning 2017-2018 and 2019-2020. Exclusion criteria were applied as follows: 10,050 participants were omitted due to incomplete data necessary to ascertain cholelithiasis status, 1,238 had missing BMI data, 141 were pregnant, and a further 255 were excluded for having a BMI over 50 kg/m². After these exclusions, a total of 13,130 participants remained for the final analysis, representing a weighted population estimate of approximately 229,869,148, calculated using the formula 1/2*wtmec2yr (Fig.1).
Definition of cholelithiasis in NHANES
Cholelithiasis was defined as the “mcq_j/p_mcq” information from the “mcq550” questionnaire in the NHANES data, which was labelled as “has a doctor or other health professional ever told you that you had gallstones?” with the answer of “yes”.
Definition of Frailty Index in NHANES
Frailty was constructed using the standard procedure introduced by Searle and colleagues[28]. It comprised 49 deficits that affected multiple systems, including cognition, dependence, depression, and comorbidities. The severity of the deficit was used to assign a value between 0 and 1, enabling the combination of continuous and categorical variables. The calculation of the frailty index value involves dividing the number of acquired deficits by the total number of potential deficits[29]. A frailty index score of 0.21 has been established as the threshold for identifying 'frail' individuals who are at an increased risk of hospital-related events[30, 31]. The variables and their corresponding scores for the frailty index can be found in Supplementary Table 1 in the Appendix.
Assessment of Covariates
In both binary logistic regression models, we adjusted for a set of potential confounders based on previous studies. These included age, gender, body mass index, race/ethnicity, education level, marital status, poverty, recreational activity, smoking status, and drinking status. BMI was classified as '<25', '25-30', and '>30' kg/m². Race/ethnicity was categorized as non-Hispanic White, non-Hispanic Black, Hispanics, and other races. Family poverty was divided into '<13', '13-35', and '>35' by the state’s specific poverty thresholds [32]. Education level was categorized as 'Below high school', 'High school or equivalent', or 'College or above'. Marital status was classified as 'Married or with a partner', 'Unmarried', or 'Widowed/Divorced/Separated'. Smoking status was classified as 'Never', 'Former', or 'Now' smoker according to the NCHS classification. Drinking status was categorized as either 'Yes' or 'No' based on whether participants had consumed at least 12 alcoholic drinks within the past year at the time of the interview. With regards to physical activity, participants were classified as either 'Yes' or 'No' based on whether they engaged in regular moderate or vigorous recreational activities. The presence of diabetes mellitus was determined by fasting blood glucose levels of ≥7.0 mmol/L, glycosylated hemoglobin levels of ≥6.5%, glucose tolerance test results of ≥11.1 mmol/L, or self-reported diagnosis of diabetes mellitus[33]. Impaired Fasting Glucose (IFG) is a state of abnormal blood glucose levels, between normal fasting blood glucose and diabetes. It is a pre-diabetic stage, meaning that an individual has higher than normal blood glucose levels, but has not yet reached the threshold for diabetes. Hypertension was defined as systolic blood pressure of ≥140mmHg, diastolic blood pressure of ≥90mmHg, use of blood pressure medication, or self-reported diagnosis of hypertension[34]. The study assessed the presence of hyperlipidemia among participants based on their self-reported use of cholesterol-lowering medications or meeting certain lipid level criteria: triglycerides ≥ 150 mg/dL, total cholesterol ≥ 200 mg/dL, low-density lipoprotein ≥ 130 mg/dL, or HDL ≤ 40 mg/dL in males and ≤ 50 mg/dL in females[35].
Statistical analysis
The complex sampling design and sampling weights of NHANES were incorporated into our analysis. Then, the normality of distribution of continuous variables in the dataset was assessed by the Kolmogorov-Smirnov test, which showed that all continuous variables included in this study satisfied or almost satisfied the assumption of normality. The statistical analysis presents continuous variables as means ± standard errors SE for normal distribution. Categorical variables were described as frequencies with percentages. To evaluate the differences between groups with and without cholelithiasis, Student’s t-test and Pearson’s chi-square χ2 test were used. To verify the relationship between FI and the risk of cholelithiasis, logistic regression models were utilized to calculate odds ratios (OR) and 95% confidence intervals (CI). These models were adjusted for age, gender, race, education, marital status, activity, drinking, smoking, hypertension, diabetes, hyperlipidemia, and CVD. Subsequently, further adjustments were made for activity, drinking, smoking, model 2, and finally additional adjustments were made for hypertension, diabetes, hyperlipidemia, and CVD (model 3). Based on these results, a restricted cubic spline RCS was created to estimate individual-specific probabilities of cholelithiasis. Additionally, calibration curves were used to verify discrimination and clinical validity of this model. All statistical analyses were performed using R version 4.3.2. https://www.r-project.org/.