The NHANES is a population-based cross-sectional survey designed to assess health and nutritional status within the US population (14). Since 1999, approximately 5,000 nationally representative individuals have been surveyed annually, and related data have been published biennially. Demographics, methodology, examination data, dietary data, questionnaire data, and laboratory data were included in the survey; specific information could be found on the website https://www.cdc.gov/nchs/nhanes. The National Center for Health Statistics (NCHS) Ethics Review Board approved the NHANES study protocol. In this study, data for six periods were included, from 2007 to 2018, in our collection. This period received attention because data on kidney stones became available in 2007 and the coronavirus disease 2019 (COVID-19) had not yet become epidemic in 2018. We explored the relationship between smoke exposure and kidney stones in terms of participants' smoking status and serum cotinine concentrations. The following were the exclusion criteria for this study: (1) missing/without data about kidney stones (n = 25,163); (2) missing/without data about smoking status (n = 22); and (3) missing/without data about serum cotinine concentrations in a population with a complete kidney stone and smoking status data (n = 3,305). Questionnaires on kidney stones were conducted only for those older than 20 years. Therefore, the population included in this study was all adults. In terms of smoking status, 34,657 participants out of 59,842 participants were screened in our study population. On the other hand, in terms of serum cotinine concentrations, a total of 31,352 participants were included. Figure 1 shows details of the inclusion and exclusion processes for this study.

For the definition of smoking status, all participants needed to answer this question first, “In your lifetime, have you smoked more than 100 cigarettes?” Those who answered “No” were labeled as “Non-smokers,” while those who answered “Yes” were required to answer the following question further: “Do you smoke now?” The person who answered “Some days” or “Every day” was labeled as a “Current smoker,” and those who answered “Not at all” were labeled as “Former smokers.” In general, participants were divided into “non-smokers,” “former smokers,” and “current smokers.”

Cotinine, a major nicotine biomarker, was the first choice for assessing smoke exposure by measuring cotinine concentrations in blood, urine, and saliva (15). Compared with other diagnostic tools, cotinine was the best indicator of tobacco use because of its high sensitivity, good specificity, and long half-life (16). The liquid chromatography/atmospheric pressure ionization tandem mass spectrometric method was used to measure serum cotinine concentrations (17). Based on previous studies (18–20), we used 0.05 and 3.00 ng/ml as cutoff values to further convert serum cotinine concentrations into categorical variables. The class of serum cotinine was embodied in <0.05 ng/ml, 0.05–2.99 ng/ml, and ≥3.00 ng/ml. In addition, a log₂ transformation was used for serum cotinine because of its skewed distribution, and then it was analyzed as a continuous variable.

Participants answered the question “Have you ever had kidney stones?” in the kidney condition questionnaire, and those who answered “Yes” were identified as having a clear history of kidney stones (4).

Through previously published studies, we screened factors associated with smoke exposure, cotinine, or kidney stones. These included age, sex, race, marital status, education, income-to-poverty ratio in the family (family PIR), body mass index (BMI), physical activity, serum uric acid, hypertension, coronary heart disease, diabetes, and gout (4, 21–25). Continuous variables included age, family PIR, BMI, and serum uric acid. Categorical variables included sex, race, marital status, education, physical activity, hypertension, coronary heart disease, diabetes, and gout. Vigorous or moderate work for at least 10 min continuously was labeled as an active physical activity. Hypertension was established when one of the following three conditions was present: systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg or being on antihypertensive medications. Fasting blood glucose ≥126 mg/dl or glycohemoglobin ≥6.5% was labeled as diabetes. With or without coronary heart disease and gout were divided by answering the corresponding questionnaire “Yes” or “No.”

We used a genome-wide association study to obtain genetic instrumental variables for serum cotinine, which involved 5,185 current smokers (serum cotinine >10 ng/ml) of European ancestry (26). We screened the SNPs using the following steps, which were also the basic conditions of MR analysis: first, SNPs closely related to serum cotinine were screened out with a threshold P < 5 × 10⁻⁶. Then, under the parameters, r² <0.001 and kb = 10,000, SNPs with linkage disequilibrium (LD) were removed. Finally, the F statistic was calculated for each SNP. Those F statistics >10 were considered strong instrumental variables (27) and strongly associated with serum cotinine. The flowchart of MR is shown in Figure 2. In addition, the three core assumptions about MR are shown in Supplementary Figure S1.

GWAS summary data on kidney stones were available in the FinnGen consortium (https://finngen.gitbook.io/documentation/), which was a growing project among the Finns. We used the Release 6 (R6) version of data on kidney stones, which included 5,985 cases and 253,943 controls.

Complex sampling designs need to be considered to make our study population nationally representative. During the analysis, we used the sample weights from the NHANES database. By stratifying smoking status and serum cotinine concentrations, the weighted baseline characteristics of the overall population and participants under the corresponding stratification were described. Categorical variables and continuous variables were represented by percentages and the mean ± standard deviation (SD), weighted chi-square test, and weighted linear regression model were used to calculate the P-value, respectively. When analyzing serum cotinine concentrations as a continuous variable, a log₂ transformation was implemented to meet a normal distribution. In addition, the variance inflation factor (VIF) was calculated to avoid the existence of multicollinearity, and a VIF of <10 could be considered to indicate that there was no multicollinearity between covariates (28). Univariate analysis was performed on participants based on smoking status and serum cotinine to judge the association between covariates and kidney stones. Then, three multivariate logistic regression models were constructed to explore the relationship between smoke exposure and kidney stones. Age, sex, race, and BMI were adjusted in Model 1. Model 2 adjusted marital status, education, family PIR, and physical activity based on Model 1. Model 3 adjusted serum uric acid, hypertension, coronary heart disease, diabetes, and gout based on Model 2. Based on Model 3 adjustments, a possible non-linear relationship was explored by constructing a restricted cubic spline (RCS) between serum cotinine (log₂ transformation) and the risk of kidney stones. The number of knots in RCS was set to 5 because of the large sample size, and the serum cotinine concentration with OR = 1 was set as the reference value. In addition, to avoid bias in the results caused by too many missing values, the multiple imputation (MI) method, which was done via the “MICE” package in R software (29), was used to supplement the missing values and verify the robustness of the results. Five complete sets of data were obtained through MI, and finally, the effect values of the five datasets were integrated. Model 4 represented the effect values integrated after MI based on the fully adjusted model.

In this MR analysis, we used inverse-variance weighting (IVW) as our main research method. The presence or absence of heterogeneity determined whether a random-effects model or a fixed-effects model was used in the analysis. In addition, we used the MR-Egger, the weighted mode, and the weighted median for sensitivity analysis. Previous studies had described the details of these methods (30–32). The MR-Pleiotropy RESidual Sum and Outlier (MR-PRESSO) method was used to rule out outliers and test for horizontal pleiotropy. Furthermore, directional pleiotropy was determined by the MR-Egger intercept method. Heterogeneity was detected by the Cochrane Q test. MR-robust adjusted profile score (MR-RAPS) could correct pleiotropy with adjusted file scoring and make our results more reliable when many weak instrumental variables existed (31). Eventually, if the potential effects of SNPs were found in the “leave-one-out” analysis, then we needed to draw conclusions with caution (27). A scatter plot was created to visually observe the association between serum cotinine and the risk of nephrolithiasis.

All statistical analyses were performed using R Software (Version 4.2.1, http://www.R-project.org, The R Foundation) and Empowerstats Software (Version 2.0, http://www.empowerstats.com, X&Y Solutions, Inc., Boston, MA). In our study, a p-value of <0.05 was considered statistically significant.

In the aspect of smoking status, the weighted baseline characteristics of the study population are shown in Table 1. Non-smokers, former smokers, and current smokers consisted of 19,414 (56.02%), 8,219 (23.72%), and 7,024 (20.27%) participants, respectively. We found statistically significant differences in age, sex, race, marital status, education level, family PIR, BMI, physical activity, history of hypertension, history of coronary heart disease, history of diabetes, history of gout, and history of kidney stones in different smoking status groups (all P-values <0.001). The average age of the overall participants was 47.48 ± 17.02 (mean ± SD) years. The majority of the study population was female (51.91%) and non-Hispanic white (65.89%). Compared to the other groups, the current group of smokers was younger, had a lower family PIR, had lower serum uric acid, had a lower BMI, had active physical activity, and had no history of hypertension, coronary heart disease, diabetes, or gout. In particular, the population had a high prevalence of kidney stones (9.72%).

As for grouping according to serum cotinine concentrations, the weighted baseline characteristics are shown in Table 2: 16,650 (53.11%), 6,426 (20.50%), and 8,276 (49.71%) participants in the group with serum cotinine of <0.05 ng/ml, 0.05–2.99 ng/ml, and ≥3.00 ng/ml, respectively. The average age of the overall participants was 47.57 ± 16.94 (mean ± SD) years. We did not observe a statistically significant difference in whether participants had a history of coronary heart disease (P-value = 0.176) or kidney stones (P-value = 0.067). The remaining characteristics in the serum cotinine ≥3.00 ng/ml group were similar to those of the participants in the current smoker group. Similarly, a high prevalence of kidney stones (10.19%) in participants with serum cotinine concentrations of ≥3.00 ng/ml was discovered.

The correlation between covariates and kidney stones based on smoking status is shown in Supplementary Table S1. Among participants with or without a history of kidney stones, we found that age, sex, race, marital status, BMI, education level, and serum uric acid were statistically significant (P < 0.001). On the other hand, hypertension, coronary heart disease, diabetes, and gout were closely related to the occurrence of kidney stones. The above features were also observed in participants based on serum cotinine concentrations, which are demonstrated in Supplementary Table S2.

The association between smoking status and kidney stones is presented in Table 3. We found that the current smoker group was strongly associated with the occurrence of kidney stones compared to the non-smoker group regardless of the adjustment model [Model 1: OR = 1.19 (1.07–1.32), P < 0.001; Model 2: OR = 1.18 (1.05–1.31), P = 0.004; Model 3: OR = 1.17 (1.04–1.31), P = 0.009], while these features were not observed in former smokers (all P > 0.05). Similar results were obtained in former smokers (P = 0.249) and current smokers [OR = 1.16 (1.05–1.28), P = 0.005] in Model 4. Therefore, current smokers have a higher risk of developing kidney stones (P for trend in Model 1, Model 2, and Model 3: <0.001, 0.004, and 0.010, respectively).

Table 3 shows the association between serum cotinine as a continuous or categorical variable and the occurrence of nephrolithiasis. Log2-transformed serum cotinine was linked with an increased risk of kidney stones [the effect values were the same for either model: OR = 1.02 (1.01–1.03), P < 0.001]. No matter which adjustment model they were under, those with serum cotinine of 0.05–2.99 ng/ml and ≥ 3.00 ng/ml had varying degrees of increased risk of developing kidney stones compared with the control group (serum cotinine <0.05 ng/ml), especially in participants with higher serum cotinine concentrations [≥3.00 ng/ml; Model 1: OR = 1.25 (1.14–1.38, P < 0.001; Model 2: OR = 1.25 (1.12–1.39), P < 0.001; Model 3: OR = 1.22 (1.10–1.37), P < 0.001; Model 4: OR = 1.23 (1.11–1.36), P < 0.001]. As the category of serum cotinine levels increased, the prevalence of kidney stones also increased (P for trend was all <0.001 in Model 1, Model 2, and Model 3). Effect values for each of the five complete sets of data after MI, based on smoking status and serum cotinine concentrations, are presented in Supplementary Tables S3, S4, respectively.

In RCS, we observed a non-linear relationship between log2-transformed serum cotinine and kidney stone risk (Figure 3; P for non-linearity = 0.028). The reference value is 5.206 ng/ml (serum cotinine = 36.91 ng/ml). When log₂-transformed serum cotinine was below the reference value, the risk of kidney stones was little changed or even decreased, and when log₂-transformed serum cotinine was above the reference value, the risk increased rapidly.

A total of 10 SNPs were eventually selected after a series of rigorous screenings, the details of which are shown in Supplementary Table S5. Genetically predicted serum cotinine was found to be associated with an increased risk of kidney stones by method IVW [Figure 4; OR = 1.09 (1.00–1.19), P = 0.044] and MR-RAPS [OR = 1.10 (1.01–1.21), P = 0.038], but the MR-Egger, the weighted mode, and the weighted median did not find a causal relationship between the two. Furthermore, heterogeneity, directional pleiotropy, and horizontal pleiotropy were not observed in this study (see Supplementary Table S6). The results of MR analysis were influenced by potential SNPs, as could be seen by the “leave-one-out” plot, so conclusions needed to be drawn with caution (Supplementary Figure S2). The scatter plot showed that as serum cotinine levels increased, so did the risk of nephrolithiasis, which is demonstrated in Figure 5.