The study adhered rigorously to the guidelines set forth in the Strengthening the Reporting of Observational Studies in Epidemiology Mendelian randomization (STROBE-MR) framework (Supplementary Table S1) (Skrivankova et al., 2021). MR is predicated on three fundamental assumptions: IVs must exhibit a robust association with exposure, they should be unaffected by confounding variables, and they should influence the outcome exclusively through the exposure (Burgess et al., 2015). Our research implemented MR analyses in strict accordance with these core principles, as illustrated in Figure 1.

To ensure the stability of the causal relationship between exposure and outcome, IVs were selected based on the following principles (Zhu et al., 2021): We established genome-wide significance thresholds for exposure at p < 5 × 10⁻⁸ (Hájková et al., 2021). Cluster analysis was conducted to address linkage disequilibrium (LD) among the selected IVs (r ² < 0.001, kb = 10,000) (Marić et al., 2019). To mitigate bias from weak IVs. The strength of the IVs was quantified using the F value (β²/SE), with those having F < 10 being excluded. Here, β represents the effect size of exposure and SE represents the standard error of the effect size.

Genetic summary data on smoking initiation, cigarettes per day, and alcohol consumption measured in drinks per week were acquired from the Sequencing Consortium of Alcohol and Nicotine use (GSCAN) (Saunders et al., 2022). This consortium’s dataset encompasses information from 2,669,029 individuals of European descent. The genome-wide association study (GWAS) dataset on the lifetime smoking index included details on smoking duration, intensity, and quitting, which were combined to create a simulated half-life (τ) constant and a lifetime smoking index (n = 462,690) (Wootton et al., 2020). GWAS data on the frequency of alcohol intake were obtained from the questionnaire “How often do you drink alcohol?” with ordered categorical variables (n = 462,346) (Hemani et al., 2018).

Genetic association summary data for upper urinary calculi were obtained from the ninth release of the FinnGen Consortium database, encompassing 376,406 individuals of Finnish ancestry, including both males and females (Kurki et al., 2023). The analysis excluded individuals exhibiting extreme heterozygosity (±4 SD), a genotyping deletion rate (>5%), ambiguous gender, or non-Finnish ancestry. The genetic associations’ effect sizes were estimated using logistic regression, controlling for age, sex, and principal genetic components. Additionally, cases of upper urinary calculi in the United Kingdom Biobank were identified based on the International Classification of Diseases, Ninth Revision (ICD-9) and 10th Revision (ICD-10) criteria. Correlation tests were adjusted for confounding variables, including age at enrolment, gender, and the first ten principal genetic components. Table 1 shows detailed information about the GWAS data.

We conducted an a priori power calculation with an α of 5% (Brion et al., 2013). Ensuring that we had at least 80% power to detect Odds Ratios (ORs) for upper urinary calculi of 1.30, 1.42, 1.27, 1.25, and 1.10 for the respective variables of smoking initiation, cigarettes per day, lifetime smoking index, drinks per week, and alcohol intake frequency in the FinnGen data. Similarly, in the United Kingdom Biobank data, we could detect ORs of 1.11, 1.09, 1.51, 1.32, and 1.41 for these variables.

To mitigate potential pleiotropy arising from indirect pathways, we utilized MVMR analyses. By consulting the PhenoScanner database, we pinpointed associations of IVs with education and obesity-related traits that met the genome-wide significance threshold (p < 5 × 10⁻⁸). Consequently, we meticulously selected SNPs linked to education (n = 1,131,881) (Lee et al., 2018) and BMI (n = 681,275) (Yengo et al., 2018) for our multivariate analysis.

Our primary analysis utilized the robust inverse-variance weighted (IVW) method (Burgess et al., 2013). For validation, we applied supplementary methods including weighted median, MR-Egger regression, and weighted mode. The weighted median approach, renowned for its reliability, provided consistent results by prioritizing the influence of the most powerful instrumental variable, which carried a 50% weight (Bowden et al., 2016). To tackle potential directional pleiotropy, we further employed MR-Egger regression and weighted mode methods (Bowden et al., 2015; Hartwig et al., 2017).

We assessed the genetic correlation for upper urinary calculi between the United Kingdom Biobank and FinnGen consortium using LD Score Regression (LDSC) software, which revealed high consistency (rg = 0.80; p = 3.75 × 10⁻²²) (Bulik-Sullivan BK. et al., 2015; Bulik-Sullivan B. et al., 2015). Thus, we integrated the data using a fixed-effects model. I² > 50% was interpreted as indicative of high heterogeneity.

Sensitivity analysis plays a critical role in evaluating heterogeneity and potential biases in MR studies. We first assessed heterogeneity using Cochran’s Q test, which calculates the weighted sum of squared differences between individual study estimates and the overall IVW estimate (Burgess et al., 2017). To detect and adjust for potential outliers, we applied the MR Pleiotropy RESidual Sum and Outlier (MR-PRESSO) detection method (Verbanck et al., 2018). Additionally, MR-Egger regression was utilized to test for potential horizontal pleiotropy by examining the regression intercept. Furthermore, the Steiger test was employed to rule out potential inverse associations (Hemani et al., 2017).

Statistical analyses were conducted utilizing R version 4.2.2, employing the “TwoSampleMR”, “MRPRESSO”, “meta” and “MVMR” packages. Odds ratios (ORs) with corresponding 95% confidence intervals (CIs) quantified the MR analysis, with a p < 0.05 denoting statistical significance.

After the initial screening of SNPs, and subsequent removal of variants in potential linkage disequilibrium (threshold: r ² = 0.001, 10,000 kb) and applying Steiger filtering, a total of 433, 84, 117, 232, and 90 SNPs were used as IVs for the number of smoking initiations, cigarettes per day, lifetime smoking index, drinks per week, and alcohol intake frequency, respectively. These SNPs explained 1.5%, 1.27%, 1.05%, 0.83%, and 2.13% of the phenotypical variance, respectively. Importantly, all the included SNPs had F-values exceeding 10, indicating a minimal likelihood of weak IVs bias (Supplementary Tables S2–S11).

The absence of heterogeneity and the absence of directional pleiotropy were demonstrated in all analyses, as shown in Table 2. Furthermore, the MR-PRESSO analyses did not identify any outliers, indicating a robust dataset (all p for Global test >0.05) (Table 3).

Genetic predisposition to increased drinks per week was found to be significantly associated with a decreased risk of upper urinary calculi in both the FinnGen consortium and United Kingdom Biobank study. The combined OR for upper urinary calculi was 0.6 (IVW 95% CI: 0.49–0.74; p = 1.31 × 10⁻⁰⁶) per standard deviation decrease in drinks per week. Interestingly, the genetically predicted alcohol intake frequency was associated with a significantly increased risk upper urinary calculi (IVW combined OR = 1.27; 95% CI: 1.11–1.45; p = 0.0005). Our study found no association between smoking initiation, the number of cigarettes per day, and the lifetime smoking index and the risk of upper urinary tract stones (Figure 2). These results were consistently reproduced in supplementary analyses (Supplementary Figure S1).

The PhenoScanner search identified links between IVs and traits related to education and obesity. After conducting multivariate IVW analyses, considering several potentially relevant adjustments including education and BMI, we observed that drinks per week remained significantly associated with a lower risk of upper urinary calculi (IVW combined OR = 0.52; 95% CI: 0.36–0.75; p = 0.001). However, the association between alcohol intake frequency and upper urinary calculi became non-significant (IVW combined OR = 1.07; 95% CI: 0.97–1.17; p = 0.174). Furthermore, when adjusting for the lifetime smoking index, alcohol intake frequency was found to be associated with a higher risk of upper urinary calculi (IVW combined OR = 1.90; 95% CI: 1.27–2.83; p = 0.002) (Figure 3).

Our MR analysis has the following advantages. Firstly, this is the first large-scale use of GWAS data to infer causal associations between smoking, alcohol consumption and upper urinary calculi, thereby reducing confounders and reverse bias. Secondly, the study population included only individuals of European origin, minimizing population stratification interference. Finally, sensitivity analyses and different model estimations were used to ensure the reliability of the results.

However, certain limitations are unavoidable. Firstly, the results of the study failed to be validated in other populations. Moreover, smoking and drinking habits are often combined, making it difficult to analyze them as independent factor. Secondly, there was a lack of data on alcohol type and consumption levels. Also, there was a failure to differentiate comparisons between genders, and multivariate analyses were unable to overcome bias because of multiplicity of effects in pathways other than education or obesity. Finally, future studies require larger sample sizes and precise stratified analyses to identify underlying physiopathological mechanisms.