Causal Association between Chronic Kidney Disease and Risk of 19 Site-Specific Cancers: A Mendelian Randomization Study

Chronic kidney disease (CKD), depicted by reduced estimated glomerular filtration rate (eGFR) or presence of albuminuria, is a challenging global health issue. Cancer is another widely recognized problem, causing around 10 million deaths per year and ranking as the second leading cause of death following behind cardiovascular disease (1). Its substantial disease burden, lack of solution, and high mortality risk, are needed for early detection and effective controlling.

Previous observational studies indicated potential adverse effect of CKD on both overall (2–6) and site-specific cancers, including genitourinary (4–12), hematopoietic (5, 6, 8), skin (5, 13), digestive (4, 6, 9), lung (2, 6, 8, 14), breast (11, 15), and thyroid cancer (9), but yet yielded conflicting conclusions on which types of cancer patients with CKD are predisposed to. For example, according to a meta-analysis of six prospective studies by Wong and colleagues, in dialysis patients, the incidence of urinary tract and hematologic cancers reached 4.1 and 1.5 per 1,000 person-years, higher than 1.0 per 1,000 person-years for those with eGFR > 75 mL/minute/1.73 m² for both cancers; and the mortality rate were both 0.8 per 1,000 person-years, while it was 0.1 and 0.6 per 1,000 person-years for those with eGFR > 75 mL/minute/1.73 m², respectively. Adjusted HRs revealed a trend towards increased risk of urinary tract cancers among patients with declining renal function (9). Nevertheless, data from the Atherosclerosis Risk in Communities study showed that, for site-specific cancers, only albuminuria, but not decreased kidney function, showed a significant association with lung cancer (14).

While providing paramount insights, these associations cannot be directly interpreted as causality, partly due to the inevitable existence of unmeasured confounding factors. Besides, one cannot rule out the possibility that reverse cancer-to-CKD causation leads to the comorbidity status, since cancer itself and cancer therapy can induce renal toxicity (16). Mendelian randomization (MR), based on Mendel's law of independent assortment, is an epidemiologic approach that strengthen the inference of causal relationship between modifiable exposures and health-related outcomes (17). Recent genome-wide association studies (GWAS) have identified a number of single-nucleotide polymorphisms (SNP) significantly associated with renal phenotypes (18, 19). Given the random allocation of genetic variants during conception, using these genetic variants as proxies representing long-term exposure of kidney function and albuminuria unaffected by environmental influences, minimizes the impact of confounding and reverse causation on the inference of CKD-cancer association.

In this study, in order to explore whether CKD is causally related to the elevated risk of overall and site-specific cancers that have been identified in previous observational studies, we conducted an MR study using SNPs strongly correlated with decrement in eGFR and albuminuria as the genetic instruments, and cancer data from the UK Biobank, the FinnGen consortium as well as several large cancer-specific consortia.

In this study, exposures of interest included kidney function, a continuous variable measured by eGFR, and albuminuria (i.e., urine albumin to creatinine ratio ≥ 30 mg/g), a binary variable defined by presence or absence of the condition. To ensure the validity of genetic instruments, we searched recent large GWAS for qualified SNPs significantly correlated with the exposures.

A meta-analysis of European-ancestry GWAS of the CKDGen consortium, independent from the UK Biobank, identified 243 SNPs associated with eGFR at the genome-wide significance (i.e., P < 5 × 10^–8) in both marginal and conditional analysis (18). Correlated pairs of variants were further detected using a window of 10 Mb and maximal linkage disequilibrium of r² = 0.01 as threshold in the MR Base “clumping” procedure, and the variant with lower P value within each pair was retained.

In order to maintain the strength of genetic instruments for albuminuria, we selected 17 SNPs representative of albuminuria from a meta-analysis of GWAS combining CKDGen and UK Biobank participants (constituted of over 95% of Europeans; ref. 19). All 17 SNPs are independent (i.e., r² < 0.01). Four SNPs obtained from the sample completely independent of the UK Biobank were also used for sensitivity analysis (20).

Palindromic SNPs (i.e., SNPs with A/T or C/G) with intermediate allele frequency (i.e., 0.42–0.58) were excluded from the analysis. If chosen SNPs were not presented in the outcome data source, proxy SNPs (i.e., r² > 0.8) were used. Lists of instrumental SNPs for kidney function and albuminuria are presented respectively in Supplementary Tables S1 and S2. The corresponding F-statistics were 59 to 74, indicating sufficient strength (21).

The outcomes of interest were genitourinary (2, 3, 6–12), hematopoietic (2, 3, 8), skin (3, 13), digestive (2, 6, 9), lung (2, 5, 8, 14), breast (11, 15), and thyroid cancers (9) that were reported to be associated with CKD in previous observational studies. Characteristics of these observational studies are listed in Supplementary Table S3. In the primary analysis, cancer-specific consortia, if available, were preferred as data sources; otherwise, the population-based UK Biobank was used.

We retrieved summary statistics of the association between SNPs and lung cancer from the International Lung Cancer Consortium (11,348 cases and 15,861 controls; ref. 22), breast cancer from the Breast Cancer Association Consortium (122,977 cases and 105,974 controls in females; ref. 23), ovary cancer from the Ovarian Cancer Association Consortium (25,509 cases and 40,941 controls in females; ref. 24), endometrium cancer from the Endometrial Cancer Association Consortium (12,906 cases and 108,979 controls in females; ref. 25), prostate cancer from the Prostate Cancer Association Group to Investigate Cancer Associated Alterations (79,148 cases and 61,106 controls in males; ref. 26), and renal cell carcinoma (RCC) from the International Agency for Research on Cancer (3,227 cases, 4,916 controls in males, and 1,992 cases, 3,095 controls in females; ref. 27), respectively. All participants included in the aforementioned consortia were of European ancestry.

For cancers without publicly available summary statistics from external data sources, the individual-level genetic data from the UK Biobank was used to derive SNP-cancer associations. The UK Biobank is a large-scale biomedical database, containing in-depth genetic and health information of 500,000 participants aged 40 to 69 years from England, Scotland, and Wales (28). Health-related outcomes were obtained periodically from external data providers, with the consent of participants (29). Participants were genotyped by UK BiLEVE and UK Biobank Axiom array, which share 95% common markers, and variants were imputed using Haplotype Reference Consortium, as well as merged UK10K and 1000 Genomes phase III reference panels (30).

In this study, we excluded UK Biobank participants: (i) who withdrew consents (n = 48); (ii) who were not of validated European genetic background (n = 92,875); (iii) with first or second degree of relatedness to others (i.e., with kinship coefficients over 0.0884, n = 60,768; ref. 31); (iv) who did not pass the quality control of genetic data (i.e., with inconsistent self-reported and genetic sex, high rate of missingness, and heterozygosity, n = 1,272). Therefore, we derived a cohort of 347,408 participants.

In the UK Biobank, cancer cases, identified by using International Classification of Diseases, ninth and tenth Revision codes, were ascertained from hospital inpatient records linked to Hospital Episode Statistics for England, Scottish Morbidity Record, Patient Episode Database for Wales, as well as cancer and mortality data available from National Health Service (NHS) Digital in England and Wales, NHS Central Register in Scotland. Self-reported information from the verbal interview conducted by a trained nurse was also collected (Supplementary Table S4). In total, 68,871 overall and 19 site-specific cancer cases were identified. We then calculated log odds and corresponding standard errors of SNP-cancer associations using –glm command of PLINK 2.0 with an addictive model adjusting for age, sex, third degree of relatedness, first 10 principal components, and genotyping arrays. The detailed results of SNP-cancer associations are presented in Supplementary Table S5. In addition, estimates derived from generalized linear mixed model performed by Yang Lab, which accounts for sample relatedness, were also used for confirmation (32).

All the UK Biobank participants gave written informed consent before data collection. The UK Biobank has full ethical approval from the NHS National Research Ethics Service (16/NW/0274). The study conformed to the Declaration of Helsinki. And this study was approved by the biomedical research ethics committee of West China Hospital (Chengdu, Sichuan, 2019–1171).

In the primary analysis, inverse variance weighted (IVW) method, under random-effect model, was first applied. The IVW approach essentially fits a weighted regression model of SNP-outcome effects on SNP-exposure effects where the intercept is constrained to zero. To be specific, Wald ratios of β coefficients of SNP-cancer associations divided by those of SNP-kidney function or albuminuria associations were calculated and then meta-analyzed (33). We also used likelihood-based method, which assuming a different distribution of variants.

To address the potential directional pleiotropic effects, we conducted MR-Egger regression, detecting and correcting pleiotropy by allowing a nonzero intercept term (34). Alternative median- or mode-based weighted methods, which allow some invalid instrumental variables and increase robustness when pleiotropy exists, were also used (35, 36). In addition, we performed MR pleiotropy residual sum and outlier analysis to detect and remove outliers with statistically significant horizontal pleiotropy (37).

To evaluate the direct independent effect of kidney function and albuminuria on cancers, we performed multivariable MR simultaneously including both biomarkers. Smoking and body mass index (BMI) were further adjusted to control for the influence of other confounders (38). We chose 378 SNPs significantly associated with smoking initiation (i.e., ever smoked regularly) in the GWAS & Sequencing Consortium of Alcohol and Nicotine use consortium and 77 with BMI in the Genetic Investigation of ANthropometric Traits consortium (39, 40).

Combing these genetic instruments together, SNPs were checked for duplication and clumped by selecting the SNP within a region most strongly associated with eGFR from the summary statistics (41). The Sanderson–Windmeijer conditional F-statistics indicated the joint strength of instruments was adequately strong for kidney function, smoking initiation and BMI, however, did not reach 10 for albuminuria as recommended (42). Therefore, the results of multivariable MR incorporating albuminuria should be interpreted with caution due to the potential weak instruments bias.

For cancers causally associated with kidney function or albuminuria in the above-mentioned primary analysis, we performed bidirectional MR to examine whether reverse causation from cancers to CKD also existed. We used SNPs detected by Laskar and colleagues as the genetic instruments representative of RCC (27). For other cancers, we performed GWASs in the UK Biobank following the instruction reported by Marees and colleagues (43). Briefly, we applied quality control on the genetic data of UK Biobank by removing SNPs with genotyping rate < 0.01, minor allele frequency < 0.01, and departure from Hardy-Weinberg Equilibrium at P value < 10^–6. GWASs were ran in PLINK 2.0 among 347,408 UK Biobank participants with qualified genetic data, using an addictive model and adjusting for age, sex, third degree of relatedness, first 10 principal components, and genotyping arrays. For multiple SNPs reaching the genome-wide significance (i.e., P < 5 × 10^–8) within the same region, we clumped them using a window of 10 Mb and maximal linkage disequilibrium of r² = 0.01 to ensure that SNPs were independent. The list of genetic instruments, along with the magnitude of their associations with cancers, are presented in Supplementary Table S6.

To confirm the robustness of the results, we applied the individual-level allele score approach among UK Biobank participants. To be specific, genetic risk scores (GRS) were derived by aggregating the number of the risk allele at each locus weighted by its corresponding beta coefficient. Estimates of GRS-cancer associations were obtained from the logistic regression, adjusting for age, sex, third degree of relatedness, first 10 principal components, and genotyping arrays.

To avoid sample overlapping that may introduce bias and inflated type I error (44), we used four SNPs discovered among European participants independent of the UK Biobank as the proxies of albuminuria to validate the observed significant associations in the primary analysis (20).

In the primary analysis of multivariable MR, we clumped the SNPs according to their strength of associations with eGFR. This may mean that the instruments for kidney function were stronger than those for albuminuria. So we repeated the analysis by clumping SNPs according to their strength of associations with albuminuria (41).

Beside the UK Biobank, a replication analysis was performed using GWAS summary data derived from the FinnGen consortium (R6 data release including up to 260,405 participants). Detailed description of FinnGen is presented elsewhere (https://finngen.gitbook.io/documentation/).

The original estimate of MR (β) corresponds to change in log odds for cancer per unit increase of log-transformed eGFR or log odds of albuminuria, and is further scaled to OR per 10% decrease in eGFR, or per doubling in the odds of albuminuria [i.e., 0.9^β and exp (0.693 × β) respectively; ref. 45]. Power of MR to detect the true effect of CKD-cancer association was calculated on a web tool (https://shiny.cnsgenomics.com/mRnd/) and results are presented in Supplementary Table S7 (46).

Analysis were performed in PLINK 2.0 and R 4.0.2 using TwoSampleMR and MVMR packages. P values of IVW method less than 0.05 following the FDR correction (Benjamini–Hochberg method) were considered as statistically significant (47). P values that did not pass multiple testing but were less than 0.05 were considered as suggestive evidence of the associations.

The data underlying this article is available or can be applied from: http://www.ukbiobank.ac.uk/register-apply; https://www.finngen.fi/en/access_results; http://ckdgen.imbi.uni-freiburg.de; https://gwas.mrcieu.ac.uk; https://www.ebi.ac.uk/gwas and cancer consortia as mentioned above.

In the univariable analysis using IVW method, both continuous decrease in kidney function, as measured by reduced eGFR, and albuminuria did not predispose patients to the increased risk of overall cancer [OR_eGFR = 0.99; 95% confidence interval (CI), 0.96–1.03; OR_alb = 1.03; 95% CI, 0.98–1.08; Supplementary Fig. S1]. With respect to site-specific cancers, higher odds of leukemia (OR = 1.23; 95% CI, 1.06–1.43; P = 0.007; P_FDR = 0.1), and in females, cervical cancer (OR = 1.22; 95% CI, 1.04–1.43; P = 0.017; P_FDR = 0.22), RCC (OR = 1.4; 95% CI, 1.12–1.77; P = 0.004; P_FDR = 0.091), per 10% decreased in eGFR was observed (Fig. 1). There was suggestive evidence that increasing odds of albuminuria was a risk factor of colorectal cancer (OR = 1.21; 95% CI, 1.07–1.36; P = 0.002; P_FDR = 0.091), however, inversely associated with non–Hodgkin lymphoma (NHL) incidence (OR = 0.76; 95% CI, 0.62–0.92; P = 0.006; P_FDR = 0.092; Fig. 2).

To address the potential directional pleiotropic effects and weak instruments bias, other MR approaches were also conducted. We did not detect significant heterogeneity, directional pleiotropy and outlier in complementary methods pertaining to the above-mentioned associations (Supplementary Tables S8 and S9). The direction of effect was consistent across all methods, but with broader CIs (Figs. 3 and 4; Supplementary Tables S10 and S11). For the albuminuria-cervical cancer association, observing the presence of substantial heterogeneity (I² = 38.7; Cochran Q = 26; P = 0.053) and pleiotropy (P_intercept in MR-Egger = 0.04; Supplementary Table S11), larger effect sizes were identified in weighted median (OR = 1.39; 95% CI, 1.04–1.83; P = 0.019) and weighted mode (OR = 1.37; 95% CI, 1.04–1.81; P = 0.04) methods which were more robust for the inclusion of certain invalid instrumental SNPs, though the result was insignificant in IVW method (OR = 1.09; 95% CI, 0.84–1.4; Fig. 4; Supplementary Table S9).

Kidney function and albuminuria share one common variant (i.e., rs4410790). After additionally adjusting for albuminuria, the effect size of eGFR–leukemia association decreased (OR = 1.14; 95% CI, 0.97–1.34; P = 0.11), while the inverse associations of eGFR and cervical cancer (OR = 1.21; 95% CI, 1.01–1.45; P = 0.036) and RCC (OR = 1.36; 95% CI, 1.06–1.73; P = 0.015) in females were virtually identical to the univariable analysis (Table 1; Supplementary Table S12). Further adjusting for smoking initiation and BMI to assess the direct impact of kidney function on cancers, only eGFR–cervical cancer associations reached the significance (OR = 1.47; 95% CI, 1.14–1.89; P = 0.003; Table 1; Supplementary Tables S13 and S14). In the multivariable MR, albuminuria was associated with neither colorectal cancer nor NHL.

In addition to genetic instrument for RCC selected from the International Agency for Research on Cancer, a total of 10 SNPs representative of cancers in the UK Biobank were identified in GWASs, including one for leukemia, two for cervical cancer, six for colorectal cancer, and one for NHL (Supplementary Table S6). No reverse causation was observed (Supplementary Table S15).

In the UK Biobank, GRS_eGFR and GRS_alb explain 3.36% and 1.54% of the variance of corresponding renal phenotypes respectively. eGFR increased by 3.1% (P < 0.001) per SD of GRS_eGFR. Participants in the intermediate and high albuminuria risk groups according to GRS_alb were respectively 24% (OR = 1.24; 95% CI, 1.19–1.3; P < 0.001) and 70% (OR = 1.7; 95% CI, 1.62–1.8; P < 0.001) more likely to develop albuminuria. Using combined GRS_eGFR and GRS_alb as exposures, ORs of leukemia and cervical cancer per 10% decrease in eGFR were respectively 1.22 (95% CI, 1.05–1.42; P = 0.01) and 1.18 (95% CI, 1–1.38; P = 0.046), ORs of colorectal cancer and NHL among participants with highest genetic liability to the development of albuminuria were respectively 1.13 (95% CI, 1.03–1.23; P = 0.006) and 0.89 (95% CI, 0.77–1.01; P = 0.078; Supplementary Figs. S2 and S3).

Using four SNPs discovered among European participants completely independent of the UK Biobank as the proxies of albuminuria, the direction of association of albuminuria and colorectal cancer (OR_IVW = 1.08; 95% CI, 0.98–1.19; P = 0.13) and NHL (OR_IVW = 0.91; 95% CI, 0.78–1.05; P = 0.2) was the same as the primary analysis (Supplementary Table S16). In the sensitivity analysis of multivariable MR, results were identical when clumping SNPs by albuminuria in the sensitivity analysis (Supplementary Tables S17 and S18).

Using GWAS summary data of cancer from the FinnGen consortium, the direction of eGFR–leukemia (OR_IVW for lymphoid leukemia = 1.17; 95% CI, 0.91–1.51; P = 0.21; OR_IVW for myeloid leukemia = 1.05; 95% CI, 0.52–1.72; P = 0.51) and eGFR–cervical cancer (OR_IVW = 1.16; 95% CI, 1–1.35; P = 0.052) associations was the same as the primary analysis, though the results did not reach the statistical significance (Supplementary Table S19 and S20). Based on data released by Yang Lab, we found the results remained largely unchanged (Supplementary Tables S21 and S22).

In this study, on the basis of appropriate genetic instruments for the exposures, we investigated the potential causal associations between CKD and cancers using MR approach. We found that genetic liability to decreased eGFR may increase the risk of several cancers of genitourinary and hematological systems. There was also suggestive evidence that albuminuria is a risk factor of colorectal cancer, and may be protective to NHL. No reverse effect of these cancers on kidney function and albuminuria was detected in the bidirectional MR.

It is, to the best of our knowledge, the most comprehensive MR study exploring the causal associations between different clinical manifestations of CKD and overall, as well as 19 site-specific cancers integrating both summary- and individual-level data from large cancer-specific consortia, the UK Biobank and FinnGen consortium. A major strength of the study lies in its MR design, which diminishes confounding factors and reverse causation, thus offering more convincing conclusions than conventional observational studies. The reliability of our results was further ensured by applying multiple complementary MR methods accounting for potential bias.

Our findings of site-specific cancers partly corroborate observational studies in that (2, 3, 7), impaired kidney function may increase the risk of some genitourinary cancers, such as RCC and cervical cancer, and the effect was independent of albuminuria. Among a cohort including more than 1 million participants, graded inverse eGFR–RCC relationships were found when eGFR fell below 60 mL/minute/1.73 m² (7). In this study, sex-specific relationship between decreased eGFR and RCC is indicated in the current study. An MR study examining the urea-cancer association also reported a higher risk of urea-related RCC in females but a null finding in males (48). Possible explanations lying behind the phenomenon may be the physiologically less kidney function reserve of female compared with male. When we further adjusted for smoking initiation and BMI, effect sizes of eGFR–RCC associations decreased but eGFR–cervical cancer associations were further intensified, suggesting a strong direct effect of decreased kidney function on cervical cancer.

It is also notable that the incidence of hematopoietic cancers, such as leukemia increased when kidney function declined, which is aligned with observational studies (5, 6, 8). A large population-based study enrolling 719,033 Swedes ages more than 40 years without prior history of cancer reported that eGFR 30 to 59 and less than 30 mL/minute/1.73 m² associated with risk of hematologic cancer (i.e., NHL, multiple myeloma, and leukemia) with HRs of 1.23 (95% CI, 1.09–1.38) and 1.71 (95% CI, 1.28–2.28; ref. 5). In this study, a higher odds of multiple myeloma per 10% decrease in eGFR was found (OR = 1.23; 95% CI, 0.95–1.61), but failed to reach statistical significance probably due to a lack of power. In multivariable MR, estimates of eGFR–leukemia associations were attenuated with wider CIs overlapping null, suggesting other factors, such as albuminuria, smoking initiation, and BMI, may confound the association.

In this study, genetic predisposition of albuminuria was found to have a detrimental effect on colorectal cancer. Mok and colleagues have reported increased risk of rectal cancer among participants with proteinuria (6). In contrast, we surprisingly observed a preventive effect albuminuria exerts on NHL, and the results were consistent across all statistical methods. It is notable that SNPs we used for albuminuria genetically correlated with hyperlipidemia as confirmed in the study of Teumer and colleagues (19). Previous observational studies have indicated a protective effect of lipids and lipoprotein on NHL, and an MR study further found a positive tendency of high-density cholesterol and an inverse one of triglyceride on NHL, though the results did not pass the multiple testing (49). Therefore, we presumed that the altered lipid metabolism profile mediate the association between albuminuria and NHL. Further research is warranted for in-depth elucidation.

The observed carcinogenic effect of CKD is pathophysiologically plausible, since the DNA impairment, oxidative stress, proinflammatory microenvironment, endothelial dysfunction, and activated renin-angiotensin system involved in the progression of CKD (50–52), are all promotors of tumor growth (53–55). Our finding that albuminuria associated with cancer risk distinctly from kidney function merits attention. This may be attributed to the fact that separate molecular pathways exist through which they link to cancer. Among selected genetic instruments, only rs4410790, a variant near AHR gene is shared between kidney function and albuminuria.

The relationship between kidney disease and cancer has been reported to be “circular”. Patients with cancer are vulnerable to declined kidney function, probably due to paraneoplastic kidney injury or the nephrotoxicity of anticancer treatment (56). In this study, we did not observe bidirectional association of kidney damage and cancers, indicating that these cancers per se may not suffice to cause impaired kidney function.

Several limitations have to be recognized as well. First, considering effect sizes of SNPs and some cancers were obtained from the UK Biobank which recruited “healthier” volunteers with low prevalence and incidence of cancer (57), the possibility of imprecise estimation of causal-association due to inadequate power cannot be ruled out. Second, it can provide deeper insight to use sex-specific genetic instruments of kidney function and albuminuria when measuring the effect of kidney damage on cancers in females and males separately. Nevertheless, such data with sufficient statistical power are largely unavailable. Third, the sample overlap in the analyses of albuminuria brought about the propensity of type I error. The sensitivity analysis, in which another GWAS conducted completely outside the UK Biobank was used, showed the same direction of effect as the primary analysis, however, the estimates did not reach the statistical significance probably due to a lack of power. So the possibility of false positive results still could not be excluded. Fourth, in the multivariable MR, the conditional F-statistics of albuminuria was less than 10, indicating that genetic instruments were not strong enough. Besides, UK Biobank participants were included in both the smoking initiation and cancer outcome datasets, introducing bias into estimates. Finally, participants involved in this study were mainly of European ancestry in order to minimize population stratification, extrapolation of the results to other ethnic backgrounds is confined.

In conclusion, the current MR study provided some evidence for the participation of CKD in carcinogenesis, including leukemia and colorectal cancer, and in females, cervical cancer and RCC. The sex-specific effects of CKD on cancer, as well as the potential mechanisms for the CKD-cancer causal association are worth further investigation.

No disclosures were reported.

L. Tang: Conceptualization, data curation, software, formal analysis, investigation, methodology, writing–original draft. C. Li: Formal analysis, investigation. W. Chen: Formal analysis, investigation. Y. Zeng: Formal analysis, investigation. H. Yang: Formal analysis, investigation. Y. Hu: Software, investigation. H. Song: Funding acquisition, investigation, methodology, writing–review and editing. X. Zeng: Conceptualization, funding acquisition, investigation, methodology, project administration, writing–review and editing. Q. Li: Investigation, writing–review and editing. P. Fu: Investigation, writing–review and editing.

Summary-level data for SNPs associated with CKD-related traits were extracted from the CKDGen Consortium. Summary-level data for genetic associations with the cancers were contributed by the International Lung Cancer Consortium, the Breast Cancer Association Consortium, the Ovarian Cancer Association Consortium, the Endometrial Cancer Association Consortium, Prostate Cancer Association Group to Investigate Cancer Associated Alterations and International Agency for Research on Cancer, the UK Biobank, and FinnGen Consortium. The analyses of the UK Biobank data were conducted under application no. 54803. The authors thank all participants and investigators for sharing these data, and the editors and the reviewers for their suggestions to improve the work.

Among the authors, X. Zeng was supported by the 1.3.5 project for disciplines of excellence, West China Hospital, Sichuan University (ZYJC18010); Science and Technology Department of Sichuan Province (2021YF0035); and Chengdu Science and Technology Bureau (2020-YF09–00117-GX). H. Song was supported by the 1.3.5 project for disciplines of excellence, West China Hospital, Sichuan University (ZYYC21005).

A copy of the code used in this analysis is available at: https://github.com/TangLei97/CKD_cancer_MR.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.