Association Between Diabetes and the Risk of Kidney Cancer: Systematic Review and Meta-Analysis

We accomplished the recommendations described by the Cochrane Collaboration and PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) for conducting and writing systematic reviews.

The inclusion criteria for our study were: 1) articles describing case-control and cohort studies, 2) with the exposure variable type 1 and/or type 2 diabetes, and 3) with information available with which to calculate odds ratios (ORs), relative risks (RRs), or hazard ratio (HRs) with 95% CIs. We excluded studies that included the diagnosis of gestational diabetes, those with participants <18 years of age, and those not calculating OR, RR, or HR as an association measure.

We performed a search in the MEDLINE, Embase, LILACS (Latin American and Caribbean Health Sciences Literature), and CENTRAL (Cochrane Central Register of Controlled Trials) indexes from inception to the present (Supplementary Material). To ensure literature saturation, we scanned references from relevant articles identified through the search, conferences, thesis databases, and the Open Gray, Google Scholar, and ClinicalTrials.gov websites, among others. There were no language or setting limits.

All the authors independently screened the results for a relevant title/abstracts and then full-text articles according to the inclusion and exclusion criteria. We then manually searched for references listed in the selected articles.

We extracted the following information: first author’s last name, year of publication, country, and continent of origin; duration of follow-up; cases in exposed and unexposed groups by sex; diabetes type, method of exposure determination, diagnostic criteria for diabetes, and time of exposure for diabetes; association measure estimates and their corresponding 95% CIs controlling for the significant potential confounders; and covariates controlled for in the analysis.

We assessed the risk of bias based on the New Castle–Ottawa Scale (NOS). This scale evaluates three categories within each study: study design, comparability, and event, using a numeric grading scale from 0 to 9. Studies graded 0–3 were determined to be of low quality, those graded 4–6 were considered to be of intermediate quality, and those graded 7–9 were deemed high-quality studies (10).

For meta-analysis calculation purposes, we selected HRs with 95% CIs as the summary measure for results of all included studies. This ratio was calculated using the method described by DerSimonian and Laird (11), assuming a random effect model. Statistical heterogeneity between studies was calculated using the I² statistic (12), considering 25, 50, and 75% as cutoffs for low, intermediate, and high heterogeneity, respectively. All statistical analyses were performed using Review Manager, v. 5.3, software.

Publication bias was evaluated using the funnel plot graphical method. Sensitivity analyses were conducted by type of study, with adjustment for the covariates of obesity, hypertension, and cigarette smoking because these factors are strongly related to kidney cancer occurrence (13–15). Subgroup analyses were performed to investigate the possible origin of heterogeneity between studies considering the variables of continent of origin, sex, and time of exposure.

Through our search of databases, we found 42 potential articles for full-text analysis. We included 17 studies for the final analysis (Figure 1) (16–32).

Of the included studies, 12 were cohort studies (16–19,25–32) and five were case-control studies (20–24) (Tables 1 and 2). In total, we analyzed 2,296,993 people diagnosed with diabetes and 5,425,793 people without diabetes from the cohort studies. Also, we included for analysis 3,037 patients diagnosed with renal cancer, and 7,309 control subjects without kidney cancer.

With regard to the geographical distribution of the populations analyzed, one of the studies included populations from different regions (24); five included populations from Europe, including Italy (22,25), Iceland (26), Greece (21), and Sweden (23); six were conducted in Asia, including China (32), Israel (31), Japan (18), and Taiwan (20,27,28), and five were from North America, all of which were conducted in the United States (16,17,19,29,30).

With regard to type of diabetes included in the studies, six studies did not report type of diabetes (16,21,22,24,30,31), four included people with type 1 or type 2 diabetes (20,23,25,28), and seven included only people with type 2 diabetes (17–19,26,27,29,32). None of the studies were conducted in a population that was exclusively diagnosed with type 1 diabetes (Tables 1 and 2).

With regard to covariate adjustments for hypertension, obesity, and cigarette smoking, five studies adjusted by the three variables simultaneously (29%), of which one was a case-control study (20) and four were cohort studies (15,25,26,29); nine adjusted based on at least one of the three variables (53.9%), of which three were case-control studies (22–24) and six were cohort studies (17–19,28,29,32); two adjusted by other covariates (11.7%), both of which were cohort studies (25,31); and one case-control study did not have any adjustments (21) (Tables 1 and 2).

Among the studies included, we assessed 13 as being of high quality based on the NOS (16–19,21,23,26–32). We rated the remaining studies as being of intermediate quality (20,22,24,25). None of the studies included were deemed to be of low quality (Tables 1 and 2).

We found an association between diabetes and kidney cancer (HR 1.36, 95% CI 1.25–1.48, I² = 60%) (Figure 2).

In the cohort study subgroup, we included 12 studies (16–19,25–32), and in the case-control subgroup, we included five studies (20–24). The cohort studies yielded an HR of 1.38 (95% CI 1.25–1.52, I² = 71%), whereas the case-control subgroup yielded an HR of 1.29 (95% CI 1.08–1.53, I² = 0%). There were no statistically significant differences between the groups.

We divided the included studies into two subgroups: those with adjustment for the covariates of obesity, hypertension, and/or cigarette smoking (16–24,26–30,32) and those with no adjustment (25,31). We found an HR of 1.26 (95% CI 1.20–1.33, I² = 10%) in the covariate adjustment group and an HR of 1.64 (95% CI 1.38–1.95, I² = 49%) in the group with no adjustment. There were no significant differences between groups.

We performed a publication bias analysis using the visual exploration of a funnel plot. It resulted in a symmetrical graph, which indicates no publication bias (Figure 3).

We included seven studies for meta-analysis (17,18,24,25,28,29,31), as the remaining 10 studies did not included results by sex. We found an HR of 1.41 (95% CI 1.13–1.75, I² = 80%) in women and an HR of 1.29 (95% CI 1.14–1.46, I² = 56%) in men. There were no significant differences between groups (χ² = 0.44, df = 1, P = 0.51, I² = 0%) (Figure 4).

Sixteen studies were grouped in three categories according to their continent of origin: five from Europe (21–23,25,26), six from Asia (18,20,27,28,31,32), and five from North America (16,17,19,29,30). One study was from multiple countries and therefore was not included in this analysis. We found an HR of 1.21 (95% CI 0.99–1.19, I² = 0%) in studies from Europe, an HR of 1.46 (95% CI 1.28–1.66, I² = 80%) in studies from Asia, and an HR of 1.23 (95% CI 1.08–1.39, I² = 16%) in studies from North America. There were no significant differences between groups (χ² = 3.98, df = 2, (P = 0.14, I² = 49.8%) (Figure 5).

We created two subgroups: five studies with <5 years of diabetes exposure (17,19,23,31,32), for which we found an HR of 1.93 (95% CI 1.13–3.31, I² = 98%), and six studies with >5 years of diabetes exposure (17,19,23,25,31,32), for which we found an HR of 1.38 (95% CI 1.26–1.52, I² = 16%). Six studies were not included, as they did not report exposure time. Nevertheless, there were no statistically significant differences between groups (χ² = 1.42, df = 1, P = 0.23, I² = 29.5%) (Figure 6).

The main finding of this systematic review and meta-analysis, which included cohort and case-control studies from different geographical areas of the world, suggests that diabetes increases the risk of kidney cancer. This association seems to be higher in women than in men, in Asia and North America than in Europe, and in people with <5 years of diabetes exposure, although these differences were not statistically significant. The sensitivity analysis showed that the main result was slightly attenuated by studies that adjusted for obesity, hypertension, and cigarette smoking, but it remained statistically significant after adjustment. This finding suggests that diabetes is an independent risk factor for the incidence of kidney cancer.

Our results are consistent with those described in the systematic reviews of Bao et al. (8), which showed an increased 40% risk of kidney cancer in people with diabetes, and Larsson and Wolk (9), which showed an increased risk of 42% in this population.

Elevated blood glucose levels can trigger metabolic processes such as increased insulin and IGF-1 levels and promote the presence of proinflammatory cytokines in tissues that activate signaling pathways involved in cell growth and thus the development of tumors. Therefore, diabetes may increase the risk of cancer in a biologically plausible manner (5–7). None of the included studies evaluated the dose-response relationship between A1C in people with diabetes and the risk of renal cancer. Future research on this topic is needed.

The covariate adjustments could explain the high heterogeneity observed within the meta-analysis. However, this finding should be taken cautiously because there were few studies included in the no-adjustment subgroup. Nevertheless, adjustment for covariates did not affect the association between diabetes and kidney cancer risk.

The subgroup analysis did not reveal any statistically significant differences. Therefore, the heterogeneity might have resulted from the type of diabetes, which authors did not describe in six of the 17 included studies; in the other studies, participants were described as having type 2 diabetes, diabetes, or either type 1 or type 2 diabetes. Similarly, other variables such as the diagnostic criteria for diabetes used in the included studies or the sources of the data collected in these studies may have influenced the results. In addition, other variables such as the use of metformin in the populations with diabetes, which has demonstrated a possible protective effect for the development of various cancers (33), may have mitigated the real effect of diabetes on the risk of kidney cancer.

With regard to sex, this systematic review and meta-analysis showed an increased risk of kidney cancer in both sexes, but a slightly higher risk among women. This finding can be explained by differences in glucose homeostasis and the prevalence of metabolic syndrome, as women tend to have higher glucose intolerance and a higher prevalence of metabolic syndrome than men (31). This occurrence accounts for the possible biological plausibility that exists between glucotoxicity and oncogenesis (34).

This study also found a statistically significant association between diabetes and the risk of kidney cancer in populations from Asia and North America, with a higher risk in Asian populations. However, no association was found for European populations. These findings are consistent with those previously described in the study by Bao et al. (8). Genetic differences between these populations could account for these results. Nevertheless, further investigation is needed.

Finally, this systematic review and meta-analysis found a more significant association between diabetes and kidney cancer in people with a lower exposure to diabetes. In this regard, there is a high probability of detection bias in this association because there was more active surveillance (more imaging diagnostic techniques used) in patients with chronic kidney disease in the population recently diagnosed with diabetes (9).

This review has several strengths. First, the studies included in the analysis were of high or intermediate quality based on NOS assessment. Second, the subgroup analyses included potential cofounders in the summary association measure between diabetes and risk of kidney cancer, making the study result reliable. Third, the included studies accounted for different populations around the world, ensuring adequate external validity for the results.

This study also has limitations. First, there was a high degree of statistical heterogeneity. There were also covariates present in studies other than obesity, hypertension, and cigarette smoking. The studies included were observational in design and thus could have had various types of bias, as well as under-registration of variables that may have interfered with the outcomes they measured. Additionally, data on diagnoses of diabetes and kidney cancer were obtained from clinical records or external sources such as cancer or population registries. This strategy may have introduced selection and information biases. Also, type 2 diabetes is the form of the disease most consistently associated with kidney cancer in epidemiological studies. Thus, inclusion of individuals with type 1 diabetes in some of the studies could have led to attenuation of the real association between diabetes and kidney cancer. Finally, as stated in the meta-analysis summary, diabetes is an underdiagnosed disease, which may have resulted in bias due to misclassification of exposure.

Based on the results of this study, diabetes may be considered an independent risk factor for kidney cancer. Additionally, we suggest incorporating a renal ultrasound into the regular follow-up care of patients with diabetes in primary care.

Diabetes is an independent risk factor for the development of kidney cancer. Future studies are needed to evaluate a possible dose-response interaction of A1C and kidney cancer risk and to assess kidney cancer risk in people with diabetes from other geographical locations such as Latin America, Africa, and Oceania.

No potential conflicts of interest relevant to this article were reported.

A.B.-S. and J.R.-M. researched data, wrote the manuscript, and contributed to the discussion. H.A.G.-P. researched data, reviewed and edited the manuscript, and contributed to the discussion. A.B.-S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

This article contains supplementary material online at https://doi.org/10.2337/figshare.19131905.