Introduction

The incidence of kidney cancer has been increasing worldwide over the past few decades [1]. Besides obesity, cigarette smoking and hypertension, which are associated with greater risk of kidney cancer [2–5], the causes of this disease are largely unknown.

Like kidney cancer, type 2 diabetes is a serious and growing health problem worldwide. Obesity and cigarette smoking are also positively associated with risk of type 2 diabetes [6, 7]. Evidence from epidemiological studies indicates that diabetic patients have an increased risk of cancer at several sites, including cancers of the colon and rectum [8], pancreas [9], liver [10], bladder [11], breast [12] and endometrium [13]. Whether diabetes is also a risk factor for kidney cancer remains unclear. We used meta-analysis to systematically review published cohort studies examining the association between diabetes and incidence of kidney cancer.

Methods

Retrieval of studies

We conducted a computerised search of the PubMed database for studies published in any language between January 1966 and December 2010 using the search term ‘diabetes’, combined with ‘kidney cancer’, ‘renal cell cancer’ or ‘renal cell carcinoma’. We also manually reviewed the references of relevant publications to search for further studies and reviewed publications on diabetes and risk of total and specific cancers to search for results on kidney cancer incidence.

Inclusion and exclusion criteria

In this meta-analysis, we included studies that fulfilled the following criteria: (1) publication presented original data from a cohort study; (2) the exposure of interest was diabetes mellitus; (3) the outcome of interest was incidence of kidney or renal cell cancer; and (4) RR estimates with CIs or enough data to calculate them were reported. We summarised results from cohort studies because these are less prone to bias than case–control studies. We did not consider studies in which the exposure of interest was mainly or solely type 1 diabetes, defined as diagnosis before 30 years of age. Studies on mortality rates from kidney or renal cell cancer were not included. In the event of multiple publications from the same study population, the most recent publication with the largest number of kidney cancer cases was included in the meta-analysis.

Data extraction

We extracted the following information from each publication: the first author’s last name, publication year, study location, age and sex of participants, sample size (cases and cohort size), diabetes assessment and type of diabetes, variables controlled for in the analysis and RR estimates with CIs for the association between diabetes and kidney cancer incidence. From each study, we extracted the RR estimate that was adjusted for the greatest number of potential confounders.

Statistical analysis

Summary RRs and their 95% CIs were calculated with the method of DerSimonian and Laird using the assumptions of a random-effects model that considers within-study and between-study variation [14]. We investigated possible heterogeneity in results across studies by using the Q and I 2 statistics [15]. I 2 is the proportion of total variation contributed by between-study variation [15]. To assess for publication bias, we applied regression methods to determine funnel plot asymmetry as proposed by Egger et al. [16]. A p value of less than 0.10 was considered statistically significant. All statistical analyses were performed with STATA version 10.1 (StataCorp, College Station, TX, USA).

Results

We identified 15 potentially relevant cohort studies [17–31]. Six studies were excluded because of overlapping publications from the same study population [17, 19] or because the outcome was cancer mortality [28–31]. Thus, a total of nine cohort studies met our inclusion criteria (Table 1). Four studies were conducted in Europe, three in Asia and two in North America. The study population in six studies consisted of both sexes [18, 20, 23–25, 27], two studies included men only [22, 26] and one study included women only [20]. The outcome was renal cell carcinoma in three studies [20, 23, 25], and total or unspecified kidney cancer in six studies [18, 21, 22, 24, 26, 27]. Two studies used a 1 year latency period in all analyses [18, 26].

Table 1 Characteristics of studies of diabetes and kidney cancer incidence, ordered by year of publication
Full size table

Analysis of all studies showed that, compared with individuals without diabetes, those with diabetes had a statistically significant increased risk of kidney cancer (1.42, 95% CI 1.06–1.91; Fig. 1). There was statistically significant heterogeneity among studies (p < 0.001 for heterogeneity; I 2 = 95.2%). In analysis stratified by sex, a significantly stronger (p = 0.01) positive association was observed in women (RR 1.70, 95% CI 1.47–1.97; p = 0.98 for heterogeneity; I 2 = 0%) than in men (RR 1.26, 95% CI 1.06–1.49; p = 0.01 for heterogeneity; I 2 = 69.9%). The association between diabetes and kidney cancer was significantly stronger (p = 0.05) in studies with a standardised incidence ratio as the measure of relative risk (RR 1.70, 95% CI 1.21–2.40; p < 0.001 for heterogeneity; I 2 = 94.3%) than in studies with a rate ratio as the measure of relative risk (RR 1.17, 95% CI 1.00–1.37; p = 0.25 for heterogeneity; I 2 = 23.1%). In stratified analysis by outcome, the RRs were 1.12 (95% CI 0.57–2.22; p = 0.18 for heterogeneity; I 2 = 37.1%) in studies of renal cell carcinoma and 1.50 (95% CI 1.09–2.08; p < 0.001 for heterogeneity; I 2 = 96.9%) in studies of total or unspecified kidney cancer. The difference in risk estimates by outcome was not statistically significant (p = 0.45). We found no evidence of publication bias (Egger’s test: p = 0.75).

Fig. 1
figure 1

RRs for the association between diabetes mellitus and kidney cancer incidence in cohort studies. The black squares and horizontal lines represent the study-specific relative risk and corresponding 95% CI, respectively. The size of the black squares reflects the weight of the respective study, measured by the inverse of the variance. The diamond represents the pooled relative risk with its 95% CI. Test for heterogeneity among studies: Q = 229.85; p < 0.001; I 2 = 95.2%. M, men; W, women

Full size image

Obesity and cigarette smoking are risk factors for diabetes and for kidney cancer, and thus potential confounders of the relationship between diabetes and risk of kidney cancer. When we restricted the analysis to the three studies that had adjusted for body mass index or obesity [23, 24, 26], the RR was 1.12 (95% CI 0.99–1.27; p = 0.39 for heterogeneity; I 2 = 2.4%). Among the three studies that controlled for cigarette smoking [22–24], the pooled RR was 1.29 (95% CI 1.05–1.58; p = 0.66 for heterogeneity; I 2 = 0%).

Six studies examined the risk of kidney cancer by interval between start of follow-up and kidney cancer diagnosis [18, 23–27]. When the RR estimates obtained after excluding 1, 2 or 5 years of follow-up were combined [18, 23–27], the pooled RR of kidney cancer was 1.48 (95% CI 1.13–1.94; p < 0.001 for heterogeneity; I 2 = 91.5%). Three studies examined the effect of excluding the first 5 years of follow-up [24] or used a latency period of 5 years between diabetes diagnosis or hospitalisation and diagnosis of kidney cancer [26, 27]. When the risk estimates from those studies were combined, the pooled RR of kidney cancer associated with a long history of diabetes (≥5 years) was 1.35 (95% CI 0.98–1.86; p = 0.005 for heterogeneity; I 2 = 76.5%).

Discussion

The findings of this meta-analysis of nine cohort studies indicate that diabetes is associated with a 42% increased risk of kidney cancer. The positive association was stronger in women than in men. It was also stronger in studies with a standardised incidence ratio as the measure of relative risk than in studies with a rate ratio, possibly due to lack of adjustment for potential confounders such as obesity and cigarette smoking.

Whether the association between diabetes and kidney cancer is causal is unclear. In this respect, several points should be considered. First, because our analyses were based on observational studies, confounding cannot be excluded as a potential explanation for the observed positive association between diabetes and risk of kidney cancer. Obesity has consistently been associated with increased risk of diabetes [7] and kidney cancer [4]. Only three studies controlled for body mass index or obesity [23, 24, 26]. The relation between diabetes and kidney cancer was weaker and not statistically significant when we restricted the analysis to those studies, suggesting that the observed association may in part be due to confounding from obesity. Cigarette smoking is another shared risk factor between type 2 diabetes and kidney cancer [2, 6]. When risk estimates from the three studies that adjusted for smoking were combined [22–24], the association between diabetes and kidney cancer was weaker (RR 1.29) than the overall result including all studies (RR 1.42).

Second, detection bias may also have contributed to the observed association between diabetes and kidney cancer risk because diabetic patients are under increased medical surveillance. If medical surveillance bias is present, kidney cancer would tend to be diagnosed at an earlier stage in diabetic patients than in those without diabetes. Two studies examined the risk of kidney cancer by interval between diabetes diagnosis [26] or last hospitalisation for type 2 diabetes [27] and diagnosis of kidney cancer. In the study by Atchison et al. [26], the risk of kidney cancer declined over time and was no longer significantly increased at 10 years after diagnosis of diabetes. The RRs of kidney cancer for an interval between diagnoses of diabetes and kidney cancer were 1.24 (95% CI 1.11–1.38) after 2 to 5 years, 1.09 (95% CI 0.97–1.22) after 6 to 10 years and 1.00 (95% CI 0.91–1.10) after >10 years [26]. In the study by Hemminki et al. [27], the standardised incidence ratio of kidney cancer declined from 2.50 with no latency period to 1.44 for a 5 year latency period. In the study by Ogunleye et al. [25], the RR of kidney cancer comparing diabetes with no diabetes was 0.52 (95% CI 0.21–1.27) when including all cases and 0.25 (95% CI 0.06–1.06) when excluding cases diagnosed within the first year of follow-up. In contrast, two other studies found a stronger positive association between diabetes and kidney cancer risk after excluding the first 2 or 5 years of follow-up [23, 24]. Because of these inconsistent results, it remains unclear whether the observed positive association between diabetes and kidney cancer risk is due to detection bias.

Because diabetes is an underdiagnosed disease, some misclassification of exposure is likely, which would tend to attenuate any true association between diabetes and kidney cancer. Diabetes was diagnosed and assessed differently in the studies. In some studies, diabetes was self-reported [20, 23, 24], whereas other studies assessed diabetes through diabetes hospitalisation or diagnosis [18, 21, 25–27]. Some studies included type 1 and type 2 diabetes, which would lead to an attenuation of the risk estimates if only type 2 diabetes is a risk factor for kidney cancer. In fact, a population-based cohort study in Sweden found no excess risk of kidney cancer in hospitalised type 1 diabetes patients [32]. Moreover, because total kidney cancer was assessed in most studies, the overall association would be attenuated if diabetes were a risk factor for renal cell (parenchyma) or renal pelvis cancer only. In this meta-analysis, diabetes was more strongly associated with total and unspecified kidney cancer than renal cell cancer. Only one study reported results for renal cell and renal pelvis cancer [19]. In that study, the association between diabetes and risk of kidney cancer was similar for cancers of the parenchyma (standardised incidence ratio 1.5; 95% CI 1.3–1.7) and pelvis (standardised incidence ratio 1.4; 95% CI 1.0–1.9) [19].

There was heterogeneity among study results. The heterogeneity of risk estimates may be due to different adjustment for potential confounders and different mixtures of type 1 and type 2 diabetic participants. Furthermore, some studies included both sexes, whereas others included only men or only women. If the association is stronger in women, as suggested in this meta-analysis, different proportions of men and women in the studies may in part account for the observed heterogeneity.

Inherent in any meta-analysis of published data is the possibility of publication bias, that is small studies with null results tend not to be published. However, we found no evidence of publication bias in this meta-analysis.

Diabetes may affect the risk of kidney cancer by increasing insulin resistance and blood levels of insulin and IGF-I [33, 34]. Experimental evidence indicates that insulin could affect malignant transformation or tumour growth through its anabolic and anti-apoptotic growth factor effects or by increasing the levels of bioactive IGF-I [35, 36]. IGF-I may stimulate cell proliferation and inhibit apoptosis [37]. IGF-I levels have been positively associated with risk of several cancers (e.g. breast and prostate cancer) [38, 39]. However, a study among older middle-aged male smokers found that low serum IGF-I levels were associated with increased risk of kidney cancer [40]. This finding suggests that IGF-I may not explain the positive association between diabetes and kidney cancer risk. Diabetes may also potentially increase the risk of kidney cancer through hypertension, which has been positively associated with risk of kidney cancer [3, 5]. Indeed, type 2 diabetes and hypertension frequently coexist [41].

In summary, the available epidemiological evidence indicates that individuals with diabetes have an increased risk of developing kidney cancer. The possibility that the association may be due to confounding or detection bias cannot be ruled out. Additional large cohort studies that adjust for obesity and other potential confounders as well as assessing latency between diabetes and kidney cancer are needed to establish a potential causal relationship between diabetes and kidney cancer risk. Future research should also assess whether the association differs for renal cell and renal pelvis cancer.