Diabetes and Risk of Bladder Cancer: Evidence from a Case-Control Study in New England

Abstract

Background

Diabetes is an emerging public health issue in the US, affecting 11% of Americans over the age of 20, and with long-term complications that include cardiovascular disease, retinopathy, neuropathy and nephropathy. A recent meta-analysis found that bladder cancer incidence was approximately 40% higher in individuals with diabetes; however, few studies considered duration or type of therapy and had limited adjustment for potentially confounding factors.

Methods

To further investigate the potential role of diabetes in risk of bladder cancer, we analyzed data from a case-control study conducted in New Hampshire in which patients with bladder cancer and controls sampled from the population completed an interview regarding history of diabetes along with cigarette smoking history, height, weight and history of urinary tract infections.

Results

The study consisted of 331 cases and 263 controls on whom information regarding diabetes was ascertained. History of diabetes was related to an increased bladder cancer risk (adjusted odds ratio = 2.1, 95% CI: 1.1 to 4.1). The association was strongest in those who had diabetes for the longest duration (OR for 16 or more years = 3.6, 1.1 to 11.2) and in those taking oral hypoglycemic medications (OR=3.3, 1.5 to 7.1).

Conclusions

Our findings support an association between bladder cancer incidence and diabetes and further suggest that the risk may be greater among patients taking oral hypoglycemics, and those with diabetes of longer duration.

Introduction

Diabetes is an emerging public health issue in the US, affecting 11% of Americans over the age of 20 [1], and with long-term complications that include cardiovascular disease, retinopathy, neuropathy and nephropathy. Additionally, diabetes has been implicated as a risk factor for certain malignancies, including cancers of the liver, colon and pancreas [2–5].

Bladder cancer is the sixth most common cancer in the United States [6]. Several potential risk factors have been identified for this cancer, including tobacco use, occupational exposure to aromatic amines and polycyclic aromatic hydrocarbons, drinking water contaminants such as chlorinated byproducts and arsenic, and treatment with cyclophosphamide [7]. A meta-analysis of epidemiologic studies suggested an increased risk of bladder cancer in relation to diabetes [8]. The meta-analytic odds ratio was 1.4 (95%CI: 1.0 – 1.8) based on 7 case-control studies, 1.4 (1.2 –1.7) based on 3 cohort studies and 1.0 (0.9 – 1.1) based on 6 studies comparing cohorts of diabetics with external populations. In the meta-analysis, only 3 of the 16 studies examined the differential impact of diabetes therapies and only 2 considered duration of diabetes. The greatest potential confounding factor, smoking, was considered in only half of these studies, while body mass index (BMI) was only used in 3 of the 16. A history of urinary tract infections (UTI) is also a potential confounder [9–11]; however, none of the 16 studies in the meta-analysis adjusted for UTIs.

To further investigate the potential role of diabetes in risk of bladder cancer, we analyzed data from a case-control study conducted in New Hampshire in which patients with bladder cancer and controls completed an interview regarding history of diabetes, UTIs, cigarette smoking, as well as current height and weight (from which to compute BMI).

Methods

A detailed description of the study design appears in earlier reports [12]. Briefly, newly diagnosed, histologically confirmed bladder cancer cases were identified by the New Hampshire State Department of Health and Human Services' Cancer Registry and included New Hampshire residents, aged 25–74 years, first diagnosed between July 1, 1998 and December 31, 2001. The data quality and completeness of this registry meet the standards of the North American Association of Central Cancer Registries [13] [14]. During the study period, the Registry was estimated to capture more than 90% of incident cancers in New Hampshire. Of the 472 potentially eligible cases that were contacted, a total of 398 (84%) were interviewed. For efficiency, we shared a control group with a study of non-melanoma skin cancer in New Hampshire covering a diagnostic period of July 1, 1997 to March 30, 2000. Controls less than 65 years of age were selected using population lists obtained from the New Hampshire Department of Transportation. Controls 65 years of age and older were chosen from data files provided by the Centers for Medicare & Medicaid Services (CMS) of New Hampshire. Cases and controls were eligible if they were English speaking New Hampshire residents ages 25–74, with a working telephone number. A total of 526 controls (76%) were interviewed from a potential 694 confirmed as eligible participants. Participating controls did not appreciably differ by age or sex from those who declined to participate (data not shown).

Consenting participants underwent a personal interview covering socio-demographic factors (i.e., level of education), lifestyle characteristics (i.e., use of tobacco products), and medical history prior to the reference date (diagnosis date of cases and comparable date for controls). Cases and controls were coded as having a history of UTIs if they answered yes to either of the questions, “Did you ever have a bladder infection?” and “Did you ever have a kidney infection?”.

As part of their medical history, participants were asked: “Have you ever been told by a physician that you have diabetes” Those who responded “yes” were asked about age at diagnosis and type of treatment, e.g., use of insulin and oral hypoglycemic medications prior to the reference date. Duration of diabetes was determined as the difference in reference age and age at diagnosis. Questions regarding diabetes were phased into the study; consequently, of the 398 cases and 526 controls who were interviewed, 263 cases and 331 controls had the diabetes questions administered.

We used logistic regression to calculate the odds ratios for bladder cancer risk in relation to history of diabetes. Bladder cancer was regressed on diabetes with and without adjustment for age, gender, smoking status, BMI and UTI. To account for missing BMI data, we use multiple imputation [15] whereby BMI was randomly imputed using a linear regression of BMI on diabetes, age, gender, smoking and UTI. In addition to history of diabetes (yes/no), we computed odds ratios by duration of diabetes and type of diabetes treatment. We did not collect data on type of diabetes (i.e., type 1 versus type 2) and were unable to stratify on age at onset of diabetes because less than 1% of the cases and of the controls had developed diabetes during childhood (defined as onset prior to age 18 years). Therefore, we examined the association with diabetes, both including and excluding, those with diabetes diagnosed during childhood.

Results

Overall, there were a greater proportion of males, smokers, and those with a history of urinary tract infections among the cases than in the controls (Table 1), but cases and controls were similar with respect to BMI.

Table 1.

Selected characteristics of bladder cancer cases and controls

Characteristic	Controls N=263	Cases N=331	P-value
Sex
Female	41% (107)	25% (82)
Male	59% (156)	75% (249)	<0.0001
Mean Age (SD)	60 (11)	62 (9)	0.07
History of Cigarette Smoking
Never Smoked	37% (97)	16% (54)
Former Smoker	44% (116)	50% (166)
Current Smoker	19% (50)	34% (111)	<0.0001
History of Urinary Tract Infections
Yes	25% (65)	33% (108)
No	75% (196)	67% (219)	0.04
Mean Body Mass
Index (SD) *	27.0 (5.0)	28.0 (4.8)	0.07

^*BMI obtained for 260 controls and 122 of cases.

A history of diabetes was reported in 21% of the bladder cancer cases, and 10% of the controls (Table 2) (adjusted OR = 2.4; 95% CI = 1.4, 3.9). The odds ratio remained the same when the three individuals with childhood-onset diabetes were excluded. The proportion of subjects who had diabetes treated by diet only was similar in cases and controls (OR =1.2; 95% CI = 0.5 to 3.2). In contrast, diabetes treated by oral hypoglycemic drugs or insulin was associated with an increased risk of bladder cancer (OR for oral hypoglycemic drugs = 3.3; 95% CI = 1.7, 6.9 and OR for insulin = 2.2; 95% CI = 0.9, 5.5). The odds of bladder cancer were somewhat higher among those with longer duration of diabetes; compared to those without diabetes, those with diabetes of 16 years or more had an adjusted OR of 3.6 (95% CI=1.2 to 11.1), whereas among those with a duration of 5–15 years the adjusted OR was 2.1 (95% CI=1.0 to 4.5), and among those with a duration of less than 5 years the adjusted OR was 2.4 (95% CI=1.1 to 5.2). Adjustment for age, gender, smoking, BMI (by multiple imputation) and UTI did not substantially alter the odds ratios. Type of diabetes treatment and duration of diabetes are related variables. Of those who reported having diabetes for 16 or more years, 62% took insulin, compared to 32% of those with a duration of 5–15 years and 3% of those with a duration less than 5 years. After adjustment for duration, use of oral hypoglycemics remained significantly associated with risk of bladder cancer (adjusted OR=7.5, 95% CI 1.6 – 36.1) and the adjusted odds for use of insulin did not change (data not shown). We considered more detailed smoking variables than the ones reported in Table 1 (number of years smoked, as well as average cigarettes smoked per day) and for each the adjusted odds ratio was within 10% of the estimates reported in Table 2. When we restricted our analysis to cases and controls for which BMI was available, there were no appreciable differences in the crude and adjusted odds ratios for diabetes and diabetes duration. In subgroup analyses, we did not observe appreciably differences in the ORs for diabetes by age or sex (data not shown). The magnitude of the association with diabetes was stronger for non-invasive cancer, adjusted OR=2.8 (1.6 – 4.9) than invasive cancer, adjusted OR=1.5(0.7 – 3.2); however, this difference could have been due to chance.

Table 2.

Odds Ratios and 95% Confidence Intervals for Bladder Cancer and Diabetes

Diabetes	Controls N=263	Cases N=331	Crude Odds Ratio	Adjusted Odds Ratio*	Adjusted Odds Ratio**
Variable	N (%)	N (%)	(95%CI)	(95%CI)	(95%CI)
History of Diabetes
No	90% (238)	79% (255)	(reference)	(reference)	(reference)
Yes	10% (25)	21% (66)	2.4 (1.4, 3.9)	2.5 (1.4, 4.2)	2.2 (1.3, 3.8)
Duration of Diabetes (yrs) ***
1 – 4	4% (10)	8% (27)	2.4 (1.1, 5.2)	2.4 (1.1, 5.4)	2.1 (0.9, 4.8)
5 – 15	4% (10)	7% (23)	2.1 (1.0, 4.5)	2.2 (1.0, 4.9)	2.0 (0.9, 4.6)
16+	2% (4)	5% (16)	3.6 (1.2, 11.1)	3.8 (1.2, 12.2)	3.6 (1.1, 11.2)
			P(Trend)=0.0008	P(Trend)=0.0009	P(Trend)=0.003
Therapy ***
Diet Only	3% (8)	3% (11)	1.2 (0.5, 3.2)	1.2 (0.4, 3.2)	1.0 (0.4, 2.8)
Oral Only	4% (10)	11% (37)	3.3 (1.7, 6.9)	3.6 (1.7, 7.9)	3.3 (1.5, 7.1)
Insulin	3% (7)	5% (17)	2.2 (0.9, 5.5)	2.2 (0.9, 5.7)	2.2 (0.8, 5.6)

^*Adjusted for age, gender, and smoking

^**Adjusted for age, gender, smoking, BMI and UTI

^***Referent group for odds ratios is subjects without diabetes

Discussion

Our findings suggest an association between history of diabetes and bladder cancer risk in New England. We detected an increased risk for individuals with diabetes using either oral or insulin therapy, but not for individuals with diabetes using diet therapy alone. Further, risk was highest among those with the longest duration of diabetes.

The overall odds ratio of 2.1 (95% CI: 1.1 to 4.1) in our study is higher than, although consistent with, the findings of a recent meta-analysis [8]. Aside from methodologic or population differences, heterogeneity of results could be due to differences in prevalence or duration of diabetes treatment, both of which were associated with risk of bladder cancer in our study.

A recent prospective study of 45,000 Swedish men followed for over 9 years, of whom 414 developed bladder cancer, found bladder cancer incidence to be 1.2 times higher in men reporting a history of diabetes, but this was not statistically significant [16]. This and another prospective study of over 100,000 Japanese adults with up to 10 years of follow-up were not included in the meta-analyses. In the Japanese cohort, in which 105 developed bladder cancer, a history of diabetes was associated with a 60% increase in bladder cancer among men and 40% decrease among women, Again, these findings were not statistically significant [5]. Neither the meta-analysis, nor the Swedish or the Japanese studies, stratified by diabetes treatment or adjusted for urinary tract infections.

While a unique aspect of our study is that we adjusted for UTI due to evidence of an association with bladder cancer and diabetes [10–11, 17], this adjustment did not substantially alter the odds ratio. Indeed, adjusting for all covariates did not alter the odds ratio by more than 10%. Because obesity has been associated with bladder cancer risk [8, 18], we attempted to adjust for BMI, which was available for a subset of our cases and controls, unlike the majority of the 16 studies reported in the meta-analysis. In our study, BMI was elevated in cases compared to controls but not significantly, and adjustment for BMI did not alter the odds ratio for bladder cancer associated with diabetes.

Insulin has been hypothesized to be a cancer growth promoter, which could explain an increased cancer risk in adults with type 2 diabetes [2, 19]. Type 2 diabetes is characterized by an extended period of elevated endogenous insulin levels, which compensates for insulin resistance. Fasting blood glucose (FBG) and post challenge glucose levels also have been investigated in relation to bladder cancer risk, but with mixed results. Risk of bladder cancer was 1.6 times higher for those with FBG above 126 mg/dL compared to those with less than 90 mg/dL in a prospective study from Korea [20], although this was not statistically significant. In contrast, a similarly designed cohort study of over 140,000 Austrians found no dose-response association of post-challenge glucose load with bladder cancer, or all cancers combined [21]. In one study, impaired glucose tolerance was found to not be associated with bladder cancer [4]. Thus, further mechanistic data are needed.

In our study, subjects who reported having diet-controlled diabetes had no apparent increased risk compared to subjects without diabetes. The reason for this is not known. Aside from simply serving as a better measure of diabetes itself (i.e., diminishing misclassification), it raises the possibility that the association between bladder cancer and diabetes may be due to treatment for diabetes. In our study, subjects who reported taking oral hypoglycemics had the highest risk, followed by those who reported taking insulin. Most of those classified as taking insulin likely received oral hypoglycemics before they began insulin therapy, as this is the usual course in type 2 diabetes (the predominant form of diabetes). There is some evidence that a common anti-diabetic drug, metformin (a biguanide), may reduce cancer risk, whereas sulfonylureas may increase risk [2]. While there has been some interest in thiazolidinediones as chemopreventive agents, a recent study in rats found that rosiglitazone increased the incidence of bladder cancer [22]. A cross-sectional study of individuals with diabetes found that use of rosiglitazone was associated with self-reported history of cancer [23]. Additionally, the increased cancer risk in diabetics taking insulin in our study could be due to the possible cancer growth promoting effect of insulin [2]. On the other hand, treatment could be a marker for severity of disease. Clearly, further data are needed to understand this association more fully.

Greater than 90% of individuals with diabetes have type 2, and this form of diabetes is believed to have multiple etiologies, including obesity [24]. Type 1 diabetes is an autoimmune disease with an unknown trigger, characterized by a younger age at onset and insulin-dependence. Individuals with type 2 diabetes may also take insulin. Of the 95 individuals with diabetes in our study, only 3 reported being diagnosed during childhood, but it is likely that some of those diagnosed as adults have type 1 diabetes.

Thus, a limitation of our study is our inability to discriminate between type 1 and type 2 diabetes. Another limitation is use of self-reported diabetes and the potential for misclassification, although evidence suggests that self-reported diabetes has good agreement with medical records [25–26]. As diabetes is not a known cause of bladder cancer, the possibility of differential recall (i.e., between cases and controls) seems unlikely. We attempted to minimize reporting bias by asking subjects about their type of diabetes treatment. Indeed, stronger associations were observed among those treated with oral agents and, to some extent, insulin. As we did not have information on glucose levels or other markers of glucose control, we were unable to assess whether the increased risk of bladder cancer found among those receiving oral anti-diabetic drugs or insulin was due to the medication, or if the medication was a marker for more severe diabetes. Additionally, we did not have detailed drug information, which precluded our ability to examine dose or specific type of therapy (e.g., for oral agents). Finally, we adjusted for BMI using multiple imputation because BMI was measured in only about one third of the cases due to the absence of BMI in the first wave of data collection. The presence of missing data and the use of multiple imputation does not cause bias because a missing BMI was not informative [15]. Moreover, BMI did not appear to be a strong bladder cancer risk factor in our data.

In conclusion, we found that risk of bladder cancer was elevated among those with a history of diabetes patients, in particular, those taking oral hypoglycemics or insulin. Additionally, the association increased with duration of diabetes. More research is needed to determine the differential effects of diabetes medications and diabetes severity on the pathogenesis of bladder cancers.