Risk of Breast Cancer After Onset of Type 2 Diabetes: Evidence of detection bias in postmenopausal women

The dataset used for this study has been described in detail previously (4). In brief, this was a population-based retrospective cohort study, using the British Columbia Linked Health Databases (BCLHD) from British Columbia, Canada.

Women who were likely to have a clinical diagnosis of diabetes were identified based on the established case definition for the Canadian National Diabetes Surveillance System. The index date for diabetes was defined as the first of 1) a hospital admission for diabetes (ICD-9 code 250) or 2) the second of two medical fee-for-service claims coded ICD-9 250 within a 2-year period. Physician visits and hospital admissions were classified according to ICD-9. Women with claims for gestational diabetes (ICD-9 648.8) were excluded.

From 1 April 1996 to 31 March 2006, we identified women with incident T2D as those 30 years or older and meeting the definition of diabetes after 2 consecutive years of not meeting the criteria (N = 84,506). Nondiabetic women registered for health coverage on the diabetes index date were then matched, with replacement, on age and index year. After a minimum 2-year cancer washout period prior to index date, first breast cancers were identified prospectively in both cohorts, via linkage to the British Columbia Cancer Agency database. Censoring occurred at death, departure from British Columbia, or 31 March 2006, whichever was earliest.

Incident rates and 95% CIs were determined for breast cancer using Poisson regression. Cox regression was then used to estimate the hazard ratios (HRs) and 95% CIs for the association between incident diabetes and breast cancer. We stratified this analysis, using age as a proxy, for pre- (<55 years) and postmenopausal (≥55 years) status. We split the follow-up time into earlier (0–3 months) and later (3 months to 10 years) intervals, after diabetes index date, to account for time-varying HRs. SAS version 9.2 (SAS Institute, Inc.,Cary, NC) was used for analyses and R version 2.8 (Wien, Austria) for plotting.

There were 84,506 women in each of the incident diabetes and matched nondiabetes cohorts, and the mean (SD) age was 61.8 (14.2) years. The mean cancer-free period before index was 6.9 (2.9) and 7.1 (2.9) years, and the mean (SD) follow-up was 4.4 (2.9) and 4.5 (2.9) years for the diabetes and nondiabetes cohorts, respectively.

There were 2,381 incident breast cancer cases identified throughout the full follow-up period. Incidence rates (per 1,000 person-years) for breast cancer were 1.83 (1.61–2.08) and 1.92 (1.68–2.21) for premenopausal women with and without T2D, and 4.02 (3.77–4.29) and 3.89 (3.62–4.19) for postmenopausal women with and without T2D, respectively (Table 1).

During the first 3 months after diabetes index date, we observed a trend toward an increased risk of breast cancer in postmenopausal women with T2D (HR 1.31; 95% CI 0.92–1.86; P = 0.14), compared with women without T2D (Table 2). Relative risk of breast cancer in premenopausal women with T2D during this earlier time period was 0.95 (0.48–1.86; P = 0.88). In the later time window of 3 months to 10 years after diabetes index, the risks of breast cancer were 0.92 (0.75–1.13; P = 0.45) and 1.00 (0.90–1.11; P = 0.93) in pre- and postmenopausal women with T2D, respectively, compared with women without T2D (Table 2).

Our observation of a trend toward an elevated risk of breast cancer among postmenopausal women with T2D is consistent with previously published studies (1). However, there is little evidence in the literature examining detection bias as a possible explanation for the observed association between T2D and breast cancer (2). Therefore, previous epidemiologic associations of breast cancer risk in women with diabetes may have been overestimated, if they did not take into consideration this initial detection bias at the time of diabetes diagnosis.

Strengths of this study include the use of a very large population-based dataset, linkage with the BC Cancer Registry, matching our cohorts to balance potential confounders, and examining risk of breast cancer during different time windows after diabetes index date to address potential detection bias. The main limitation to our research surrounds the potentially tenuous index date assigned to diabetes onset. Although it is possible for a clinical diagnosis of diabetes to have preceded the case date in the BCLHD, it is less likely that a cancer diagnosis would have actually preceded the clinical diagnosis of diabetes, particularly given the minimum 2-year cancer-free period we imposed prior to index date, which coincided with the same time window for the diabetes case definition. In previous validation studies of the diabetes case definition, ∼85–90% of case subjects were identified through physician services claims. Furthermore, the median duration between qualifying physician visits is only 39 days (interquartile range 12–150 days; unpublished data).

In summary, we observed a trend toward an increased risk of breast cancer, but only in postmenopausal women with T2D. As well, these findings were restricted to the time period immediately after diabetes index date, reflecting a potential detection bias. Perhaps screening initiatives for breast cancer in women with T2D should focus on postmenopausal women in this earlier time period after a diagnosis of diabetes. Additional research is required to further explore this potential for detection bias in different patient populations.

J.A.J. is a Senior Scholar with Alberta Innovates—Health Solutions and holds a Canada Research Chair in Diabetes Health Outcomes. C.A.M. holds a Canada Research Chair in Pharmaceutical Outcomes. This work was supported in part by an operating grant from the Canadian Institutes of Health Research (CIHR) (Reference MOP-82737) and a CIHR team grant to the Alliance for Canadian Health Outcomes Research in Diabetes (Reference OTG-88588), sponsored by the CIHR Institute of Nutrition, Metabolism, and Diabetes. The funding source played no role in the study design or the collection, analysis, or interpretation of data.

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

S.L.B. conceived of the study, planned the data analysis, prepared a first draft, and edited subsequent versions of the manuscript. K.R. conducted the data analysis and contributed to the writing of the manuscript. C.A.M. obtained the data from the BCLHD and British Columbia Cancer Agency database and edited subsequent versions of the manuscript. J.A.J. obtained funding, planned the data analysis, and edited subsequent versions of the manuscript.

Parts of this study were presented in abstract form at the annual European Association for the Study of Diabetes meeting, Lisbon, Portugal, 12–16 September 2011, and at the Canadian Diabetes Association meeting, Toronto, Ontario, Canada, 26–29 October 2011.