Association between diabetes and subsequent malignancy risk among older breast cancer survivors

Abstract

Type II diabetes is associated with cancer risk in the general population but has not been well studied as a risk factor for subsequent malignancies among cancer survivors. We investigated the association between diabetes and subsequent cancer risk among older (66-84 years), 1-year breast cancer survivors within the linked Surveillance Epidemiology and End Results (SEER)-Medicare database using Cox regression analyses to quantify hazard ratios (HR) and corresponding 95% confidence intervals (95% CI). Among 133 324 women, 29.3% were diagnosed with diabetes before or concurrent with their breast cancer diagnosis, and 10 452 women developed subsequent malignancies over a median follow-up of 4.3 years. Diabetes was statistically significantly associated with liver (HR = 2.35, 95% CI = 1.48 to 3.74), brain (HR = 1.94, 95% CI = 1.26 to 2.96), and thyroid cancer risks (HR = 1.38, 95% CI = 1.01 to 1.89). Future studies are needed to better understand the spectrum of subsequent cancers associated with diabetes and the role of diabetes medications in modifying subsequent cancer risk, alone or in combination with cancer treatments.

Type II diabetes is the most diagnosed metabolic disease, occurring in less than 10% of the US population (1). Among cancer survivors, the prevalence increases to approximately 30% (2,3). Diabetes has been associated with increased risks for esophageal (3), pancreatic (4-6), liver (7-9), breast (10,11), stomach (12), colorectal (9,13), kidney (14), urinary tract (15), endometrial (16,17), cervical (18,19), and thyroid (20,21) cancers in the general population. Shared risk factors (eg, increased body weight, inactivity, advancing age) (22) and independent pathways (eg, links between insulin growth factor and estradiol levels) are hypothesized to explain these observed associations (23,24). The role of diabetes in the etiology of subsequent malignancies is not well characterized, having been evaluated in only a few studies focusing on contralateral breast cancer risk in breast cancer survivors and second cancer risk in colorectal cancer (CRC) survivors (25,26). Understanding the relationship between diabetes and subsequent malignancies can inform treatment and cancer screening strategies among the growing population of cancer survivors. Herein, we investigated the association between history of diabetes and subsequent primary cancers among one-year survivors of a first primary breast cancer, a population accounting for approximately 20% of cancer survivors in the United States (27).

We conducted a retrospective cohort study of subsequent solid malignancies in breast cancer survivors within the linked Surveillance, Epidemiology, and End Results (SEER)-Medicare dataset (28). Our study included female breast cancer survivors who were diagnosed with an invasive, nonmetastatic first primary breast cancer between 2000 and 2017 at age 66-84 years, who survived at least 12 months without developing a subsequent cancer, and who had at least 12 months of continuous Part A and B and non-HMO (Health Maintenance Organization) Medicare coverage before and after breast cancer diagnosis to optimize ascertainment of diabetes status at breast cancer diagnosis and breast cancer-directed therapy (Supplementary Material, Supplementary Figure 1, available online). Follow-up began 1 year after initial breast cancer diagnosis and continued until the earliest of age 85, loss to follow-up, death, or end of study (December 31, 2018). For analyses of all solid subsequent malignancies, women were followed to their first solid subsequent malignancy, whereas for analyses of specific subsequent cancer types, follow-up continued past intervening malignancies of other types. For analyses of subsequent breast cancers only, we excluded women with a bilateral mastectomy at the time of first breast cancer diagnosis. Medicare claims were used to identify patients who had been diagnosed with diabetes before or within 30 days of their breast cancer diagnosis (ie, history of diabetes), based on either 1 inpatient or 2 outpatient diabetes (“Type I,” “Type II,” or “Other”) claim(s) 30 days or more apart. Hazard ratios (HRs) and associated 95% confidence intervals (95% CIs) were estimated from Cox regression analyses adjusted for age, race and ethnicity, year of diagnosis, stage, hormone receptor status, and Charlson comorbidity score (modified to exclude diabetes), as classified in Table 1. We also adjusted for time-dependent radiotherapy, chemotherapy, and intervening subsequent malignancy diagnosis. Subsequent malignancy sites were selected based on association with diabetes in previously published literature in the general population (3-17,20,21) or sites with 50 or fewer patients in the study population. Two-sided P values (P < .05) were considered statistically significant.

The study population included 133 324 women; 39 012 (29.3%) had a known history of diabetes at the time of initial breast cancer diagnosis (Table 1). Median follow-up from 1 year after initial breast cancer diagnosis until exit was 4.3 years (range 0.003–18.0 years), and the median age at first primary breast cancer diagnosis was 73.6 years. Breast cancer survivors with a history of diabetes were more likely to be diagnosed with breast cancer at an older age, higher stage, and in more recent calendar years compared with women without a history of diabetes (Table 1). We also observed differences in the distribution of race and ethnicity by diabetes status. Specifically, there was a higher proportion of non-Hispanic Asian, non-Hispanic Black, and Hispanic (all races) breast cancer survivors with a history of diabetes compared to those without diabetes. Additionally, women with diabetes at breast cancer diagnosis were more likely to have other comorbidities (comorbidity score above 0) and less likely to have received radiotherapy within the first year after breast cancer diagnosis.

There was no association between history of diabetes and risk of all solid subsequent cancers combined (N patients with subsequent cancer_{diabetes/total}=2646/10452, 25.3%; HR = 1.03, 95% CI = 0.98 to 1.08, P = .189) or for subsequent breast cancers (N = 761/3142, 24.2%; HR = 1.03, 95% CI = 0.95 to 1.12, P = .481; Figure 1). For site-specific cancers, a history of diabetes was associated with statistically significant increased risks of subsequent liver (N = 44/92, 47.8%; HR = 2.35, 95% CI = 1.48 to 3.74, P < .001), brain (N = 33/97, 34.0%; HR = 1.94, 95% CI = 1.26 to 2.96, P = .002), and thyroid cancers (N = 62/203, 30.4%; HR = 1.38, 95% CI = 1.01 to 1.89, P = .046). Conversely, a history of diabetes was associated with a lower risk of subsequent lung cancer (N = 500/2089, 23.4%; HR = 0.89, 95% CI = 0.80 to 0.99, P = .032). HRs for additional cancers that have been previously reported to be associated with diabetes in the literature, including esophagus, stomach, kidney, pancreatic, uterine, and colorectal cancers, were greater than 1.00 but not statistically significant. Within the subset of women with diabetes for whom Medicare Part D data were available, metformin was associated with subsequent CRC risk (HR = 0.61, 95% CI = 0.39 to 0.97, P = .037; Supplementary Figure 2, available online), whereas no association was observed for other sites investigated.

The elevated HR for liver cancer was consistent with previous reports in the general population (7,8), whereas the lack of statistically significant associations for colorectal, kidney, pancreas, and uterine cancer differed from prior findings. Improvements in diabetes management with increased medical surveillance of cancer survivors could potentially explain the lack of association for these cancers in our study population. Specifically, the diabetes drug metformin has been shown to reduce cancer risk in the general population (29). Although based on a small subset of our population for whom Medicare Part D information was available, metformin was associated with reduced CRC risk. It is also possible that by considering diabetes status at or before initial breast cancer diagnosis and not accounting for diabetes diagnosed after breast cancer diagnosis, misclassification of diabetes status during follow-up biased our results toward the null. Additionally, we did not have data regarding hysterectomies; thus, inclusion of women who underwent hysterectomy could have attenuated the HR for uterine cancer.

For both brain and liver cancers, we considered the possibility that misclassified metastases could explain our results given that these are frequent sites for metastatic breast cancer (30,31). However, this seems unlikely as misclassification would need to differ by diabetes status. Additionally, reported histologic types were most commonly glioblastoma (70/97 brain cancer patients) and hepatocellular carcinoma (72/92 liver cancer patients), which seem unlikely to be misclassified metastatic disease. Finally, these malignancies occurred primarily in women diagnosed with stage I and II breast cancer; only 8.2% of subsequent brain and 10.8% of subsequent liver cancer cases occurred in individuals diagnosed with stage III or unknown stage at breast cancer diagnosis; therefore, metastasis seems unlikely.

Our observation of lower lung cancer risk was unexpected based on previously reported null or increased risk of lung cancer in patients with diabetes (9,32-36) and may reflect confounding by smoking status or other unmeasured confounders. Notably, a previous study of smoking prevalence among cancer survivors found that older cancer survivors with more comorbidities had more frequent interactions with the healthcare system and additional opportunities to access smoking cessation programs than those with fewer comorbidities, possibly contributing to a lower prevalence of smoking (37). Unfortunately, we did not have reliable information about smoking history and cessation status or other individual modifiable risk factors (https://healthcaredelivery.cancer.gov/seermedicare/considerations/measures.html) in the linked SEER-Medicare dataset. For example, we could not evaluate the potential influence of body mass index (BMI). In SEER-Medicare, obesity data are highly incomplete, and BMI is not available.

Although the linked SEER-Medicare dataset is a rich data source, there are some additional limitations to this analysis. The SEER-Medicare population is 65 and older; thus, the results may not be generalizable to a younger population of breast cancer survivors. We also lacked information on time since first diabetes diagnosis, given that claims data do not provide exact dates of diagnosis and Medicare does not capture a patient’s full medical history (38). We adjusted analyses for radiotherapy and chemotherapy but not for other systemic therapy (immunotherapy, hormone therapy) because these are not available from SEER in the linked dataset and many oral formulations are not covered by Medicare Parts A and B. Lastly, although Type II diabetes accounts for most of the diabetes in our study population, differentiating between Type I and Type II was difficult because Medicare claims often included both types or nonspecific coding. Distinguishing the two types may further elucidate relationships with subsequent cancer risk in cancer survivors.

In conclusion, this large population-based study suggests that breast cancer survivors with diabetes may be at higher risk for developing several types of subsequent cancers compared with survivors with no history of diabetes. Future studies are needed to better understand the spectrum of subsequent cancers associated with a history of diabetes and the role of diabetes medications in modifying subsequent cancer risk, alone or in combination with cancer treatments.

Data availability

The data underling this article are available through application to the National Cancer Institute Division of Cancer Control and Population Sciences (https://healthcaredelivery.cancer.gov/seermedicare/).

Author contributions

Kaitlyn Noel Lewis Hardell, PhD, MPH (Conceptualization; Formal analysis; Investigation; Methodology; Validation; Visualization; Writing—original draft; Writing—review & editing), Sara J. Schonfeld, PhD (Conceptualization; Formal analysis; Methodology; Validation; Writing—original draft; Writing—review & editing), Cody Ramin, PhD (Conceptualization; Methodology; Writing—review & editing), Jacqueline B. Vo, PhD, RN, MPH (Conceptualization; Methodology; Writing—review & editing), Lindsay Morton, PhD (Conceptualization; Funding acquisition; Methodology; Project administration; Resources; Supervision; Writing—original draft; Writing—review & editing).

Funding

This study was supported by the Intramural Program of the National Cancer Institute (NCI), National Institutes of Health. The funder did not play a role in the design of the study; the collection, analysis, and interpretation of the data; the writing of the manuscript; or the decision to submit the manuscript for publication.

The collection of cancer incidence data used in this study was supported by the California Department of Public Health as part of the statewide cancer reporting program mandated by California Health and Safety Code Section 103885; the National Cancer Institute’s Surveillance, Epidemiology and End Results Program under contract HHSN261201000140C awarded to the Cancer Prevention Institute of California, contract HHSN261201000035C awarded to the University of Southern California, and contract HHSN261201000034C awarded to the Public Health Institute; and the Centers for Disease Control and Prevention’s National Program of Cancer Registries, under agreement U58DP003862-01 awarded to the California Department of Public Health. The ideas and opinions expressed herein are those of the author(s), and endorsement by the State of California Department of Public Health, the National Cancer Institute, and the Centers for Disease Control and Prevention or their contractors and subcontractors is not intended nor should be inferred.

Conflicts of interest

None to declare.

Acknowledgments

This study used the linked SEER-Medicare database. The interpretation and reporting of these data are the sole responsibility of the authors. The authors acknowledge the efforts of the National Cancer Institute; the Office of Research, Development, and Information, CMS; Information Management Services (IMS), Inc; and the Surveillance, Epidemiology, and End Results (SEER) program tumor registries in the creation of the SEER-Medicare database.

References