Similar Breast Cancer Risk in Women Older Than 65 Years Initiating Glargine, Detemir, and NPH Insulins

Long-acting insulin analogs, insulin glargine (glargine) and insulin detemir (detemir), are structurally altered human insulins designed to overcome the limitations of neutral protamine Hagedorn (NPH) insulin, namely, its short half-life and risk of nocturnal hypoglycemia. However, altering human insulin may influence mitogenicity, and so concern was raised about the carcinogenic potential of glargine. Indeed, in vitro studies showed that glargine had more potent mitogenic properties (up to eightfold) than regular insulin (1) and a substantially higher proliferative effect on breast cancer cells compared with other insulins (2). However, further studies suggested that in vivo metabolites of glargine, M1 and M2, were not mitogenic and are most measurable after insulin glargine administration with very little to none of the intact molecule (M0) detectable (3–5).

Due to concerns about carcinogenicity, several European observational studies examined the possibility of an association between glargine and cancer, although results were inconsistent (6–9). Since then, a study conducted at Kaiser Permanente Northern California (KPNC) reported an increased risk of breast cancer among glargine users who used the drug for ≥2 years compared with NPH (hazard ratio [HR] 1.6; 95% CI 1.1–2.4) (10). Another study, conducted using U.S. prescription claims data, reported no increased risk of breast, prostate, or colon cancer among patients initiating glargine compared with NPH (11).

A recent study in the Clinical Practice Research Datalink (CPRD) reported that compared with NPH use, long-term use of glargine (≥5 years) was associated with an increased risk of breast cancer, although this elevated risk was restricted to prior insulin users (HR 1.53; 95% CI 1.10–2.12) and was not seen in new initiators or in detemir users (12). This study represented an update on a previous study in the same database published in 2011, which reported similar findings with a significantly increased risk of breast cancer seen only among long-term users of glargine who were prior insulin users (HR 2.7; 95% CI 1.1–6.5) and not in new initiators (13).

The Outcome Reduction With Initial Glargine Intervention (ORIGIN) trial, a multicenter international trial that examined cardiovascular and cancer outcomes in patients with dysglycemia assigned to glargine versus standard care, found no association between insulin glargine use and breast cancer (HR 1.01; 0.60–1.71) or any other cancer (14).

Given the inconsistent findings and small numbers of long-term glargine and detemir users in previous observational studies, uncertainty remains as to whether longer-term glargine use influences breast cancer risk. Additionally, insulin glargine is widely used in the U.S., accounts for 81.3% of the long-acting analog insulin prescription share (15), and was second from the top in terms of drug expenditures in 2017 (16). Therefore, we used Medicare data to assess whether long-term use of glargine, specifically, compared with use of NPH or detemir, was associated with an increased risk of breast cancer in women with diabetes.

Female Medicare beneficiaries, aged ≥65 years, enrolled in fee-for-service Medicare Part A (hospitalization), Part B (office-based medical care), and Part D (prescription drugs) were eligible for study inclusion if they initiated a study insulin (glargine, detemir, or NPH) between September 2006 and September 2015 and if in the 270 days prior to the date of the qualifying prescription (index date) they had continuous enrollment in Medicare, had no prior cancer diagnosis, and did not receive a prescription for a study insulin. Patients were excluded if they were in a skilled nursing facility or nursing home or if they received the index prescription from a long-term care pharmacy or were receiving hospice care on the index date. Kidney transplant recipients, patients undergoing dialysis, and anyone who entered Medicare for reasons other than age were also excluded (Fig. 1).

The primary exposure definition was ever use of glargine, detemir, or NPH. Duration of insulin use, defined as cumulative days’ supply in years (0 to <3, 3 to <5, and ≥5), and insulin dose, defined as cumulative units of insulin dispensed (0–20,000, 20,000–60,000, and ≥60,000), were also examined, as was length of follow-up time in years (0 to <3, 3 to <5, and ≥5). In addition, dose in combination with duration of follow-up (e.g., 0 to <3 years and 0 to <20,000 units) was investigated.

An intention-to-treat approach to follow-up time was applied in our primary analyses. Follow-up began the day after cohort entry (index date) and continued until one of the following: death, disenrollment from Medicare, end of study (31 May 2017), switching to other study insulin, switching to insulin degludec, or diagnosis of breast cancer.

Breast cancer was identified using a previously validated algorithm developed by Setoguchi et al. (17). Breast cancer case subjects were defined as those who had two or more diagnoses of breast cancer recorded within 2 months. The date of the second breast cancer diagnosis was used as the outcome date. This definition resulted in high specificity (99.62%) and a good PPV (76.56%) in the validation study.

Preexisting medical conditions, medication use, an adapted Diabetes Complications Severity Index (aDCSI) (18), and health care utilization covariates were identified in the 9-month baseline period prior to cohort entry. Logistic regression was used to estimate the probability of receiving glargine versus NPH and glargine versus detemir and to further calculate inverse probability of treatment weights (IPTW). Average treatment effects among treated (ATT) IPTW weights, with glargine considered the reference treatment, were applied. The distributions of propensity scores and ATT IPTW were inspected for outliers. Weight truncation at the 99th percentile was conducted for extreme weights in the detemir and NPH cohorts. Covariate balance between the weighted cohorts was assessed using standardized mean differences, with a value ≤0.1 indicating a negligible difference between groups (19).

All analyses were based on IPTW-adjusted cohorts and therefore accounted for potential confounding by baseline factors. Crude breast cancer incidence rates per 1,000 person-years (PY) and 95% CIs were calculated from standard (unweighted) Cox proportional hazards models. Weighted Cox proportional hazards regression with robust SEs was used to estimate weight-adjusted HRs and 95% CIs for incidence of breast cancer among glargine users compared with users of NPH and detemir. As there were no significant differences in observed confounders across cohorts after the weighting, the weighted Cox model included a treatment indicator as a sole covariate leading to estimation of the marginal treatment effect on the breast cancer risk. Separate weighted Cox proportional hazards models were used to examine ever use, cumulative duration of use, cumulative dose of insulin, length of follow-up time, and a combination of cumulative dose and length of follow-up.

Secondary analyses were restricted to insulin users with ≥5 years of follow-up time. Within each insulin cohort, we compared women with low cumulative dose with those with medium or high cumulative dose for the occurrence of breast cancer.

A prespecified sensitivity analysis repeated the primary analyses using untruncated ATT IPTW weights.

This study was classified as public health surveillance by the U.S. Food and Drug Administration (FDA) and was exempted from review by the FDA’s Research in Human Subjects Committee in accordance with the updated Common Rule. Analyses were performed using SAS, version 9.4 (SAS Institute, Cary, NC) and R 3.4.3 (R Foundation for Statistical Computing, Vienna, Austria).

A total of 203,159 glargine, 67,012 detemir, and 47,388 NPH initiators contributed 691,342, 198,731, and 143,628 PY of follow-up, respectively. The median follow-up time in years for glargine, detemir, and NPH was 2.9 (interquartile range 1.44–4.93), 2.48 (1.22–4.21), and 2.32 (0.93–4.51), respectively. A median of four prescriptions per year was received by women in each of the insulin cohorts.

Before IPTW adjustment, there were minor differences in clinical characteristics between the glargine and detemir cohorts related to diabetes medications, comorbidity score, hospitalizations, and physician specialty. Differences in demographics and clinical characteristics between NPH and glargine cohorts were more frequent and substantial than those between glargine and detemir cohorts and included race, income level, use of statins and diabetes drugs, numbers of physician visits, and physician specialty (Supplementary Table 1). After IPTW adjustment, all cohorts were closely balanced for all covariates (Table 1).

During up to 10.7 years of follow-up, there were 6,267 breast cancers identified: 4,170 (66%) in glargine initiators, 864 (14%) in NPH initiators, and 1,233 (20%) in detemir initiators. The crude breast cancer incidence rates per 1,000 PY were similar across the insulin cohorts, ranging from 6.02 to 6.20 (Tables 2 and 3).

Ever use of glargine was not associated with an increased risk of breast cancer compared with NPH (HR 0.97; 95% CI 0.88–1.06) or detemir (HR 0.98; 95% CI 0.92–1.05). No increased risk was seen with glargine use compared with either NPH or detemir when stratifying by duration of insulin use, by length of follow-up, or by cumulative dose of insulin dispensed (Tables 2 and 3 and Supplementary Fig. 1).

On examination of both cumulative dose and length of follow-up combined, no increased risk of breast cancer was found among glargine users compared with NPH or detemir users, except among those using insulin for 3–5 years and with a cumulative dose of ≥60,000 units. An HR of 1.62 (95% CI 1.02–2.58) was observed for glargine versus NPH, and an HR 1.53 (95% CI 0.99–2.36) was observed for glargine versus detemir. Despite this, among those with the longest follow-up time and highest cumulative doses of glargine combined (≥5 years and ≥60,000 units), there was no difference in breast cancer risk for glargine versus NPH (HR 0.99; 95% CI 0.63–1.57) or for glargine versus detemir (HR 0.89; 95% CI 0.65–1.21) (Table 4, Supplementary Table 2, and Supplementary Fig. 1).

Among patients with ≥5 years of follow-up, there was no difference in breast cancer risk at different cumulative doses of glargine or detemir. For NPH, risk was reduced in patients with intermediate but not high cumulative exposure (Supplementary Table 3). The results of the sensitivity analysis using untruncated weights were consistent with those from the main analysis (data not shown).

In this large new-user, active comparator, cohort study, which included almost 320,000 insulin initiators, no association was seen between ever use of glargine and breast cancer risk, compared with use of NPH or detemir. This null association persisted across levels of dose, duration of use, and length of follow-up. There was also no increased risk of breast cancer observed in those who had ≥5 years of follow-up and who used medium- or high-dose glargine compared with low-dose users.

These results conflict with those of a recent study by Wu et al. (12), which reported that compared with NPH, glargine use was associated with an increased risk of breast cancer (HR 1.44; 95% CI 1.11–1.85), especially after 5 years of use (HR 2.23; 95% CI 1.32–3.77) and with ≥30 prescriptions (HR 2.29; 95% CI 1.26–4.16). However, these differences are likely the result of differences in study design, population size, and potential imbalances in the cohorts studied. First, the primary results of Wu et al. were derived from the entire cohort of insulin initiators, using a prevalent new-user design, which has been proposed as a way of including in a study new initiators of a drug who have also used the older comparator drug and so are not treatment naïve to both study drugs (20). Prevalent new-user designs, though, may be limited by selection bias, immortal time bias, and issues with properly timing covariates in the baseline period. When the analysis by Wu et al. was restricted to new users only, as we did in our study, no increased risks were seen for ever use of glargine compared with NPH (HR 1.18; 95% CI 0.77–1.81) or with ≥5 years of use (HR 1.60; 95% CI 0.66–3.84). The association between glargine and breast cancer seemed to be particularly concentrated among prior insulin users (HR 1.53; 95% CI 1.10–2.12). Second, there were just 4,148 new glargine users in that study compared with our 203,159 glargine initiators, increasing confidence in our estimates. Third, in the study by Wu et al., the breast cancer incidence rate per 1,000 person-years decreased among NPH users as duration of use increased, from 3.6 (95% CI 2.8–4.5) in those with <3 years’ exposure to 2.2 (95% CI 1.4–3.5) for those with ≥5 years’ exposure, while in glargine users it increased slightly. The decreased breast cancer incidence in longer-term users of NPH may have contributed to a potentially spurious increased HR for glargine users. NPH users tended to be older than glargine users (mean age 70 years vs. 64 years, respectively) and so were less likely to get screened for breast cancer according to U.K. National Health Service breast screening guidelines (21), which may explain the lower incidence of breast cancer in these users across time.

A retrospective cohort study conducted in the Inovalon Medical Outcomes Research for Effectiveness and Economics Registry (MORE²) registry by Stürmer et al. (11), which used claims data similar to those used in our study, also found no association between glargine use and breast cancer risk (HR for ever use of glargine compared with NPH was 1.07 [95% CI 0.65–1.75] and for duration ≥2 years was 0.67 [95% CI 0.18–2.54]).

While the findings were similar, our study was larger than the study by Sturmer et al. (11), with 203,159 glargine and 47,388 NPH initiators compared with 43,306 glargine initiators and 9,147 NPH initiators, and had longer median durations of follow-up: 2.9 in the glargine cohort and 2.3 in the NPH cohort compared with 0.9 and 0.8, respectively. We also conducted detailed analyses on dose of glargine used and dose and duration of use combined.

A meta-analysis of seven cohort studies, published in 2012, also found no difference in breast cancer incidence rates in glargine users compared with users of other types of insulin, although considerable heterogeneity (I² = 74%) was seen across the included studies, which may have influenced the summary estimate (22).

Our findings are also similar those of the ORIGIN trial, which reported no association between insulin glargine use and breast cancer compared with standard care. However, that trial had only 56 breast cancer cases (28 in each arm), while our study included 6,267 breast cancers with 4,170 (66%) occurring in glargine initiators, 864 (14%) in NPH initiators, and 1,233 (20%) in detemir initiators.

We did not find any evidence of increased breast cancer risk in glargine users in analyses examining ever use, dose, duration of use, and length of follow-up individually; however, in our analysis that examined length of follow-up and dose simultaneously, an increased risk of breast cancer was seen in glargine users compared with NPH who used high-dose insulin, ≥60,000 units, over a relatively short period of time (3–5 years’ follow-up). Despite this finding, there was no increased risk seen among those using high-dose insulin across either shorter or longer time periods (0–3 years or ≥5 years of follow-up) and so this may be due to chance given the large number of analyses conducted. As a post hoc analysis, at the request of a reviewer, we conducted an adjustment for multiple comparisons in our primary analyses for each of our cohort comparisons (glargine vs. NPH and glargine vs. detemir) using the Benjamini-Hochberg procedure. No statistically significant results were found after the adjustment. Specifically, in the analysis of both cumulative dose and length of follow-up combined among those with 3–5 years of follow-up and ≥60,000 units, the Q value from the adjustment was 0.81, suggesting that our finding of an increased risk of breast cancer in this category was due to chance. A numerically increased risk was also seen in analyses comparing glargine and detemir, among those using high-dose insulin over a relatively short period of time (3–5 years of follow-up and 1–3 years of follow-up), although statistical significance was not achieved and there was no suggestion of increased risk in the period of ≥5 years of follow-up.

Our study has several strengths. First, it is the largest study to date to examine the association between glargine use and breast cancer risk and had almost five times as many glargine users as the next-largest study that examined this question. Up to 10.7 years of follow-up time was available, with ∼20% of all users having >5 years of follow-up. A new-user design was applied, rather than the prevalent new-user design used in some other studies, to avoid missing potential early effects of drugs and to accurately account for baseline confounders. An active comparator was used to reduce unmeasured confounding, and this was the first study to use detemir, also a long-acting analog insulin, in addition to NPH, as an active comparator for glargine. Our study was also the first study to date that examined insulin dose according to units of insulin dispensed rather than counting of numbers of prescriptions and the first that assessed the influence of cumulative dose and follow-up time combined. Unlike some previous studies, that used a multivariate adjustment approach to estimate risks, conditioning on confounders, our study used a propensity score method to estimate marginal risks across cohorts that correspond with the effect measures from randomized trials.

Our study did have some limitations. It was observational and therefore may be subject to confounding by factors not adjusted for in the analysis. The study population was aged ≥65 years, and so the generalizability of findings to those outside that age range may be limited. The Medicare claims data used in this study are ideal for capturing drug exposure but may be limited in identifying cancer diagnoses and potentially important covariates such as BMI and smoking. To address this, we used a validated algorithm (17) that has been applied in previous study of insulin glargine and cancer risk (11) that used claims data. That study, by Stürmer et al. (11), used two external validation studies to assess the potential for unmeasured confounding by BMI and showed that BMI did not affect the decision to initiate insulin treatment with glargine versus NPH. We also used ICD-9, National Drug Code, and Healthcare Common Procedure Coding System to identify, as much as possible, evidence of smoking and smoking status in the claims data. Diabetes has been associated with a slightly increased risk of breast cancer. A recent meta-analysis of 18 studies reported a summary relative risk of 1.13 (95% CI 1.04–1.24) (23). While we did not have information on diabetes duration, we did estimate a modified diabetes severity index and included it in the propensity score model to account for any potential confounding by diabetes severity.

In summary, insulin glargine use was not associated with an increased risk of breast cancer compared with NPH or detemir use, in female Medicare beneficiaries with diabetes, irrespective of dose, duration of use, or length of follow-up.

Funding. This study was funded by the FDA through an interagency agreement with the Centers for Medicare & Medicaid Services.

The views expressed in this manuscript are those of the authors and do not necessarily reflect the views of the FDA.

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

Author Contributions. M.C.B. designed the study, interpreted the results, and wrote and edited the manuscript. H.L., P.B., M.S., J.A.K., and D.J.G. designed the study, interpreted the results, and reviewed and edited the manuscript. Y.C., X.W., S.P., M.W., and T.E.M. designed the study, analyzed the data, and reviewed and edited the manuscript. M.C.B. 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.

Prior Presentation. Parts of this study were presented in abstract form at the 34th International Conference on Pharmacoepidemiology & Therapeutic Risk Management, Prague, Czech Republic, 22–26 August 2018.