Findings From the Danish Sex Hormone Register Study
The study utilized data from a Danish cohort of all women living in Demark aged 15 to 79, excluding those with cancer (except non-melanoma skin cancer), venous thromboembolism, and women who had received infertility treatment. Registry data provided information of filled prescriptions for hormonal contraception, cancer diagnoses, and confounders.
The data were adjusted for time-dependent information and level of education, parity, polycystic ovarian syndrome, endometriosis, and family history of breast or ovarian cancer. The mean follow-up of the 1,797,932 women included in the study was 10.9 years (19.6 million person-years).
The overall increased relative risk of breast cancer was 20% among current or recent users of any hormonal contraceptive compared with women who never used hormonal contraception (relative risk [RR], 1.20; 95% CI, 1.14-1.26). The risk increased with longer duration of use. Use of any hormonal contraception, for example, for less than 1 year was associated with an RR of 1.09 (95% CI, 0.96-1.23), whereas use for 1 to 5 years was 1.18 (95% CI, 1.10-1.27), 5 to 10 years was 1.24 (95% CI, 1.15-1.34), and longer than 10 years was 1.38 (95% CI, 1.26-1.51).
The increased risk of breast cancer was present for different types of hormonal contraceptives, including COCs (which include both estrogen and progestin), POPs, and intrauterine devices. The RR for COCs ranged from 1.05 to 1.62, depending on the progestin included in the formulation, and ranged from 1.00 to 1.93 for POPs. The RR of breast cancer with the levonorgestrel-releasing intrauterine system was 1.21 (95% CI, 1.11-1.33).
Though there was no significant association for the patch and the implant, these were “statistically uncertain due to the low number of women exposed to these formulations of administration. So, we need further data before we come out with messages about that,” Dr Lidegaard said.
Though this study supports findings from many other epidemiologic reports, including the analysis of the Nurses’ Health Study and the Collaborative Group on Hormonal Factors in Breast Cancer, there are limitations.1,3 The authors note, for example, that they were not able to adjust the data for age at menarche, breastfeeding, alcohol consumption, or physical activity. David A. Grimes, MD, of the University of North Carolina at Chapel Hill, noted that “more than 20 risk factors for breast cancer are known; you can see how many of these potential confounders could be examined and controlled for in the report.”
Dr Grimes told Cancer Therapy Advisor that he found the study “uninformative” because the “administrated database was not designed for research, and has a discriminatory limit (relative risk) of about 2.0 or its reciprocal 0.5. Thus, a study of this type cannot discern a relative risk of 1.2; this finding is likely due to noise.” Bias and residual confounding can result in small but significant RR or odds ratios that are likely to be false positives.4
A strength of administrative databases is, however, their large size and, in the case of the Danish registry, the ability to link registers with various data, such as prescription and diagnostic databases.5 Yet some weaknesses may include poorly recorded confounders, such as comorbidities, or misclassifications. The large size can also lead to spurious statistical associations.
Dr Lidegaard disagreed with Dr Grimes, stating that “in these registries, we have access to information about more confounders than any other data sources, and therefore are able to control for more confounders than almost all previous studies. We achieved very precise estimates due to the large number of included women.” Another strength of the Danish study was that a large number of current hormonal contraceptive users were included and the women used modern contraceptive formulations compared with previous studies.3