Hormone Replacement Therapy and the Risk of Melanoma in Post-menopausal Women

B.M. Hicks; K.B. Kristensen; S.A. Pedersen; L.R. Hölmich; A. Pottegård


Hum Reprod. 2019;34(12):2418-2429. 

In This Article

Materials and Methods

Data Sources

We obtained data from six nationwide registry sources: the Danish Cancer Registry (Gjerstorff, 2011), the Danish Pathology Registry (Bjerregaard and Larsen, 2011), the National Prescription Registry (Wallach Kildemoes et al., 2011), the National Patient Registry (Schmidt et al., 2015a), Registers in Statistics Denmark on educational level (Jensen and Rasmussen, 2011) and the Civil Registration System (Schmidt et al., 2014). A description of these registries is provided in Supplementary Table SI, with codes for diagnoses, drug exposure and covariates in Supplementary Table SII. All linkages were performed by Statistics Denmark.

Virtually, all medical care in Denmark is funded by the Danish National Health Service, allowing true population-based register linkage studies covering all residents of Denmark (Thygesen et al., 2011). Data were linked by a unique personal identification number, assigned to all residents. Linkages were performed by Statistics Denmark.

Investigation of Melanoma Risk: Case Control Study

Selection of Melanoma Cancer Cases and Population Controls. From the Danish Cancer Registry, we identified cases as all women with a primary, histologically verified diagnosis of invasive cutaneous melanoma between 1 January 2000 and 31 December 2015. The date of diagnosis corresponded to the index date. We included only patients between the ages of 45–85 years at the index date and excluded patients with any residency outside of Denmark within 10 years prior to the index date. We further excluded those who had a history of primary ovarian failure, radical hysterectomy or bilateral salpingo-ophorectomy/oophorectomy, those patients with a previous history of cancer (excluding non-melanoma skin cancers) and those with xeroderma pigmentosum. Finally, we excluded those with a history of organ transplantation, HIV diagnosis or use of azathioprine, cyclosporine or mycophenolate mofetil, as immunosuppression has been associated with an increased risk of skin cancer (Dahlke et al., 2014; Olsen et al., 2014b; Fattouh et al., 2017).

Controls were selected using risk set sampling. For each case, we selected 20 controls amongst Danish women matched by age and calendar time, applying the same selection criteria as for cases. Controls were assigned the same index date as the case to whom they were matched. Subjects were eligible for sampling as controls before they became cases. Thereby, the calculated ORs provide direct estimates of the incidence rate ratios from a cohort study utilising the source population (Rothman and Lash, 2008).

Exposure Definition: Systemic HRT. Based on prescriptions dispensed since 1995, ever-use of HRT was defined as having filled at least one prescription for HRT prior to the index date. HRT included all systemic agents available in Denmark during the study period, including oestrogen only, progestogen only and oestrogen and progestogen combination therapies. Hormonal intrauterine devices were not included. Intravaginal oestrogens were also not included in our HRT exposure definition as the primary indications for intravaginal oestrogens are local complaints including vaginal atrophy. Furthermore, doses administered with intravaginal therapy are markedly lower than systemic oestrogen and have minimal systemic absorption, with previous studies finding use of low-dose intravaginal oestrogens does not result in sustained serum oestrogen levels exceeding the normal menopausal range (Rigg et al., 1978; Simunić et al., 2003; Santen, 2015). However, intravaginal oestrogens were investigated in sensitivity analyses. High levels of HRT use were defined as filled prescriptions equivalent to ≥1000 defined daily doses (DDDs) of HRT corresponding to ~3 years of cumulative use. This corresponded to a cumulative use of, for example, 200 mg of estriol or 5000 mg of norethisterone (World Health Organisation, 2019). For all analyses, prescriptions filled in the year prior to the index period were disregarded. This 1-year lag period was introduced to allow for a minimum latency time window and to minimise reverse causality (Rothman and Lash, 2008). The length of the lag period was varied in sensitivity analyses.

Potential Confounders. We defined potential confounders as the following: (i) drugs suggested to have photosensitising properties including oral retinoids, topical retinoids, tetracycline, macrolides, flourquinolones and aminoquinolines, amiodarone, methoxypsoralene and hydrochlorothiazide (Stern et al., 1984; Kaae et al., 2010; Schmidt et al., 2015b); (ii) oral contraceptive use; (iii) drugs suggested to potentially modify the risk of cancer including low-dose aspirin, non-steroidal anti-inflammatory drugs and statins (Jensen et al., 2009; Muranushi et al., 2015, 2016; Lin et al., 2018); (iv) history of comorbidities (defined by diagnosis codes and related medications) including diabetes, chronic obstructive pulmonary disease (COPD), chronic renal insufficiency, diseases associated with heavy alcohol consumption, inflammatory bowel disease, psoriasis, sarcoidosis and stroke (Henderson et al., 2015; Dąbrowski et al., 2016; Tseng et al., 2016; Groothoff et al., 2018); (v) Modified Charlson Comorbidity Index (CCI) score (0 low; 1–2 medium; ≥3 high) based on the prevalence of 19 chronic conditions (Charlson et al., 1987) and (vi) highest achieved education (basic, medium, higher or unknown). Exposure to the drugs outlined above was defined as two or more filled prescriptions prior to the index date and hospital histories of comorbidities were defined as a primary or secondary discharge or outpatient diagnosis. For all covariates, information within 1 year prior to the index date was disregarded.

Statistical Analyses. Conditional logistic regression was used to calculate ORs and 95% confidence intervals (CIs) for malignant melanoma associated with the use of HRT compared with never-use, adjusting for all potential confounders outlined above. We also performed secondary analyses to examine a potential dose-response association, stratifying cumulative HRT use by predefined categories (1–99 DDDs, 100–499 DDDs, 500–999 DDDs, 1000–2000 DDDs and >2000 DDDs). Analyses were carried out by HRT type (including oestrogen, progestogen and oestrogen/progestogen combinations, not restricted to exclusive use) and by route of HRT admission including oral HRT (oestrogen, progestogen and oestrogen/progestogen) and transdermal (oestrogen and oestrogen/progestogen). Analyses were also conducted to investigate associations with recent high use, defined as a cumulative use of ≥1000 DDDs (including the one-year lag-time) amongst users with a filled prescription in the 2 years prior to the index date. In all analyses, never use of HRT served as the reference category.

Subgroup and Sensitivity Analyses. A number of pre-specified subgroup and sensitivity analyses were also conducted. To examine potential effect measure modification, we stratified the main analyses according to age at index date, melanoma subtype (superficial spreading melanoma, nodular melanoma, lentigo maligna melanoma, acral lentiginous melanoma and other melanomas) and site of melanoma (i.e. skin of head and neck, trunk, upper limb, lower limb and site unspecified) and stage at diagnosis (i.e. localised [stage I or II, depending on thickness of the tumour] or non-localised [stage III or IV]). Tests of effect measure modification was carried out by conducting a likelihood ratio test of a conditional logistic regression model without a interaction term nested in a model with an interaction term corresponding to the patient characteristic defining the subgroup of interest. To get an impression of the magnitude of the influence of individual factors on the overall risk estimate, subgroup analyses were conducted by excluding individuals with certain characteristics that have been reported as potential risk factors for melanoma, i.e. a history of diabetes, chronic renal disease and history of non-melanoma skin cancer. In addition, three sensitivity analyses were conducted. First, as a control exposure, we examined use of intravaginal oestrogens with the reference category of never-use of HRT and intravaginal oestrogens. Second, we applied a new-user design excluding prevalent users of HRT during 1995–1996. Finally, the lag time was varied between 0 and 5 years in 6-month intervals.

Investigation of Melanoma Prognosis: Cohort Analysis

Study Population. We conducted a nationwide cohort study to investigate the risk of a second primary melanoma associated with the use of HRT amongst women aged 45–85 years and diagnosed with a previous melanoma. From the cases identified previously for case-control analyses, we identified those with incident melanoma between 1 January 2000 and 31 December 2013 (to ensure sufficient follow-up time). Follow-up time began 1 year after melanoma diagnosis and continued until a new melanoma diagnosis, death from any cause or end of the study period (31 December 2015), whichever occurred first. The first year of follow-up was excluded for latency purposes, to minimise detection bias due to increased contact with healthcare professionals and to ensure a true second primary melanoma diagnosis.

Exposure Definition. Exposure to HRT was defined into five mutually exclusive groups (Supplementary Figure S1); new users of HRT were those patients who had filled at least one prescription for HRT in the year post-diagnosis of melanoma but not in the 5 years prior to cohort entry.

Continuous users were defined as those who received at least one prescription for HRT in the 2 years prior to cohort entry and in the year post cohort entry; past users were those who filled at least one prescription for HRT between 5 and 2 years pre-diagnosis but not since then, and pre-diagnostic users were defined as those patients who received at least one prescription for HRT in the 2 years prior to cohort entry but not in the year post cohort entry. Non-users were those patients who did not use HRT (excluding intravaginal oestrogens) in the 5 years prior to diagnosis and to 1 year after diagnosis and were considered the reference category for all analyses. Secondary analyses investigated HRT type and route of HRT admission. For analyses investigating associations with intravaginal oestrogen, the reference category was non-use of all HRT, including intravaginal oestrogen.

Statistical Analyses. Cox proportional hazard models, using time from diagnosis as the time scale, were used to estimate hazard ratios (HRs) and 95% CIs of second primary melanoma associated with the use of HRT compared with non-use. Models were adjusted for the confounders listed previously with the addition of melanoma stage (TNM Classification of Malignant Tumors). The proportional hazards assumption was assessed using Schoenfeld residuals. Analyses also investigated the association between HRT and the secondary outcome of all-cause mortality. Sensitivity analyses were conducted restricting the follow-up period to 5 years.

All analyses were performed using Stata release 15.1. According to Danish law, ethical approval is not required for registry-based studies.