Association between menopausal hormone therapy use and mortality risk: a Swedish population-based matched cohort study

Abstract Background The net effect of menopausal hormone therapy on the risk of death is understudied, and current evidence is conflicting. Our aim was to investigate the association between menopausal hormones and risk of all-cause, cardiovascular, and cancer-specific mortality, based on the Swedish Prescribed Drug Registry and National Patient Registry. Methods This Swedish population-based matched cohort study included all women, 40 years or older, who had received at least one prescription of systemic menopausal hormone therapy between 2005–2014 (n = 290,186), group level matched 1:3 to non-users (n = 870,165). Multivariable conditional logistic regression models estimated the relative risk of all-cause and cause-specific mortality, adjusting for several clinical factors and comorbidities. Results Ever-use of menopausal hormones was associated with a slightly lower overall odds of all-cause (OR = 0.97, 95%CI 0.95–0.98) and cardiovascular (OR = 0.97, 95%CI 0.95–0.99) mortality, whilst 30% lower overall odds of cancer-related mortality (OR = 0.70, 95%CI 0.68–0.72) was shown. The odds of all-cause and cancer-related mortality were consistently reduced among women who began menopausal hormone therapy ≤60 years, whereas the association with cardiovascular mortality was inconsistent. In contrast, oestrogen-only therapy was associated with elevated odds of all-cause (OR = 1.14, 95%CI 1.11–1.16) and cardiovascular mortality (OR = 1.04, 95%CI 1.01–1.06) among women who began treatment at ≥70 years. Among current users, oestrogen-only therapy was associated with higher odds of all-cause (OR = 1.48, 95%CI 1.44–1.52) and cardiovascular mortality (OR = 1.24, 95%CI 1.20–1.28), whereas past use of oestrogen-only therapy suggested lower odds of mortality. Conclusions Our generalisable data suggest that early menopausal hormone treatment initiation does not increase the odds of mortality. However, the role of oestrogens in particularly cardiovascular mortality remains to be investigated.


Introduction
Menopausal hormone therapy (MHT) encompasses various treatments prescribed primarily to prevent vasomotor symptoms during menopause [1]. MHT was one of the most commonly prescribed drugs in the United States during the 1990s. The decline in MHT prescriptions over the past two decades might have resulted in a whole generation of women missing the beneficial effects of MHT. Menopausal hormones are still to date the most effective licenced prescription drugs to alleviate menopausal symptoms, experienced by 50% up to 75% of women [2][3][4]. Cumulative evidence indicates MHT use is associated with both favourable and deleterious outcomes, varying across different treatment types, formulations and regimens, [5][6][7][8][9] also depending on the individual risk profile. Whereas several studies have focussed on the association between MHT use and the risk of various cancer types and cardiovascular events, [10][11][12][13][14][15] evidence for the net effect of MHT on the risk of all-cause and cause-specific mortality remains inconclusive [16][17][18][19][20][21]. Notably, the age of women at enrolment has varied in the previous studies, and direct comparisons are challenging considering that the formulations and doses have changed over time, and the prescription patterns.
Currently, the latest guidelines of the US Preventive Services Task Force advise against MHT use for primary prevention of chronic diseases associated with ageing, as current evidence is insufficient to outweigh potential health benefits from the potential risks clearly [22]. The updated Swedish guidelines advocate using MHT primarily to treat vasomotor symptoms among women younger than 60 years or within 10 years from menopausal onset (median 51 years). A distinctive feature of Sweden is the availability of various MHT treatment options [23,24]. Therefore, we conducted this population-based matched cohort study including virtually all Swedish MHT ever-users, investigating the net effect of contemporary MHT use on the risk of all-cause and causespecific mortality.

Study design
This nationwide population-based matched cohort has been used extensively to investigate the risk of breast, [25] gastrooesophageal, [26] colorectal, [27] biliary tracts, [28] pancreatic, [29] and ovarian cancer [30], and as a source cohort for a study on colorectal cancer survival [31]. The study design and the group-level matching procedure (1:3) have been thoroughly described elsewhere [25]. In short, the Swedish Prescribed Drug Registry was used to identify virtually all Swedish MHT ever-users between 1 July 2005 and 31 December 2014. The unique Swedish personal identity number ensured a valid data linkage between other national registries supervised by the National Board of Health and Welfare [32,33]. The Regional Ethical Review Board in Stockholm granted the ethical approval (2014/1291-31/4) without a need for informed consent.
This population-based matched cohort study included all Swedish women, 40 years or older at first prescription, who received at least one dispensed prescription of systemic MHT between 1 July 2005 and 31 December 2013/2014, as ascertained from the Swedish Prescribed Drug Registry. The index date was the date of the first dispensed prescription of menopausal hormones. The Swedish Causes of Death Registry was used to ascertain the date and cause of death, defined by the International Classification of Diseases (ICD) 10 th edition. Information on the cause of death was available until December 2013 and for all-cause mortality until December 2014. Clinical data were ascertained from the National Patient Registry (in-and outpatient care) [34].

Exclusion criteria
All women with a history of any malignancy (apart from non-melanoma skin cancer, ICD-10 codes C43.0-9) were excluded before the matching. We also excluded MHT preparations with local effects (i.e., vaginal creams) because they are unlikely to have systemic effects [35].

Exposure
The Swedish Prescribed Drug Registry, which is over 99% complete for all outpatient care drugs, was used for exposure ascertainment based on Anatomical and Therapeutic Chemical (ATC) classification codes (Supplementary Table 1). Notably, MHT is not sold over the counter in Sweden, and injectable MHT is not used [3,25]. Women who received one or more dispensed progestin prescriptions during the study period were classified as EP-MHT users. Oestradiol accounted for over 99% of the oestrogenic component in EP-MHT prescriptions. Furthermore, MHT users were classified as current versus past users. Current users were defined as women who had received at least one MHT prescription within six months before their death or end of the study period. All other women were classified as past users. Notably, one prescription in Sweden covers three months, allowing for a potential washout period.

Statistical analyses
Multivariable conditional logistic regression models were used to estimate the relative risk of death due to all-cause and cause-specific mortality (i.e., cardiovascular and cancer mortality), providing adjusted odds ratios (ORs) with 95% confidence intervals (CIs). These analyses were adjusted for all the eight matching variables (i.e., year of birth, hysterectomy, delivery, thrombotic events, smoking-and alcoholrelated diseases, obesity, diabetes), and osteoporosis (i.e., confounding by indication), comparing MHT ever-users with women who did not receive MHT during the study period. Because the group-level matching procedure counterbalances the effect of follow-up time between the groups, no substitute date was created for the non-exposed women to evaluate the duration of follow-up [25,30]. The analyses were stratified by women's age at first prescription (categorised as <60 years, 60-69 years, and 70 years and older, based on the updated Swedish guidelines) [23,24], MHT types, and current versus past use. Subgroup analyses investigated the different oestrogen formulations and EP-MHT regimens separately. Women who switched therapy types during the study period were excluded from the subgroup analyses. Women were followed up until death or the end of the study period (December 2014), whichever occurred first. Interaction by age was modelled on an additive scale, with the youngest group constituting the comparison group. A p-value <0.05 was considered indicative of statistical significance. To assess the robustness of our results, we conducted sensitivity analyses excluding all women who died within the first year (i.e., between July 2005 and June 2006). All statistical analyses were performed on STATA MP4 15.1.

Patient and public involvement
Due to the registry-based nature of the study, no patients or the public were directly involved in the study.

Results
The results of this study are reported in line with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [37]. A total of 1,160,351 women were included in this nationwide cohort, including 290,186 MHT ever-users and 870,165 group-level matched MHT non-users. During the study period, 33,650 (11.7%) of MHT ever-users died and 104,740 (12.0%) of non-users died ( Table 1). The median age at death was 87 years among MHT ever-users and 86 years among MHT non-users.

All-cause mortality
Compared with non-users, ever-use of MHT was associated with 3% lower odds of all-cause mortality (OR ¼ 0.97, 95%CI 0.95-0.98) ( Figure 1, Supplementary Table 3). However, important differences were revealed by age at treatment initiation, MHT types, and current versus past use. Compared with non-users, higher all-cause mortality was shown among E-MHT users, particularly among women who began treatment at !70 years (OR ¼ 1.14, 95%CI 1.11-1.16, p-interaction <0.0001). In contrast, E-MHT user among younger women was associated with lower odds of mortality, and this association was strongest among women 60 years (OR ¼ 0.51, 95%CI 0.45-0.57, p-interaction <0.0001). Among current E-MHT users, the odds were 48% increased (OR ¼ 1.48, 95%CI 1.44-1.52), while past E-MHT use was associated with lower odds of all-cause mortality (OR ¼ 0.83, 95%CI 0.80-0.86) ( Table 2). Considering all women, tibolone was associated   Table 3.

Cardiovascular mortality
Of MHT ever-users 4.8% (n ¼ 13,985) died due to cardiovascular events and respectively 5.0% (n ¼ 43,328) of MHT non-users died due to cardiovascular cause (

Formulations and regimens
Among E-MHT users, the different oestrogen formulations were associated with varying odds of mortality (Supplementary Table 4 contrast, sequential testosterone-derived regimens were associated with an increased odds of all-cause mortality (OR ¼ 1.24, 95%CI 1.11-1.38) and higher odds of cardiovascular mortality (OR ¼ 1.56, 95%CI 1.01-2.41).

Sensitivity analyses
The association remained similar in the sensitivity analyses, excluding women who died within the first year from the beginning of the study (Supplementary Table 5).

Discussion
In this nationwide and population-based study, we observed a minimally reduced overall odds of all-cause and cardiovascular mortality among MHT ever-users compared with MHT non-users. In contrast, we found a 30% likelihood of overall cancer-related mortality among MHT users. Furthermore, the overall all-cause and cancer-specific mortality risks were consistently lower among women who began treatment <70 years. However, no association with cardiovascular mortality was found among women who initiated oestrogen-only or tibolone treatment 60 years. The observed increased mortality appeared to be driven by oestrogen-only therapy and particularly by oestriol but not oestradiol which accounted for over 99% of the oestrogenic component in all EP-MHT treatments. Moreover, an increased odds of all-cause and cardiovascular mortality was shown among older women on E-MHT (!70 years). Among current users of E-MHT, but not past users, 48% increased odds of all-cause mortality and a 24% higher cardiovascular mortality was shown. For tibolone and EP-MHT use, no overall association was found.
Notably, continuous EP-MHT regimens were associated with lower odds of mortality, whereas sequential administration of testosterone-derivatives increased the odds of all-cause mortality by 24%. This study has several strengths, including populationbased and nationwide design, enhancing statistical power to detect differences, counteracting selection bias, and facilitating the generalisability of our results. The unique Swedish personal identity number ensured valid data linkage between the high-quality Swedish health data registries, with high coverage and completeness of follow-up [33,38]. Information on the exposure was ascertained from the virtually complete Swedish Prescribed Drug Registry, and systemic MHT is not available over the counter in Sweden, minimising the risk of misclassification of exposure. The outcome was ascertained from the Swedish Causes of Death Registry, which is 100% complete.
Our study has some limitations as well. Whereas the population-based methodology counteracts potential selection bias, we cannot eliminate the possibility of healthy user bias. A link between MHT use, socioeconomic status and potentially better survival and higher adherence to treatment has been described [3,39]. To alleviate these concerns, we stratified the analyses by the different MHT types and current versus past use. Nevertheless, clinicians may be more prone to prescribe MHT to healthier women. The use of menopausal hormones is contraindicated among women at high risk of breast cancer, with current or a history of cardiovascular disease or among women with severe liver disease [23,24]. A closer follow-up is favoured among women using MHT, which could potentially lead to earlier detection of diseases selectively among MHT users. A sensitivity analysis excluding women who died within the first year showed similar results.
Another concern is potential left censoring, as the Swedish Prescribed Drug Registry was established in July 2005, and we lacked information on potential MHT use before the index date. A large proportion of the cohort (59.8%) was already on MHT from the start of the study; [25] however, any exposure misclassification should be at random between the groups. Whereas duration of treatment should be considered, it was more informative to compare current versus past users of menopausal hormones, given the concern of left censoring.
Although we could adjust for several potential confounding factors, we lacked data on oral contraceptive use. However, this group of women would be very young concerning the outcomes considering most Swedish oral contraceptive users are young adults [40]. Furthermore, whereas we lacked data on BMI, our analyses were adjusted for obesity. However, it should be acknowledged that whereas the function of ovaries ceases at menopause, heavier women with more adipose tissues might have different circulating hormone levels, which could influence the cumulative exposure to endo-and exogenous oestrogens [41]. Our analyses were not adjusted for unilateral/bilateral oophorectomies without concomitant hysterectomy because hysterectomy has a significant influence on deciding the type of menopausal hormones women will receivewhich is not the case for oophorectomies. Whereas unilateral/bilateral oophorectomies without a concomitant hysterectomy were rarely recorded during the study period, [42] it is also less likely that ovaries are left in among women who are already menopausal. Additionally, several indications for oophorectomies are single-sided, yet the Swedish codes for laterality are inconsistent and inadequate to determine with certainty whether both ovaries were removed [43].
The results of this nationwide and population-based cohort study are in line with the previous two meta-analyses, including 31 randomised controlled trials, [17] and another meta-analysis pooling together 30 studies [19]. Both of these systematic reviews and meta-analyses revealed a 28% lower mortality risk (n ¼ 16 studies) [17], respectively, 39% lower risk of mortality (n ¼ 13 studies) [19] among women younger than 60 years, while no association was found among older women (the mean age at treatment initiation was 66.9 years, [17] respectively >60 years) [19]. Our results suggest that oestrogen-only therapy among women !70 years or older could be associated with higher odds of all-cause and cardiovascular mortality. In the extended analysis of the Women's Health Initiative trial, no impact of either combination therapy (HR ¼ 1.02, 95%CI 0.96-1.08) nor unopposed oestrogens (HR ¼ 0.94, 95%CI 0.88-1.01) were observed compared to placebo on the risk of all-cause mortality. The results were similar for cardiovascular and cancer mortality during a cumulative 18 years of follow-up [44]. In contrast, in a population-based study including 14,361 women (aged !45 years and enrolled between 2003À2007), MHT ever-use initiated at an early menopausal age was associated with a higher risk of all-cause mortality (HR ¼ 1.31, 95%CI 1.10-1.56), whereas no association was shown among MHT never-users (HR ¼ 1.01, 95%CI 0.85-1.20), without consideration of the smoking status of ethnicity in the analyses. 20 In a prospective Danish study including fewer than 30,000 women (aged 50-64 years), the effect of lower cardiovascular mortality was evident only after five years of follow-up [21]. In our data, current oestrogen-only use, but not past use, was associated with a higher risk of all-cause and cardiovascular mortality. Timing of treatment initiation could explain the observed increased odds among current users, as an increase in cardiovascular mortality has been described shortly following MHT treatment initiation, only to flatten out after a few years of MHT use [45]. This could partly explain the lower cardiovascular mortality found among past users, who may have passed the time window of increased risk, and therefore, our results should be interpreted cautiously. Another possibility is that oestrogens may reduce the risk of coronary disease, [46] and past users may have received MHT for a longer duration. Currently, the primary indication of MHT is the alleviation of the vasomotor symptoms of menopause [22]. Yet, it is plausible that women prescribed MHT are experiencing more severe symptoms. These women may have different circulating hormone levels or changes in levels over the menopausal transition, influencing the cumulative exposure to endo-and exogenous oestrogens [47]. Whereas this study cannot determine the underlying pathophysiological mechanisms, our findings highlight the need for more mechanistic studies, particularly on the role of oestrogens in cardiovascular mortality.

Conclusions
In conclusion, our results indicate that MHT use initiated before 60 years is not associated with higher odds of mortality. Notably, current use of oestrogen therapy, but not past use, and oestrogen-only therapy initiation at an older age were associated with higher odds of all-cause and cardiovascular mortality.