All-cause mortality

Included studies

Six publications were included for this review: four randomised controlled trials (RCTs) (Chlebowski 2017, Mulnard 2000, Os 2000; PEPI 1995) and two systematic reviews of RCTs (Kim 2020 and Nudy 2019).

This review was limited to RCT data only as observational studies would be subject to residual confounder bias, which is a particular problem for mortality.

For the systematic reviews, individual RCT data which matched the protocol was extracted and meta-analysed (see Appendix E). The individual RCTs incorporated from the systematic reviews are listed below:

• From Kim 2020 data from 11 RCTs were individually extracted: Cherry 2014, Collins 2006, Herrington 2000, Hulley 2002, Hodis 2003, Hodis 2016, Manson 2017, Tierney 2009, Veerus 2006, Vickers 2007, Viscoli 2001.
• From Nudy 2019 data from 11 RCTs were individually extracted: Angerer 2000, Giske 2002, Guidozzi 1999, Hall 1998, Harmann 2014, Hodis 2001, Jirapinyo 2003, Komulainen 1999, Kyllonen 1998, Nachtigall 1979, Samaras 1999.

Manson 2017 reported results from the same trial as Chlebowski 2017 however different subgroups were reported in each publication.

Combined effect estimates from the systematic review were not used as there was overlap with the RCT data from the two systematic reviews. The systematic reviews were primarily used to aid data extraction and risk of bias assessment. Therefore, effect estimates were reported separately for the individual RCTs and meta-analysed as appropriate according to the criteria set out in the protocol of this review. The included studies are summarised in Table 2, please refer to the systematic review extraction tables in Appendix D for details of the incorporated RCTs.

See the literature search strategy in Appendix B and study selection flow chart in Appendix C.

Excluded studies

Studies not included in this review are listed, and reasons for their exclusion are provided in Appendix J.

Comparison 1. Oestrogen + progesterone (any combination) versus placebo or no HRT

For the comparison oestrogen plus progesterone versus placebo or no HRT, high quality evidence showed no important difference in overall all-cause mortality. However, there were some differences when analysed by subgroup. Low to high quality evidence showed there were no important differences for the progestogenic constituents synthetic progestins, medroxyprogesterone and noresthisterone acetate on all-cause mortality. When looking at the subgroup age at first use, high quality evidence showed there was no important difference if age at first use was 50–59, 60–69 or over 69. Some of the evidence of very low to low quality was analysed separately due to low event numbers, and there was no important difference for all the subgroups for these studies.

Comparison 2. Sequential combined oestrogen + progesterone versus placebo or no HRT

There were relatively few studies contributing to sequential combined regimens and very low to low quality evidence showed no important difference for all subgroups.

Comparison 3. Continuous combined oestrogen + progesterone versus placebo or no HRT

Most of the evidence for oestrogen plus progesterone combined was in a continuous combined regimen and was of moderate to high quality. The evidence showed that there were no important differences between oestrogen plus progesterone and placebo or no HRT, overall and for all the subgroups. Some of the evidence of very low to low quality, was analysed separately due to low event numbers, and there was no important difference for all the subgroups for these studies.

Comparison 4. Oestrogen-only versus placebo or no HRT

For the comparison oestrogen-only versus placebo or no HRT, high quality evidence showed no important difference in overall all-cause mortality. The evidence showed no important differences for most of the subgroups. For constituent, moderate to high quality evidence showed there was no important difference for all-cause mortality with oestradiol, or equine oestrogen. The test for subgroup differences showed that there was not a statistically significant difference between the different ages at first use. Low quality evidence showed there was no important difference between arms for black ethnicity and high-quality evidence showed no important difference for white ethnicity. Some of the evidence of very low to low quality was analysed separately due to low event numbers, and there was no important difference for all the subgroups for these studies.

See Appendix F for full GRADE tables.

Included studies

A systematic review of the economic literature was conducted but no economic studies were identified which were applicable to this review question.

A single economic search was undertaken for all topics included in the scope of this guideline. See Supplement 2 for details.

Excluded studies

Economic studies not included in this review are listed, and reasons for their exclusion are provided in Appendix K.

The outcomes that matter most

The critical outcome was all-cause mortality. This was chosen because HRT could have a variety of different positive and negative effects on health, but any serious overall positive or negative effect should be apparent as a difference in overall mortality.

The quality of the evidence

The quality of the evidence for outcomes was assessed with GRADE and was rated as very low to high. Where the evidence was downgraded, there were mainly concerns around imprecision when 95% confidence intervals crossed 1 or more decision-making thresholds. Findings were also downgraded due to risk of bias for example around deviations for the intended intervention, as prescription registries or women’s self-reporting may indicate the use of HRT, but it cannot be fully confirmed that they took the HRT. The evidence was not downgraded for inconsistency or indirectness.

Benefits and harms

Due to the number of possible known and unknown confounders for the outcome all-cause mortality (life expectancy) only RCT data were included.

The committee discussed that some of the evidence was from people with underlying health conditions which may have impacted on the results. They also noted that the length of follow-up was relatively short in many of the studies and there may not have been sufficient time for a difference in mortality to become apparent. However, they noted that the study findings from the Women’s Health Initiative trial were given most weight and this included a population without underlying health conditions, and a longer follow-up period of 18 years. The committee discussed that overall high-quality evidence showed no difference in mortality, with either oestrogen-only, or combined HRT.

The committee looked at the analysis by subgroup and noted that for most of the subgroups there was no difference between HRT user and placebo. Whilst they noted that there was an isolated statistically significant decreased rate of all-cause mortality in a subgroup of people starting oestrogen-only HRT between 50 and 59, they noted that the confidence intervals in all age groups overlapped, there was no clear age trend, and the test for subgroup differences was not significant, meaning that there were no differences between subgroups by age. There was no such difference in the combined HRT analysis by age group of starting HRT.

The committee discussed some of the specific aspects of the Women’s Health Initiative. They considered that the regimen in the Women’s Health Initiative included a dose of equine oestrogen that is no longer used in practice, and that the route of administration was oral, whereas transdermal is more commonly used in practice. They discussed that the results were therefore specific to these characteristics of HRT. However, the committee agreed that there was high quality evidence to support a recommendation that would help to counsel a person who is considering HRT, and the results of the trial were still important.

The committee recommended that, when discussing HRT as a treatment option, it should be explained to people that HRT is unlikely to increase or decrease their life expectancy. This recommendation was based on the evidence which showed no important differences in risk of death from any cause with either oestrogen-only or combined HRT. The committee agreed that people should be aware of this when deciding whether or not to start or continue HRT, and that it would help them make a more informed decision. Though HRT treatment does not affect overall life-expectancy, the committee agreed that discussions should aim to establish the best balance between effectively treating symptoms and alleviating risks from the treatment (see Tables 1 and 2 in the NICE guideline), taking into account the person’s age, symptoms, medical history, preferences and personal circumstances. See also the section on discussing treatment options (see evidence review D) which highlights what should be discussed and taken into account in relation to HRT treatment.

In the absence of any good evidence to suggest that different types of progestogens have different effects on all-cause mortality, it was assumed that the findings could be generalised to all HRT.

There was also a lack of evidence on how HRT treatment in trans men and non-binary people can affect the risk all-cause mortality (or any other health outcomes). The committee assumed that the existing evidence discussed above could be generalised to transgender men and non-binary people who have never taken gender affirming hormone therapy. However, there was uncertainty about transgender people or non-binary people who have taken gender affirming hormone therapy in the past. The committee also noted that there was no evidence from people from minority ethnic family backgrounds. They agreed to make research recommendations for these groups to fill this evidence gap. The descriptions of the research recommendations can be found in appendix K of evidence report C.

Other factors the committee took into account

Whilst it is unclear how HRT might affect long term health outcomes (such as breast and endometrial cancer, CVD, and stroke) in trans men and non-binary people who have previously taken as gender affirming hormone therapy because evidence is lacking, the committee agreed that it is important to improve access to services for them. They therefore recommended that it should be ensured that they can discuss their menopause symptoms with a healthcare professional with expertise in menopause. The discussion of this is described in further detail in ‘the committee’s discussion and interpretation of the evidence’ section of evidence review C.

Cost effectiveness and resource use

No previous economic evidence was identified for this topic.

The recommendations made for this review topic centre around the impact of HRT on the risk of mortality. Whilst recommendations in this area will lead to people being better informed about treatment decisions, it is unclear how such information will change the treatment decisions made and how these will impact overall resource use. It would however be unethical to prevent such information being discussed with patients even if it did lead to an increase in resource use through changes in treatment decisions.

Excluded effectiveness studies

Table 9

Excluded studies and reasons for their exclusion.

Excluded economic studies

No economic evidence was identified for this review. See Supplement 2 for further information.