The goal of this study was to assess changes in health-related quality of life associated with discontinuation of HRT in a sample of elderly women. To our knowledge, no studies to date have specifically examined the impact of HRT discontinuation on quality of life. Given the large number of women who have either discontinued or will discontinue HRT, it is important to understand the associations between HRT cessation and HRQOL. The results of this study suggest that HRT discontinuation is associated with changes in HRQOL, but the direction and magnitude of these changes vary according to age. While HRT discontinuation was associated with apparent HRQOL declines among younger women in this study, women aged 85 or older appeared to experience improvements in HRQOL following HRT cessation.
The HRT discontinuation rate of 43% observed in this study is consistent with other recent reports. A recent study of U.S. national HRT prescribing rates found that overall HRT use declined by 38% between 2002 and 2003, with higher declines observed for combination therapy than for unopposed estrogen therapy [7
]. The majority of HRT users in the present study used unopposed estrogen, warranted only for women without uteri. The high observed rate of unopposed estrogen use in this sample is supported by a 2000 PACE survey which found that, while 40% of all female respondents reported a prior hysterectomy, 75% of current HRT users reported having had a hysterectomy. These results suggest that hysterectomy history may play an important role in prescribers' decisions to initiate or continue HRT in elderly women.
Depending on the specific measure examined, HRT discontinuers aged 65–74 averaged an increase of one to two days per month in which HRQOL was suboptimal. The results described here do not explain what factors may have mediated this decline. Possibilities include acute vasomotor symptoms, such as flushing, which have been shown in prior studies to be important mediators of depression in menopausal women [31
]. Alternatively, there may be more complex physiological effects of HRT discontinuation that affect mood, pain perception, or other aspects of perceived health.
The BRFSS HRQOL measures employed in this study have been previously shown to predict morbidity and mortality in the PACE population [32
]. Based on the significant associations of these measures with HRT discontinuation it is plausible that HRT cessation may affect risk for some adverse outcomes, although data needed to address this question are not yet available. Recent work suggests that HRT discontinuation is associated with reduced bone density and increased fracture risk [33
]. On the other hand, given the risks for breast cancer and cardiovascular disease that appear to be attributable to HRT, discontinuation may result in reduced risk for these outcomes. Clearly, more research is needed to explore the complex relationships among age, HRT cessation, HRQOL, and specific health outcomes.
A strength of this study is that it takes advantage of a natural experiment afforded by the availability of repeated survey data, which were collected during the general time period in which large numbers of women discontinued HRT. However, an important caveat regarding our findings is that our study design compared women who had either continued or discontinued HRT as of their follow-up survey date, but did not model the time course of HRQOL change following HRT discontinuation. Therefore, although the time interval between the baseline and follow-up surveys was approximately one year for all study subjects, the interval between HRT cessation and the follow-up survey response date could range from approximately one month to approximately one year. We conducted additional analyses to examine the impact of time since HRT cessation on HRQOL among discontinuers. Those analyses found no gradient in HRQOL change associated with the time elapsed since discontinuation. However, sample size considerations limited our statistical power to detect such differences. Further research is needed to model the time course of HRT discontinuation and its impact on HRQOL, and to examine whether short-term changes in HRQOL following cessation persist over time.
Due to the greater age and burden of illness present in the PACE population, the results of this study are limited in their generalizability to other populations, including younger women. These unique features of PACE, however, provide a valuable opportunity to examine the impact of HRT use and HRT discontinuation on elderly women. Most studies of HRT use have focused on women in the immediate postmenopausal years rather than elderly women. The present study, therefore, provides new information about the impact of HRT use and cessation on elderly women. The mean age of HRT users in this study was 75.5 years, and HRT users' ages at baseline ranged from 65 to 102. Although the large number of very old HRT users may be somewhat surprising, it is consistent with unpublished research indicating that 28% of PACE HRT users in 2000 were aged 80 or older. The same study found that, among current HRT users, the mean self-reported duration of use was 17.8 years, and 25% of current users reported that they had used HRT for 28 years or longer. These results highlight the need for awareness regarding the prevalence of very long-term HRT use by elderly women for whom menopause may have occurred several decades in the past.
A number of statistical limitations that could have a bearing on the results reported here should be noted. In this observational study, we sought to reduce selection bias through the use of propensity scores and by matching continuers with discontinuers on the basis of multiple factors. Comparison of the baseline HRQOL means for continuers and discontinuers suggested that the groups were well balanced in terms of baseline HRQOL. One limitation of this study, however, is that despite the statistical procedures employed to reduce bias, it is still possible that women who discontinued HRT differed in unmeasured ways from women who continued HRT. Conversely, because we matched on multiple variables as well as the propensity score, another statistical limitation relates to possible overmatching, as discussed by Rothman and Greenland [35
]. To the extent that matching may have been performed on variables related to HRT use but not to HRQOL, statistical efficiency may have been reduced. Of greater concern is the possibility that some matching variables may have been directly related to HRQOL change, which could lead to additional bias [35
Another limitation of this study is one that is inherent in any study relying on pharmacy claims data – HRT usage was inferred from prescription claim records. It is not known to what extent women who filled HRT prescriptions actually took the medication, or if the data recorded by the pharmacy at the point of sale accurately described patients' dosing instructions from their physicians. It is also possible that some women who did not fill prescriptions for HRT had access to the medication through other sources, such as samples received from physicians. Another important consideration relates to potential age-related measurement errors, such as recall bias, which may have affected women's perceptions of their HRQOL over the last 30 days. Age differences in women's expectations regarding the anticipated effects of HRT discontinuation could also be a factor. These limitations underscore the need for further studies to explicate the determinants and outcomes of HRT discontinuation.
Despite these limitations, the pattern of results observed in this study is an important reminder that even populations defined on the basis of age – such as PACE, for which the minimum eligible age is 65 – may include a broad range of ages with associated heterogeneity. Our results suggest that the response to HRT discontinuation among women aged 85 or older may be quite different from that of women in their 60's or 70's. The etiology of the age differences seen in this study is not known, and may reflect cohort effects, including the age-related measurement issues discussed above, or alternatively, the results may reflect physiological differences related to aging. Other recent work suggests that there are important age differences in the effects of estrogen on various physiological systems. For example, Brownley and her colleagues have recently reported differential associations between HRT and blood pressure according to the time elapsed since menopause [36
]. There is also growing evidence from animal studies that the effects of estrogen replacement on neurological function may be attenuated with increasing age [37
]. As discussed by Savonenko and Markowska [37
], such findings suggest that aging processes may modulate the mechanisms by which estrogen exerts physiological effects.
Regardless of the mechanisms that may explain the pattern of HRQOL changes reported in this study, the declines observed among younger HRT discontinuers emphasize the need for communication between clinicians and patients throughout the discontinuation process. Based on current evidence obtained from clinical trials, HRT increases risk for breast cancer, stroke, and other adverse health outcomes. On that basis, HRT discontinuation is rational and may provide important health benefits. Nevertheless, short-term changes in HRQOL may occur following HRT cessation, and strategies to optimize the discontinuation process are needed. For example, current recommendations for HRT discontinuation advocate a gradual cessation in which dosing is tapered over a three to six month period [38
]. Ideally, future research efforts will evaluate differences in HRQOL change according to the intensity and duration of the cessation process.