Prior to Update #3, electronic searches identified 1,426 citations: 94 from the Cochrane Library, 735 from MEDLINE, 479 from Embase, 28 from hand searching of reference lists, 58 from pharmaceutical company submissions, and 32 from PreMEDLINE.

Results of literature searches for Update #3 are shown in Figure 1. Forty-four new studies were included: 6 head-to-head trials with hot flash or other symptom outcomes, 16 placebo-controlled trials with hot flash or other symptom outcomes, 9 placebo controlled trials with bone mineral density outcomes, 4 placebo-controlled trials with data about harms, 7 reports from the Women’s Health Initiative, and 2 recent systematic reviews. Dossiers were submitted by one pharmaceutical company (Wyeth, for Prempro, Premarin, and Premarin Vaginal Cream), but these dossiers did not contain any new studies not previously identified.

Figure 1

Figure 1

Results of literature search for Update #3

Key Question 1. What is the comparative effectiveness of different hormone therapy preparations when used by postmenopausal women or women in the menopausal transition stage for reducing symptoms?

Numbers of included studies are summarized in Table 2. Additional data on these trials are provided in Evidence Tables 1 (head-to-head trials) and 2 (placebo-controlled trials), and quality scores are provided in Appendix E. Quality ratings of studies added for Update #3 are shown in Appendix G.

Table 2

Table 2

Number of studies of estrogens and menopausal symptoms

A hot flash or flush refers to the spontaneous sensation of warmth, often associated with perspiration, resulting from a vasomotor response to declining estrogen levels. Although the term “flash” indicates a prodromal phase and “flush” the vasomotor dilation phase, they are combined in this report because they were reported inconsistently among the trials. These episodes are reported in many ways in the included studies. Most commonly, study participants recorded the number of episodes over a day or week period of time and changes indicated treatment responses. Other trials used measures such as percentage of participants experiencing symptoms or severity of symptoms, for example. A cumulative symptom score, the Kupperman Index [14], was used in some studies to classify the severity of menopausal symptoms. This index is based on the severity and intensity of hot flashes, paresthesias, insomnia, nervousness, melancholia, vertigo, weakness, arthralgia/myalgia, headache, palpitation, and formication. The maximum score is 51; a value of more than 20 indicates moderate to severe symptoms and a score of 10 describes mild complaints. Hot flashes are the most important symptom in the index. The use of this score is controversial, however, as it has not been validated.

Head-to-head comparisons reporting hot flash/flush outcomes

Twelve trials compared estrogen preparations head-to-head (Table 3, Evidence Table 1). Five trials compared different oral preparations, including one trial of CEE compared to oral E2 [15], one trial of CEE compared to E2 acetate and micronized E2 [16], two trials of oral E2 compared to E2V (one rated poor quality) [17, 18], and one trial of conjugated synthetic estrogen compared to E2 [19]. In three trials, the type of progestin was different in different estrogen treatment groups; [17–19] two trials used unopposed oral estrogen preparations [15, 19]. Symptoms improved from baseline for all treatment groups in these trials, but none found one oral estrogen preparation to be superior to another.

Table 3

Table 3

Head-to-head trials with hot flash/flush or other symptom outcomes

Five trials compared oral CEE to transdermal E2 (gel or patch) [3, 20–23]. One of these was rated poor quality [22] and the others were fair. All trials reported improved number and/or severity of hot flashes for all of the estrogen treatment groups compared to placebo or baseline. The poor quality trial found more patients had improvement in vasomotor symptoms at one year with E2 transdermal gel or patch than with oral CEE at one year [22]. However, because of flaws in the study’s design (high withdrawal rate, no intention-to-treat analysis, patients not masked), these results are not reliable. There were no statistically significant differences in treatment effects in any of the head-to-head estrogen comparisons in any of the other trials. Two trials were combined in a meta-analysis [20, 21], and one excluded because data was provided in graph form [3]. The pooled weighted mean difference in hot flashes was not significantly different between E2 and CEE treatment groups, thereby favoring neither agent (−0.3; 95% CI: −3.4, 2.7).

In a good quality trial of 159 women receiving either a vaginal ring releasing 50 or 100 mg of E2 compared to 1 mg oral E2 per day, the number of hot flushes/night sweats at 24 weeks was reduced in all groups and there were no significant differences between groups [24]. A fair quality trial of postmenopausal women with symptoms of vaginal atrophy compared an E2 vaginal ring with an E2 vaginal tablet [25]. There were no significant differences between treatments on self-reported vaginal symptoms at week 4, or on investigator-assessed vaginal signs at week 48. Urogenital quality of life was improved for both groups at week 48, but there were no differences between groups. Hot flashes were not assessed in this study.

Dose-response trends were demonstrated in three trials, with higher doses corresponding to bigger treatment effects [15, 20, 24]. In the intravaginal E2 ring trial, a dose response pattern was seen at 12 weeks, but not at 24 weeks [24]. Too few dose comparisons were conducted between estrogens to determine if differences exist.

Placebo-controlled trials reporting hot flash/flush outcomes

Thirty-six RCTs comparing an eligible estrogen preparation with placebo met criteria for this review through Update #2 (Evidence Table 2). Among 16 new trials added for Update #3, 13 (in 15 publications) were rated fair quality [26–39], 1 was fair to poor [40], and 2 (in 3 publications) were rated poor [41–43].


  • Trials were conducted predominantly in the U.S. or Western Europe and most often recruited participants from general or gynecology practices.
  • In general, each trial enrolled small numbers of participants and had multiple comparison groups.
  • Entry criteria varied: some stated that “most” or a percentage of participants had symptoms, some required a certain threshold of symptoms such as “5 or more vasomotor symptoms per day”.
  • Trials often enrolled both peri- and postmenopausal women but did not separate them in the analysis so comparisons between them cannot be made. Ages ranged from the mid 40s to 60’s; most trials reported mean ages in the early 50’s.
  • Hysterectomy status was clearly reported if the study criteria called for women either with or without hysterectomy. For trials including both types, the data were not separately reported so comparisons cannot be made.
  • No trial specifically addressed treatment in women with premature ovarian failure. A limited number of trials focused on women with recent hysterectomy and oophorectomy, although ages varied.
  • Reporting of concurrent medications, co-morbidities, or other potential confounders was minimal, although inclusion criteria generally focused on healthy, symptomatic women.
  • Many different outcomes were reported and lack of standardization makes them difficult to compare. Frequency of hot flashes was the most common measure and there were enough trials to combine them in a meta-analysis. Other outcomes are described in Evidence Table 2.
  • All estrogen preparations generally improved vasomotor symptoms among symptomatic women compared to placebo.
  • Women in placebo groups usually also had an improvement in symptoms, as the natural history of the estrogen withdrawal effect is gradual resolution of symptoms.
  • Women with the most frequent or severe symptoms most often had the biggest treatment effect and trials that enrolled highly symptomatic women tended to have large mean treatment effects.
  • Data on the effects of estrogen preparations on sleep were sparse and inconsistent.
  • Studies reporting health-related quality of life reported conflicting results.
  • In the WHI, CEE-only and CEE/MPA study, vasomotor symptoms improved; the small improvement in sleep was not likely clinically significant and health-related quality of life was not different from placebo at 3-year follow-up.

Eleven of twelve trials of oral E2 demonstrated statistically significant improvements in hot flash frequency and/or severity compared to placebo [44–54]. The one trial that reported no difference between groups was conducted in Chinese women in Hong Kong after oophorectomy [55]. Approximately 66% of women in this trial had vasomotor symptoms at baseline and 23–35% considered them “moderate to severe,” a lower level than in some of the other trials. One trial reported that women in early (3–12 months amenorrhea) as well as late menopause (>12 months amenorrhea) had benefit [44]. Eight trials included concomitant progestin/progesterone use (continuous and cyclic norethindrone acetate [56], cyclic nomegestrol) [44–47, 49, 52–54].

Three trials of E2V reported statistically significant improvements in hot flash frequency and/or severity compared to placebo [57–59]. All three trials included concomitant progestin/progesterone use (continuous medroxyprogesterone acetate [MPA], cyclic and continuous cyproterone acetate).

All six trials of CEE reported statistically significant improvements in hot flash frequency and/or severity compared to placebo [60–65]. Two trials included treatment groups with concomitant progestin/progesterone use (cyclic and continuous MPA, cyclic micronized progesterone) as well as unopposed CEE and reported no differences in treatment effects [64, 65]. One trial included three doses of CEE (0.3, 0.45, 0.626 mg/day) and noted dose-response relationships with higher doses corresponding to bigger treatment effects [65].

A 12-week trial of synthetic conjugated estrogens B compared with placebo in 281 US women included three doses of conjugated estrogen (0.3 mg, 0.625 mg, 1.25 mg/day). Significant reduction in frequency of hot flashes occurred at all dosage strengths compared with placebo (−72%, −85%, −87%, −47% for 0.3 mg, 0.625 mg, 1.25 mg, and placebo, respectively) but a dose-response relationship was not reported [66]. This study was rated fair quality. Adequate randomization and allocation concealment methods were used, intention-to-treat results are not reported, but only 5 patients were excluded from the analysis. A relatively high number of women discontinued treatment (19% for 0.3 mg, 15% for 0.625 mg, 17% for 1.25 mg, and 24% for placebo), but discontinuation rates were not significantly different between groups. Percent reductions differed from placebo (P<0.05) at 4, 8, and 12 weeks for all dosage strengths. Dose-response relationship was not reported.

One trial of estropipate indicated statistically significant improvements in hot flash frequency compared to placebo [67]. Women enrolled in this trial differed from the others because they had symptoms of depression as well as hot flashes.

All 11 trials of transdermal E2 reported statistically significant improvements in hot flash frequency and/or severity compared to placebo [20, 68–76]. Two trials included concomitant progestin/progesterone (cyclic NETA, continuous transdermal levonorgestrel) [71, 74].

There is one fair quality placebo-controlled trial of a transdermal vaginal ring releasing E2 for treatment of vasomotor symptoms [77]. Three hundred thirty-three women with at least 7 moderate to severe hot flushes per day, or at least 56 moderate to severe vasomotor symptoms per week, were randomized to a vaginal ring delivering the equivalent of 50 or 100 mcg E2 per day or a placebo vaginal ring. Symptoms were recorded by women on daily diary cards using a 4-point scale (0=no flushes, 1=mild, 2=moderate, and 3=severe). The efficacy analysis was not intention-to-treat; it included only women with a baseline measurement of moderate to severe vasomotor symptoms who had a vaginal ring inserted and who had at least one evaluation during the study (325/333 randomized). At 13 weeks, the percentage reduction from baseline in number of moderate to severe vasomotor symptoms per week was 79.9% in women randomized to the E2 50 mcg ring, 90.6% in those randomized to the E2 100 mcg ring, and 49.1% in those using a placebo vaginal ring (p<0.05 for both E2 groups compared to placebo).

For Update #3, we identified eight new fair-qual ity studies (in 11 publications) which examined symptoms (Table 4) [27, 28, 30, 33–39, 78]. An additional two studies (in three publications) were rated poor quality [41–43]. All of the new studies focused on postmenopausal women except one which examined a mix of postmenopausal women and women in the menopausal transition(Newton 2006). This latter study did not examine these two population subgroups separately. The number of flushes and/or the severity of symptoms decreased in all fair-quality studies of oral estrogen preparations: estradiol acetate [37], conjugated equine estrogen [34, 35], estradiol with norethisterone [28], oral estradiol with drospirenone [36], and ethinyl estradiol with norethindrone [38]. Transdermal estradiol 50mcg/day with norethindrone acetate decreased hot flashes compared to placebo [33], as did transdermal estradiol with oral tibolone (not available in the US) [27], whereas the UltraLow Transdermal estRogen Assessment trial (ULTRA) (n=417) did not demonstrate an improvement in postmenopausal symptoms among older, asymptomatic women compared with placebo at 2-year follow-up [78].

Table 4

Table 4

Placebo-controlled trials reporting symptoms or quality of life outcomes (new for Update #3)

Women’s Health Initiative Hormone Replacement Study

The Women’s Health Initiative (WHI), begun in 1993, was designed to examine major causes of morbidity and mortality in postmenopausal women (Tables 5 and 6). Details of hormone replacement studies from the WHI are shown in and Evidence Tables 3 (outcomes) and 4 (quality assessment). It encompasses two large, randomized, controlled, double-blind studies of estrogen therapy in postmenopausal women. In addition, there is a dietary trial and a calcium and vitamin D supplementation trial [79]. Women between the ages of 50 and 79 years were recruited form 40 clinical centers in the U.S. The WHI estrogen plus progesterone trial randomized 16,608 postmenopausal women with an intact uterus assigned to 0.625 mg of conjugated equine estrogen (CEE) plus 2.5 mg medroxy progesterone acetate (MPA) (Prempro, Wyeth) or to placebo [4]. This trial was stopped early due to an unfavorable global risk-benefit profile at 5.2 years, rather than the planned 8.5 years of duration [4]. The WHI CEE-only trial involved 10,739 women who had had a hysterectomy. This study was also stopped early (at 6.8 years) due to a lack of overall health benefit and an increased risk of stroke similar to that seen in the estrogen-only trial [80].

Table 5

Table 5

Women’s Health Initiative hormone replacement studies

Table 6

Table 6

Women’s Health Initiative: Summary of benefits

Barnabei and colleagues [81] reported that women with an intact uterus and moderate-to-severe hot flashes, night sweats, or vaginal or genital dryness at baseline who took CEE and MPA had improvements in these symptoms, as well as improvements in joint pain and stiffness (p<0.001 for each of these outcomes) at 1-year follow-up. Women who were younger, thinner, and closer to the menopause experienced more relief of hot flushes and night sweats. Among women asymptomatic at baseline, treatment-related beneficial effects included prevention of hot flushes (p<0.001), night sweats (p=0.003), and vaginal or genital dryness (p<0.001) and reduction in the incidence of new musculoskeletal symptoms (p<0.001).

A subgroup (8.6% of randomized population, oversampled for minorities) of women was examined at 3-year follow-up [81]. Among women who had moderate-to-severe symptoms at baseline, there were no significant differences between treatment groups for hot flashes or for various genital and musculoskeletal symptoms. Among women who were asymptomatic at baseline, vasomotor symptoms were not prevented, but these women were less likely to report vaginal or genital dryness and joint pain or stiffness than women on placebo.

The WHI was a good-quality study with high follow-up rates for most outcomes, intention-to-treat analyses, and baseline comparability of treatment groups. Adherence rates were low, however. In the CEE/MPA study, 42% of the treatment group and 38% of the placebo group stopped taking the study drug during the follow-up period [4]. In the estrogen-only study, 54% stopped the study medication [80].

Data informing the question of the applicability of the study to broad U.S. population are reported by Stefanick and colleagues [82]. The hormone replacement therapy study of the WHI involved a very large and diverse cohort: over 16,000 women in the estrogen/progesterone study and over 10,000 in the estrogen-alone cohort. The ethnic distribution of participants was similar to that of the U.S. census for women aged 50 to 79 years.

There were important differences between study participants and the general U.S. population, however [82]. Family household income and percentage with a college degree were higher in the study population than among general populations. The WHI hormone therapy participants contained fewer smokers and fewer women reporting no leisure time physical activity each week. There were more obese women in the study and the average intake of dietary calcium was above average. Study participants also appeared to be at fairly low risk for coronary heart disease, including low rates of hypertension, diabetes, and elevated cholesterol requiring drug therapy.

In addition, there are important differences between the populations of the estrogen-only study (post hysterectomy) [80] and the estrogen/medroxyprogesterone study (intact uterus) [82]. The estrogen-only study subjects were at higher risk for coronary heart disease, were more obese and less active, and had a slightly higher incidence of pre-existing cardiovascular disease than the estrogen/progesterone study subjects [82]. It is not possible to determine if the differences between the two study groups is due to uterine status, and data are not available to determine if demographic and other characteristics vary between women with and without a uterus [82].

Meta-analysis of placebo-controlled trials examining symptoms

Of 12 trials of oral E2 compared to placebo for treatment of hot flashes, five met criteria for the meta-analysis [44, 46–48, 52]. The pooled weighted mean difference in hot flashes is −16.8 (95% CI: −23.4, −10.2) per week compared to placebo. Combining only the four trials that included E2 and progestin/progesterone did not significantly change results (−19.1; 95% CI: −29.6, −8.6) [44, 46, 47, 52]. Trials were excluded from analysis if they did not provide data on frequency of hot flashes [45, 49, 53–55, 83] or did not provide standard deviations [50, 51].

Three trials of oral estradiol valerate did not meet criteria for the meta-analysis because they did not provide data on frequency of hot flashes [57–59].

Of six trials of CEE compared to placebo, one met criteria for the meta-analysis [63]. This trial reported a mean reduction of −19.1 (95% CI: −33.0, −5.1) of hot flashes per week after treatment compared to placebo. The other five trials were excluded from analysis if they did not provide data on frequency of hot flashes [62, 64], provided data in a graph form [60], or did not provide standard deviations [60, 61, 65].

One trial of estropipate compared to placebo was identified from the search and met inclusion criteria [67]. This trial reported a mean difference in hot flashes of −11.4 (95% CI: −22.6, −0.2) per week.

Of 11 trials of transdermal E2 compared to placebo, six met criteria for the meta-analysis [20, 68, 70, 72–74]. The pooled weighted mean difference in hot flashes for these trials is −22.5 (95% CI: −39.4, −4.8) per week compared to placebo. Only one trial included E2 and progestin/progesterone and results were not significantly different than the others [74]. Trials were excluded if data was provided in a graph form [71, 75] or the trials did not provide standard deviations [51, 75].

In Update #3, we were unable to obtain a pooled estimate of effect for any outcome, including hot flashes/flushes (the most frequently reported outcome in our review) as there was marked heterogeneity of relevant outcomes measures, including vasomotor composite scores, mean number of flashes/flushes per week, mean change in number of flashes/flushes, and percentage improved. In addition, very few studies reported measures of dispersion (standard deviation or standard error). We therefore used a qualitative approach to synthesis of these data.

Comparison with Cochrane meta-analysis

The results of this review and meta-analysis are consistent with a Cochrane review and meta-analysis of oral estrogens and menopausal hot flashes that includes trials published prior to 2000 [8]. The Cochrane review included double-blind, randomized, placebo-controlled trials of all forms of oral estrogen, alone or with progestin/progesterone, for at least 3 month’s duration. The meta-analysis reported weekly hot flash frequency and symptom severity. References were checked against the results of the OHP search. The OHP review differs from the Cochrane review because OHP defined a narrower range of oral agents, included transdermal forms, captured studies published after 2000, and included head-to-head comparisons.

The Cochrane meta-analysis indicated a significant reduction in the weekly hot flash frequency for estrogen compared to placebo with a pooled weighted mean difference of −17.5 (95% CI: −24.7, −10.2; 6 trials) per week, equivalent to a 77% reduction in frequency (95% CI: 58.2, 87.5). Severity of symptoms was also significantly reduced compared to placebo (odds ratio=0.13; 95% CI: 0.08, 0.22; 13 trials). Differences between types of estrogens were not determined, although trials of E2 and CEE predominated.

The review also found that the reduction in weekly hot flash frequency was similar for opposed and unopposed estrogen regimens compared to placebo (opposed: 77.1% reduction; 95% CI: 49.1, 89.7; unopposed: 76.8%; 95% CI: 59.4, 86.7). Symptom severity seemed to be better treated by opposed (odds ratio=0.10; 95% CI 0.06, 0.19; 10 trials) than by unopposed estrogen (odds ratio=0.35; 95% CI: 0.22, 0.56; 4 trials). However, differences between trials could also contribute to this discrepancy.

Sleep disturbances/night sweats

A trial of CEE in women with hot flashes and nighttime awakening at baseline indicated improvement in menopausal symptoms and measures of psychological well-being, but not in parameters of sleep quality such as total sleep time, sleep onset time, number of awakenings, and REM sleep duration compared to placebo [84]. Sleep disturbances were measured along with other quality-of-life measures in a subset of 1511 women enrolled in the WHI [85]. At one year of follow-up there was a small improvement (0.4 point on a 20-point scale) from baseline in women taking CEE compared with placebo, and no difference from placebo at 3 years.

A trial of transdermal E2 indicated significant improvement in sleep quality, sleep onset, and decreased nocturnal restlessness and awakenings compared to placebo [86]. In this trial, participants on E2 were less tired in the daytime and had associated alleviation of vasomotor, somatic, and mood symptoms. Women with the worst insomnia had the best improvement with E2. Two other trials of transdermal E2 indicated significant declines in night sweats compared to placebo [68, 70].

A head-to-head trial of an intravaginal ring delivering E2 compared with oral E2 [24] found improvement on the combined endpoint of hot flushes/night sweats in both groups, but night sweats are not reported separately, so it is not possible to determine the effect of the interventions on this outcome alone.

The WHI reported night sweats, as noted above under the section Hot Flashes/Flushes [81]. For Update #3, four new studies were identified. A small, fair-quality trial of postmenopausal women taking oral conjugated equine estrogens did not find significant improvement in sleep symptoms [29] and a study of transdermal estradiol found an improvement in sleep at 12 weeks (p=0.046) [33]. Two other studies were of poor quality [41, 42].

Mood changes

Nine trials of estrogen reporting mood outcomes met eligibility criteria, including one trial comparing E2 and E2V [17], one of oral E2 compared to placebo [45], two of transdermal E2 compared to placebo [87, 88], and five of CEE compared to placebo [34,64, 89–91].

In the head-to-head comparison trial of E2 and E2V, women were asked if symptoms of irritability, nervousness, anxiety, or depression were present or not before and after treatment cycles. Mood disturbances were more frequently reported by the E2 group (82%) than the E2V group (68%) at baseline [17]. At the end of treatment, symptoms were reduced to 52% in the E2 group compared to 44% in the E2V group (p=0.039).

In placebo-controlled trials, one study that randomized early postmenopausal women to oral E2 reported significantly improved scores after one year on the Beck Depression Inventory (21 items) as well as on the manic-depressive melancholia subscale (12 items) and the anxiety subscale (14 items), but not on the asthenia subscale or mania subscale [45].

One trial of transdermal E2 enrolled 50 women meeting DSM-IV criteria for major depressive disorder (26 women), dysthymic disorder (11 women), or minor depressive disorder (13 women) [87]. Remission of depression, measured by the Montgomery-Asberg Depression Rating Scale, was observed in 68% of women using E2 compared with 20% using placebo (p=0.001). Another trial of 87 women diagnosed with major depression, dysthymia, or minor depression compared changes in Hamilton Depression Scale (HAM-D) and Center for Epidemiologic Studies Depression Scale (CESD) scores after 8 weeks of treatment with low dose transdermal E2 (0.1 mg per day) or placebo. Both groups had improvements in depressive symptoms and the differences between placebo and E2 were not significant [88].

Five trials of CEE indicated mixed results. One trial reported significantly positive effects of CEE measured by an overall symptom rating scale and depression and feelings of inadequacy subscales, but not other subscales relating to neuroticism and effects of life events [89]. Another trial of psychologically well-adjusted women reported significant improvement on the Beck Depression Inventory with CEE (p<0.05) [90]. Women enrolled in the Heart and Estrogen/Progestin Replacement Study (HERS) with flushing who used CEE had significantly improved mental health and fewer depressive symptoms than those who used placebo, although women without flushing did not [91]. In the Postmenopausal Estrogen/Progestin Interventions Trial (PEPI), women on CEE did not differ from those on placebo for anxiety and affective symptoms [64]. However, many women in PEPI were also taking progestins that have independent effects on mood. Another trial indicated that CEE did not improve scores on the Beck, General Health Questionnaire, or Eysenck personality scales compared to placebo [61].

For Update #3, a small fair-poor quality study [40] found no significant differences between treatment with transdermal estradiol and placebo for depressive symptoms measured with the BASDEC (brief assessment scale depression cards). In a poor-quality study, Heinrich and colleagues [42] found no significant effects of treatment with estradiol on mood or depression, both measured with self-administered, German questionnaires.

Urogenital symptoms and sexual function

A head-to-head trial comparing CEE and transdermal E2 indicated that the majority of women reported either no change or improvement in vaginal dryness and itching, dyspareunia, and urinary pain and burning in all treatment groups with no major differences between groups [3]. All treatment groups demonstrated improved vaginal cytology, measured by the maturation index, with the biggest improvement in the higher dose E2 group (0.1 mg/day).

A head-to-head trial compared continuous low dose E2 released from a vaginal ring with CEE vaginal cream among women with signs and symptoms of urogenital atrophy [92]. Results indicated that the two agents were comparable for relief of vaginal dryness and dyspareunia, resolution of atrophic signs, improvement in vaginal mucosal maturation indices, and reduction in vaginal pH. The only outcome that differed significantly between agents was that participants found the ring more acceptable and preferred it to the cream. Similar findings were reported in another trial of the E2 vaginal ring and CEE cream [93] and a trial of the E2 tablet and CEE cream [94].

A head-to-head trial of an intravaginal ring releasing E2 versus oral E2 that was designed to assess vasomotor symptoms also reported urogenital symptoms as a secondary outcome [24]. The mean intensity of vaginal dryness, involuntary loss of urine, and pain during intercourse decreased from baseline to 24 weeks in both groups.

A placebo-controlled trial [77] examined urogenital symptoms in women randomized to a vaginal ring releasing the equivalent of 50 mcg or 100 mcg E2, or a placebo vaginal ring. There were some baseline differences among groups in vaginal irritation and itching (more severe in placebo group) and vaginal dryness (greater in placebo and 100 mcg vaginal ring groups). There was significant improvement in vaginal dryness at 4 and 8 weeks in the E2 vaginal ring 100 mcg group, and significant improvement in pain during intercourse at week 4 in both E2 groups and at week 13 in the E2 100 mcg group. There was a nonsignificant trend toward greater improvement of other urogenital symptoms in both E2 groups compared with placebo. In a subgroup of 60 women (18% of total) with signs and symptoms of vaginal atrophy at baseline, the maturation index was improved in both E2 groups compared with placebo at week 13.

A trial of transdermal E2, utilizing responses on the McCoy Sex Scale Questionnaire, indicated improvement in responses to five of nine items compared to placebo [95]. A correlation between improved sexual life and a quality-of-life questionnaire was also reported in this study. These findings were supported by another trial of transdermal E2 that indicated improvement in sexual problems and dysfunction as measured with the McCoy Sex Scale compared to placebo [76]. Another trial of transdermal E2 indicated improvement in vaginal dryness, but not dyspareunia, frequent urination, dysuria, stress incontinence, and nocturia, compared to placebo [96]. Another trial comparing transdermal E2 and placebo indicated no differences between groups for symptoms of vaginal discomfort, loss of libido, and incontinence [73].

There are two brief reports from one head-to-head study that measured sexual functioning and sexual quality-of-life in 186 women randomized to transdermal E2 or oral E2. One of these is an abstract [97] and the other a poster presentation [98]. On some, but not all, measures of sexual function and sexual quality of life, there was more improvement in women who used transdermal E2 compared with oral E2. This study is not published in full-text form and the brief reports do not provide sufficient detail to assess quality.

A trial of CEE reported significantly improved vaginal dryness and urinary frequency, but no significant improvement on six other items related to sexual function on a General Health Questionnaire compared to placebo [61]. The HERS trial found that women with at least one episode of incontinence per week at baseline who received CEE/MPA had worsening incontinence after approximately 4 years of follow up compared to women taking placebo [99].

The WHI reported on genital symptoms, as noted above under the section ‘Hot flashes/flushes’ [81].

In Update #3, the ULTRA study found no differences between treatment with low-dose transdermal estradiol on vaginal dryness [30] or on urinary incontinence [39]. There was a reduction in investigator-assessed vaginal atrophy, dryness, and friability for estradiol acetate compared with placebo (p<0.05) in a large, fair-quality study [37].

A Cochrane systematic review compared efficacy and safety of intra-vaginal estrogen preparations (creams, pessaries, tablets, and estradiol-releasing ring) for the relief of symptoms of vaginal atrophy (vaginal dryness, itching, discomfort, and painful sexual intercourse) [100]. Overall, the author concluded that the preparations appear to be equally effective for the symptoms of vaginal atrophy. CEE cream caused more side effects compared to estradiol tablets (uterine bleeding, breast pain, and perineal pain) or estradiol vaginal ring (endometrial overstimulation). For the comparison of the estradiol ring to CEE vaginal cream, there was no difference between groups in patient assessment of vaginal dryness or withdrawals due to adverse events, but there was more improvement in pruritis with the ring. For the comparison of estradiol ring versus estradiol tablet, vaginal dryness was improved more with tablets, but there was no difference between groups in genital pruritis or withdrawals due to adverse events. Symptom improvement was similar for tablet versus cream, but there were fewer withdrawals due to adverse effects with tablets compared with cream. There was no difference among all treatment comparisons for dysuria, nocturia, urgency, urge incontinence, participant symptom improvement in dryness, soreness, and irritation, loss of libido, and vaginitis.


A head-to-head comparison of CEE vs. transdermal E2 utilizing the Menopause Specific Quality of Life Questionnaire indicated improvement in all areas with no significant differences between groups in any of the domains at baseline or after treatment [101]. A trial comparing oral E2 and intravaginal ring E2 found significant improvement on the Greene Climacteric Scale among both treatment groups but no between-group differences [102].

Six placebo-controlled trials of oral E245 44, 47, 52–54 and one trial of E2V [59] reported significant improvements compared with placebo on various quality-of-life scales, including Kupperman index, Greene climacteric score, and General Health Questionnaire. One trial of oral E2 conducted in HRT-naive women in Thailand observed no difference in mean Greene score improvement compared with placebo after 12 months of treatment [83]. A trial of low-dose oral E2 (1 mg per day) [103] reported significant improvement from baseline at 6 and 12 weeks on six of nine domains of the Women’s Health Questionnaire (vasomotor symptoms, sexual behavior, depressed mood, somatic symptoms, anxiety/fear, and sleep problems). There was no difference between control and treatment groups on the memory concentration, menstrual symptoms, and attractiveness items of the scale.

Seven trials of transdermal E2 and placebo indicated improved health related quality-of-life and well-being measured by various instruments: Nottingham Health Profile, Psychological General Well-Being Index, Women Health Questionnaire, Kupperman’s index, McCoy Sex Scale, and psychological general well-being index [68, 70, 73, 74, 76, 96, 104]. One trial indicated that women with high well-being and no vasomotor symptoms at baseline had no improvement with treatment as measured by the Psychological General Well-Being Index [105].

The HERS trial (CEE), using non-validated quality of life instruments (Duke Activity Status Index, RAND Mental Health Inventory, among others), found that quality of life scores were significantly lower among women who were older, had diabetes, hypertension, chest pain, or heart failure, and that use of CEE had little effect [91]. One trial found a significant decrease in Kupperman’s index among women treated with E2V compared with placebo [59]. A trial of esterified estrogens reported improvement in the Quality of Life Menopause Scale compared to placebo [106].

Health-related quality of life (HRQL) measures were collected on a subgroup of women enrolled in the WHI randomized to CEE plus MPA or to placebo (n=16,608) [85]. Quality of life and functional status were assessed using the RAND 36-item Health Survey, which includes items about general health, physical functioning, limitations on usual role-related activities due to physical health problems, bodily pain, energy and fatigue, limitations on usual role-related activities due to emotional or mental problems, social function, and emotional or mental health. At 1-year follow-up, there were small but statistically significant positive effects of CEE/MPA on physical functioning (0.8 units on a 100-point scale), bodily pain (1.9 points on a 100-point scale), and sleep disturbance (0.4 units on a 20-point scale) compared with placebo. There were no differences from placebo in any other HRQL measure and by 3 years of follow-up (n=1511) there were no significant differences from placebo on any HRQL measure. Subgroup analyses detected no statistically significant interactions between baseline age, race, ethnicity, body mass index, or menopausal symptoms and HRQL. In a post hoc analysis of women 50 to 54 years of age who reported moderate-to-severe vasomotor symptoms at baseline, there was a positive effect on sleep disturbance, but no effect on other HRQL measures, despite significant improvement in vasomotor symptoms.

HRQL was also examined in the WHI estrogen-only study (n=10,739) [107]. At 1-year follow-up, there was a small positive effect of CEE on sleep disturbance (0.4 on a 20-point scale, p<0.001) and a negative effect on social functioning (1.3 on a 100-point scale, p=0.003). At 1-year follow-up of women who had moderate-to-severe vasomotor symptoms at baseline, 72.4% of the CEE group no longer reported these symptoms, compared to 55.6% of the placebo group (p<0.001). In a subsample (n=1,189) examined at 3-year follow-up there were no significant differences in any HRQL measure between treatment groups.

For Update #3, none of the three new studies reporting HRQL or related outcomes showed significant effects between the treatment and placebo groups. The ULTRA study of low-dose transdermal estrogen [78] reported no significant improvements in the SF-36 subscales of physical and mental function. The findings of Dayal and colleagues [29] were similar in that conjugated equine estrogen did not improve vitality, general health status, or quality of life at 12-week follow-up. A third study of women over 70 years randomized to oral estradiol or placebo also did not report significant changes in a “SF-36 score.” [26]

Key Question 2. What is the comparative effectiveness of different hormone therapy preparations when used by postmenopausal women or women in the menopausal transition stage for preventing low bone density and fractures?

Outcomes include bone density measurements at lumbar spine, forearm, and hip sites and/or fracture data from one or more sites. Numbers of included studies are summarized in Table 7 below; trials are described in Evidence Tables 5 (head-to-head trials) and 6 (placebo-controlled trials), and quality ratings are presented in Appendix F. Quality ratings of studies added for Update #3 are shown in Appendix G.

Table 7

Table 7

Number of studies of estrogens with bone density or fracture outcomes

Characteristics of the trials included:

  • Three trials with bone density outcomes compared estrogens head-to-head.
  • 68 trials with bone density outcomes compared an estrogen preparation to placebo.
  • 12 trials with fracture outcomes compared an estrogen preparation to placebo.
  • Trials often included concurrent calcium and vitamin D supplementation for both estrogen and placebo groups.
  • Five different forms of estrogen were used in these trials.
  • X-rays verified all fracture outcomes.
  • Bone density was measured in grams per centimeter or grams per centimeter squared by single-photon absorptiometry, dual-photon absorptiometry, dual x-ray absorptiometry (DXA), or quantitative computed tomography (QCT) at the lumbar spine, forearm, or hip sites.
  • Both prevention and treatment trials are included. Treatment refers to studies of women with pre-existing fractures or a diagnosis of osteoporosis at baseline.
  • The majority of studies were 1 or 2 years in duration although the longest trial was 5.2 years.
  • Both open and double-blinded studies are included because bone density and fracture outcomes are less prone to bias than self-reported symptom outcomes.

Bone density

Head-to-head comparisons

We identified no new head-to-head trials with bone density or fracture outcomes in this update. Four head-to-head trials compared different estrogen preparations, including three trials of CEE compared to transdermal E2 [108–110], and one trial of transdermal E2 compared to estradiol valerate (Table 8 and Evidence Table 5) [111].

Table 8

Table 8

Head-to-head trials with bone density outcomes

Two trials comparing CEE to transdermal E2 (0.05 mg/day for 25 days/month) evaluated two regimens of CEE (0.625 mg/day for 30 vs. 25 days/month) [108, 109]. All groups also received 2.5 mg/day of MPA for the last 12 days of treatment each month. In one trial, women using either CEE for 30 days or transdermal E2 for 25 days/month had an increase in lumbar spine bone mineral content compared to placebo (CEE: +4.4%, p<0.05: E2: +7.1%, p<0.01) [108]. Use of CEE for 25 days/month did not show a significant change (+1.3%, NS). Similar results were found when using these regimens in 118 women with prior hysterectomies [109].

A third trial comparing oral CEE (0.625 mg/day) with transdermal E2 (0.05 mg twice weekly) further evaluated the addition of alendronate (10mg/day) to each form of estrogen treatment. Increases in bone mineral density (BMD) occurred in all treatment groups after one year, and the increases did not differ significantly between the CEE and E2 groups. The addition of alendronate to either form of hormone therapy increased BMD significantly more than did hormone therapy alone [110].

One study of 73 healthy postmenopausal women age 45 to 54 years compared the effects of oral E2 and E2V on forearm and spinal BMD [111]. Both groups significantly gained bone density compared to placebo, and no significant differences between groups were found at any site.

Placebo comparisons

Sixty-four RCTs comparing an eligible estrogen preparation with placebo and reporting BMD outcome data met criteria for this review. These studies are described in Evidence Table 6. New studies added for Update #3 are shown in Table 9.

Table 9

Table 9

Placebo controlled trials with bone density outcomes (new for Update #3)

Characteristics of the trials include:

  • Trials were conducted predominantly in the U.S. or Western Europe and most often recruited participants from general or gynecology practices.
  • Both prevention and treatment trials were included and a broad patient population was provided for this review by including healthy postmenopausal women as well as those with pre-existing fractures.
  • Hysterectomy status was sometimes reported. For trials including both types, the data was not separately reported so comparisons could not be made.
  • The number of study subjects in trials ranged from 21 to over 16,000; trials ranged from 1 to over 5 years in duration.
  • 36 trials of estradiol in three forms were included: 16 trials of oral E2, 15 trials of transdermal E2, and 5 trials of E2V.
  • 26 trials of CEE and one trial of esterified estrogen were included.
  • One trial of conjugated synthetic estrogen plus medroxyprogesterone was included.
  • All estrogen preparations generally increased bone density or slowed its loss when compared to the placebo group.
  • Most results were reported as the mean difference between treatment and placebo groups or as percent change from baseline.

Fourteen of 16 studies of oral E2 demonstrated statistically significant improvements in bone density compared with placebo [83, 112–124]. One trial did not report treatment and placebo group differences, but stated that forearm bone density in the treatment group was statistically significantly increased from baseline while the placebo group showed no change [125]. Another trial reported a trend in E2 groups towards increased bone density, however statistical significance was not reached for between group comparisons [126].

All 15 trials of transdermal E2 reported statistically significant improvements in bone density compared to placebo [127–139]. Only three trials did not use concomitant progestin/progesterone [129, 134, 138, 140, 141].

Five trials of E2V with concomitant progestin/progesterone reported bone density outcomes [111, 142–145]. Four of the five trials noted improvement in treatment groups compared to placebo [111, 142–144], and one did not [145].

Twenty-six trials evaluated the effect of CEE on bone density outcomes [146–171]. All trials reported significant within-group changes in bone density at multiple sites for various doses with higher doses showing greater changes. In a good-quality trial comparing combination treatment with CEE (with or without medroxyprogesterone) plus alendronate to either treatment alone, patients on combination therapy had a significantly greater increase in total hip BMD than those on either ALN or HRT alone after 3 years (p<.01) [170, 171]. In one small (N=135) trial [151], CEE 0.625 mg increased bone density over 3 years at the femoral neck (p=0.02), total femur (p<0.001), and trochanter (p<0.001), but not at the lumbar spine (0.84% increase from baseline compared with placebo, p=0.39). Some trials reported that doses lower than 0.625 mg were less effective in maintaining or increasing bone density [147, 154, 158–160]. A more recent substudy of the Women’s HOPE trial found that most women on lower doses of CE with or without medroxyprogesterone (0.625, CE 0.625/MPA 2.5, CE 0.45, CE 0.45/MPA 2.5, CE 0.45/MPA 1.5, CE 0.3, CE 0.3/MPA 1.5) had less continued bone loss over 2 years than women randomized to placebo (See Table 9) [172].

The WHI study of CEE plus MPA [149] demonstrated consistent positive effects on BMD: hip BMD increased a mean of 1.7% and 3.7% by year 3, compared with a loss of 0.44% at year 1 and 0.14% improvement in the placebo group (p<0.001). Similar improvements were found in the lumbar spine. In subjects with 6-year follow-up BMD data (n=443), the percentage increase in lumbar spine BMD was 7.5% in the CEE plus MPA group compared with 2.6% in the placebo group. The CEE-only study of the WHI produced modest but consistent positive effects on bone mineral density [173].

One study of esterified estrogen [174] examined dosages of 0.3, 0.625, and 1.25 mg daily, and all doses showed statistically significant increases in lumbar spine and total hip bone density compared to placebo (p<0.05). The 1.25 mg/day dose was significantly more effective in increasing bone density at the lumbar spine than the lower doses [174].

Effect of discontinuation of estrogen on bone density

Two studies reported the effect on bone density after discontinuing the use of estrogen to determine if bone density gains were sustained after discontinuation, or if there was evidence that bone loss was accelerated in women who had used estrogen therapy when compared with those who had not used it [175, 176]. Both studies found the rate of bone loss after stopping estrogen was similar to that of women who did not receive estrogen treatment, as described below.

A follow-up study from the PEPI trial [175] measured bone density for an average of 4 years in women using CEE for 3 years. Further bone density gains were not observed in women after discontinuation of estrogen therapy, but there was also no evidence of accelerated bone loss when compared with those who had taken placebo. The second study reported the effect on bone mineral density of discontinuation of estrogen therapy for one year after 5 years of treatment in women enrolled in a randomized placebo-controlled trial of raloxifene and estrogen for prevention of postmenopausal bone loss [176]. This study also found that changes in bone density after one year of discontinuation were not significantly different in women using CEE compared with women randomized to placebo.

Comparison with other meta-analyses

A Cochrane review and meta-analysis published in 2002 on estrogen and bone density and fractures was reviewed for this report [7]. Fifteen of the trials included in the Cochrane review did not meet inclusion criteria for this review because they used ineligible estrogen preparations [177–191].

Results of the Cochrane meta-analysis included:

  • The pooled percent change in bone density was statistically significantly increased with estrogen compared to placebo at all measurement sites when combining results for all prevention and treatment trials and for both opposed and unopposed regimens.
  • After 1 year, the percent change in bone density was higher in the estrogen groups compared to placebo (5.4% at the lumbar spine, 3.0% at the forearm, and 2.5% at the femoral neck).
  • After 2 years of treatment, the estrogen groups had further increases in bone density compared to placebo (6.8% lumbar spine, 4.5% forearm, and 4.1% femoral neck).
  • At each of the sites, the percent differences between trials for prevention and treatment were not statistically significant.
  • There were no significant differences when opposed and unopposed estrogen trials were compared at 1 and 2 years.
  • A dose-response relationship was identified at each site at 2 years when low, medium, and high doses were compared.
    • For low-dose estrogen (equivalent to 0.3 mg CEE), the percent change in bone density was 3.9% at the lumbar spine, 3.1% at the forearm, and 2.0% at the femoral neck.
    • For high-dose estrogen (equivalent to 0.9 mg CEE) the percent change was 8.0% at lumbar spine, 4.5% at forearm, and 4.7% at femoral neck.
  • When different estrogen preparations were evaluated, including CEE, oral E2, and transdermal E2, they all demonstrated significantly improved bone density compared to placebo and there were no significant differences between them. For the lumbar spine, the differences between estrogen and placebo groups were:
    • 5.45% (95% CI: 3.31, 7.59) for transdermal E2;
    • 5.36% (95% CI: 3.99, 6.75) for oral E2;
    • 5.62% (95% CI: 4.64, 6.60) for oral CEE.

Another meta-analysis, published in 2003 [192], similarly found that different estrogen preparations, including CEE, oral and transdermal E2, E2V, and EE, were equally effective in the maintenance or gain of BMD at the lumbar spine and hip. This study was restricted to placebo-controlled trials of at least 2 year’s duration and enrollment of at least 60 subjects. Although the study did not report a systematic assessment of the quality of the trials selected for review, the number of dropouts in each trial and use of intention-to-treat results were assessed. The 2-year mean changes in lumbar spine BMD (weighted for the ratio of sample size/dropouts) are summarized as follows:

  • 7.6% (range 1.5% to 13.4%) for CEE;
  • 7.2% (range −1.5% to 20.0%) for oral E2, E2V, EE, and estrone sulphate;
  • 7.5% (range 3.4% to 14.4%) for non-oral estrogens.


Head-to-head comparisons

No head-to-head trials were found.

Placebo comparisons

We identified 11 studies of estrogen that included outcome data on fractures (Evidence Table 6). Seven were included [128, 135, 144, 155, 156, 168, 193] in a recent Cochrane meta-analysis [7], and the remainder were more recently published [4, 117, 149, 194].

Only one study of oral E2 evaluated fracture outcomes and found a statistically significant risk reduction for forearm fractures (RR=0.45; 95% CI: 0.22, 0.90) but not for overall fractures (RR=0.82; 95% CI: 0.53, 1.29) [117]. Both studies of transdermal E2 indicated no significant improvement in vertebral [128, 135] and non-vertebral fractures [128]. One trial of E2V in early postmenopausal women reported a significant decrease in nonvertebral (RR=0.29; 95% CI: 0.10, 0.90) but not vertebral fractures [144].

Seven studies examined CEE preparations [4, 155, 156, 163, 168, 193, 194]. Although some of these studies showed a trend toward reduction of fractures at various sites in the treatment groups, only the WHI showed a significant result [4]. When compared with the placebo group, total fractures for women on CEE were significantly reduced (HR=0.76; CI: 0.69, 0.85) [4]. Risks were also reduced for site-specific fractures of the hip and vertebra, although confidence intervals adjusted for multiple comparisons included 1.0.

In a more recent update of fracture data from the WHI [149] with average follow-up of 5.6 years, 8.6% of women in the CEE plus MPA group compared with 11.1% in the placebo group had a fracture at any site (HR 0.76; 95% CI, 0.69–0.83) and CEE plus MPA reduced the risk of hip fracture by 33% (HR ratio 0.67; 95% CI, 0.47–0.96). This effect did not differ in women stratified by age, body mass index, smoking status, history of falls, personal and family history of fracture, total calcium intake, past use of hormone therapy, bone density, or summary fracture risk score.

The WHI study of CEE use in women post hysterectomy [80] also reported a decrease in total fracture rates at mean follow-up interval of 6.8 years (HR 0.70, 95% CI, 0.63–0.79, 95% CI adjusted for multiple comparisons 0.59–0.83) (p<0.001). Hip fractures and clinical vertebral fractures were also decreased, although 95% confidence intervals adjusted for multiple comparisons overlapped a HR of 1.0 [hip fractures HR: 0.61 (adjusted 95% CI, 0.33 – 1.11); vertebral fractures HR 0.62 (adjusted 95% CI, 0.34–1.13)]. Additional data on fractures recorded through the study termination (average 7.1 years of follow-up) [173] also showed a reduction in incident fractures at the hip, spine, and wrist. These positive effects occurred largely irrespective of baseline risk factors for osteoporosis or fracture. The global index of overall health risks and benefits was balanced, however, with no evidence of overall benefit or risk noted even for women in the highest tertile of risk for fracture.

Comparison with Cochrane meta-analysis

Seven studies [128, 135, 144, 155, 156, 168, 193] reporting fracture outcomes were included in a Cochrane review published in 2002 [7]. Two trials indicating significant fracture risk reduction, including the WHI, were not included because they were published after the Cochrane analysis [117]. Findings included:

  • Four of five studies measuring vertebral fracture outcomes indicated non-statistically significant reductions in estrogen groups (RR=0.66; 95% CI: 0.41, 1.07) [131, 151 ,164, 188].
  • Five studies measured the effect of estrogen on nonvertebral fractures [128, 144, 156, 168, 193].
    • One study indicated a statistically significant relative risk reduction for nonvertebral fractures with estrogen use [144].
    • Three of the other studies had a risk reduction that was not statistically significant [128, 156, 193] and the other had a RR of 1.0 [168].
  • When all studies were pooled, there was a nonsignificant reduction in nonvertebral fractures (RR=0.87; 95% CI: 0.71, 1.08).

Key Question 3. What is the comparative safety of different hormone therapy preparations for short-term use (<5 years)?

Summary points

  • Breast tenderness and vaginal bleeding increase with all estrogen preparations.
  • In the few studies reporting on endometrial hyperplasia, no cases were identified with estrogen treatment.
  • The incidence of venous thrombosis was not increased in a large study of healthy women given estradiol and norethisterone.
  • All of the trials of symptoms and most of the trials of bone density and fractures were less than 5 years in duration and few enrolled more than 200 participants.
  • The WHI CEE+MPA study reported an increased rate of vaginal bleeding, breast tenderness, headaches or migraines, and vaginal or genital discharge than women in the placebo group at 1-year follow-up in women asymptomatic at baseline.
  • Cognitive function was not significantly affected in four fair-quality studies with follow-up between 12 weeks and 3 years.

Head-to-head trials

Adverse events reported in short-term head-to-head trials of different estrogen preparations are shown in Evidence Tables 7 (trials with symptom outcomes) and 8 (trials with bone outcomes). Head-to-head comparison trials provided insufficient evidence to determine the relative adverse effects of different estrogens. One trial of CEE and oral E2 reported that the incidence of possible drug-related adverse experiences ranged from 20% in placebo, E2 1 mg/day, and CEE 0.625 mg/day groups to 35% in E2 2 mg/day and CEE 1.25 mg/day groups, with no statistically significant differences between groups [15].

Most head-to-head trials reported similar rates of specific adverse events and withdrawals due to adverse events between treatment groups, with a few exceptions. In one trial, a significantly greater incidence of breast tenderness was found in women randomized to oral E2 2 mg plus NETA versus CE 5 mg plus MPA, and more women in the E2/NETA group withdrew from the trial during the first 3 months (17.1% vs. 4.1%; p<0.001) [19]. A trial of a vaginal ring releasing E2 compared with an E2 vaginal tablet found more withdrawals in the vaginal ring group, mainly occurring during the first 3 months of treatment and due to abdominal discomfort, lower back pain, and slippage of the ring [25]. In a head-to-head trial of an intravaginal ring delivering E2 compared with oral E2 for treatment of vasomotor symptoms, there were no significant differences between groups in the frequency of the most common adverse events [24].

Placebo-controlled trials

Withdrawals due to adverse effects and withdrawals due to specific adverse effects in placebo controlled trials are summarized in Evidence Table 9 for trials of hot flashes and Evidence Table 10 for trials of bone density and fractures. Specific adverse effects include atypical bleeding and endometrial hypertrophy, nausea and vomiting, breast tenderness, headache, weight change, dizziness, venous thromboembolic events (VTE), cardiovascular events, rash and pruritus, cholecystitis, liver effects, and others including breast cancer and additional problems. These outcomes were reported unevenly across studies and could not be combined in summary statistics.

Table 10

Table 10

Women’s Health Initiative: Summary of the adverse effects

Among trials with placebo groups, comparisons between types of estrogens cannot be made with the data provided. The most notable differences between estrogen and placebo groups were breast tenderness and vaginal bleeding; both symptoms were more frequent among women with higher compared to lower doses of estrogen regardless of type of estrogen. Reports of bleeding varied depending on concomitant progestin/progesterone use and regimen (cyclic or continuous). Several of the other symptoms, such as headache and mood changes, were common for both estrogen and placebo groups. Adverse skin reactions were most common among women using transdermal forms of E2. Withdrawals were often high among the placebo group in the hot flash trials because of lack of treatment effect among women who were enrolled based on the presence of symptoms.

In Update #3, among placebo-controlled trials examining efficacy/effectiveness of estrogen projects on symptoms, five studies reported harms [31, 34–36, 38]. All efficacy trials with bone density outcomes reported some information on harms. We identified an additional four studies which reported adverse effects without reporting efficacy or effectiveness (Evidence Tables 9 and 10) [195–198].

Oral estrogen/progesterone regimens increased vaginal spotting and atypical vaginal bleeding compared with placebo [31, 34, 38, 121, 122, 141, 170, 198, 199]. Withdrawal rates due to vaginal bleeding specific to treatment group were not reported in most studies, however. In a trial of BMD outcomes, 18% of women taking E2 1 mg plus intermittent norgestimate withdrew due to uterine bleeding [121]. A study of estradiol/drospirenone reported one woman with severe bleeding requiring hysterectomy, revealing adenomyosis and leiomyomata [36]. Langer and colleagues [198] reported no cases of endometrial hyperplasia in the treatment or placebo group; one case of endometrial cancer occurred in the placebo group.

In the ULTRA study [196] of low-dose unopposed transdermal estradiol (14 ug per day), vaginal bleeding occurred at year 1 in 5–6% of participants in both groups. Rates in year 2 were also similar (between-group p-value 0.03). Focal atypical endometrial hyperplasia developed in 1/188 women in treatment group and in 0/177 in the placebo group. One adenosarcoma of the uterus developed in the treatment group and none with placebo.

Breast tenderness was reported significantly more frequently with conjugated equine estrogen with medroxyprogesterone than with placebo [170, 171]. Headache [34] and dizziness or disorientation [31, 34] were reported at similar rates between estrogen users and the placebo group, as was the percentage of study subjects gaining weight [31]. Greenspan and colleagues [31] reported that the incidence of venous thromboembolic disease, endometrial and colon cancer, hospitalizations, myocardial infarction, and clinical fractures was similar between subjects receiving conjugated equine estrogen with medroxyprogesterone and placebo. Speroff and colleagues [38] reported that rates of headache, breast tenderness, vaginal bleeding, and palpitations were evenly distributed between treatment with ethinyl estradiol/norethindrone acetate and placebo (n=266). In a small study (n=40), similar rates of unspecified gastrointestinal adverse effects and headache were reported between the group using CEE and the placebo group (no statistics provided) [35].

In the WHI (see Table 11, summary table), vaginal bleeding was frequent among the CEE plus MPA treatment group [81], occurring in 42.5% and 51.0% of subjects in the first 6 weeks and 6 months of the trial, respectively. At year 5, 13.0% of the treatment group reported bleeding. In contrast, reports of bleeding never exceeded 8% in the placebo group. Among women asymptomatic at baseline, the treatment group also reported more breast tenderness at 1-year follow-up (CEE+MPA 9.3%, placebo 2.4%, p=0.026), particularly among thinner or older women and those further from menopause. This group also reported more vaginal discharge (CEE+MPA 4.1%, placebo 1.0%, p<0.001), more headaches or migraines (p=0.003), but less vaginal dryness (p<0.001) [81].

In Update #3, six new trials were identified which examined the effects of hormone therapy on cognitive function with follow-up between 12 weeks and 3 years [26, 31, 32, 40, 43, 78], all demonstrating no differences between groups at up to 2-year followup. The fair-quality ULTRA trial [78] found no significant differences at 2-year follow-up between treatment with low-dose transdermal estradiol and placebo for multiple measures of cognitive function: Mini Mental State Exam, logical memory, Brief Visual Spatial Memory Test, memory and recall of words, and verbal fluency. Similar negative results were found in a small, fair-quality study of estradiol patch [32]. In the fair quality study (n=373), the Folstein Mini-mental State Examination did not differ between the hormone replacement group (conjugated equine estrogen with or without progesterone depending on uterine status) and placebo at 3-year follow-up. Self-reported function and physical activity levels also did not differ significantly between groups [31]. Almeida and colleagues [26] also did not find significant differences between treatment with oral estradiol and placebo for a battery of cognitive tests among women 70 years of age and older. A poor-quality study did not find differences between treatment group for multiple measures of cognitive function either [43]. A fair-poor quality study reported improvement in 1 of 5 measures of cognitive function at 12-week follow-up (p=0.05) [40].

Key Question 4. What is the comparative safety of different hormone therapy preparations for long-term use (5 or more years)?

Summary points

  • In the WHI CEE/MPA study, coronary heart disease (CHD) events increased significantly, although CHD mortality did not at 5.2-year follow-up. In the WHI CEE-only study, CHD events were not increased.
  • The risk of stroke and venous thromboembolism were increased in both the CEE-only and CEE/MPA WHI studies.
  • The incidence of probable dementia increased in the WHI study of CEE/MPA, but not in the CEE-only study. Global cognitive function and mild cognitive impairment did not differ from placebo groups in either WHI study.
  • No head-to-head studies were available that compared adverse effects of different estrogen preparations after 5 or more years of use.
  • The WHI and HERS/HERS II studies provided the best evidence of long-term adverse effects for postmenopausal estrogen use and both used continuous regimens of CEE/MPA [4, 194, 200].
  • In the WHI CEE plus MPA study, a significant increase was noted in the hazard ratio (unadjusted for multiple comparisons) for cardiovascular events, stroke, venous thromboembolism, invasive breast cancer (p=0.05), and probable dementia. Rates of cardiovascular mortality were not increased
  • In the WHI CEE-only study, a significant increase was noted in the hazard ratio (unadjusted for multiple comparisons) for stroke and venous thromboembolism. Rates of probable dementia, cardiovascular events or mortality, and invasive breast cancer were not increased.
  • The WHI is the largest trial to evaluate the potential harms of postmenopausal estrogen use for both continuous CEE, MPA [4], and CEE only among woman post hysterectomy (Table 10) [80]. The WHI was designed as a primary prevention trial, not a trial of menopausal symptom treatment.

Cardiovascular events

The WHI demonstrated a statistically significant increase in coronary heart disease (CHD) events among users of CEE and MPA without known heart disease at a mean follow-up interval of 5.2 years (HR 1.29, 95% CI, 1.02, 1.63) [4]. Mortality from coronary heart disease events was not elevated, however (HR 1.18, 95% CI, 0.70 – 1.97). Events occurred early in the trial and persisted throughout the 5.2-year follow-up period. No interaction was found for age, race, BMI, smoking status, blood pressure, diabetes, statin use, or the effect of CEE/MPA on CHD events. Absolute increases in coronary heart disease cases were estimated at 7 per 10,000 person-years. Among the small subgroup with established CHD at baseline (n=400), the HR was 1.29 (95% CI, 0.64 – 2.56) and was similar to the group without known CHD.

Among women in the WHI using CEE alone (post hysterectomy) [80], no significant effect on CHD rates was observed compared with placebo at a mean follow-up of 6.8 years (5 fewer events per 10,000 person-years with CEE, HR 0.91, 95% CI, 0.75 – 1.12). Total mortality was also not significantly different between treatment groups. Among women with prior myocardial infarction or revascularization procedures, the effect of CEE compared to placebo on CVD event rates did not differ from the effect among women without known CHD.

The WHI study examined a global index of risks and benefits which was defined for each subject as the time to the first event among the monitored outcomes, including CHD, stroke, pulmonary embolism, breast cancer, colorectal cancer, hip fractures, and death [80]. This measure was used to assess the overall balance of risks and benefits [4] and was balanced overall (HR 1.01, 95% CI 0.91 – 1.12). CEE did not affect total mortality or cause-specific mortality [80].

The Heart and Estrogen progestin Replacement Study (HERS) [193] was a RCT of 2,763 women with a uterus comparing 0.625 mg CEE plus 2.5 mg MPA to placebo. All participants had documented coronary heart disease at randomization. The unadjusted relative hazard (HR) for CHD events was not different from placebo (HR 1.00, 95% CI, 0.84 – 1.17) over the mean follow-up of 6.8 years [200]. In post hoc analyses, the HR for the first year of treatment was 1.52 (95% CI, 1.01 – 2.29), with lower rates in subsequent years.


Risk for stroke was elevated in the WHI for CEE/MPA compared to placebo (HR 1.41, 95% CI, 0.97 – 1.85; adjusted 95% CI, 0.86 – 2.31) [4] and in HERS/HERS II (RR 1.09, 95% CI, 0.88 – 1.35). A systematic review and meta-analysis of other studies of estrogen and stroke reported a significant increase in stroke risk (RR 1.12, 95% CI, 1.01 – 1.23) [201]. Absolute increases in stroke are estimated at 8 per 10,000 person-years using WHI estimates [4].

In the CEE-only WHI study [80] the risk for stroke was increased by 39% in the CEE group (p=0.007; HR 1.39, 95% CI, 1.10 – 1.77). The differences in cumulative hazards for stroke began to emerge early after randomization and persisted throughout the follow-up period (mean 6.8 years, range 5.7 to 10.7 years). A greater risk of stroke was estimated among study participants who complied with study medications, taking more than 80% of study drugs, compared to the intention-to-treat population.

Venous thromboembolism

Risk for venous thromboembolism (including both deep vein thrombosis and pulmonary embolism) were elevated with long-term use of CEE/MPA in the WHI (HR 2.11, 95% CI, 1.26 – 3.55) [4]. Absolute increases in venous thromboembolic events are estimated at 18 per 10,000 using WHI estimates [4]. In the CEE-only WHI trial [80], active treatment increased venous thromboembolic disease (p = 0.03, HR 1.47, 95% CI, 1.04 – 2.08; adjusted 95% CI, 0.87 – 2.47).

Venous thromboembolic events were elevated during the HERS study [193] with 4.1 years of follow-up (HR 2.66, 95% CI, 1.41 – 5.04), however, during follow-up to a mean of 6.8 years in HERS II, this elevated risk decreased (p-value for time trend =0.08). The overall risk for all 6.8 years was 1.08 (95% CI, 1.28 – 3.4) [194]. A review and meta-analysis of studies of estrogen and venous thromboembolic events confirmed these findings, although studies with several different estrogen preparations were included and data were not stratified by preparation [202].

The incidence of venous thrombosis was not increased in a large study (n=2016) with 1-year follow-up of healthy postmenopausal women treated with sequential estradiol and norethisterone acetate; the only three cases were in the placebo and non-treatment groups [197]. New or worsening urinary incontinence increased with CEE among post-hysterectomized women (n=619). At 3-year follow-up, rates were 7.0% with treatment and 1.3% with placebo (p<0.02) [195].

Breast cancer

The WHI of CEE/MPA reported increased risks for invasive breast cancer at 5.2 years of follow-up (HR 1.26; 95% CI, 1.00 – 1.59) [4]. On the other hand, HERS/HERS II indicated no increase after 6.8 years (RR=1.27; 95% CI: 0.84, 1.94) [194]. Mortality from breast cancer was not elevated in either of these studies.

This increased risk of breast cancer with estrogens in the WHI CEE/MPA trial is consistent with estimates based on a meta-analyses of other studies (RR 1.23 to 1.35) [201]. Absolute increases in invasive breast cancer cases were estimated at 8 per 10,000 with CEE/MPA using WHI estimates [4]. Comparisons among estrogen preparations have not been conducted because of the limited data about types of preparations provided in the studies.

In the WHI study of CEE alone, the incidence of invasive breast cancer, the primary safety outcome for this trial, was decreased [80] over a mean follow-up duration of 6.8 years (HR 0.77, 95% CI, 0.59 – 1.01; 26 versus 33 cases per 10,000 person-years, p=0.06). This differential effect became apparent beginning in year 2.

A cohort study followed 3,175 French women, users (89% estrogen with progesterone) and non-users of estrogen, for 8.9 years for incidence of breast cancer [203]. Women who had used any type of estrogen therapy were eligible for the study; the most commonly prescribed regimen in France is transdermal E2 combined with oral progesterone or progestins. The relative risk of breast cancer associated with HRT use, adjusted for calendar period of treatment, date of birth, and age at menopause was 0.98 (95% CI, 0.73 – 1.75) compared with non-users. The risk was similar in the subgroup using combined therapy (adjusted relative risk 1.10, 95% CI, 0.73 – 1.66). Results are not presented by type of estrogen, so this study does not provide additional information about comparative risk.

Cognition and dementia

In the WHI Memory Study (WHIMS), an ancillary study to the WHI study, examined the effect of postmenopausal CEE with and without MPA on dementia and cognitive impairment in healthy women 65 years of age and older [5, 204–208]. The incidence of probable dementia among participants with an intact uterus taking CEE and MPA for mean duration 4 years (n=4532) was increased (HR 2.05, 95% CI, 1.21 – 3.48). Risk increased with age and with lower Mini Mental State exam scores at baseline [5]. Mild cognitive impairment was not significantly increased (HR 1.07, 95% CI, 0.74 – 1.55). Global cognitive function increased in both treatment and placebo groups for year 1 through 4 (likely due to a practice effect repeated testing) and then decreased in both groups with no significant differences between groups at year 5 and 6 [205]. Mean rates of change in cognitive function over time did not vary significantly between treatment groups for age, education, race, BMI, diabetes, or use of aspirin when multiple comparisons were taken into account (significant p<0.003) [205].

The WHI Study of Cognitive Aging (WHISCA) was an ancillary study to the WHI and WHIMS [206] and started 3 years after WHI randomization. CEE/MPA appeared to have different effects on various cognitive domains in older women after 4.35 years of treatment, with a negative effect on verbal memory (p< 0.01) and a positive effect on figural memory (p=0.012). There were no significant differences between treatment and placebo for other cognitive domains, depressive symptoms, and affect.

In the CEE-only study of the WHI, the incidence of probable dementia was not significantly increased at mean follow-up of 5.2 years (HR 1.49, 95% CI, 0.83 – 2.66) and was not significantly different from rates with CEE/MPA [207]. Rates of mild cognitive impairment were also not significantly increased [207]. Similar to patterns in the CEE/MPA trial, global cognitive function increased for the first 4 years, then decreased, with no significant differences between treatment groups [204]. Subjects with lower baseline scores in cognitive function had the greatest decline in cognitive function (p<0.01) [204]. The largest declines in scores occurred more frequently in CEE than in placebo, and the relative risk of decline of 10 units in the Mini Mental State exam with CEE compared to placebo was 1.47 (95% CI, 1.04 – 2.07) [204].


HERS/HERS II reported increased risks for biliary tract surgery among estrogen users with long-term use (mean follow-up of 6.8 years; RR=1.44; 95% CI: 1.10, 1.90) [194]. The Nurse’s Health Study also reported an increased risk with long-term use (RR=2.5; 95% CI: 2.0, 2.9) [209]. Data from this study also suggests that risk for cholecystitis increases with duration of estrogen use.

The HERS/HERS II trial reported increased risks for biliary tract surgery among estrogen users early in the study (RR 1.39 over the first 4.1 years, 95% CI, 1.00 – 1.93) [194]. Follow-up of 6.8 years revealed an overall HR of 1.48 (95% CI, 1.12 – 1.95). This outcome is supported by results of the Nurse’s Health Study, a large prospective observational study of estrogen users compared to nonusers (RR 1.8, 95% CI, 1.6 – 2.0) [209].

Ovarian cancer and endometrial cancer

The WHI and HERS/HERS II reported no increase in ovarian or endometrial cancer with CEE and MPA [4, 194]. Other studies of unopposed estrogen have indicated increased endometrial cancer for a woman with a uterus [210]. Observational studies of estrogen imply an increased risk for ovarian cancer [211, 212] while others do not [213].

Systematic review

A recent Cochrane systematic review assessed the effect of long-term hormone therapy on mortality, heart disease, venous thromboembolism, stroke, transient ischemic attacks, breast cancer, colorectal cancer, ovarian cancer, endometrial cancer, gallbladder disease, cognitive function, dementia, fractures, and quality of life [214]. Searches were conducted through November 2004. Fifteen randomized controlled trials were included, but the WHI and HERS, the largest trials, contributed most of the data. This review concluded that combined continuous hormone therapy significantly increased the risk of both venous thromboembolism and coronary events after one year, stroke after 3 years, breast cancer after 5 years, and gallbladder disease. In women over age 65, the incidence of dementia was also increased. In younger women (age 50 to 59 years) taking either combined regimens or estrogen-only hormone therapy, there was an increased risk of venous thromboembolism, but the absolute risk was low.

Key Question 5. Are there subgroups of patients for which one medication or preparation is more effective or associated with fewer adverse effects?

Age groups

Trials of estrogen and menopausal symptoms were usually conducted among women ranging in age from 40 to 60 years old with the mean age in the early 50’s. Data was not stratified by age and direct within-study comparisons cannot be made. Generally, women with the most symptoms had the most benefit. Trials of estrogen and bone density and fractures were conducted predominantly in older women in order to detect significant treatment effects because the prevalence of low bone density and fractures is higher among older women.

The most comprehensive trials of adverse effects (WHI and HERS/HERS II) enrolled older women with mean ages of 63 and 67 at baseline respectively. Data were not stratified by age in HERS/HERS II. In the WHI [149], there was no evidence that the effect of CEE in reducing fracture risk differed by age or time since menopause. It is not clear how well the findings of these trials relate to younger women using estrogen for short-term relief of symptoms. Younger post-menopausal women (50–54 years of age) who reported moderate-to-severe vasomotor symptoms at baseline were examined in the WHI CEE/MPA study [85] and there was a positive effect on sleep disturbance, but no effect on other HRQL measures, despite significant improvement in vasomotor symptoms. These data were consistent with results in older women

For Update #3, several studies examined older women and results were similar to studies among younger women. A study of oral estradiol compared to placebo in women over 70 years with an intact uterus did not report significant changes in a “SF-36 score” or in cognitive function [26]. In a study of community-dwelling women 65 years of age and older, Greenspan and colleagues [31] reported no significant difference between treatment with CEE 0.625 mg with or without MPA (depending on uterine status) and placebo for self-reported functional assessment including instrumental activities of daily living at one year and cognitive function at 3 years. The ULTRA study examined the use of low-dose transdermal estradiol in women 60 to 80 years and found that active treatment did not improve menopausal symptoms, urinary incontinence, or cognitive function at 2 years [30, 39, 78].

Racial/ethnic groups

Most trials enrolled white women in the U.S. or W. Europe who were recruited through clinical practices. The few trials conducted in nonwhite women took place in countries where differences in lifestyle factors could influence outcomes. The WHI reported a subanalysis by race [149]. Among black women (N=1124), CEE plus MPA reduced the risk of total fractures by 42%. This was not statistically significant because of the small number of fractures in this subgroup. There was no evidence of an interaction between treatment and race/ethnicity.

For Update #3, we identified no additional information on the effectiveness or harms in racial or ethnic groups.


The WHI reported that risks for breast cancer were not different among estrogen users with high risk compared to average risk, as defined by the Gail score or family history [4, 149]. No trials consider smokers, women at high-risk for ovarian cancer, or other risk factors and co-morbidities separately. The bone density trials include populations of women with and without pre-existing osteoporotic fractures and indicate that both groups benefit.

In the WHI CEE/MPA study [4], rates of CHD events were elevated to a similar degree in the small subgroup with established CHD at baseline (n=400) (HR 1.29, 95% CI, 0.64 – 2.56) compared to the main study group without known CHD.

In the WHIMS study [5, 204–208], the increased incidence of probable dementia among participants taking CEE (+/−_ MPA) was positively related to increasing age and lower Mini Mental State exam scores at baseline [5].

For Update #3, we identified no other additional information.

Early oophorectomy (<45 years) or premature menopause (<35 years)

No trials compare women with early oophorectomy or premature menopause with women undergoing menopause at an older age.

For Update #3, we identified no additional information.