Nonhormonal Therapies for Menopausal Hot Flashes:  Systematic Review and Meta-analysis FREE

Hot flashes are the most common symptom related to menopausal transition.¹ They are experienced by more than 50% of menopausal women, can persist for several years after menopause, and for some women can interfere with activities or sleep to such a degree that treatment is requested.¹ A hot flash, or flush, is the spontaneous sensation of warmth, often associated with perspiration, palpitations, and anxiety, resulting from a vasomotor response to decreasing estrogen levels. It is hypothesized that endorphin concentrations in the hypothalamus decrease with declining estrogen production, enhancing release of norepinephrine and serotonin. This lowers the set point in the thermoregulatory nucleus and leads to inappropriate heat-loss mechanisms.^2- 4

Estrogen has been used as a hormone supplement for nearly 60 years to treat menopausal symptoms, and randomized controlled trials (RCTs) indicate that estrogen reduces the frequency of hot flashes by 77%⁵ or by approximately 2.5 to 3 hot flashes daily⁶ compared with placebo. However, recent studies reporting adverse effects of estrogen,^7- 9 such as cardiovascular events and breast cancer, raise important concerns about its use. Following the publication of the Women's Health Initiative trial in 2002 comparing estrogen with placebo, annual dispensed prescriptions for estrogen therapy in the United States declined from 91 million users in 2001 to 57 million users in 2003.¹⁰ Treatment of menopausal symptoms remains an indication for estrogen, although the US Food and Drug Administration advises physicians to use the smallest effective dose for the shortest duration possible.¹¹

Concerns about the adverse effects of estrogen have led to increased interest in alternative therapies for improving menopausal symptoms. Women with breast cancer who cannot take estrogen are also seeking alternatives. A survey of women aged 45 to 65 years enrolled in a health maintenance organization conducted before the Women's Health Initiative publications reported that 22% of women used alternative therapies to manage menopausal symptoms.¹² These included stress management, nonprescription remedies, chiropractic and naturopathic care, massage therapy, dietary soy, herbs, and acupuncture.¹² The idea that therapies targeting various nonestrogen components of the vasomotor response could influence symptoms is reasonable; however, evidence of their efficacy and adverse effects is generally lacking or unclear.

The purpose of our review is to compare the efficacy and adverse effects of therapies other than agents primarily composed of estrogen, progestin or progesterone, or androgen for menopausal hot flashes based on published RCTs. For simplicity, these are referred to as nonhormonal therapies, although some therapies such as plant isoflavones have weak estrogenic and antiestrogenic activities. This review uses a methodology that determines the strengths and limits of available evidence to provide a comparative assessment of treatment options.

Therapies included in this review and meta-analysis were selected from a broader systematic review of treatments for menopausal symptoms conducted by the investigators.¹ Trials reported from November 2004 through October 2005 were added to the previous review.¹ To compare efficacy across therapies, including estrogen, we included trials using ingested agents and comparable placebo treatments that reported hot flash outcomes. These included trials of antidepressants, other prescribed medications such as clonidine, and isoflavone extracts derived from red clover and soy. Trials of other therapies, such as dietary forms of isoflavones (flour, powder, food items), herbs and other dietary supplements, acupuncture, energy therapies, and behavioral therapies such as relaxation techniques, were not included because of nonstandardization of doses and forms and dissimilar approaches to blinding and using controls. Available trials of these therapies do not support their efficacy in improving hot flashes and are described elsewhere.¹

We identified relevant trials from comprehensive searches of MEDLINE (1966-October 2005), PsycINFO (1974-October 2005), and the Cochrane Controlled Clinical Trials Register Database (1966-October 2005). We also searched MANTIS (1880-July 2004) and Allied and Complementary Medicine Database (AMED) (1985-August 2004) databases, but these yielded no relevant trials. Specific search terms are available from the authors.¹ Additional articles were obtained by manually searching recent systematic reviews,^13- 15 reference lists of relevant articles and Web sites, and by consulting experts. We entered citations into an electronic database using EndNote version 6.0 (Thomson ISI ResearchSoft, Carlsbad, Calif).

Only published English-language, randomized, double-blind, placebo-controlled trials providing data on treatment of menopausal hot flashes using 1 or more nonhormonal therapies as described above were included. Head-to-head trials without a placebo group that compared nonhormonal therapies with estrogen or other medications were excluded because of difficulty interpreting results without a placebo.

Participants included women experiencing menopausal hot flashes who were recruited from health care settings or the general population. Trials enrolling women with breast cancer were included and additional data unique to them, such as concomitant use of tamoxifen or other selective estrogen receptor modulators (SERMs), were obtained. Trials of women with other major diseases or estrogen use within 1 month of commencement of the study were excluded.

Trials were included if they measured frequency or severity of hot flashes. Frequency was obtained by self-report using symptom diaries. Severity measures were also self-reported but in a variety of ways, such as using a graded scale or a composite measure (frequency × severity).¹⁶ Hot flash frequency and composite measures have demonstrated validity and reliability, and are highly correlated.¹⁶ Some trials used cumulative symptoms scores, such as those obtained with the Kupperman Index¹⁷ and Greene Climacteric Scale,¹⁸ to describe hot flashes as well as various other menopausal symptoms. However, use of these scores is problematic because some have not been validated, they contain different components and cannot be directly compared, and measures of vasomotor symptoms may not be specifically expressed. Outcomes were determined by the differences in hot flashes measured at baseline compared with the end of the trial. Treatment effects were defined as the differences in outcomes between the treatment and placebo groups at the end of the trial.

From each trial, we abstracted patient characteristics, eligibility criteria, interventions (treatment, dose, duration), comparisons, numbers enrolled and lost to follow-up, method of outcome ascertainment, results for each outcome, and adverse effects. Intention-to-treat results were recorded if available. Benefits and adverse effects of therapies were considered equally important and both types of outcomes were abstracted and summarized.

We reviewed all eligible trials for quality and applied a best-evidence approach that emphasized trials with the highest quality and most rigorous design.¹⁹ Two reviewers independently rated the quality of trials based on internal and external validity using criteria developed by the US Preventive Services Task Force.²⁰ These criteria incorporate key elements of the CONSORT checklist.²¹ We also considered additional criteria relating to trial size (≥50 patients per group) and duration (≥4 weeks) based on methodological research on studies of hot flashes.¹⁶ Interreviewer agreement was high overall (κ = 0.78). When reviewers disagreed, a final rating was reached through consensus with a third reviewer (H.D.N.).

We performed a meta-analysis to determine combined estimates of the efficacy of therapies in reducing hot flash frequency, the most uniformly reported outcome measure across trials. Trials eligible for meta-analysis provided adequate data for hot flash frequency outcomes. In some cases, we obtained additional data by contacting investigators.^22- 23 The mean difference (SE) in number of hot flashes per day between treatment and control groups was obtained or estimated from each trial. When the mean difference (SE) was estimated, we assumed correlation between baseline and end points to be zero to produce the most conservative estimates. Most crossover trials in this review did not report adequate crossover analysis and did not include washout periods, limiting interpretation of summary statistics. To combine crossover trials with trials of parallel design in a single meta-analysis, we used data from the first period only.²⁴

We used a random effects model²⁵ because there was evidence of heterogeneity between trials based on the χ² test for heterogeneity at a significance level of P=.10 and the I² statistic, percentage of total variation across studies due to heterogeneity rather than chance, of more than 25%.²⁶ Trials were stratified by type of therapy and duration of follow-up when data allowed. We performed sensitivity analysis to evaluate the effects of dose, concurrent use of SERMs, quality of trials (eliminating trials rated poor quality), and degrees of correlation between baseline and outcome measures.

We assessed publication bias by using funnel plots and Egger's linear regression method.²⁷ No publication bias was detected by these methods, however, their interpretation is limited by the small numbers of trials for each therapy.²⁸ All analyses were performed using Stata version 9.0 (StataCorp LP, College Station, Tex).

Of 4249 abstracts, 43 trials met inclusion criteria, which included 10 trials of antidepressants, 10 trials of clonidine, 6 trials of other prescribed medications, and 17 trials of isoflavone extracts (Figure 1).

Ten trials of antidepressant medications met inclusion criteria (Table 1).^{22- 23,29- 36} These included 6 trials of selective serotonin reuptake inhibitors (SSRIs) or serotonin norepinephrine reuptake inhibitors (SNRIs) (paroxetine,^29- 30 venlafaxine,^22,31 fluoxetine,^23,32 and citalopram²³), 3 trials of the antidopaminergic drug veralipride,^33- 35 and 1 trial of a selective monoamine oxidase-A inhibitor moclobemide.³⁶

Two trials reported significantly reduced hot flash frequency and severity with paroxetine compared with placebo.^29- 30 Women enrolled in a good-quality trial taking paroxetine controlled release (12.5 mg/d or 25 mg/d) experienced fewer daily hot flashes than women taking placebo (3.2-3.3 vs 1.8 fewer episodes, P = .01).²⁹ In a fair-quality trial that included predominantly women with breast cancer using tamoxifen, those women taking 10- and 20-mg/d doses of paroxetine also had fewer daily hot flash episodes compared with placebo (50%-51% vs 16%, P<.001).³⁰ Composite scores (frequency × severity) were reduced by 27% to 35% more than placebo in the 2 trials.^29- 30 Although treatment effects were comparable between dose levels, more women taking higher doses experienced adverse effects in both trials, including headache, nausea, and insomnia or drowsiness.^29- 30

A good-quality trial comparing 3 doses of venlafaxine extended release (37.5, 75, or 150 mg/d) with placebo in women with breast cancer or with self-perceived high risk of breast cancer reported decreased hot flash frequency (30%-58% vs 19%, P<.001) and decreased composite scores (37%-61% vs 27%, P<.001) with all doses of venlafaxine.³¹ Effects were greater with the 2 higher doses compared with the lowest dose, although it is not clear from the data that these differences were statistically significant.³¹ Women using tamoxifen had similar results as nonusers. A smaller fair-quality trial of women without breast cancer reported no differences in frequency or composite score for women using venlafaxine extended release (75 mg/d) compared with placebo.²² In this trial, venlafaxine users had a greater reduction in a measure of their perception of how hot flashes interfere with daily living (51% improved with venlafaxine vs 15% with placebo, P<.001). Adverse effects of venlafaxine included dry mouth, decreased appetite, nausea, constipation, and sleeplessness, and were more common at higher doses.^22,31

A fair-quality trial of fluoxetine (20 mg/d) in women with breast cancer or with self-perceived high risk of breast cancer indicated no significant differences in hot flash frequency or composite score between treatment and placebo groups before crossover.³² Another fair-quality trial in women without breast cancer compared fluoxetine (30 mg/d), citalopram (30 mg/d), and placebo and found no significant differences in frequency of hot flashes and overall Kupperman Index scores.²³ Main adverse effects included nausea and dry mouth.²³

Veralipride was compared with placebo in 3 poor-quality trials.^33- 35 Trials were conducted in the 1980s, enrolled 50 or less patients, and reported limited data. Two trials of veralipride (100 mg/d) reported reduced hot flash composite scores compared with placebo,^34- 35 and in 1 trial, more women noted subjective improvement with veralipride compared with placebo (85% vs 50%, P<.05).³⁴ Another trial reported changes from baseline consistent with the other trials but did not report between-group comparisons.³³ Mastodynia, galactorrhea, and gastrointestinal complaints occurred more frequently in the veralipride group in these trials.^33- 35

A poor-quality trial of moclobemide reported reduced hot flash composite scores at 2 doses (70% reduction at 150 mg/d and 35% reduction at 300 mg/d), although comparisons with placebo (24% reduction) were not reported.³⁶ Two patients withdrew from the trial because of somnolence from moclobemide.

Sixteen RCTs of other prescribed medications met inclusion criteria, including 10 trials of the centrally active α-adrenergic agonist antihypertensive clonidine,^37- 46 3 trials of the α-adrenergic agonist antihypertensive methyldopa,^47- 49 2 trials of the γ-aminobutyric acid analog anticonvulsive gabapentin,^50- 51 and 1 trial of the antispasmodic/sedative combination ergotamine, phenobarbital, and levorotatory alkaloids.⁵²

Of 10 trials comparing clonidine with placebo,^37- 46 all except 3 fair-quality trials^38,45- 46 met criteria for poor quality due to few patients, lack of clear inclusion and exclusion criteria, high attrition or loss to follow-up, no washout period in crossover trials, lack of data for precrossover comparisons, and short treatment duration (Table 2). Four trials^{39,42,45- 46} reported reduced frequency of hot flashes with clonidine compared with placebo and 5 did not.^{37- 38,41,43- 44} Two trials^45- 46 reporting reduced hot flash frequency included women with breast cancer taking tamoxifen; reduced severity⁴⁵ and composite score⁴⁶ were also reported. Two other trials^41- 42 reported reduced hot flash severity with clonidine compared with placebo, although 3 trials found no differences.^37- 38,40 Adverse effects were reported in 8 trials; dry mouth^38- 40,45 and insomnia or drowsiness^{39- 40,45- 46} occurred more frequently in women taking clonidine than placebo. Blood pressure was not adversely affected by clonidine.^{37- 39,41,43- 44}

Three poor-quality crossover trials compared methyldopa with placebo (Table 3).^47- 49 All trials provided precrossover statistics and none found significant differences in hot flash frequency.^47- 49 One trial⁴⁹ reported that methyldopa was more effective than placebo in improving scores on a visual analog scale. Fatigue or drowsiness, dizziness, and dry mouth occurred more often among women using methyldopa than occurred with placebo.^47- 49 No significant changes in blood pressure were noted^47,49; however, 1 study⁴⁸ reported orthostatic hypotension in 1 patient.

Two trials of gabapentin,^50- 51 a fair-quality trial⁵⁰ and a good-quality trial of women with breast cancer using tamoxifen,⁵¹ reported reduced hot flash frequency and severity compared with placebo (Table 3). Women taking a gabapentin dose of 900 mg/d experienced reductions in hot flash frequency in both trials (45% vs 29% for placebo, P = .02⁵⁰; 44% vs 15% for placebo, P<.001⁵¹) but not when taking a 300-mg/d dose.⁵¹ Somnolence, fatigue, dizziness, rash, heart palpitations, and peripheral edema were reported with gabapentin use.

A poor-quality trial compared Bellergal Retard, a combination of 0.6-mg ergotamine, 40-mg phenobarbital, and 0.2-mg levorotatory alkaloids, with placebo (Table 3).⁵² The number of hot flashes, severity, and sweating decreased in both the Bellergal Retard and placebo groups, but these reductions were not significantly different between groups. Adverse effects were similar between groups and included dry mouth, dizziness, and sleepiness.

Seventeen RCTs of isoflavone extracts met inclusion criteria. Trials evaluated red clover isoflavones,^53- 58 which contained genistein, daidzein, formononetin, and biochanin, and soy isoflavones,^59- 69 which contained predominantly daidzein, genistein, and their glucoconjugates.

Red clover isoflavones were compared with placebo in 6 trials; 1 trial rated good quality, 3 trials rated fair quality, and 2 trials rated poor quality (Table 4).^53- 58 All 6 trials used promensil, which contained a higher proportion of biochanin and genistein, and 1 trial used rimostil,⁵⁷ which contained a higher proportion of formononetin and daidzein. Only 1 fair-quality trial⁵⁸ indicated improved hot flash frequency with promensil compared with placebo (44% reduction vs 0%, P = .02); no trials reported differences in severity scores on the Greene Climacteric Scale^54,56- 58 or in a symptom diary.⁵³ Reduction in hot flashes was significantly faster with promensil than with placebo in 1 trial.⁵⁷

Soy isoflavones were compared with placebo in 11 trials (Table 5).^59- 69 Three trials, 2 rated fair and 1 rated poor, reported reduced hot flash frequency with soy isoflavones compared with placebo using genistein (54 mg/d),⁵⁹ genistein and daidzein (70 mg/d),⁶¹ and genistein and daidzein (50 mg/d).⁶⁷ Three other fair-quality trials found no differences in hot flash frequency compared with placebo using similar preparations.^65- 66,68 Severity scores compared with placebo improved in 3 trials using genistein, diadzein, and glycitein (33 mg/d)⁶² or genistein and daidzein (50 mg/d).^67- 68 Five other trials,^{60,63- 64,66,69} 4 trials of women with breast cancer,^{63- 64,66,69} found no differences in severity scores with soy isoflavones compared with placebo.

Adverse effects did not differ between isoflavone and placebo groups, although they were not well characterized in several trials. Gastrointestinal symptoms were generally the most common adverse effect with both isoflavone and placebo. Six trials evaluated endometrial thickness and found no differences between isoflavone and placebo groups during the course of the trials.^{54,59,62,65,67- 68}

Trials of SSRIs or SNRIs, clonidine, gabapentin, red clover isoflavones, and soy isoflavones provided data for the meta-analysis. Trials were excluded from the meta-analysis because they lacked adequate frequency data, did not provide precrossover statistics, or because there were inadequate numbers of fair- or good-quality trials to combine.

Seven comparisons from 6 trials of SSRIs or SNRIs were included in the meta-analysis.^{22- 23,29- 32} These included paroxetine (10 mg/d, 12.5 mg/d [controlled release], 20 mg/d, 25 mg/d [controlled release]), venlafaxine extended release (37.5 mg/d, 75 mg/d [2 trials]), fluoxetine (20 mg/d), and citalopram (20 mg/d). Data for 150-mg/d dose of venlafaxine extended release were not included because this dose was substantially higher than the other dosages. The combined weighted mean difference in the number of daily hot flashes for trials of SSRIs or SNRIs compared with placebo was −1.13 (95% confidence interval [CI], −1.70 to −0.57) (Figure 2). Sensitivity analysis indicated that the 4 trials enrolling women with breast cancer and SERM use reported significantly decreased hot flashes (mean difference, −1.40; 95% CI, −1.97 to −0.82),^29- 32 although the 2 trials of women without breast cancer and SERM use did not (−0.17; 95% CI, −1.41 to 1.07).^22- 23

Four trials of clonidine provided data for the meta-analysis^{38,41,45- 46}: 3 trials used 0.1-mg/d dose^41,45- 46 and 1 trial used a range of doses (0.05-0.15 mg/d).³⁸ The combined weighted mean difference in the number of daily hot flashes for clonidine compared with placebo in 4 trials was −0.95 (95% CI, −1.44 to −0.47) after 4 weeks use and in 2 trials was –1.63 (95% CI, −2.76 to −0.50) after 8 weeks use (Figure 3). In the 4-week trials, eliminating the poor-quality trial⁴¹ from the analysis did not influence results (weighted mean difference, −0.95; 95% CI, −1.45 to −0.46). The 2 trials enrolling women with breast cancer and SERM use reported significantly decreased hot flashes (weighted mean difference, −1.00; 95% CI, −1.51 to −0.49),^45- 46 although the 2 trials of women without breast cancer and SERM use did not (−0.53; 95% CI, −2.09 to 1.04).^38,41

Two trials of gabapentin (900 mg/d),^50- 51 including a trial of women with breast cancer using tamoxifen,⁵¹ provided a combined weighted mean difference in the number of daily hot flashes of –2.05 (95% CI, −2.80 to −1.30) compared with placebo.

Six trials of 2 types of red clover isoflavones provided data for the meta-analysis, including promensil^53- 58 (40 mg/d, 80-82 mg/d, 160 mg/d) and rimostil (57 mg/d).⁵⁷ The combined weighted mean difference in the number of daily hot flashes for red clover isoflavones compared with placebo was −0.44 (95% CI, −1.47 to 0.58) (Figure 4). Quality of trials and type of red clover isoflavone did not influence results. No trials were conducted in women with breast cancer using SERMs.

Six trials of soy isoflavones (50-70 mg/d and 150 mg/d) provided data for the meta-analysis.^{59,61,65- 68} The combined weighted mean difference in the number of daily hot flashes for soy isoflavones compared with placebo was −1.15 (95% CI, −2.33 to 0.03) after 4- to 6-weeks use in 5 trials, −0.97 (95% CI, −1.82 to −0.12) after 12- to 16-weeks use in 4 trials, and −1.22 (95% CI, −2.02 to −0.42) after 6 months in 2 trials (Figure 5). In the 4- to 6-week trials, results shifted slightly when removing a trial of women with breast cancer and SERM use⁶⁶ (weighted mean difference, −1.48; 95% CI, −2.49 to −0.48) but not when removing the single poor-quality trial.⁶⁷

Sensitivity analysis evaluating the effects of dose and degree of correlation between baseline and outcome measures did not significantly influence results for any of the therapies. Interpretation of the sensitivity analysis for any of the variables considered is limited because of the small number of trials available.

This systematic review and meta-analysis of double-blind, randomized, placebo-controlled trials of nonhormonal therapies provides supportive evidence for the efficacy of SSRIs or SNRIs, clonidine, and gabapentin in reducing the frequency and severity of menopausal hot flashes based on a small number of fair-good trials (SSRIs or SNRIs and gabapentin) or poor-fair trials (clonidine) (Table 6). The trials do not support the efficacy of red clover isoflavone extracts and present mixed results for soy isoflavone extracts. Evidence for other therapies is limited due to the small number of trials and their deficiencies. Few trials compare different therapies head-to-head and relative efficacy cannot be determined.

Although only 3 trials of the 7 comparisons of SSRIs or SNRIs with placebo indicated statistically significant reductions in hot flash frequency, the combined estimate indicated a reduction of approximately 1 hot flash per day. A meta-analysis of similarly designed RCTs of estrogen indicated reductions of approximately 2.5 to 3 hot flashes per day.⁶ However, direct comparisons with estrogen are not available and the study populations may not be similar. The SSRI or SNRI estimate is influenced by 2 large trials of paroxitine^29- 30 and a large trial of venlafaxine³¹ that showed significant reductions in hot flash frequency. Whether this effect can be expected from the other SSRIs is not known. Trials evaluating multiple doses of paroxetine^29- 30 found no differences in the effects of dose on hot flashes, although higher doses of venlafaxine were more effective than lower doses in 1 trial.³¹ Adverse effects were greater with higher doses supporting the use of the lowest effective dose for individual patients. The mechanism of action of SSRIs or SNRIs in relieving hot flashes is not known, but it has been hypothesized that temperature increases associated with hot flashes could be linked to an overloading of serotonin-receptor sites in the hypothalamus.²⁹ In these trials, hot flashes improved earlier than expected for psychiatric symptoms and were effective regardless of coexisting depression and anxiety, supporting a separate mechanism.³⁰

Ten trials of clonidine reported inconsistent results, with approximately half of trials demonstrating significantly reduced hot flash frequency or severity. Although most trials met criteria for poor quality, the 3 fair-quality trials provided a combined estimate of approximately 1 hot flash per day reduction with clonidine.^38,45- 46 Clonidine may relieve hot flashes by reducing peripheral vascular reactivity.

The combined estimate from 2 trials of gabapentin (900 mg/d) indicated a reduction of 2 hot flashes per day compared with placebo,^50- 51 although the use of 300-mg/d dose did not provide benefit.⁵¹ The mechanism of action of gabapentin, effective for treating seizure disorders and neuropathic pain, is unclear. Increased activity of neurotransmitters in the hypothalamus as a consequence of up-regulation of the gabapentin binding site from estrogen withdrawal has been proposed.⁵¹

Hot flash frequency was not reduced when all trials of red clover isoflavone extracts were combined, and results for soy isoflavone extracts were contradictory even among the largest and highest quality trials.^{59,61- 66,68} These results are consistent with other recent systematic reviews.^13- 14 We also reviewed trials of other forms of soy isoflavones, such as flour, powder, and food items, in another study.¹ However, these trials are difficult to compare because of variability of components and doses. Overall, evidence does not support benefit in relieving hot flashes. Isoflavones derived from red clover and soy have weak estrogenic and antiestrogenic activities and are able to bind to estrogen receptors, potentially mediating hot flashes triggered by estrogen deficiency.^70- 71 This mechanism of action could also cause adverse effects. Although 6 treatment trials included in this review reported no endometrial thickening with short-term therapy, a trial reporting adverse effects of isoflavones indicated an increased rate of endometrial hyperplasia after 5 years of 150 mg/d of soy isoflavone supplement use.⁷²

Eleven trials in this review included women with breast cancer, many of whom were using tamoxifen.^{29- 32,45- 46,51,63- 64,66,69} Hot flashes in women with breast cancer are particularly problematic because they are common in this population,⁷³ estrogen therapy is contraindicated, and some agents such as paroxetine may interfere with tamoxifen.⁷⁴ Although breast cancer was not active during the course of the trials, it is not known if breast cancer itself influences vasomotor symptoms. Hot flashes are an adverse effect of tamoxifen, and it is not clear if patients enrolled in treatment trials were experiencing hot flashes induced primarily by menopause or tamoxifen. This difference may be irrelevant if mechanisms are similar. Trials considering tamoxifen users separately from nonusers found comparable results.¹⁶

Our review and meta-analysis is limited by potential publication bias and bias from including only English-language publications. Most treatment trials of hot flashes, regardless of therapy, share methodological challenges. Placebo effects are large and drop out rates are often high in placebo groups due to lack of effect. This must be considered when designing trials and recruiting patients. Other limitations include use of highly selected small samples of women; relatively short and variable lengths of follow-up; inadequate reporting of loss to follow-up, maintenance of comparable groups, contamination, and methods of analysis; use of dissimilar measures and outcomes that are often not standardized or validated; and unclear inclusion and exclusion criteria. Reporting of adverse effects varies by study and usually involves self-reported symptoms experienced during the course of the trial. As a result, data on adverse effects are limited and not sufficiently systematic to draw conclusions. Comprehensive methods including long-term follow-up are needed to more completely determine adverse outcomes. Larger, more rigorous, and more standardized trials, which include head-to-head as well as placebo comparisons, are needed to accurately determine relative benefits and adverse effects of a wider array of therapies.

Despite increasing interest in therapies for menopausal hot flashes that avoid use of estrogen, the efficacy and safety of other options currently are not well supported. The SSRIs or SNRIs, clonidine, and gabapentin trials provide some evidence of efficacy. However, effects are less than for estrogen therapy, few trials have been published and most have methodological deficiencies, and generalizability beyond the small clinical populations studied could be limited. Adverse effects and cost may prohibit use for many women. Although these therapies may be most useful for highly symptomatic women who cannot take estrogen, they are not optimal choices for most women.