Treatment of Vasomotor Symptoms of Menopause with Black Cohosh, Multibotanicals, Soy, Hormone Therapy, or Placebo: A Randomized Trial FREE

—The Editors

Hormone therapy remains the recommended treatment for vasomotor symptoms, but trials have shown serious risks with even short-term use (1 - 2). The use of herbs, particularly black cohosh, multibotanical supplements, and dietary soy for menopausal symptoms has grown dramatically (3 - 6). Few of these approaches have been scientifically evaluated. Women and providers are seeking safe, effective alternatives to hormone therapy. We designed the Herbal Alternatives for Menopause Trial (HALT) to provide rigorous evidence on the efficacy and short-term safety of commonly used naturopathic approaches for management of vasomotor symptoms.

HALT was a 1-year double-blind, randomized, controlled trial designed to investigate the effects of 3 naturopathic approaches for vasomotor symptom relief and hormone therapy compared with placebo. Study methods have been described elsewhere (7). The Group Health Institutional Review Board approved this study, and a data and safety monitoring committee monitored it. The study was conducted at Group Health, an integrated health plan in Washington State.

Eligibility criteria were as follows: age 45 to 55 years; late menopausal transition (≥1 skipped menses within the preceding 12 months) or postmenopausal (no bleeding within 12 months, or follicle-stimulating hormone level > 20 IU/mL if patient had undergone hysterectomy without bilateral oophorectomy); and 2 or more vasomotor symptoms per day over 2 weeks (≥6 moderate to severe symptoms). Women in menopausal transition were included because many are highly symptomatic and trial data are lacking for this group. Exclusion criteria were the following: contraindications to hormone therapy; use of hormone therapy or oral contraceptives within 3 months before the trial; use of herbal medicines for menopausal symptoms within 1 month before the trial; soy allergy; bilateral oophorectomy; history of breast cancer; and nonadherence during the run-in period (<80% of capsules taken).

From May 2001 through August 2003, women were recruited by using direct mail. Screening calls determined initial eligibility. Women attended an orientation visit at which eligibility was confirmed, physical measurements were collected, and placebo medication and questionnaires for the 2-week run-in period were provided.

Participants were randomly assigned by using SAS software (SAS Institute, Inc., Cary, North Carolina), stratified by previous hormone therapy and hysterectomy; block sizes within strata ranged from 5 to 25. Treatment assignments were sent to the University of Washington Research Pharmacy, where medications were bottled and labeled with a sequential identification number without treatment group indication. At the randomization visit, vasomotor symptom diaries and medication counts were examined to confirm eligibility and adherence. The study nurse determined the appropriate stratum, assigned the participant the next study number in that stratum without knowledge of group assignment, and distributed study medications.

The publication of results from the Women's Health Initiative (WHI) estrogen–progestin trial (2) raised new concerns about the safety of estrogen therapy. New study participants were given the choice of 5-arm (including hormone therapy) versus 4-arm (no hormone therapy) randomization. Current participants gave consent again, incorporating risk estimates from the WHI, and were given the option of finding out whether they had been assigned to hormone therapy; 16 were unblinded, 1 discontinued use of the study drug, and all remained enrolled. Following publication of the WHI Memory Study (8 - 9), we informed participants of those findings and restricted randomization to herbs and placebo. No further women were unblinded or discontinued use of the study drug.

Naturopathic medicine provided the model for study interventions. The herbal products, doses, and soy diet were based on approaches used by naturopaths when the study was designed (7). The study groups were as follows: 1) black cohosh (Actaea racemosa or Cimicifuga racemosa, 160 mg daily; 2.5% triterpene glycosides; 70% ethanol extract); 2) multibotanical; 3) multibotanical plus soy diet counseling; 4) conjugated equine estrogen, 0.625 mg daily, with (for women with a uterus) or without (for women without a uterus) medroxyprogesterone acetate, 2.5 mg; and 5) placebo. The multibotanical delivered daily doses of the following: black cohosh, 200 mg; alfalfa (Medicago sativa), 400 mg; boron, 4 mg; chaste tree (Vitex agnus-castus), 200 mg; dong quai (Angelica sinensis), 400 mg; false unicorn (Chamaelirium luteum), 200 mg; licorice (Glycyrrhiza glabra), 200 mg; oats (Avena sativa), 400 mg; pomegranate (Punica granatum), 400 mg; Siberian ginseng (Eleutherococcus senticosus, standardized constituents 0.8% eleutherosides E and B), 400 mg.

Black cohosh was provided by Pure World, Inc., (Hackensack, New Jersey). The multibotanical, ProGyne, was purchased from Progena Professional Formulations (Albuquerque, New Mexico) and encapsulated to study specifications. Both companies follow current good manufacturing practices, and single batches were used. ConsumerLab.com (White Plains, New York) tested the products after the study commenced. Dong quai, false unicorn, and pomegranate were not detected, suggesting that they were of poor quality or did not contain the tested marker compounds. Other marker compounds were detected in the approximate doses as labeled (7). The Appendix Table describes the herbal products in detail.

To facilitate blinding, medications and lactose placebo were encapsulated to provide 2 white and 2 blue capsules to each woman, and medication boxes were labeled with a unique identification number that did not indicate study group.

The soy food intervention was modeled after a successful 5-a-day intervention (10). We chose soy foods because of uncertainty about the efficacy of isoflavone supplements and because naturopaths commonly recommend whole soy foods. Participants received 5 telephone calls from a clinical dietitian and a 34-page booklet recommending 2 soy food servings per day (12 to 20 g of soy protein) (7). Other participants received 1 telephone call and a similar booklet reinforcing fruit and vegetable intake. Participants were instructed not to discuss dietitian calls with study nurses and were unaware that soy counseling was linked to the multibotanical.

The primary outcomes were change from baseline (measured over 2-week run-in period) to 3, 6, and 12 months (each measured for 4 weeks) and change from baseline to the average for all follow-ups with regard to the mean frequency and intensity of vasomotor symptoms (daytime hot flashes plus night sweats) and the mean Wiklund Vasomotor Symptom Subscale score (11 - 12). We also evaluated change from baseline to follow-up (months 3, 6, and 12 and average change) for daytime hot flash rate, night sweat rate, and the total Wiklund Menopause Symptom Scale score. Symptom diaries and global ratings of menopause symptoms are almost universally used in studies of vasomotor symptoms. Participants used a vasomotor symptom diary to record daytime hot flashes and night sweats, rating intensity as mild, moderate, or severe (scale, 1 to 3), as recommended by the U.S. Food and Drug Administration (13). Women completed the Wiklund Menopause Symptom Scale, rating the severity of 13 menopausal symptoms (sweats, hot flashes, sleep disturbance, fatigue, vaginal dryness, depression, headache, irritability, muscle/joint pain, breast tenderness, nervousness, palpitations, and dizziness/fainting on a scale of 0 (none) to 10 (severe) (11). Soy food intake was monitored by using a self-reported, validated soy food questionnaire (14).

Participants returned to the research clinic at 3, 6, and 12 months. Questionnaires were collected, and study medication was dispensed. Medication adherence counts were conducted by staff without participant contact and without knowledge of treatment assignment. Study nurses encouraged adherence and monitored adverse events during monthly telephone calls and at visits. Adverse events were identified through participants' responses to the question, “Have you had any medical problems or been hospitalized?” Responses were recorded and adverse events were followed until they were resolved. The study nurse determined intensity (mild, moderate, or severe) and whether the event was serious (yes or no). Events were reviewed by the study physician, who determined whether the event was study related, without knowledge of group assignment. Adverse events were coded by using the Coding Symbols for Thesaurus of Adverse Reaction Terms(15).

The Data and Safety Monitoring Committee reviewed study progress and unblinded outcome and safety data 5 times.

The study was powered to detect an effect of herbs halfway between the expected effects of hormone therapy and placebo, an outcome that we hypothesized would be meaningful.

Treatment effects, the difference between each treatment group and the placebo group with regard to the mean change from baseline, and the associated 95% CIs and P values were estimated by using a multivariate mixed model (PROC MIXED in SAS). We used an unstructured covariance matrix for the repeated measures, with separate parameter estimates for women in the hormone therapy group, since this structure best fit the data. Mixed models increase statistical power because of their ability to use all follow-up data and to better handle missing data. Although retention rates were very high, mixed-model analysis allowed us to use a true intention-to-treat approach, including data from all 351 randomly assigned women (n = 349 in adjusted analyses because of missing covariate data). The two Wiklund measures were analyzed after square-root transformation to normalize their distributions.

Mixed models were evaluated with and without adjustment for covariates. All models included a term for randomization protocol (4-arm vs. 5-arm). The adjusted models also controlled for age, body mass index (BMI), hysterectomy, menopausal status (menopausal transition vs. postmenopausal), and previous hormone therapy. All covariates except for BMI were selected a priori because of their hypothesized correlations with study outcomes and exposures. Results from the adjusted and unadjusted models were identical; we present only the adjusted results.

We also tested whether treatment effects differed by 4-arm versus 5-arm randomization (arm-by-treatment interaction). Since they did not differ, we present the results based on all randomly assigned women. An “as-treated” sensitivity analysis, restricted to women who took at least 80% of their study medications, was conducted, but results were similar and are not presented. Finally, we used mixed models to test whether the effect of each treatment varied in a statistically significant manner with BMI (nonobese [BMI < 30 kg/m²] vs. obese), hysterectomy, menopausal status, previous use of hormone therapy, and baseline symptom rate (<7 symptoms per day vs. ≥7 symptoms per day).

Adverse events rates were compared between each group and placebo by using chi-square tests or Fisher exact test (if expected count was <5). The study biostatistician conducted all analyses.

This study was funded by the National Institute on Aging and National Center for Complementary and Alternative Medicine. These agencies did not participate in the design, conduct, or analysis of the study or in decisions to submit the manuscript for publication. The National Institute on Aging did participate in decisions related to recruitment redesign in response to the release of the WHI findings and had a representative who attended all Data and Safety Monitoring Committee meetings.

We mailed 157 493 informational brochures and received 3443 responses (Figure 1). The baseline visit was attended by 509 women; of the 398 eligible women, 351 consented and were randomly assigned as follows: black cohosh (n = 80); multibotanical (n = 76); multibotanical plus soy counseling (n = 79); conjugated equine estrogen with medroxyprogesterone acetate (n = 29 women with a uterus) or without medroxyprogesterone acetate (n = 3 women without a uterus, all receiving unopposed estrogen); or placebo (n = 84). We enrolled 159 women under the 5-arm randomization scheme and 192 under the 4-arm randomization scheme; 147 of 183 women who were given a choice selected the 4-arm protocol. Ninety-two percent of women completed the study (327 of 351), and 87% (306 of 351) were taking study medication at 12 months.

Baseline characteristics were similar across treatment groups, with the exception of BMI (Table 1). On average, BMI was lower in the black cohosh group than in the placebo group and was higher in the hormone therapy group than in the placebo group. Average age was 52.2 years; 93% of participants were white, and all had at least a high school education. Women averaged 6.5 vasomotor symptoms per day (SD, 3.7; range, 1.4 to 24), and 34% averaged at least 7 symptoms per day at baseline. Average symptom intensity was 1.8 (on a scale of 1 to 3); 29% of participants reported symptom intensity averaging at least 2.0. The average Wiklund Menopause Symptom Scale score was 2.3 (SD, 1.2; range, 0.2 to 6.5) and the average Wiklund Vasomotor Symptom Subscale score was 4.5 (SD, 2.0; range, 0.8 to 10). Of 183 women (52%) who were in menopausal transition at baseline, 79 (46.6%) achieved 12 months of amenorrhea during the study.

Among women assigned to the soy food intervention, 77% completed 3 or more telephone calls (mean, 3.6). At baseline, women reported an average of 0.6 serving of soy per day. On average, women in the multibotanical plus soy intervention increased dietary soy by 1.1 servings per day between baseline and 3 months, compared with 0.1 serving per day in the other 4 groups.

The average adjusted number of vasomotor symptoms per day (Figure 2) and the Wiklund Vasomotor Symptom Subscale score (Figure 3) decreased between baseline and 3 months in all groups. There were no statistically significant differences in the average adjusted change in vasomotor symptoms per day or in vasomotor symptom intensity between the herbal interventions and placebo at 3, 6, or 12 months, or for the average over all the follow-up time points, with 1 exception: At 12 months, the multibotanical plus soy intervention had higher (worse) symptom intensity relative to placebo (P = 0.016) (Table 2). The average difference in vasomotor symptoms per day between the placebo and herbal treatments groups was less than 1 symptom per day at 3 months and less than 0.6 symptom per day for the average over all the follow-up time points. The average adjusted difference for hormone therapy compared with placebo was −4.55 (95% CI, −6.51 to −2.59) vasomotor symptoms per day at 3 months (P < 0.001) and −4.06 (CI, −5.93 to −2.19) vasomotor symptoms per day for the average over all the follow-up time points (P < 0.001) (Table 2).

There were no statistically significant differences in the Wiklund Vasomotor Symptom Subscale score between any of the herbal interventions and placebo at 3, 6, or 12 months or for the average over all the follow-up time points (Table 2). The Wiklund Vasomotor Symptom Subscale score was statistically significantly lower with hormone therapy than with placebo at all follow-up time points.

There were no statistically significant differences in hot flashes per day or night sweats per day between any of the herbal interventions and placebo at 3, 6, or 12 months or for the average over all the time points, with 1 exception: At 3 months, the black cohosh group had 0.38 less night sweat per day than the placebo group (CI, −0.72 to −0.04; P = 0.030) (Table 3). The difference between the herbal treatments and placebo was less than 0.6 hot flash per day and less than 0.4 night sweat per day at any time point; the differences in the average over all the time points were even smaller. The differences in hot flashes per day and night sweats per day between hormone therapy and placebo were statistically significant at all times points; over all follow-up time points, the average difference was nearly −3 hot flashes per day and nearly −1 night sweat per day (Table 3).

There were no statistically significant differences in the Wiklund Menopause Symptom Scale score between any of the herbal interventions and placebo at 3, 6, or 12 months or for the average over all the time points (Table 3). All differences between hormone therapy and placebo were statistically significant at all times (Table 3).

As-treated analyses, limited to women with at least 80% adherence, and separate analyses for women in the 4-arm and 5-arm randomization schemes, yielded similar results (data not shown). Results did not vary (no statistically significant treatment by subgroup interaction) when results were examined by baseline number of vasomotor symptoms (<7 vs. ≥7 per day), 4-arm versus 5-arm randomization scheme, baseline menopausal status, previous use of hormone therapy, hysterectomy (yes or no), or BMI (not obese [<30 kg/m²] vs. obese) (data not shown).

Women assigned to hormone therapy reported more breast pain (P < 0.001) and menstrual disorders (P = 0.04) compared with placebo (Table 4). There were no statistically significant differences between any of the 4 groups and placebo in the proportion of women with upper or lower gastrointestinal symptoms; nausea and vomiting; fatigue, asthenia, or malaise; or headaches or migraine. Too few severe adverse events occurred to make meaningful group comparisons (1 case of endometrial cancer in the multibotanical plus soy group 2.8 months after randomization; 1 case of breast cancer in the multibotanical group 4.7 months after randomization).

Over the 12-month follow-up period, the overall study sample on average took 86% of their pills; among individual study groups, adherence was 88% for black cohosh, 80% for multibotanical, 87% for multibotanical plus soy, 87% for hormone therapy, and 82% for placebo. These adherence figures include all 351 women; adherence was set to 0 for those who dropped out or stopped using study medication (Table 4). The primary reasons for discontinuation of study medication use or study withdrawal were no symptom relief (n = 10), other symptoms (n = 8), or change in health status (n = 8) (Figure 1).

In this large, randomized, double-blind trial, none of the 3 herbal treatments had clinically meaningful effects on any of the primary outcomes. As expected, hormone therapy resulted in a clinically important decrease in vasomotor symptom frequency and Wiklund scores throughout 1 year of treatment.

At least 5 randomized, placebo-controlled trials of black cohosh and menopausal symptoms have been published (16 - 20). All were short-term (≤12 weeks), and most were small, typically randomly assigning 30 to 60 women per group. Among 5 trials that examined frequency of hot flashes (16 - 19,21), the only trial reporting a positive effect for black cohosh also found no effect of conjugated equine estrogen versus placebo (18); this result raises concerns about the validity of the findings. When menopause rating scales were used, 2 trials reported a positive effect of black cohosh (16,18), 3 reported no difference from placebo (17,19,21), and 1 did not report main trial effects (20). The only trial that investigated a multibotanical containing black cohosh found no differences in vasomotor symptom frequency or menopause scale scores compared with placebo (22). Thus, the totality of the evidence does not consistently support a short-term effect of black cohosh on menopausal symptoms.

Effects of herbal products, such as black cohosh, may be sensitive to dose, extraction method, plant type, and coadministration of other herbs. The total daily triterpene glycoside dose in our black cohosh product was 5 mg, comparable to the 2 to 4 mg found in Remifemin (Schaper & Brümmer GmbH, Salzgitter, Germany), the most widely used product. Like Remifemin, our black cohosh (fingerprinted and verified to be Actaea racemosa) (7) was standardized to 27-deoxyactin. Remifemin is an isopropyl alcohol extract, whereas the products we tested are ethanol extracts. The implications of these different extraction techniques are unknown.

The literature on the other ingredients in the multibotanical is limited. Randomized trials have shown no improvement in vasomotor symptoms with dong quai (23) or Siberian ginseng (24). We are unaware of any study that has examined how the other components of the multibotanical affect vasomotor symptoms, although similar formulas are prescribed by naturopathic clinicians (7) and sold as over-the-counter supplements.

We reviewed 16 randomized clinical trials that tested whole soy or soy isoflavone supplements for vasomotor symptoms (25 - 40). Most were 12-week trials (range, 4 to 52). Eight studies found statistically significant improvements in at least 1 menopause symptom measure (26 - 28,31,33,35,37,39). The magnitude of benefit was a 25% to 55% decrease in frequency or severity of menopause symptoms. Only 3 short-term studies evaluated dietary soy and vasomotor symptoms (38 - 40); 1 found a statistically significant improvement in vasomotor symptoms after 12 weeks of a soy- and flax-enriched diet (39).

To our knowledge, ours is the longest and largest placebo-controlled, double-blind trial to date, and we included both placebo and hormone therapy as benchmark comparison groups. Findings from our study indicate that trials longer than 12 weeks are necessary to evaluate the sustainability of effects. Adherence and retention were high. We included women in the menopausal transition, and women with 2 or more vasomotor symptoms per day (compared with the 7 to 8 required in Food and Drug Administration–monitored drug trials (13)), because women with fewer symptoms are a key target for herbal therapies. Our findings were similar for women with 7 or more versus fewer than 7 symptoms per day. Although we found no evidence of significant side effects, only a larger and longer trial could provide reassurance in this regard. Our results are generalizable to white, relatively well-educated women who have an average of at least 2 vasomotor symptoms per day. We conducted independent testing of the herbal products; no contaminants were found, and key constituents were present in the amounts specified by the manufacturers (7).

An important question in a trial that does not find statistically significant treatment effects is whether the negative findings were due to a true lack of clinically important effects or a lack of statistical power. The confidence intervals for our primary outcome of symptoms per day can be used to determine the effect size that can be “ruled out” by our results (Table 2). Over all 3 follow-up time points, one can rule out reductions beyond 1.5 symptoms per day from black cohosh and 1.0 symptom per day from the multibotanical treatments. Whether these are clinically important effects is debatable, but we would argue that most women would not think so. It is important to emphasize that these are the largest possible effects that are consistent with our data. Our best estimates of effect are far less; the average reduction in symptoms per day over the entire follow-up period was less than 0.01 symptom per day, combining results for the 3 herbal groups, all of whom received black cohosh.

Treatments used in naturopathic practice motivated our choice of interventions. The whole-person approach used by most naturopathic physicians differs significantly from the treatments selected for our study, and this might have affected response to therapy. Time spent with the patient on counseling about diet, exercise, and emotional issues related to menopause; dose revisions; and additional supplements are important aspects of the naturopathic strategy for managing menopausal symptoms. We could not replicate this approach.

In summary, there is a pressing need for safe and effective interventions for vasomotor symptoms. Regrettably, this trial and the totality of the evidence indicates that black cohosh used in isolation, or in a multibotanical product, has little potential to play an important role in relief of vasomotor symptoms.