The ASRM Menopause Special Interest Group (MSIG) Addresses the Women's Health Initiative (WHI) Study Results

April 19 , 2007
by: ASRM
Orginally published in ASRM Announcement

By this time, most members of ASRM are familiar with the first report from the WHI, published in July 2002 in the JAMA. This portion of the WHI, a multi-center, double-blind, randomized placebo controlled  trial comparing conjugated equine estrogen (CEE) 0.625 mg plus medroxy-progestrone acetate (MPA) 2.5 mg [Prempro™ (Wyeth, Philadelphia, PA)] versus placebo included a population of 16,608 women with an intact uterus. The trial was terminated early when the incidence of breast cancer reached a preset safety limit in the original protocol design.  The written report included the following major findings:

Women's Health Initiative Study Outcomes  

Outcomes Hazard Ratio Increased Risk in 10,000 Women Taking Prempro for 1 Year
Coronary heart disease (CHD) +29% 7 more CHD events
Stroke +41% 8 more strokes
Venous thromboembolism (VTE) (blood clots) +111% 18 more VTEs
Breast cancer +26% 8 more invasive breast cancers
Colorectal cancer -37% 6 fewer colorectal cancers
Hip fracture -34% 5 fewer hip fractures






Adapted from Writing Group for the Women's Health Initiative Investigators. JAMA. 2002; 288:321.

With a mean observational period of 5.2 years, the global index, a composite risk assessment including the 7 major outcomes of the study, yielded a 15% excess risk in the Hormone Therapy (HT) users.

Those who attended the NIH conference heard some additional clarification of the data and some new insights   into the design, conduct and outcomes.    Diana B. Petitti, M.D. discussed the safety oversight of the WHI. She reviewed the role of Institutional Review Boards (IRBs), Informed Consent, and Data Safety and Monitoring Boards (DSMBs) in clinical trials.  She pointed out that within this study there were 41 IRBs. The DSMB consisted of 12 people who met every 6 mo from 1997 forward.  The safety monitoring issues were cardiovascular disease, (CVD), breast cancer, and the global index.  The increase in stroke was not anticipated.  The rate of deep venous thrombosis was greater than expected. The breast cancer risk showed up sooner than expected, and the anticipated CVD benefit became an adverse event.  Overall the Global index was calculated to be a negative outcome.  Marcia Stefanick, Ph D discussed levels of statistical power set for outcomes for harm and benefit.  The study design anticipated adverse events (AEs) to decline or disappear at 2-year mark based on numbers from the HERS trial, but in point of fact, excess AEs continued through 4 years.

Back to Top

Garnet Anderson, Ph D, principle investigator Fred Hutchinson Cancer Center in Seattle discussed the statistical methods used in the WHI.  She discussed the methods used to predict event rates and the weighting system used in calculations.  Events were weighted for determining the z score and were based on the time of the event since randomization. When expecting a reduction in CV events and fractures, adverse events in the earlier years of observation were weighted less heavily, assuming the treatment affect disease rates.  Events during the first 3 years received on ½ weighting, while those after 3 years received full weight.  Breast cancers were expected to occur at 9 or 10 years and received full wt at that time point.  Cancers that occurred at 4,5,6 years received half weighting when determining the z score.  Despite only 1/2 the weighting applied to early breast cancers, the z score for breast cancer risk rapidly passed the preset z score and thus required early discontinuation of the trial.   The preset z score for breast cancer risk was  -2.32, and actual score at the time the Data Safety Monitoring Board met was -3.19.  If early breast cancers had been fully weighted, the trial would have been stopped even earlier.  While the hazard ratio of 1.26 had confidence intervals from 1.00-1.59, they are nonetheless significant, given that this trial was a prevention trial, and threshold for significant harm lowered, according to a comment by Dr. Elizabeth Barrett Connor. 

Marian Limacher, MD from University of Florida discussed cardiovascular outcomes. She acknowledged that researchers and clinicians had every reason to believe that HT was cardioprotective. HT appeared to meet all the criteria for proof of a hypothesis on causality.  Animal models and physiologic measures offered support.  The construct was biologically plausible and the epidemiological evidence was extensive showing a strong correlation across different study populations, all with similar risk reduction. She then presented the age-stratified data on CVD.

The WHI has been severely criticized for having a mean age at entry of 63 years of age.  The study was specifically constructed and weighted with 1/3 of the population under 60 and 2/3 over age 60.   In contradiction to the speculations by critics of the WHI, the greatest increase in cardiovascular risk was seen in the youngest women.  The hazard ratios by age were:

50-59 years   RR= 1.67
60-69 years   RR= 1.26
70-79 years   RR= 1.18 

Interestingly,  the rate of CVD events  in women over 60, approximately –0.7% - was only 1/10th   the rate for women in the general population, where the expected incidence of CV events would be 7-8%. 

Rebecca D. Jackson, M.D. from Ohio State University discussed the osteoporosis data at length. The population within the WHI, using WHO T score for diagnosis of osteopenia and osteoporosis was stratified within the age ranges 50 to 65 and 65 to 79.  The percentages for normal, osteopenia and osteoporosis where 60.8, 37.1 and 2.1 % for the younger group, and  39.7,   50.9, and 9.4% for the older group of women.   While the data did not show statistically significant correlations, there were two trends identified.  The reductions in hip fracture were greatest in the women over age 70 and in women with a body mass index less than 25.  Women with BMI over 30 had no hip fracture benefit from HT.

Back to Top

Some interesting messages were:

  1. The confidence intervals for the sensitivity analysis of compliant women were presented.   For the group of women who had been 80% compliant with HT the  risks were higher than in the intent to treat analysis and were all statistically significant. 

            CHD                 RR= 1.51 (CI 1.13-2.01) 
            Stroke               RR= 1.67 (CI 1.17-2.40) 
            VTE                  RR= 3.29 (CI 2.25-4.82) 
            Breast Cancer   RR= 1.49 (CI 1.10-2.02)
  2. Further information and analysis will be forthcoming regarding women who were 100% compliant with medication.  
  3. Dr. Otis Brawley, a gynecologic oncologist from Emory University provided some perspective on the seemingly small amount of excess events in the WHI.  While HT use resulted in only 8 extra cases of breast cancer per 10,000 women years of use,  the number needs to be put into some perspective, relative to other health outcomes.  The number need to treat to prevent one case of breast cancer with mammographic screening between ages 50-70 is 17,135 screens per year of life saved.  For women age 40-49,  25,088 screens need to be done per year to save one life. Clearly the risk incurred with HT is not trivial compared benefits of screening.
  4. Deborah Grady, MD from University of California San Francisco designed a model for individual risk assessment utilizing the WHI data.  The model included relative risk or hazard ratio, absolute risk, absolute risk reduction and number needed to treat.  Some illustrative examples of cases showed that in most clinical situations,  the risks of HT, even for  a woman at high-risk osteoporosis, would outweigh the benefits.

The public comments were wide ranging, leveling both praise and criticism of the WHI.  Most of the critical comments centered on the age of the participants.  Given that the mean age at enrollment was  63, and given that most of the women were not newly menopausal,  the value of the study as a primary prevention trial was repeatedly questioned.  The second and perhaps harshest criticism centered on the manner in which the initial results were announced and the media coverage that followed.  Some  attendees felt that the medical community did not receive adequate information prior to public release of the findings,  and therefore was ill prepared to handle the onslaught of patient calls that followed.  The WHI investigators explained  that the timing of the notice to research subjects through study sites presented  great logistical challenge.  The WHI Data Safety Monitoring Board and the investigators had to balance patient safety issues with publics need to know.  Any attempt to inform professionals before the participants would have undermined the paramount ethical dictate of the study protocol – the safety of the study subjects. 

The WHI writing group is current working to produce over 20 publications, presenting more detailed analysis of the published data and additional material from other trials within the WHI.  Additional  information  will include: 

  • Early versus late cardiac disease in the trial of HT
  • Prior illness and biomarkers in the women with adverse outcomes in the HT arm of the study
  • WHI memory study
  • WHI cognitive assessment
  • WHI sight exam (macular degeneration)
  • Gall bladder and other biliary disease report
  • Osteoarthritis and joint replacement assessment
  • Genitourinary effects and sexual function
  • Physical function (healthy aging study)

This Web page contains links to miscellaneous ASRM announcements that do not fall into a typical publication category (e.g., Press Releases or News Briefs).

Also see: FDA announcements and NIH announcements

Back to Top