Stanley Feld M.D. FACP.MACE
This blog entry is dedicated to Joseph Goldzieher M.D. one of the giants in Reproductive Endocrinology.
The Women’s Health Initiative is perhaps the most famous example of a study that was released to the press before the data was published in a medical journal. The conclusions of the study changed forever the way peri-menopausal and post menopausal women are treated.
The pre-published press conclusion was that estrogen causes heart disease, breast cancer, stroke and pulmonary embolism. The media announced the WHI’s findings before the medical profession had a chance to study the data.
The conclusions frightened every peri-menopausal and post menopausal woman in this country. Over the years observational data supported the conclusion that estrogen was of great value in treating symptoms associated with the acute menopausal syndrome, namely hot flashes, vaginal dryness, urinary tract irritation, skin changes and emotional instability. Estrogen also seemed to protect against heart disease, osteoporosis, weight gain and promote a general sense of well being. There was no good evidence for or against breast cancer.
This NIH sponsored double blind placebo controlled study (WHI) was performed to develop proof with a level A(double blind placebo controlled) study to test the validity of observational data reports of estrogen effects. The WHI reported results that concluded the opposite effects of estrogen reported by many observational studies. The WHI conclusions were that conjugated estrogen caused breast cancer, heart disease, stroke, and pulmonary embolism. The WHI claimed that conjugated estrogen did protect against osteoporosis. You will recall the media is the message and these were the results the media frenzy reported.
Prior to release of the study results many women were afraid to take estrogen on general principles alone. Many felt that estrogen deficiency was part of the aging process. However, women had a life expectancy of 50 years in the early part of the 20th century. Women today live much longer and observational data suggests healthier, as a result of estrogen therapy.
There are many problems with the WHI study that have not been discussed in the popular press. These problems have not been discussed in the medical literature either. Practicing physicians were confused and enraged by the WHI study results and the manner in which they were presented. Patients taking estrogen were angry at their physicians.
There are many defects in the study from a statistical point of view.
1. Age Distribution: 66.6% of the patients were between 60 and 70 years old. 87% of the patients were 60 to 80 years old. The majority of the patients receiving Hormone Replacement Therapy (HRT) for the first time were at least10 years post-menopausal. This age group population does not represent the usual population for starting HRT. HRT is usually started just prior to the onset of menopause or at menopause (48 years old).
2. The drop out rate in the placebo and HRT group was 40%. The impact of the dropout rate was not addressed in the validity of statistical analysis section of the paper. Maximal tolerable dropout rate should not be greater than 20% in a statistically significant protocol.
3. The unblinding of 3000 women represents a departure from the protocol and biased the findings of treatment difference.
4. A hazards ratio (HR) should be greater than 2 in order to have for a result to have convincing difference and should not be expressed to two decimal places. A hazards ratio of less than two can not discriminate causality from bias and confounding of variables.
5. Power of the study was disrupted by the 40% drop out rate. The study was not sufficiently powered to have significant results
6. The traditional approach to presenting a nominal confidence interval is valid when one outcome is being studied against a placebo. Adjusted confidence intervals must be used when studying multiple outcomes with multiple confounding variables.
The WHI’s conclusions were based on the use nominal confidence intervals. The nominal confidence intervals were significant. However they came close to touching the magic number one (1). All of their published adjusted confidence intervals were non significant because they crossed 1.
CHD HR 1.29 Nom CI (1.02-1.63) Adj CI 0.85-1.97
Breast cancer HR 1.26 Nom CI(1.00-1.59) Adj CI 0.83-1.92
Stroke HR 1.41 Nom CI (1.07-1.85) Adj CI 0.86-2.31
PE HR 2.13 Nom CI(1.39-3.25) Adj CI 0.99-4.56
The adjusted confidence intervals were published in the original paper.
Media blitz publicity of the results of the study created a high level of certainty for the results of the study in the public’s mind prior to any peer discussion of the data or the weaknesses in the data. Few physicians were in a position to dispute the statistical weakness of the data. The results the media reported were to change forever the way physicians practice medicine for menopausal women. In my view, the results led to a great disservice to women. The publicity also had a devastating impact on the physician patient relationships and the patient confidence in clinical research.
The estrogen only leg of the study showed no significant difference in breast cancer or heart disease. These results and the facts related to the result was less publicized by the media.
The conclusions of the data should have been that the study results were not related to the combination of conjugated estrogen and progesterone in PremPro or the conjugated estrogen alone in Premarin. Even though estrogen might cause heart disease, pulmonary embolism, stroke, and breast cancer, the Women’s Health Initiative did not have the statistically significant evidence to prove it. Once again media published conclusions disrupted the therapy regime of millions of patients as well as their confidence in their physicians. Once again, physicians contributed to the dysfunction of the healthcare system.
Freedom of the press is vital to our freedom of speech, but manipulation of the media’s tendency to sensationalize issues prior to proper judgment is not helpful.