I have expanded the paper about mortality in WHI (Limitations and Misuse of Multicenter Trials: A Critique of The Women's Health Initiative (WHI) Study (II) beyond my original intent into two papers. [see WHI- Better Survival? 10/2004 for the first part]
In the meantime there is new weight to the WHI critiques, and the following are excerpts from recent ones. The statistical techniques used in WHI are complex and need more evaluation. Conclusions coming out from WHI (and there are many papers) need to be "taken with a grain of salt". Call me in the morning.
Several points were emphasised in a paper by William Creasman September 2003.
WHI: Now that the dust has settled William T. Creasman, M.D.
American Journal of Obstetrics and Gynecology, 2003Sep;189(3)621-6
"One arm of the WHI study was lauded as a prospective, double-blind, randomized clinical trial... terms have been used to describe this study, such as "vigorously designed," robust,'" and "statistically valid." Therefore, the WHI results have been widely accepted without careful analysis... In our review of WHI study results, we are troubled by many questions and concerns."
Coronary Heart Disease (CHD)
"A recent review by the US Preventive Task Force of all published reports of women who are current HRT users showed a significant decrease in CHD incidence [(relative risk RR=.68-.95), mortality (RR=.40-.60), overall cardio-vascular mortality RR=.44-.93]. The HERS study suggested that if a woman was on estrogen and sustained an MI she was less likely to die from her CHD than those not on estrogen."
"The excess occurrence of breast cancer was given as the reason the WHI study was terminated. Yet, when the data were reviewed, the hazard ratio of 1.26 is not statistically significant, indicating the finding could be simply due to chance... There are no data to suggest that estrogen causes normal breast tissues to become cancers... [and] there was no increase in the diagnosis of carcinoma in situ" which would be expected if E+P increased breast cancer risk.
"Venous thromboembolism (VTE) was the only item that retained its statistical significance between the E+P and placebo arms at both the nominal and the adjusted levels. Actually, when DVT and PE are considered separately, only DVT remains statistically significant at the adjusted level. Diagnosis of PE and DVT required clinical symptoms supported by relevant diagnostic studies. What relevant study? The diagnosis of PE and DVT based on clinical symptoms is very unreliable... The WHI study showed a 10-fold increase in the baseline incidence compared with commonly quoted figures."
"It would appear that the mortality benefit to colorectal cancer protection is greater than the increased risk in breast cancer. The committee that closed the study apparently ignored this fact... The trial was stopped for breast cancer with seemingly no consideration of the benefits to colorectal cancer."
"There is an inconsistency as to incorporating risk factors in the Cox model but not the stopping rule: the stopping rule uses a log rank statistical test that typically does not incorporate covariables. The Cox regression, which estimates the hazard ratio, uses covariates. Normally the results from a log rank and a Cox proportional hazard model would give similar results. Why are these so different?"
"To publish data that may or may not be entirely true or certainly premature is a disservice to the medical profession and, most important, to our patients. The majority of the data that were published is not statistically significant even at the nominal level. For several decades, data from univariate analysis that have not remained significant in multivariate analysis [ie., corrections to confidence intervals for multiple testing] have been disregarded in managing patients. Because much of the data presented in the WHI study falls into this category, why the big change to modify or change practice?... Data in which the difference could be due to chance alone do not satisfy this time-tested edict.
The final publications on this study that may satisfy our many concerns could well pass medical scrutiny. Until then, this, as well as other like publications, should be taken with a grain of salt."
Leon Speroff wrote an editorial that introduced the paper:
WHI trial: It's time to be critical
Leon Speroff, M.D.
American Journal of Obstetrics and Gynecology, 2003Sep;189(3)620
"The challenge is to identify the reasons for the disagreement between the results of the WHI trial and the previous literature. It is too simplistic to merely accept the results of the WHI trial because it is a randomized clinical trial... It is appropriate (indeed essential for the care of our patients) to be critical and to excercise our medical judgement."
*Risks and Benefits of Estrogen Plus Progestin in Healthy
Postmenopausal Women: Principle Results from the Women's
Health Initiative (WHI) Randomized Controlled Study,
Writing Group for the Women's Health Initiative (WHI) Investigators
JAMA July 17, 2002 288(3):p321-349