A Computer Decision Analysis About Postmenopausal Hormone Therapy
ABSTRACT & COMMENTARY
Col and colleagues from Tufts University School of Medicine, the University of Massachusetts Medical Center, and the Fallon Health Care System present a computer-generated decision analysis regarding the impact of postmenopausal hormone therapy on life expectancy. This was presented at the 19th annual meeting of the Society of General Internal Medicine, where it was given a distinguished research award. In this analysis, two hypothetical cohorts of identical healthy 50-year-old postmenopausal women were followed over time (of course by computer simulation) with only one cohort receiving hormone therapy. The computer simulation continued until all women in each cohort died. Simulation estimates for life expectancy and lifetime disease incidence for each cohort were based upon data derived from published epidemiologic studies. In this simulation, the hormone therapy assessed was combination estrogen and progestineither a sequential method or a daily continuous combinationand unopposed estrogen alone in hysterectomized women.
For the purposes of this simulation, Col et al assumed that current users of estrogen would have a 40% reduced risk of coronary heart disease, an increasing risk (up to a 46% increased risk after 5 years) of breast cancer, and a 54% reduced risk of hip fractures. Simulation projected that postmenopausal hormone therapy would increase the life expectancy of most women with gains up to 41 months (15% gain) for women at the greatest risk for coronary heart disease and the lowest risk for breast cancer. The only women not expected to increase life expectancy were those who were at the lowest risk for coronary heart disease and at the highest risk for breast cancer. Col et al present a diagram which, based upon their assumptions, would identify women who can expect to gain at least six months of life expectancy from hormone therapy. Not surprisingly, the greater the risk for developing coronary heart disease, the greater the gain in life expectancy with hormone therapy. The analysis was extended to data derived from black women, and it was concluded that black women could expect 10-25% less gain in life expectancy compared to white women. Col et al conclude that most women would live longer using postmenopausal hormone therapy. (Col NF, et al. JAMA 1997;277:1140-1147.)
COMMENT BY LEON SPEROFF, MD
There are no surprises in this computer simulation. Clinicians familiar with the literature regarding postmenopausal hormone therapy understand the major impact that one can expect in preventing coronary heart disease and osteoporosis-related fractures. Indeed, as Col et al point out in their report, they deliberately selected data that would underestimate the benefits of postmenopausal hormone therapy in an effort to avoid biasing the results in favor of hormone therapy. Thus, I expect that long-term current use of postmenopausal hormone therapy has an even greater impact than indicated by this computer modeling.
I have to quarrel with the choices made by Col et al regarding breast cancer. Col et al state that most well-designed studies examining long-term estrogen use have found a small increase in risk. However, they then refer only to those studies (with one exception) that have concluded there is an increased risk. Most importantly, they rely heavily on the Nurses’ Health Study. They further refer to the fact that their conclusion of an overall 30% increased risk was similar to that found in previous meta-analyses. However, they refer to the two meta-analyses that concluded that there was an increased risk and ignored the three meta-analyses that found no increased risk. Once again, this is probably an effort to avoid biasing the conclusions in favor of hormone therapy. Nevertheless, I am disappointed because this high- visibility report will support the argument that long-term use of hormone therapy increases the risk of breast cancer.
Another important point emphasizes the underestimation of the impact of hormone therapy. Col et al conclude that hormone therapy would not affect survival after patients have had coronary heart disease diagnosed. There is growing evidence that this is not the case. Col et al state that this is inconclusive, but they refer to Jay Sullivan’s data, which are not inconclusive but dramatically demonstrate greater survival in the estrogen users. (Sullivan JM, et al. Arch Intern Med 1990;150:2557-2562.)
An important observation emerges from this computer simulation. This effect on life expectancy can be expected only with long-term if not life-long postmenopausal hormone therapy. Half the gain in the computer modeling was not achieved until after 10 years of treatment, and it took 20 years of treatment to achieve 75% of the gain.
In their discussion, the authors point out that even with this pessimistic view of breast cancer, patients who have only one risk factor for coronary heart disease can expect the balance to be tipped in favor of hormone therapy. The authors believe that their model provides a practical approach for physicians to help patients make decisions regarding therapy. I’m not convinced that this works in real life.
In my experience, any suggestion that a treatment may cause breast cancer is a message that has a tremendous impact on the patient and becomes an obstruction to the patient’s hearing the whole story. I would much rather use the computers in our brains and our patients’ brains to assimilate information and make decisions than to use this mathematical and relatively inflexible model. Indeed, as objective as this model appears to be, it is based upon assumptions made by Col et al that may not be true. This is especially the case regarding breast cancer.
The impact of this article will be on physicians. A sentence stands out in the authors’ discussion. They estimate that less than 1% of healthy perimenopausal women would fail to benefit from hormone therapy. For those physicians who continue to be skeptical that postmenopausal hormone therapy offers important preventive health care benefits, this computer model provides a convincing argument in favor of hormone therapy.