Current evidence suggests that postmenopausal hormone replacement therapy (HRT) decreases morbidity and mortality from coronary disease (CAD) and osteoporosis, while increasing morbidity and mortality from breast and uterine cancer. The overall risk/benefit ratio associated with HRT for a specific individual will depend on that individual's a priori risks for these conditions. Thus, a woman at high risk for CAD and osteoporosis and at low risk for breast and uterine cancer would be more likely to benefit from HRT than a woman with an opposite risk profile.
The best way to estimate risks and benefits of a therapy is to make use of the results of randomized clinical trials, but these results are not yet available for HRT. This study presents a computer model designed to estimate the effect on life expectancy of HRT in women, according to their risks for coronary disease and breast cancer.
In order to model the disease processes being studied here, the authors used a Markov state transition model. Markov analysis is a way of analyzing complex systems. It assumes that the system can be described by a number of variables that can each be in one of several "states", and that these variables make transitions from one state to another at discrete time intervals and with probabilities that depend on the values assumed by other variables. When analyzing clinical situations, the appropriate selection of variables and the determination of probabilities governing the transitions from one state to another (such as from the "free of coronary disease" state to the "coronary disease" state) are key issues.
The model considers two initially healthy cohorts of 50 year-old women, one cohort taking HRT and one cohort not taking it. It was assumed that women with an intact uterus would take combined estrogen/progesterone, whereas women without a uterus would take unopposed estrogen (thus, HRT should not increase the risk for endometrial cancer, and this cancer was not explicitly modelled).
The model assumes that, each year, women are at risk for developing breast cancer, coronary disease and hip fractures. Once a woman has developed one (or more) of these conditions, she is at risk for death from that condition. Women are also at risk for dying from other causes. It is assumed that the effect of HRT is primarily on the likelihood of developing breast cancer, CAD or an osteoporotic hip fracture, not on the likelihood of dying from it once such an event has occurred.
The model is "run" until all women have died, and the life expectancies and disease incidences in the two cohorts are then compared. In order to investigate the effect of a priori risk on outcomes, the model is run in cohorts of women with different risk factor profiles.
When determining the overall effect of HRT on mortality, it is necessary to take into account the fact that breast cancer, CAD and hip fractures occur at different ages. The authors did this by calculating the average life-expectancy that a 50 year old woman would have, according to this model, if she were or were not taking HRT. This comparison was calculated for women with varying combinations of risk factors for CAD, breast cancer and hip fracture.
As expected, HRT provides most benefit (in terms of life expectancy gained) for women with more risk factors for CAD and fewer risk factors for breast cancer. These results are illustrated in the article with a 3-dimensional graph. The model predicts that, among women without risk factors for hip fracture, in the low/lowest risk categories for CAD who are also at an increased risk for breast cancer, HRT produces a shortening in life expectancy. For the rest (the majority), however, HRT produces an increase in life expectancy, up to a maximum of 41 months (depending on risk factor profile).
The number of incident cases of CAD or hip fracture averted for every case of breast cancer caused is also graphed. This number approaches 9:1 in women at highest risk for CAD and lowest risk for breast cancer.
The authors present two possible methods for applying their results to specific patients. The first is a decision tree which helps determine whether or not a woman would gain at least 6 months of added life expectancy, based on certain risk factors. The second is a graph that helps determine whether the gain in life expectancy from HRT would be 0, 3, 6 or 12 months in women with certain combinations of risk factors. Of note, the estimated advantage for HRT is 10% to 25% less for black women.
A sensitivity analysis was performed, in order to determine how sensitive the model was to changes in the main assumptions and estimates. There were no major changes in the results when most of the parameters were varied.
The authors make a number of important points about their study:
This study was based on data from non-randomized studies, so patients on HRT could have been healthier. Studies that attempted to control for this potential bias found similar results, however.
The authors attempted to be conservative in their assumptions, thus biasing their results against HRT and strengthening their conclusion.
They believe their results indicate that HRT may be indicated for a broader range of women than has previously been advocated (all except those at high risk for breast cancer and without risk factors for CAD, which they estimate to be around 1% only). They note their results compare favorably with the results for smoking cessation and cholesterol lowering, in terms of added life expectancy.
They also note that their model does not take into account patient preferences and issues of quality of life, which may tip the balance in one direction or another, particularly in cases where the added life expectancy from HRT is expected to be small.
It must be reiterated that this is not a randomized trial, or even a cohort study. It is a computer model of the anticipated gain in life expectancy associated with HRT, based on data from other, non-randomized trials. Nevertheless, the results certainly seem reasonable.
One point should be made. The end-point of increased life-expectancy is a combined one, and can hide some important information. A year of life lost to coronary artery disease is not the same as a year of life lost to breast cancer. Not because of the difference between the diseases (which the authors allude to when they discuss quality of life issues), but because of the difference in age of onset of these two diseases.
To take a fictitious but not implausible example: Assume, for a given set of risk factors, HRT is expected to prolong life by one year and, for those same risk factors, there would be 5 fewer patients developing coronary disease for every one developing breast cancer. We might prevent 5 deaths at age 70 of coronary disease in women who would otherwise die at age 75 (net gain: 25 years) while causing one death at age 55 from breast cancer, in a woman who would otherwise have died at age 79 (net loss: 24 years).
The exact numbers are not important. The principle remains: since coronary disease tends to strike at a later age than breast cancer, and since the incidence of coronary disease is higher than the incidence of breast cancer, the "average" gain in life expectancy may be the result of prolonging several lives by a few years at higher ages, balanced against decreasing one life by a greater number of years at a younger age.
The type of model described here could be adapted to give information on this specific issue (by reporting the average number of years of life lost to breast cancer or CAD in both groups). But even then, interpretation will remain subjective. Furthermore, we need to take into account the fact that HRT produces other benefits, including quality of life benefits.
It is hard enough to come up with valid numbers, in the absence of randomized trials. When the results of these trials are in, their interpretation will still leave room for controversy and many editorials.
April 29, 1997
It must be emphasized that your provocative observation about the amount of years gained or lost respectively for CAD and cancer applies for prophylactic use of ERT.
In women with established CAD (and older, in their vast majority) the benefits derived from ERT may largely overcome any of its risks. Even though this may seem obvious, there is still a lot of reluctance among doctors (and patients), to prescribe ERT for women suffering from CAD. There is not yet clear evidence that this approach reduces mortality in such women, but the results of pooled analyses suggest a reduction of more than 50% in "hard" events.
Submit a comment about this article
Return to Journal Club home page