There is no more confusing medical decision for women than whether to use hormone replacement therapy (HRT). The letters and phone calls on the subject are endless.
I hesitate to tackle this one, but it's a good example of a larger issue: Are medical studies asking the right questions?
Estrogen replacement therapy was first available in 1942. By now its effects should be obvious, yet it continues to be controversial. Is the problem a lack of sufficient information, or is there something inherently unresolvable at the heart of the estrogen story?
I've done my reading and talked to the experts, yet remain confused and uncertain. My politically incorrect, shoot-from-the-hip conclusion is that more studies aren't going to solve the problem. The controversy is here to stay because the problem is in the nature of medical studies in general. Questions for which there are few confounding factors and absolute, objective measures are relatively simple to answer. But once there are multiple interacting variables or the effects being measured are not entirely objective, all bets are off.
Let's take a look at HRT and see what is and isn't possible to understand. Estrogen helps prevent osteoporosis. We know this because we have a specific measure of bone density and can compare post-menopausal women who take and don't take HRT. We can argue about measurements of bone density, but these are minor quibbles that can be resolved.
Similarly, we know that estrogen reduces the incidence of cardiovascular disease. Using objective tools such as EKGs, cardiac enzymes and MRI scans, we can objectively determine the incidence of heart attacks and strokes.
But what about HRT and breast cancer? Breast tissue contains estrogen receptors; estrogen stimulates cell growth. Most breast-cancer cells (perhaps two-thirds) are estrogen-receptor positive (respond to estrogen). There is a statistically increased incidence of breast cancer in patients on HRT. Tamoxifen, which blocks the effects of estrogen on breast tissue, is a major advance in chemoprevention of breast cancer, though not without significant side effects, including an increased incidence of endometrial cancer.
So isn't estrogen bad for women at risk of, or who've had, breast cancer? It's not that simple. High-dose estrogen (stilbestrol) was once a treatment for metastatic breast cancer, with large tumors undergoing dramatic, though temporary, reduction when exposed to estrogen. Several recent studies have shown that women taking HRT when their breast cancer was diagnosed had a better prognosis than those not on HRT. (Many doctors in the U.K. now feel that HRT is to be recommended for patients with a history of breast cancer.)
My first impulse is to wonder if some studies are better than others, or whether some overarching study might be able to answer the question. But I stop myself. The problem may lie elsewhere, in hidden possibilities and the unpredictability of interacting influences.
Though we are great at planning and calculating, we all recognize the profound role of random or inconsequential events -- the missed planes, serendipitous meetings and odd mistakes -- in shaping our lives. Such events, though we try to control them, also shape the outcome of medical studies.
A weather forecaster understands this dilemma, shrugging when we ask if it's going to rain tomorrow. "Maybe. I'd say an 80 percent chance." Because we can cover our bases by carrying an umbrella and a sun hat, we do not fret over the limitations of the weather report. Not so with medical studies. Too much is at stake and we need to feel in control of our health, so we act as though medical problems have more definite solutions than weather predictions.
(If you think that, with a proper design, all medical questions can be answered, consider the field of nutrition. Every week there's a new study with different recommendations. I confess that it's beyond me how to determine what is the optimal combination of vitamins, minerals, herbal supplements and fat and fiber content. It would be like predicting world weather for the next millennium.)
Which leaves women with this impossible question: If you are at risk for breast cancer, would you prefer to have a statistically greater chance of developing the disease or possibly a better prognosis if you already have it? Are the two statements in any way comparable or is this apples versus bananas? And how would we know?
Even more confusing is the relationship of estrogen to memory. At least with breast cancer, one can construct studies in which there are objective endpoints such as tumor recurrence or death. But what are we measuring when we study memory?
Neuropsychologists act as though memory is an entity that can be quantified via testing. Down deep we all know differently. Memory isn't a white-blood-cell count or an X-ray. Anyone who has given any thought to the false memory/recovered memory controversy understands the slippery nature of the very process of remembering. Anyone who has taken speed or much caffeine or stayed up all night cramming for an exam or had major test anxiety understands how memory itself is affected by a wide variety of biological, psychological and environmental factors.
Compare the following: Psychologists at Montreal's McGill University have recently concluded that estrogen maintains verbal memory in women, may prevent or forestall the deterioration in memory that occurs with normal aging and may decrease the incidence of Alzheimer's disease and/or retard its onset.
Conversely, doctors from the Yale University department of internal medicine and a group of Finnish doctors feel that there is inadequate evidence to say that HRT improves cognitive function in post-menopausal women or women with Alzheimer's disease.
Is one study is better than the other? What should we conclude? A study from Holland may shed some light on the problem. Sixty-two healthy post-menopausal women were given either HRT or a placebo and were told the purpose of the study. Those given HRT reported better sleep patterns as well as improved psychological and memory function than before taking the HRT, yet didn't do better on memory testing than did the controls (the women who didn't get the estrogen). Then the researchers repeated the study, but without telling the women the purpose of the experiment. This time no positive effects of HRT could be detected.
There are many possible interpretations, but one jumps out at me. Estrogen therapy can't be given blindly -- there are too many physiological changes that are easily detectable by the subjects, such as relief from hot flashes and/or vaginal dryness. So, in the first part of the study, the subjects knew what they were taking and what the researchers were looking for. They felt better (though not measurably better than the controls). When they knew what they were taking, but not what the researchers were looking for, however, they didn't feel better.
Is the difference between the various studies I've cited due to a placebo effect, merely a reflection of how the researchers asked the questions, what kind of perfume the head nurse was wearing or whether free coffee and doughnuts were passed out along with the questionnaires? Were the subjects subliminally prodded toward the proper responses?
Besides the notoriously difficult-to-analyze placebo effect, I wonder how those in the control group felt about not getting estrogen, about having more hot flashes than those in the other group? Was the control group resentful, glad not to be taking estrogen, pleased to be part of a study, indifferent or a combination of the above?
What if one member of the control group read an Internet article about some new negative effects of estrogen and passed the word to the others in the group? If expectation influences results, what role does control-group psychology play? Is there such a thing as a control-group effect? And how would you know? Certainly not by comparing control groups -- which would result in an infinite regression.
Surely there must be some shred of objectivity we can bring to the question.
A Yale pediatrician, Sally E. Shaywitz, has used functional magnetic resonance imaging to compare brain patterns of post-menopausal women taking estrogen supplements and those on a placebo. While on estrogen, the women showed brain patterns similar to those seen in functional MRI scans of young people. The areas of the brain that lighted up more extensively than predicted correspond to regions of the brain known to be rich in estrogen receptors.
Aha, you might think -- at last a definitive objective measure of increased brain activity when administering estrogen. So far, so good. But Shaywitz made the mistake of testing the patients to see if they also performed better. Unfortunately, they didn't. Objective evidence of brain activation wasn't accompanied by any detectable change in memory function. Her conclusion: More subjects needed to be studied. Nevertheless, on April 7, 1999, the Associated Press headline about the study read: "Estrogen may improve memory in post-menopausal women" (a true but potentially misleading statement).
This problem of subjectivity extends to much of medicine, whether one is analyzing treatments for chronic pain, depression, social phobias, chronic fatigue or even the effects of zinc lozenges on the common cold. Symptom relief makes a lousy scientific endpoint.
We are stuck. We need a new way of asking medical questions and we need a new way of reporting results. For starters, I'd suggest that each medical article clarify whether the effects being studied were objectively measured (bone density) or were based upon a subjective endpoint (changes in memory, reduction in back pain, feelings of well-being). The article should address the very knotty problem of whether the question being asked can be answered. As consumers, we need to be comfortable with these distinctions, to know when an article represents a true scientific determination and when it simply represents subjective opinion. Only then can medical news avoid the horrible to-and-fro swings in opinion that do nothing but enhance our already considerable health anxieties.
I think of the judges at an Olympic diving competition. Each holds up two score cards -- one for technical difficulty and one for artistic merit. Maybe each new medical study should be similarly judged. The connection between estrogen and bone density is a simple question. That between estrogen and memory is profoundly more difficult. We should have different expectations, depending upon what is possible.
And we should know what parameters scientists are using. An old story illustrates this: A scientist is seen crawling around under a streetlight. A bystander walks up to him and asks what he's doing. The scientist says that he's looking for his keys. The bystander asks him where he lost them, and the scientist points into the darkness beyond the pool of light from the street lamp. "So, why don't you look out there?" the bystander asks. "Because I can see here," the scientist says, continuing to search in the pool of light.
It is scientists' obligation to tell us why they are looking where they are. It is up to us to know whether the keys can be found where the scientists are looking. Meanwhile, here is an interesting, well-balanced Web site dealing with HRT.