This speech was delivered on November 6, 1999, before the 22nd annual convention of the Freedom From Religion Foundation, St. Anthony Hotel, San Antonio.
Richard Sloan is the lead author of a recent article in the prestigious medical journal Lancet cautioning physicians not to prescribe religion as medicine.
Thank you very much for inviting me to join this group of risk-takers. I was unaware of how much a group of risk-takers you were. It’s one thing to support atheism–but that breakfast! I mean, bacon, eggs, cheese, and fried tomatoes!
I want to tell you a little bit about why I became interested in this, and what we’ve done. You can’t go a week without finding some article in the popular press or in the popular broadcast media about a new study demonstrating that religious activity promotes health. Sometimes it’s mental health outcomes, sometimes it’s physical health outcomes. But it’s ubiquitous in the press.
It’s an interesting question. Is it really possible that religious activity could be associated with better health outcomes? Inundated by this media deluge, my colleagues and I became interested in whether or not it was actually so, and so we started to look at the literature.
The literature is enormous. There is an enormous number of papers published in the scientific and medical literature in one way or another assessing the relationship between religious activity and health outcomes.
You will hear proponents of this point of view say there are thousands of studies, and three-quarters of them, as one of them says, indicate beneficial outcomes of religious activity. In fact, there are thousands of studies, but at least half of the studies don’t look at religion as a factor influencing health outcomes, but rather look at religion as a consequence of health conditions. So, for example, someone mentioned earlier the old folk wisdom, “There are no atheists in foxholes.” Well, in a medical foxhole, presumably that may be true as well. An enormous number of studies supposedly about religion and health actually look at whether people become more religious when confronted with health crises. Whether or not that’s so, I don’t know. We didn’t look at that literature.
What we were interested in was looking at the literature that purports to show that being religious or engaging in religious activities in one way or another promotes better health. Because there are still hundreds and hundreds of papers about this topic, we restricted our analysis only to cases in which physical disease outcomes–which are more easily measurable–were the supposed outcome of religious activities. I can’t speak to the literature on mental health outcomes because I haven’t reviewed that literature.
Interest in religion is widespread in this country. There are national newspaper articles all the time about the substantial fraction of people in the United States who profess a belief in God, and attend religious ceremonies on a regular basis. Within the medical community, there is also a considerable and growing interest. There are NIH-funded seminars and meetings sponsored by the National Institute of the Aging, which is one of the National Institutes of Health, on religion and spirituality and medicine. Other national medical meetings also have focused on this topic.
A survey conducted by the American Academy of Family Physicians in 1996 indicated, astonishingly, that 99% of the participants professed strong religious belief–these are physicians. And 75% of them believe that prayers of others could promote a patient’s health. Harvard cardiologist Herbert Benson, who is most noted for his work on the relaxation response, has written in one of his popular books, Timeless Healing, that humans are “wired for God.” That’s what he said, “wired for God,” meaning it is hardwired somewhere in the brain that we believe in God.
Physicians like Dale Matthews of Georgetown and David Larson, who has an adjunct appointment at Duke and is the director of a misleadingly named organization, the National Institute for Healthcare Research, have written that it is time for us to “tear down the wall of separation between the twin traditions of medicine and religion.” Matthews has gone even further to say, in an article in the New York Times magazine section about 18 months ago, that the future of medicine is “prayer and Prozac.” Matthews and his colleagues recommend asking patients who respond favorably to whether religion is helpful in handling their illness: “What can I do to help support your religious faith or commitment?” There is a concerted effort on the part of a relatively small number of well-funded researchers to bring religion into medicine.
To address this, you have to look at two aspects of this problem. One: You have to look at the state of empirical evidence–is the evidence good that religion is associated with better health outcomes? Two: Irrespective of the answer to that question, you have to consider the substantial ethical issues raised by bringing religious activity into medicine. We sought to do that.
To begin with the empirical evidence, you need to understand there are typically two kinds of studies done in medicine research. There are epidemiologic studies in which you look at large populations and attempt to determine associations between characteristics and outcomes. So, for example, you might look at the association between frequency of attendance at religious ceremonies and health outcomes. That’s done all the time in this literature.
The problem with that kind of study is that it takes people who are already inclined to be religious, and compares them to people who are not inclined to be religious. Those two groups of people may differ in fundamental ways, in addition to whether or not they’re religious or not. It may be that those other, fundamental ways in which they differ influence the outcome. Nonetheless, for example, the epidemiologic studies demonstrating that smoking is associated with increased rates of cancer and heart disease were central and fundamental to understanding how smoking affects health, and led to our current approach to smoking within medicine.
So epidemiological studies are very important, but without experimental studies of one sort or another, they are only part of the picture. By an experimental study, I mean taking a group of research subjects and randomly assigning them to different conditions. Typically, intervention trials in medicine take the form of experiments in which people are randomly assigned to either receive a medication or not receive a medication, to receive a surgical procedure or not receive a surgical procedure, receive certain kinds of advice or not receive certain kinds of advice. Then you examine the difference between those two groups after some interval to determine whether that treatment has an impact.
So, for example, you could, in the case of smoking cessation, conduct a very simple experiment. You could take a group of physicians and have them on a random basis, either give or not give their patients the following advice: “I, as your physician, recommend that you stop smoking, because the health consequences of smoking are very serious.” That’s the intervention.
You could determine whether patients who receive that intervention stop smoking at a higher rate than patients who don’t receive that intervention. That seems like a very simple thing, but in fact study after study has demonstrated that simply telling your patients that you recommend that they do not smoke doubles the quit rate. That’s an intervention. We can assume, because the groups are randomly assigned, that there’s no difference between the groups and that therefore, it is the “intervention” that has the impact.
The problem with most of the studies in the religion and health literature is that they are not intervention studies, rather they are epidemiologic studies which look at associations between variables, between church attendance, or frequency of prayer, or self-reported reading of the bible, and then health outcomes. Of course, the people who read the bible on a regular basis, or the people who report that they pray or go to religious activities on a regular basis, differ in that respect from people who don’t, but they may differ in many other respects. So it’s essential for these studies which purport to show relationships between religious activity and health to look for the other possible confounding variables that may account for the relationship.
Here is a hypothetical example of an egregious case of failing to account for confounders. There is no doubt that people who carry matches in their pockets develop lung cancer more frequently than people who don’t. [laughter] How many of you believe that it’s the matches? [more laughter] Obviously, there is another factor at work. Matches are associated with smoking cigarettes and it’s cigarettes that are the active agent.
We want to know if it is true that people who are regular attenders of religious services, for example, show increased longevity and better health. Is that what the operative factor is, or is there something else? What you must do as a methodologist is to attempt to identify other factors which may contribute to this outcome. In fact, most of the studies in the literature simply fail to do that. I don’t mean to be critical of studies that were conducted 20, 30 or 40 or more years ago, because our methodological standards today are different from those of an earlier era. But nonetheless, studies that fail to control for confounders and covariates cannot be taken–despite the fact that there are hundreds of them, maybe even thousands of them–as seriously as evidence that religion is associated with better health outcomes.
Here’s an example. In 1971, George Comstock, a very senior epidemiologist at Johns Hopkins, published a paper showing that attendance at church was associated with reduced mortality at a follow-up seven years later. This study is cited over and over by proponents of this position. What these proponents never report is that seven years later, in 1978, Comstock retracted that finding, on the following basis. He said that he failed to account for the fact that by looking at people who go to church and contrasting them with people who don’t go to church, he missed the effect of previous illness. That is, people who are already too sick, i.e., are functionally incapacitated, can’t go to church, and people who are already too sick die at a higher rate than people who aren’t so sick. So the effect of church attendance on mortality was entirely wiped out by considering functional status. Comstock publicly retracted this finding in a paper published in a major journal in 1978.
That 1978 paper is never ever cited by the proponents of this point of view, which leads me to another issue. There are differences between primary and secondary sources in science. A primary source is the account of the actual experiment, the paper itself. Then there are reports and reviews, which are the secondary sources. Problems exist in both arenas. In some of the primary sources, studies are very poorly conducted, and the reviews often are selective, even when people review their own work. It’s quite astonishing!
I recently reviewed a manuscript in this general area submitted to the journal Psychosomatic Medicine. One of the authors also was an author of a well-conducted paper published last year in the journal Demography. It showed that religious attendance was associated with reduced mortality. The paper I was reviewing asserted that this effect of religious attendance on health outcomes pertained to all causes of mortality. For example, it would pertain to heart disease, cancer, respiratory diseases, infectious diseases, diabetes, etc. But that wasn’t true, despite the fact the same author appeared on both papers. The author of the Demography paper had, in the manuscript under review, misreported his own findings, published only a few months earlier! The effect was barely true for cardiovascular diseases, wasn’t true at all for cancer, wasn’t true at all for diabetes. In fact, the strongest effect was only for respiratory diseases. This is a particularly graphic example of how secondary sources often misreport what is published in previous papers.
One of the most important things to consider when you hear reports of association between religious activity and health outcomes is to ask whether potential other variables–confounders and covariates–actually can account for the findings. In the best studies, the most recent studies with very large databases and thousands of subjects, you can statistically control for the possibility that other factors may account for the findings. When you do that, sometimes you see an effect, and sometimes you don’t. But there is by no means an overwhelming evidence base that religious activity is associated with better health outcomes. So one of the major problems in the literature is the failure to control for these other factors which may account for the findings.
Another problem in the literature is what we have referred to as the problem of multiple comparisons. Science operates not out of certainty, but out of probability. So when you conduct an experiment, when you report a finding, you indicate the degree to which you are certain that it is not attributable to chance, or the degree to which you are certain that the effect is due to what you say it’s due to. Science typically takes as its criterion 95% certainty. If you say that A is greater than B, it’s to the degree that you are 95% certain that A is greater than B. But every time you make a comparison, you increase the likelihood of discovering that A is greater than B merely by chance.
The group at Duke which publishes a lot in this area typically fails to address this problem. And in fact, they report that they fail to do this. For example, they published a study in 1997 in the International Journal of Psychiatry and Medicine, ranked as the 41st most prestigious journal within the field of psychiatry [laughter], but nonetheless gets a lot of press, showing that even after controlling for relative covariates and confounders, that subjects who regularly report going to church have lower levels of Interluken-6, an index of immune function. This finding was statistically significant at the 95% certainty level. But what the authors also report is that they measured eight different measures of immune function. Basically, what they were doing was fishing until they found something that was significant. Then they reported it. If you’re going to measure eight variables, then you have to adjust what constitutes an adequate level of statistical significance by the fact that you are making eight different comparisons, and you are increasing by eightfold the possibility of finding something significant purely by chance. They actually reported that they did this but the paper was published nevertheless. This is the second problem with the empirical literature.
The third problem is that the findings are all over the map. In some cases, you find that church attendance is associated with reduced mortality and increased longevity, and in other studies, equally large, equally well-conducted, you find that it’s not. If there is any regular effect, you should expect some consistency of result. Sometimes you find self-reported frequency of prayer is associated with beneficial outcomes and sometimes you don’t.
A paper recently published in the journal Psychosomatic Medicine showed that among elderly patients undergoing coronary artery bypass surgery, those who reported feeling increasing levels of strength and comfort survived three times longer than those who didn’t. That’s interesting. However, the authors also reported that church attendance did not predict, and frequency of bible-reading did not predict, and all the other indices of religious activity did not predict any differences in mortality. Only the finding about strength and comfort provided by religion made the popular press.
For there to be any basis whatsoever in making religious activity an adjunctive medical treatment, there has to be solid and consistent evidence of the sort you see from the smoking literature. We can demonstrate that smoking is associated with health outcomes both at the epidemiologic level but we can also do it at the tissue level, the animal level, the cellular level. We can show in many different ways how smoking is associated with poorer health outcomes.
We cannot do that with religion, because there is no consistency of findings and moreover we can’t conduct animal studies or cellular studies, because you can’t make monkeys religious or not religious. [laughter] And of course you cannot conduct an experiment because you cannot randomly say, “Okay, this group is going to become religious, and this group is not going to become religious.” So we are stuck with the problem that there are pre-existing differences between those who practice religion and those who do not.
Those are the three general problems of the scientific literature in the area: the failure to control for confounding variables; the failure to control for fishing for the data and making multiple comparisons, and the inconsistent findings.
I want to talk for a moment about the study that was published two weeks ago in the Archives of Internal Medicine on the impact of intercessory prayer in the coronary care unit. This is actually not an epidemiologic population study; it was a real experiment in which half of the patients in the coronary care unit in a Kansas City hospital were assigned to receive prayer from Christian intercessors, and half were assigned not to receive prayer from these intercessors. Patients did not know that they were receiving prayer, or even that they were in an experiment–which raises some ethical issues in and of itself.
They reported, and I’m sure most of you saw it in the news, that there was about a 10% advantage in the medical course these patients had in the coronary care unit, compared to those who didn’t receive prayer. Most of the work on religion and health is published in lower level journals, but this is the Archives of Internal Medicine, an esteemed, AMA publication and–at least until this publication–deservedly so. [laughter]
Here’s what did not receive attention in the press. The Christian intercessors who were to pray for the subject in the study were instructed to pray for a speedy recovery and no complications. The authors themselves pointed out that the two groups, those who received prayer and those who didn’t, did not differ in length of stay in the coronary care unit, and they didn’t differ in length of stay in the hospital. In other words, there was no difference in the speed of recovery! They didn’t even comment. It’s an astonishing omission and the reviewers apparently didn’t pick it up either.
The only difference between the two groups was in the scale that they constructed to assess the course in the coronary care unit. Now, scale construction, in the medical sciences, is not only an art, it’s a science. You don’t just make up scales that purport to measure things.
The scale consisted of the following. All of the patients were given values if they had certain events. So, for example, if they were in CCU and they required an antibiotic, they got 1 point. If they required an antianginal agent, like nitroglycerine, they got another point. If they required cardiac catheterization, they got 3 points, and if they died they got 6 points. [laughter]
Presumably, there is a relationship between the events and the values. Presumably, catheterization, valued at 3 points, is three times as bad as requiring an antibiotic, valued at 1 point. How do we know that it’s three times as bad? Maybe it’s four times as bad. Maybe it’s six times as bad. We have no evidence, so the scale fails on that ground. The most astonishing aspect of the study is that the authors reported the difference between the two groups in two ways: first, using the weighted scale, meaning that the different values from different events were summed, and second, using the unweighted scale, i.e., simply counting the number (but not the weights) of the events. That is, they counted how many bad events the prayer group had and compared it to the number the control group had. This is completely ridiculous! It means somebody who died–one bad event, but a bad one [laughter]–is better off than someone who required antianginal agents, antibiotics, and cardiac catheterization! [laughter] Now I suppose we could give patients the choice of which they would prefer. [laughter] As soon as the paper came out, we sent a letter to the editor of the Archives pointing this out.
Those are the empirical issues. It’s very clear there is nowhere near enough evidence to justify tearing down the wall of separation between religion and medicine, and there is nowhere near enough evidence to support Matthews’ assertion that the future of medicine is going to be “prayer and Prozac.”
But forgetting about the empirical issues, there are significant ethical issues that are raised by attempts to bring religion into medicine. We focus on three. There are certainly more.
The first one is the physician-patient relationship, and the power physicians have, even in these days of consumerism in medicine. You see a physician because you seek medical expertise, in the same way you would go to a tax attorney because you seek expertise in tax law. That means that the physician is, by the nature of the relationship, entitled to assume that you will follow his or her recommendations. That’s the whole point. The patient seeks expertise and the physician wants the patient to follow the medical recommendations. That’s all well and good, and appropriate, in the context of medicine. But when physicians depart from a medical agenda to pursue a nonmedical one, then they abuse their relationship as experts. That raises the risk of religious coercion.
If people want to be religious, they ought to have the opportunity to be so, and if they don’t want to be religious, they ought to feel free to do so as well. We don’t need physicians telling us whether we should do that or not. It’s just not appropriate. One of the ethical problems is this implicit coercion and it is certainly not hard to imagine that that occurs.
The proponents of this point of view–Koenig and Larson and Matthews–all say that they probe first, as part of taking a history: “Do you smoke? drink? how much exercise do you get? what’s your prayer life like?” I think it’s important for physicians to know about all aspects of patients, including whether or not they’re deeply religious. It’s important to understand. There’s a very big difference between taking into account religious factors, and taking them on as objects of intervention.
I want to come back to one point about the logical problem that proponents of this point of view find themselves in, although they don’t realize it, of course. They say–although I’m not sure that we should believe it–that they would not force religion onto anybody. They will only recommend religious activity, or engage in religious activity, with their patients, if the patients clearly indicate a willingness to do this. But then they also assert that the evidence is overwhelming that religious activity promotes health. It seems to me that by taking the former stance, that they will only engage in religious activity if their patients are open and receptive to it, they are derelict in their duties as physicians. It’s like saying to a patient: “You’ve got pneumonia. What’s your feeling about antibiotics? Are you in favor of them, or not?” [laughter] Physicians don’t do that. They say: “I recommend that you take antibiotics,” because there’s a consensus that antibiotics are an appropriate treatment for pneumonia. Nobody disputes that. If they’re saying the evidence is so strong that religion is associated with good health outcomes, then they’re derelict in their duty by not recommending religious activity to every patient, regardless of their feelings!
The second ethical problem, which in some ways is the most interesting, is the limits of medical intervention. There is no end to the number of factors, personal and socioeconomic, that influence health outcomes. For example, it is well-established that marital status confers benefits to health. While this marital effect may be stronger for men than for women, in general people who are married live longer and they are more healthy than people who are not. If you as a single person were to visit a physician, what would you say if the physician said, “You know, Bob, there’s this massive amount of evidence suggesting that marital status is good for your health, so I as your physician recommend that you get married.” [laughter] You all laugh. Why do you laugh? It’s as consistent as expecting that if people who are religious live longer, that you should engage in religious activity. The reason physicians don’t do it in the case of marriage, and in the case of financial and socioeconomic status, which are also associated with good health, is because we believe there are certain aspects of our lives that are private and personal, and even if they have an impact on health, are out-of-bounds from medicine.
The third ethical problem, and perhaps the one that is most serious, is the possibility of actually doing harm. Patients, even in these days, confront the age-old folk wisdom that illness is due to moral failure. Let me give you an example that comes from my own experience as a researcher.
I was visiting a research patient in an oncology unit. The patient was in a semi-private room, and she had just had a biopsy, and she was waiting for the biopsy results. The other patient, at that time surrounded by her family, also was waiting for her biopsy results. While I was there, the biopsy for the second patient came back, and it was negative. I’ll never forget this. Her father said, “We’re good people. We deserve this.” Now how was the patient I was visiting supposed to feel when her biopsy came back positive? Was she supposed to say, “I’m not a good person, that’s why I got cancer? Am I insufficiently devout, am I insufficiently faithful?” When you suggest that religious activity is associated with better health, you implicitly suggest quite the opposite: that poor health is a product of insufficient devotion, insufficient faith. It’s bad enough to be sick. It’s worse still to be catastrophically ill. To add the burden of guilt and remorse on top of that is simply unconscionable. And that happens all the time.
So for a variety of ethical reasons, it seems clear to me, regardless of what the empirical evidence is, that bringing religion into medicine not only makes no sense, it’s simply wrong to do, even if there were solid evidence–which, of course, there isn’t.
So . . . there’s just no solid evidence, and the ethical problems are so serious they have to be addressed. We should simply not tolerate attempts to bring religion into medicine until these matters have been resolved.
Audience Questions
Do a lot of failed studies not get reported?
That’s usually referred to as the “file-drawer effect” in science. It’s impossible to know, because the bias in scientific publications is toward studies that show effects. Every once in a while you do see publication of studies in which the findings are negative or at least counter-intuitive. I did just come across a paper in Social Science in Medicine, published in May of this year, following 250 hospitalized patients in London who were identified as either having spiritual beliefs or not. Those who had spiritual beliefs, nine month later, were worse off than those who didn’t. That didn’t make the national news, but did make Salon.com, on the internet. It wasn’t on ABC, I can tell you that.
Question about any studies on Christian Scientists who rely exclusively on religious treatment as opposed to conventional medical treatment.
I don’t know whether they live longer than others. There are certainly celebrated cases of Christian Scientists or their children who have died in pursuit of their religious beliefs.
Could the placebo effect account for many of the findings?
The answer certainly is “yes.” I find it ironic though, that that point was made by proponents of this point of view, that it might be the placebo effect. It seems to me that religious people should be deeply offended by the suggestion that religion is a placebo. [laughter]
What can be done when a physician oversteps his or her bounds and intrudes on the realm of the personal and the private?
I don’t know that you can report it to any regulatory board. What you can do is find a new physician.
Given the evidence of methodological shortcomings in this literature, is there an agenda behind this?
I don’t know the answer to that. So far as I know, there has been no investigation of that. I was actually delighted to be on the bill with Barbara Ehrenreich, because I thought she’s the kind of person who could do that, following the money that supports a lot of this research. I think that should be done.
Question about whether medical protocol was violated in the 1988 study by Randolph Byrd published in the Southern Medical Journal on cardiac patients.
I know the Byrd paper very well. It makes all kinds of methodological flaws, but I don’t know anything about protocol violations, and haven’t seen anything to that effect. When you talk about protocol violations, you are talking about treating patients. And there are relatively few studies that talk about treating patients. Most of the studies are population studies: comparing two groups, one of which is religious, and one which is not, or varying degrees of religion. So the medical protocol is irrelevant. There are very few experiments in which people are randomly assigned to either receive some kind of intervention and not receive some kind of intervention which is religious based.
Are there any different results from religious differences in the people assigned to pray for patients?
Great question! Let’s just talk about the intercessors. What would we ever do with a finding that prayers to Allah were more efficacious than prayers to Jesus? [laughter, clapping] That question highlights the absurdity of some of these investigations. Even if you showed an effect, what would you do? We raised the point in the Lancet paper: what if it were unequivocally shown that it was better to be Protestant than Catholic? Better to be Protestant than Jewish? What then? Should physicians counsel conversion?
Comment about it’s being almost impossible to go into a doctor’s or dentist’s office in South Carolina without finding bibles there, and having a doctor who admitted he doesn’t believe in evolution.
Find another doctor!
How do the researchers know that the people in the control group, presumably those not receiving the prayers of the intercessors, weren’t receiving prayers from anybody else?
After all, there are people who pray every day for all the sick, right? They actually address this question, but don’t really give a good answer. But the only answer can be that if the effect is real, there is basically a dose response effect. The more you get, the better! What else can it be? [laughter]
The question is if religion is demonstrably efficacious, if it really influences longevity, morbidity and mortality, and the quality of life, why don’t the insurance companies get in on it?
That really cuts to the chase. The answer is obvious!
Richard P. Sloan, Ph.D, is director, Behavioral Medicine Program, Columbia-Presbyterian Medical Center, New York, NY. He is an associate professor, Department of Psychiatry, at the College of Physicians and Surgeons, Columbia University. He is also Chief of the Department of Behavioral Medicine at New York State Psychiatric Institute. He received his B.S., Biology, Union College, Schenectady, and his M.A. and Ph.D Psychology, New School for Social Research, New York City. He is a New York State Licensed Psychologist. He and colleagues have explored and criticized the purported links between religion, spirituality and health appearing in popular and medical journals.
