Religion, Spirituality, and Medicine:
A Rebuttal To Skeptics
By Harold G. Koenig, M.D., M.H.Sc.; Ellen Idler, Ph. D; Stanislav Kasl, Ph.D, et al.

International Journal of Psychiatry in Medicine 29 (2), 123-31

A recent article by Sloan et al. in the Lancet has presented the skeptical side in the scientific debate on the religion-health relationship [1]. The interest in this topic and its relevance to medicine is underscored by the fact over 60 to 126 medical schools in the United States have initiated courses on religion/spirituality, and more are planning to do so. While we agree with Sloan et al. that the evidence needs careful evaluation, we find the review a highly selective one that over-emphasizes the negative aspects of many highly credible studies published in some of the best epidemiological journals, and minimizes or dismisses their substantial positive findings. While largely methodological, their critique is based on inadequate understanding of epidemiological methods, as described below.

We have recently completed a systematic review of over 1200 studies on the religion health relationship [2]. These studies have been conducted by different investigators, working at different institutions, studying different clinical and community populations located in different parts of the United States and world over the span of a century. The vast majority of these studies show a relationship between greater religious involvement and better mental health, better physical health, or lower use of health services. Many are not perfect because of the difficulty in studying this topic, but what is especially notable is that there are relatively few showing no relationship, and even fewer studies showing a significant negative relationship between religion and health. Clinicians have long known about the placebo effect­if patients believe they are going to get well, or are hopeful and positive about their situation, then they are likely to get better. Could not religion profoundly impact health through the power of belief? Instead of prematurely dismissing the promise of the evidence, as the Sloan et al. article does, we feel that additional research is needed to narrow down exactly how religion affects health, what aspects of religion do so, and how these findings can be sensitively applied to clinical practice.

POINTS OF AGREEMENT

We agree with Sloan and colleagues that religious or spiritual activities provide many people with comfort in the face of illness, and that there is no ethical objection to discussing medical issues between doctor and patients in the context of a shared faith tradition. We agree the characteristics of populations such as age, sex, education, race, socioeconomic status, and health status must be controlled using multivariate methods to reduce the likelihood of confounding. We also agree that studies should control for multiple comparisons particularly when multiple outcomes are being examined without a single guiding hypothesis. Finally, we completely agree that between endorsing the idea that religion and faith can bring comfort to patients and rejecting the view that physicians should actively promote religion among patients lies a vast uncharted territory where guidelines for appropriate behavior are needed.

POINTS OF DISAGREEMENT

In our opinion, the Lancet article was not a very even-handed review. We divide our concerns into three sections, arguing that this review 1) omitted evidence, 2) made erroneous, incorrect, or misleading statements, and 3) argued ethical issues based on personal opinions and strawman arguments rather than scientific evidence.

Omitted Evidence

In the last sentence of the introductory paragraph, Sloan et al. note that their review is a "comprehensive, though not systematic, review of the empirical evidence...." On careful analysis, however, we come to a different conclusion. The review was highly "selective" in discussing only 24 of the approximately 325 studies of religion's relationship to physical health and none of the nearly 900 studies on mental health [2]. Only 17 of nearly 100 mortality studies were examined in the article, and nine high-quality studies were not even mentioned. Eight of these nine prospective studies found a significant inverse relationship between a religious measure and mortality after controlling for multiple covariates [3-10]; only one study found no association [11].

Four of the most recent studies, employing state-of-the-art statistical methodology, demonstrate substantially longer survival for frequent church attenders ­ two studies using national samples and two using large regional samples on opposite coasts of the United States [7-10]. While these studies were completed since the Sloan et al. review, they illustrate the pace of recent research and the convergence of evidence. After examining the entire body of research, most scientists would not conclude that the findings are either weak or inconsistent.
Moreover, while this review acknowledges that religion can provide comfort to patients with medical illness, it completely excludes research on the relationship between religion and mental health. This is relevant because one of the strongest rationales for religion's effects on physical health lies in its connection with psychological and social functioning. Finally, there is only a cursory mention of religion's effects on health behaviors, and only as evidence to dismiss correlations between religious involvement and physical health or mortality outcomes. Thus the studies chosen for the review were selective from the standpoint of omitted evidence and also because they relied on a narrow conception of health.

Erroneous, Incorrect, or Misleading Statements

The reviewers emphasize the need to control for behavioral, psychosocial, health, and demographic characteristics in the models. They also imply that if after controlling for these variables no association remains, then the relationship between religion and health outcome is trivial or spurious. This line of argument begins well but is taken too far and misses a crucial distinction. Two very different types of variables are often controlled when examining the relationship between religion and health. One set of variables are the true confounders that disguise or exaggerate a relationship between two variables (i.e., age, race, sex, education, baseline health status, and functioning, etc.). If groups in the population that differ in their religiousness are also different with respect to these characteristics at baseline, follow-up differences could be due to the pre-existing differences, and not to religiousness at all.

The other set of variables are potential explanatory or mediating variables that described the mechanisms by which religion might affect health. These may include behavioral variables such as smoking or alcohol use, psychosocial variables such as social support or depression, or intervening biological variables such as hypertension or immune status. If religion increases social support, buffers stress, reduces depression, decreases smoking, and lowers blood pressure, and all these variables are also related to mortality, then controlling for these variables may completely account for the relationship between religious involvement and mortality. A positive relationship between religion and survival may weaken or lose significance altogether. This does not mean that there is no relationship between religion and health, only that a study has exposed some of the mechanisms through which religious involvement affects health.

For example, the reviewers minimize the finding that "In another large study, attendance at religious services was associated with increased functional capacity in the elderly but after control for appropriate covariates, this relation held for only 3 of the 7 years in which outcome data were collected." In fact, investigators followed subjects for twelve years, finding significant associations between church attendance in 1982 and better functioning in 1983, 1984, 1985, 1986, 1987, and 1988 after controlling for multiple confounders [12]. Sloan et al.'s comments however, refer to a later phase of the analysis, which introduced explanatory variables. Adjustment for behavioral and psychosocial variables which were themselves strongly associated with religious practices did in fact reduce the association, but this helped explain the relationship ­ not "explain it away." To conclude that relationships were significant in only three of the seven years (again, after controlling for the mechanism of effect) ­ is misleading the reader.

Even the most serious form of confounding in such studies, of religious attendance with health status and functioning at baseline, may be inflated as an issue. Idler and Kasl found that physical functioning had only a slight impact on later attendance; religiousness had a far greater impact on later functioning [12]. Likewise, Strawbridge et al. found that frequent church attenders had more mobility problems at baseline than did infrequent attenders [13]. Some very ill people attend services every Sabbath, arriving in wheelchairs and with walkers.

Elsewhere, the reviewers criticize a study simply because it is a case-control study, "the deficiencies of which are widely known." It is a weak argument to dismiss a study because of a design ­ especially a frequently used one ­ without discussing what specific weakness of the design renders the results questionable. A case-control design is appropriate for cross-sectional data. Years would have passed before discovering the association between cigarette smoking and lung cancer if it were not for the early Doll case-control studies. The remarkable results of the Friedlander et al. study (showing that male unbelievers had a 4.2 times increased likelihood of having an acute myocardial infarction and females a 7.3 times increased likelihood) cannot be ignored ­ particularly when they support the findings of prospective cohort studies [14].

Erroneous or misleading statements also indicate careless scholarship on the reviewers' part. For example, Sloan et al. note "a multivariate model that predicted mortality from coronary heart disease included standard risk factors but omitted religion, and no information on risk-ratio or confidence intervals or even level of statistical significance was provided." Quite to the contrary, Goldbourt et al. did include religion in a multivariate analysis (which included age, systolic blood pressure, total cholesterol, cigarette smoking, diabetes, body mass index, and baseline heart disease) and did note a statistically significant reduced risk associated with greater religious involvement (i.e., >=20%) [15]. A 20 percent reduced risk of coronary heart disease mortality in 10,000 persons over a period of twenty-three years is not a weak effect. Similarly, Sloan et al.'s discussion of the Kark et al. study implies that differences in conventional risk factors such as blood pressure, dietary habits, serum cholesterol, smoking, etc., might have accounted for the lower mortality among members of religious vs. secular kibbutzim [16]. Sloan et al. failed to mention, however, that the potential effect of risk factor differences was assessed and found to have a trivial effect on mortality (RR = 1.0-1.2).

Findings from important studies are given less value in the review simply because effects are particularly strong in one subgroup. For example, the review notes "In a smaller, study, religiousness predicted mortality in the elderly poor but only among those in poor heath." Finding a relationship only among sick elderly people does not make the finding irrelevant. It only indicates that a statistical interaction or effect modification is present, and that the effects are strongest the among elderly with health problems ­ indeed, that subgroup of the population most likely to see medical doctors. Likewise, reviewers note for two major studies that "after control for all relevant covariates, this relation held only for women." Again, this simply indicates an effect modification, not that results are irrelevant and can be dismissed (13, 17). In the Alameda County study involving over 5000 participants, Strawbridge et al. found that frequent attendance at religious services in 1965 predicted a one-third reduction in mortality for women during a twenty-eight year follow-up (after controlling for confounders ­ including multiple measures of physical functioning and chronic illness, explanatory variables, and time-varying covariates) [13]. That the findings are sometimes stronger for women, who are known to be more religiously observant, should not be a basis for downplaying their importance.

Sloan et al. referred to the findings of the Byrd study on several occasions (which tested the efficacy of distant intercessory prayer in double-blinded conditions) as evidence for a poorly conducted study [18]. We too are skeptical about the design and findings of distant intercessory prayer studies (in which the subjects are unaware of the prayers) and consider this design quite different from the remainder of the epidemiological studies with which we are concerned. The beneficial effects in such prayer-intervention studies supposedly derive from mechanisms for which there is no scientific explanation, which is quite different from the observational studies of populations discussed in the paper (where mechanism is being understood in terms of well-known and established psychosocial, behavioral, and biological pathways).

Another major point made by the reviewers is that the findings are inconsistent. They note the different results associated with measures of public religious involvement (church attendance) vs. private religious activity (prayer, religious coping). For example, while private religiousness was not related to mortality in one study [19] one of two items that made up this scale (religious comfort and strength) was related to lower mortality in another study [20]. The charge of inconsistency, however, should be applied only when near identical study designs yield different results, not when studies of very different populations using different dimensions of religiousness are relevant to different health outcomes. The two studies chosen for comparison in the review were a healthy elderly population and a clinical sample of cardiac surgery patients. Rather than weak evidence, inconsistent findings may be suggesting specificity of processes.

Reviews of research on the health effects of other psychosocial variables like socioeconomic status, social support, depression, stress, etc., seldom reveal highly consistent results across all studies. Eventually, the evidence mounts in spite of earlier lack of uniformity. In this regard, Sloan et al. appear to be holding religion to a higher standard than other psychosocial variables. Even a drug company must only submit two studies to the FDA that show an effect for a drug over placebo. There are usually more than two studies submitted and they don't ALL have to show an effect. Most of us would agree that drugs are more dangerous than religion, so why hold religion to a higher standard?

Religion is not a single homogenous construct where different religious measures all assess the same thing. The many aspects of religion, which include public ritual observances, private devotional practices, as well as attitudes, beliefs, and feelings, make it difficult to study. The dimensions of religion relevant to seriously ill patients are virtually certain to be different from those that affect healthy people living in the community. Moreover, the different dimensions very likely have different pathways in their effects on health. The only sound conclusion is that the complexity of the subject requires a great deal more research before it will be fully understood.

Ethical Issues or Personal Opinions?

Sloan et al. note that doctors should not depart from their area of established expertise to promote a non-medical agenda. If, however, it is established that religious/spiritual involvement is connected with health behaviors, medical compliance, and coping with illness, this intertwines it with a medical agenda. The primary task of the physician is "to cure sometimes, to relieve often, to comfort always." If some proportion of patients utilize religious beliefs and practices as their primary way of coping with medical illness and the stresses associated with it, then "to comfort always" must include the support and recognition of what the patient finds comforting.

There are many areas of private behavior that physicians delve into routinely. For example, smoking is a relatively private affair ­ particularly if the person respects others' health by smoking outdoors. The same is true for diet, exercise, and substance use, and sexual activity since these behaviors have been linked to health. There is ample evidence (ignored in this article) that religion is associated with lower rates of cigarette smoking, excessive drinking, drug abuse, and depression. We would argue that physicians should see religion as a substantial health resource that already exists in the lives of many of their patients, and treat it with the respect it deserves.

Finally, Sloan et al. are concerned that if religion is suggested as a possible road to better health, people may begin to see illness as a result of insufficient faith or moral failure. It could be similarly argued that doctors should refrain from telling patients to quit smoking for fear that patients would blame themselves in years to come if emphysema or lung cancer developed because they hadn't quit. This comment is generic to most of the advice that physicians give patients about any health-related attitude or behavior.

Concern about moral failings has long been a problem in other areas of social and behavioral medicine. If their cancer recurs, patients often feel that they did not deal adequately with life stressors or practice their relaxation techniques faithfully enough. People who don't or can't lose weight often feel similarly. This problem with guilt over becoming sick in not unique for religion. Health professionals have also had to refrain from "judging" patients for decisions that do not promote health ­ even in morally sensitive areas such as behaviors that increase the risk of AIDS, other STDs, or teenage pregnancy. While health-care providers should clearly never shoulder patients with additional burden or guilt for moral failure or inadequate faith simply because their physical health fails, many patients may already be taking this perspective. Discussion of such issues with their physician will at least get them out into the open, and allow for referral to a chaplain or other clergy who can help the patient deal appropriately with such concerns.

CONCLUSION

We represent the disciplines of medicine, epidemiology, public health, and the social sciences, and a host of differing religious beliefs, including having few or none. When we look at the best epidemiological studies that have been conducted in this field, we see evidence of a strong association between religious involvement and health that puts religion in the same category with other psychosocial factors where the research has had more time to accumulate. Religion may affect health through many mechanisms including its influence on health practices, its provision of social supports, and its offer of comfort even in situations of extreme suffering. Our thoughts about the relevance of this research to medical care, however, are as varied as our own beliefs. We all agree that physicians should "take account of" their patients' religious beliefs, but then so do Sloan et al. We differ among ourselves about whether physicians should or can effectively take the lead in providing spiritual guidance to patients. Nevertheless, we are strongly convinced, as Sloan et al. are not, that the evidence regarding religion and health, while still emerging, is neither weak nor inconsistent, and that religion is a factor that should not be overlooked in describing influences on the health of populations.

--------------------------------------------------------------------------------

REFERENCES

• Sloan RP, Baglella E, Powell T. Religion, spirituality, and medicine. The Lancet 1999; 353:664-667.
• Koenig HG, McCullough M, Larson DB. Religion and health: A century of research reviewed. New York, NY: Oxford University Press, in press.
• Schoenbach VJ, Kaplan BH, Fredman L, Kleinbaum DG. Social ties and mortality in Evans County, Georgia. American Journal of Epidemiology 1986; 123:577-591.
• Seeman TE, Kaplan GA, Knudsen L, Cohen R, Guralnik J. Social network ties and mortality among the elderly in the Alameda County Study. American Journal of Epidemiology 1987; 126:714-723.
• Bryant S, Rakowski W. Predictors of mortality among elderly African Americans.
  Research on Aging 1992; 14:50-67.
• Goldman N, Korenman S, Weinstein R. Marital status and health among the elderly. Social Science and Medicine 1995; 40:1717-1730.
• Oman D, Reed D. Religion and mortality among the community-dwelling elderly.
  American Journal of Public Health 1998;88: 1469-1475.
• Hummer R, Rogers R, Nam C, Ellison CG. Religious attendance and mortality in the U.S. adult population. Demography 1999.
• Koenig HG, Hays JC, Larson DB, George LK, Cohen HJ, McCullough M, Meador K, Blazer DG. Does religious attendance prolong survival?: A six-year follow-up study of 3,968 older adults. Journal of Gerontology: Medical Sciences 1999; 54A: M370-M377.
• Musick MA, House JS, Williams DR. Attendance at religious services and mortality in a national sample. Paper presented at the 1999 Annual Meeting of the American Sociological Association, Chicago, Illinois, August 6.
• Koenig HG, Larson DB, Hays JC, McCullough ME, George LK, Branch PS, Meador KG. Kuchibhatla M. Religion and survival of 1010 male veterans hospitalized with medical illness. Journal of Religion and Health 1998; 37:15-29.
• Idler EL, Kasl, SV. Religion among disabled and nondisabled elderly persons, II: Attendance at religious services as a predictor of the course of disability. Journal of Gerontology 1997; 52B:s306-s316.
• Strawbridge WJ, Cohen RD, Shema SJ, Kaplan GA. Frequent attendance at religious services and morality over 28 years. American Journal of Public Health 1997; 87: 957-961.
• Friedlander Y. Kark JD, Stein Y. Religious orthodoxy and myocardial infarction in Jerusalem: A case-control study. International Journal of Cardiology 1986; 10:33-41.
• Goldbourt U, Yaari S, Medalie JH. Factors predictive of long-term coronary heart disease mortality among 10,059 male Israeli civil servants and municipal employees. Cardiology 1993; 82:100-121.
• Kark JD, Shemi G, Friedlander Y, Martin O, Manor O, Blondheim SH. Does religious observance promote health? Mortality in secular vs. religious kibbutzim in Israel. American Journal of Public Health 1996;86:341-346.
• House JS, Robbins C, Metzner H. The association of social relationships and activities with mortality: Prospective evidence from the Tecumseh community health study. American Journal of Epidemiology 1982; 116;123-140.
• Byrd RC. Positive therapeutic effects of intercessory prayer in a coronary care unit population. Southern Medical Journal 1988;81:826-829.
• Idler EL, Kasl, SV. Religion disability, depression, and the timing of death. American Journal of Sociology 1992;97:1052-1079.
• Oxman TE, Freeman DH, Manheimer ED. Lack of social participation or religious strength and comfort as risk factors for death after cardiac surgery in the elderly. Psychosomatic Medicine 1995; 57:5-15.
  

Return To Faith & Health Connection Main Menu | Go To Next Article