In the general public and within medicine, interest in the possibility that religious and spiritual activity may confer health benefits is increasing, as the article by Anandarajah and Hight¹ in this issue of American Family Physician clearly demonstrates. The empiric support required to convert this interest into recommendations for medical practice is weak and inconclusive at best, with most studies having numerous methodologic shortcomings.² Even if there were methodologically solid findings demonstrating associations between religious and spiritual activities and health outcomes, problems would still exist.

First of all, such associations would occur at the epidemiologic level, not at the level of clinical interventions. To convert epidemiologic findings to clinical practice, randomized trials of interventions must be conducted. Without such trials, no scientific basis exists for making recommendations to engage in religious activities in clinical settings. We have no idea, for example, whether recommending that patients attend religious services will lead to increased attendance and, if so, whether attendance under these conditions will lead to better health outcomes. The difference between going to church on one's own and doing so on the advice of a physician may be substantial. There is a qualitative difference between engaging in religious activity and, for example, following a low-fat diet. And even if we had convincing evidence that these interventions worked, they still raise substantial ethical concerns.

First, many factors that influence health are nonetheless regarded as beyond the domain of medical practice. For example, marital status is strongly associated with health effects, but we would recoil at a physician recommending marriage because of its association with health. This is because we regard such decisions about marriage as personal and private, even if they have associations with health.

Second, the physician-patient relationship is asymmetric: physicians expect patients to comply with their recommendations, and patients generally accede to this authority. Recommending religion to patients in this context may be coercive.

Third, by linking religious activity with better health outcomes, the physician also implies the converse: that poor health outcomes are associated with insufficient devotion. Even in the 21st century, patients still confront the age-old folk wisdom that illness is a punishment for moral failure. By linking religion to health outcomes, physicians may actually cause harm.

Finally, by probing the religious interests of patients and distinguishing those for whom religion and spirituality are important from those for whom they are not, physicians run the risk of discriminating by encouraging only the former group to engage in religious activity. If, as proponents of making religious activity an adjunctive medical treatment indicate, the evidence of the health benefits of religious activity is overwhelming, then by this assessment two classes of patients are created: those who will receive this important advice and those from whom it will be withheld.

Thus, the absence of compelling empiric evidence and the substantial ethical concerns raised suggest that, at the very least, it is premature to recommend making religious and spiritual activities adjunctive medical treatments.