Scientists are cautious against making a Type I research error, of accepting as valid a therapy that is not truly effective.

We know that therapies may be effective because patients and doctors expect they will work rather than because they are intrinsically beneficial. This is called the placebo effect. Roughly 30% of patients will show some symptomatic improvement with any treatment whatsoever, in almost any illness. This appears to be the result of psychological expectations combined with people's self-healing capacities.

Many medical practitioners have regarded placebo responses as something of a nuisance, feeling that they interfere in efforts to determine the effects of what conventional medicine considers real therapeutic interventions, particularly with medications.

To guard against Type I research errors, researchers have devised the randomized, double-blind, controlled trial (RCT). Before being accepted as a legitimate therapy, a treatment modality (drug, surgical procedure, psychotherapeutic method, complementary therapy) must pass stringent RCTs. A group of patients with the same diagnosis and severity of illness is randomly divided into sub-groups. One of the subgroups is given the therapy under study, a second group is given a known placebo (such as a sugar pill). Another group may be given a therapy of proven intrinsic value (a medication, massage, etc.). The groups receiving comparison treatments are called control groups.

Therapists, researchers and subjects in the study are kept blind to which therapy (experimental, placebo, or intervention of known value) is being given to any one subject. This minimizes the effects of suggestion that might bias subjects to feel better or worse according to the beliefs of the therapists or researchers, and according to the subjects' own expectations of improvement with a given therapy. Assessments of symptomatic change, for better or worse, are also made by diagnosticians who are blind to the particular therapy being given, for similar reasons.

Therapies are given within a standard protocol that is determined prior to the start of the study so that subjects within treatment groups receive similar doses and durations of interventions. This assures that the treatments under study are of known quality and quantity, and allows other researchers to replicate the experiment under a similar protocol.

Statistical analyses are calculated for the results. These provide estimates of the possibility that the results of the RCT might have occurred by chance. It is commonly accepted that a treatment must demonstrate itself effective beyond the statistical probability of five times in a hundred (abbreviated as p less than .05, or p < .05) if it is to be accepted as a valid finding rather than a chance occurrence. Naturally, if the results could occur by pure chance less than one time in a hundred (p < .01) or one time in a thousand (p < .001), and so on, they are that much more impressive.

The aim of science is not to open the door to infinite wisdom, but to set a limit to infinite error.                                           Bertolt Brecht (1939)

At the end of the study subjects in the various groups are checked to see that there were no differences between groups in any clinically important variables. Otherwise there might be one group that had subjects with more severe symptoms than another group, and these differences could have biased the results between groups in response to the therapies that each received. For instance, if one group had a lot of younger subjects, they might have had more resilience and could have recovered from whatever condition they had than the older group. The initial randomization of subjects into the various groups is meant to guard against this possibility. By assigning subjects randomly to each group it is statistically likely that equal numbers of subjects with any given condition or symptom will be assigned to each group, and that therefore the groups will be equivalent at the start of the study.

In some studies the same subjects are consecutively given active treatments and placebos, also under double-blind conditions. Subjects thereby serve as their own control group, minimizing differences that could occur between groups on the basis of the assignment of different subjects to treatment and control groups.

The difficulty with this approach is that there may be effects of a treatment that carry over from one period of intervention to the next, thereby confounding the picture.

Despite the best efforts of scientists to establish a research protocol to protect against Type I research errors, it is still possible that errors might occur. There could be confounding differences between groups that are present but not identified and that bias the responses of some or all of the subjects for better or worse, ending up misleading researchers into believing that an effect of therapy was observed or not, when it really was due to unequal distribution of symptom severity or other factors between groups.

If such an error led to the rejection of a therapy as being ineffective when it is actually a potent intervention, it would be classed as a Type II research error, of rejecting as false something that is actually true. The tendency in conventional medicine is to err in the direction of caution, and not to accept new therapies until there is a convincing body of evidence to minimize Type I research errors.

Even outside of any consideration of complementary therapies such as spiritual healing, many are questioning the validity of relying so heavily on controlled studies. They point out that the focus in RCTs is too narrow, and leads too readily to Type II errors.

At the next level you will find a list of Randomized, Controlled Studies of spiritual healing.

An annotated review and discussion of all the studies I could locate up to March, 2000, is presented in Healing Research, Volume I.

For those interested in designing a randomized controlled study, there is a generic outline for a RCT of healing. While this may appear simple and straight-forward, there are many ways in which confounding factors can influence such a study. Consultation in designing studies and in reporting them is strongly advised.     Scientists are cautious against making a Type I research error, of accepting as valid a therapy that is not truly effective.