Let Us Pray That People Stop Praying by James W. Williamson, M.D. (April 2002)

Since the terrorist attacks of September 11, already keen interest in prayer has increased as part of the frenzied upsurge in religion in general. Topping The New York Times’ bestseller list is a small book called The Prayer of Jabez, with “its message that lives can be profoundly changed by the power of prayer.” Images bombard us on television regularly of masses of humanity kneeling in supplication, praying to some all-powerful deity. These images only increase the perception that prayer is a potent force in dealing with the world’s problems that is endorsed by almost all of humanity.

Many religious people want to cling to the ancient belief in the supernatural, including prayer, and yet accept the conclusions and benefits of modern science. They can’t have it both ways. To study the natural world, scientists must have an implicit assumption that it operates only by natural, predictable processes, which cannot be affected arbitrarily by an all-powerful deity.

One of the major ways that scientists provide proof of theories is through well-designed studies, of which the “gold standard” is the large, randomized, prospective, controlled, double-blind type. If such a study could be influenced by a personal god who responded to prayers to change the results, science would be in shambles. There would be no way ever to do a valid experiment since investigators couldn’t be sure that someone, somewhere, hadn’t uttered a specific or generic prayer that would affect the study. In short, science by its very nature, rejects any influence of prayer on the physical world.

Even though prayer is an irrational concept, could it nevertheless be tested scientifically? Francis Galton, the brilliant and eccentric cousin of Charles Darwin, thought so and gave the idea scientific legitimacy. Galton was the father of biometry and a central figure in the founding of modern statistical analysis. He argued that regardless of how the prayers “may be supposed to operate,” the efficacy of prayer . . . is a perfectly appropriate and legitimate subject of scientific inquiry” because it can be tested statistically. He then proceeded to set up such studies.

In one statistical study, Galton examined the longevity of clergy. He reasoned that clergy should be the longest lived of all since they were the most “prayerful class” of all and among the most prayed for. When Galton compared the longevity of eminent clergy with eminent doctors and lawyers, the clergy were the shortest lived of the three groups. In this study of the clergy, he cited a previous study by Guy (Galton wasn’t the first to think of analyzing prayer statistically but usually gets the credit) where Guy found prayer did not protect royalty, who were much prayed for, when compared to other members of the aristocracy. In analyzing the data on royalty, Galton concluded: “Sovereigns are literally the shortest lived of all who have the advantage of affluence.”
Galton looked for other statistical data. He examined the insurance rates for ships. He reasoned that ships carrying missionaries and pilgrims should have lower rates since frequent praying by the occupants should decrease the number of accidents. He found that the rates were the same; ships carrying missionaries and pilgrims sank just as often as other ships.

Following up on Galton’s statistical studies on prayer, Rupert Sheldrake, a Cambridge-trained plant biologist, did one of his own, examining the effects of prayer in India. Most people there prefer having a son, and a tremendous amount of praying goes into the effort to produce one. Sheldrake examined statistics of live male births in India and used data from England as a control where the preference for sons was less strong. He found that in both England and India there were 106 males to 100 females, just as in every other country. He stated, “if this enormous amount of psychic effort and praying of holy men were working, you would expect on average the percentage of live male births to be higher.”

Although these statistical studies from the nineteenth century strongly suggest that prayer is not effective, they do not meet the “gold standard” of a completely valid scientific study. The media regularly mention a large number of contemporary studies that supposedly scientifically validate the beneficial effects of prayer on human health. So what is the truth in this matter? Actually, there are only three that meet the “gold standard.” Happily, the fact that there are only three studies considerably reduces the amount of information freethinkers need to acquire to refute frequent and erroneous claims.

When we say that a finding in a scientific study is statistically significant, “significant” has a specific statistical meaning. To be considered significant, a finding must be (.05) or less, which means the probability that it could be due to chance is 5 in 100. The main point to appreciate is that this figure, although reasonable, is strictly arbitrary. Therefore, the figure of (.05) is borderline significant, .04 (a probability of 4 in a 100 of being due to chance) is considered significant, and .06 (6 in 100) is considered not significant.

The figure (.05) is the one accepted for “ordinary” scientific studies. But what criterion should be applied in proving a supernatural finding? After all, as the old saw goes, extraordinary claims should require extraordinary proof, and this requirement should especially apply to claims of the supernatural.

The James Randi Educational Foundation has a standing offer of one million dollars to anyone who can demonstrate any supernatural event under carefully controlled scientific conditions. The foundation has never had a single person who even got past the preliminary testing. Its members think that a study that would prove a claim of the supernatural should eliminate the possibility that the result could be by chance, in the range of 1 in 10,000,000, a far cry from 5 in 100.

Robert Park, in his excellent book, Voodoo Science, observes that a characteristic of voodoo science is that there are always very small differences in studies, just barely detectable, and that can’t be amplified in further investigations. These barely detectable positive results usually indicate flaws in the studies themselves rather than real findings.

Let’s examine in some detail the three studies on intercessory prayer that were large, prospective, randomized, double-blind ones–the only three that pass muster as valid scientific investigations of the effects of prayer on human health. Intercessory prayer (prayer at a distance) was chosen so that the placebo effect of direct prayer would be eliminated. All of these studies were done on coronary care unit (CCU) patients.

The first study was entitled “Positive Therapeutic Effects of Intercessory Prayer in a Coronary Care Unit Population” by Randolph Byrd, M.D., published in the Southern Medical Journal, July 1988. Dr. Byrd stated:

“My study concerning prayer and patients in a general hospital coronary care unit was designed to answer two questions: (1) Does intercessory prayer to the Judeo-Christian God have any effect on the patient’s medical condition and recovery while in the hospital? (2) How are these effects characterized, if present?”

Over ten months, 393 patients admitted to the CCU at San Francisco General Hospital were randomized to an intercessory prayer group (192 patients) or to a control group (201 patients). After randomization, each patient in the prayer group was assigned to three to seven intercessors, who were all “born-again Christians (according to the Gospel of John 3:3)” of various denominations.

Dr. Byrd wrote: “The patients’ first name, diagnosis, and general condition, along with pertinent updates on their condition, were given to the intercessors. The intercessory prayer was done outside the hospital daily until the patient was discharged from the hospital. Under the direction of a coordinator, each intercessor was asked to pray daily for a rapid recovery and for prevention of complications and death, in addition to other areas of prayer they believed to be beneficial to the patient.”

The results were summarized in “Table 2” of the Byrd study entitled “Results of intercessory Prayer.” There was no statistically significant difference between the prayer and control group in these measurements: days in CCU after entry; days in hospital after entry; number of discharge medications. Only when a list of 26 “New Problems, Diagnoses, and Therapeutic Events After Entry” was compiled was any statistically differences found and then only in 6 of the items: congestive heart failure (.03); diuretics (.05); cardiopulmonary arrest (.02); pneumonia (.03); antibiotics (.005); intubation/ ventilation (.002). When Dr. Byrd subjected these items to multivariate analysis (a statistical method of analyzing the overall significance when multiple factors are positive), he found the prayer group to better the control group at the statistically significant level of (.0001).

In “Table 3,” “Results of Scoring the Postentry Hospital Course,” he constructed three categories, “Good, Intermediate, and Bad,” using a self-designed and previously not scientifically validated method. The prayer group bettered the control group at a level of (.01).

Although this study appears to meet the “gold standard” of a large, prospective, randomized, double-blind investigation, scientists have pointed out a number of flaws:

The study was not “blinded’ in two respects: 1) Janet Greene, the coordinator of the study, on whom Dr. Byrd depended for the collection of data, knew exactly who was being prayed for, and interacted regularly with the patients in the study. 2) “Table 3” was formulated by Dr. Byrd at the request of editors who initially evaluated his paper after the “blinding” had been removed.

There was no difference in clear-cut end points such as days in the CCU, days in the hospital, or mortality between the two groups. Only when complicated statistical analyses were done on a long list of items do any data emerge that favor the prayed-for group–hardly evidence of an all-powerful deity. Also, if prayer had any effect, an overall improvement would be expected. Of the six items where the prayer group did better, four were of borderline statistical significance and only two were clearly significant. Are we to conclude that the deity is only concerned with reducing antibiotic use and ventilating patients in the CCU? This study provides no information on the physicians involved in this study. This information could be important since certain physicians use antibiotics and intubate patients much more readily than others.

The method that Dr. Byrd used in his scoring in “Table 3” had not been validated by any previous studies.

When Irwin Tessman, Ph.D., professor of biological sciences at Purdue University, requested of Dr. Byrd that Dr. Tessman be allowed to review the raw data that went into the study, he was refused. Since Dr. Byrd’s claim is one of the supernatural, it would seem appropriate that all aspects of the study be reviewed by independent investigators.

The degree of obvious religiosity communicated by Dr. Byrd raises doubts that he could be completely objective on a scientific investigation of prayer, something that he deeply believes is effective. Under “Acknowledgments” at the end of the paper, he writes: “I thank God for responding to the many prayers made on behalf of the patients.”

The second study that appears to meet the “gold standard” for scientific studies is “A Randomized, Controlled Trial of the Effects of Remote, Intercessory Prayer on Outcomes in Patients Admitted to the Coronary Care Unit” published in the October 25, 1999 edition of the Archives of Internal Medicine. The investigators were William S. Harris, Ph.D., plus eight others of the Mid America Heart Institute. The study was conducted at Saint Luke’s Hospital, Kansas City, Missouri, a private, university-associated hospital.

“The purpose of the present study was to attempt to replicate Byrd’s findings by testing the hypothesis that patients who are unknowingly and remotely prayed for by blinded intercessors will experience fewer complications and have a shorter hospital stay than patients not receiving such prayer,” admitted the investigators.

The intercessors (five to pray for each patient compared to three to seven in Byrd’s study), were to pray for “a speedy recovery with no complications” plus “anything else that seemed appropriate to them.” 1013 patients were randomized, 484 to the prayer group and 529 to the usual care group. After removal of those patients who spent less than 24 hours in the CCU (prayer was not started until 24 hours after admission), 524 remained in the usual care group and 466 in the prayer group (a high drop-out rate).
A list of events after entry into the study was compiled, much like the one in the Byrd study, but with 34 events instead of the 26 in the Byrd Study. Again, a scheme was devised to evaluate the overall hospital course, a totally new and untested system, but different from the also new and untested one devised by Byrd. The Harris study scheme was called the Mid America Heart Institute Cardiac Care Unit (MAHI-CCU) Scoring System, and its criteria are presented in “Table 1” of his paper.

The only finding in the Harris study that indicated the prayer group outperformed the control group was in using the MAHI-CCU Scoring System and then only at a probability level of (.04), a figure very close to the cut-off level of (.05).

The Harris study is a much better study than the Byrd study because the number of patients is larger, it appears to be completely blinded, and the degree of religiosity of the investigators appears to be lower (although Dr. Harris supposedly supports the idea of “intelligent design”). Nevertheless, scientific investigators have noted flaws:

1) As already noted, the MAHI-CCU Scoring System has never been previously scientifically validated. Without such validation, any result produced by it is subject to question.

2) The much higher dropout rate in the first 24 hours in the prayer group is a very serious criticism of the study. The statistical probability that this finding would appear by chance is (.001), or 1 chance in a 1000, a statistically very significant finding. This higher dropout rate, since the mortality rate in the two groups was the same, suggests that the prayer group, for unknown reasons, was not quite as ill as the control group since patients discharged within a day often turn out not to have serious problems. If they were a little less ill at the start, we would expect them to have a more favorable course.

3) The conclusions stated in this investigation, as I’ll describe shortly, are not justified by the data.

Positive findings in a scientific study are not considered valid until replicated by independent investigators. So did the Harris study replicate the positive findings of the Byrd study? The answer is a resounding no! Of the 6 items in the list of 26 items previously described in the Byrd study where the prayed-for group did better, not one was statistically significant in the Harris study. When the Harris study subjected its data to the same scheme that Byrd had used in his evaluation of the hospital course of the patients (Table 3 in the Byrd study), the Harris study found the difference between the two groups of (.29) was not even close to being statistically significant. The Harris study did replicate the negative findings from the Byrd Study. There was no statistical difference in days in the CCU, days in the hospital, or mortality.

In remarks at the end of the Harris study, the investigators stated: “Our findings support Byrd’s conclusions despite the fact that we could not document an effect of prayer using his scoring system.” This statement is erroneous. Not only do these findings not support Byrd’s conclusions, they directly refute them.

The most recent study and, I believe, the best designed one, was published in the Mayo Clinic Proceedings in December 2001, entitled “Intercessory Prayer and Cardiovascular Disease Progression in a Coronary Care Unit Population: A Randomized Controlled Trial.” This third “gold standard” study should settle the matter once and for all scientifically. The investigators were Jennifer M. Aviles, M.D., and six others. This trial was done on patients immediately after discharge from the Coronary Care Unit, a time when the intensity of extraneous intercessory praying by family and friends would generally be waning.

Here is their summary of the findings:

“Patients and Methods: In this randomized, controlled trial conducted between 1997 and 1999, a total of 799 coronary care unit patients were randomized at hospital discharge to the intercessory prayer group or to the control group. . . . The primary end point after 26 weeks was any of the following: death, cardiac arrest, rehospitalization for cardiovascular disease, coronary revascularization, or an emergency department visit for cardiovascular disease. Patients were divided into a high-group based on the presence of any of 5 risk factors (age > or = 70 years, diabetes mellitus, prior myocardial infarction, cerebrovascular disease, or peripheral vascular disease) or a low-risk group (absence of risk factors) for subsequent primary events.”
The investigators summarized their findings as follows:

“Conclusions: As delivered in this study, intercessory prayer had no significant effect on medical outcomes after hospitalization in a coronary care unit.” Not even one difference showed up between the control group and the prayed-for group.
The statistical studies from the nineteenth century, and the three CCU studies on prayer are quite consistent with the fact that humanity is wasting a huge amount of time on a procedure that simply doesn’t work. Nonetheless, faith in prayer is so pervasive and deeply rooted, you can be sure believers will continue to devise future studies in a desperate effort to confirm their beliefs.

Now that you have the scientific information, don’t let the statement that the efficacy of prayer has been proven by scientific studies go unchallenged. It’s simply untrue.

Freedom From Religion Foundation