The Prayer Prescription: Considering the New Evidence

Years ago, in 1988, it was reported in The Southern Medical Journal that it had been scientifically proven that prayer caused patients in a coronary care unit (CCU) to get better. There were very serious methodological flaws in that study. The physician who conducted it, Randolph Byrd, also refused to share his materials with those who wanted to look at the work in more detail.

Since then, even the National Institutes of Health (through the Alternative Medicine center that Congress found millions of dollars to fund) has been paying for medical "research" to study the effects of prayer. Now, 11 years after the Byrd report, it's being claimed that a new research study has replicated its findings and so validated "the healing power of prayer."

There are many reasons why the idea of prayer as medicine is nonsensical. The logical consequences of it ought to frighten even devout believers. But do the facts say that prayer really does "work?" The Kansas City Star of October 24th quoted the lead author of the new study — a PhD named William Harris — as saying that, "The patients who were prayed for just did better."

The Star also quoted a professor of psychiatry and medicine at Duke University School of Medicine as saying that, "This is very significant."

Well, I got a copy of the actual report, which came out in The Archives of Internal Medicine the week of the Star article and other media coverage. Basically, 990 patients admitted to a CCU were either prayed for or not prayed for by a team of "intercessors." The "intercessors," all Christians of one sort or another, were only given the first name of their "targets" and prayed for "a speedy recovery with no complications." The study subjects were completely ignorant of the study. In fact, the whole project was conducted without any direct patient contact. The hospital charts were the only measure used to determine how the subjects had done during their hospital stay.

On one measure, called the Mid America Heart Institute Cardiac Care Unit Score (MAHI-CCU Score), which is a composite of clinical data, the prayed-for patients had scores that were, on average, 10% better. That's not a very good number. An all-powerful creator of the universe ought to be able to do better than that. The benefits of aspirin are more impressive. In addition, the statistical significance of the 10% difference was poor. It was calculated that there was one chance in 28 that the difference could be coincidental. Again, in medical studies this is not impressive. Standard medical interventions are usually shown to be effective with statistical significance values of 1 in 100, 1 in 1000, or better. Still, it's not unreasonable to ask what might have caused the small, weakly significant 10% improvement among the prayed-for patients.

Well, to begin with, you have to realize that these sorts of studies are statistical games to some degree. As you might guess, if there are 28 different ways to get the result you want or expect, it's close to certain that you can find it if you are willing to settle for a statistical significance of 1 in 28. That is what appears to have happened here.

To begin with, it's not even a replication of the Byrd study. Using the same scoring system that Byrd used, there was no difference in the course of patients who were prayed for or not prayed for. So right there you have doubled the chance of finding that prayer "works."

But guess what? There was also no significant difference between the two groups in 34 of the 35 measures that make up the MAHI-CCU Score. In the 35th, there was a difference that was significant at 1 in 33. Again, this is not surprising. Similar situations arise in medicine all the time. For example, it is quite common to find "abnormal" lab values in healthy people if only you measure enough different things. This is because many lab values vary over a wide range and so "normal" is defined as the 95% of values around the mean.

How about the amount of time the prayed-for patients in this study spent in the CCU? How about the total number of days they were hospitalized? Remember, the "intercessors" prayed for "a speedy recovery." Yet there were no differences in the two groups' length of stay, either in the CCU or counting their entire hospitalization. In fact, there were two prayed-for subjects who spent almost twice as long in the CCU as any of the other patients. But even when these were removed from the calculations there was still no significant difference between the two groups' length of stay.

In these kinds of studies it's also important to have an idea of how a treatment group compares to the control group. That is, even if the subjects are randomly assigned, it can't be assumed that they're comparable with one another with respect to factors known to affect the result you want to measure. You have to actually make sure that they are comparable. In this study, for example, out of 31 "comorbid conditions" — variables known to be associated with poor outcomes — there was no statistically significant difference for any of them. But there was a trend towards the prayed-for subjects being less likely to have serious pre-existing conditions.

In addition, with a study such as this there is the problem of what is known as "publication bias." This is the tendency of researchers to write up and publish their positive results but not their negative results. Considering the interest in prayer and the funding that has gone to researchers purporting to study it, it is strange that there haven't been more published studies on prayer. So there is almost certainly negative data out there that no one will ever know about. And, obviously, if you get to keep the data you like and throw out the data you don't like, it ends up looking as if you've proved your claims.

The most outrageous part of this work is the authors' discussion of the significance of their findings. For although they admit that, "chance still remains a possible explanation of our results," they obviously think that pleading with a supernatural being should be taken seriously as a medical intervention. Yet they say that other prayers on behalf of or by the study subjects themselves in both groups were "presumed similar." How can a study of an intervention be taken seriously when such presumptions are made? For the result is that there was no actual measurement of the intervention under study. Yet if prayer can really be thought of as comparable in some way to a medical intervention, then there ought to be a demonstrable dose-response relationship. But for all anyone knows, the not-prayed-for subjects in this study may have had many more prayers offered on their behalf than the patients prayed for by the "intercessors."

Although the authors admit that their work does not prove that god(s) answers prayers or even that god(s) exist, they certainly do seem to think that they have demonstrated some sort of paranormal phenomenon. They compare themselves to James Lind, who discovered that lemons and limes cured scurvy aboard the HMS Salisbury in 1753. They say that, "Lind's inability to articulate ... a natural explanation for his findings ... did not invalidate his observations." But the natural explanation in that case was obviously that there was something in the lemons and limes that cured scurvy. In addition, Lind's results were dramatic and undeniable. This is not the case with this study.

As is usual with all such reports, the authors call for more research in their closing sentence. Fine. Let us suppose that the results in this case (and others) were so unimpressive and statistically weak because the study design really didn't involve any trust in God. It is, after all, a principle article of the Christian religion(s) that God values absolute faith in him and his power above all else. Matthew (17:20) has Jesus telling his followers, for example, that if they would only have "faith as a grin of mustard seed" that they could cause mountains to move from one place to another. And the Gospel Jesus frequently accomplishes his healing miracles by saying that the sick person's faith is what has actually cured them.

So why not conduct a real test of prayer consistent with these well-established doctrines? Why not give prayer every conceivable opportunity to show its superiority to the fallen works of human beings?

A large series of CCU admissions could be randomized to either a no-prayer, standard medical treatment group or to a group whose subjects would be immediately discharged to home with no medical treatment, there to be prayed over incessantly. Although it is doubtful that every potential subject would agree to participate in the study unless they got to stay in the CCU, surely there would be many Christians who would be willing to help demonstrate the power of prayer under conditions that could not fail to impress. Then, once this research showed strong positive results — and it should certainly be continued until it did show strong positive results — the "Our Father" prayer could be compared to the "Hail Mary." Next, praying while facing Mecca could be tested to see if there is a stronger effect than praying while facing Salt Lake City. Praying to Jesus could be compared to praying to the Holy Spirit, or to the Virgin Mary, or to Allah or Krishna.

In this way, tens or hundreds of millions dollars, if not more, could be frittered away in attempting to supply various religions and their sects with the "evidence" they need to be able to make a media splash with stories of "my god beats up your god!" For that seems to be the whole point of "research" work such as this, a veneer of academic scholarship notwithstanding.

© 1999 by Dr. Tim Gorski