Does prayer do anything?

The STEP trial randomised 1,802 cardiac bypass patients at six American hospitals into three groups: prayed for and uncertain, not prayed for and uncertain, and prayed for and told so. Three Christian prayer groups, two Catholic and one Protestant, prayed for fourteen days from a list of first names and last initials, starting the night before surgery.

Complications occurred in 52% of the prayed-for uncertain group and 51% of the not-prayed-for uncertain group, a relative risk of 1.02. In the group certain of receiving prayer it was 59%: against the prayed-for uncertain group, a relative risk of 1.14 whose confidence interval excludes 1. Thirty-day mortality was similar across all three. Major events were not: 18% in the prayed-for uncertain group against 13% in the not-prayed-for one, which the authors call a borderline excess that may also be chance.

The authors state the conclusion flatly: intercessory prayer itself had no effect on complication-free recovery, and certainty of receiving it was associated with more complications.

A Cochrane review pooled ten randomised trials covering 7,646 patients and found no clear effect on death, on general clinical state, on readmission or on rehospitalisation.

Its conclusion is unusually candid: the findings are equivocal, the evidence supports no recommendation either way, and the authors "are not convinced that further trials of this intervention should be undertaken" and would rather the money went elsewhere.

The same review states, in its opening, that whether any of this bears on the existence of God is a philosophical question outside its scope. That sentence is the most important one in it.

Matthew has the instruction in the negative: pray in a room with the door shut, and "don’t use vain repetitions as the Gentiles do; for they think that they will be heard for their much speaking", because "your Father knows what things you need before you ask him".

A practice whose own founding instruction says the hearer already knows, and warns against volume, is not obviously the practice a trial of added prayer is measuring.

The same collection also promises results. James tells the sick to call the elders to pray over them, and says "the prayer of faith will heal him who is sick" and "the insistent prayer of a righteous person is powerfully effective". That is much nearer to what was tested, though it puts the elders at the bedside rather than strangers working from a list of first names.

Paul writes that he begged three times for a thorn in the flesh to be removed and reports the answer as "my grace is sufficient for you, for my power is made perfect in weakness". The request is denied and the denial is the point of the passage.

On the Mount of Olives, the night he was arrested, the prayer is that the cup pass, followed by "nevertheless, not my will, but yours, be done". Whatever the tradition takes prayer to be, two of its most quoted instances are requests that were not granted.

A group told that strangers were praying for them did worse. The authors say they have no clear explanation for it. They note that postoperative atrial fibrillation and flutter accounted for much of the excess, and that the excess may be a chance finding.

They did not claim to have shown that being prayed for is dangerous, and neither should anyone else. Understanding it, they write, will require further study that, on the review’s advice, will probably not be attempted.

Every arm of such a trial is contaminated: the patients in the control group have families, and most patients in STEP expected, at enrolment, to be prayed for by others anyway. There is no blank condition, only more and less.

Beyond that lies a problem about what is being modelled, and the Cochrane reviewers raise it themselves: a believer "may suggest that the nature of divine intervention could be subtle", more subtle "than is likely to be revealed by the results of a randomised trial". Three critics answered in the review’s published feedback that this reservation is "no more applicable to a scientific investigation of prayer than to trials of any other medical intervention".

This is a separate question with better prospects, though the evidence cited here is on meditation rather than prayer. A meta-analysis of 163 studies found medium average effects, strongest for emotion and relationships and weaker for attention, and it discarded nearly three quarters of the literature it found, mostly for methodological problems.

No trial addresses this and the best of them says so. A null result is consistent with nobody listening and equally consistent with somebody listening who declines to run a randomised trial on request.

That is not a dodge. It is the structure of the claim, and it is why this question is filed as uncertain rather than as settled in either direction.

The Catholic catechism calls prayer a "vital and personal relationship with the living and true God", and the definition it takes from John of Damascus has two halves: "the raising of one’s mind and heart to God or the requesting of good things from God". It lists the basic forms as blessing, petition, intercession, thanksgiving and praise, and defines intercession as asking on behalf of another. It is one of five and the list does not begin with it.

On that account a trial measuring intercession for strangers is testing one form among several, and testing it in the manner the texts are most wary of. That is a fair reading and it is also, conveniently, unfalsifiable, which is worth saying out loud.

The evidence about outcomes for strangers is mostly null, and its own reviewers call it equivocal and grade it low to moderate. It is not evidence about whether prayer is worth doing on any account its practitioners give of it.

A trial of petitions for strangers and a practice its own texts describe as relationship as much as request are not the same subject, though they share a word. Both results are real. They are answers to different questions.