Friday, March 31, 2006

Prayer doesn't heal

A New York Times article started off with a sentence that should come as no surprise to most skeptical thinkers:
"Prayers offered by strangers had no effect on the recovery of people who were undergoing heart surgery, a large and long-awaited study has found."
The study under discussion (Benson et al. 2006) is a multi-year, multi-thousand participant study that examined the effect of prayer on patients undergoing heart surgery. The study treats its subject matter with deadly seriousness; I'm not a medical doctor, but from what I can tell the methods in the study were clean, precise, and exactly what I'd look for in any study of a possible medical treatment.

The study included patients who were undergoing coronary artery bypass graft (CABG) surgery. The study's coordinators asked more than 3,000 patients at six different hospitals in the United States to participate, and they ended up with more than 1,800 patients in the study. The patients were randomly divided into three groups:
  • Patients who were prayed for, but were told that they "may or may not be prayed for." (Group 1)
  • Patients who were not prayed for, but were told that they "may or may not be prayed for." (Group 2)
  • Patients who were prayed for, and were told that they "will be prayed for." (Group 3)
Those patients who were prayed for had their names sent to three different congregations (two Catholic and one Protestant), where congregants were instructed to pray for the patients for 14 days.

The study's methodology had a number of good points:
  • Patients were randomly assigned to each treatment group.
  • Patients were informed of their treatment group via a sealed letter; hospital staff were not informed of the patient's treatment group, and patients were told not to inform hospital staff of their assigned treatment.
  • Data collection and analysis were conducted without knowledge of the treatment group of the patients.
  • The treatment groups were sized to be able to statistically detect 10% changes in outcomes between the groups.
  • The authors collected a tremendous amount of data about all participants (e.g., age, religion, cardiovascular history statistics, operation time) and showed that "[t]here were no important differences in baseline or operative characteristics across the 3 groups."
  • Standard statistical analyses were used to analyze all the data.
  • The statistical comparisons to be made appear to have been decided before the study commenced (a procedure that prevents data fishing).
The outcome the study authors examined was the number of complications each patient exhibited within 30 days of the surgery. The authors analyzed their data by doing two pairwise comparisons:
  • Comparing the two groups (1 and 2) who were informed they might be prayed for; this tested whether prayer had any effect, since one group was prayed for and one group was not prayed for.
  • Comparing the two groups (1 and 3) who were prayed for; this tested whether patient knowledge of prayer had any effect, since one group was certain they'd be prayed for while the other group was not certain.
Figure 2 from the paper
A modified version of Benson et al.'s (2006) figure 2; their statistical analyses (chi-square tests) are summarized in the figure. The black bars at the top of each bar represent the number of patients who either did not undergo surgery or did not complete the study. [Axis labels were redrawn by me for clarity; everything else is original.]

Here's a quick summary of the results:
  • Prayer had no significant effect on the number of patients experiencing complications after surgery (comparing groups 1 and 2, 52% vs. 51% of patients respectively; P=0.67). In other words, prayer did nothing.
  • Patients who knew they were being prayed for had significantly more complications after surgery than patients who were prayed for but weren't sure they were being prayed for (comparing groups 1 and 3, 52% vs. 59% of patients respectively; P=0.025). In other words, knowing you were being prayed for appeared to hurt, not help.
Considering that there's no scientific evidence supporting the existence of a higher power in the first place, the first result is not unexpected. The second result is more interesting; it appears that knowing you are being prayed for actually makes you more likely to have complications post-surgery. This is probably not what religious backers of the study were hoping to find.

The authors did more data analysis than what is summarized above, but everything else follows the same trends identified above.

I am impressed by these authors - they appear to have done a first-class study on a very controversial topic. However, what's distressing is that the study will probably not change anyone's opinion. Those who already believe that prayer is useless didn't need a $2.4 million study to show as much, and those who believe in prayer will probably just rationalize this study away. Examples of this rationalization can be already found in many of the newspaper articles covering this study:
  • "Working in a large medical center like Mayo, Mr. Marek [a hospital chaplain] said, 'You hear tons of stories about the power of prayer, and I don't doubt them.'" (NY Times)
  • "A person of faith would say that this study is interesting," Mr. Barth [spiritual director of Silent Unity] said, "but we've been praying a long time and we've seen prayer work, we know it works, and the research on prayer and spirituality is just getting started." (NY Times)
  • "Dr. David Stevens, executive director of the Christian Medical and Dental Associations, said he believes intercessory prayer can influence medical outcomes, but that science is not equipped to explore it. 'Do we control God through prayer? Theologians would say absolutely not. God decides sometimes to intervene, and sometimes not,' he said." (Houston Chronicle)
  • "'Maybe the people weren't praying very hard,' Higgins said of the study. 'I have no doubts that intercessory prayer works, (just) not all the time.' Even Christ's prayers in the garden of Gethsemane weren't answered, Higgins said. 'At the end, it's God's will be done. ... I don't know how you can measure those things.'" (St. Petersburg Times)
It'd be nice if they had some data to back up their claims.

If there were a good rational basis for why prayer should work (and/or multiple other well-controlled studies showing that prayer did actually help), I could see listening to these critics and attempting to argue that maybe this study wasn't quite perfect and thus we should do more study. However, there is no good evidence that prayer does anything, and not even an evidence-supported mechanism for why prayer should work, so let's just call it a day and focus on remedies that might actually help people.

So, can prayer now be considered an altie treatment?

H Benson, Dusek JA, Sherwood JB, Lam P, Bethea CF, Carpenter W, Levitsky S, Hill PC, Clem DW, Jain MK, Drumel D, Kopecky SL, Mueller PS, Marek D, Rollins S, and Hibberd PL. 2006. Study of the Therapeutic Effects of Intercessory Prayer (STEP) in cardiac bypass patients: A multicenter randomized trial of uncertainty and certainty of receiving intercessory prayer. American Heart Journal 151 (4): 934-942 (abstract)

Thanks to PZ Myers for helping locate the original journal article.

No comments: