This is the study that really caught my attention, in the early 1990’s.  I didn’t really do anything about it until I had an opportunity to move to a new institution that had a large cardiac care unit.  This was Saint Luke’s Hospital in Kansas City.  We decided to perform a trial that would basically replicate, hopefully replicate, Byrd’s trial.  At least give it a try.  These were our co-investigators, several cardiologists, statistician, psychologist here.  Jerry Kolb is a chaplain for the hospital, he was very involved.

And as I said, the purpose of the study is to attempt to replicate Byrd’s trial. Certainly, in science, if you get one study and get an outcome, and nobody replicates it, you really don’t have much confidence that what they found is real.  Until it can be replicated in some other setting, by other investigators, using other people, other methods.  Science is very dependent on replication.  So, we wanted to try and replicate Byrd’s study.  But we wanted to use what we thought would be an improved design; didn’t really like his “good,” “intermediate,” and  “bad” scale; thought something could be better, improved.  We also hoped to develop a new tool to assess CCU hospital course.  It really wasn’t a tool at a scale that we could find, that would do this. We wanted to be able to summarize in one number how people did.

The hypothesis of the trial, going into it, was that blind, remote, intercessory prayer, for patients in the coronary care unit, would result in fewer complications, shorter length of stay.  That’s what we were proposing.  Of course, blind prayer means, the people who did the praying didn’t know who they were praying for, the people who were being prayed for did not know anybody was praying for them.  And remote, of course, means outside of the hospital.  Intercessory prayer means to say a prayer for somebody else. To intercede, for somebody else. 

Our primary end points, the things we were primarily interested in measuring, were relatively fewer complications.  We created a composite score, which I’ll describe in a minute. We had a weighted, and an unweighted score.  Severity weighted, severity unweighted. And I’ll explain that.  The shorter stays, simply we were looking at the number of days in a CCU Unit, the number of days in a hospital.  CCU stands for Coronary Care Unit.  So these were the primary outcomes we were interested in measuring.   

This is the way the study flowed.  We began, it took about a year, fifty weeks, of randomizing, folks.  People who were admitted to the coronary care unit at Saint Luke’s Hospital, about a thousand patients over a fifty week period were randomized, or, excuse me, were admitted. We decided to eliminate some people, up front, to begin with, that we weren’t going to consider.  There were a few that were weight listed for transplant medium, meaning they were admitted to the hospital, admitted to the Coronary Care Unit, their heart is dying; they are waiting there for a transplant.  And many people don’t ever get into to die in a hospital.  But we know that anybody who is going to be weight listed for transplant is going to be there for a long time.  At least that’s the history.  And one could make a case that it would be interesting to pray for those people.  But we didn’t. 

So they were eliminated.  We then randomized, or randomly assigned by, based on medical record number, even and odd, people to the prayer group, or to the usual care group.  And we ended up with 484 randomized to prayer, and 529 to usual care.  I would have thought we’d be even more even than that, but that’s the way it came out.  This P value here tells you that there’s a one chance in five of getting that kind of distribution.  So it’s not an unusual distribution. 

The other people we eliminated up front were folks that we said if they were admitted to the Coronary Care Unit for less than one day: in and out, in the morning, out in the afternoon, that we weren’t going to consider them in the trial.  And that’s primarily because we couldn’t start the praying for them in that short a period of time.  Because telephone calls had to be made, people had to be contacted, et cetera, et cetera.  So we had reason up front that we weren’t going to deal with the people who admitted for less than a day.

As it turned out, and I think Dr. Tessman may have noted this, and other people may have noted this, we didn’t see this thing.  Point of fact, that there were eighteen people who were admitted for less than one day into the prayer group, five who were admitted for less than a day in this group, and that is a highly statistically significant difference.  There are more people now.  I have not yet been able to, for a variety of reasons, find out what happened to these people.  So, for the purposes of this discussion, they’re just out, we didn’t include them in the trial.  So our final group, is here, the final prayer group, at 466 patients, the final usual care group, 524.  And those are the people upon which the data I’m going to describe to you is collected.

Who are the intercessors?  We recruited folks from the christian community, who worked at Saint Luke’s, it’s an Episcopalian hospital.  Lots of contacts, of course, in that community.  It was easy to use those kind of people, and it was also, as I said, our intent was to replicate the Byrd trial, and Byrd used just christian intercessors. This meant obviously a lot of discussion about why we couldn’t use other people. And one certainly could use other kinds of believers in different denominations, in different belief structures, and actually one of the studies I’ll show you did that.

The intercessors randomly, and blindly, were organized into fifteen teams, of five people each.  They didn’t know each other, they’d never met each other, the people who were on these teams.  One person was picked to be a team leader, that’s the person who got the phone call.  In picking intercessors, we didn’t want to have any particular denomination necessarily, but we did ask them to believe in this statement, or at least affirm this statement: “I believe in god.  I believe that He is Personal, and is concerned with individual lives.  I further believe that he is responsive to prayers for healing on behalf of  the sick.”  People who like to do intercessory prayer obviously believe things like this. So this was not a hard sell.

How did we do this study? The protocol for the trial was as follows.  Every morning in the hospital the computer system of course tells you what patients are on what floor and what units.  And the chaplain’s office, of course, has a computer.  And the chaplain needs to know who’s in the hospital, so they can go visit people who want to be visited. So we have a list every day that comes up on the computer in the chaplain’s office.  The study coordinator, who was the chaplain’s secretary, assigned patients to prayer or usual care, based upon the last digit of their medical record number, whether it was even or odd. She was the only one who knew the even/odd code.  She did not have any contact with the patients, she did not have any contact with the doctors, she did not have any contact with statisticians, she didn’t even know where the Coronary Care Unit was of the hospital. The chaplain’s office is off in a corner somewhere, she doesn’t know. So all she knows is to assign prayer or usual care. If someone was assigned to the prayer group, she looked on her list, she found what the next prayer team was that needed someone to pray for, she would call the team leader on the telephone, and give them the first name of the patient.  “Bob.” “Mary.” “Susan.” Whatever.  Then the team leader would contact others in the team by telephone, there’s five people on a team.  And they all put the name down on a little log sheet.  And we asked them to pray for twenty-eight days.  Regardless of what happened to these people.  May’ve died the next day. Didn’t make any difference. They were going to pray for twenty-eight days.  Too logistically complicated to tract every single patient and get back to people when they should stop and start, so we just said, “pray for twenty-eight days.” 

And what did they pray for? They prayed for speedy recovery, with no complications.  Anything else seemed inappropriate for that.  That’s what we asked them to pray for.  There are some that would say that in this kind of research what you ought to do is pray undirected prayer, a “thy will be done” type prayer.  We choose a directed prayer, a prayer for healings specifically.  Again, partly because we were trying to replicate Bird’s trials, what he did,  and secondly, we figured people are in a hospital because they want to get well, so, ask for healing.

How did we assess how they did at the hospital? This is the important part of the study. How do you tell how people do?  The way we did it, was we looked at basic charts blindly after patients had been dismissed from the hospital.  We tallied all the new events, procedures, prescriptions that occurred after the first day, we see them.  Remember, the first day we said, after the first day, we aren’t going to take any data from that first day.  Because prayer couldn’t get started till after that first day.  And as it turned out, it took one-point-two days on average for anybody to start praying for somebody.  So, we eliminate all the first day evidence, or events, and then from then on, we track what happened.  And based on this, we determined a score, and we call it the MAHI score, which I’ll explain. MAHI stands for Mid-America Heart Institute, that’s the heart hospital that’s part of Saint Luke’s.  It’s a continuous variable, meaning it could be from one to fifty.  In integers, that’s continuos variable, it’s not a good-intermediate-bad-yes-no. It’s a scan of variables. Which is a better way to analyze them statistically if you have this kind of data.  The idea was a high number, is bad, a low number is good, play golf.  And the events, every event that happened to someone was assigned a value, and the values were added together to get a score. And this is how it worked.  This is kind of small, I hope you can see this.  The idea is that here we have five categories, one, two, three, four, five, six, and then we have death.  Six categories, excuse me. And these were taken to some extent from Byrd’s paper, also, from a knowledge of the kind of adverse events that happened in our hospital.  And our cardiologists ranked all these things, and they came up with a scale.  So, if you had any of these events happen to you, number one, that was kind of low rent stuff, not a big deal, you get a small room for it, these are more serious, these are more serious yet, more serious yet, cardiac arrest, more serious, then death, of course, is the most serious.

The scale’s intuitively reasonable. It hasn’t been validated. This is a criticism (it is a valid criticism) of the study, meaning that we don’t know that someone who has a high number actually does badly over time. We presume that would be the case.  Important thing to remember is that this scale was applied equally to both groups. So if there’s a problem with the scale, it’s not going to upset one or the other. It’s valid with both groups, if it’s applied blindly and evenly in both treatment groups.

How did this work? Just to give you an example, two examples. Here’s one patient, said that they, after one day in the CCU unit, they developed unstable angina, they developed heart chest pains, and that gets you one point.  They were treated with an anti-anginal agent, they were given another point.  That means the angina was bad enough to actual require being treated.  They were sent for heart catharitization, where they stick the dye into your heart, the blood vessel, to take a picture of it.  And then they were successfully revascularized, meaning that they had a balloon, stuck into their artery, pumped up, and opened the vessel.  And that’s worth three points.  Scandena, you get one, two, three and previous six points, so that person gets six points.

The unweighted score for this person: this is the latest score, severity adjusted.  Unweighted score is four.  Because there’s simply four events, one, two, three, four.  So their weighted score is (inaudible) . . .  Count the events, or count the events and give them weight. 

Another example would be here a patient who had a cardiac arrest, it’s five points for that, and dies, six points, so it’s eleven points.  But the unweighted score is only two. So that’s the reason for doing a weighted score.  I think it makes somewhat more sense, that this is a somewhat more serious set of events.  Now one can philosophize all you want about how this, does this make sense or not, should death be twenty, should death not even be in it.  Some people think, at one point, another death is not a bad outcome, depending on the situation. You can talk about all this, but, we did it this way, and it seemed to work out.

Statistical evaluation: this is how we did our statistics.  Categorical data was analyzed by chi square efficiency exact test, and this is, basically, how many people had another heart attack, how many people needed a balloon, how many people needed a diuretic for strengthening. And this was adjusted for multiple comparisons, because there were several, three or four different events that we looked at.  So we set the P value high, to make sure that we weren’t just finding things by chance. 

We did try to do the Byrd score. We tried to replicate what he did, as best we could. We don’t really think this is a great test.  But we did it anyway, thinking it would be appropriate to do that.  P values of five.  Continuous data, this is primarily our, primary end points, mean of point-oh-five is required, this is a composite test.  Attention to pre-analysis means we analyzed everybody who was randomized into the study.  Regardless of whether they finished it or not.  Now I know you can’t see this, but, long as you can see it, you can’t read it. This is a list of how sick were these people to begin with. You do a study like this, you’ve got to be sure that that the two groups are equally sick, to begin with.  Because you’re going to look at outcomes down the road.  And if one group is sicker than the other one, naturally the outcomes going to be worse, in the one that was sicker to begin with.  When you randomize four or five hundred people in two groups, you get pretty even groups.  And this is what we got.  Without obviously going through all this, none of these P values is statistically significant.  They’re all quite high.  And the per cent of patients that have all these different events is about the same. 

If anyone wants copies of the paper, I’ve got copies of the paper.  And I mean it,  look over this stuff.

No individual component of this MAHI-CCU use score was significantly different between groups. Meaning that, these are all the things that happened to them after they got admitted to the hospital.  And, the percent of people in each group, the usual care, and the prayer group, was tabulated here.  These are the individual events.  It was out of these events that we calculated our composite score.  The only thing that came close to being statistically significant was this one here, which was a certain kind of catheter, point-oh-three, but we said that’s not statistically significant, because we set our standard at point-oh-oh-five, which means it’s much higher bar one has to jump to find significance.  So we say there’s no significant difference in any specific event. 

This is the primary result of the trial.  The MAHI-CCU score, this is the severity adjusted score, for the usual care group, was six-point-nine-seven, six-point-two-four for the prayer group, this is about eleven percent difference, that’s the effect size.  And that P value is point-oh-five.  Means there’s one chance out of twenty, that this is just a random effect, not a real effect.  The unweighted score, is three versus two-point-seven, that’s a ten per cent difference, remember, this is just a count of things that happened, average number of things that happened.  Point-oh-four was the significance there.  These are considered statistically significant in all bio-medical research.  That level, point-oh-five, or below, means that it was statistically significant.  The chances of it happening by chance, are pretty small, and we will accept that that uncertainty and say that this is a real thing.

Length of stay? Length of stay was not different.  See here that the P values were over point-oh-five then the difference of finding is not statistically significant, could have happened by chance, pretty good chance that it happened by chance.  So we say no effect on length of stay, either total hospital length, or the CCU length of stay.  So the only effects we found were on these medical course scores. 

Byrd’s score?  We again have the usual care/prayer group, we did the same thing he did, tried to organize them into good-intermediate-bad.  As we can see with usual care, the prayer group had slightly more, 67% versus 65%.  This is not statistically significant, but there were a few more goods,  the same intermediates and a little bit less bad, 19% versus 22%, in the prayer group.  But by this test, it was not statistically significant, so we did not confirm Byrd.  Using Byrd’s score we did not find the effect.

Differences in design between our two studies I think are interesting to point out.  The blinding: in Byrd’s study, patients were blind to treatment, meaning they didn’t know if they were getting prayed for or not, but they knew they were in a study. Because they’d signed a piece of paper saying “I agree to be in a study.” In our study, they were blind to the existence of the trial.  And this was a point of some controversy.  But we felt . . . it’s always a balancing act in science.  It’s better that people don’t know that they’re being prayed for, don’t know they’re being intervened upon.  The other side, is the requirement for informed consent. We went through our institutional review boards, and they agreed to let us do the study, without telling the patients, in light of the fact that they thought that somebody praying for them five miles away from their home was probably not going to be associated with any averaging of the facts.

So, consent was gotten in this study the same (unintelligible) as not contained in this study. Information of the intercessors: they were given updates of how people were doing in the Byrd study; we didn’t give them any feedback.  Different kinds of assessments in our sample size was about longer or twice as much, as Byrd.  Couple of caveats of the  trial need to be appreciated.  To the extent that patients were being prayed for, by family, by friends, by clergy, whatever prayer was offered in this study was supplemental. There’s this unknown, an uncontrollable, an unmeasurable amount of background prayer for everybody in this trial.  Both groups.  We assume when you randomized four to five hundred people into two groups, that the amount of background prayer is the same. A fair assumption, we think. I mean, that’s the whole reason for randomizing.  So what we’re looking at here is the addition of some supplemental prayer on top of  whatever noise is going to run back there. 

Second, is the fairly low dose of prayer.  You think about this in terms of a medication.  Three to five people, kind of, praying their first name, for twenty-eight days, out, and not knowing you or not knowing anything about your family, or what you’re like, or no emotional contact whatever.  I would consider that sort of low dose prayer, as opposed to different kinds of prayer that can be administered, that are much more direct. Limitations to the study, as I mentioned, the score has not been previously validated, we didn’t do studies first to show that score means something.  It was an intuitive start on trying to get some kind of chi-rational score.

What’s the clinical significance?  What does a ten per cent reduction in score mean, in terms of patients. What does it mean to a health care system?  We weren’t able to track costs, costs would have been an interesting end point, as well, to save money.  Certainly, the chaplain’s office would like to know, if praying for people saves money, to help justify their existence. Of course, another limitation, there’s no known biological mechanism.  Or physical mechanism, that can explain this finding, and so this leaves them a little bit up in the air. Nevertheless, I don’t think the findings themselves are at risk.  Beyond that, there’s a lot of little questions that come up.  Let me do . . . Patrick, when did I start, do you know? Fifteen? All right, I’ll try to wrap it up in fifteen. 

I want to go over this third study.  This was published in 1998. This was a study done on AIDS patients.  Again, this was a very small study, only forty patients. Twenty and twenty.  Two groups.  So this is pretty small.  This was a distant healing, called D.H. This is not just prayer, this is distant healing.  And they were, and what happened is, folks were randomized, who had AIDS, into two different groups. And then there were distant healers, from all over the country, working on the people in the distant healing group, for ten weeks, five days a week, one distant healer at a time. They kind of mixed it all up.  And they measured impact on medical and social kinds of things. 

Who were the healers? Distant healers were recruited from professional associations and distant healing schools.  I didn’t know there were distant healing schools, but apparently there are.  So they found these people, to do this.  The religious traditions, christian, jewish, buddhist,  american indian, shamanic, Secular, Non-Religious perspectives.  And they practiced distant healing one hour a day, for six days a week, each patient had five healers.  And, these were the outcomes.  So many outcomes, couldn’t get them all on the slide. But if they look, here’s the distant healing group, twenty people, here’s the control group, twenty people.  Outpatient visits, visits to the hospital, over the ten week period we studied.  Nine in this group, thirteen in this group, that was statistically significant, lower in the distant healing group.  Number of hospitalizations per patient, point-one-five-point-six, it was four times more hospitalizations, in the group that didn’t get . . . how do I say it . . . distant healing assistance.  And again, it wasn’t just prayer.  Days in the hospital, point five versus three point four, significant difference. Illness severity was significantly lower.  C D four cells, these are certain white blood cell types that measure in AIDS. There was no difference, no effect in that case. P value here, which shows there is no significance difference, it’s just noise.  This is a PMOS (I have to look) Profile of Mood Spacey Habits.  A survey, of  “how ya feelin,’” you know. Before and after, this went down in the healing group.  And it went up in this group.  And that was statistically significant.  And, MOS is Medical Outcome Scores, another medical test. Improved a little bit, unimproved there, not statistically significant. 

So, here’s a trial with just forty people in it, where there was distant healing applied, whatever that means.  And it appears to have had an effect.  That, statistically, it would be improbable. 

This is what I want to wrap up on, this is a summary of distant healing trials.  This was published in 2000, The Annals of Internal Medicine.  These folks, these are from the University of Maryland.  They looked at, they went to the literature and picked all the trials, they thought, of the best studies, the best designs of all the trials. And they said “okay, they’ve got to have blinding, they’ve got to have double blind, control, all this kind of stuff.”  They found twenty-three trials, twenty-three studies, five in the intercessory prayer category, eleven with therapeutic touch (this is where people are, [I’ve never seen this done] I guess, they stand over the body and they draw stuff out with their hands or something like this. I don’t know how it works).  But it’s definitely alternative medicine, to be sure, as are all of these.

Other distant healing techniques, like the one I just presented, there were seven studies.  They found that of these twenty-three trials, thirteen of the trials, fifty-seven percent, reported statistically significant benefits to the patient remaining in the trial. Nine of them, or thirty nine per cent, showed no benefit. And one, four per cent, had a statistically significant detriment.  So overall, actually, there was no effect whatever.  All this stuff is completely sham.  There’s nothing to it. You would expect that ninety-five per cent of the studies would show no effect.  And that maybe one of the studies would show positive effect and one of the studies would show negative effect.  That’s statistically what you would expect.  To find that fifty-seven per cent of them showing benefit, suggests that there’s something here.  I’m not going to go over all that, they’re conclusions were, methodological limitations of several studies, make it difficult to draw definite, definitive conclusions, about the efficacy of distant healing, however, given that approximately fifty-seven per cent of the trials showed a positive treatment effect, the evidence thus far merits further study.  And I think that’s where we are today, that this is an area that merits further study.  Clearly, if distant healing, if prayer can impact health, we ought to know about it.  And we need to learn about it.  And figure out how it works.  And how to make it, apply it to help our patients do better.  I think that, a slide that suggests that testing complementary medicine techniques is going to continue, in the future.  Okay. I’ll stop.

(Applause)

Moderator:  Thank you, Dr. Harris.  And now for a response, I would like to introduce Dr. Irwin Tessman, who holds a Ph.D. from Yale, and is professor of biological sciences, at Purdue University. His primary areas of scientific research are molecular genetics of DNA repair, utagenesis, recombination, and transposition in viruses and bacteria. And the evolution of altruism, by natural selection.  Dr. Tessman, criticized the procedures and conclusions of Harris, et al.  In an article co-authored by Jack Tessman, for the March/April 2000 issue of Skeptical Inquirer, entitled “Efficacy of Medical Prayer: A Critical Examination of Claims.”  Dr. Tessman . . .

But, Galton showed you don’t have to be interested in the mechanism.  And, there’s a famous story, about the physicist Niels Bohr, some of you may have heard it. It’s probably fictitious. Niels Bohr was the inventor of the quantum theory of the atom.  And a visitor to his office in Copenhagen noticed that nailed over his door was a horseshoe, with the open end up, proper orientation to bring good luck.  And the visitor said, “surely, Professor Bohr, you don’t believe in such superstition?”  And Bohr said “of course not, I don’t believe in the superstition, but fortunately for me, I’m told that it works, whether you believe in it or not.”  And the idea is, you don’t have to worry about what the reason is, the question is, does it work. And the same applies to intercessory prayer. I’m here to tell you that I don’t think it works. 

Galton responded to a challenge that was given in his time. Someone said, “can anyone design a method for testing whether prayer works?”  And Galton was a very ingenious person.  He’s one of the founders of the field of statistics.  So he was eager to show off the field of statistics, in solving this problem, in approaching this problem.  He was also the founder of biometry, the measuring of all sorts of biological things, and some of which I’ll mention.  He pointed out that public prayers are most often said for royalty. They have prayers like, praying for healthy life and of long life.  And Galton said “aha! all this prayer for a long life for royalty; that should extend the life of royalty.”  And so he went about looking for data on the subject, and he found the following.

Now don’t despair if you can’t read this.  I’m going to read it for you.  The data on the mean age attained by males of various classes but all affluent classes so they all have the advantages of affluence.  And they all, the ages that they attained may seem high for you for the Seventeenth and Eighteenth and Nineteenth Centuries.  But the only people considered were those who lived at least to the age of thirty, so infant mortality and the youthful indiscretions did not affect the ages that these people reached. 

The first line, are members of royal houses.  He had lots of data, good numbers, ninety-seven of royalty.  And they lived to be average age of sixty-four years.  He also examined numbers of clergy, lawyers, medical profession, English aristocracy, that’s pretty close to royalty.  Gentry, trading, commerce, officers in the Royal Navy.  You’d think they wouldn’t live very long in the Royal Navy. English literature and so forth.  Members of the royal houses had the lowest mean age.  Not only did they not live longer, they lived shorter.  The prayers that were offered seemed to be counter-productive.

Well, this is not a very definitive study.  And, British royalty, to this day, ignore this study.  They defend their privileges. You might be interested in knowing that when Diana and Charles got divorced, Queen Elizabeth saw to it that her name was removed from the list of beneficiaries of the public prayer. 

Galton looked for other data.  Rather interesting what he found. He looked at the premiums that were paid to insurance companies, or the premiums that the insurance companies charged to insure ships. And Galton reasoned, that missionaries and pilgrims prayed a great deal.  And ships that carried missionaries and pilgrims should have a great deal of prayer for the safety of the ship. And therefore the insurance premiums should be lower for such ships.  But apparently the insurance premiums were exactly the same.  Insurance companies found that ships carrying pilgrims and missionaries sank just as often as other ships.  So apparently the prayers didn’t help there. 

Well, I mention all this about Galton because he set the stage really.  He made it respectable to study supernatural phenomenon.  He made it respectable and he showed that if you make a scientific issue out of this, you can study it statistically.  And that’s what Prof. Harris has done, and Dr. Byrd that he referred to, in 1988. 

Randolph C. Byrd worked in San Francisco General Hospital.  And he studied, as Professor Harris did, the people who were admitted to the coronary care unit in San Francisco General Hospital.  And his work was labeled a landmark by the journal Alternative Therapies in Health and Medicine.  They republished his paper, which was originally published in the Southern Medical Journal.  They republished it in 1999 to honor the paper, because they felt it set a mark for future studies. And indeed the two studies Professor Harris talked about followed Byrd.  And Professor Harris’ study was meant to imitate Byrd and check it carefully.  And Byrd’s study, as you heard, had a large number, a fairly large number of patients, three hundred and ninety three.  And the study was planned to be double blind, Professor Harris told you how his studies were double blind.  Double blind means that the physicians, the nurses, anyone on the staff didn’t know whether the patient was in the prayer group, the test group, or whether they were in the control group.  The people, the doctors who evaluated the results, were not to know who was in either group.  No one was to know except someone who kept the code, who identified which person goes into which class, the test group or the control group.  To give an example of this, this is what was written (I’ll read it out for you). This is how Byrd describes the double blinding: “Patients, the staff, and doctors in the unit, and I (that’s Dr. Byrd) remained blinded throughout the study.  As a precaution against biasing the study the patients were not contacted again.”  That’s very important, because you don’t want any bias.  If people know which category you’re in, it’s not a question of just honesty, but a question of being consistent.  It’s very difficult to be absolutely objective if you know who is in which category.  So this double blinding is very important. 

But unfortunately, it was not double blind. . . .

   (end of side one of first audiotape; beginning of side two of first audiotape)

Well, there was another problem with the blindness, which I will get to in a moment.  Well let me go over . . . you’ve seen the results before, but I’m going to go over it again, to make my point.  Here is the table that Professor Harris showed.  And its not focused in the light. Hmm. Here we go. Well, you can’t have the whole thing in focus.  And I’ll start in back.  All right, the first thing we see is the first question that Professor Harris answered, “did you speed up the recovery?” And that’s the subject, and that’s the subject of our debate, that I’ll be flexible and say there’s more than just that in our debate.  Days in the coronary care unit after entry.  Days in the hospital after entry. The numbers are very similar and they’re not statistically, the difference between them is not significantly different.  And they were given medications on discharge, and the number of medications given to those on intercessory prayer, the test group, and those given to the control group, are slightly different but not significantly different.  Then there were all these other . . . So that answers one question that Professor Harris asked and that I’m asking. “did you speed up the healing of any of the people, of the people, in the test.” And the answer is, the speed was not increased.  There was no improvement.

Then there were those lists of new problems, diagnoses, therapeutic events after entry.  And there were twenty-six of these. And, I believe, Professor Harris had about thirty-five. And covered all of the ones here.  He gives the per cent of events of those in the test group, eleven per cent, twenty-one individuals, ten per cent in the control group, nineteen individuals, nineteen to twenty-one, ten per cent, eleven per cent,  not statistically significant.  The P value was a large number, less than one, but between point-oh-five and one, probably around point-four, point-five. I’m going to get back to this.  I want to emphasize the meaning of the P value.

Anything less that are able than point-oh-five is considered significantly significant.  That’s an arbitrary point.  But it’s a reasonable point.  And I’ll explain why in a moment. It was six conditions. Here is congestive heart failure.  Where the test group fared better, statistical significance in the control group.  Here’s mortality.  They did about the same. That’s a very simple thing to measure. And there are six of these. Now what does . . . I’ll tell you what I think is needed to understand about the P value. Let’s take a P value of around point-oh-five. What it means is, if these two numbers were really the same, but due to the limitations of the measurements, fluctuations and so on, pure chance, they came out different.  One was three per cent, and one was eight per cent.  Could that have happened purely by chance?  Of course, anything could happen by chance.  The further apart the two values are, the less the chances are that that would happen purely by chance.  Could happen.

But when it gets so low, as in point-oh-five, that’s a five per cent chance, once in twenty that will happen purely by chance.  That’s low enough to say, that’s a significant difference.  It’s a marginal difference.  But it’s a significant difference. Point-oh-three is a little bit lower than that, so it’s a little bit more significant.  Point-oh-oh-five, five chances in a thousand, that that would happen purely by chance, that’s unreasonable. This must be really different.  Point-oh-oh-two: this must be really different between the test group, zero per cent, and six per cent.  So they found significant differences, in all these new problems, all these events that happened during hospitalization.  Not very many differences.  Not many statistically significant differences.

But as Professor Harris pointed out, in almost every case, there was a small advantage of the test group over the control group.  Not individually big enough difference to count statistically.  But when you add them all up, as he did, you see that there’s significant difference of the test group from the control group.  And they found, by analyzing, I won’t go into the method they used.  But by analyzing all the data together, by what’s called a multivariate analysis, they came up, saying, that the probability that you’ve got these two sets of results, is less than point-oh-oh-one. Very improbable that it would happen by chance. So this must be a significant result in favor of the test group, those that were prayed for. 

Then something interesting happened in his study. He wrote up these results, sent it in to a journal, and they sent it back, saying “interesting study, but we’d like you to do a further analysis. We’d like you to score the course of events in the hospital.” And physicians like to do this.  They like to talk about “was the hospitalization stay good, with the course of events good, were they intermediate, or were they bad?”  And he set up a scoring system. The course of events was good if only one of the following occurred: left heart catheterization, and so on.  All these events that are listed here: ventricular tachycardia, heart blocked requiring temporary pace maker, and so on.  And you can’t read this very well, but I’ll read it, well, I’ll try to focus it . . .

(brief breaking in recording, then voice resumes)

 . . . individual patients to having had a good, intermediate, or bad outcome, he knew who they were, he knew which class they were in.  So this table was not done blind, he knew who was who when he did the table.

Well, to summarize, what Dr. Byrd found, he found that the length of stay in the hospital was unchanged, whether you were prayed for or not.  That is one of the significant things one wants to look for.  And that’s very straightforward, does not involve any intricate statistical analysis.  You know when they came in, when they went out, and the average there.  He compared different conditions that occurred in the hospital that the patients suffered.  And found in most cases there was no significant difference except in six cases that I pointed out.  And he did this very important scoring of the course of events in the hospital.  And found the difference between the prayed group and the control group at a level of P value of zero-point-oh-one.  I could have erased my comments that I had on the manuscript, but I wanted to show you that I was diligent enough to control a miscalculation. 

But then there’s the Janet Greene problem.  We don’t know how much interaction she had between the staff and the patients. She keeps going back into the wards.  Someone’s going to figure out which patients are in which group.  Because she’d have to ask for the records and so on.  That’s a problem today.

Well, something interesting happened at that point. The person who did the statistical analysis at San Francisco General Hospital, who I was in contact with, said he found the original tapes that contained all the data for this study.  And I asked Dr. Byrd, would he mind if I had temporary access to this data, I’d like to re-analyze it.  He thought about it for a while, and then said he’d rather not.  And the reason he gave was, I think, was a reasonable one. He said, “it’s unusual for people to ‘rummage through your notebooks’ so to speak,” and he didn’t think there was any reason why people should rummage through his notebooks.  I would not like it for people to rummage through my notebooks, so I cheerfully accepted his reasoning.

Except for one reservation.  I didn’t tell him that reservation. But I’ll tell you my reservation.  And you’re going to be skeptical of what I’m going to say, but I mean it sincerely.  This was not the usual situation. This was quite an unusual situation. If what Dr. Byrd claimed he found was true, then he found evidence for a supernatural event.  Sometimes called paranormal event. If he found proof of a supernatural event, then that was the greatest scientific discovery of the last century.  Better than the special and general theory of relativity.  Better than Watson and Crick’s discovery of the structure of DNA. That was something unheard of.  People claim to have evidence, but he had hard scientific evidence.  And if you have something of that sort, then maybe people should rummage through your notebook and look at it.  Anyway, I don’t consider what I’ve said to be hyperbole, I really mean that that would have been the greatest scientific discovery of the last century. 

Dr. Byrd said something that resonated with me.  He says he’s done his job, he’s done the experiment, he’s set the path, why don’t other people now come and try and reproduce his results.  And if they can’t, if they don’t reproduce it, fine. And maybe he was wrong. If they do, that’s the confirmation that any scientist looks for.

And that brings us to Professor Harris and colleagues' work. They did set out to reproduce Byrd’s work and followed his protocol in many ways.  Except I believe they really did have a double blind study, whereas I don’t feel that Byrd did have a double blind study.   I made up my overhead before I heard Professor Harris’ talk of course.  And I’m emphasizing certain things that he doesn’t deny.  But I took the precaution of preparing an overhead. 

Well, lets go back to that table that replicated Byrd’s study.  This time, it’s the Harris, et al study.  I counted thirty-five, I may have been off by one.  Here we have the control group that (the order is reversed, from Byrd’s table). Control group comes first.  The test group comes second.  And here are the P values.  The probability that the difference could have happened by pure chance. And here is sixty-two per cent, that’s very high. Means it could easily have occurred by pure chance. And so forth.  And there are no significantly different . . . well, here’s one, point-oh-three.  But one out of thirty-five itself could happen by pure chance.

What happened to those six that Byrd found?  Let’s look at that again.  Did this confirm or did it contradict Byrd’s study?  Seems to contradict. Here are the six events that were significantly different in Byrd’s study. Point-oh-three, that’s not so significant, that’s marginal.  Point-oh-five is marginal.  That’s the line. The arbitrary line, of course there’s no sharp line. But that’s the arbitrary line that we set between, to say, perhaps it’s significant.  Point-oh-three’s a little better.  But Harris’ study finds point-six-oh, meaning it could have happened by pure chance.  Nothing statistically significant about that.  Same with point-oh-five, point-eight-nine. You may not be so impressed by these, but when you get to point-oh-oh-five, that’s very unlikely to happen by chance.  But, yet Harris will find it could easily happen by chance. Their data was very different.  Let me say, not because he’s sharing the podium with me and the colleagues here.  I think his study was better because he used far more patients in his study.  He had a thousand patients compared to the four hundred of Byrd.  So I think their data is better but it contradicts in every case the results that Byrd got.  So it puts in doubt, in my mind, the results that Byrd got.  And I keep thinking of the Janet Greene effect.  If any of you know Janet Greene, please tell her I would like to speak to her (laughter in the audience.)

Well, these data were analyzed further, and you’ve seen this, in Professor Harris’ presentation.  I want to go over it and I will give it a different spin.  Here is the weighted score that they used.  They were weighted for all those results and there’s some controversy about whether that’s a valid weighting or not.  I don’t care about that controversy.  The probability of that happening by chance is point-oh-four.  I call that marginal.  It’s marginally significant.  The unweighted score; these are the same sets of data just analyzing different ways, so it’s really just one result with a point-oh-four. Claims of staying in the hospital, major stays in a cardiac care unit: no significant difference besides the point-three-six.  He had that right.  No significant difference.  So, one of the two turned out to be no statistical significance.  One turned out to have marginal statistical significance.  When you do two studies of that sort, it’s not so significant that one of them was marginal, when the other one was not at all.

Can you remember the table that Dr. Byrd was asked to make up, like the referees of his original manuscript, where they want him to analyze the data further and come up with a scoring of the course of events that occurred in the hospital.  Was the course good, was it intermediate, was it bad.  And as you remember, there was a difference between the two groups.  And here are the criteria. Professor Harris and his group adopted the exact same criteria, I believe, but they came out with a different result on their data.  Same criteria, but their own data. And they found P equaled point-two-nine between the two groups.  Byrd found P less than point-oh-one.  Pretty significant.  Harris said all find P value point-nine-nine not significant at all.  So these two studies do not agree in most cases. 

Well, to summarize, I hope you got the feeling that the results were mixed. Both Byrd and Harris agree that the prayers did not speed up the recovery.  They stayed the same length of time in the hospital.  They both agree that the list of events that occurred in the hospital favored, or disfavored, those who were not prayed for, those who were prayed for seemed to have fewer events occurring in the hospital.  But when you look at the overall course of events they disagree again. Byrd found that the overall course of events in the hospital favored the prayer group.  Harris and his colleagues found that it had no effect, no statistically significant effect.

So I think that after all this work the evidence is very slight, where there is evidence, a marginal point P value of point-oh-four in favor of the prayer group.  But most of the evidence is against any significant effect.

Carl Sagan said “extraordinary claims require extraordinary proof.”  Just because he said it doesn’t make it so.  But I suggest to you that it’s a reasonable approach to take.  I think these are extraordinary claims, as I indicated before.  Again, I’ll repeat, I meant it seriously.  If these results are true, they suggest the existence, they prove the existence, of a supernatural power. I don’t deny there is or isn’t, I don’t see any is or isn’t, but this would be scientific proof.  And that’s an extraordinary claim.  I don’t think there’s any extraordinary proof.  I don’t think there’s really any significant proof.  And that’s what I attempted to show.  Thank you.

(applause)

(voice begins abruptly)

And I think there’s actually a study that’s coming out of Duke, that’s been accepted by  The American Journal of Cardiology, that is a pilot study looking at different modes of complementary medicine. They found in that study that the best effect, this is of people who’ve had balloon angioplasties, a group of a hundred and fifty, and they’re going to report in that trial that there was the, although not statistically significant because they had five different treatment groups, the best effect was seen in the group that got intercessory prayer. And this is a study that they doing to basically going to develop pilot data so they can enlarge multi-center clinical trial with intercessory prayer. And Herb Benson is a, from Harvard, Mind/Body Institute will be presenting data, in the next several months of a study he’s been doing, in cardiac bypass surgery, with prayer.

 And so, we’ll see, they’ll be more data coming.  I think that it’s important to keep looking at this stuff, as an effect. You don’t have to know a mechanism to study whether phenomena is real or not.  There’s no point in looking for a mechanism until you know a phenomenon is real.  Then you can start exploring mechanism.  We might not be able to get to the mechanism on this one, in a scientific way.  That’s fine.  Let’s just go as far as we can.

 I wanted to put this up.  It’s a quote from the editor of The Archives of Internal Medicine regarding our paper.  He was interviewed in a newspaper when our paper was published. And he said it’s a very well designed study, well written, “in this particular study, patients benefited, if this had been a medication, the conclusion would be that this medication helped.” I think that was a very honest statement from him. And this is true. There’s plenty of drug trials in the medical literature where the difference between the treatment group and the drug group is less than ten per cent.  But it’s statistically significant, and it helped in a positive study.  So I think he acknowledged that it is a positive study.

 I’m going to skip that.

 I’m going to skip that.

 I want to show a couple three quotes to end. Now we’re in on my closing comments here. We’re looking for closure. 

A quote from Elizabeth Targ, who was really the primary driver on that study with AIDS patients I told you about.  The forty patients with AIDS, she was the primary researcher in this.  She said: “No experiment can prove or disprove the existence of god.  But if in fact mental intention can be shown to facilitate healing at a distance, this would clearly imply that human beings are more connected to each other and more responsible for each other than previously believed.  These connections can be actuated through the agency of god, consciousness, love, electrons, or combination of god knows what. The answers to such questions await further research.”  That’s a great attitude. We don’t know how this could happen but further research should be done.

Here’s a quote from William Crookes, who discovered thallium (didn’t know about this guy, this is from the Nineteenth Century, he’s talking about Michael Faraday): “to quote Faraday, Faraday said ‘before we proceed to consider any question involving physical principles, we should set out with clear ideas of the naturally possible and impossible;’ (and this was Crookes take on that quote, Crookes said:) this appears like reasoning in a circle; we are to investigate nothing until we know it to be possible; while we cannot say what is impossible (outside pure mathematics) until we know everything. In the present case I prefer to enter upon the inquiry with no preconceived notions whatever as to what can or cannot be.”  That’s a scientist talking. You don’t set predetermined limits as to what you can find. You open the door, you look at the data, you follow the data where it goes. And if it goes places that you don’t like philosophically, that’s tough. That’s where the data goes.  And that’s how we have to be.

And this is my last quote, from (Alfred North) Whitehead (1946, I think you’ll enjoy this one): “the universe is vast; nothing is more curious than the self-satisfied dogmatism with which mankind, in each period of his history, cherishes the delusion, of the finality of it’s existing modes of knowledge. Skeptics and believers are all alike. At this moment, scientists and skeptics are the leading dogmatists.” (I think nothing has changed much.) “Advances in details are admitted, but fundamental novelty is barred.” Major changes are not allowed. Okay. This dogmatic common sense is the death of philosophical adventure. The universe is vast.  I think we need keep this in mind. Thank you.

(Applause)

(Sounds of motion)

Robert Park wrote a book called Voodoo Science.  And he says this is one of the characteristics of voodoo science, that there are always small effects, they can never amplify them, so the result stares you in the face, you can see it without elaborate statistical analysis.  So I’m very curious about those small advantages, in every case down the line (I’m not going to show the overhead again), but down the line, the prayed group, the test group, tiny advantage over the control group.  I worry about some systematic error in the way they did it. I’m worried about the possibility that they weren’t fully blinded.  I told you the evidence that it wasn’t blinded.  It looked as if it was terribly unblinded.  That a person who was devoted, Janet Greene, keeps coming up, who was devoted to proving that prayer helps.  She wanted people to pray harder, she called them and told them “so and so is having a crisis, pray harder.” She had heart and soul to get prayer to work.  Such a person is not objective, when they are looking at the data.

Professor Harris mentioned that another study is coming out. I heard Benson, of Harvard Medical School.  He probably doesn’t, he showed Irving Tessman and Jack Tessman, both of the Skeptical Inquirer article that criticized his work and Dr. Byrd’s work.  He probably is not aware that we wrote a book review  together, published in the journal Science, about the work of Herbert Benson, and he is doing a study, supported by the Templeton Foundation (that’s a man whose made billions of dollars) and he’s doing, he’s supporting a lot of work on the interface between science and religion, a lot of it very good.  And he’s supporting Dr. Benson’s work.  We wrote a book review of his, in which we criticized him for making claims that were patently untrue. We talked about things that they showed by their studies, his featured procedures to have a mantra, people repeated over and over again a few times a day, and it solved all sorts of medical problems. For example, one problem he claimed in his book, and that they studied, at Harvard Medical School, that there’s no hospital there, was the effect on fertility.  They had a fertility clinic.  And they studied whether saying this mantra improved fertility. These were women who were infertile.  And were going to fertility clinics in addition to his group, his Mind/Body Institute.  And he claimed in his book that more women became pregnant as a result of the mantra. But then you go and read the paper, which he was a co-author, and the paper said we cannot draw that conclusion, because we didn’t test the controls.  They didn’t realize that they had controls in the paper, and my brother and I found the control, and the control showed the same rate of conception, as the group that was saying the mantra. So it would be interesting to see what comes out of this, and how their results conform with Byrd’s, and Harris’.

I just conclude very briefly by saying the following.  If you ask me does intercessory prayer shorten anyone’s stay in purgatory, I don’t know of any evidence. People often say “I don’t know of any evidence” and you think they mean there’s evidence, and evidence that doesn’t support it.  But this is an example where there is no evidence, and I don’t know how to study it.  But if you ask me, “does intercessory prayer shorten the stay in the hospital,” I say there is evidence, and the evidence shows that it does not shorten the stay in the hospital. Thank you.

(applause)

Harris: Again, I’m not awfully familiar with therapeutic touch. But those studies were blinded studies, and the patient didn’t know if the practitioner was actually doing whatever therapeutic touch is. Again, I think it is holding hands over certain parts of the body and if the patient was lying on their belly and someone was working on their back they wouldn’t necessarily know what’s going on there. But, I can’t say for sure how they did it.  But, I know that in that particular review article, they were very picky about not putting in things there were no blindings.  I presume it was.

Moderator: The next question is for Dr. Tessman.  They ask: “If the vast majority of the world believes that prayer heals, assuming that belief in a god means belief in healing prayer, is the vast majority of the world delusion?”

(laughter)

Tessman: Yes.

(laughter)

(applause)

(laughter)

Moderator: Okay. The next one is for Dr. Harris. They ask, “How did god know who they were praying for? If the intercessors were just given a name, how did they not know he was just, when they prayed for someone, just Bob, how did they . . .  ”

Harris: Bob’s in the hospital real name?

Moderator: Bob’s in the hospital real name.

Harris: I’ll follow his lead: I don’t know.  

(laughter)

Moderator: Okay. This is sort of a follow-up question for you Dr. Harris, same person. “The assumption was that the patients wanted to get better. What was the reason behind (the assumption) that god wanted them t