Thursday, February 07, 2008

Diabetes Study Partially Halted After Deaths

This is the headline of Gina Kolata's article in the New York Times about a large study on diabetics. One part of the study compared the benefits from very rigorous control of blood sugar to a program of somewhat looser control. The obvious hypothesis was that trying to get the blood sugar levels to as close to nondiabetic levels as possible would be a good thing, in this case a good thing in preventing heart attacks. What the study found instead is stunning:

For decades, researchers believed that if people with diabetes lowered their blood sugar to normal levels, they would no longer be at high risk of dying from heart disease. But a major federal study of more than 10,000 middle-aged and older people with Type 2 diabetes has found that lowering blood sugar actually increased their risk of death, researchers reported Wednesday.

The researchers announced that they were abruptly halting that part of the study, whose surprising results call into question how the disease, which affects 21 million Americans, should be managed.

The study's investigators emphasized that patients should still consult with their doctors before considering changing their medications.

Among the study participants who were randomly assigned to get their blood sugar levels to nearly normal, there were 54 more deaths than in the group whose levels were less rigidly controlled. The patients were in the study for an average of four years when investigators called a halt to the intensive blood sugar lowering and put all of them on the less intense regimen.

The results do not mean blood sugar is meaningless. Lowered blood sugar can protect against kidney disease, blindness and amputations, but the findings inject an element of uncertainty into what has been dogma — that the lower the blood sugar the better and that lowering blood sugar levels to normal saves lives.

Medical experts were stunned.

Note that the patients were randomly assigned to each treatment group. This means that the differences in the death rates are not because one group was in worse state of health to begin with, say. The differences must be in the treatment regimes themselves, and if you read the linked article you will find that the researchers worked very hard to find the reason for the excess deaths, other than the lowered blood sugar, but failed to do so.

This study reminds us of the possibility that something "everybody knows to be good" may in fact not be good at all. It also reminds us of the need to do studies of this type, even in areas where "everybody knows what works."

This post might be a good place to talk about the importance of randomized controlled trials in medicine. Such trials consist of randomly allocating patients into one treatment regime or another or of randomly allocating them into a treatment regime and a control regime with no treatment but perhaps a placebo. Ideally these trials should be "double-blind", too, meaning that neither the patients nor the people who administer the treatment can tell who is in which group. That way the trial avoids finding differences just because the very act of getting treated could make a difference in either the patients' self-assessment of health or in the way the providers view their health.

The initial random allocation of patients is very important, because only by random allocation can we guarantee that the differences we might find later on are not caused by initial differences between the patient groups.

Some of you may wonder how ethical such studies could be. After all, isn't it unethical not to give patients a treatment? The usual answer to that question is that these trials should be held when nobody really knows if one treatment is any better than another treatment or if a treatment is better than doing nothing. In that state of real "not-knowing" a participating patient faces the same initial risk of not getting the best possible treatment, whichever group he or she might be assigned to. If the new treatment turns out to be superior to the alternative the trial is stopped and all patients transferred to the new treatment regime. If, on the other hand, the new treatment turns out to be inferior to the alternative (or to have severe side effects), the trial is also stopped and all patients transferred to the control regime.

This diabetes study brings the point home rather forcefully. We might have well assumed that it would be wrong not to put all the patients on the most rigorous regime of blood sugar control. But the study proved otherwise.