Bias in medical research
The death of an infant is a soul-wrenching tragedy. We must avoid such tragedies through whatever responsible means possible.
Physicians have improved the care of premature infants dramatically, although areas of uncertainty remain. Physicians long wondered whether targeting higher oxygen levels harmed sick infants' eyes without helping them, but there was no way to know without a randomized study.
Finally, the National Institutes of Health funded a study, led by University of Alabama researchers and performed at 22 hospitals. Researchers divided the standard range of oxygen targets for premature infants into a low and a high range, and demonstrated that high oxygen was associated with eye damage but better survival.
After the study was published, a government regulatory body censured the researchers. Essentially they wanted the consent form, approved at the time by appropriate oversight bodies, to incorporate knowledge obtained after the study was completed and to discuss the risk of routine clinical care as if it were a risk specific to the study. This unjustified censure has cast a pall over clinical research.
The unfounded criticism of the so-called SUPPORT study is counterproductive. We will all suffer if cognitive biases prevent ethical, high-quality research from saving the lives of future patients.
Human beings naturally make facts fit their overall picture of the world. Such biases are reasons to doubt the criticism of SUPPORT and are the reason the study was required in the first place. We tend to judge the past by present knowledge. Once a study's results are known, we struggle to imagine the world before the study was performed. But studies must be interpreted on the basis of knowledge when they were performed, not after their results are known.
On the other hand, the bias toward seeing what we believe is why we need randomized studies. Conventional wisdom needs to be rigorously tested. Physicians once believed that medically suppressing dangerous heartbeats after a heart attack was good. That is, until the CAST study, in which more patients died with treatment than those whose dangerous heartbeats were not treated.
Similarly, physicians came to believe that "less is more" for treating lung failure, that giving tiny breaths was safer than the usual large breaths on the ventilator. This ARMA study proved that tiny breaths saved lives.
Both CAST and ARMA had decent science and strong opinions behind them, but the randomized study provided the answer: In CAST, the good idea was wrong; in ARMA, the good idea was right. Both studies improved medical care.
The critics also mistakenly attribute the inherent risks of being a premature infant to the SUPPORT study. The only choice facing parents at the time was what kind of randomization to undergo. Parents could choose meaningful randomization as part of SUPPORT. Or they could choose meaningless randomization on the basis of their physician's preference.
We respect our personal physicians and generally we should. But when physicians lack solid evidence to guide them, they follow a random combination of life experience and exposures to pharmaceutical salespeople, medical professors and colleagues.
In retrospect, babies had long been dying when their physician believed that low oxygen levels safely prevented eye damage. Meaningful randomization allows others to benefit from our experience; meaningless randomization does not.
The courage and grace of the parents who enrolled their precious children in SUPPORT and the researchers' careful work mean that future infants will get the right amount of oxygen.
We must avoid the cognitive biases that threaten such important research.
Samuel Brown is assistant professor of pulmonary/critical care medicine and medical ethics and humanities at the University of Utah, and attending physician in the Shock Trauma ICU at Intermountain Medical Center.