Data monitoring committee ends behavioral COPD trial early

Scheduled safety monitoring should occur during behavioral intervention trials in case of unexpected outcomes, according to the authors of one such study for chronic obstructive pulmonary disease (COPD) that recorded excess deaths in the intervention group.

The stated purpose of the study was to determine the efficacy of a comprehensive care management program in reducing the risk for COPD hospitalization. Researchers created a randomized, controlled trial comparing behavioral interventions with guideline-based usual care at 20 Veterans Affairs outpatient clinics for patients hospitalized for COPD in the past year. The primary outcome was time to first COPD hospitalization.

Results, and a consideration of the ethics of such trials, appeared in the May 15 Annals of Internal Medicine.

Intervention patients received a written, individualized action plan for flare-ups that included prescriptions for prednisone and an antibiotic chosen in consultation with the primary care physician, with instructions to start treatment within 48 hours after onset of exacerbation symptoms.

In contrast to other behavioral intervention studies, which typically use a safety officer, VA trials require a data monitoring committee. The study began enrollment in January 2007. A scheduled meeting of the committee in January 2009 revealed an imbalance in mortality between the study groups.

After failing to find any reason for the difference, the data monitoring committee recommended that study enrollment and the intervention be stopped immediately, and that all patients have all baseline studies repeated with follow-up in an observational study lasting six months.

By that point, 426 (44%) of the planned total of 960 patients were enrolled. There were 28 deaths from all causes in the intervention group versus 10 in the usual care group (hazard ratio, 3.00 [95% CI, 1.46 to 6.17]; P=0.003).

Cause could be assigned in 27 (71%) deaths. Deaths due to COPD accounted for the largest difference: 10 in the intervention group versus 3 in the usual care group (hazard ratio, 3.60 [95% CI, 0.99 to 13.08]; P=0.053).

Not all educational and care management programs are appropriate for all patients, the authors concluded.

“This study underscores the critical role of the DMC [data monitoring committee] in ensuring the safety of study patients and questioning current clinical trial practices,” the authors wrote.

An editorialist discussed the consequences of stopping a trial early, including that it may exaggerate the true effect because the trial may stop on a “random high,” whereas the observed difference might well have regressed. Furthermore, there was no clear pattern consistent with a plausible reason for harm, and the primary end point and other outcome measures show no evidence of a treatment effect one way or another, the editorialist said.

The editorialist compiled the results of this trial and two others of behavioral interventions for COPD. All three trials combined showed no difference in mortality rates between behavioral interventions and usual care.

“It may be best to think of this trial as having stopped for futility—there was no hint of any beneficial effect on the primary outcome,” the editorialist wrote. “On the other hand, the possibility that genuine harm was done by this behavioral intervention cannot be dismissed. Perhaps insufficient evidence was collected by the investigators about the detailed consequences of the educational package, thus denying us any causal insight into the excess mortality.”