Sunday, June 18, 2006

Could the SMART trial have been better designed to bring about the results that GSK did not want?

More than one pulmonary disease specialist has referred to the SMART trial as the [not so] SMART trial. Its design is subject to criticism both in terms of what was offered to the patient volunteers and what was likely to happen in terms of results?

GSK was faced with the task of designing and executing a large clinical trial that had been brought about because of concern that salmeterol might be harmful to some patients with asthma. An earlier trial in Great Britain had shown a numerical increase in salmeterol treated patients that was not statistically significant. One lingering popular explanation for any possible increase in salmeterol treated patient death is that these patients were either under poor control at the beginning of the trial and/or perhaps used the medications improperly or inadequately.

So, what did they do?They recruited patients by advertisements, insuring a high proportion of patients who might not have been receiving good physician care, gave them either salmeterol or placebo and then provided only monthly telephone followup. At this point we have a group of patients, many of whom were not under good control or receiving regular physician hands-on follow up, who were given a medication that was purportedly possibly harmful and not monitored other than by telephone calls. They could not be accused of "cherry-picking" their study subjects in this trial.

This "phase one " recruitment ended when it appeared the trial would need over 60,000 participants (over twice what the original power computation indicated because of the relatively low mortality rate in both the control and treatment arms) and the recruitment process shifted to study investigators who recruited their own patients.This change brought about a situation at least resembling how the drug might be used in clinical practice.

There was a marked difference in the outcomes based on the method of recruitment. 13 of the 16 asthma related deaths occurred in patients recruited in phase one ( figure 2 in the SMART article in CHEST).There were about 15,00 patients in phase one versus about 11,000 in phase
2. The report's table 7 indicates that there were no statistically significant differences between salmeterol and placebo in any of the primary or secondary outcome measures in the phase 2 patients.

If there had been no phase 1 recruitment, there would likely be no black box warning on salmeterol containing products. It should be noted that this trial was terminated early by GSK. Only half of the subjects had been enrolled and the predefined criteria for terminations had not been met.The reasons given were the disproportionate number of events in African Americans and difficulty in enrollment.

What about the effect of prior or concomitant treatment with an inhaled corticosteroid (ICS) on salmeterol treatment outcomes? This is how salmeterol is currently used according to guidelines. As the study authors said " SMART was not adequate to determine whether or not ICS use affected the incidence of the key outcome events."

To me, the most robust finding of SMART is that asthma treatment is much better when a patient in under a physcian's care than when a prescription is given to a patient and he is told that they will be contacted by phone to see how they are doing.

No comments: