This tendency toward low power is accentuated by the recommendation, to which we adhered, of excluding patients who were sure that theophylline was helpful following an open trial. However, our rate of stopping one half of the n of 1 trials before completing at least three treatment pairs is consistent with other studies, and it suggests that the failure of patients to complete multiple treatment crossovers is an inherent limitation of n of 1 trials. Our decision to exclude patients who were sure that theophylline was helpful following open therapy should not have invalidated conclusions regarding the effects of n of 1 trials or standard practice on group outcomes because of differing baseline perceptions about the drug’s value. This conclusion is based on our randomizing patients to n of 1 trials or standard practice only after confirming that they were uncertain about the effects of theophylline. However, this approach does limit the generalizability of our findings. Specifically, we did not assess the impact of initiating theophylline by either n of 1 trials or standard practice, and without an open run-in, in patients with CAL.
A final possible reason for the negative result may reflect our decision to return patients to their own physician following the first 3 months of the study. In so doing , we adopted an effectiveness (“real world”) rather than efficacy (“ideal world”) strategy to assess the impact of n of 1 trials. In seven patients, nonstudy physicians subsequently chose to resume theophylline despite conclusions during the n of 1 trials or standard practice approach that it was ineffective so buy ventolin inhaler. We cannot determine the impact of these reversals on the comparisons of quality of life and exercise capacity between the n of 1 trial group and the standard practice group beyond noting that (1) there was not a large difference between the two groups in the number of patients who subsequently resumed theophylline (five and two patients, respectively), and (2) excluding these patients did not yield different results from the primary, intent-to-treat analysis. It is, therefore, less likely that these switches materially affected our overall findings. However, investigators comparing outcomes between patients randomized to n of 1 trials or standard practice in future studies should consider limiting subsequent treatment changes (by directly controlling patient therapy beyond the initial phase) or allowing such reversals to occur but maintaining study power to detect differences after excluding patients who switch (by randomizing a larger number).