Impossibility to eliminate observer effect in the assessment of adherence in clinical trials

PURPOSE: To utilize the Travoprost Dosing Aid (DA) in the assessment of patient medication adherence, while also determining whether or not altering the functionality of the DA in three randomized subject groups can reduce observer effect. METHODS: Forty-five subjects were randomized into three groups: two with monitored DAs and one without monitoring. One group of subjects was given a DA that both monitored drop usage and had visual and audible alarms, while the other monitored group included subjects given a DA that had no alarms but continued to monitor drop usage. The third group was given a DA that had no alarm reminders or dose usage monitoring. Subjects were informed that some monitors would not be functional, in an attempt to reduce observer effect, or the effect of being monitored on subject behavior and adherence. A six-item questionnaire was also utilized to assess how the subjects felt about their adherence and DA use. RESULTS: The overall adherence rates were found to be 78% in the fully functional group (95% confidence interval: 70–88) and 76% in the no alarms group (95% confidence interval: 65–89). No association was seen between questionnaire response and medication adherence. The patients in the DA group without alarms had a significantly higher odds ratio of medication adherence if they reported on the questionnaire that using the DA did affect how much they used their drops. CONCLUSION: Though the use of DA was expected to reveal different rates of adherence depending on the functionality of the DA between groups, patients with a nonfunctioning DA did not have a significant difference in medication adherence compared to those given a fully functional DA. This supports that an observer effect was not reduced despite these interventions, and that the subjects adhered to taking their medications as if they had a functioning DA and were being monitored.