2002 OPEN FORUM Abstracts
ARE PATIENTS REALLY BLINDED IN A DOUBLE-BLIND RANDOMIZED CONTROLLED STUDY
Timothy Myers BS, RRT and Carolyn Kercsmar MD. Rainbow Babies and Childrens Hospital and Case Western Reserve University. Cleveland, OH
Introduction: The concept of experimental validity is a concern with conclusion accuracy in a research design. Often studies are conducted in a double-blind approach to eliminate threats to external validity from study-effect or investigator biases. Another hypothetical threat to study validity is the patient(s) becoming unblinded in a randomized controlled trial and tipping off the investigators. In a double-blind, randomized trial comparing the acute treatment of pediatric asthma with two types of albuterol (racemic albuterol and levalbuterol), we sought to determine the ability of patients or their parent(s) to correctly break blinding of their randomized albuterol type.
Methods: In a double-blind, randomized control trial of 552 asthmatic patients, we sought to determine if levalbuterol resulted in improved clinical outcomes compared to racemic albuterol. Patients who consented for participation were randomized in a double-blind fashion to receive either 2.5 mg albuterol or 1.25 mg levalbuterol delivered by a high-density Nebutech (Salter Labs, Irvine, Ca.) nebulizer. We utilized our assessment-driven Asthma Carepaths to standardize treatment between groups. For the purpose of this study, a small convenience sample (25% or 113) of patients and/or their parents were provided a questionnaire asking them to attempt to identify the form of albuterol (racemic or levalbuterol) prescribed to them/their child for the acute treatment of their asthma exacerbation. Questionnaire respondents were also given a list of qualifying reasons for their selection of a specific type of albuterol. Nominal data was analyzed for sensitivity and specificity, positive and negative predictive value, and diagnostic accuracy. Kappa and Phi coefficients were used to assess for reliability between respondents.
Results: All 113 surveys were returned. The figure below is a breakdown of the predictive test for the respondents? answers to type of medication they/their child was receiving. Thirty-four (30.1%) questionnaire respondents stated they were ?not sure? of the albuterol type. Fifty-two (46%) reported taking levalbuterol, while 27 (23.9%) patients reported taking racemic albuterol. Percentages for sensitivity and specificity were 31.8% and 62.8% respectively. Positive predictive value of the responses was 51.9%, while the negative predictive value was 42.3%. Diagnostic accuracy was 52.1%. Reliability testing by Kappa coefficients yielded a value of 0.18. Reliability testing by Phi coefficients yielded a value of -0.056.
|Questionnaire Response||Racemic Albuterol||Levalbuterol|
Conclusion: From the results above, the respondents? ability to accurately predict the type of albuterol they/their child was randomized to be statistically no more predictive than that of chance guessing. In this study, patient blinding was adequately achieved in a sub-sample that was surveyed.