Recurrent Episode Analyses of Pulmonary Exacerbations in the Infant Study of Inhaled Saline (ISIS) Trial

Abstract

The Infant Study of Inhaled Saline (ISIS) randomized, controlled clinical trial evaluated the use of daily inhaled hypertonic saline treatment for 48 weeks to reduce the rate of pulmonary exacerbations in children of less than 6 years of age with cystic fibrosis. In an earlier trial among older children and adults with cystic fibrosis, the group randomized to inhaled hypertonic saline had significantly fewer pulmonary exacerbations and a significantly higher proportion of participants without exacerbations. In ISIS, the primary analysis of rate of pulmonary exacerbations using Poisson regression had a null result for treatment effect (the rates were 2.3 per person-year for both groups; adjusted rate ratio = 0.98 with a 95% confidence interval of 0.84, 1.15), as did an additional analysis of exacerbation-free survival using Cox proportional hazards regression. However, these analyses have limitations. The primary analysis fails to account for the timing of exacerbations, while the second analysis ignores all but the first exacerbation. This thesis applies extensions of the Cox proportional hazards model and a relatively new method, temporal process regression, to provide a more comprehensive picture of the ISIS trial pulmonary exacerbation data in an exploratory setting. Conclusions regarding the effect of treatment from these additional analyses are consistent with those of the original ISIS analyses. An example dataset is constructed to illustrate how extensions of the proportional hazards model and temporal process regression may reveal evidence of a treatment effect that goes undetected by the Poisson regression and simple Cox model analyses.