Abstract
RATIONALE: The efficacy of inhaled tobramycin on chronic Pseudomonas aeruginosa infections in patients with cystic fibrosis (CF) has been established in clinical trials. However, little is known about its clinical effectiveness on lung function outside randomized controlled trial settings; conventional analysis of existing registry data has heretofore been confounded by treatment selection bias. OBJECTIVE: To determine effectiveness of inhaled tobramycin on FEV1 decline in patients with chronic P. aeruginosa infections using observational data from the Cystic Fibrosis Foundation Patient Registry. METHODS: Patient-level tobramycin use was measured at first chronic P. aeruginosa infection (n = 13,686 patients; age, 6-21 yr). Decline in FEV1 2 years after infection was estimated for patients treated with tobramycin and compared with untreated patients. Multiple linear regressions with confounder adjustment and propensity scores were used to estimate mean FEV1 decline for each group. Because care is organized by centers, we used center-specific prescription rates as an instrument to reduce treatment-by-condition bias. MEASUREMENTS AND MAIN RESULTS: Using center-level prescribing rates, instrumental variables analysis showed less FEV1 decline for patients who received tobramycin when first eligible compared with those who did not receive tobramycin (difference, 2.55% predicted; 95% confidence interval, 0.16-4.94; P = 0.0366). CONCLUSIONS: Inhaled tobramycin is effective in reducing lung function decline among patients 6 to 21 years of age with CF. Because CF care is organized by center, using center-specific prescription rates as an instrumental variable is a feasible approach to using the Cystic Fibrosis Foundation Patient Registry to determine treatment effectiveness. More generally, this approach can correct for treatment-by-condition bias arising from observational studies.