Background: Tobramycin solution for inhalation (TOBI; TSI) is indicated to treat patients with cystic fibrosis who are infected with Pseudomonas aeruginosa. Preliminary findings from a randomized trial indicate that patients who received TSI had about half the mortality rate of those assigned to placebo.
Methods: We used the Cystic Fibrosis Foundation registry data to conduct a retrospective cohort study of the risk of death among cystic fibrosis patients in 1999 according to their use of TSI during 1998. We controlled for age, lung function, height and infection with; other factors were not important confounders.
Results: The crude risk of death among those who received TSI therapy for 4 or more months was 3.5 times greater than that among those who received no TSI (90% confidence interval = 3.0-4.2). In general, increased TSI use was related to progressive increases in the risk of death. Such a relation is expected because TSI is used for those who are close to death, resulting in strong confounding by indication. After control of the previously mentioned confounders, the estimated risk ratio was reduced from 3.5 to 1.2. Unfortunately, it is difficult to remove confounding by indication in its entirety. Using a method that estimates the magnitude of uncontrolled confounding, we show that the actual relation between TSI and the risk of death is likely to be protective, and may well be consistent with the results from the randomized trial.
Conclusions: These data illustrate strong confounding by indication and the extent to which the interpretation of data can rest on assumptions about the data and its residual biases.