Inhaled Corticosteroids and Mortality in COPD: Conclusion

Inhaled Corticosteroids and Mortality in COPD: ConclusionOur study has a number of strengths. It examined a large unselected population of patients, including those < 65 years of age, using a comprehensive database. We attempted to assess and adjust for comorbidities and disease severity by examining the use of other respiratory drugs and physician exposure. Excluding patients with a previous diagnosis of asthma, or those who had previously used ICSs did not affect our results. We avoided immortal time bias” by excluding deaths within the first 90 days of hospital discharge. We were further able to examine mortality in relation to the timing of the receipt of ICSs. starlix 60 mg
There were no records of drug prescription during hospitalizations in our database, giving rise to two potential biases. First, people who were hospitalized in the initial 90 days of observation had less opportunity to receive outpatient medications such as ICSs and were potentially a high-mortality group. We repeated our analysis after eliminating people who were rehospitalized during the initial 90 days, and ICSs were associated with reduced mortality in the remainder (relative risk, 0.73; 95% CI, 0.58 to 0.92), so this issue did not confound our cohort analysis. A second potential bias concerns people dying in hospitals who had received unknown drugs, introducing an error in estimating the potential influence of outpatient prescriptions in regard to death. We repeated our case-control study examining only people who had not been hospitalized at death. The same relationships shown in Figure 3 were observed (ie, the effect of ICSs was most evident in the 30 days before death), so we do not think that this bias influenced our result in an important way.
Our study had limitations that are common to most database analyses. We relied on physician claims for diagnoses, which are often inaccurate, but are less so when derived from hospital discharge information as we did. We attempted to control for disease severity and comorbidities but cannot be certain that this was successful. We used drug dispensing as a surrogate for drug use and recognized that the two were not necessarily the same, although they were most likely to be similar immediately after drug receipt. Further, discrepancies between drug dispensing and use would have tended to blur the time effects associated with the former. We relied on death certificates to ascertain causes of death, which may have been misclassified but not biased with respect to treatment group. All-cause mortality data were robust, and the main causes of death (ie, COPD and cardiovascular disease) were consistent with the literature.

This entry was posted in Pulmonary Function and tagged , , , .