In the cohort analysis, subjects were classified into ICS users and nonusers on the basis of drug dispensation in the 90 days following discharge from the hospital. Subsequently, the two groups differed substantially in terms of the receipt of ICSs; 79.5% of those classified as users at 90 days had filled a prescription for ICSs between 90 and 365 days after hospital discharge compared with 12.0% of nonusers. Each month, between 90 days and the 12th month, approximately 40 to 45% of ICS users received additional ICSs compared to 5 to 10% of initial nonusers.
Among subjects > 65 years who were not treated with ICSs, 1,326 (55.4%) received bronchodilators (ie, P-agonists, ipratropium bromide, or theophylline) within 90 days following hospital discharge. Surprisingly, the remaining 1,067 subjects (26.5% of the total) did not receive bronchodilators, although they could have received antibiotics or oral corticosteroids. Only 7% of patients who received ICSs during the first 90 days after hospital discharge did not also receive bronchodilators. These patients and those who received only systemic steroids and/or antibiotics constituted groups that were too small for meaningful analysis. Between the 4th and the 12th month, approximately 60% of the ICS group, 55% of the bronchodilator group, and 10 to 15% of the no-treatment group filled a prescription for a bron-chodilator. We repeated the Cox model comparing the following three treatment groups: ICSs; bron-chodilators; and neither. We used bronchodilators as the reference treatment because this is currently recommended for all symptomatic COPD patients. Therapy with ICSs reduced the risk of death by 23% (95% CI, 6 to 37%) in comparison with bronchodi-lator treatment. Reduction was significant for cardiovascular deaths (38%; 95% CI, 11 to 57%) but not for COPD (Fig 2). Mortality reductions with ICSs were similar to those observed in patients who had received neither bronchodilators nor inhaled steroids. Presumably, this group of patients had less serious disease; in addition to their lower mortality rate and lower drug use, in the year prior to the initial hospitalization they had fewer physician visits for COPD (mean, 2.4 physician visits; patients subsequently given inhalers, 4.4 physician visits).
Figure 2. Risk of death (total and cause-specific) after 90 days in patients > 65 years of age who were segregated by treatment within 90 days of discharge from hospital. HRs with 95% CIs were adjusted for age, sex, number of physician visits in the year prior to hospitalization, and Charlson comorbidity score. The reference group was patients who were treated with bronchodi-lators (BDs) but not ICSs.