Figures & data
Table 1. Baseline characteristics by drug-use pattern.
Figure 2. Adjusted survival functions for LB plus ICS versus LB only users. LB: long-acting bronchodilators. ICS: inhaled corticosteroid. Adjusted Hazard Ratio: 0.83; 95% CI: 0.72 – 0.97; p-value: 0.024.
![Figure 2. Adjusted survival functions for LB plus ICS versus LB only users. LB: long-acting bronchodilators. ICS: inhaled corticosteroid. Adjusted Hazard Ratio: 0.83; 95% CI: 0.72 – 0.97; p-value: 0.024.](/cms/asset/a52c5b2b-87d6-4908-b27a-78ea27ad2899/icop_a_1044861_f0002_oc.gif)
Table 2. Baseline characteristics in patients with recent out-of-hospital exacer-bations.
Figure 3. Adjusted survival functions for LB plus ICS versus LB only users in patients with recent out-of-hospital exacerbations. LB: Long-acting bronchodilators. ICS: Inhaled corticosteroid. Adjusted Hazard Ratio: 0.63; 95% CI: 0.44–0.90; p-value: 0.012.
![Figure 3. Adjusted survival functions for LB plus ICS versus LB only users in patients with recent out-of-hospital exacerbations. LB: Long-acting bronchodilators. ICS: Inhaled corticosteroid. Adjusted Hazard Ratio: 0.63; 95% CI: 0.44–0.90; p-value: 0.012.](/cms/asset/f8201c1f-f137-4418-8bf3-893b9315091c/icop_a_1044861_f0003_oc.gif)
Table 3. Sensitivity analyses on the length of grace periods.
Figure 4. Sensitivity analysis of residual confounding: the array approach. The symbols represent the “fully” adjusted Hazard Ratios (HR, including the unmeasured confounder ‘smoking habits’), assuming a smoking prevalence of 0.45 in the exposed group (Pc1, “LB plus ICS”). The prevalence of the unmeasured confounder in the unexposed group (Pc0, “LB alone”) is varied between 0.45 to 0.55 on the x-axis. The strength of the confounder-disease association (RRcd) is assumed to be 1.57.
![Figure 4. Sensitivity analysis of residual confounding: the array approach. The symbols represent the “fully” adjusted Hazard Ratios (HR, including the unmeasured confounder ‘smoking habits’), assuming a smoking prevalence of 0.45 in the exposed group (Pc1, “LB plus ICS”). The prevalence of the unmeasured confounder in the unexposed group (Pc0, “LB alone”) is varied between 0.45 to 0.55 on the x-axis. The strength of the confounder-disease association (RRcd) is assumed to be 1.57.](/cms/asset/196ffadb-0e5b-458a-8de2-ef989202cb63/icop_a_1044861_f0004_oc.gif)
Appendix 1 Coding algorithms for defining patient co-morbidities and medical procedures
Appendix 2 Baseline co-morbidities by drug-use pattern