Inhaled corticosteroids for asthma: impact of practice level device switching on asthma control

This 2-year retrospective matched cohort study included a baseline period of 1 year before the date of ICS device switch (index date) and an outcome period of 1 year after the index date. Patient data were drawn from 1990–2004 records in the General Practice Research Database (GPRD), a large computerized database of anonymized longitudinal medical records from primary care in the Uk [15]. The years of the switch and matched cohort analyzed in this study are 1992 to 2003. This database includes information for over 13 million patients, including 3.4 million active patients for whom data are currently being gathered. Reports for each patient include demographic data and event history, including date and type of event with a description of the event and the corresponding code. Information about the reason for a switch in device is not recorded. Ethical approval was received from the GPRD Scientific and Ethical Advisory Committee (Protocol 706R, 26.4.05).

Patients eligible for inclusion in this analysis were receiving ICS, were 6–65 years of age on the index date, and had a recorded diagnosis of asthma and no record of chronic obstructive pulmonary disease (COPD). We used the GPRD to identify primary care practices where ICS inhaler devices were switched without an accompanying consultation on the same day for five or more patients within a 3-month period. Patients had to be registered at the practice for at least 1 year before and 1 year after the index date. We included switches to inhaler devices that require training to use, namely, from a dry powder inhaler to a metered-dose inhaler, another brand of dry powder inhaler, or a breath-actuated inhaler or from a breath-actuated to a metered-dose inhaler or another brand of breath-actuated inhaler. Patients changed to a device recorded as 'generic' were excluded.

Each patient included in the analysis was individually matched to another patient (from any practice) who was receiving the same ICS device on the index date and met the following six criteria: same sex, age within 5 years, same smoking status, match on presence or absence of rhinitis history, and, during the prior year, same number of oral corticosteroid prescriptions and hospitalizations for asthma.

Outcome measures were analyzed for the year before (baseline period or year 1) and the year after the index date (outcome period or year 2). The primary outcome measure was a composite measure of asthma control, as defined in Table 1. Secondary outcome measures included several individual elements of asthma-related health resource utilization. The number of prescriptions/patient per year was analyzed for inhaled short-acting β-agonist and oral corticosteroids; in addition, the number of doses/day of short-acting β-agonist was determined for each patient based on the number of inhalers prescribed/year and an assumed standard dose of salbutamol 200 μg and terbutaline 500 μg. We also analyzed the number of general practice consultations (physician and practice nurse) for asthma, the number of hospital admissions for asthma, and the number of hospital admissions for possible asthma, which were defined as nonspecific hospitalization code and asthma-related code within a 1-week window.

Other data recorded for each patient included assigned socioeconomic status of the primary care practice and other respiratory or confounding diagnoses. The daily dose of ICS was calculated from the prescribing instructions on the first available prescription after the index date.

Baseline characteristics of switched and matched control patients were compared using t and χ² tests as appropriate. Median intended ICS dose, daily short-acting β-agonist dose, and asthma consultation pattern during the baseline year were compared for the two cohorts using the Mann-Whitney U test. For comparison of change between year 1 and year 2, the paired t test (mean changes in short-acting β-agonist use and asthma consultations) and χ² test (change or not in ICS dose or device from year 1 to year 2) were used.

For comparisons of the primary outcome measure, in addition to the unadjusted analyses, we used logistic regression to correct for potential baseline confounders. Binary regression models were constructed using as the three outcome variables 1) successful vs. partially successful/unsuccessful treatment, 2) unsuccessful treatment vs. successful/partially successful treatment, and 3) subsequent change in ICS device or dose and, as explanatory variables, cohort membership plus possible confounding factors. Possible baseline confounders entered into the model included 1) short-acting β-agonist use (average daily use categorized as none, ≤0.5 doses/day, >0.5–2 doses/day, >2 doses/day); 2) asthma consultation rate (categorized as: none, 1, 2, ≥3); 3) socioeconomic status (in quintiles); and 4) prescribed ICS dose at switch (in sensitivity analysis as incomplete data due to some patients receiving vague prescribing instructions, categorized as ≤400 μg/day, 401–800 μg/day, 801–2000 μg/day, >2000 μg/day).

We identified 112,090 patients with asthma in the GPRD database who had a switch of ICS device (Figure 1). There were 55 primary care practices in the database where a switch in device without an accompanying consultation occurred on five or more occasions over 3 months. Overall, at these 55 practices, 835 patients had a device switch, and 824 could be matched according to our criteria. The number of switches ranged from 5 to 127 per practice; ≥10 patients were switched at 24 practices, and ≥30 patients were switched at 7 practices. Over half (53%) of the device switches were from a dry powder to a metered-dose inhaler (Table 2).

The switched and control cohorts were identical or comparable, as per the study protocol, with regard to the criteria used for matching (Table 3). The rate of admissions associated with a respiratory consultation was also similar at baseline in the switched and control cohorts (5 vs. 9 patients with 6 vs. 10 admissions, respectively), as was the presence of recorded gastroesophageal reflux or cardiovascular disease (see Table 3).

The median intended dose of ICS at the index date was similar for the two cohorts (see Table 3). Data for 601 of 824 patients (73%) in each cohort were used for these calculations; it was not possible to quantify the daily dose for the other 223 patients (27%) in each cohort because the instruction 'as directed' was recorded for their ICS prescriptions. Table 3 highlights that there were some differences (p < 0.05) between cohorts at baseline.

Fewer patients experienced successful treatment and more patients experienced unsuccessful treatment in the switched cohort than in the control cohort (Figure 2). The likelihood of successful treatment for the switched group was half that for the control group and fell to less than one third that for the control group after adjusting for confounding factors (Table 4). No other variables were predictive of success. Including the ICS dose in the model made no substantial difference: cohort membership remained the only predictor of success, with switched having significantly lower odds of success than control patients (OR, 0.30; 95% CI, 0.20 to 0.45; p < 0.001).

The likelihood of unsuccessful treatment among switched patients was almost double that for control patients (see Table 4). Other variables predictive of unsuccessful treatment were a greater use of short-acting β-agonist and more asthma consultations at baseline. Including the ICS dose in the model made no substantial difference (OR, 1.93; 95% CI, 1.46 to 2.55; p < 0.001).

Patients in the switched cohort were also more likely than those in the control cohort to experience a subsequent change in ICS device or dose in year 2 (see Table 4). Other variables predictive of ICS change were short-acting β-agonist use, socioeconomic status, and rate of asthma consultation at baseline. Results were not substantially different when the ICS dose was included in the analysis (OR, 1.89; 95% CI, 1.48 to 2.40; p < 0.001). Subsequent changes in ICS device or dose during year 2 were recorded for 273 patients (33.1%) in the switched cohort and 183 patients (22.2%) in the control cohort (p < 0.001).

During year 2, the daily use of short-acting β-agonist increased in the switched cohort and decreased in the control cohort (Figure 3), and short-acting β-agonist use was greater in the switched than in the control cohort (Table 5). The switched cohort used 0.38 extra doses/day of short-acting β-agonist, compared with baseline use, than the control cohort (95% CI, 0.22 to 0.53; p < 0.001).

Most patients did not require a course of oral corticosteroids during the outcome year, and there were no significant differences between patient groups in the use of oral corticosteroids during this time (Table 5) or in the change in use from baseline. Similarly, there was no significant difference in asthma consultation rate between patients in switched and matched cohorts. Between year 1 and year 2, consultation rates fell slightly by a mean of 0.09 (SD, 1.33) in the switched cohort and by 0.35 (SD, 1.59) in the matched cohort. Thus, as compared with the baseline period, the switched cohort had 0.26 more asthma consultations in the outcome period than the matched cohort (95% CI, 0.11 to 0.40; p < 0.001).

There were very few hospitalizations for asthma during the study. Moreover, there was no significant difference between cohorts in admissions for possible asthma (Table 5) or in the change from baseline to outcome periods in admissions for possible asthma.