Because these results are based entirely on computerized prescription drug data of dispensed medications, they may not coincide precisely with the actual intake of the medications, potentially resulting in some misclassification of drug use. It is also possible that some patients may acquire extra canisters of an SA P-agonist for security reasons. Although this is a limitation of using administrative prescription data for any pharmacoepidemiologic study, the abundance of previous studies specifically investigating asthma using this methodology suggests that it is an acceptable limitation. Using annualized prescription data and assuming that the receipt of a prescription approximates usage patterns eliminates the potential recall bias and attenuates any effect of the seasonal variation in drug utilization. The large sample size and longitudinal analysis in this study should minimize the impact of this potential prescription bias.
By utilizing a repeated-measures analysis and including only subjects in the data set each year, we controlled for this potential confounding and were able to more accurately assess drug utilization patterns, independent of asthma prevalence. Despite this, a limitation of this analysis is that this is not an incident cohort and, in fact, is likely composed of mostly prevalent
SA P-agonist users. Patients receiving long-term therapy may be less likely to alter their therapy or to have it changed, whereas newly treated asthmatic patients may be more apt to be treated in accordance with the guidelines. It was not possible to test this hypothesis using these data.
Furthermore, we identified a paradoxical trend of decreasing ICS use, which is illustrated by an increasing prevalence of not receiving an ICS and a greater likelihood of not receiving an ICS after having received one previously than of starting to receive one. Most disconcertingly, these trends were most pronounced in patients who are the most likely to benefit from concomitant ICS therapy (ie, the excessive SA P-agonist users).
A subgroup analysis of factors associated with increasing use identified increasing ICS dose as the strongest risk factor for increasing SA P-agonist use, which likely is indicative of increasing asthma severity. These analyses therefore demonstrate different associations of age, gender, and receipt of social assistance with increasing SA P-agonist use in controlled and uncontrolled patients, independent of disease severity.
This decrease may have been at least partially attributable to the increase in the deductible that became effective April 1, 1998. The prevalence of SA P-agonist use in this sample was greater than the prevalence of asthma in the general population (17% vs 5 to 10%, respectively). This was not unexpected given the apparent higher prevalence of asthma in lower social classes.
In a previous Canadian study, Habbick et al demonstrated an increase in the annual prevalence of ICS use between 1989 and 1992, with a subsequent decrease in 1993. This reversal may have occurred due to the weakening of the effect of the 1990 publication of the guidelines. Although we expected to identify some effect following the republication of the guidelines, we were unable to identify any trends toward improved asthma management in terms of either decreasing SA P-agonist use or increasing ICS use.
Of 4,237 patients who received more than four canisters of an SA P-agonist in 1996, 584 and 1,332, respectively, met the criteria of increasing and decreasing use (Table 3). Patients whose use decreased received significantly more canisters of SA P-agonists (13.0 vs 9.3, respectively; p < 0.0001) and had a higher mean daily ICS dose (475 vs 326 ^g/d. respectively; p < 0.0001) at baseline than those whose use increased. More plan C patients than plan E patients also increased their use over the 3 years (33.0% vs 25.2%, respectively; p = 0.004). Although there was no relationship between age and increasing use in uncontrolled patients, and the association between gender and increasing use was attenuated in the adjusted model, the receipt of social assistance remained significant. Patients on Pharmacare plan C were 1.3 times more likely than patients on plan E to increase their SA P-agonist use by at least 50%.
Because this finding could be related to decreasing asthma severity, we repeated the analysis in the 966 patients who received at least 13 canisters in all 3 years and found that their probability of not receiving an ICS increased from 0.16 to 0.20 (p = 0.002) between 1996 and 1998.
Figure 2 illustrates the comparison of the proportion of patients who filled at least one prescription for an ICS in 1996 but did not do so in 1998 to the proportion that did not fill an ICS prescription in 1996 but did so in 1998. Although there was no difference in proportions in the low-use group, significantly more patients appear to have discontinued ICS therapy than were initiated on it over the 3 years. The most salient discrepancy occurred in the excessive SA P-agonist users, 16.4% of whom received a prescription for an ICS in 1996 but did not do so in 1998 compared to only 4.7% who initiated ICS therapy (p < 0.0001).
Although the majority of patients received < 4 canisters of SA P-agonist each year, > 2,000 patients (> 6%) received > 13 canisters, > 40% of whom received > 20 canisters (Table 1). Among these patients, 1,147 received an average of > 20 canisters per year, of whom 152 (13.2%) averaged > 40 canisters per year. We had hypothesized that, over time, fewer patients would receive inappropriately high doses of SA P-agonists. However, the proportion of patients in each SA P-agonist use group varied < 1% between years.
The mean daily ICS dose at baseline also was evaluated and was included in the adjusted models if significant. The dependent variable used for the analysis of both subgroups was increasing use, resulting in the determination of odds ratios (ORs) for increasing use relative to remaining in the low-use group over the 3 years in controlled users, and for increasing use relative to decreasing use in uncontrolled users. All variables significant in univariate analyses at a level of a = 0.10 were included in the multiple regression models. Model fit was assessed using —2 Log Likelihood. Crude and adjusted ORs and 95% confidence intervals (CIs) are reported for each variable.
Factors associated with increasing SA P-agonist use were assessed in two discrete samples of this population, as follows: (1) users of low doses of SA P-agonists (ie, four canisters or fewer) in 1996 who could be considered controlled; and (2) users of more than four canisters in 1996 whose utilization could increase, decrease, or remain static but could be deemed uncontrolled. Controlled subjects who consistently received four or fewer canisters of SA P-agonist each year were compared with those who were controlled in 1996 but received more than four canisters in 1997 and 1998, and at least four more canisters in 1998 relative to 1996.
Comparative dosages were derived from dosage comparison data and recommended dosages. Salbutamol administered by dry powder inhaler was considered to be half as potent as administration by MDI, and fenoterol and salbu-tamol were considered equipotent when administered by MDI and nebulizer. Each dosage form of ICS also was standardized to the equivalent dose of beclomethasone dipropionate (BDP) administered by MDI. BDP was considered to be equi-potent to budesonide administered by MDI, and half as potent when administered by dry powder inhaler. Triamcinolone and fluticasone were considered to be half and twice as potent, respectively, as BDP when administered by MDI.