Abstract
Despite the proven efficacy of inhaled corticosteroids in reducing airway inflammation and their increasing use for the treatment of asthma since the mid 1980s, hospitalization for asthma has been increasing in frequency in several countries. Only few studies, reporting contradictory results, have investigated the role of inhaled corticosteroids in the prevention of hospitalizations for asthma. Using a cohort of 2,059 hospitalized asthmatic patients between 5 and 54 yr of age, we estimated the effectiveness of inhaled corticosteroids in preventing a readmission to hospital for asthma as a function of the duration of therapy. The cohort was selected from the databases of Saskatchewan Health from 1977 to 1993. The rate ratio (RR) of a readmission for asthma varied with duration of regular therapy with inhaled corticosteroids. During the first 15 d of regular therapy, users of inhaled corticosteroids were as likely as nonusers of these medications to be readmitted for asthma with a RR of 1.2 (95% CI: 0.8–1.8). Subjects treated regularly with inhaled corticosteroids for at least 16 d and as long as 6 mo were 40% less likely to be readmitted for asthma (RR = 0.6; 95% CI: 0.4–0.9), while after 6 mo of regular treatment the protective effect disappeared (RR = 1.3; 95% CI: 0.7–2.4). We conclude that regular therapy with inhaled corticosteroids can substantially reduce the risk of a readmission for asthma after only 15 d of use. Confounding by severity appears as the most likely explanation for the disappearance of the beneficial effect after 6 mo of regular therapy.
Inhaled corticosteroid therapy was initially introduced to reduce the need for oral corticosteroids in patients with severe asthma. In the late 1980s and early 1990s numerous randomized clinical trials demonstrated the efficacy of inhaled corticosteroids for the treatment of patients with all levels of asthma severity, including mild asthma (1-6). The efficacy of these medications was demonstrated using intermediate clinical outcomes such as the reduction of airway hyperresponsiveness, the decrease in the need for bronchodilators as well as the improvement of airway caliber. Following the results of these trials, inhaled corticosteroids have been recommended as the medication of choice to bring and keep asthma under control (7-10).
Despite the availability of efficacious medications to treat asthma, mainly inhaled corticosteroids, hospitalization for asthma has been increasing in frequency over the last decades in several countries, including Canada and the United States (11-17). Hospitalization is an important outcome in asthma due to the associated increased risk of a subsequent fatal attack as well as being costly for the health care system (18-20). Moreover, it has been shown that patients hospitalized for asthma are prone to readmissions to hospital with increasing probability of a readmission as the number of previous hospitalizations increases (21-23). Hospitalizations for asthma are, however, sufficiently rare that randomized clinical trials have been unable to investigate them as an outcome. Only few nonexperimental studies have examined the role of inhaled corticosteroids in reducing the risk of hospitalization for asthma and these have provided contradictory results (22, 24-29). The only study that addressed the risk of a readmission for asthma found no association between the use of inhaled corticosteroids and the rate of readmissions (22).
In this study, using a large cohort of asthmatics between 5 and 54 yr of age, we assess the effectiveness of inhaled corticosteroids given after a hospitalization for asthma in preventing a readmission to hospital for asthma. The effectiveness is estimated for different intervals of duration of therapy with inhaled corticosteroids.
Using the computerized health insurance databases of Saskatchewan, covering approximately 95% of the residents of this Canadian province (30), we selected a cohort (the posthospitalization cohort) of 2,059 asthmatics who had at least one hospitalization for asthma between 1977 and 1993. The posthospitalization cohort, which is the object of this report, was selected from a source population of 19,168 newly treated asthmatics, ranging in age from 5 to 44 yr at entry into the source population.
The source population was formed by linking the Health Insurance Registration file, which contains demographic and health coverage data for each beneficiary, with the Out of Hospital Prescription Drug data file. The source population consists of all newly treated asthmatics between 5 and 44 yr of age who were identified in the Drug database from September 1977 to December 1991. Operationally, we selected the beneficiaries who received at least three antiasthma prescriptions in a 365-d period and who were enrolled in the database for at least 2 yr before the date of their first claim for an asthma medication. This 2-yr criterion to identify newly treated asthmatics was selected for consistency in order to provide the same minimum 2-yr interval without asthma treatment in all subjects included in the source population. This was necessary because the drug database was only created in 1975. For the three-prescription criterion, we considered all antiasthma medications that were covered by the health insurance plan, i.e., beclomethasone, budesonide, triamcinolone acetate, flunisolide, sodium cromoglycate, ketotifen, nedocromil, albuterol, fenoterol, terbutaline, isoproterenol, metaproterenol, procaterol, epinephrine bitartrate, ipratropium bromide, and any compound of theophylline, except for oral corticosteroids. Oral corticosteroids were not included in the list because patients receiving only this class of medications are more likely to suffer from conditions other than asthma. Thus subjects receiving only oral corticosteroids were not included into the source population, but subjects receiving oral corticosteroids in addition to any other antiasthma medications were included if they met the three-prescriptions criterion. The 19,168 selected subjects entered the source population on the date of their third prescription dispensed during the 365-d period.
To minimize the effects of preferential prescribing of inhaled corticosteroids to more severe patients, we selected, from the source population, the cohort of 2,059 asthmatics who had experienced at least one hospitalization for asthma after they entered the source population and who were therefore more homogeneous with respect to disease severity. The 2-yr criterion without asthma prescription required before entry into the source population was not reapplied before entry into the posthospitalization cohort. Selection of the members of the posthospitalization cohort was not based on drug consumption, it was only based on the occurrence of a hospitalization for asthma.
The date of discharge of the first hospitalization for asthma occurring after entry into the source population (initial hospitalization) was taken as the date of entry into the posthospitalization cohort. The subjects were then followed until a readmission for asthma (the outcome), their 55th birthday, the date of death, emigration from the province, or end of coverage of the health insurance plan or December 31, 1993, whichever occurred first. The analyses that follow were all performed within the posthospitalization cohort.
The initial hospitalization and the outcome of interest, a readmission for asthma, were defined as a hospitalization with a primary discharge diagnosis of asthma (ICD-9 codes 493.0, 493.1, or 493.9) or a hospitalization with a secondary discharge diagnosis of asthma and a primary discharge diagnosis that was either an alternative code for asthma (e.g., bronchitis), a precipitating factor of asthma (e.g., upper respiratory tract infection) or a complication of asthma (e.g., pneumothorax) (ICD-9 codes 426.4, 427.0, 427.1, 427.3, 427.5, 427.8, 427.9, 460, 461.9, 462, 463, 464.2 to 465.9, 466.0, 466.1, 472.0, 472.2, 476.0 to 478.9, 487.0, 487.1, 487.8, 490, 491.2, 491.9, 492, 494, 496, 508.0, 508.8, 512, 516.8, 518.0 to 519.9, or 786.0 to 786.6).
Inhaled corticosteroids regroup beclomethasone, budesonide, triamcinolone acetate, and flunisolide. “Regular users” of inhaled corticosteroids were defined as subjects who had filled at least one prescription of these medications every 90 d, assuming that every prescription of inhaled corticosteroids could last up to 90 d. This assumption was based on the minimal recommended daily dose of inhaled corticosteroids which implies that one canister should last around 50 d. The duration of a canister was then extended to 90 d to account for noncompliance of the patients to their treatment (8). According to our definition of “regular use,” a subject who was readmitted for asthma 1 mo after the initiation of a treatment with inhaled corticosteroids was considered as a regular user. Subjects were no longer considered as regular users if 91 d after the last prescription of inhaled corticosteroids they had not renewed their prescription.
Disease severity was measured with the following variables: dispensing of at least one prescription of oral corticosteroids, of antiallergic agents (sodium cromoglycate or nedocromil), of inhaled, oral or nebulized β-agonists, of any theophylline compound and of ipratropium bromide in the 6 mo prior to the initial hospitalization; dispensing of oral corticosteroids in the month following the initial hospitalization; the number of asthma hospitalizations prior to entry into the source population. Dispensing of inhaled corticosteroids in the month preceding the initial hospitalization was treated both as a confounder and as an effect modifier. The calendar period in which the subjects initiated their therapy with inhaled corticosteroids was considered as a potential confounder. The study period was divided into two categories, i.e., 1977–1987 and 1988–1993; 1988 corresponds to the increasing use of higher dose formulations of inhaled corticosteroids for the treatment of asthma (7). The season, age (5–9, 10–14, 15–19, 20–24, 25–29, 30–39, 40–54 yr), and gender of the subject at the initial hospitalization were also considered as potential confounders.
An intent-to-treat analysis and an analysis assessing the effectiveness of regular therapy with inhaled corticosteroids were performed within the posthospitalization cohort. In the intent-to-treat analysis the use of inhaled corticosteroids was assessed only within the first month following the initial hospitalization and was not updated thereafter. Subjects were considered as users of inhaled corticosteroids if they had at least one dispensed prescription of that medication during the month following discharge from the initial hospitalization and were considered as nonusers otherwise. Although medications taken while hospitalized (including the initial hospitalization) are not recorded in the Drug database, medications prescribed during the initial hospitalization and dispensed after discharge were recorded. These data would therefore capture regular practice, which is to give oral corticosteroids during hospitalization and to initiate a therapy with inhaled corticosteroids after discharge.
The subjects were followed until a readmission for asthma (outcome), for a maximum of 1 yr after the initial hospitalization or the end of the study period, whichever came first. The first year after the initial hospitalization was divided into five consecutive intervals: 1–15 and 16–90 d as well as 4–6, 7–9 and 10–12 mo of follow-up. For each interval, the rate ratio (RR) of a readmission for asthma contrasting users and nonusers of inhaled corticosteroids was estimated using a logistic regression model, with adjustment for confounding variables. A stepdown approach was used to find the most parsimonious model (31).
The second analysis also performed within the posthospitalization cohort consisted in estimating the effectiveness of regular use of inhaled corticosteroids on a readmission for asthma. The effect of inhaled corticosteroids was estimated as a function of the duration of regular continuous therapy, i.e., the risk of a readmission for asthma while on therapy with inhaled corticosteroids. Users of inhaled corticosteroids were allowed to initiate their therapy at any time within the first 2 yr after the initial hospitalization. The subjects were considered as nonusers, and used as such in the analysis, until they received their first prescription of inhaled corticosteroids or until the end of the follow-up if they were not treated with these medications. Subjects who started a treatment with inhaled corticosteroids were followed until they stopped taking this medication regularly, reached 1 yr of regular treatment, or reached the end of the follow-up, whichever came first. The first year of regular therapy with inhaled corticosteroids was divided into seven intervals: 1–15, 16–30, 31–60, 61–90 d, 4–6, 7–9 and 10– 12 mo of therapy. For example, a subject who started a treatment with inhaled corticosteroids 2 mo after the initial hospitalization and took it regularly for 8 mo would first contribute 60 person-days in the no-use category and then contribute person-days in the first six intervals of regular use of inhaled corticosteroids.
For each interval of duration of therapy, regular users of inhaled corticosteroids were compared with all nonusers who had been followed for a period of time at least as long since the initial hospitalization. For example, subjects who had been treated with inhaled corticosteroids for 2 mo were compared with all nonusers of these medications who were still in the cohort 2 mo after their initial hospitalization. This analysis was restricted to the first 24 mo of follow-up (i.e., 24 mo after the initial hospitalization) and to a maximum of 1 yr of regular treatment with inhaled corticosteroids to minimize the difference in severity between users and nonusers of inhaled corticosteroids that was expected to increase over time. A shorter duration of follow-up (e.g., 12 mo) may have further minimized the difference in severity but would not have provided adequate power. Twenty-four months was a trade-off between precision and validity.
For regular users of inhaled corticosteroids, crude rates of readmission for asthma were estimated for the intervals of duration of treatment. For nonusers, crude rates were estimated for 11 intervals of time covering the first 2 yr of follow-up (the first seven intervals defined as for the duration of treatment, plus four further 3-mo intervals) in order to assess the variability of the rate of readmission to hospital over time. For each interval of treatment duration, the crude and adjusted RRs of a readmission for asthma, contrasting regular users and nonusers of inhaled corticosteroids, were estimated using logistic regression models. To control for the fact that therapy with inhaled corticosteroids was allowed to start within 2 yr following the initial hospitalization, follow-up time (time since the initial hospitalization) was considered as a potential confounder. The covariates described earlier were also considered as potential confounders and a stepdown approach was used to find the most parsimonious model (31).
To assess the differences in severity over time between users and nonusers of inhaled corticosteroids, we computed the distribution of the use of other antiasthma medications at the initiation of therapy with inhaled corticosteroids and at 3, 6, and 12 mo afterwards. These were compared with the distribution among nonusers of inhaled corticosteroids.
Within the first year of follow-up of the posthospitalization cohort, 482 of the 2,059 subjects were readmitted for asthma for a rate of 29 readmissions per 100 persons per year. Table 1 shows selected demographic characteristics and utilization of medications for all cohort members and those readmitted for asthma. In this cohort 45% of the subjects were males and 44% had their initial hospitalization before the age of 15 yr. The incidence of asthma hospitalization was higher in the fall. Only 10% of the subjects had received a prescription of inhaled corticosteroids in the month preceding the initial hospitalization. Inhaled β-agonists constitute the most prevalent class of medication, with 66% of subjects receiving at least one prescription from this class of medication in the 6 mo preceding the initial hospitalization. For 84% of the subjects, the initial hospitalization after entry into the source population represents their first hospitalization for asthma. More than 95% of the subjects readmitted for asthma had a primary discharge diagnosis of asthma (ICD-9 493.0, 493.1 or 493.9).
| Characteristics (%) | All Subjects (n = 2,059) | Subjects Readmitted for Asthma | ||||
|---|---|---|---|---|---|---|
| Intent-to-treat Analysis (n = 482) | Effectiveness Analysis (n = 484) | |||||
| Male gender | 45.4 | 41.5 | 42.1 | |||
| Age at initial hospitalization, yr | ||||||
| 5–14 | 44.0 | 46.7 | 51.6 | |||
| 15–29 | 28.9 | 25.9 | 24.6 | |||
| 30–54 | 27.1 | 27.4 | 23.7 | |||
| Month of initial hospitalization | ||||||
| Dec.–Feb. | 19.5 | 17.6 | 16.7 | |||
| March–May | 26.1 | 24.1 | 22.7 | |||
| June–Aug. | 22.0 | 22.2 | 21.9 | |||
| Sep.–Nov. | 32.4 | 36.1 | 38.6 | |||
| Treatment initiation prior to 1988* | 48.8 | 62.4 | 64.7 | |||
| Inhaled corticosteroids dispensed in the month prior to initial hospitalization | 9.8 | 9.1 | 9.7 | |||
| Oral corticosteroids dispensed in the month after initial hospitalization | 28.5 | 29.0 | 23.1 | |||
| Medication received in the 6 mo prior to initial hospitalization | ||||||
| Oral corticosteroids | 13.8 | 18.0 | 15.7 | |||
| β-agonists | ||||||
| Inhaled | 66.1 | 70.1 | 66.7 | |||
| Oral | 20.1 | 23.4 | 25.4 | |||
| Nebulized | 7.5 | 9.5 | 8.3 | |||
| Theophylline | 35.7 | 44.2 | 43.0 | |||
| Antiallergic | 21.0 | 22.2 | 23.6 | |||
| Ipratropium bromide | 2.0 | 2.7 | 2.5 | |||
| Asthma hospitalization(s) prior to entry in incident cohort | 15.7 | 20.1 | 21.3 | |||
The results of the intent-to-treat analysis are presented in Table 2. This table provides the interval-specific crude and adjusted RRs comparing the subjects who initiated a treatment with inhaled corticosteroids in the month following the initial hospitalization with subjects who did not initiate such therapy in the same period. Only the dispensing of oral corticosteroids and inhaled β-agonists in the 6-mo period preceding the initial hospitalization and the calendar period of the initiation of the treatment were retained for adjustment in the logistic regression model. Table 2 shows that within the first 15 d after the initial hospitalization, users of inhaled corticosteroids were as likely as nonusers to be readmitted for asthma. Between the 16th and the 90th day following the initial hospitalization, users of inhaled corticosteroids were three times less likely to be readmitted for asthma than nonusers (RR = 0.3; 95% CI: 0.2– 0.6). After 3 mo had passed since the initial hospitalization, the subjects who had initiated a treatment with inhaled corticosteroids were no longer protected against a readmission for asthma, however.
| Follow-up Time* | Users | Nonusers | Rate Ratio | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Readmission | Person-days | Rate† | Readmission | Person-days | Rate† | Crude | Adjusted‡(95% CI) | |||||||||
| 1–15 d | 25 | 7,113 | 10.7 | 59 | 22,961 | 7.8 | 1.4 | 1.4 (0.9–2.2) | ||||||||
| 16–90 d | 13 | 33,777 | 1.2 | 121 | 108,080 | 3.4 | 0.3 | 0.3 (0.2–0.6) | ||||||||
| 4th–6th mo | 22 | 37,905 | 1.8 | 79 | 118,349 | 2.0 | 0.9 | 0.9 (0.5–1.4) | ||||||||
| 7th–9th mo | 22 | 34,941 | 1.9 | 64 | 108,789 | 1.8 | 1.1 | 1.1 (0.7–1.8) | ||||||||
| 10th–12th mo | 16 | 32,021 | 1.5 | 61 | 100,665 | 1.8 | 0.8 | 0.8 (0.5–1.4) | ||||||||
| Total | 98 | 145,757 | 384 | 458,844 | ||||||||||||
The crude results of the analysis investigating the effectiveness of regular use of inhaled corticosteroids are presented in Table 3. This table provides the crude rate of a readmission for asthma as a function of the duration of treatment among regular users of inhaled corticosteroids and the crude rate of a readmission for asthma as a function of the time since the initial hospitalization among nonusers. Among regular users, the rate oscillated around 1.5 readmissions per 100 persons per month except during the first 15 d of treatment with inhaled corticosteroids where the rate attained 5.4. Among nonusers, the rate of a readmission was highest during the first 15 d after the initial hospitalization and decreased thereafter.
| Regular Use of Inhaled Corticosteroids | No Use | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Time Since Initiation of Therapy* | No. of Readmissions | Person-days | Rate‡ | Time Since Initial Hospitalization† | No. of Readmissions | Person-days | Rate‡ | |||||||
| 1–15 d | 24 | 13,480 | 5.4 | 1–15 d | 78 | 25,233 | 9.4 | |||||||
| 16–30 d | 8 | 13,201 | 1.8 | 16–30 d | 27 | 23,050 | 3.6 | |||||||
| 31–60 d | 14 | 25,693 | 1.7 | 31–60 d | 44 | 41,815 | 3.2 | |||||||
| 61–90 d | 12 | 24,826 | 1.5 | 61–90 d | 43 | 37,467 | 3.5 | |||||||
| 4–6 mo | 8 | 22,658 | 1.1 | 4–6 mo | 59 | 97,653 | 1.8 | |||||||
| 7–9 mo | 8 | 9,724 | 2.5 | 7–9 mo | 45 | 82,152 | 1.7 | |||||||
| 10–12 mo | 3 | 5,608 | 1.6 | 10–12 mo | 37 | 70,688 | 1.6 | |||||||
| 13–24 mo | NA§ | 13–24 mo | 74 | 217,328 | 1.0 | |||||||||
| Total | 77 | 115,190 | 407 | 595,386 | ||||||||||
Figure 1 provides the adjusted RRs of a readmission for asthma as a function of the duration of treatment with inhaled corticosteroids (the intervals of treatment duration correspond to those presented in Table 3). We retained in the model only the covariates that were acting as confounders, i.e., the follow-up time, dispensing of inhaled β-agonists and of oral corticosteroids in the 6 mo preceding the initial hospitalization as well as the subject's age at initial hospitalization. The dispensing of inhaled corticosteroids in the month preceding the initial hospitalization was not a modifier of the RRs of a readmission for asthma. Figure 1 shows that the adjusted RRs were of similar magnitude for the intervals covering the period from 16 d to 6 mo of regular use of inhaled corticosteroids. We therefore collapsed these four intervals and estimated only one RR. Similarly we collapsed the two last intervals that included, respectively, only eight and three cases still regularly treated with inhaled corticosteroids at the time of readmission (see Table 3).
Fig. 1. Adjusted rate ratios of a readmission for asthma contrasting regular use of inhaled corticosteroids with no use; effectiveness analysis.
[More] [Minimize]Table 4 provides the crude and adjusted RRs of a readmission for asthma where the duration of treatment was regrouped in only three intervals. The adjusted RRs were estimated with a logistic regression model including the same variables as in the model used to estimate the RRs presented in Figure 1. In the first interval of treatment duration, users of inhaled corticosteroids were compared with all nonusers, while in the second and third intervals, users were compared with nonusers who had been followed for at least 16 d and 6 mo, respectively. Table 4 shows that subjects who had been using inhaled corticosteroids for less than 15 d were as likely as nonusers to be readmitted for asthma (RR = 1.2; 95% CI: 0.8–1.8). The subjects who had been treated regularly with inhaled corticosteroids for 16 d up to 6 mo were, however, 40% less likely to be readmitted for asthma than nonusers of these medications (RR = 0.6; 95% CI: 0.4–0.9). After 6 mo of regular use of inhaled corticosteroids the RR returned to 1.3.
| Duration of Treatment | Regular Use | No Use* | Rate Ratio | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Readmissions | Person-days | Readmissions | Person-days | Crude | Adjusted† | 95% CI | ||||||||
| 1–15 d | 24 | 13,480 | 407 | 595,386 | 2.6 | 1.2 | 0.8–1.8 | |||||||
| 16 d–6 mo | 42 | 86,378 | 329 | 570,153 | 0.8 | 0.6 | 0.4–0.9 | |||||||
| 7–12 mo | 11 | 15,332 | 156 | 370,168 | 1.7 | 1.3 | 0.7–2.4 | |||||||
In order to investigate the presence of confounding by severity in these RR estimates we compared the distribution of other antiasthma medications between users and nonusers of inhaled corticosteroids. The results presented in Table 5 show that regular users of inhaled corticosteroids tended to receive more prescriptions of oral corticosteroids, inhaled and nebulized β-agonists, theophylline, and ipratropium bromide than nonusers and this trend increased over time as shown by the increasing values of the ratio measure.
| Use of Inhaled Corticosteroids | ||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Initiation of Therapy* | After 3 mo | After 6 mo | After 12 mo | |||||||||||||||||||||
| Yes | No | Yes | No | Yes | No | Yes | No | |||||||||||||||||
| Number of subjects | 1,098 | 1,820 | 997 | 1,199 | 175 | 985 | 61 | 742 | ||||||||||||||||
| Other antiasthma drugs | Percentage† | Ratio‡ | Percentage† | Ratio‡ | Percentage† | Ratio‡ | Percentage† | Ratio‡ | ||||||||||||||||
| Oral corticosteroids | 22.9 | 9.4 | 2.4 | 25.7 | 8.6 | 3.0 | 17.1 | 2.6 | 6.6 | 9.8 | 1.7 | 5.8 | ||||||||||||
| β-agonists | ||||||||||||||||||||||||
| Inhaled | 64.9 | 45.3 | 1.4 | 76.8 | 40.4 | 1.9 | 82.3 | 36.1 | 2.3 | 80.3 | 34.9 | 2.3 | ||||||||||||
| Oral | 10.8 | 11.4 | 0.9 | 6.9 | 9.0 | 0.8 | 2.9 | 6.8 | 0.4 | 3.3 | 6.7 | 0.5 | ||||||||||||
| Nebulized | 10.6 | 6.1 | 1.7 | 10.2 | 6.3 | 1.6 | 10.3 | 4.4 | 2.3 | 9.8 | 4.2 | 2.3 | ||||||||||||
| Theophylline | 30.2 | 26.3 | 1.1 | 30.3 | 24.1 | 1.3 | 30.9 | 16.7 | 1.9 | 32.8 | 15.2 | 2.2 | ||||||||||||
| Antiallergic agents | 16.9 | 17.6 | 1.0 | 16.5 | 18.0 | 0.9 | 13.7 | 14.0 | 1.0 | 21.3 | 12.2 | 1.7 | ||||||||||||
The intent-to-treat analysis shows that asthmatics who initiated a treatment with inhaled corticosteroids in the month following their initial hospitalization were three times less likely to be readmitted for asthma in the 16 to 90 d after the initial hospitalization. During the first 15 d and during the period of 4 to 12 mo after the initial hospitalization, inhaled corticosteroid use initiated within 1 mo of hospitalization did not affect the risk of a readmission for asthma. The results from a more precise analysis investigating the effectiveness of regular use of inhaled corticosteroids initiated within 2 yr of hospitalization are similar. It shows that inhaled corticosteroids required a minimum of 15 d to protect against readmissions for asthma; that subjects who have been treated regularly with this medication for at least 16 d and as long as 6 mo were 40% less likely to be readmitted for asthma and that the effect waned after 6 mo of regular treatment.
Our results point in the same direction as those reported by others who found on average 50% reduction in asthma hospitalizations among users of inhaled corticosteroids (24, 25, 29). The first two studies contained only few subjects and a very limited number of hospitalizations for asthma. The third study was much larger in size but did not take into account the changing effect of inhaled corticosteroids over time, nor did the investigators control for previous hospitalizations for asthma which are recognized as a strong predictor of asthma hospitalization. Using an ecological design, Wennergren and coworkers and Garrett and coworkers found in Sweden and in New Zealand, respectively, a decrease in the rate of hospitalization for asthma (26, 27). The authors attributed this decrease to the increased use of inhaled corticosteroids observed at the same time. In a case-control study it was also found that inhaled corticosteroids reduce the risk of fatal or near fatal asthma (32). On the other hand, Godfrey and coworkers as well as Mitchell and coworkers could find no association between use of inhaled corticosteroids and the risk of hospitalization for asthma (22, 28). The lack of precision in the assessment of the use of inhaled corticosteroids, the lack of an appropriate control group as well as inappropriate control of confounding may explain the negative results of these studies.
In our effectiveness analysis, the exposure to inhaled corticosteroids was continuously updated, and regular users of inhaled corticosteroids were compared with nonusers of these medications. The patient's severity was assessed prior to the initial hospitalization. Because our cohort was selected from a source population of newly treated asthmatics, it was also possible to adjust for previous hospitalizations for asthma as well as previous treatment with inhaled corticosteroids. Moreover, our study is the first to estimate the effectiveness of inhaled corticosteroids as a function of the duration of therapy and to show that patients may have to wait up to 15 d before being protected against a readmission to hospital for asthma. This result is consistent with the minimum length of time required to substantially affect chronic inflammation (33).
The disappearance of the preventive effect of inhaled corticosteroids after several months of use deserves further discussion. The most plausible explanation is that this result is biased. In the intent-to-treat analysis, the exposure to inhaled corticosteroids was assessed within the first month after the initial hospitalization and was not updated thereafter. Misclassification of exposure could easily have resulted in an underestimation of the effect of inhaled corticosteroids, as exposed subjects may have stopped their therapy in the first year while unexposed subjects may have started to take inhaled corticosteroids. An analysis limited to subjects initiating regular therapy with inhaled corticosteroids in the month immediately after the initial hospitalization might have provided more accurate results, but was not possible to perform because of lack of power.
In the effectiveness analysis, where the use of inhaled corticosteroids was assessed much more precisely, we also believe that the estimate of the RR is biased, but that the bias is of a different nature than the one operating in the intent-to-treat analysis. Subjects using inhaled corticosteroids regularly for 6 mo or more seem to have more severe asthma than patients who have discontinued their treatment or subjects not treated with these medications. The differences between groups in use of other asthma medications over time illustrated in Table 5 support the hypothesis that the RRs for the 7 to 12 mo period is biased by the severity of asthma and that the bias resulted in an underestimation of the protective effect of inhaled corticosteroids. It was, however, not possible to adjust (by regression) this RR estimate for the differences in severity that we observed because there were only 11 subjects who were readmitted for asthma after 6 mo of regular therapy with inhaled corticosteroids. The distribution of the use of other asthma medications among these 11 cases did not have enough variability to allow for statistical adjustment.
Our study was based entirely on computerized data obtained from administrative files of the Health Ministry of Saskatchewan. To ensure the quality of the computerized data, a series of checks are made on each prescription submitted for reimbursement and a random sample of reimbursed prescriptions is selected every week for validation with the beneficiary (30, 34). The data, however, represent dispensed prescriptions and may not correspond exactly to the medications actually taken. This may lead to nondifferential misclassification of the exposure, and an underestimation of the RRs. In addition, the variables used to adjust for the severity of asthma at baseline (at initial hospitalization) were selected from those available in computerized databases. We did not have access to clinical or physiologic measures of asthma severity. However, some of the medications for asthma, and especially use of inhaled β-agonists, are well recognized to be markers of asthma severity or uncontrolled disease (8-10). Recently, Cockcroft and Swystun stated that the level of asthma severity should be measured by the quantity of medication needed to reach and keep asthma under control (35).
We also did not have access to the cause of death of the 15 persons who died during the follow-up of the posthospitalization cohort, because this information is not recorded in the Saskatchewan databases. Censoring persons who died from asthma may have underestimated the RRs by decreasing the number of nonusers of inhaled corticosteroids who were readmitted to hospital, if we assume that subjects who died were more likely to be nonusers of inhaled corticosteroids.
Because of the lack of information on the duration of prescriptions in the databases, we had to assume a fixed duration for each prescription of inhaled corticosteroids (i.e., 90 d). The results of the analysis where we examined the effect of regular use of inhaled corticosteroids are, to a certain extent, dependent on the “duration” that we retained. In order to assess the impact of this assumption, we reestimated the seven time-specific RRs of inhaled corticosteroids, assuming that the duration of each prescription of inhaled corticosteroids was either 60 d or 120 d. The estimates of the RRs were unaffected by this sensitivity analysis. However, due to our definition of “regular use” it was not possible to distinguish between regular and irregular use of inhaled corticosteroids within a treatment period of 90 d. This implies that any difference between regular and nonregular uses can not be made during the first 90 d of therapy, leading probably to an underestimation of the RRs. Moreover, because of missing information on the duration of prescriptions, we were not able to assess precisely the dose of inhaled corticosteroids and perform a dose–response analysis. In any case, this analysis would have been difficult to perform because we did not have sufficient power to simultaneously study the effect of the dose and the duration of therapy.
The results of our study are important in that the frequency of hospitalizations for asthma has increased over the last decade in several countries (11-17) and represent an important part of the resources allocated to the treatment of this disease. In Canada, the rate of hospitalization for asthma has increased by about 40% from 1980 to 1988, to attain a rate of 216 hospitalizations per 100,000 population per year for Canadians between 15 and 34 yr of age (36-38). In the United States the costs related to asthma hospitalizations in 1990 were estimated at 1.6 billion dollars (in 1990 dollars), representing 24% of the total yearly costs associated with asthma (19), while in Canada, these costs were estimated at 84 million dollars (in 1990) representing 28% of the direct costs of asthma (20). Moreover, some studies have reported that the increase in the rate of hospitalization for asthma, especially among children, may be due to an increase in the readmission rate (13, 39). The results of our study may suggest that the high rate of hospitalization for asthma is due, at least in part, to an underutilization of inhaled corticosteroids. Only 40% of the subjects in our cohort had received a prescription of inhaled corticosteroids in the first year following their initial hospitalization.
In summary, we found that regular use of inhaled corticosteroids can substantially reduce the rate of readmission for asthma, events which are severe and costly. These results further support recommendations made by recent guidelines for the treatment of asthma: inhaled corticosteroids are the treatment of choice to bring and keep asthma under control (8-10). Moreover, these results suggest that patients discharged from hospital should also be treated with oral corticosteroids for the first 2 wk to improve the prevention of a readmission to hospital for asthma while waiting for the beneficial action of inhaled corticosteroids to take effect.
The authors thank Winnane Downey and MaryRose Stang of the Saskatchewan Research Services, Population Health Branch for assistance with the data.
| 1. | Kerrebijn K. F., van Essen-Zandvliet E. E., Neijens H. J.Effect of long-term treatment with inhaled corticosteroids and beta-agonists on the bronchial responsiveness in children with asthma. J. Allergy Clin. Immunol.791987653659 |
| 2. | Juniper E. F., Kline P. A., Vanzieleghem M. A., Ramsdale E. H., O'Byrne P. M., Hargreave F. E.Effect of long-term treatment with an inhaled corticosteroid (budesonide) on airway hyperresponsiveness and clinical asthma in nonsteroid-dependent asthmatics. Am. Rev. Respir. Dis.1421990832836 |
| 3. | Bel E. H., Timmers M. C., Hermans J., Dijkman J. H., Sterk P. J.The long-term effects of nedocromil sodium and beclomethasone dipropionate on bronchial responsiveness to methacholine in nonatopic asthmatic subjects. Am. Rev. Respir. Dis.14119902128 |
| 4. | Haahtela T., Jarvinen M., Kava T., Kiviranta K., Koskinen S., Lehtonen K., Nikander K., Persson T., Reinikainen K., Selroos D., Sorijärvi A., Stenius-Aarniala B., Svahn T., Tammivaara R., Laitinen L. A.Comparison of a beta 2-agonist, terbutaline, with an inhaled corticosteroid, budesonide, in newly detected asthma [see comments]. N. Engl. J. Med.3251991388392 |
| 5. | van Essen-Zandvliet E. E., Hughes M. D., Waalkens H. J., Duiverman E. J., Pocock S. J., Kerrebijn K. F.Effects of 22 mo of treatment with inhaled corticosteroids and/or beta-2-agonists on lung function, airway responsiveness, and symptoms in children with asthma: the Dutch Chronic Non-specific Lung Disease Study Group [see comments]. Am. Rev. Respir. Dis.1461992547554 |
| 6. | Meltzer E. O., Orgel H. A., Ellis E. F., Eigen H. N., Hemstreet M. P.Long-term comparison of three combinations of albuterol, theophylline, and beclomethasone in children with chronic asthma. J. Allergy Clin. Immunol.901992211 |
| 7. | Hargreave F. E., Dolovich J., Newhouse M. T.The assessment and treatment of asthma: a conference report. J. Allergy Clin. Immunol.85199010981111 |
| 8. | AnonymousInternational consensus report on diagnosis and treatment of asthma. National Heart, Lung, and Blood Institute, National Institutes of Health. Bethesda, MD. Publication No. 92-3091, [see comments] [Review]. Eur. Respir. J.51992601641 |
| 9. | Ernst P., Fitzgerald J. M., Spier S.Canadian Asthma Consensus Conference summary of recommendations. Can. Respir. J.3199689100 |
| 10. | AnonymousGuidelines on the management of asthma: statement by the British Thoracic Society, the British Paediatric Association, the Research Unit of the Royal College of Physicians of London, the King's Fund Centre, the National Asthma Campaign, the Royal College of General Practitioners, the General Practitioners in Asthma Group, the British Association of Accident and Emergency Medicine, and the British Paediatric Respiratory Group [see comments] [published errata appear in Thorax 1994 Jan;49(1):96 and 1994 Apr;49(4): 386]. Thorax481993S1S24 |
| 11. | Infante-Rivard C., Esnaola S., Sukia, Roberge D., Baumgarten M.The changing frequency of childhood asthma. J. Asthma241987283288 |
| 12. | Evans R., Mullally D. I., Wilson R. W., Gergen P. J., Rosenberg H. M., Grauman J. S., Chevarley F. M., Feinleib M.National trends in the morbidity and mortality of asthma in the U.S.: prevalence, hospitalization and death from asthma over two decades: 1965– 1984. Chest91198765S74S |
| 13. | Friday G. A., Fireman P.Morbidity and mortality of asthma. Pediat. Clin. North Am.35198811491162 |
| 14. | Buist A. S., Vollmer V. M.Reflections on the rise in asthma morbidity and mortality [comment]. J.A.M.A.264199017191720 |
| 15. | Horwood L. J., Dawson K. P., Mogridge N.Admission patterns for childhood acute asthma: Christchurch 1974–89. N.Z. Med. J.1041991277279 |
| 16. | Strachan D. P., Anderson H. R.Trends in hospital admission rates for asthma in children. B.M.J.3041992819820 |
| 17. | Hyndman S. J., Williams D. R., Merrill S. L., Lipscombe J. M., Palmer C. R.Rates of admission to hospital for asthma [see comments]. B.M.J.308199415961600 |
| 18. | Rea H. H., Scragg R., Jackson R., Beaglehole R., Fenwick J., Sutherland D. C.A case-control study of deaths from asthma. Thorax411986833839 |
| 19. | Weiss K. B., Gergen P. J., Hodgson T. A.An economic evaluation of asthma in the United States [see comments]. N. Engl. J. Med.3261992862866 |
| 20. | Krahn M. D., Berka C., Langlois P., Detsky A. S.Direct and indirect costs of asthma in Canada, 1990 [see comments]. C.M.A.J.1541996821831 |
| 21. | Crane J., Pearce N., Burgess C., Woodman K., Robson B., Beasley R.Markers of risk of asthma death or readmission in the 12 months following a hospital admission for asthma. Int. J. Epidemiol.211992737744 |
| 22. | Mitchell E. A., Bland J. M., Thompson J. M.Risk factors for readmission to hospital for asthma in childhood. Thorax4919943336 |
| 23. | Li D., German D., Lulla S., Thomas R. G., Wilson S. R.Prospective study of hospitalization for asthma: a preliminary risk factor model. Am. J. Respir. Crit. Care Med.1511995647655 |
| 24. | Adelroth E., Thompson S.Advantages of high-dose inhaled budesonide [letter]. Lancet11988476 |
| 25. | Agertoft L., Pedersen S.Effects of long-term treatment with an inhaled corticosteroid on growth and pulmonary function in asthmatic children. Respir. Med.881994373381 |
| 26. | Garrett J., Kolbe J., Richards G., Whitlock T., Rea H.Major reduction in asthma morbidity and continued reduction in asthma mortality in New Zealand: what lessons have been learned? [see comments] [Review]. Thorax501995303311 |
| 27. | Wennergren G., Kristjansson S., Strannegard I. L.Decrease in hospitalization for treatment of childhood asthma with increased use of antiinflammatory treatment, despite an increase in prevalence of asthma. J. Allergy Clin. Immunol.971996742748 |
| 28. | Godfrey, S., L. Balfour-Lynn, and M. Tooley. 1977. Beclomethasone dipropionate aerosol in childhood asthma: a three- to five-year follow-up. Brit. J. Clin. Pharmacol. 4(Suppl. 3):273S. |
| 29. | Donahue J. G., Weiss S. T., Livingston J. M., Goetsch M. A., Greineder D. K., Platt R.Inhaled steroids and the risk of hospitalization for asthma. J.A.M.A2771997887891 |
| 30. | Malcolm E., Downey W., Strand L. M., McNutt M., West R.Saskatchewan Health's linkable data bases and pharmacoepidemiology. Post Marketing Surveillance61993175264 |
| 31. | Greenland S.Modeling and variable selection in epidemiologic analysis. Am. J. Pub. Hlth.791989340349 |
| 32. | Ernst P., Spitzer W. O., Suissa S., Cockroft D., Habbick B., Horwitz R. I., Boivin J. F., McNutt M., Buist A. S.Risk of fatal and near-fatal asthma in relation to inhaled corticosteroid use. J.A.M.A.268199234623464 |
| 33. | Kraan, J., G. H. Koeter, Th. W. v. d. Mark, H. J. Sluiter, and K. de Vries. 1985. Changes in bronchial hyperreactivity induced by 4 weeks of treatment with antiasthmatic drugs in patients with allergic asthma: a comparison between budesonide and terbutaline. J. Allergy Clin. Immunol. 76:628–636. |
| 34. | Strand, L. M., and W. Downey. 1994. Health databases in Saskatchewan. In B. L. Strom, editor. Pharmacoepidemiology, 2nd ed. Wiley, New York. 217–229. |
| 35. | Cockcroft D. W., Swystun V. A.Asthma control versus asthma severity. J. Allergy Clin. Immunol98199610161018 |
| 36. | Wilkins K., Mao Y.Trends in rates of admission to hospital and death from asthma among children and young adults in Canada during the 1980s. C.M.A.J1481993185190 |
| 37. | Bates D. V., Baker-Anderson M.Asthma mortality and morbidity in Canada. J. Allergy Clin. Immunol.801987395397 |
| 38. | Mao Y., Semenciw R., Morrison H., MacWilliam L., Davies J., Wigle D.Increased rates of illness and death from asthma in Canada. C.M.A.J.1371987620624 |
| 39. | Mitchell E. A., Burr D.Comparison of the characteristics of children with multiple admissions to hospital for asthma with those with a single admission. N.Z. Med. J.1001987736738 |
Disclaimer : This study is based in part on data provided by the Saskatchewan Department of Health. The interpretation and conclusions contained herein do not necessarily represent those of the Government of Saskatchewan or the Saskatchewan Department of Health.
Funded by grants from the Medical Research Council of Canada (Grant MA-13477), Boehringer Ingelheim Canada, and Zeneca Canada.
Lucie Blais is the recipient of a postdoctoral scholarship from the National Health and Research Development Program (NHRDP) of Canada.
Samy Suissa is a Senior Scientist of the Medical Research Council of Canada (MRC).
