INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

During the 1980s, Australia and New Zealand led the world in deaths from asthma. In Australia, the standardized mortality rate (SMR) for asthma rose to 5.8/100,000/yr in 1989 and then fell to 3.0/100,000/yr in 1996 (1). The epidemic of asthma deaths and near-fatal attacks has been investigated in both countries through case series (2, 3), case-control studies (4), and cohort studies (9, 10). Preventable factors have included inadequate assessment or therapy, discontinuity of medical care, poor adherence with therapy, delay in seeking or receiving help, and inadequate treatment of the fatal attack (3, 4).

There has been controversy about the possible relationship between inhaled -agonists and deaths in asthma. Several case- control studies in New Zealand suggested that the potent and relatively non-₂-selective agonist fenoterol doubled the risk of death from asthma (5). The rise and fall in asthma mortality in New Zealand has been attributed to the rise and fall in the prescription of fenoterol inhalers (11). However, a cohort study conducted in the Canadian province of Saskatchewan suggested that increased asthma deaths and near deaths could be a class effect of -agonists rather than being due to the toxicity of fenoterol alone (12). Nonetheless, further analysis of the Saskatchewan data indicated that the risk of death was much greater among those patients who had only fenoterol prescribed than among those who had only salbutamol prescribed (13).

One problem with both the New Zealand and Canadian studies was in the assessment of exposure to -agonists. The case-control studies determined the exposure of cases from medications prescribed by general practitioners (5) or recorded in hospital notes (6, 7). In Saskatchewan, the investigators obtained exposure information from a computerized prescribing data base (12). It is well recognized that misclassification of exposure can substantially bias estimates of risk. Measurement of drug concentrations in blood should give more accurate information about actual exposure to -agonists. Increased -agonist concentrations have been reported both in patients with acute severe asthma (14) and in cases of death from asthma (15).

Another concern has been the potential for confounding by severity of asthma. After performing a stratified analysis, utilizing markers of chronic asthma severity, Beasley and coworkers (16) argued that confounding alone could not explain the association between fenoterol and asthma death. The increased risk of death found in Saskatchewan also persisted after adjustment for confounding by available markers of asthma severity (17). However, a cohort study in Auckland found that the excess risk of life-threatening attacks associated with fenoterol disappeared after control was applied for the number of hospital admissions, continuous oral steroid use, severity of previous attacks, and race (9). It has since been suggested that this discrepancy may have arisen from differences in the population (older patients attending a single institution) and outcome (predominantly intensive care unit [ICU] admissions rather than deaths) (13).

Good management of asthma involves much more than the prescription of -agonists. The Australian Asthma Management Plan (AAMP) advises physicians to assess asthma severity, achieve best lung function, maintain best lung function by avoiding trigger factors and optimizing medication, develop an action plan, and educate and review patients regularly (18). Key features of the AAMP have been increasingly adopted in clinical practice in Australia. Between 1990 and 1993, the proportion of adult asthma patients with both a peak flow meter (PFM) and an action plan increased significantly, as did the proportion using inhaled corticosteroids (19). However, it is not known whether these clinically desirable changes are in any way responsible for the decline in asthma mortality.

We therefore undertook a case-control study to compare blood -agonist levels in cases of death from asthma with those in asthmatic controls; to estimate the risk of death associated with the reported use of specific classes of medication in asthma; and to estimate the risk of death associated with particular patterns of asthma management, after allowing for confounding by severity.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Cases

Cases of death from asthma were identified through the Victorian Registry of Births, Deaths and Marriages from 1994 to 1996. The inclusion criteria were that asthma appeared on the death certificate and that the age at death was < 60 yr. The death certificates were reviewed by two clinicians to confirm that asthma was the underlying cause of death. We found very good interrater agreement ( = 0.77 for adults and = 0.94 for children) (1). There were 66 potential cases in which one clinician was unsure of the cause of death and which were reviewed by a clinical panel that had access to interviews with general practitioners (GPs) and coroner's records, including postmortem reports when these were available.

In 89 cases the panel decided that asthma was the principal cause of death. A further 116 potential cases were excluded either by both clinical reviewers or by panel because death was attributed to acute myocardial infarction (n = 21), ischemic heart disease with congestive cardiac failure (n = 11), other cardiac conditions (n = 5), chronic obstructive pulmonary disease (COPD) (n = 21), pneumonia (n = 14), pulmonary embolism (n = 5), cystic fibrosis (CF) (n = 2), malignancy (n = 20), injury or poisoning (n = 9), an acute cerebral event with aspiration (n = 4), renal or hepatic failure (n = 3), or sudden infant death syndrome (n = 1).

Questionnaire

The GPs and/or next of kin of the case patients were interviewed with the questionnaire previously used in the South Australian asthma mortality study (20). Both interviews were completed for 41 cases, a GP interview only for 28 cases, and a next-of-kin interview only for 10 cases. The next of kin declined to be interviewed in 16 cases and were lost to follow-up in 12 cases. The next of kin participated in four cases in which the GP declined and in six cases in which no GP could be identified. Thus, some information was available for 79 (89%) of the asthma deaths. In the analyses presented in this report, the next-of-kin questionnaires served as the primary data source for demographic information, psychosocial problems, asthma morbidity, chronic asthma severity, asthma medications, and management.

The next of kin were asked whether the case patients had used any inhaled medicines to help their breathing at any time in the month before death. If the answer was "yes," the next of kin were shown photographs of asthma medications available in Australia at the time, and were asked two further questions: (1) Which of these medications had the patient usually taken in the last month? (2) Did the patient use any of these medications for the last attack of asthma? The replies were coded to distinguish between usual medication and medication used for the last asthma attack.

Medications were classified into the five categories of: (1) inhaled symptomatic medication, which included the bronchodilators salbutamol (albuterol), fenoterol, terbutaline, ipratropium, salmeterol, adrenaline (epinephrine) or isoprenaline delivered by metered dose or dry powder inhaler; (2) nebulized symptomatic medication, which included salbutamol, ipratropium, fenoterol, or terbutaline solutions; (3) inhaled preventive medication, which included beclomethasone, budesonide, fluticasone, or cromoglycate delivered by metered dose or dry powder inhaler; (4) oral symptomatic medication, which included theophylline, salbutamol, terbutaline, or ephedrine tablets or syrups; and (5) oral preventive (steroid) medication, which included prednisolone, prednisone, betamethasone, or dexamethasone tablets.

Controls

The controls were patients presenting with a principal complaint of acute severe asthma who required nebulized bronchodilators in the emergency departments of four participating hospitals in Melbourne between 1994 and 1996. The exclusion criteria were age > 60 yr or documented COPD. A total of 322 potential controls were identified between 1994 and 1996. The panel excluded two controls who subsequently died from asthma and a further seven patients who were deemed to have presented primarily for treatment of COPD, bronchiectasis, CF, hyperventilation, or hernia.

Two hundred and two controls (65%) and 92 of their GPs were interviewed with the South Australian questionnaire (20). Thirty-four controls declined to participate and 79 were lost to follow up; in 13 instances the controls participated but the GP declined; and in 15 instances the controls participated but no GP could be identified. The remaining GPs were not approached. Markers of acute asthma severity (arterial carbon dioxide tension [Pa_CO2], peak expiratory flow, and serum potassium level) were extracted from hospital records whenever these investigations had been conducted. Details of ambulance attendance to both cases and controls were obtained from the records of Ambulance Service Victoria.

₂-Agonist Assays

Blood was available in 35 cases referred to the coroner, either on reception of the body or during postmortem examination at the Victorian Institute of Forensic Medicine (VIFM). In addition, 229 controls underwent venipuncture for clinical purposes. ₂-Agonist assays were undertaken at the VIFM, where whole blood or plasma specimens were analyzed with gas chromatography-mass spectrometry for the simultaneous determination of salbutamol, terbutaline, and fenoterol. A solid-phase extraction procedure was used, giving a limit of quantitation of 1 ng/ml for each analyte in 1 ml of blood and plasma (21).

Statistical Analysis

The questionnaire responses, data extracted from hospital records, and blood salbutamol concentrations were entered into data bases generated with the dBase IV software system (Ashton Tate Corporation, Torrance, CA). The data bases were merged by subject identifiers, and range and logic checks were performed. All subsequent statistical analyses were performed with the SAS version 6.12 software package (SAS Institute, Cary, NC) (22). Categorical variables were compared in contingency tables and associations assessed with chi-square tests. The means (or medians) of continuous variables were compared with student's t tests or equivalent nonparametric tests.

Estimates of the risk of death were adjusted by multiple logistic regression. All models included demographic variables (age, sex, country of birth, employment, residential address), smoking status, and psychosocial factors (drinking and family problems). The models for usual medications included hospital admissions in the last year. The models for written action plans and medications for an attack included both hospital admissions in the last year and usual oral preventive medication in the last month, to control for confounding by asthma severity.

Because the distribution of blood salbutamol concentrations was highly skewed, values were log-transformed before further analysis. Multiple linear regression models were developed to adjust salbutamol concentrations for age, sex, mode of administration (inhaled/intravenous), ambulance attendance, and markers of asthma severity (hospital admissions in the last year and usual oral preventive medication). A value of p < 0.05 was accepted as statistically significant.

Ethics

The study was approved by the Standing Committee on Ethics in Research on Humans at Monash University and by the ethics committees of the Alfred Hospital, Austin and Repatriation Medical Centre, Box Hill Hospital, Monash Medical Centre, and Victorian Institute of Forensic Medicine.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Description of Cases and Controls

Demographic and psychosocial characteristics of the cases and controls are described in Table 1. The case patients were significantly older and more likely to be male than were the controls. There were no significant differences in employment or country of birth. The case patients were significantly more likely to have smoked and to have drunk alcohol upon arising in the morning, and their next of kin reported that they had more family problems than did controls.

Both cases and controls had high levels of asthma morbidity (Table 2). Although controls were more likely to have lost days from work, school, or usual activities, the median days lost in the preceding year was significantly higher among cases (17.5 d versus 8.5 d, p = 0.01). Cases were significantly more likely to have had persistent or very frequent asthma (more than once a week) than were controls. However, the two groups were reasonably well matched in markers of chronic asthma severity, such as hospital admissions during the preceding 12 mo. The controls were significantly more likely than the cases to have been admitted to the hospital during the month preceding death or the study index admission.

Cases excluded from analysis because of missing data were significantly more likely than those included to have had ambulance attendance (75% versus 36%, p = 0.01). However, there were no significant differences between cases included and those excluded with respect to age, sex, frequency of symptoms, psychosocial problems, hospital admissions, asthma medications, or management. Similarly, there were no significant differences in these variables between cases from Melbourne and those from regional or rural Victoria. Controls excluded from analysis were slightly older than those included (mean age: 32.4 yr versus 29.4 yr, p = 0.06), but there were no significant differences in sex, hospital admissions, Pa_CO2, peak expiratory flow rates, serum potassium, or salbutamol concentrations.

Asthma Management and Medications

Key features of asthma management for the cases and controls are summarized in Table 3. Although case patients were marginally more likely than controls not to have consulted a GP within the 12 mo preceding the study, controls were significantly more likely to have consulted multiple GPs than were case patients. The case patients were marginally less likely than controls to have a peak flow meter at home or to have kept a record of their results. Importantly, the case patients were significantly less likely than controls to have actually used a peak flow meter during the preceding 12 mo. The case patients were also significantly less likely than controls to have a written asthma action plan, and were marginally more likely to have only been given verbal instructions about what to do during a severe asthma attack. After adjustment was made for demographic and psychosocial factors and asthma severity, written action plans were found to be associated with a 70% reduction in the risk of death (Table 4).

Reported use of asthma medications is presented in Table 5. Case patients were nearly four times more likely than controls to have usually taken nebulized symptomatic medication, but only a third as likely to have usually taken other inhaled symptomatic medication during the month preceding death or the study index admission. Either inhaled or nebulized symptomatic medication was usually taken by almost all subjects, and there was no significant difference between cases and controls in this respect. Case patients were nearly three times more likely to have usually taken oral steroids during the preceding month, but there were no other significant differences in usual medication. Conversely, cases were significantly less likely than controls to have used either inhaled symptomatic or nebulized medication or inhaled or oral steroids for their last attack of asthma.

After adjustment for demographic and psychosocial factors and disease severity, each of the usual nebulized symptomatic medications and usual oral steroid treatments were found to be separately associated with a significantly increased risk of death (Table 4). Reported usual use of nebulized symptomatic medication either during the month preceding death or the index admission or for an attack did not change this increased risk (Table 4). The risk associated with either inhaled symptomatic or inhaled preventive medication for an attack was no longer significantly protective after adjustment. However, the use of oral steroids for an attack reduced the adjusted risk of death by 90% (Table 4). Taking oral steroids either usually during the preceding month or for an attack (thus including subjects who increased their usual dosage during the last attack) was still associated with a 30% reduction in the risk of death, but after adjustment this was no longer significant (Table 4).

Blood ₂-Agonist Concentrations

Blood salbutamol concentrations were significantly higher in cases (n = 35) than in controls (n = 229). Geometric mean concentrations (95% CI) were 33.5 ng/ml (95% confidence interval [CI]: 2.4 to 464 ng/ml) in cases and 11.5 ng/ml (95% CI: 1.2 to 107 ng/ml) in controls. The highest concentrations recorded among the cases were 785, 512, 262, and 257 ng/ml, whereas the highest level among the controls was 224 ng/ml. The distributions of blood salbutamol concentrations are shown in Figure 1. The median blood terbutaline level was only 3.9 ng/ml and the maximum level was 49.3 ng/ml among controls (n = 12). Terbutaline was not detected in the cases. Fenoterol was not detected in any blood specimens from either cases or controls.

View larger version (12K): [in this window] [in a new window]   Figure 1.   Distributions of blood salbutamol concentrations in cases (n = 35) and controls (n = 229).

The final complete multivariate model of blood salbutamol concentrations is presented in Table 6. The adjusted ratio of geometric mean concentrations in cases to that in controls was 2.54. Significant independent predictors of salbutamol concentrations were the dosage of nebulized or intravenous salbutamol administered in the emergency department, and ambulance attendance. Although usual oral preventive medication was left in the model to control for possible residual confounding by asthma severity, neither it nor hospital admissions during the preceding 12 mo were significantly related to blood salbutamol concentrations. The timing of specimen collection in relation to the administration of further treatment in the emergency department was not related to blood salbutamol concentrations in the controls.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Our results suggest that written asthma action plans can reduce the risk of death from asthma. Although these plans are a key feature of many national (18) and international guidelines, their benefit with regard to mortality is unlikely to ever be definitively confirmed by a randomized controlled trial. Nonetheless, a systematic review of the available trials concluded that written asthma management plans resulted in a significant reduction in hospitalization (23). We do not believe that written action plans were more likely to be recalled by controls simply because the latter were more likely to have been admitted to a hospital during the month preceding death or the index admission. At the time our study was conducted, there was no systematic inpatient asthma education in any of the participating hospitals.

We also found that monitoring of asthma by using a peak flow meter was less likely in cases of fatal asthma than among controls, although this was no longer significant after adjustment for potential confounders. This could represent poorer care, less interest in self-monitoring, or slower appreciation of deterioration among the cases. There is an ongoing debate about the value of peak flow monitoring as compared with symptom-based action plans (24, 25). Although we collected considerable detail about the elements of subjects' action plans, it is not known whether the plans were based on peak flow or symptom monitoring.

Blood salbutamol concentrations in many cases and some controls were much higher than have been previously reported. These results negate the view that many asthma deaths occur as a result of undertreatment with bronchodilators. The accepted target range for plasma salbutamol concentrations is between 10 and 20 ng/ml in stable asthma. In one study, plasma salbutamol concentrations > 14 ng/ml were found in 12% of patients presenting for emergency treatment of asthma (14). Mean plasma salbutamol concentrations of 22.7 ng/ml have been found in children with acute severe asthma who were given nebulized salbutamol (26). Adults given high-dose, continuously aerosolized salbutamol for acute asthma were found to have mean serum salbutamol concentrations of 37.7 ng/ml (27). Postmortem salbutamol concentrations > 47.8 ng/ml, or postmortem terbutaline concentrations > 110 ng/ml, were found in 14 (21%) of 68 patients dying from asthma or COPD (15).

The difference in salbutamol concentrations between cases of fatal asthma and controls in our study remained highly significant after we adjusted for further administration of salbutamol by ambulance personnel or in the emergency department. This suggests that some cases may have died from salbutamol toxicity caused by unsupervised self-administration of escalating doses. In high concentrations, salbutamol is not entirely ₂-selective, and can cause adverse cardiac (27) and metabolic (28) effects, although not to the same extent as fenoterol (29). We believe that toxic concentrations may have caused lethal tachyarrhythmias in the context of hypoxia and hypokalemia. It is quite possible that in the cases of fatal asthma, escalating doses of salbutamol were self-administered because the fatal attack was refractory to treatment at home. An alternative explanation is that the measured concentrations may have been affected by postmortem redistribution of the drug. However, salbutamol is not a lipophilic drug, and we have previously found no evidence of its significant postmortem redistribution (30).

It could be argued that high salbutamol concentrations alone cannot prove that salbutamol played a role in the asthma deaths in our study, and that the most likely explanation for this finding is confounding by the acute severity of disease. Unfortunately, it was not possible to control for the severity of the fatal attack, because virtually none of the deaths occurred in settings in which objective markers such as Pa_CO2, peak expiratory flow rates, or serum potassium (as a surrogate for acidosis) could be measured. However, we would argue that even in this situation, the chronic severity of the underlying asthma is a more relevant potential confounder. The difference in blood salbutamol concentrations between cases and controls remained significant after we controlled for validated markers of chronic asthma severity, such as hospital admissions during the preceding 12 mo and usual use of oral steroid medication.

Case patients were significantly less likely than were controls to use oral steroids for attacks of asthma. This finding is consistent with a metaanalysis that established the use of systemic steroids early in acute asthma attacks as reducing the need for hospital admission and preventing relapse (31). Case patients were also less likely to use inhaled steroids for attacks, although this was no longer significant after adjustment was made for confounders. Clinical trials have shown that several weeks of treatment with high-dose inhaled corticosteroids is necessary to ease asthma symptoms and improve spirometric results, peak expiratory flow (32), bronchial hyperreactivity, and airway inflammation (33), and that maximal improvement can take months (34). Thus, inhaled steroids are not generally recommended in acute asthma, and our findings confirm that if used alone, they would probably not significantly reduce the risk of death. However, it is quite likely that some of our cases would have survived if they had received oral steroids early in the course of their ultimately fatal attacks.

One disconcerting finding was that patients in cases of fatal asthma appeared to be less likely to use most classes of medications for the last attack. This could in part be explained by their greater use of nebulized symptomatic and oral preventive medication during the preceding month. However, it also raises the possibility of information bias, in that information about cases was obtained from the next of kin, whereas controls were personally interviewed. In the South Australian study (20), proxy reports agreed well with self-reports for observed behaviors. In our study, however, the fatal attack was rarely witnessed by the next of kin, who were therefore unlikely to have been aware of exactly which medications were actually used by our cases. For this reason, we place greater emphasis on the measured salbutamol concentrations in our study than on the next-of-kin reports of symptomatic medication inhaled during attacks as reflecting the use of such medication. In fact, the agreement between reported use of salbutamol and blood concentrations of this drug was poor (data not shown).

A high prevalence of smoking and psychosocial problems among patient with fatal and near fatal asthma has been previously reported (3, 4, 35). However, family problems may have been overreported by next of kin in our study as a consequence of their bereavement. The high levels of asthma morbidity and discontinuity of medical care were also not unexpected, and the two study groups were well matched for the key severity marker of hospital admissions in the year preceding death or study index admision (4, 36). The greater admission rate during the preceding month among controls probably reflects their greater use of health care services, but may have been exaggerated by recall bias. It is also possible that some cases may have delayed their presentation to the hospital, thereby delaying the institution of effective therapy and making resuscitation unavailable in the event of a cardiorespiratory arrest. Although the greater usual use of oral steroids by cases suggests some residual confounding by asthma severity, allowance for this was made in the analysis.

Although there were some missing data, we do not believe that the main findings in our study were substantially affected by selection bias. There were few differences between subjects included and those excluded from the analyses. If anything, the more frequent ambulance attendance on excluded cases might be expected to have resulted in higher blood salbutamol concentrations after unsuccessful resuscitation.

The excluded controls were marginally older than those included, and although this increased the age difference between cases and controls, allowance was made for it in subsequent analyses. There were no important differences in the severity of the acute asthma attack prompting the index admission between included and excluded controls. Unfortunately, this could not be assessed in the cases because of a lack of data (see the previous discussion). Furthermore, excluded controls did not have higher blood salbutamol concentrations than those included in the analysis. The inclusion of two controls who subsequently died from asthma did not substantially alter the results.

Another potential source of selection bias was that although cases were identified from the entire state of Victoria, controls were recruited only from hospitals in metropolitan Melbourne, for logistic reasons. In fact, some of the controls also had residential addresses in regional or rural Victoria. There were no important differences between metropolitan and regional or rural cases. Furthermore, fitting of area of residence in the final multivariate models did not substantially alter the key study findings.

Although this case-control study confirms the value of some key components of modern asthma management plans, further studies are clearly required. Randomized controlled trials of selected components of such plans in high-risk asthma patients would better inform clinical practice. Our findings support the current practice of using inhaled -agonists as required, rather than giving them in regular doses on fixed schedules. We would urge physicians to be cautious about prescribing ever-increasing doses of -agonists, and urge patients not to self-administer escalating doses of -agonists during acute severe asthma attacks without first seeking assistance. Patients who require such intensive therapy should be under close medical supervision, and should also receive oxygen and systemic steroids. More widespread adoption of such sound asthma management practices should further reduce the mortality from asthma.