Confounding by indication and channeling over time: the risks of beta 2-agonists

Does influenza vaccination prevent asthma exacerbations in children?

OBJECTIVE

Influenza can exacerbate asthma, particularly in children. The effectiveness of influenza vaccine in preventing influenza-related asthma exacerbations, however, is not known. We evaluated influenza vaccine effectiveness in protecting children against influenza-related asthma exacerbations.

STUDY DESIGN

We conducted a population-based retrospective cohort study with medical and vaccination records in 4 large health maintenance organizations in the United States during the 1993-1994, 1994-1995, and 1995-1996 influenza seasons. We studied children with asthma who were 1 through 6 years of age and who were identified by search of computerized databases of medical encounters and pharmacy dispensings. Main outcome measures were exacerbations of asthma evaluated in the emergency department or hospital.

RESULTS

Unadjusted rates of asthma exacerbations were higher after influenza vaccination than before vaccination. After adjustment was done for asthma severity by means of a self-control method, however, the incidence rate ratios of asthma exacerbations after vaccination were 0.78 (95% CI: 0.55 to 1.10), 0.59 (0.43 to 0.81), and 0.65 (0.52 to 0.80) compared with the period before vaccination during the 3 influenza seasons.

CONCLUSIONS

After controlling for asthma severity, we found that influenza vaccination protects against acute asthma exacerbations in children.

Beta agonists, inhaled steroids, and the risk of intensive care unit admission for asthma

Although inhaled corticosteroid (ICS) use is associated with a decreased risk of hospitalization for asthma, the impact of ICS on the risk of life-threatening asthma exacerbation is less clear. The effect of ICS and inhaled beta agonist (IBA) dispensing on the risk of intensive care unit admission for asthma, a surrogate for life-threatening exacerbation, is evaluated. Using computerized International classification of diseases (ICD)-9 discharge diagnoses, a cohort of all 2,344 adult Northern California members of a health maintenance organization hospitalized for asthma over a 2-yr period were identified. Computerized pharmacy data was used to ascertain asthma medications dispensed during the 3-,6-, and 12-month intervals preceding index hospitalization for asthma. During the 3-months preceding hospitalization, a minority of subjects had no IBA units dispensed (34%), with 14% receiving low level (1 unit), 20% medium level (2-3 units), and 32% high level (> or = 4 units) therapy. A substantial proportion received no ICS units (55%), whereas 13% had low, 16% medium, and 15% high level therapy. In multiple logistic regression analysis, high level IBA use was associated with a greater risk of intensive care unit (ICU) admission for asthma after controlling for asthma severity. There was no relationship, however, between low or medium level IBA use and ICU admission. Conversely, medium level and high level ICS use were associated with a reduced risk of ICU admission. Analysing 6- and 12-month medication dispensing data, similar risk patterns were observed. Inhaled corticosteroid dispensing was associated with reduced risk of intensive care unit admission among adults hospitalized for asthma, whereas the opposite applied for high dose beta agonist usage. This suggests that ICS prescription to adults with moderate-to-severe asthma could reduce the risk of life-threatening exacerbation.

Characteristics of users of inhaled long-acting beta 2-agonists in a southern European population

We characterized the population of users of inhaled long-acting beta 2-agonists in the region of Friuli-Venezia Giulia, in Italy, and assessed changes in asthma treatment and control after initiating long-acting beta 2-agonists. All residents using formoterol or salmeterol between 1992 and 1996 were identified in the regional Health Databases. Utilization rates of asthma medications and hospitalization rates for asthma were computed for the year before and after the date of the first long-acting beta 2-agonist prescription. There were 3803 users of formoterol and 20,054 users of salmeterol. Overall, 65% of users were older than 54 years of age. All formoterol users and 86% of salmeterol users received their first prescription for the respective drug during the study period (new users). Among these new users, 50% had not received any asthma drug during the 4 months preceding the start of long-acting beta 2-agonist administration. Prior 1 yr utilization rates of asthma medications and hospitalization rates for asthma were greater among new users of long-acting beta 2-agonists than among new users of salbutamol and xanthines. In addition, formoterol new users had higher prior use of asthma drugs than new users of salmeterol. One year prior hospitalization rates for asthma were also higher among formoterol than salmeterol new users with rate ratios of 1.7 (95% CI 1.3-2.2) for patients younger than 45 and 1.5 (1.2-1.9) for older patients. Use of short-acting beta 2-agonists, oral steroids and xanthines significantly declined after starting formoterol, whereas the use of inhaled steroids increased after the start of either formoterol or salmeterol. Asthma hospitalizations decreased by 32% in patients under age 45, by 43% in older patients, during the year following the start of formoterol, and by 15% and 24%, respectively, after the start of salmeterol. We conclude that long-acting beta 2-agonists were mainly prescribed to middle-aged and elderly patients and that formoterol appeared to be preferentially prescribed to patients with more severe asthma than salmeterol. Changes in asthma treatment and reduction in hospitalization rates for asthma after starting formoterol and salmeterol are compatible with an improvement in the control of asthma.

Assessment and control for confounding by indication in observational studies

In the evaluation of pharmacologic therapies, the controlled clinical trial is the preferred design. When clinical trial results are not available, the alternative designs are observational epidemiologic studies. A traditional concern about the validity of findings from epidemiologic studies is the possibility of bias from uncontrolled confounding. In studies of pharmacologic therapies, confounding by indication may arise when a drug treatment serves as a marker for a clinical characteristic or medical condition that triggers the use of the treatment and that, at the same time, increases the risk of the outcome under study. Confounding by indication is not conceptually different from confounding by other factors, and the approaches to detect and control for confounding--matching, stratification, restriction, and multivariate adjustment--are the same. Even after adjustment for known risk factors, residual confounding may occur because of measurement error or unmeasured or unknown risk factors. Although residual confounding is difficult to exclude in observational studies, there are limits to what this "unknown" confounding can explain. The degree of confounding depends on the prevalence of the putative confounding factor, the level of its association with the disease, and the level of its association with the exposure. For example, a confounding factor with a prevalence of 20% would have to increase the relative odds of both outcome and exposure by factors of 4 to 5 before the relative risk of 1.57 would be reduced to 1.00. Observational studies have provided important scientific evidence about the risks associated with several risk factors, including drug therapies, and they are often the only option for assessing safety. Understanding the methods to detect and control for confounding makes it possible to assess the plausibility of claims that confounding is an alternative explanation for the findings of particular studies.

Optical illusions from visual data analysis: example of the New Zealand asthma mortality epidemic

The abundance of health-related statistics routinely collected worldwide invites their misuse from haphazard associations between secular trends of these data. This misuse is often compounded by assessing these associations simply on the basis of a visual inspection of the data. The visual approach to data analysis, known to have several pitfalls, is particularly tempting in the context of asthma where it has often been used. For example, the epidemic of asthma deaths that occurred in New Zealand during the last two decades has been imputed to fenoterol, a medication for asthma, on the basis of a visual assessment of ecological data. The simultaneity of time trends in the asthma death rate and fenoterol market share in that country formed an important part of the statistical basis of the evidence. We verified whether the results of such visual analyses are corroborated by more objective quantitative statistical methods of analysis. We reanalyzed these same data, namely the time trend data of New Zealand asthma death rates, fenoterol market share, sales of beta-agonists and inhaled corticosteroids, measured yearly for the 16-year span 1976-1991, using Poisson weighted loglinear regression. We found that the protective effect of inhaled corticosteroids (rate ratio 0.5 per canister per month; 95% confidence interval 0.4 to 0.7; p = 0.0001) was more closely associated with changes in asthma mortality than either fenoterol (RR 2.7 per canister per month; 95% CI: 0.9 to 7.5; p = 0.06) or all beta-agonists combined (RR 1.6; 95% CI: 0.8 to 3.0; p = .19). We conclude from this quantitative analysis that these ecological asthma mortality data provide evidence of a stronger association with inhaled corticosteroids, little used in New Zealand at the onset of the epidemic but used abundantly at its termination, than with fenoterol. This conclusion is diametrically opposite to that found by the visual approach. The quantitative analysis demonstrates that the visual approach to the analysis of ecological data, although seemingly convincing, can be misleading by creating an optical illusion. This purely visual approach to data analysis may thus have serious implications when the resulting scientific information is used to make vital public health and policy decisions.