Azithromycin for Bronchial Asthma in Adults
After 3 weeks of receiving study medication, 13% of those randomized to placebo and 29% randomized to azithromycin correctly guessed their allocation; 34% of placebo and 29% of azithromycin subjects guessed incorrectly, and the remainder were unsure of their allocation (P = .27). Subjects randomized to azithromycin or placebo had no significant differences in overall asthma symptoms, AQL, or asthma control (Figure 2 and Table 3). At 1 year, subjects randomized to azithromycin were more likely than placebo subjects to have an AQL score ≥1-unit increase compared with baseline (36% vs 21%). This difference was not statistically significant (P = .335). Compared with subjects randomized by PBRN members (n = 69), subjects randomized by the community allergist (n = 6) were more likely to have been skin tested (49% of PBRN subjects skin tested vs 100% of allergist subjects; P = .024) but otherwise were similar in baseline characteristics, including comparable distribution of skin test results. Removing the allergist-recruited subjects from the RM ANOVA did not alter the outcome results (data not shown).
(Enlarge Image)
Figure 2.
Differences from baseline for asthma symptoms, quality of life, and control. A: Symptoms—rating of overall asthma symptoms for the past 24 hours (negative numbers indicate decreased symptoms and hence improvement). B: Quality of life— Juniper Asthma Quality of Life Questionnaire (AQLQ); positive numbers indicate higher quality of life scores and hence improvement). C: Control—Juniper Mini-Asthma Control Questionnaire (ACQ; negative numbers indicate better control and hence improvement). See Methods for details. Symbols represent the mean paired differences from baseline. Bars represent 95% confidence intervals. *P < 0·05, **P < 0·01, ***P < 0·001 (t tests, placebo versus open label).
Comparing the OL and placebo groups, RM ANOVA found that the effect of intervention and the interaction of intervention and time were significant for symptoms, AQL, and asthma control (P < .05 for each). In univariate analyses, OL group asthma symptoms were significantly improved from month 4.5 to month 12 compared with placebo and AQL was significantly improved from month 6 to month 12, whereas asthma control improvements were less consistent (Figure 2). In multivariate analyses (controlled for age, sex, ever-smoking, and concurrent controller medication use) the OL group reported significant improvements in symptoms, AQL, and asthma control that were maximal at study month 9 and waned somewhat at month 12 (Table 3).
AQL score change of 1 unit or more was a prespecified secondary outcome. AQL ≥ 1-unit improvement was achieved significantly more often in OL than in placebo subjects at the 3-, 6-, and 9-month time points. A similar but less consistent pattern was noted for asthma control scores with a ≥1-unit improvement (Table 3). After adjustment for age, sex, ever-smoking, and asthma controller medication use at 9 months (6 months after treatment completion), 80% of OL patients versus 22% of those enrolled in the placebo arm reported AQL score changes of ≥1-unit improvement (P = .001; number needed to treat [NNT] = 2) and 67% versus 21% reported asthma control score changes of ≥1-unit improvement (P = .023; NNT = 3). AQL and asthma control score improvements of ≥1 unit also were correlated significantly with self-reported asthma improvement at all time points (P < .01 for each).
We performed further exploratory analyses of different increments of change in AQL score up to and including a change of ≥2 units (Figure 3). The results showed that changes in AQL after azithromycin (both randomized or OL) assumed U-shaped distributions, whereas changes in AQL for placebo were skewed to the left, suggesting a binary "all or none" response to azithromycin. Finding patient characteristics that predict a treatment response are potentially important. Other than asthma severity, however, our data yielded no indications that patients' clinical characteristics were predictive of an azithromycin treatment response. For example, we analyzed OL and placebo subjects in a logistic model of AQL change of ≥2 units from baseline as the dependent variable and included age, sex, smoking status, and "infectious asthma" as other possible predictors. In this model, only azithromycin treatment was a significant predictor of AQL ≥2 (P = .026).
(Enlarge Image)
Figure 3.
Improvement in asthma quality of life (AQL) after azithromycin treatment may be "all or none." AQL change scores from baseline to 12 months after enrollment (9 months after treatment completion) are defined as follows: "AQL change <0" = AQL change worse than baseline; "AQL change 0 < .5" = change between 0 to <.5 units; "AQL change 0.5 < 1" = change between 0.5 to <1 (change of 0.5 unit is considered the minimum clinically important change); "AQL change 1 < 2" = change between 1 and 2 (change >1.5 units represents a large change); "AQL change ≥2" = change of 2 units or more. The contrasting patterns between placebo and open-label azithromycin were statistically significant, as noted in the text. The differences between randomized azithromycin and placebo were not statistically significant, as noted in the text.
This study was not powered to detect significant differences in exacerbation frequency. During the 12-month study period, 51% of subjects reported one or more asthma exacerbations. There were no significant differences between the 3 study groups in exacerbation frequency at any time point or cumulatively.
One subject allocated to placebo was hospitalized for acute coronary syndrome. Another subject allocated to placebo discontinued study medication because of side effects. Compared with placebo, subjects taking azithromycin (randomized and OL combined) reported significantly more nausea (33% vs 9% for placebo), stomach pain (42% vs 12% for placebo), and diarrhea (42% vs 15% for placebo). The majority of these side effects were mild to moderate in severity and no subject taking azithromycin (either randomized or OL) reported discontinuation because of side effects. There were no significant differences in side effect frequency or severity when the arm randomized to azithromycin was compared with the cohort that elected OL azithromycin (Table 4).
Outcomes
Randomized Trial
After 3 weeks of receiving study medication, 13% of those randomized to placebo and 29% randomized to azithromycin correctly guessed their allocation; 34% of placebo and 29% of azithromycin subjects guessed incorrectly, and the remainder were unsure of their allocation (P = .27). Subjects randomized to azithromycin or placebo had no significant differences in overall asthma symptoms, AQL, or asthma control (Figure 2 and Table 3). At 1 year, subjects randomized to azithromycin were more likely than placebo subjects to have an AQL score ≥1-unit increase compared with baseline (36% vs 21%). This difference was not statistically significant (P = .335). Compared with subjects randomized by PBRN members (n = 69), subjects randomized by the community allergist (n = 6) were more likely to have been skin tested (49% of PBRN subjects skin tested vs 100% of allergist subjects; P = .024) but otherwise were similar in baseline characteristics, including comparable distribution of skin test results. Removing the allergist-recruited subjects from the RM ANOVA did not alter the outcome results (data not shown).
(Enlarge Image)
Figure 2.
Differences from baseline for asthma symptoms, quality of life, and control. A: Symptoms—rating of overall asthma symptoms for the past 24 hours (negative numbers indicate decreased symptoms and hence improvement). B: Quality of life— Juniper Asthma Quality of Life Questionnaire (AQLQ); positive numbers indicate higher quality of life scores and hence improvement). C: Control—Juniper Mini-Asthma Control Questionnaire (ACQ; negative numbers indicate better control and hence improvement). See Methods for details. Symbols represent the mean paired differences from baseline. Bars represent 95% confidence intervals. *P < 0·05, **P < 0·01, ***P < 0·001 (t tests, placebo versus open label).
Open-label Cohort
Comparing the OL and placebo groups, RM ANOVA found that the effect of intervention and the interaction of intervention and time were significant for symptoms, AQL, and asthma control (P < .05 for each). In univariate analyses, OL group asthma symptoms were significantly improved from month 4.5 to month 12 compared with placebo and AQL was significantly improved from month 6 to month 12, whereas asthma control improvements were less consistent (Figure 2). In multivariate analyses (controlled for age, sex, ever-smoking, and concurrent controller medication use) the OL group reported significant improvements in symptoms, AQL, and asthma control that were maximal at study month 9 and waned somewhat at month 12 (Table 3).
AQL score change of 1 unit or more was a prespecified secondary outcome. AQL ≥ 1-unit improvement was achieved significantly more often in OL than in placebo subjects at the 3-, 6-, and 9-month time points. A similar but less consistent pattern was noted for asthma control scores with a ≥1-unit improvement (Table 3). After adjustment for age, sex, ever-smoking, and asthma controller medication use at 9 months (6 months after treatment completion), 80% of OL patients versus 22% of those enrolled in the placebo arm reported AQL score changes of ≥1-unit improvement (P = .001; number needed to treat [NNT] = 2) and 67% versus 21% reported asthma control score changes of ≥1-unit improvement (P = .023; NNT = 3). AQL and asthma control score improvements of ≥1 unit also were correlated significantly with self-reported asthma improvement at all time points (P < .01 for each).
We performed further exploratory analyses of different increments of change in AQL score up to and including a change of ≥2 units (Figure 3). The results showed that changes in AQL after azithromycin (both randomized or OL) assumed U-shaped distributions, whereas changes in AQL for placebo were skewed to the left, suggesting a binary "all or none" response to azithromycin. Finding patient characteristics that predict a treatment response are potentially important. Other than asthma severity, however, our data yielded no indications that patients' clinical characteristics were predictive of an azithromycin treatment response. For example, we analyzed OL and placebo subjects in a logistic model of AQL change of ≥2 units from baseline as the dependent variable and included age, sex, smoking status, and "infectious asthma" as other possible predictors. In this model, only azithromycin treatment was a significant predictor of AQL ≥2 (P = .026).
(Enlarge Image)
Figure 3.
Improvement in asthma quality of life (AQL) after azithromycin treatment may be "all or none." AQL change scores from baseline to 12 months after enrollment (9 months after treatment completion) are defined as follows: "AQL change <0" = AQL change worse than baseline; "AQL change 0 < .5" = change between 0 to <.5 units; "AQL change 0.5 < 1" = change between 0.5 to <1 (change of 0.5 unit is considered the minimum clinically important change); "AQL change 1 < 2" = change between 1 and 2 (change >1.5 units represents a large change); "AQL change ≥2" = change of 2 units or more. The contrasting patterns between placebo and open-label azithromycin were statistically significant, as noted in the text. The differences between randomized azithromycin and placebo were not statistically significant, as noted in the text.
Exacerbations
This study was not powered to detect significant differences in exacerbation frequency. During the 12-month study period, 51% of subjects reported one or more asthma exacerbations. There were no significant differences between the 3 study groups in exacerbation frequency at any time point or cumulatively.
Serious Adverse Events and Side Effects
One subject allocated to placebo was hospitalized for acute coronary syndrome. Another subject allocated to placebo discontinued study medication because of side effects. Compared with placebo, subjects taking azithromycin (randomized and OL combined) reported significantly more nausea (33% vs 9% for placebo), stomach pain (42% vs 12% for placebo), and diarrhea (42% vs 15% for placebo). The majority of these side effects were mild to moderate in severity and no subject taking azithromycin (either randomized or OL) reported discontinuation because of side effects. There were no significant differences in side effect frequency or severity when the arm randomized to azithromycin was compared with the cohort that elected OL azithromycin (Table 4).