Page 76 - ADULT-ONSET ASTHMA PREDICTORS OF CLINICAL COURSE AND SEVERITY

ADULT-ONSET ASTHMA – PREDICTORS OF CLINICAL COURSE AND SEVERITY
from 21% to 81%. The diagnos c accuracy was evaluated for three markers; two of them in more than one study (Table 1): FeNO (6 studies; 349 pa ents) and blood eosinophils (3 studies; 192 pa ents) had pooled AUC es mates of 0.81 (range 0.56-0.89) and 0.78 (0.74- 0.81), respec vely.
META-ANALYSIS: SENSITIVITIES AND SPECIFICITIES
Adults
Su cient data (≥4 studies) to perform meta-analysis were only available for induced sputum as the reference standard. Forest plots of FeNO, blood eosinophils and IgE for detec ng sputum eosinophils ≥3% and ≥2% are presented in Figure 2a-b, with summary ROC curves in Figure 3. Almost all studies had used the “op mal cutpoint” of sensi vity and speci city on the ROC curve to de ne the posi vity threshold of the markers. These thresholds varied widely. For example, the “op mal” threshold for FeNO to detect sputum eosinophils ≥3% ranged from 10ppb to 41ppb.
Summary es mates of sensi vity and speci city of FeNO, blood eosinophils, and IgE for detec ng sputum eosinophils ≥3% and ≥2%, obtained by meta-analysis, are presented in Table 2. They ranged from 0.63 to 0.76 for sensi vity, and from 0.59 to 0.83 for speci city.
When pooling direct comparisons, FeNO was found to be signi cantly more accurate than IgE in detec ng sputum eosinophils ≥2% (4 studies; p=0.025), but not in detec ng sputum eosinophils ≥3% (5 studies; p=0.34). Pooling of other direct comparisons (FeNO vs. blood eosinophils and IgE vs. blood eosinophils) revealed no signi cant di erences.
Sta s cal tes ng for funnel plot asymmetry revealed no evidence of publica on bias (Appendix 9). Forest plots of sensi vity and speci city of FeNO, blood eosinophils, and IgE for detec ng sputum eosinophilia in subgroups based on smoking, treatment, and asthma severity status are provided in Appendix 10.
Children
The forest plot and summary ROC curve of FeNO for detec ng sputum eosinophils ≥2.5-3% are presented in Figure 2c and 3. Summary es mates of accuracy based on ve studies (318 pa ents) were 0.72 (95%CI 0.24-0.95) for sensi vity and 0.77 (0.20-0.98) for speci city, again without evidence of publica on bias (Appendix 9).
74