Page 82 - ADULT-ONSET ASTHMA PREDICTORS OF CLINICAL COURSE AND SEVERITY

ADULT-ONSET ASTHMA – PREDICTORS OF CLINICAL COURSE AND SEVERITY
DISCUSSION
We systema cally reviewed studies on the diagnos c accuracy of minimally invasive markers for detec ng airway eosinophilia in asthma. In adults, FeNO, blood eosinophils, and total IgE have been extensively inves gated, but their ability to dis nguish between pa ents with and without airway eosinophilia seems limited, with summary es mates of AUC, sensi vity, and speci city never exceeding 0.8. Other markers, such as VOC-analysis, were reported to be more accurate in single studies, but these results have not yet been replicated. Studies in children are scarce, but ndings for FeNO and blood eosinophils are comparable to those in adults.
Several considera ons deserve a en on. Almost all studies showed risk of bias. These sources of bias are likely to overes mate diagnos c accuracy10, which would mean that the extracted accuracy es mates, although usually moderate, may be even too op mis c. Subop mal repor ng, a common phenomenon for diagnos c accuracy studies15, o en withheld us from a proper evalua on of risk of bias.
Failure to publish is a common phenomenon in diagnos c accuracy studies16. We aimed to reduce the risk of publica on bias by searching trial registries, and by contac ng authors of published studies that seemed to have data from which accuracy es mates could be calculated. This approach was successful. More than one third of the included results were unpublished by the me of our searches. However, this approach also has its limita ons. First, only a minority of diagnos c accuracy studies is currently registered17. Second, most of the included unpublished data came from studies that were, at least par ally, reported and had included at least 50 pa ents. These may di er from smaller studies, or those that do not get published at all. Though we did not observe any di erences between accuracy es mates obtained from published and unpublished data (Appendix 8) and we observed no funnel plot asymmetry (Appendix 9), we cannot completely exclude the possibility of repor ng bias. Drivers of non-publica on are unknown in diagnos c research, but it is likely that studies with lower accuracy es mates have lower chances of ge ng published. Should this be the case, this may have led to further overes ma ons of accuracy.
Overall, nine di erent de ni ons of airway eosinophilia were used across studies, based on di erent thresholds for eosinophilia in induced sputum, BAL and/or EBB. These three airway compartments do not exhibit strong correla ons with regard to eosinophil counts18. Although the diagnos c accuracy of markers may vary across di erent eosinophilia de ni ons, we observed that the summary AUCs were stable when comparing studies using any de ni on of airway eosinophilia, sputum eosinophils ≥3%, or sputum eosinophils ≥2%. There was also considerable heterogeneity regarding the study popula on and test methods. Some studies only included smokers, for example, while others only included non-smokers, and at least four di erent FeNO devices were used. Many studies analysed both pa ents with childhood- and adult-onset asthma. In the la er group, dis nguishing asthma from COPD and ACOS (Asthma-
80