The number needed to treat provides additional insight on the performance of detection points of asthma exacerbations in self-management plans.

With interest we have read the paper of Kupczyk et al. It is an important contribution to the current understanding of the course and early detection of asthma exacerbations, especially with the advent of internet-based self-management programmes using daily or weekly monitoring. In the current research the authors found that the combination of a 20% decrease in peak expiratory flow (PEF) on 2 consecutive days or a 20% increase in day symptoms on 2 consecutive days optimally detects an exacerbation with 65% sensitivity and 94.9% specificity. However, the authors have not addressed an important aspect of the performance of detection points, namely how often it will lead to a false positive prediction for one true positive prediction. This measure of applicability can be captured by the number needed to treat (NNT). In a recent study on early detection of asthma exacerbations we showed that using 70% of personal best PEF resulted in a NNT of minimally 10, which means that for each predicted exacerbation, 10 false positive signals were given. When using an increase in symptoms of two SDs above the mean symptom score during baseline, the NNT was 17. Based on only sensitivity and specificity the authors conclude their optimal detection point is the combination of PEF and symptoms in an ‘OR’ fashion. However, this worsens the NNT of PEF or symptoms solely, indicating a higher number of false positive predictions of exacerbations per correctly identified exacerbation. In contrast, our study showed that combining information on PEF (70%PB) ‘AND’ symptoms (>2 SD) yielded the best performance with sensitivity 85.1%, specificity 97.2% and relatively low NNTof 6. Another important issue is when an exacerbation is detected. In the current study the authors use values for PEF and symptoms after the first visit to the clinic. Ideally, in a self-management plan, a detection point would identify a future exacerbation before it actually occurs, rather than classify it as such afterwards. Therefore it would have been interesting to know which combination would be optimal using the symptoms and PEF of the assessed values at −2, −5 and −7 days before the exacerbation. In our study the optimal detection point detected future exacerbations 4.1 days before occurrence. The applicability of detection points in practice can be best assessed by including the NNT and the number of days before the start of the exacerbation as measures of performance.