Applying GRADE for diagnosis 31 2 Study selection, data collection and risk of bias assessment Two authors (MT, HdB) screened abstracts and full-text articles for inclusion. Both read the full text of included studies. One reviewer (MT) completed predefined data extraction tables (Appendix 2) by extracting detailed information about study type, patient characteristics, methods, outcomes and risk of bias. The second reviewer (HdB) checked this process. Discrepancies were resolved by discussion. We used study-design appropriate checklists for risk of bias assessment (table 2). Table 2. Detailed methods per part of the test-treatment strategy Literature search (see Appendices) and selection eligibility criteria Method of risk of bias/quality assessment Diagnostic accuracy [6] Cross-sectional studies (or systematic reviews) that compare sIgE-test with nasal provocation Exclusion: case-control studies McMaster search filters for best balance of sensitivity and specificity in diagnosis [13] QUADAS-2 [14] Test burden Systematic reviews of at least moderate quality [15], reporting on adverse effects of venipunctures Search: MeSH with adverse events as freefloating subheading AMSTAR-2 [15] with appraisal of risk of bias of RCTs and nonRCTs Management Systematic reviews of at least moderate quality, consisting of RCTs (positive score on AMSTAR-2 items 4, 9, 11 and 15) McMaster search filters for best balance of sensitivity and specificity in reviews AMSTAR-2 with appraisal of risk of bias of RCTs and nonRCTs Natural course Prospective cohort studies FPIN prognosis search filter [16] Adapted QUIPS [17] Link between test and management Follow-up studies (reviews, scoping) JBI Critical Appraisal Checklist for Studies Reporting Prevalence data [18] RCT: randomised controlled trial Data Analysis We planned to pool results about diagnostic accuracy with RevMan 5.3. In case of substantial heterogeneity, we planned to present ranges of sensitivity and specificity. For the evidence elements ‘test burden’ and ‘management’ (avoidance measures, antihistamines, corticosteroids) we planned to calculate pooled (standardised) mean differences (MD or SMD) (in continuous outcomes) and risk ratios (RR) (in dichotomous outcomes). For the evidence elements ‘natural course’ and ‘the link between test results and management’, we used descriptive statistics (mean, ranges).
RkJQdWJsaXNoZXIy MTk4NDMw