A32-30 Re-Evaluating the Prognostic Value of Symptom Control in Predicting Asthma Attacks: A Patient-Level Analysis of Control Groups of 22 Randomised Trials (ORACLE2)

Introduction Stratifying asthma attack risk is a key management strategy. Established risk factors include prior attacks, elevated type 2 inflammatory biomarkers (blood eosinophils and fractional exhaled nitric oxide [FeNO]), and symptoms measured by the Asthma Control Questionnaire (ACQ). However, the incremental prognostic value of the ACQ-5 total score and its five items has not been clearly quantified after accounting for other risk factors. Moreover, the performance of ACQ-5 alone or as part of a multivariable model for predicting attacks remains uncertain. We hypothesized that certain ACQ-5 symptom items have greater prognostic importance and that ACQ-5, whether alone or combined with other predictors, provides limited incremental prognostic value for asthma attack risk. Methods We analyzed data from the OxfoRd Asthma attaCk risk scaLE (ORACLE2) patient-level meta-analysis, including 6,513 control-arm participants aged ≥12 years from 22 randomized controlled trials of fixed treatment regimens ≥24 weeks. Severe attacks were defined as events requiring ≥3 days of systemic corticosteroids. Univariable and multivariable negative binomial regression models, adjusted for trial and baseline covariates, estimated rate ratios (RRs) with 95% confidence intervals for associations between baseline ACQ-5 mean score (per 0.5-point increase; the minimal clinically important difference) and annualized attack rates. ACQ-5 increments corresponding to clinically meaningful RRs of 1.2 and 1.4 were derived, and each ACQ-5 item was assessed separately. Predictive performance was evaluated using changes in Nagelkerke R² when excluding ACQ-5, attack history, or biomarkers, and by comparing models using each predictor alone. Results The ORACLE2 cohort included 6,513 patients (4,615 attacks over 5,482 person·years); 3,558 had item-level ACQ-5 data. Each 0.5-point ACQ-5 increase was associated with a 9% higher attack rate (adjusted RR 1.09 [1.06–1.12]). Clinically meaningful 20% and 40% risk increases corresponded to ACQ-5 increases of 1.1 [0.8–1.5] and 1.9 [1.5–2.8] points. Wheeze and dyspnea items showed the strongest univariable associations but were not statistically significant after adjustment. Removing ACQ-5 reduced model R² by 0.02, compared with greater reductions when excluding attack history (0.06) or biomarkers (0.04). Models using single predictors showed lower discrimination (R² < 0.22). Conclusions Asthma symptoms measured by the ACQ-5, particularly wheezing and dyspnea, are associated with increased attack risk. However, ACQ-5 adds limited incremental information to a fully adjusted clinical prediction model. Symptom-only and single predictor approaches poorly stratify the risk of severe asthma attacks. Registration : PROSPERO-CRD42021245337 Funding : APQ; FRQS; AIRS; AcMedSci; NIHR This abstract is funded by: APQ; FRQS; AIRS; AcMedSci; NIHR