Refining the value of secretory phospholipase A2 as a predictor of acute chest syndrome in sickle cell disease: results of a feasibility study (PROACTIVE)

Abstract

Acute chest syndrome (ACS) is defined as fever, respiratory symptoms and a new pulmonary infiltrate in an individual with sickle cell disease (SCD). Nearly half of ACS episodes occur in SCD patients already hospitalized, potentially permitting pre-emptive therapy in high-risk patients. Simple transfusion of red blood cells may abort ACS if given to patients hospitalized for pain who develop fever and elevated levels of secretory phospholipase A2 (sPLA2). In a feasibility study (PROACTIVE; ClinicalTrials.gov NCT00951808), patients hospitalized for pain who developed fever and elevated sPLA2 were eligible for randomization to transfusion or observation; all others were enrolled in an observational arm. Of 237 enrolled, only 10 were randomized; one of the four to receive transfusion had delayed treatment. Of 233 subjects receiving standard care, 22 developed ACS. A threshold level of sPLA2 ≥ 48 ng/ml gave optimal sensitivity (73%), specificity (71%) and accuracy (71%), but a positive predictive value of only 24%. The predictive value of sPLA2 was improved in adults and patients with chest or back pain, lower haemoglobin concentration and higher white blood cell counts, and in those receiving less than two-thirds maintenance fluids. The hurdles identified in PROACTIVE should facilitate design of a larger, definitive, phase 3 randomized controlled trial.

Fig 1. Disposition of Patients in Study

Tx, transfusion, SC, standard care; sPLA2, secretory phospholipase A2; ACS, acute chest syndrome.

Fig 2. Classification Rates for predicting ACS in adults and children

For each threshold level of sPLA2, true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN) were tabulated and used to compute the following rates: accuracy (ACC = (TP+TN)/(TP+TN+FP+FN)), sensitivity (TPR = TP/(TP+FN)), specificity (1 – FPR = TN/(FP+TN)), positive predictive value (PPV = TP/(TP+FP)), and negative predictive value (NPV = TN/(FN+TN)).

Fig 3. Classification trees

Within each age group, data was recursively partitioned into two groups based upon a splitting rule that effectively optimized the prevalence within one of the subgroups. The candidate splits included all possible threshold levels for clinical predictors that were determined to have high predictive ability in this data. The number of subjects and the rate of ACS are reported within each node.