39·0°C versus 38·5°C ear temperature as fever limit in children with neutropenia undergoing chemotherapy for cancer: a multicentre, cluster-randomised, multiple-crossover, non-inferiority trial

Fever in neutropenia is the most frequent complication of chemotherapy for cancer. The temperature limit defining fever used clinically varies. A higher limit can avoid unnecessary diagnoses in patients spontaneously recovering from fever. This trial primarily aimed to determine if a limit of 39·0°C ear temperature is non-inferior to 38·5°C regarding safety.This cluster-randomised, multiple crossover, non-blinded, non-inferiority trial was done in six Swiss Paediatric Oncology Group centres (clusters) in Switzerland. Patients (aged 1 to <18 years) with any malignancy and treated with myelosuppressive chemotherapy expected to last 2 months or more were repeatedly randomly assigned (1:1), at the cluster level, to either monthly 39·0°C or 38·5°C ear temperature limits for diagnosis of fever in neutropenia. Diagnosis below the randomised limit was allowed for clinical reasons. Such a diagnosis implied emergency hospitalisation, examinations (including blood culture), as-needed antipyretics, and empirical intravenous broad-spectrum antibiotics. The primary outcome was the rate of fever in neutropenia with safety relevant events (SRE) per chemotherapy year; we also assessed efficacy in terms of rate of fever in neutropenia. The non-inferiority margin was 1·33 for safety, and for effiacy, the superiority margin was 1·00. This trial is registered at ClinicalTrials.gov, number NCT02324231.269 patients were recruited between April 28, 2016, to Aug 27, 2018, until the trial was stopped for success after the second interim analysis. Patients were repeatedly randomly assigned, with 1210 (48%) of 2547 randomisation periods and 92 (47%) of 195 chemotherapy years randomised to 39·0°C. SREs were diagnosed in 72 (20%) of 360 fever in neutropenia episodes (zero deaths, 16 intensive care unit admissions, 22 cases of severe sepsis, and 56 cases of bacteraemia). In 92 chemotherapy years randomised to the 39·0°C fever limit, 151 episodes of fever with neutropenia were diagnosed (1·64 per year), including 22 (15%) with SRE (0·24 per year). In 103 chemotherapy years randomised to 38·5°C, 209 episodes were diagnosed (2·03 per year), including 50 (24%) with SRE (0·49 per year). The mixed Poisson regression rate ratio (RR) of fever in neutropenia with SRE in 39·0°C versus 38·5°C was 0·56 (95% upper confidence bound 0·72). The corresponding RR of fever in neutropenia was 0·83 (95% upper confidence bound 0·98).In children with neutropenia and chemotherapy for cancer, 39·0°C ear temperature was safe and seemed efficacious. For Switzerland and comparable settings, 39·0°C can be recommended as new evidence-based standard fever limit except for patients with acute myeloid leukaemia or haematopoietic stem cell transplantation.Swiss Cancer League (KLS-3645-02-2015).