This study is a further analysis of the TOPPS trial data and assesses patients' baseline characteristics at enrolment, burden of thrombocytopenia, presence of fever and presence of minor hemorrhage as potential risk factors for WHO grade 2 to 4 bleeding. Recurrent event analyses have been used to allow assessment of bleeding over the 30-day trial period and to account for multiple bleeds per patient.
MethodsDetails of the UK and Australian TOPPS randomized trial have been published previously .2,8,10 Six hundred patients were randomized to receive either prophylactic platelet transfusions if the platelet count was <10x10 9 /L (n=299), or no prophylaxis (n=301). The trial was approved by independent ethics committees. Although the median number of days with complete data was 30, bleeding data were not complete for all 30 days for all patients. The vast majority of bleeds were inpatient bleeds 8 and, as incomplete records often related to when the patient was at home, no bleeding was assumed on days for which no data were reported.Bleeding types were grouped to describe characteristics of bleeding: skin bruising included petechiae, purpura, and bruising; gastrointestinal bleeding included blood in the stool, melena and hematemesis; mouth bleeding included oral blood blisters and oropharyngeal bleeding; and other included musculoskeletal or deep tissue bleeding, eyes and visual impairment, pleural tap, and cerebral bleeds.For the modeling (Figure 1), patients were grouped by treatment plan [autologous hematopoietic stem cell transplant (HSCT) and chemotherapy (induction and consolidation)/allogeneic HSCT (myeloablative and reduced intensity)] and diagnosis [acute myeloid leukemia (AML), myeloma, lymphoma and other diagnoses]. In contrast to the TOPPS analysis, only the first day of consecutive bruising or petechiae was counted as a bleed, unless bleeding on the subsequent day was worse.
Modeling analysisNegative binomial regression 11 was used to model baseline characteristics associated with the number of days of grade 2-4 bleeding (analysis 1). The risk factors considered are shown in Table 1. Previous HSCT (defined as any prior transplant) and relapsed disease were also considered.Recurrent event analysis 12 was used to model the hazard of a grade 2-4 bleed on any one day, accounting for previous bleeds (analyses 2a, 2b and 2c). Robust sandwich variance estimates 12 were used for the hazard confidence intervals. Analysis 2a was the main recurrent event analysis. Analysis 2b was performed on a subset of bleeding records in which the platelet count from all 3 previous days was reported. Analysis 2c was a further subset analysis in which patients' temperature from all 3 previous days Risk factors for bleeding and platelet transfusion haematologica | 2015; 100(6) 741 Figure 1. Shows the number of patients (Np) and number of bleeding records (Nr) included in each part of the analysis.Eight of the 600 patients in the TOPPS trial dataset were not included in the analyses described in this paper: two patients w...