Predictors of the corrected count increment after platelet transfusion, a secondary analysis of the PACER randomized controlled trial.

Platelet transfusion effectiveness is measured by the corrected count increment (CCI), which is the difference between posttransfusion and pretransfusion platelet counts, corrected for body surface area and platelet dose.1 The CCI can be determined at any time posttransfusion but is usually measured after either one or 24 h. Platelet refractoriness is a consistently poor CCI and is usually defined as a one hour CCI below 5.0 or a 24 h CCI below 2.5 on two consecutive occasions.1 Many different patient- and transfusion product-related factors have been associated with lower CCI.2-4 However, transfusion practice has changed over the years, which has likely also changed the relative importance of some of these associations. For example, AB0-incompatibility was consistently shown to lower the CCI, which resulted in the standard practice to give AB0-matched PCs. The majority of previous studies on this subject were specifically on patients with hematologic malignancy, and some were on selected subpopulations of critically ill patients. We aimed to find predictors of lower CCI in patients that received a prophylactic platelet transfusion using PACER trial data. The PACER trial was a multicenter randomized controlled trial on severely thrombocytopenic patients at the hematology ward and the intensive care unit (ICU), in which patients were randomized to receive either no prophylactic platelet transfusion or one unit of prophylactic platelet transfusion prior to central venous catheter (CVC) placement, with catheter-related bleeding as primary outcome.5 The PACER trial was approved by the Amsterdam UMC, University of Amsterdam, medical ethics committee; study conduct was overseen by a data safety and monitoring board; and the trial was registered in the Dutch Trial Register (NL5534). For this analysis, patients in the transfusion arm that did not receive any further platelet transfusions in the 24 h following CVC placement were selected. There were no restrictions in the type of platelet concentrate (PC) transfused, although the Dutch standard during the study period consisted of five-donor pooled buffy coat platelets. Apheresis and/or human leukocyte antigen-matched PCs were transfused whenever indicated. PCs were stored in 100% plasma until June 2018 and in 35% plasma and 65% platelet additive solution (PAS) E from July 2018 onwards. The maximum PC storage time was 7 days. Pretransfusion platelet counts were obtained at most 24 h before transfusion, and posttransfusion platelet counts were obtained 24 h after transfusion to determine the CCI.1 A list of possible CCI predictors was compiled based on a literature search, and the 14 most important potential predictors were selected, limited by sample size and availability of data. Patient-related variables included age, gender, BMI, presence of allo-antibodies, bone marrow depression as primary cause of thrombocytopenia, disseminated intravascular coagulation (DIC), liver failure, renal failure, fever, sepsis, and corticosteroid use. Product-related variables included PC storage length, PC irradiation, and PC storage medium. Bone marrow depression was considered the primary cause of thrombocytopenia if the decrease in platelet count coincided either with disease progression of a hematologic malignancy or with the initiation of chemotherapy. DIC was considered present when the International Society on Thrombosis and Haemostasis DIC score was 5 or higher. Liver failure was defined as a Child-Pugh C classification. Renal failure was considered present when the estimated glomerular filtration rate was below 15 mL/min/1.73m2 and/or renal replacement therapy was provided. Fever was defined as a temperature of 38.5°C or higher, and sepsis was defined according to the Sepsis-2 criteria. The primary analysis was a linear regression model with variable selection using augmented backward elimination. Estimates were adjusted using parameter-wise shrinkage. To assess model stability and bias and variance inflation caused by the variable selection procedure, a bootstrap sensitivity analysis was performed. To account for missing data, a multiple imputation sensitivity analysis was performed. We included 131 transfusion episodes in the complete case primary analysis and an additional 35 in the multiple imputation sensitivity analysis (Table S1). In the primary analysis, sepsis, bone marrow depression, and DIC were the only significant patient-related predictors of a lower CCI (Tables 1 and S2). PC irradiation was the only significant product-related predictor of a lower CCI. Sepsis was associated with a mean (95% CI) lower CCI of 13.5 (6.7–20.2), bone marrow depression was associated with a mean (95% CI) lower CCI of 10.0 (3.4–16.6), DIC was associated with a mean (95% CI) lower CCI of 11.2 (4.2–18.2), and PC irradiation was associated with a mean (95% CI) lower CCI of 5.6 (1.0–10.2). Renal failure was included in the final model but was no longer significant after parameter estimate shrinkage, with a mean (95% CI) lower CCI of 5.1 (−1.6–11.9). The bootstrap sensitivity analysis arrived at the same predictors and comparable parameter estimates as the primary analysis (Tables 1 and S3). In the multiple imputation sensitivity analysis, renal failure was also a significant predictor of lower CCI (Tables 1 and S4). Moreover, presence of allo-antibodies, storage length, and storage medium were selected in three, four, and seven imputed datasets, respectively, but not in the final model. Parameter estimates were similar to the primary analysis and bootstrap sensitivity analysis. Additional details of both the primary and sensitivity analyses, and pre and posttransfusion platelet counts (Table S5) are provided as online supplement. In this prespecified, prospective, observational sub-study of the PACER trial, we found evidence for an association with lower CCI of sepsis, DIC, bone marrow depression, and PC irradiation. These associations held up after thorough model diagnostics. We found ambiguous results for the association between CCI and renal failure, presence of allo-antibodies and PC storage length and medium, and no evidence for an association with age, gender, BMI, liver failure, corticosteroid use, or fever. Previous studies on predictors of CCI were mostly in patients with hematologic malignancy or in a subset of critically ill patients, but not in a combination of both. We were able to identify bone marrow depression as negative predictor of CCI by including a combination of patients with and without bone marrow depression. Moreover, the use of thorough model diagnostics allowed us to be both more certain about the most significant predictors of CCI and provide a nuanced view on the remaining variables included. The fact that the parameter estimates for all selected predictors were similar in both the presumably unbiased global model, the selected model, the bootstrap model, and the multiple imputation model provides extra validity to our results. Although it is recommended to always determine posttransfusion platelet counts, in clinical practice, often prophylactic platelet transfusions before invasive procedures are given without follow-up measurement. Our data may be helpful to select patients that are more likely to have a lower platelet count increment after platelet transfusion, both in clinical practice and for further research. In general, in patients with characteristics known to be associated with a lower platelet count increment, a follow-up platelet count should be measured at all times. There are several limitations to the observations presented here. First, we were unable to account for all previously identified predictors of CCI. This was in part due to the limited sample size, which restricted the number of variables that we were able to include in model building. Another limiting factor was the type of data included in our database. For instance, we did not collect extensive data on co-medication and were thus unable to account for specific antimicrobial medications like amphotericin B and vancomycin. Also, we did not collect data on fluid balance, nor did we account for bleeding. Platelet loss through bleeding and hemodilution or -concentration could have influenced CCI. Because we selected patients that did not receive further platelet transfusions in the 24 h following the initial transfusion, clinically significant bleeding seems not to be a likely explanation for the lower CCI seen in some of our patients. However, we cannot account for differences in fluid balance. Especially patients with sepsis are likely to have large shifts in fluid balance, which is usually positive during early treatment and negative later on. Second, because the proportion of patients with allo-antibodies and the proportion of patients with liver failure were low, our study was underpowered to draw strong conclusions on their relationship with CCI. Third, this was an observational sub-study, which precludes any conclusions on causality. Finally, our results are based on observations from a specific subset of hematology ward and ICU patients requiring CVC placement. While patients at the hematology ward and in the ICU are among those most likely to require platelet transfusions, there are likely to be some differences in baseline characteristics between patients that do and do not require CVC placement, leading to selection bias. On the other hand, previous studies on predictors of CCI did not focus on platelet transfusion prior to invasive procedures. Our results should be more applicable to this specific population, which is representative of many thrombocytopenic patients seen in real-life practice. In conclusion, we identified predictors of 24-h CCI in a prospective sub-study of the PACER trial. The study population included severely thrombocytopenic hematology ward and ICU patients requiring CVC placement for which prophylactic platelet transfusions were given in one arm of the study. Significant predictors of a lower 24-h CCI included sepsis, DIC, bone marrow depression, and PC irradiation. The data will be available on reasonable request; please contact [email protected]. Data S1. Supporting Information. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.