T4P (Threshold for Platelets)

Research question

In critically ill patients, below what platelet count should platelets be transfused for a planned procedure that might cause bleeding? Background

Intensive care units (ICUs) are the highest platelet transfusion users after cancer services. Platelet transfusions are given to prevent procedure-induced bleeding in patients with low platelet counts. However recent trials outside adult ICUs show giving platelets may cause harm.

Systematic reviews and a literature search for this application reveal a lack of evidence for prophylactic platelet transfusions in critically ill patients. Recent practice guidelines, unable to recommend a platelet count below which giving platelets is beneficial, call for new research. Aims

• To define the platelet count below which platelets should be transfused in critically ill patients undergoing low risk planned invasive procedures. • To explore whether the threshold varies with patient characteristics. Objectives

• To model the platelet transfusion threshold-mortality curve for critically ill patients with planned low-risk invasive procedures. • To evaluate whether the curve varies with the cause of low platelets.

• To evaluate the net monetary benefit (NMB) of practice standardisation at the optimum threshold versus current practice. Methods

An open-label, randomised, Bayesian adaptive, comparative effectiveness trial with an internal pilot across five equally-spaced platelet count thresholds (<10 - <50x10^9/L) in critically ill patients. Participants Patients accepted for admission or within an ICU, requiring an invasive procedure, with a current platelet count <50x10^9/L. Intervention/comparator

Patients will be randomised to one of five platelet thresholds below which they will receive a single adult equivalent dose of platelet transfusion for the index procedure and subsequent procedures during their ICU stay. Outcomes (90 days after randomisation) Primary clinical effectiveness outcome: all-cause mortality

Primary cost-effectiveness outcome: NMB Data collection We will use: • data linkage for routinely collected patient data • additional in-hospital data collection into a case report form • 3 and 12 month post-hospital questionnaires. Statistics We will:

• estimate a continuous non-linear relationship between the allocated platelet count threshold and mortality. The primary estimate will be the location of the optimum of the threshold-mortality curve with a 95% credible interval.

• compare the estimated NMB at the optimum threshold with the NMB across the range of thresholds in the clinician survey. Internal pilot

Over the first 9 recruitment months we will assess recruitment, willingness to randomise, protocol adherence and data quality. Timelines 60 months (6 months set-up, 42 months recruitment, 12 months follow-up, analysis and dissemination). Impact and dissemination

With our patient representatives, the ICU patient group “ICUsteps”, and stakeholder groups including the National Blood Transfusion Committee and the Intensive Care Society, we will write outputs for a range of audiences in different formats. Our results will inform guideline-writing groups internationally (with which we are involved) to rapidly translate our findings into clinical practice.

Demonstrating the threshold-response curve methodology will enable application to many other settings where treatment decisions are made based on physiological parameters.