MedicalResearch.com Interview with:
Barnaby C. Reeves, D.Phil.
Professor of Health Services Research, Clinical Trials & Evaluation Unit
School of Clinical Sciences, University of Bristol
Bristol Royal Infirmary Bristol
Medical Research: What is the background for this study?
Response: Variable decisions are made about when to transfuse patients after cardiac surgery. The circumstances of particular patients influence decisions about whether to give a transfusion.
Transfusion is a life-saving intervention when a patient is experiencing life-threatening bleeding but most patients have only one or two units of red cells transfused. These transfusions are given at varying levels of anaemia. Some doctors prefer to give a transfusion after cardiac surgery when a patient is only mildly anaemic, believing that the transfusion will promote recovery, while other doctors prefer to wait to transfuse until a patient is substantially anaemic, believing that a transfusion may do more good than harm and is wasteful if it is not needed. Therefore, we carried out a randomized controlled trial comparing restrictive (transfuse when haemoglobin <7.5 g/dL) and liberal transfusion thresholds (transfuse when haemoglobin <9.0 g/dL).*
Medical Research: What are the main findings?
Response: We obtained written informed consent before surgery but only randomized participants after surgery, in intensive care, if their Hb dropped below 9 g/dL. (Hence, we recruited over 3,500 patients but randomized only 2007.) This design avoids ‘diluting’ any difference between groups by including participants who would not usually be ‘considered’ for transfusion.
The primary outcome was the occurrence of one or more serious complications: heart attack, stroke, acute kidney injury, bowel infarction, infection; this included/involved 35.1% of the patients in the restrictive-threshold group and 33.0% of the patients in the liberal-threshold group. This slight difference – more in the restrictive group – was not statistically significant.
We then compared the percentages of patients who died; these were 4.2% in the restrictive group and 2.6% in the liberal group. The difference in this secondary outcome was of borderline statistical significance. Frequencies of other secondary outcomes (infections, ischaemic events, days in critical care and hospital, pulmonary complications) were not different in the two groups.
We also carried out some pre-specified sensitivity analyses for the primary outcome and all-cause mortality. The two most important ones aimed to avoid dilution of the difference between groups as a result of patients having transfusions or outcome events before randomization. Excluding patients who were transfused before randomization shifted the treatment effect to favour the liberal threshold more strongly, for both the primary outcome and mortality. Excluding patients who experienced an outcome event in the first 24 hours after randomization did not change the treatment effect for either outcome.
Medical Research: What should clinicians and patients take away from your report?
Response: The headline finding of no difference in the primary outcome allows us to reject the hypothesis that a restrictive threshold is better. The finding can be considered “supportive” of either a restrictive or a liberal threshold. Given the cost of red-cells (and the very small risks of a transfusion-related harm), not giving a transfusion should logically be preferred. However, some findings pointed to patients in the liberal threshold group having better outcomes, including fewer deaths, and there were no adverse consequences of the liberal threshold; these findings have prompted us to hypothesize that the liberal threshold may be better for patients.
Medical Research: What recommendations do you have for future research as a result of this study?
Response: Researchers responsible for four other liberal vs restrictive transfusion trials in cardiac surgery patients should re-analyse their data, excluding patients in both groups who did not breach the liberal threshold. These re-analyses should increase the ability of the trials to detect a difference and make available the results of all of the trials in a consistent and optimal way. These estimates could then be pooled in a meta-analysis (accepting that the thresholds studied varied in the trials).
Researchers could do a larger trial powered for mortality. We think this is relevant because doctors would be likely to act on the difference in mortality that we observed (50% higher risk of death at 90 days) if they were confident that the difference really existed. In fact, a new trial, TRICS-III, has started to recruit in Canada and the USA with a target sample size of 3,592. It has a composite primary outcome of death, heart attack, kidney failure requiring dialysis or stroke. A meta-analysis combining the mortality data from this trial with TITRe2 and other trials (ideally analysed as proposed above) would have sufficient power to detect a difference of the magnitude observed in the TITRe2 trial.
* The trial was funded by the National Institute for Health Research (NIHR) Health Technology Assessment program (ref: 06/402/94). The views and opinions expressed are those of the authors and do not necessarily reflect those of the NIHR Health Technology Assessment program, the National Health Service or the Department of Health.
Gavin J. Murphy, F.R.C.S., Katie Pike, M.Sc., Chris A. Rogers, Ph.D., Sarah Wordsworth, Ph.D., Elizabeth A. Stokes, M.Sc., Gianni D. Angelini, F.R.C.S., and Barnaby C. Reeves, D.Phil. for the TITRe2 Investigators
MedicalResearch.com Interview with: Barnaby C. Reeves, D.Phil. (2015). Is Liberal Transfusion Threshold After Cardiac Surgery Preferable?