Restrictive Versus Liberal Transfusion Trials: Are They... : Anesthesia & Analgesia (original) (raw)

We recently conducted an Overview of Systematic Reviews and Meta-analyses reporting mortality outcomes from clinical trials comparing red blood cell (RBC) transfusion strategies.1 The overview identified 19 systematic reviews pooling data from 68 unique randomized controlled trials (RCTs). Within these systematic reviews, 33 meta-analyses reported mortality outcomes. Overall, the 68 RCTs pooled were published between 1956 and 2017, with the number of patients randomized ranging from 30 to 5243, and the number of trial sites ranging from 1 to 73. Albeit in different clinical settings, these 68 RCTs are designed to answer the research question: What is the optimal hemoglobin concentration at which to transfuse RBCs?

In general, the results from RCTs in most clinical settings point to no difference in mortality between patients assigned to restrictive or liberal transfusion strategies. These results have led some in the medical community to conclude that there is no dose-dependent relationship between RBC transfusion and increased mortality and morbidity,2 and therefore the observational studies finding such an association are a result of confounding. However, this conclusion assumes the RBC transfusion is the intervention of interest in these RCTs. As these studies do not compare RBC transfusion to placebo or another intervention, they are not designed to “specifically test the efficacy of transfusion.”3 In other words, they do not answer the question: What effect does transfusion have on mortality?

In addition, our overview uncovered under-recognized issues limiting interpretation from RCTs. Our objective was to draw attention to 4 common and significant issues limiting the conclusions drawn. In doing so, we also challenge the dogma that RCTs, by default, automatically provide the best evidence for patient blood management (PBM).

DISCUSSION

A recent systematic overview of systematic reviews investigated mortality in patients assigned to restrictive and liberal transfusion strategies.1 This sizeable overview uncovered 4 important and underappreciated issues common to RCTs. These are small differences in units of RBC transfused between restrictive and liberal groups; small actual differences in hemoglobin concentrations between restrictive and liberal groups; RBCs transfused before randomization; variation in posttransfusion hemoglobin targets.

Units of Blood Transfused

Many RCTs comparing restrictive and liberal transfusion strategies suffer from small differences in the number of units of blood transfused between groups. Of the 68 RCTs included in our overview, 38 (56%) did not report the number of red cells transfused in the restrictive and liberal arms. Figure 1 demonstrates almost half (n = 14) of trials reporting the units transfused had a mean difference of <1 unit transfused per patient.

Figure 1.:

Mean difference in red cells units transfused between restrictive and liberal strategy groups from 30 trials that reported the number of red cells transfused in each trial arm.

Of the trials with a mean difference below 1 unit, half (n = 7) had a mean difference between groups of half a unit of red cells or less. For example, in 1 trial, 109 patients randomized to the restrictive group were transfused a total of 89 red cell units, while 109 patients randomized to the liberal group received a total of 119 units. Therefore, this trial compared outcomes in a group receiving a mean of 0.82 units per patient to a group receiving 1.09 units per patient.4 It is difficult to comprehend how a difference between groups of <0.5 units of red cells would result in any clinically significant differences in patient outcome.

The trial with the smallest mean difference in units of red cells transfused between groups was 0.08. In this trial, 299 patients in the restrictive group were transfused a mean of 0.78 units of red cells per patient, compared with 0.86 units in the 304 patients in the standard care group.5

Observational studies consistently demonstrate that adverse outcomes associated with red cell transfusion are dose-dependent. With such small differences in the mean amount of blood transfused between groups in clinical trials, it is unlikely that a dose response would be identified. Therefore, it is likely that smaller than expected differences in units of blood transfused between groups will skew overall patient outcomes toward no significant differences.

Hemoglobin Concentration

“It seems hardly probable that a hemoglobin difference <1 g/dL (<10 g/L) may influence the clinical outcome.”6 Although this quote from Leal-Noval et al6 refers to the results of 1 clinical trial, their statement draws attention to a wider issue with RCTs comparing transfusion thresholds. This issue is the small actual differences in hemoglobin thresholds between restrictive and liberal groups. Focusing our attention on 2 points can help better understand this issue. First, the difference in hemoglobin thresholds as planned in the study protocol. Second, the actual difference in hemoglobin thresholds between groups. The actual is usually narrower than the planned.

Figure 2 presents the planned hemoglobin threshold protocols by RCT. Of the 68 RCTs included in our overview, 17 (25%) did not provide enough information to determine what the pretransfusion hemoglobin threshold protocol was. Of trials reporting hemoglobin thresholds, almost half (n = 24) compared thresholds with a difference of 20 g/L. This was made up of 9 trials comparing a restrictive threshold of 70 g/L with a liberal threshold of 90 g/L, 14 trials comparing 80 g/L with 100 g/L, and 1 trial comparing 100 g/L with 120 g/L.

Figure 2.:

Planned restrictive and liberal hemoglobin threshold (g/L) protocols by randomized controlled trial. Blue boxes represent the restrictive group, and red boxes refer to the liberal group.

While the majority of trials have planned pretransfusion hemoglobin differences of 20 g/L between groups, the resulting actual difference between groups is usually lower. For example, 1 trial comparing thresholds of 70 g/L with 90 g/L reported the actual pretransfusion hemoglobin level was 68 g/L in the restrictive group and 79 g/L in the liberal group.7 This difference of 11 g/L is significantly lower than the planned 20 g/L. Similarly, another trial comparing pretransfusion hemoglobin thresholds of 80–100 g/L reported actual thresholds of 79 and 92 g/L, a difference of 13 g/L.8

We consistently found that trials comparing hemoglobin thresholds before transfusion with a planned difference between groups of 20 g/L had differences closer to 10 g/L. A Cochrane systematic review and meta-analysis reported that, of 16 trials presenting the actual difference in hemoglobin thresholds between groups, 6 (38%) had an average difference of ≤10 g/L.9 It is difficult to imagine that a hemoglobin difference of 10 g/L would result in significant changes in patient outcomes.

RBCs Transfused Before Randomization

The timing of randomization varied between individual trials. Some trials randomized patients commencing on admission, some intraoperatively, some after surgery, and others during a portion of a patient’s hospital stay. This creates an issue not often discussed, namely, the implications of comparing patient outcomes between transfusion strategies while ignoring the significant amounts of blood transfused before randomization.

A recent feasibility trial comparing traumatic brain injury patients admitted to intensive care highlights the serious implications of prerandomization transfusions. In this study, patients assigned to the restrictive arm received more blood than patients assigned to the liberal arm.10 Despite this, the study authors reported, “fewer RBC units were administered in the restrictive than in the liberal group.” The timing of patient randomization in this trial, like many other trials, likely masked this important point.

Leal-Noval et al6 drew attention to this overlooked point. They state, “the total number of RBC units transfused over the length of hospital stay (pre- and postrandomization) was finally higher in the restrictive group (164 U; 7.1 U/patient) than that in the liberal group (131 U; 6.2 U/patient).”6 Any clinical trial ignoring the units of blood transfused prerandomization introduces an inaccurate estimate of the transfusion’s effect on outcome.

Furthermore, blood transfused prerandomization may result in even smaller differences between groups in terms of blood administered. For example, the difference in the proportion of patients transfused between trial arms was smaller during the patients’ complete admission (pre-and postrandomization) when compared to their postrandomization hospital stay in the Transfusion Indication Threshold Reduction (TITRe2) RCT (Table).11

Table. - Difference in Transfusion Outcomes When Comparing Complete Hospital Admission to Postrandomization Hospital Stay in the TITRe2 Randomized Controlled Trial

| | Restrictive Transfusion | Liberal Transfusion | | | ---------------------------------------------- | ------------------- | ----- | | Patients transfused red cells | | | | Postrandomization | 53.4% | 92.2% | | Complete admission | 63.7% | 94.9% | | Mean red cell units transfused per participant | | | | Postrandomization | 1.49 | 2.49 | | Complete admissiona | 2.08 | 3.07 |

Values derived from Murphy et al.11

Abbreviation: TITRe2, Transfusion Indication Threshold Reduction.

aValues were derived from Stokes et al.12

It is hard to imagine clinical trials in other settings where researchers investigating the impact of a therapy ignore whether and how frequently patients received the therapy before randomization. For example, a trial studying the administration of intravenous iron in the intensive care unit excluded patients from the trial if they received intravenous iron in the 3 months prior.13 Similarly, another trial comparing antibiotic treatment strategies in intensive care excluded patients who had a course of antibiotics within the previous 24 hours.14

Variation in Posttransfusion Hemoglobin Targets

Of the 4 issues identified, variation in posttransfusion hemoglobin thresholds was the least likely to be discussed. To determine the amount of blood transfused in each trial arm some RCTs applied posttransfusion hemoglobin targets, others indicated the number of units to transfuse, and many did not provide enough information to determine the criteria applied. This variation likely explains why only 3 of 19 systematic reviews discussed posttransfusion hemoglobin targets,1 and why large variation exists in the actual number of units transfused between trials and within trials.

CONCLUSIONS

Readers of RCTs are encouraged to evaluate the research based on the quality of the study methods. Some issues are common to all RCTs, such as concealed allocation, blinding, and random allocation. In addition to these, we encourage readers of clinical trials comparing transfusion strategies to ask the following questions unique to these trials: What was the actual number of units of blood transfused in both groups? In addition to the planned hemoglobin thresholds for transfusion, what were the actual hemoglobin concentrations in both groups? Were patients excluded if they received transfusions before randomization? What was the planned posttransfusion hemoglobin target?

Some of these issues could be addressed by designing trials that achieve greater separation in pretransfusion hemoglobin thresholds and units of blood transfused between groups, and also by excluding patients transfused before randomization.

However, important questions would remain unanswered. For example, if the entry criteria for these trials is anemia, why would the cause and underlying condition leading to the anemia not be investigated or addressed?15 In addition, hemoglobin concentration does not always correlate with total red cell mass. Many factors can result in the hemoglobin concentration going up or down without any change in total red cell mass.16 Factors such as acute hypoxia, sudden cessation of tobacco use, vasodilators, intravenous fluids, stressors, disease state, and neuroendocrine responses can result in plasma volume contraction or expansion, with some of these responses potentially being protective.

We believe future RCTs in this field would benefit from incorporating pre- and postrandomization assessment and management of anemia and considering iron deficiency.17 This may be accomplished through cluster RCT designs, which are better suited to study multiple interventions and processes in an approach to patient care (eg, patient blood management) as opposed to evaluate a specific intervention (eg, transfusion threshold strategies).

Furthermore, it is important to acknowledge that observational studies are particularly valuable when a research question cannot be answered by traditional RCTs. In some clinical settings, conducting an RCT is impossible (smoking versus no smoking), unethical, or too expensive, and therefore evidence investigating the harms of interventions may need to come from large observational data sets. Demonstrating the important role observational studies play, Sir Austin Bradford Hill, the father of the medical RCT, proposed criteria for assessing causation from observational studies.18 Although well-designed cohort studies continue to rank below RCTs in the hierarchy of evidence, they are vital to answer questions like: What effect does transfusion have on outcomes? To date, a number of large prospective observational studies designed to answer this question have been published.19 In addition, data from a number of observational studies have evaluated multimodal PBM programs and provide evidence from real-world settings about the effects of implementing comprehensive bundles of patient-specific interventions.20

DISCLOSURES

Name: Kevin M. Trentino, MPH.

Contribution: This author designed, developed, and refined the manuscript with contributions from all other authors, and drafted the manuscript. They were involved in critically revising the draft, and approving the final version to be published.

Conflicts of Interest: None.

Name: Shannon L. Farmer, DHSc.

Contribution: This author contributed to the design and development of the manuscript, provided substantial contributions to the analysis and interpretation of data for the work, was involved in critically revising the draft, and approved the final version to be published.

Conflicts of Interest: S. L. Farmer reports other from National Blood Authority (Australia), personal fees from ETHICON Biosurgery, other from Thieme (Stuttgart), outside the submitted work.

Name: James P. Isbister, MB BS.

Conflicts of Interest: J. P. Isbister reports personal fees and nonfinancial support from Vifor Pharma, personal fees and nonfinancial support from National Blood Authority, personal fees and nonfinancial support from CSL Behring, outside the submitted study.

Name: Frank M. Sanfilippo, PhD.

Conflicts of Interest: None.

Name: Michael F. Leahy, MB ChB.

Conflicts of Interest: M. F. Leahy reports personal fees from Vifor Pharma, personal fees from Pfizer, grants from Amgen, outside the submitted work.

Name: Axel Hofmann, Dr rer medic.

Conflicts of Interest: A. Hofmann reports personal fees and nonfinancial support from Celgene, Belgium, personal fees, and nonfinancial support from G1 Therapeutics, nonfinancial support from South African National Blood Service, South Africa, personal fees and nonfinancial support from Takeda, South Africa, personal fees and nonfinancial support from TEM Innovations, Germany, personal fees and nonfinancial support from Vifor Pharma International AG, Switzerland, outside the submitted study.

Name: Aryeh Shander, MD.

Conflicts of Interest: None.

Name: Kevin Murray, PhD.

Conflicts of Interest: None.

This manuscript was handled by: Susan Goobie, MD, FRCPC.

GLOSSARY

PBM

patient blood management

RBC

red blood cells

RCT

randomized controlled trials

TITRe2

Transfusion Indication Threshold Reduction

REFERENCES

1. Trentino KM, Farmer SL, Leahy MF, et al. Systematic reviews and meta-analyses comparing mortality in restrictive and liberal haemoglobin thresholds for red cell transfusion: an overview of systematic reviews. BMC Med. 2020;18:154.

2. Trentino K, Farmer S, Gross I, Shander A, Isbister J. Observational studies - should we simply ignore them in assessing transfusion outcomes? BMC Anesthesiol. 2016;16:96.

3. Mazer CD, Whitlock RP, Fergusson DA, et al.; TRICS Investigators and Perioperative Anesthesia Clinical Trials Group. Restrictive or liberal red-cell transfusion for cardiac surgery. N Engl J Med. 2017;377:2133–2144.

4. Grover M, Talwalkar S, Casbard A, et al. Silent myocardial ischaemia and haemoglobin concentration: a randomized controlled trial of transfusion strategy in lower limb arthroplasty. Vox Sang. 2006;90:105–112.

5. So-Osman C, Nelissen R, Te Slaa R, Coene L, Brand R, Brand A. A randomized comparison of transfusion triggers in elective orthopaedic surgery using leucocyte-depleted red blood cells. Vox Sang. 2010;98:56–64.

6. Leal-Noval SR, Rincón-Ferrari MD, Múñoz-Gómez M. Red blood cell transfusion may be more detrimental than anemia for the clinical outcome of patients with severe traumatic brain injury. Crit Care. 2019;23:189.

7. Almeida J, Galas F, Osawa E, et al. Transfusion Requirements in Surgical Oncology Patients (TRISOP): a randomized, controlled trial. Crit Care. 2013;17Suppl 2P364.

8. Carson JL, Terrin ML, Noveck H, et al.; FOCUS Investigators. Liberal or restrictive transfusion in high-risk patients after hip surgery. N Engl J Med. 2011;365:2453–2462.

9. Carson JL, Stanworth SJ, Roubinian N, et al. Transfusion thresholds and other strategies for guiding allogeneic red blood cell transfusion. Cochrane Database Syst Rev. 2016;10:CD002042.

10. Gobatto ALN, Link MA, Solla DJ, et al. Transfusion requirements after head trauma: a randomized feasibility controlled trial. Crit Care. 2019;23:89.

11. Murphy GJ, Pike K, Rogers CA, et al.; TITRe2 Investigators. Liberal or restrictive transfusion after cardiac surgery. N Engl J Med. 2015;372:997–1008.

12. Stokes EA, Wordsworth S, Bargo D, et al.; TITRe2 Investigators. Are lower levels of red blood cell transfusion more cost-effective than liberal levels after cardiac surgery? Findings from the TITRe2 randomised controlled trial. BMJ Open. 2016;6:e011311.

13. Litton E, Baker S, Erber W, et al.; Australian and New Zealand Intensive Care Society Clinical Trials Group. The IRONMAN trial: a protocol for a multicentre randomised placebo-controlled trial of intravenous iron in intensive care unit patients with anaemia. Crit Care Resusc. 2014;16:285–290.

14. Scawn N, Saul D, Pathak D, et al. A pilot randomised controlled trial in intensive care patients comparing 7 days’ treatment with empirical antibiotics with 2 days’ treatment for hospital-acquired infection of unknown origin. Health Technol Assess. 2012;16:i–xiii, 1.

15. Faraoni D, Schaefer ST. Randomized controlled trials vs. observational studies: why not just live together? BMC Anesthesiol. 2016;16:102.

16. Isbister JP. Physiology and pathophysiology of blood volume regulation. Transfus Sci. 1997;18:409–423.

17. Spahn DR, Schoenrath F, Spahn GH, et al. Effect of ultra-short-term treatment of patients with iron deficiency or anaemia undergoing cardiac surgery: a prospective randomised trial. Lancet. 2019;393:2201–2212.

18. Hill AB. The environment and disease: association or causation? Proc R Soc Med. 1965;58:295–300.

19. Hopewell S, Omar O, Hyde C, Yu LM, Doree C, Murphy MF. A systematic review of the effect of red blood cell transfusion on mortality: evidence from large-scale observational studies published between 2006 and 2010. BMJ Open. 2013;3:e002154.

20. Faraoni D, Gross I, Shander A. Network meta-analysis of isolated patient blood management interventions leads to uncertain interpretation. Comment on Br J Anaesth 2020. Br J Anaesth. 21 August 2020 [Epub ahead of print].