The incidence of major bleeding in patients undergoing cardiac surgery requiring cardiopulmonary bypass (CPB) is between 3 and 11%.1,2,3 The transfusion rates in cardiac surgery are variable, with up to 95% of patients receiving blood products in some centres.4 In observational studies, transfusion of blood products has been associated with increased morbidity and mortality,5 as well as healthcare costs.6

Options for monitoring hemostasis in patients undergoing cardiac surgery include standard laboratory tests of hematocrit, hemoglobin, platelet count, activated partial thromboplastin time, international normalized ratio, and fibrinogen levels. While these tests are readily available, there may be substantial delays in obtaining the results, which decreases their utility in the setting of urgent and immediate management of coagulopathy.7

Point-of-care (POC) viscoelastic tests, such as thromboelastography (TEG) and rotational thromboelastometry (ROTEM) provide timely, comprehensive information regarding the coagulation status of a patient (usually within 15 min), making them attractive for monitoring hemostasis in acutely bleeding trauma and surgical patients. Unlike standard laboratory tests, TEG/ROTEM utilizes whole blood samples to assess detailed information on clot formation, strength, and lysis allowing rapid identification of the specific coagulation defect7 (Appendix 1). Point-of-care viscoelastic tests can potentially be used to more effectively tailor blood product administration, avoid transfusion of unnecessary products, and reduce morbidity and mortality.8,9

Until recently, evidence to support the routine use of TEG and ROTEM has been limited to non-randomized studies10,11,12 or small single-centre randomized controlled trials (RCTs).8,13,14 In 2016, the first large multi-centre RCT evaluating POC viscoelastic testing in cardiac surgery was published. Data from this trial pertaining to the proportion of patients transfused with blood products and mortality have yet to be considered in the context of a meta-analysis of the existing literature.9 To provide additional clarity on the utility of viscoelastic testing in adult cardiac surgery patients, we conducted a systematic review and meta-analysis that focuses on patient-centred outcomes.

Methods

To conduct this systematic review, we used the Methodological Expectations of Cochrane Intervention Reviews guidelines.15 The protocol was developed a priori and registered on PROSPERO (CRD42017065777). We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines to report our results.16 A completed PRISMA checklist is provided in the supplementary materials.

Research question

Our research question was “In adult cardiac surgery patients, compared with standard laboratory testing and/or physician discretion, does the use of POC viscoelastic hemostatic testing impact blood product transfusion, mortality, intensive care unit (ICU), or hospital length of stay?” We included RCTs involving patients > 18 yr of age, where at least 80% of the patients underwent cardiac surgery utilizing CPB. Either TEG or ROTEM had to be used as the intervention in the trial.

Our primary outcomes were the proportion of patients receiving any blood product and all-cause mortality at the longest follow-up interval. Our secondary outcomes included the proportion of patients transfused with specific blood components (red blood cells [RBC], platelets, frozen plasma [FP], and cryoprecipitate), the volume of blood products transfused, the number of patients requiring reoperation for bleeding, and ICU and hospital length of stay. We included the incidence of transfusion-associated infection, allergic reactions, and anaphylaxis as transfusion-related safety outcomes.

Search strategy and study selection

We searched Medline (Ovid), EMBASE (Ovid), CENTRAL (the Cochrane Library-Wiley), Web of Science, Biosis, Scopus, and CINAHL databases from inception to February 14, 2018. We used the Cochrane Highly Sensitive Search Strategy to create individual search strategies for each database. The strategy from the Medline search is presented in Appendix 2. We conducted a query of the World Health Organization’s International Clinical trials registry, clinicaltrials.gov, and ISRCTN to identify ongoing or planned clinical trials.

In addition to electronic database searching, to identify eligible trials, we searched abstracts and conference proceedings of the following societies from 2014 to 2017: American Society of Anesthesiology, Canadian Anesthesiologist’s Society, Society of Thoracic Surgeons, American Association for Thoracic Surgery, European Association for Cardio-Thoracic Surgery, Society of Critical Care Medicine, and Canadian Cardiovascular Society. Hand searching of reference lists from relevant citations and previously published systematic reviews was also conducted. References were managed using Endnote™ (ver. X7 Thomson Reuters, Carlsbad, CA, USA).

Two reviewers (C.L. and J.H.) independently screened titles and abstracts to determine if the study met the inclusion criteria. Each report was classified as: “include”, “exclude”, “unclear”, or “duplicate of another citation”. All full-text reports classified as “include” or “unclear” by either reviewer were retrieved for formal review. Any full-text report that was not available from library services was excluded. Next, the reviewers independently assessed each full-text report using a pilot-tested standardized form. Disagreements were resolved by discussion between the two reviewers; third-party adjudication was not necessary.

Data abstraction and management

Data were independently extracted by two reviewers (C.L. and J.H.) from all included trials using a standardized pilot-tested form. All disagreements were resolved through consensus; third-party assistance was not required. The following data were extracted: author identification, year and language of publication, source of funding, study design, population (including study inclusion and exclusion criteria), patient characteristics (age, sex, body mass index), and procedural characteristics (type and urgency of surgical procedure, CPB duration, preoperative anticoagulant use, and anti-fibrinolytic use), intervention (TEG or ROTEM, plus any other POC coagulation tests), details pertaining to comparator interventions, as well as primary, secondary, and safety outcomes.

Risk of bias assessment

To evaluate the internal validity of included trials, we used the Cochrane Collaboration Risk of Bias tool.17,18 The overall risk of bias assessment for each trial was based on the adjudication of six individual domains (sequence generation, allocation concealment, blinding, incomplete outcome data, selective outcome reporting, and “other” sources of bias). Each domain was rated “low risk”, “unclear risk”, or “high risk”. If one or more individual domains were assessed as being “high risk”, our overall assessment of the trial’s risk of bias was rated as such. For a trial to be considered “low risk”, all individual domains must have received a “low risk” rating. The risk of bias for all other studies was adjudicated as “unclear”. Publication bias assessment19,20 using funnel plot techniques was not possible given the small number of included trials.

Measures of treatment effect

We analyzed data from the included studies using RevMan (version 5.3.5, The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark). Data from a single cluster RCT were adjusted for cluster and time and presented as relative risk with the 95% confidence interval (CI). Continuous data for the parallel trials were pooled using the generic inverse variance method and expressed as a mean difference (MD) with the 95% CI. For continuous outcomes including cluster randomized data, pooling was done using the generic inverse variance method with effect measures expressed as a standard MD with 95% CI. Dichotomous outcomes, including cluster randomized data, were pooled using the generic inverse variance method and expressed as a log (risk ratio [RR]) with standard error. All data were analyzed using the random effects method. Statistical heterogeneity was explored using the I2 test.18 If significant heterogeneity was encountered with an I2 value greater than 50%, further subgroup analyses were conducted.

Subgroup and sensitivity analysis

A priori subgroups related to our primary outcomes included: risk of bias, source of funding, surgical urgency, and procedural complexity.

Results

From 1,917 identified records, we included 11 trials that enrolled 8,294 patients (Fig. 1).2,8,9,13,14,21,22,23,24,25,26 Between 1999 and 2016, ten trials were published in English and one in Turkish. Ten of the trials were single-centre trials,2,8,13,14,21,22,23,24,25,26 and one was multi-centre involving 12 hospitals.9 Four trials were conducted in North America,2,9,24,25 six in Europe,8,14,21,22,23,26 and one in Australia.13 One trial published preliminary data as an electronic abstract22 with the final data included in a recently published systematic review and meta-analysis.27 Three trials were adjudicated to be at high risk of bias8,9,25 because of lack of participant and personnel blinding; the remaining eight trials were classified as “unclear” primarily because of poor reporting (Fig. 2).

Fig. 1
figure 1

Literature review process. Flow diagram illustrating the citation and manuscript review process utilizing the Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines

Full size image
Fig. 2
figure 2

Risk of bias summary. Each trial was assessed on six different domains with the decision illustrated by circles representing “low”, “high”, and “unclear” risks of bias. The final column displays the overall judgement of the trial

Full size image

The baseline characteristics of the included trials are presented in Table 1. The mean age range of enrolled patients was 51-72 yr old. Five trials provided adequate operative details to determine the exact number of patients who had undergone “simple” vs “complex” procedures.8,9,14,21,25 Two of these trials enrolled only patients undergoing elective cardiopulmonary bypass,21,25 which was classified as a “simple” procedure. All patients in the third trial underwent complex aortic surgery,14 and the patients in the fourth and fifth trials underwent a variety of “simple” and “complex” procedures.8,9 Seven trials used TEG as the primary intervention,2,13,21,23,24,25,26 and four trials used ROTEM.8,9,14,22 The comparator group involved a transfusion algorithm based on standard laboratory tests in four trials,8,22,24,25,26 and a combination of standard laboratory tests and clinician discretion in five trials.2,13,14,21,23 The intervention (TEG or ROTEM) was implemented during re-warming on CPB and/or post protamine in all trials; three trials reported pre-CPB baseline values,23,24,26 and eight trials also utilized POC tests under various conditions in the ICU.2,8,13,14,21,22,25,26 Nine trials reported administration of blood products in both the intraoperative and postoperative period,2,8,9,13,14,21,23,24,25 while two trials examined only postoperative product use.22,26 The only multi-centre study was a pragmatic stepped-wedge cluster RCT where each centre was instructed to continue their usual institutional practice prior to implementation of the transfusion algorithm.9 Point-of-care platelet function tests were also used in addition to TEG/ROTEM in five studies.8,9,13,21,25 Postoperative follow-up duration varied among all included studies, ranging from 24 hr to six months.

Table 1 Baseline characteristics of included studies
Full size table

Primary outcomes

The use of POC viscoelastic testing was not associated with a difference in the proportion of patients transfused with any blood product (RR, 0.90; 95% CI, 0.79 to 1.02; I2 = 51%; four trials, 7,623 patients)9,14,23,24 (Fig. 3) or mortality at the longest follow up (RR, 0.73; 95% CI, 0.47 to 1.13; I2 = 5%; six trials, 7,931 patients)8,9,14,21,22,24,27 (Fig. 4). We found no difference in proportion of patients transfused with any blood product or all-cause mortality across all subgroups examined, including risk of bias, funding source, procedure urgency, and complexity (Appendices 4 and 5).

Fig. 3
figure 3

Forest plot of proportion of patients receiving any blood product. The point estimates and 95% confidence intervals of each trial are displayed in the forest plot with the summary estimate provided below. Chi2 = Chi-squared, df = degrees of freedom, I2 = I squared, P = P value, Tau2 = Tau-squared, Z= Z score

Full size image
Fig. 4
figure 4

Forest plot of all-cause mortality at longest follow-up interval. The point estimates and 95% confidence intervals of each trial are displayed in the forest plot with the summary estimate provided below. Chi2= Chi-squared, df = degrees of freedom, I2= I squared, P = P value, Tau2 = Tau-squared, Z = Z score

Full size image

Secondary outcomes

Transfusion-related variables

Compared with standard laboratory testing, the use of POC viscoelastic testing was weakly associated with a reduction in the proportion of patients receiving RBCs (RR, 0.91; 95% CI, 0.85 to 0.96; I2 = 0%; seven trials, 8,029 patients), platelets (RR, 0.66; 95% CI, 0.49 to 0.90; I2 = 65%; seven trials, 8,029 patients),8,9,14,21,24,25,26 and FP (RR, 0.58; 95% CI, 0.34 to 0.99; I2 = 87%; six trials, 7,989 patients) (Appendices 6-8). The proportion of patients transfused with cryoprecipitate was not reported directly in any trial except in the cluster RCT, where POC testing was weakly associated with an increase in the relative risk of receiving cryoprecipitate or fibrinogen concentrate (RR, 1.77; 95% CI, 1.01 to 1.86; 7,402 patients).9 To explore statistical heterogeneity in the proportion receiving FP, we performed a subgroup analysis based on trial design. In the five parallel-group RCTs, viscoelastic testing was associated with a lower proportion of patients transfused with FP (RR, 0.48; 95% CI, 0.35 to 0.65; I2 = 22%; five trials; 587 patients)8,14,21,24,25 while in the multi-centre cluster RCT, no difference in plasma use was observed (RR, 0.95; 95% CI, 0.89 to 1.03; 7,402 patients)9 (Appendix 7).

The volume of RBCs, FP, and platelets transfused was reported heterogeneously, with three trials presenting data as median and interquartile range,8,14,21 two as mean volume transfused (in mL),22,24,27 two as a median and range,2,26 one as the total number of units transfused,13 and the cluster data as relative risk with 95% CI accounting for cluster and time.9 Of the eight trials reporting the number of transfused units of RBCs and FP,2,8,9,14,21,22,24,26 only two showed a reduction in the number of units of RBCs administered2,8 (Table 2). Conversely, the majority (5/8) of trials showed a reduction in the number of transfused units of FP2,8,14,21,24 (Table 2). Viscoelastic testing was associated with a decrease in the units of platelets transfused (standard MD, −0.09; 95% CI, −0.18 to 0.00; four trials, 7,643 patients). Only one trial reported the units of cryoprecipitate used, with zero units in the TEG group and 20 units in the control group; no measures of significance were provided.13

Table 2 Units of packed red blood cells and frozen plasma transfused
Full size table

Clinical and safety outcomes

Point-of-care testing was not associated with a decrease in hospital or ICU length of stay8,13,14,21,22,27 (Table 3). The rate of reoperation was not different between the ROTEM/TEG and control groups (Table 3). Transfusion-related adverse events including infection, allergic reaction, and anaphylaxis were not reported in any of the trials included in this study.

Table 3 Summary of the meta-analysis of secondary outcomes
Full size table

Discussion

In this systematic review and meta-analysis, there was insufficient evidence to determine whether POC viscoelastic testing reduced the proportion of patients transfused with any blood product or all-cause mortality. When analyzed separately, POC testing was weakly associated with a reduction in the proportion of patients transfused with RBC, FP, and platelets. The volume of RBCs and FP transfused could not be pooled as the data were presented heterogeneously. Nevertheless, across individual studies, POC viscoelastic testing was associated with a trend towards decreased FP transfusion as reported in previous reviews.27,28 Pooled data concerning the volume of platelets transfused showed a reduction associated with viscoelastic testing.

The effect estimate of the proportion of patients transfused with FP was heterogeneous with an I2 value of 87%. To further explore this result, a post hoc sensitivity analysis was performed based on trial design (Appendix 7). This illustrated that the data from the cluster RCT9 showed no significant benefit associated with viscoelastic testing and was likely an outlier. In our analysis, the loss of significance when combining the individual products into one outcome is likely driven by the FP data from the cluster RCT, given its large size and weight within the analysis. We were unable to meta-analyze the data for the volume of RBCs and FP transfused. While there may be a trend of decreased FP transfusion across individual trials, it was not possible to arrive at a definitive conclusion regarding the influence of viscoelastic testing on the number of units transfused. Overall, the transfusion-related data were substantially heterogeneous, highlighting the variability in transfusion practice across centres that can be influenced by product availability, centre-specific transfusion culture, clinician preference, use of local transfusion protocols, and secular trends over time. Adherence to transfusion protocols may be limited by the multiple environmental factors that influence transfusion practice. Some of these factors (i.e., product availability) may be modifiable, while changes in culture and institutional tradition can be much harder to achieve.

There have been several recent systematic reviews concerning the use of POC viscoelastic testing in bleeding patients.27,28,29,30,31 Four of these reviews showed no significant difference in all-cause mortality in the ROTEM/TEG compared with the control groups.27,28,29,30 None of the reviews included data from the large cluster RCT for the pooled mortality estimate.9 POC testing primarily reduces the delay in identifying the specific mediators of an underlying medical coagulopathy, and is typically implemented as part of a transfusion algorithm. In patients undergoing complex surgery, transfusion is only one of many potential variables affecting mortality, and so it is acknowledged that it may be exceedingly difficult to establish a robust association.

Our systematic review and meta-analysis has several strengths. It is the only study focused on RCTs that are specifically concerned with adult cardiac surgery patients and it is the first analysis to include the proportion of patients receiving any blood product and all-cause mortality data from the recent multi-centre cluster RCT.9 Our review’s thoroughness is evidenced by our extensive database and grey literature search using librarian supported search strategies. All of the screening and data extraction were done in duplicate and we evaluated the internal validity of included trials using the Cochrane Risk of Bias tool.

Our review also has limitations. The reporting of primary and secondary outcomes in the literature we assessed was variable and often incomplete. Of the eleven included trials, four included data on the proportion of patients transfused with any blood product and six reviewed all-cause mortality. Safety data were especially lacking. For some transfusion outcomes, we detected substantial statistical heterogeneity necessitating exploratory subgroup analyses. It is possible that some of this heterogeneity may be related to variability in test interpretation and equipment calibration that were not accounted for in our analysis. None of the included trials were adjudicated to be at “low risk” of bias primarily because of lack of reporting consistent with the PRISMA checklist. We acknowledge that, while blood conservation can reduce cost, viscoelastic testing is also an expense. We did not examine whether the reduction in blood product use with viscoelastic testing was cost effective, but this is an outcome worthy of future consideration.

Conclusion

In cardiac surgery patients, POC viscoelastic hemostatic testing was not associated with a reduction in the proportion of patients receiving any blood product or all-cause mortality. However, viscoelastic testing was weakly associated with a reduction in proportion of patients transfused with specific blood products. At present, benefits conferred by viscoelastic testing are not sufficiently robust to recommend the universal implementation of this technology in adult patients undergoing cardiac surgical procedures.