Pediatric Thromboelastograph 6s and Laboratory Coagulation Reference Values

We conducted a prospective observational study in a single, tertiary institution with Human Research Ethics Committee approval and written informed consent obtained from parents/guardians. Healthy children undergoing noncardiopulmonary bypass surgery under general anesthesia who required intravenous access between January 3, 2017 and January 3, 2019 were eligible for inclusion. Children were excluded if they had American Society of Anesthesiologists physical status classification of 3 or greater, were clinically unwell (including inpatients), had a personal or family history of hemorrhagic or thrombotic conditions, or were receiving any anticoagulant or antiplatelet therapy.²⁸  Preterm babies born less than 37 weeks of gestation were excluded. Demographic data, personal and family past medical history, and surgical procedure information were collected. Age groups were defined to obtain discrete age-specific reference ranges (neonates < 1 month, infants 1 month–1 year, young children 1–5 years, older children 6–10 years, and adolescents 11–16 years) in agreement with previous studies.^10,14,26 

Venous blood was sampled during insertion of an appropriate gauge intravenous catheter and placed immediately into collection tubes as follows: a 3.2% sodium citrate tube (3.5-mL Vacuette; Greiner Bio-one, Kremsmunster, Austria); a 3.2% sodium citrate tube (1-mL MiniCollect; Greiner Bio-one, Kremsmunster Austria); and an EDTA tube (0.5-mL Becton Dickinson Microtainer Microtube for Automated Process [MAP microtube]; Becton Dickinson, Franklin Lakes, New Jersey). Samples were transported by a pneumatic tube system to the on-site National Association of Testing Authority Australia accredited laboratory. The pneumatic tube system was internally validated and specimen handling was in accordance with local protocols with samples checked for clots and to ensure accurate collection volumes. Inadequate samples, including those with clots or those overfilled or underfilled, were discarded and excluded from analysis.

An automated platelet count was performed from the ETDA tube on the Sysmex XN-Series Automated Hematology Analyzer (Roche, Kobe, Japan), using automated impedance techniques (and automated fluorescent techniques if required). The 3.5-mL sodium citrate sample was double-centrifuged and frozen at −80°C within 4 hours of collection, and batch tested for laboratory parameters with samples defrosted immediately before analysis. Coagulation assays were performed on each patient with HemosIL (Instrumentation Laboratory, Bedford, Massachusetts), Siemens (Siemens Healthineers, Erlangen, Germany), and Stago (Diagnostica Stago, Asnières-sur-Seine, France) reagents on ACL TOP 300, 500, and 700 analyzers (Instrumentation Laboratory), which use identical methodology (Instrumentation Laboratory) (Table 1 and Supplemental Table 1 [see supplemental digital content containing data, 5 tables, and 1 figure at https://meridian.allenpress.com/aplm in the November 2021 table of contents]). The 1-mL sodium citrate tube was used for viscoelastic hemostatic assays testing with 0.3 mL added to the TEG 6s Hemostasis Analyzer. POC testing occurred in a temperature-controlled environment within 2 hours of collection.

Planned number of participants enrolled was in agreement with previous studies outlining age-specific hemostatic reference ranges and in compliance with the Clinical and Laboratory Standards Institute C28A3 guideline.⁵  To investigate whether mean reference values varied by assay methodology, mixed-effects regression analyses were performed using age, reagent, and age-reagent interaction as fixed effects, and participant as a random effect. Mean reference values per reagent were estimated using the model at each age integer. Maximum coefficients of variation between these estimates were compared with coagulation assay-specific allowable variation according to quality requirements.^3,29  Variation by reagent was considered clinically insignificant if the maximum coefficients of variation of the predicted mean values across age values was below “allowable total error.”^3,29  Variation between different reagent types within the same patient was also explored using Bland-Altman plots, graphically separated by age groups. Two methods of defining RIs were explored. First, discrete RIs were calculated as the boundaries, including 95% of the population in each predefined age group (ie, the 2.5th and 97.5th percentiles; median values are also reported). Parameters between male and female participants were explored using Mann-Whitney U test. The Kruskal-Wallis test was used to compare the parameter between age groups. Second, a model-based method generated equations for the 2.5th and 97.5th RIs using quantile regression.^3,18  Fractional polynomial regression was first used to model the mean. Predicted values of the mean were compared with observed values; if there was a significant difference, neonates and/or infants were not included in generating the RIs. Interactions between sex and polynomial terms were explored, and the resulting quantile regression included the terms significant in fractional polynomial regression as well as relevant interaction terms. Generated RIs, along with 95% CIs, were plotted against age, separated by sex if required. Intraclass correlation coefficients (2-way, mixed-effects model) were calculated to compare the 2 methods of RI generation (upper and lower limits). Discrete and model based RIs are also graphically represented.³⁰  All analyses were conducted using StataSE version 16.0 (StataCorp Pty Ltd, College Station, Texas). Analyses were considered significant at the less than .05 level.

There were 254 samples available for laboratory parameters and/or TEG 6s analysis with complete samples for both analyses in 238 patients (93.7%). Patient demographics, number approached, included, and reasons for exclusion are shown (Figure 2 and Supplemental Digital Content).

Results from different testing methodologies were compared for activated partial thromboplastin time (aPTT) using 4 reagents, including HemosIL aPTT-SP (Instrumentation Laboratory), Stago TriniCLOT HS (Diagnostica Stago), HemosIL SynthASil (Instrumentation Laboratory), and Stago TriniCLOT HS (Diagnostica Stago) with Heparin-resistant CaCl2; and prothrombin time (PT) using 3 reagents, including Siemens Thromborel S (Siemens Healthineers), HemosIL RecombiPlasTin 2G (Instrumentation Laboratory), and HemosIL ReadiPlasTin (Instrumentation Laboratory). Mixed-effect regression models suggested significant variation in mean reference values for the different reagents used generating aPTT and PT results. Maximum coefficients of variation for mean reference values across age were 5.1% for aPTT and 21.1% for PT, both clinically significant compared with allowable total error (4.5% and 5.3%, respectively, Supplemental Table 2).^3,29  Bland-Altman plots graphically displaying agreement between assays, divided into age groups are included (Supplemental Figure 1). Although poorer at younger ages, aPTT results appeared largely in agreement. Substantial agreement was observed between fibrinogen concentration methods. A systematic difference was detected in PT tests; for the same patient, results using ReadiPlasTin were on average 4 points greater than Thromborel readings.

Discrete RIs for TEG 6s variables across Kaolin, Kaolin-Heparinase, rapid, and functional fibrinogen assays are presented (Table 2). For all TEG 6s POC tests there were no significant differences between neonates and infants, so these age-groups are combined for generation of discrete RIs (separated RIs in Supplemental Table 3). Reaction time, kinetics time, alpha angle, and maximum amplitude RIs were consistently different between age groups across all TEG 6s assays, whereas lysis 30 values and activated clotting time were similar (Table 2). Discrete age-group RIs for laboratory parameters and platelet counts are presented with significant variation observed between age-groups (Table 3). Subgroup analyses revealed no statistical differences between POC or laboratory parameters by sex.

Equations for generating the 2.5th and 97.5th percentile RIs reported as functions of age and/or sex are shown (Table 4 and the Supplemental Table 4). RI estimates and 95% CIs plotted against age are graphically displayed (Figure 3, A through J). For all parameters a substantial relationship with age was observed. RIs decreased substantially with age for reaction time, aPTT, and thrombin clotting time but increased with age for fibrinogen and kinetics time. Changes were particularly evident at younger ages with substantial reduction in RIs during earlier years for alpha angle, maximum amplitude, aPTT, and platelet count, whereas increases were observed for antithrombin and fibrinogen. For kinetics time, platelet count, PT, TCT, and aPTT the upper reference limits changed with age, but minimal changes were observed in lower reference limits. RIs did not differ by sex for kinetics time or PT; however, for reaction time, alpha angle, aPTT, TCT, and fibrinogen sex-based separation was consistently observed across the age spectrum.

We compared discrete and model-based methods graphically to explore the impact of using standard age-group defined RIs at the extreme limits for age cutoffs (Figure 4, A through J). For TEG 6s variables we included the manufacturer RIs generated using adult data.^23,31  Intraclass correlation coefficients further exploring agreement between the discrete and model-based methods of upper and lower RI generation are included (Supplemental Table 5). A significant correlation was demonstrated with moderate or good agreement observed across upper and lower limits for kinetics time, alpha angle, TCT, and antithrombin (coefficients ≥ 0.54), lower limits of reaction time, PT, and fibrinogen (coefficients ≥ 0.61) and upper limits of MA, aPTT, and platelet count (coefficients ≥ 0.77). Conversely, poorer agreement was observed for upper limits of reaction time, PT (nonsignificant), and fibrinogen, and lower limits of maximum amplitude, aPTT, and platelet counts.

In this prospective, single-center study in a population of healthy Australian children, we established RIs for the POC TEG 6s and laboratory coagulation parameters in the described technical conditions. To our knowledge, this is the first study to provide pediatric RIs for the novel TEG 6s Hemostasis Analyzer and to estimate coagulation RIs as functions of age and/or sex. Intraclass correlation coefficients confirmed that the discrete and model-based methods for generating RIs varied between parameters with variable agreement between upper and lower limits. We also observed significant differences for both laboratory and TEG 6s parameters according to the child's age, most evident at younger ages, affirming the concept of developmental hemostasis. Clinically significant variation was also observed between PT and aPTT results using different reagents. The disparity relating to demographics and reagents reinforces recommendations for pediatric-specific and reagent analyzer–specific RIs. We provide evidence of feasibility for novel applications of a sophisticated statistical approach that may overcome disadvantages inherent with age-limit cutoffs when using standard discrete RIs.

Novel viscoelastic hemostatic assays for POC testing are increasingly used in children to guide clinical decision-making, and should be subject to the same quality assurance and validation processes as existing laboratory tests to ensure accurate interpretation.^25,32,33  Our findings start to address the issue of quality, in providing pediatric values for TEG 6s substantially different from manufacturer recommended values generated from adults.²⁷  Validation steps are still required for diagnosis of pathological states or monitoring of anticoagulation, although this has been attempted in adults.³⁴  Of note, age-related differences in RIs were not identified in the earlier kaolin-activated TEG 5000 system.²⁶  While both assays measure visco-elastic properties, the platforms have several important differences underscoring why RIs need to be generated for the new version.^{23,26,31,35–37}  First, TEG 6s is an automated cartridge-based test, which runs multiple tests simultaneously through microfluidics channels. This was ostensibly designed to simplify usage and reduce operator-induced variability. Conversely, TEG 5000 requires manual pipetting, mixing, and preparation of samples, introducing potential variability.³⁷  Similarly, standard use of citrated tubes improves testing precision; however, a 15-minute incubation period is recommended, which affect parameters variably.³⁵  Second, physical methods of measurement differ between analyzers. TEG 5000 uses a suspended torsion wire in an oscillating cup to measure shear elasticity via electro-mechanical transduction. TEG 6s exposes the sample to a fixed vibration and measures vertical motion of the blood meniscus via an light-emitting diode light source and an infrared detector. It is not clear whether this influences results; however, a recent adult study demonstrated agreement between instruments.³⁵ 

We report RIs using 2 approaches: (1) the standard method applying discrete age groups defined in literature, and (2) as functions of age and/or sex using sophisticated statistical modeling, envisaging 2 key benefits from the model-based method. First, it is increasingly recognized that patients at age limit margins or the extreme cutoffs of age-group categories may not be well represented, highlighting an inherent limitation with interpretation of discrete RIs.^3,20  Our graphic comparison for RI generation methods depicts this drawback of discrete RIs; a patient aged 12 months and 1 day is grouped with a patient 4 years and 11 months. As a result, there is a paradigm shift endorsing implementation of alternate strategies, such as continuous RIs across ages.^20,38  Second, integration into electronic health systems is an obstacle necessitating innovative solutions to facilitate widespread implementation. This is a barrier to continuous RIs. Model-based methods, such as fractional polynomial and quantile regression, however, represent an opportunity to overcome these challenges.^3,18–20 

We observed substantial differences in RIs according to age, emphasizing the importance of pediatric RIs. Although there is no evidence of an increased tendency toward bleeding or thrombosis, age-related differences in coagulation parameters have been consistently reported in healthy children, particularly in the first year of life. A delicate hemostatic equilibrium evolves with age, enabling blood to circulate in intact vessels, and avoid excessive bleeding after endothelial damage via local clot formation. Use of values derived from adults or uniform results may potentially trigger unnecessary investigation or treatment in healthy children, risking complications, child and family distress, and excessive healthcare costs. An infant with an aPTT outside “normal” adult ranges may be investigated for a bleeding disorder and administered unwarranted blood products. Similarly, antithrombin replacement is common in infants treated with heparin, a population where measured activity is considerably lower, but may not reflect physiological derangement, or anticoagulant effect, risking potential harm.³⁹ 

Clinically significant variability was seen using different reagents on the same analyzer. The physiological trends in coagulation parameters we observed correlate with reported age-related variation demonstrated using similar and different technology for laboratory coagulation tests and discrete age-group RIs. Using the STA-R coagulation analyzer (China), HemosIL Assays on ACL TOP Hemostasis Testing Systems (France), and the Sysmex CA-1500 System in parallel with the Behring BCS System using reagents from Siemens (Dutch) lower fibrinogen levels and aPTT as well as TCT prolongation were observed, attributable to decreased concentrations of clotting factors in younger children.^8,10,17  Variable PT RIs are reported in the literature; our findings and results from the Chinese group observing significant differences in PT levels by age contrast with the French study. Potential explanations for observed disparity in PT results between reagents include different sensitivities to factor levels or alternate endpoints (eg, maximum acceleration versus maximum top speed). This is more pronounced with kinetics seen in younger children, owing to their developing hemostatic system, particularly for vitamin k–dependent factors and potentially reflective of liver immaturity. This inconsistency was the driver for standardization and calibration of warfarin testing through the reagent International Sensitivity Index and subsequent adoption of International Normalized Ratio.^40–42  Antithrombin activity measured using an Xa assay is lower in younger children.^8,10  However, assays measuring antigen activity have shown similar levels to adults, albeit of differing isoforms.⁴³  It is proposed that neonatal isoforms may have a greater anti-IIa activity, which may not be reflected using current methodology. This would have flow-on effects for measuring heparin effect and may explain discrepancies in monitoring tests such aPTT and anti-Xa. Differences in results between the functional fibrinogen assay in TEG 6s and measured fibrinogen concentration (Clauss) in may reflect different testing methodologies with controversy in the literature regarding correlation.^44–47  Our findings of sex-based variation need further exploration to examine if the underlying pathophysiology relates to differences in factor concentrations. Given variation according to demographics and when using different reagents/methodologies observed in our study, RIs should ideally be established for each laboratory where possible, under the same technical conditions until further harmonization studies are conducted.

Our results must be interpreted in the setting of several limitations. Clinical and Laboratory Standards Institute (C28A3) recommended sample size for establishing RIs for discrete age-groups vary from 30 to 120 depending on the setting.^2,3,5  Suitable sample size number proposed for quantile regression is 292, and our data need replication, correlation, and validation in other populations.⁴⁸  Practical and ethical challenges exist in obtaining large blood samples from healthy children, with pediatric RIs frequently based on retrospective data-mining or extrapolated from adults. As we observed, other prospective studies have experienced challenges meeting sample size goals particularly in the neonatal population.^3,10,27  Of note, 16% of samples were insufficient for testing, which poses a potential risk for POC tests where clinical staff may not be trained in checking for sample integrity. While the study protocol of a single collection attempt may not reflect normal clinical practice, sample issues and other preanalytical variables need to be considered as sources of error in both POC and laboratory coagulation testing.⁴⁹  Neonatal vitamin K administration data were not collected; however, this is standard practice in Australia. Lupus anticoagulant testing was not undertaken; however, in otherwise healthy children undergoing elective surgery this is unlikely to have played a role in our findings. Data on ethnicity were not collected, so potential differences, while not expected, were not able to be assessed.

Our findings highlight that variation between RI results may be due to physiological differences between patients, or variation within the test itself. As a result of emerging novel testing techniques, RIs need to be continuously updated and methodologic approaches to RI estimation must also evolve to overcome challenges and optimize analyses. There are several important clinical and research implications for obtaining accurate RIs. First, accurate pediatric RIs allow for appropriate interpretation of results in the context of age and the concept of developmental hemostasis. Second, modern statistical methods are a potential means to revolutionize RI estimation. They offer an opportunity to improve accuracy of laboratory test interpretation, particularly at extremes of age-group limits, and incorporate RIs into clinical practice by facilitating adoption into electronic health systems.^3,19  Third, standardization of laboratory testing methodology is vital for valid interinstitution comparisons. Published coagulation test results cannot be extrapolated unless the setup is identical. An aPTT range for heparin therapy expressed in seconds may not be generalizable to another institution using different analyzers or reagents. This underscores the importance of exhaustive reporting of coagulation testing methodologies. RI harmonization is an emerging concept necessitating further research in pediatric hematology.^3,16,50  Finally, pathological states cannot be confidently diagnosed, and treatment effect evaluated without understanding normal limits. Once a healthy reference population has been established, the test may be used to investigate a population with disease and assess interventions and is validated for these purposes. Applications include diagnosis of congenital or acquired bleeding disorders, monitoring anticoagulant therapies, and transfusion thresholds.

This study provides pediatric, reagent-analyzer RIs from a healthy population for TEG 6s and laboratory coagulation tests. These data serve as a foundation for comparison and validation of thromboelastography in pathological states and demonstrate feasibility in model-based methodologic approaches. The findings facilitate accurate interpretation of clinical results to optimize prevention, diagnosis, and treatment of thrombotic and hemorrhagic diseases.

We acknowledge the patients and families who participated in the study, the staff of Queensland Children's Hospital, and the generous funding from the Intensive Care Foundation and Getinghe group without whom this research could not have been completed. Joanne Beggs additionally assisted with laboratory testing. Monsurul Hoq and the team from the Murdoch Children's Research Institute provided invaluable advice on the statistical methodology and analyses.