Multiple electrode aggregometry and thromboelastography in thrombocytopenic patients with haematological malignancies

Abstract

Background

In thrombocytopenic patients better assessment of bleeding risk than that provided by platelet count alone is required. Multiplate^® aggregometry and thromboelastography (TEG) could be used, but information on their role in such patients is limited. The primary aim of this study was to investigate the feasibility of Multiplate^® analyses in patients with haematological malignancies. A secondary aim was to explore whether a multiple logistic regression model combining Multiplate^®, TEG, clinical and laboratory variables was associated with risk of bleeding.

Materials and methods

This was an exploratory, prospective observational study of thrombocytopenic patients with haematological malignancies. Total platelet count (TPC), white blood cell count, C-reactive protein (CRP) level, temperature and bleeding status were recorded daily. TEG and Multiplate^® analyses with four agonists were performed on weekdays.

Results

Ten patients were enrolled into the study. The median number of days in a study period was 21. Bleeding was observed on 64 of 298 study days. TPC <20×10⁹/L and <10×10⁹/L occurred on 119 and 25 days, respectively. When TPC was <33×10⁹/L, many samples showed no aggregation, regardless of bleeding status. Despite this, the odds of World Health Organization (WHO) grade 2 bleeding decreased significantly as aggregation increased and Multiplate^® had a negative predictive value (NPV) of 96% and a positive predictive value (PPV) of 19% for significant bleeding. In the multiple logistic regression model collagen-activated Multiplate^® aggregation, TEG angle, TEG reaction time and CRP significantly affected the odds of WHO grade 2 bleeding. The combined model had a NPV of 99% and a PPV of 19%.

Discussion

Our findings suggest that the markers of platelet function and haemostasis provided by Multiplate^® aggregometry and TEG may add information to support prediction of bleeding, although platelet count still remains the most accessible analysis for routine testing.

Introduction

Prophylactic platelet transfusion is a recognised component of supportive care for thrombocytopenic haematological patients during periods of bone marrow-suppressive cytostatic treatment^1,2. However, recent trials have highlighted the limited effectiveness of platelet transfusions in reducing overall rates of bleeding^3–7, particularly in some subgroups of patients such as those undergoing autologous haematopoietic stem cell transplantation. In addition, the Platelet Dose (PLADO) trial indicated a poor relationship between platelet count and bleeding risk⁸. Thus, it is important to evaluate tests or indicators that enable better assessment of bleeding risk in thrombocytopenic patients, as bleeds may be detrimental and even fatal. Indicators for predicting bleeding therefore have the potential to address patients’ safety and reduce the number of unnecessary and potentially harmful transfusions^6,9.

We investigated two available methods among the possible candidates for testing patients: Multiplate^® aggregometry and thromboelastography.

A Multiplate^® multiple electrode aggregometer is a point-of-care platelet function analyser that is commonly used to monitor antiplatelet treatment¹⁰ and exploits the principle of impedance aggregometry. Correct platelet function is essential for haemostasis and Multiplate^® aggregometry can potentially yield information on platelet function in thrombocytopenic patients. Costache and colleagues evaluated platelet function by Multiplate^® aggregometry, determining the possible role of this test in bleeding and thrombosis prediction in 42 patients with acute leucaemia¹¹. They concluded that impedance aggregometry might be a valuable tool for the prediction of both haemorrhagic and thrombotic complications. Several studies in different groups of patients, mostly those undergoing cardiac surgery, have suggested that Multiplate^® aggregometry may be useful for predicting bleeding^12–19, while others concluded that bleeding was not predicted by Multiplate^® aggregometry^20–22. In a study that evaluated the effects of platelet transfusions on Multiplate^® results the aggregation response was found to be severely reduced in thrombocytopenic patients with platelet counts below 50×10⁹/L²³. Other studies confirm these results^24–26.

Thromboelastography (TEG) is a viscoelastic haemostatic test that provides information on clot formation, stabilisation and degradation in a low shear environment. Its main use is to guide transfusion therapy in liver and heart surgery^27,28, and it has also been investigated whether the test can identify patients who are at risk of bleeding^19,29–36. In a previous study we found a significant relationship between bleeding of World Health Organization (WHO) grade 2 and the TEG parameter alpha angle in thrombocytopenic patients with haematological malignancies³². The positive predictive value (PPV) of alpha angle for predicting grade 2 bleeding was 23.1% at a cut-off of 32.15 degrees, and the negative predictive value (NPV) was 95.8%.

The primary aim of this study was to investigate the feasibility of Multiplate^® aggregometry in patients with haematological malignancies. A secondary aim was to explore whether a model which combines Multiplate^®, TEG, clinical and laboratory variables is associated with risk of bleeding. The patients were the same as in our previous study on TEG³².

Materials and methods

This exploratory prospective, observational cohort study was performed at Haukeland University Hospital, Bergen, Norway from June 2013 until February 2014, in connection with the PREPAReS study³⁷. The local ethics committee approved the study and all subjects gave their written informed consent to participation in the study.

Subjects studied

Patients were enrolled consecutively from the Section for Haematology, Department of Medicine. Patients over 18 years with haematological malignancies who were expected to need at least one platelet transfusion were approached. Exclusion criteria were congenital clotting disorders, regular use of anticoagulants in the study observation period, and immune thrombocytopenic purpura.

The study observation period started as soon as possible after hospital admission or, for patients who were not already severely thrombocytopenic at admission, when the platelet count was falling and approached 50×10⁹/L. It lasted until platelet count recovery (unsupported platelet count >50×10⁹/L), hospital discharge, or at most for 30 days of thrombocytopenia. Baseline samples were obtained on days with total platelet count (TPC) above 50×10⁹/L when possible. Patients could be re-included in subsequent treatment cycles.

Clinical data

The patients were evaluated for spleen enlargement on admission to hospital or at inclusion, mostly by abdominal examination, but some were examined by ultrasound or a computed tomography scan. Bleeding was assessed daily in the study period. Data on the administration of antibiotics to the patients were collected daily, as were the patients’ body temperature (used as a dichotomous variable with temperatures ≥38 °C defined as fever).

A detailed questionnaire, developed from a preceding pilot questionnaire, was used to collect the bleeding symptoms. The questionnaire was filled in after interviewing and examining the patient and reviewing the clinical case notes. Bleeding symptoms were classified by two people independently using the WHO criteria³⁸. Bleeding was graded 1 to 4 according to severity. There was no disagreement in the grading between the two adjudicators after joint review of the bleeding score documents.

Laboratory analyses

Blood samples were collected from a central venous access (Hickman line) or an antecubital vein. Platelet function was assessed by Multiplate^® multiple electrode aggregometry (Roche Diagnostics International Ltd., Rotkreutz, Switzerland) and blood for the analysis was drawn into 4 mL Vacuette Sodium Heparin tubes (Greiner Bio-One, Kremsmünster, Austria). Samples for TEG^39–41 were collected into 3 mL 3.2% Vacuette Sodium Citrate tubes (Greiner Bio-One) and analysed on a TEG^® Hemostasis Analyzer System (Haemonetics Corporation, Braintree, MA, USA). Blood for haematology and determination of C-reactive protein (CRP) concentration was collected into 3 mL Vacuette K₂EDTA tubes (Greiner Bio-One). A Cell-Dyn 4000 automated haematology analyser (Abbott Laboratories, Abbott Park, IL, USA) was used for daily measurements of TPC and white blood cell count (WBC), and CRP concentration was determined daily on a Roche Modular P chemistry analyser (Roche Diagnostics, Basel, Switzerland). Blood for human leucocyte antigen (HLA)-antibody analysis was collected into Vacuette K₂EDTA tubes at inclusion and then once a week until the end of the study period. HLA-antibodies were analysed using FlowPRA^™ Screening Test (One Lambda, Los Angeles, CA, USA).

Multiplate^® aggregometry and TEG were performed Monday to Friday in accordance with the manufacturer’s instructions. For the Multiplate^® studies, reference values for lithium heparin are reported, as values for sodium heparin are not available from the manufacturer. Samples anticoagulated with sodium heparin and lithium heparin give equal results for adenosine diphosphate (ADP) in the first hour after sample collection. During further storage sodium heparin conserved the results better than lithium heparin⁴². Multiplate^® area under the curve (AUC) was measured after stimulation with four commercially available agonists from Roche: ADP, collagen, ristocetin (high concentration, 0.77 mg/mL) and thrombin receptor activating peptide (TRAP).

For TEG we included the parameters reaction time, alpha angle and maximum amplitude.

Statistical analyses

Power calculations were not performed because of the lack of data on associations between Multiplate^® scores and bleeding and TEG and bleeding in patients with haematological malignancies. Descriptive statistics are reported as quartiles (Q₁, Q₂[median] and Q₃) and range.

For the analyses WHO bleeding grade 1–4 was dichotomised into significant bleeding (grade 2 or more) or not (no bleeding or grade 1) and logistic regression applied using the method of generalised estimating equations (GEE) to account for repeated measures and for four patients being included twice. For these analyses bleeding was the outcome, and the laboratory and clinical variables were predictors.

The Multiplate^® AUC values were strongly skewed towards the right for all four agonists, with many measurements under the detection limit of the test (AUC=0) and a few values at the high end of the scale, where bleeding risk is very low. These few high values would weigh a lot compared to the smaller values in analysis and could potentially mask a relationship between AUC and significant bleeding. We chose to natural log (ln) transform the AUC values to prevent this possible bias. As the log of zero is undefined, a constant, one, was added to all AUC values to allow statistical use of the results. The AUC variables are therefore ln(1+AUC), but are referred to as AUC in the text.

The AUC variables were analysed as outcome variables using mixed linear model analysis. The AUC variables were analysed without transformation, with ln transformation and with square root transformation against the predictor variables platelet transfusion and TPC to find the models that best fitted the criteria for linear mixed model analysis. From these analyses it was found that a ln transformation of the AUC variables resulted in the best adherence to the assumptions of the statistical models.

In cases in which WBC was beneath the detection limit of the analysis (<0.2×10⁹/L), the value was set at 0.1 to include it in the analysis. For the same reason TPC was set at 2×10⁹/L in the few cases in which it was below the detection limit (<5×10⁹/L).

Models for predicting bleeding were designed, and a receiver operating characteristic (ROC) curve was used to define the optimal cut-off point for the test variable. A cross tabulation of the test variable and the state variable yielded the PPV and NPV of the test variable. Where more than one predictor variable was combined in a model for predicting bleeding, logistic regression with GEE of significant bleeding as the outcome variable was first conducted. In this analysis, mean predictions were saved as a new variable, and this variable was then used as the test variable in the ROC curve analysis.

All statistical analyses were performed with IBM SPSS statistics for Windows, version 23 (IBM Corp., Armonk, NY, USA).

Model for the evaluation of bleeding risk combining thromboelastography, Multiplate^®, clinical and laboratory variables

A logistic regression model for the prediction of significant bleeding combining Multiplate^® AUC for the four agonists, the three TEG parameters, TPC, WBC, fever (yes/no), spleen enlargement (yes/no), HLA-antibodies (yes/no) and antibiotics (yes/no) was designed and analysed using GEE. This resulted in a warning that the Hessian Matrix was singular, and that some convergence criteria were not satisfied. This warning was triggered by the antibiotics variable, since all significant bleeds occurred on days with antibiotic treatment. The chi square test of independence was used to test whether it was likely that all significant bleeds would occur on days with antibiotics just by chance. The test assumes independent measurements, and p-values would be expected to be greater if repeated measures were taken into account.

Results

Ten patients (nine male) were recruited, four of whom were included during two subsequent chemotherapy cycles, giving 14 inclusions. Three patients (one male) declined participation.

The ten individuals enrolled had a median age of 45.5 years (range, 28–64). The diagnosis was acute myelogenous leukaemia (AML) in four individuals, myelodysplastic syndrome (MDS) in two, MDS-AML in two, multiple myeloma in one and histiocytic sarcoma in one patient. Of the patients included in two subsequent treatment cycles, two had AML and two had MDS. The treatment regimens for the 14 observation periods were remission induction chemotherapy in 11, consolidation chemotherapy in one, allogeneic stem cell transplantation in one and autologous stem cell transplantation in one.

The patients were followed for 298 days in total. The median observation period was 20 days (range, 7–39 days). Thrombocytopenia (TPC <50×10⁹/L) occurred on 261 study days. In six cases, the patient did not have platelet recovery before the end of the study observation period (SOP). TPC <20×10⁹/L was found on 119 study days (median duration, 8 days) and TPC <10×10⁹/L was found on 25 study days (median duration, 1 day). Overall, 189 platelet concentrates were transfused on 133 study days. A median of 11.5 (Q₁=5, Q₃=20) platelet concentrates were transfused in each study observation period.

Bleeding assessment and platelet count were available for 298 (100%) and 294 (98.7%) study days, respectively. There was 64 study days (21.5%) with bleeding, 43 with grade 1 and 21 with grade 2 bleeding. No grade 3 or 4 bleeding was observed and clinically significant bleeding is thus equivalent to WHO grade 2 bleeding. The patients had a median of 3 days with bleeding (Q₁=1, Q₃=8) in a study observation period. One patient (7.1%) did not experience any bleeding during the study observation period, in six observation periods (42.9%) there was only grade 1 bleeding, in six (42.9%) the patient experienced both grade 1 and grade 2 bleeding, and in one observation period (7.1%) the patient only had grade 2 bleeding.

The patients were febrile, defined as a body temperature of 38 °C or higher during the day, on 138 (46.3%) of 298 study days, and antibiotics were prescribed on 273 (91.6%) of the days.

In three of the 14 inclusions (21.4%), the patient had an enlarged spleen, and the patient had detectable HLA-antibodies in the study period in six of 13 inclusions. Information on HLA-antibodies was missing for one patient. One of the patients who were included twice experienced platelet refractoriness and was transfused with HLA-compatible platelet concentrates.

Multiplate^® analyses and TEG analyses were performed on 189 and 187 occasions, respectively, during the study. Descriptive statistics for the Multiplate^®, TEG and laboratory results (TPC, WBC and CRP) are summarised in Table I.

Table I.

Summary of Multiplate, TEG and laboratory results.

	N	Q1	Q2	Q3	Normal range
ADP	189	0	2	6	55–117
Collagen	189	0	2	4	61–108
Ristocetin	185	0	1	5	65–116
TRAP	189	0	3	9	92–151
TEG r	187	8.7	10.1	11.8	2–10
TEG alpha	187	38.3	47.6	55.4	55–78
TEG MA	187	46.3	51.5	56.3	51–69
TPC (×109/L)	294	15	23	31	145–387
WBC (×109/L)	294	0.1	0.3	0.5	3.5–11.0
CRP	293	35	79	120	<5

Summary of Multiplate, TEG and laboratory results in ten thrombocytopenic patients with haematological malignancies enrolled consecutively at Haukeland University Hospital in Bergen (Norway) from June 2013 until February 2014.

TEG: thromboelastography; N: number of samples; Q: quartile; ADP: adenosine diphosphate; TRAP: thrombin receptor-activating peptide; r: reaction time; alpha: alpha angle; MA: maximum amplitude; TPC: total platelet count; WBC: white blood cell count; CRP: C-reactive protein.

When platelet count fell below 33×10⁹/L, Multiplate^® AUC was zero with at least one agonist for many of the samples, regardless of bleeding status. The odds of grade 2 bleeding decreased significantly as the ln of AUC increased for all four agonists (Table II) when each was the only predictor in the model, but when all agonists were predictors, only an increase in ln of ADP was associated with a significant decrease in the odds of grade 2 bleeding. Grade 2 bleeding never occurred, for any of the agonists, when the AUC was above eight.

Table II.

Results from multiple logistic regression (using generalised estimating equations) of significant bleeding with respect to Multiplate results^a.

Predictor variable	Simple analysis				Multiple analysis
	Odds ratio	95% Wald CI		p-value	Odds ratio	95% Wald CI		p-value
		Lower	Upper			Lower	Upper
Ln(1+AUC ADP)	0.42	0.22	0.83	0.012	0.42	0.18	0.94	0.035
Ln(1+AUC COL)	0.40	0.17	0.92	0.031	0.46	0.10	2.05	0.305
Ln(1+AUC RISTO)	0.48	0.26	0.90	0.022	1.50	0.47	4.81	0.497
Ln(1+AUC TRAP)	0.51	0.30	0.85	0.009	1.29	0.58	2.83	0.532

Results from multiple logistic regression (using generalised estimating equations) of significant bleeding for 14 inclusions in ten haemato-oncologic patients with thrombocytopenia enrolled consecutively at Haukeland University Hospital in Bergen (Norway) from June 2013 until February 2014.

^aA value of 1 was added to avoid the logarithm of zero which is undefined.

CI: confidence interval; AUC: area under the curve; ADP: adenosine diphosphate; COL: collagen; RISTO: ristocetin; TRAP: thrombin receptor activating peptide.

WHO grade 2 bleeding occurred more frequently as the TPC decreased (p<0.001). Grade 2 bleeding never occurred when the TPC was above 32×10⁹/L. All four ln-transformed AUC variables were significantly correlated to TPC at the 0.01 level (2-tailed).

An increase in the ln of WBC also significantly decreased the odds of grade 2 bleeding and this finding persisted after adjusting for TPC. All four ln-transformed AUC variables were significantly correlated to ln(WBC) at the 0.01 level (2-tailed).

ROC curve analysis with WHO grade 2 bleeding as the state variable was performed for the four Multiplate^® tests combined (Figure 1a), for TPC and for Ln(WBC) (Figure 1b). The analyses showed that for Multiplate^® AUC with ADP, collagen, ristocetin and TRAP as agonists the NPV was 96.1% and the PPV was 19.4%. For TPC (with a cut-off of 15) and WBC (with a cut-off of 0.15) the NPV were 94.9% and 96.3% and the PPV 15.0% and 15.1%, respectively.

Figure 1.

Receiver operating characteristic (ROC) curve with significant bleeding as the state variable.

a) The test variable was mean predicted of a generalised estimating equations analysis with the natural logarithm of 1 + the area under the curve, Ln(1+AUC), of the four Multiplate agonists adenosine diphosphate, collagen, ristocetin and thrombin receptor activating peptide as predictor variables and significant bleeding as the outcome variable. The diagonal line is a reference line.

b) ROC curve with significant bleeding as the state variable and total platelet count (the most jagged line) and white blood cell count (the line with the long diagonal segment) as test variables, respectively.

The diagonal line is a reference line.

Model for the evaluation of bleeding risk combining thromboelastography, Multiplate^®, clinical and laboratory variables

Antibiotics were omitted from the model because all significant bleeds occurred on days with antibiotic treatment. A chi square test of independence between significant bleeding and use of antibiotics resulted in a p-value (exact, 2-sided) of 0.237, which suggests that there is no statistically significant association between antibiotics and significant bleeding. Multiplate^® AUC for the four agonists (ADP, collagen, ristocetin and TRAP), the three TEG parameters (reaction time, alpha angle and maximum amplitude), TPC, WBC, fever, spleen enlargement and HLA-antibodies were predictor variables in the model, and WHO grade 2 bleeding was the outcome variable.

The odds of significant bleeding were significantly lower when collagen-activated Multiplate^® aggregation, TEG alpha angle and TEG reaction time increased, and higher as CRP concentration increased. ADP-, ristocetin- and TRAP-activated Multiplate^® aggregation, TEG maximum amplitude, TPC, WBC, fever, spleen enlargement and HLA-antibodies did not significantly affect the odds of grade 2 bleeding (Table III).

Table III.

Results from multiple logistic regression (using generalised estimating equations) of significant bleeding.

Predictor variable	Odds ratio	95% Wald CI		p-value
		Lower	Upper
Intercept	1.51	0.00	4,158.67	0.919
Ln(1+AUC ADP)	0.88	0.21	3.71	0.865
Ln(1+AUC COL)	0.18	0.04	0.75	0.019
Ln(1+AUC RISTO)	1.10	0.25	4.74	0.902
Ln(1+AUC TRAP)	2.16	0.44	10.57	0.342
TEG R	0.72	0.54	0.97	0.030
TEG alpha	0.88	0.80	0.97	0.008
TEG MA	1.10	0.95	1.26	0.195
TPC	1.01	0.98	1.04	0.571
WBC	0.36	0.03	4.41	0.427
CRP	1.02	1.00	1.04	0.016
Fevera	1.06	0.17	6.67	0.949
Spleen enlargementa	2.81	0.47	16.85	0.259
HLA antibodiesa	0.69	0.23	2.09	0.509

Results from multiple logistic regression (using generalised estimating equations) of significant bleeding with respect to Multiplate agonists for 14 inclusions in ten thrombocytopenic patients with haematological malignancies enrolled consecutively at Haukeland University Hospital in Bergen (Norway) from June 2013 until February 2014.

^aDichotomous variable with “No” as reference category.

CI: confidence interval; AUC: area under the curve; ADP: adenosine diphosphate; COL: collagen; RISTO: ristocetin; TRAP: thrombin receptor activating peptide; TEG: thromboelastography; R: reaction time; alpha: alpha angle; MA: maximum amplitude; TPC: total platelet count; WBC: white blood cell count; CRP: C-reactive protein; HLA: human leucocyte antigen.

A ROC curve was constructed with significant bleeding as the state variable and with the mean predicted from the GEE analysis of the combined model as the test variable (Figure 2). The ROC curve had an AUC of 0.851. The optimal cut-off for the test variable gave a test sensitivity of 94.1% and specificity 58.8%. The NPV was 99.0% and the PPV was 19.0%.

Figure 2.

ROC curve with bleeding of WHO grade 2 as state variable.

The test variable was the mean predicted of a generalised estimating equations analysis of a combined model with significant bleeding as outcome variable and Multiplate, TEG, TPC, WBC, CRP, splenomegaly, HLA-antibodies and fever as predictor variables. The diagonal line is a reference line.

ROC: receiver operating characteristic; WHO: World Health Organization; TEG: thromboelastography; TPC: total platelet count; WBC: white blood cell count; CRP: C-reactive protein.

Impact of platelet transfusion on Multiplate^® area under the curve

The results of the linear mixed model analyses with Multiplate^® AUC as outcome and platelet transfusion as predictor variable are presented in Table IV.

Table IV.

Results from linear mixed model analysis of the impact of platelet concentrate transfusion on the natural logarithm of 1+Multiplate area under the curve^a, in simple analysis and adjusted for the total platelet count.

Outcome	Simple analysis				Adjusted for TPC
	N	B	95% Wald CI	p	N	B	95% Wald CI	p
Ln(1+AUC ADP)	189	−0.248	−0.429, −0.068	0.007	188	−0.079	−0.236, 0.079	0.325
Ln(1+AUC COL)	189	−0.359	−0.535, −0.182	<0.001	188	−0.205	−0.354, −0.056	0.007
Ln(1+AUC RISTO)	185	−0.356	−0.556, −0.156	0.001	184	−0.146	−0.320, 0.027	0.098
Ln(1+AUC TRAP)	189	−0.300	−0.499, −0.100	0.003	188	−0.134	−0.310, 0.042	0.134

Results from linear mixed model analysis of impact of platelet concentrate transfusion on the natural logarithm of 1+Multiplate area under the curve in simple analysis and adjusted for the total platelet count for 14 inclusions in ten thrombocytopenic patients with haematological malignancies enrolled consecutively at Haukeland University Hospital in Bergen (Norway) from June 2013 until February 2014.

^aA value of 1 was added to avoid the logarithm of zero which is undefined.

Ln: natural logarithm; AUC: area under the curve; TPC: total platelet count; N: number of samples; B: estimated regression coefficient; CI: confidence interval; ADP: adenosine diphosphate; COL: collagen; RISTO: ristocetin; TRAP: thrombin receptor activating peptide.

A platelet transfusion on the day before analysis increased the AUC significantly for all four agonists, but when adjusting for TPC the association remained significant only for collagen.

Discussion

The primary aim of this study was to investigate the feasibility of Multiplate^® aggregometry in patients with haematological malignancies, as the usefulness of this method is debated, partially due to uncertainties related to the comparison of results with those of other platelet tests^43–45. An important finding was that the odds of grade 2 bleeding decreased significantly with increasing Multiplate^® aggregation, even though many samples had undetectable aggregation when the platelet count was below 33×10⁹/L, regardless of the patient’s bleeding status. Other studies also showed that Multiplate^® aggregation was severely reduced in thrombocytopenic patients with platelet counts below 50×10⁹/L^23–26.

We identified one other study investigating Multiplate^® aggregation as a possible predictor of bleeding in patients with haematological malignancies⁴⁶. In that study Multiplate^® ristocetin-activated aggregation was significantly correlated to bleeding events in patients with chronic lymphocytic leukaemia treated with ibrutinib. Costache et al. reported that ADP-, collagen- and TRAP-activated Multiplate^® aggregation scores were significantly reduced in patients with acute leukaemia who developed haemorrhagic complications compared to those without bleeding¹¹. Wozniak et al. found that in patients on antiplatelet therapy, AUC <26 U in the ADP test was strongly predictive of serious bleeding complications after coronary artery bypass graft surgery⁴⁷, while Ranucci et al. found that AUC <22 U in the ADP test was associated with severe bleeding in patients on dual antiplatelet therapy undergoing coronary artery bypass grafting with cardiopulmonary bypass, unless the AUC was >75 U in the TRAP test⁴⁸. Ellis et al. found that a preoperative ADP result of AUC <68 U in patients undergoing coronary artery bypass grafting with cardiopulmonary bypass predicted increased risk of postoperative bleeding with a sensitivity of 0.48¹⁷. These results are parallel to our finding that Multiplate^® AUC is reduced on days with bleeding. As a part of the feasibility study, we also evaluated whether there was an effect of platelet transfusions on the results of the Multiplate^® analysis. Platelet transfusion on the preceding day was positively associated with AUC for all four agonists in simple analyses. The association remained significant for collagen (p=0.007) after adjusting for TPC, which indicates an improvement in platelet function. Several studies have demonstrated a positive relationship between TPC and Multiplate^® AUC^{10,25,49–53}. It is not known whether the improved aggregation responses after a platelet transfusion are due only to the increased platelet count, or whether the transfused platelets also function better. It is probable that the technical sensitivity of the Multiplate^® testing is limiting the value of the test in haematological patients with thrombocytopenia. Tiedemann Skipper et al. suggested that Multiplate^® testing could be a future tool for testing platelet function in thrombocytopenic patients in general²⁶. They later showed that platelets from patients with immune thrombocytopenic purpura function better than platelets from healthy individuals diluted to similar platelet concentrations⁵⁴. However, the aggregation responses in samples from thrombocytopenic cancer patients were heterogeneous, with most samples within and a few samples below the prediction interval. Despite its limitations, the widespread availability of platelet counting makes this still the most clinically relevant investigation.

The secondary aim of this exploratory study was to evaluate whether a combined model with results from Multiplate^® and TEG analyses, together with clinical and laboratory variables, might have a role in predicting bleeding in thrombocytopenic patients with haematological malignancies. In the GEE analysis that was used to generate a test variable for ROC curve analysis, the odds of WHO grade 2 bleeding were significantly lower when collagen activated Multiplate^®, TEG alpha angle or TEG reaction time increased, and higher as CRP increased. The ROC curve analysis of this model resulted in a test with a NPV of 99.0% and PPV of 19.0%. These are slightly better than TPC alone, which gives a NPV of 94.9% and PPV of 15.0%.

Our results extend earlier work by presenting combined TEG and Multiplate^® variables in the prediction of bleeding. In a study of TEG in thrombocytopenic patients with haematological malignancies, Kasivisvanathan et al. found that the TEG parameters, reaction time, alpha angle and maximum amplitude all indicated greater hypocoaguability in patients who experienced clinically significant bleeding than in patients without such bleeding. They concluded that this was related to lower fibrinogen levels, since functional fibrinogen was lower in bleeding patients, while the platelet counts were similar in both groups of patients³⁶. We have previously reported that only TEG alpha angle was significantly associated with WHO grade 2 bleeding in thrombocytopenic patients with haematological malignancies⁵⁵. With a cut-off at 32.15 degrees, the NPV of alpha angle was 95.8% and its PPV was 23.1% for WHO grade 2 bleeding. The differences in findings between the study by Kasivisvanathan et al. and our study may be explained by differences in sampling and analysis. Kasivisvanathan and colleagues analysed three samples from each patient, on three consecutive days³⁶. They summarised the results from each time-point and used a Mann-Whitney test to compare the results from patients with and without bleeding. In our study we took multiple samples from each patient, and mixed model analyses were used with bleeding as the predictor and each TEG parameter as an outcome³². In patients undergoing resuscitation with blood components, TEG parameters have been successfully used to guide transfusions^56–58. Rapid TEG also seems to be a promising technique⁵⁹.

Figure 1b shows that WBC also influences the results of Multiplate^® analyses. It is known that both polymorphonuclear cells^60–62 and in particular monocytes⁶³ express tissue factor, which is essential for in vivo coagulation. This may be clinically relevant for our group of patients, as treatment may suppress tissue factor expression⁶⁴ and also reduce formation of platelet-leucocyte aggregates⁶⁵.

In a study by Diehl and colleagues, the results of the Multiplate^® ADP-test correlated significantly with WBC when no anti-platelet medication was administered⁶⁶. Haberka et al. found that lymphocyte count correlated with platelet reactivity during clopidogrel treatment⁶⁷. It was also demonstrated that arachidonic acid (ASPI)-, ADP- and TRAP-induced aggregation was greater in patients with higher leucocyte counts⁶⁸. However, Seyfert et al. conflictingly reported that Multiplate^® was not influenced by WBC⁵². In a comprehensive report of other test methods it was described that the products released from polymorphonuclear cells contribute to both thrombotic reactions and induced lysis of thrombi⁶⁹.

Other methods have also been investigated in the search for a better way to predict bleeding in thrombocytopenic patients. Batman et al.⁷⁰ measured platelet reactivity by flow cytometry after stimulation with different agonists and found that agonist-induced platelet reactivity in thrombocytopenic haemato-oncological patients was inversely correlated to bleeding. Vinholt and colleagues also measured platelet aggregation by flow cytometry and concluded that patients with AML and MDS had reduced platelet aggregation and that low platelet aggregation identified a bleeding tendency⁷¹. They found that at a cut-point of 38% aggregation TRAP had a NPV of 100% and a PPV of 75% for bleeding. For other agonists (ADP, collagen-related peptide) the NPV was 100% and PPV 67% at a cut-point of 40% aggregation. These studies indicate that platelet function is important for bleeding risk, but that the current test battery has limited availability to identify this risk in a manner that may safely predict bleeding in haematological patients.

Our small study has several limitations. The results are pilot data, not powered for clinical outcomes, and should be considered as hypothesis testing. A high proportion of the Multiplate^® analyses had no detectable aggregation. We acknowledge heterogeneity in the diagnosis and treatment regimens in our patients. However, a total of 189 study days with Multiplate^® and TEG analyses were undertaken, and the results provide a baseline for further studies.

Conclusions

We conclude that the feasibility of Multiplate^® aggregometry in thrombocytopenic patients is debatable, as aggregation may be undetectable both on days with and without bleeding. However, the findings that no WHO grade 2 bleeding occurred when the AUC was above eight and that platelet transfusion increases AUC indicate that the test may have value in evaluating such patients. The odds of grade 2 bleeding decreased significantly as the ln of AUC increased for all four tested agonists, and Multiplate^® aggregation may have a role in evaluating bleeding risk as a model for predicting bleeding showed a high NPV (96.1%), but a low PPV (19.4%). A combined model implementing Multiplate^®, TEG, clinical and laboratory variables resulted in similar numbers with a NPV of 99.0% and PPV of 19.0% for predicting WHO grade 2 bleeding. Thus, this study underlines that no easy test to predict bleeding in haematological patients is available. However, given the small sample size, these results should be considered hypothesis-generating to inform further research evaluation.