Agreement between ACT and aPTT during extracorporeal membrane oxygenation shows intra- and inter-individual variation

Introduction

Patients undergoing extracorporeal membrane oxygenation (ECMO) support are usually anticoagulated with intravenous unfractionated heparin (UFH) to prevent clot formation through blood contact with non-biological surfaces encountered in the extracorporeal circuit. ¹ Achieving correct control of coagulation is difficult and both haemorrhage and thrombosis are leading complications of ECMO. ² The response to heparin is highly variable and no consensus exists on the tests that should be used to monitor the adequacy of anticoagulation. There are a number of additional complexities when optimizing anticoagulant control for circulatory assist circuits and devices, such as the length of tubing, surface coating of the tubing and oxygenator, the flow pattern (continuous vs pulsatile) and the intersection points of the circuit with the patient’s circulation. ³

Several assays can be used in the monitoring of unfractionated heparin. Most centres use any combination of activated clotting time (ACT), activated prothrombin time (aPTT) and anti-Xa levels. The results from these assays can vary from centre to centre, between types of patients and even within the same patient.

We present a retrospective study of the performance of aPTT compared to low range ACT in a single centre and explore possible confounding factors.

Methods

We conducted a retrospective service evaluation of data obtained in 34 consecutive adult patients who underwent veno-venous (VV) or veno-arterial (VA) ECMO support between October 2012 and August 2013 in a single centre in the UK. Given the retrospective nature and absence of treatment intervention or randomisation, ethical approval was not required by our ethics committee. Institutional approval to analyse these data was granted. Both VV and VA ECMO provide respiratory support to patients, with VA ECMO additionally providing haemodynamic support. Results from blood samples taken from all patients undergoing ECMO during this time period were extracted from the Clinical Information System and processed using Excel 2010 (Microsoft, Redmond, WA, USA). In accordance with our protocol, the ACT was sampled four times per day and the APR (aPTT ratio, calculated by dividing aPTT by the mean of the reference range) was sampled twice a day, when changes to the circuit were made and when the patient’s clinical condition changed.

The ACT was obtained at the bedside using Maxact ACT® (Helena Laboratories, Beaumont, TX, USA) containing an activator cocktail of glass particles, kaolin and cellite. The aPTT was measured in the laboratory using SynthasIL® (Colloidal Silica and synthetic phospholipid activator) reagent on an IL TOP analyser (Instrumentation Laboratory, Warrington, UK).

In order to allow direct comparison of the aPTT and ACT, results were normalized: the aPTT was normalized by dividing the actual value by the mean of the local reference range (31s) and is known as the APR (aPTT ratio) in UK clinical practice; the ACT was normalised (N-ACT) by dividing the actual result by 118s (mean of the reference range provided by the manufacturer). The APR and N-ACT were paired when taken from the same patient, with the ACT always preceding the aPTT, and the heparin infusion rate was kept constant throughout the time between samples as we used the aPTT/APR for heparin dosing in the clinical care of our patients.

Results

Comparison of N-ACT and APR

Of 34 patients identified, 15 were retained for analysis (19 were excluded as they had fewer than 5 matched pairs of N-ACT and APR available).

In the retained sample (N=15: 11 males, 4 females), the mean patient age was 45.8 (SD: 14.1). A total of 675 matched result pairs of N-ACT and APR were obtained from these 15 patients.

The data were pooled and the N-ACT was plotted against the APR (Figure 1). Table 1 shows the number of matched pairs each patient contributed to the pooled data and the correlation coefficient and results of the Bland-Altman analysis for each patient.

OPEN IN VIEWER

Figure 1.

Graph showing the relationship between APR and N-ACT. Survivors are shown by crosses and non-survivors are shown by triangles. Note that the total number of crosses and triangles on the graph is less than the total 675 as some represent data with the same paired values.

OPEN IN VIEWER

Table 1.

Table shows the number of paired APR and N-ACT values, the correlation coefficient and the Bland-Altman plot analysis data for each individual patient.

Patient	Number of contributing matched ACT and aPTT pairs	Pearson Product Correlation Coefficient	Bland Altman Mean difference (N–ACT – APR)	Upper limit	Lower Limit	Spread of data	Duration on ECMO (Hours)
1	22	0.53	−0.01	0.51	−0.52	1.03	179
2	87	0.44	0.10	0.72	−0.53	1.25	70883
3	47	0.80	0.10	0.57	−0.36	0.93	450
4	25	0.78	0.17	0.76	−0.42	1.18	5611
5	5	0.77	0.38	1.80	−0.86	2.60	17
6	24	0.70	0.04	0.62	−0.54	1.16	296
7	8	0.58	−0.4	1.40	−2.10	3.50	84
8	12	0.71	0.39	0.52	0.26	0.78	134
9	15	0.47	−0.23	0.47	−0.94	1.41	142
10	88	0.38	−0.11	0.16	−0.37	0.53	1529
11	26	0.68	0.28	0.54	0.03	0.57	341
12	84	0.03	0.19	0.71	−0.33	1.04	994
13	97	0.71	0.25	0.53	−0.03	0.56	405
14	86	0.56	−0.02	0.41	−0.46	0.87	739
15	49	0.45	0.04	0.48	−0.40	0.88	1324
TOTAL	675	0.55	−0.08	0.51	−0.68	1.19

Correlation between N-ACT and APR measurements was r=0.55 (95% CI 0.50 to 0.60; p<0.0001). Using the amended criteria (ACT and aPTT samples taken within 2 minutes of each other), a sample size of 178 was created and this demonstrated a correlation coefficient of 0.46 (95% CI 0.33 to 0.57; p<0.0001). From the analysis of the Bland-Altman plot (Figure 2), the mean difference between the calculated APR and N-ACT was −0.08 with the N-ACT being recorded (on average) as the higher ratio. The limits of agreement (mean difference +/− 1.96 standard deviations) demonstrate a wide spread of data (−0.67 to + 0.51) and suggests that agreement between APR and N-ACT is poor. When the analysis was repeated using the amended criteria, the Bland-Altman plots demonstrated a mean difference of −0.1 and a wide spread of data (−0.81 to +0.60). The mean absolute difference between paired N-ACT and APR measurements was 0.24 (SD: 0.20).

OPEN IN VIEWER

Figure 2.

Bland-Altman plot of data collected from 15 individuals. Survivors shown by crosses and non- survivors shown by triangles. Note that the total number of crosses and triangles on the graph is less than the total 675 as some represent data with the same paired values.

Discussion

We have demonstrated that, despite a moderate degree of positive correlation between the N-ACT and the APR, agreement between the assays is poor.

Both heparin-monitoring assays are based on activation of the intrinsic pathway of coagulation. Historically, the protamine titration assay is considered the gold standard assay; however, aPTT/APR measurements are now the standard assay used for the titration of UFH therapy. Most hospitals in the UK have a dosing flow chart for aPTT/APR to allow unfractionated heparin dosing per adjustment of infusion rate.

ACT is a point-of-care coagulation test, originating from the Lee-White coagulation time ⁵ and is used to monitor UFH therapy when intense short-term anticoagulation is required with levels of >1U/ml, such as during cardiac bypass. ⁵ The ACT is also used for UFH monitoring in some centres, even when UFH targeted levels are <1U/ml. The desired range for ACT varies between centres, but the aim is a similar relative prolongation of clotting time compared to baseline ACT as for the aPTT. ⁵

In a 2013 survey of Extra-corporeal Life Support Organisation (ELSO) registered ECMO centres, 97% of respondents use the ACT as a tool to monitor heparinisation. The aPTT was used by 94% in addition to the ACT on a regular basis, with a variation in therapeutic ranges between centres. ⁶ Despite being the most widely used assay, the aPTT is not standardized between hospitals like, e.g., the international normalized ratio (INR), with different sensitivities of reagents and reference ranges. ⁷

We have demonstrated for the first time in a regression model that platelet count and urea are independent predictors of agreement between the ACT and the APR.

Reducing platelet counts and increasing urea and creatinine concentrations are related to worsening of agreement between the assays. This could explain some of the discrepant results other investigators have found. ECMO patients are critically ill, with a proportion suffering from thrombocytopenia at some stage. The cause of respiratory failure and the degree of multi-organ failure will vary between patients and concomitant events, such as septic episodes, will introduce additional levels of complexity, ⁸ leading to intra- and inter-individual variation. Sepsis will upregulate clotting factors that are acute phase proteins, such as factor VIII. Microbial invasion of the bloodstream may cause disseminated intravascular coagulation, with consumption of clotting factors and platelets resulting in coagulopathy and organ failure. ⁹

Changes in the composition of whole blood and plasma volume (i.e., platelet count, haematocrit or a combination thereof), changes in platelet function and clotting factor levels over time and contact factor activation ¹⁰ are likely to affect plasma and whole blood assays in a slightly different fashion. It is well known that haemostasis is altered in uraemic patients. ¹¹ The exact contributing mechanisms are unknown. However, it is possible that the platelet-fibrinogen interaction is disturbed in uraemic states.

A few studies have previously explored the relationship between the aPTT and the ACT in adult patients undergoing ECMO. Atallah et al. investigated the relationship between the heparin dose and the ACT and aPTT and concluded that the heparin dose correlated better to the aPTT. It was the ACT that was used to adjust the heparin dose in their patients in this study and the observed correlation between ACT and aPTT was 0.41. ¹²

Others have explored the relationship of ACT, aPTT and heparin/antiXa level outside the setting of ECMO. In adults, Koerber et al. found that the aPTT showed a better correlation to the heparin infusion rate than the ACT in unselected adult patients commencing intravenous UFH therapy. They noted good correlation, but significant differences between aPTT and ACT, with the aPTT generally correlating better to the plasma heparin concentration. ¹³ DeWaele et al. have analysed patients in the intensive care unit setting and found similar results showing a significant correlation between ACT and aPTT despite significant differences that varied dependent on the intensity of anticoagulation, making it difficult to substitute one method for another. ¹⁴

Bosch et al. examined factors that determined poor heparin response on cardiopulmonary bypass of the ACT to doses of 300-400 U/kg of UFH. They found that 10% of patients showed reduced responses and that those patients often had higher platelet counts, lower haemoglobin and equivalent antithrombin levels compared to the normal responders, but that this was not statistically significant. ¹⁵

Conclusion

We have demonstrated that, despite a moderate degree of positive correlation between the N-ACT and the aPTT, agreement between the assays is poor.

We have shown for the first time in a regression model that platelet count and urea are independent predictors of agreement between the ACT and the APR.

Reducing platelet counts, increasing urea and creatinine concentrations are related to worsening of agreement between the assays.

A comparison of results should be made with caution. Centres must carefully decide on which test to base their heparin-dosing regimen and this will have to be reviewed on an individual basis.