Oral anticoagulation therapy (OAT) with coumarins (vitamin K antagonists) has been available for more than 60 years, and is prescribed for both prophylactic and therapeutic use in patients at increased risk of thromboembolism . Monitoring is based on the International Normalized Ratio (INR), which is conventionally determined on citrated plasma obtained by venepuncture.
The INR is a mathematically adjusted prothrombin time (PT) , for which reason the ‘objective true’ INR is not known [3,4]. Differences in INRs obtained with the same test sample are observed, but this fact should not be regarded as an indictment of the INR system, but as merely reflecting the variables in PT testing .
The INR system of PT standardization was originally based on manual tilt tube determination of PTs, and envisaged the assignment of a single International Sensitivity Index (ISI) value for each batch of thromboplastin reagent [5,6]. Today, however, the manual PT technique has been almost universally replaced by automated point-of-care testing (POCT) coagulometers.
Optimized management of OAT improves the clinical quality of treatment [7–9], and high-quality OAT is still an alternative to novel oral anticoagulants, e.g. dabigatran . There are different methods of managing OAT, including routine care (provided by the general practitioner), hospital outpatient clinics, highly specialized anticoagulation clinics, shared care, use of computer-assessed dosage, patient self-testing (PST), and patient self-management (PSM) [1,11–13]. PSM in OAT means that the patient analyzes a drop of blood with a portable POCT coagulometer (INR monitor). The POCT coagulometer displays the INR, which the patient then uses for coumarin dosage. PST merely means that the patient performs blood sampling and analysis, and a healthcare provider decides on dosage adjustment . These new treatment modalities are currently applied to several hundred thousand patients . The person responsible for letting patients on OAT perform PST or PSM or referring them for these treatment modalities is most often the treating physician. Accordingly, it is important for the physician to know the advantages and pitfalls of these treatment modalities, including the performance of the POCT coagulometers. A precondition for a correct dosage of coumarins is correct estimation of the INR, and the method and apparatus used for providing the INR measurements are crucial.
In this context, it is important to state that PSM is a method including patient self-dosing based on self-testing of INR, whereas the coagulometer estimates the quality of INR measurements and not the quality of OAT. However, by increasing the quality of the INR measurements, the quality of treatment can potentially be further increased.
POTC coagulometers have been investigated in a number of studies, which have led to diverse conclusions. It is difficult and challenging to perform an overview of the literature, owing to the vast amount of papers, with differences in design, statistical analysis, etc. The aim of this systematic review was to analyze the current literature, especially regarding the precision and accuracy of the POCT coagulometers, to provide recommendations for clinical use and quality control, and to point out areas for future research.
Material and methods
Publications were identified through a search of the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2011, issue 3) and PubMed (start of 1951 to September 2011). The search was supplemented by a review of personal files and a hand search of published reviews. The following strategy was used to search the CENTRAL and was adapted appropriately for PubMed:
((((“4-Hydroxycoumarins”[MeSH]) OR (acenocoumar* OR sinkumar OR sinthrome OR sintrom OR mini-sintrom OR syncoumar OR syncumar OR synthrom) OR (bishydroxycoumarin OR dicoumarin OR dicoumarol) OR (phenprocoumarol OR phenylpropylhydroxycumarine OR phenprocoumon OR falithrom OR liquamar OR marcoumar OR marcumar) OR (biscoumacetate ethyl OR ethyldicoumarol OR carbethoxy dicoumarol OR pelentan OR tromexan) OR (warfarin potassium OR warfarin sodium OR coumadin)) OR (“Anticoagulants”[MeSH])) AND (“Administration, Oral”[MeSH] OR oral*)) AND (“Self Administration”[MeSH] OR “Self Medication”[MeSH] OR home based OR self monitoring OR self monitored OR self administ* OR self medication* OR self manag* OR self care).
Articles were selected on the basis of titles and abstracts relevant to the topic and whether they contained original data. Additional relevant articles were identified by review of references in key publications. Only articles in English were included.
We extracted the following data from all included studies: author and publication year, type and working principles of POCT coagulometer, precision, accuracy, internal and external quality control, and calibration. We differentiated between low-quality and high-quality studies, and high-quality (methodologically strong) studies were defined as those studies fulfilling the following criteria: (i) number of INR measurements > 100; (ii) use of an INR comparator, such as either manual tilt tube determination or a laboratory using certified plasma/external quality control; (iii) use of Bland–Altman plots or equivalent (e.g. mean/median difference), and not merely linear interpolation, for testing accuracy.
Definition of precision and accuracy of INR
In general, the terms precision and accuracy are used when estimating the quality of a method; accuracy is the degree of veracity, whereas precision is the degree of reproducibility.
Precision is descriptive in general terms (e.g. acceptable and poor), whereas imprecision (the reciprocal of precision) is expressed by means of the standard deviation or coefficient of variation (CV) of the variability. The term precision can be replaced by the terms repeatability and reproducibility. Repeatability is the agreement between the results of consecutive measurements (within the same measuring series), and reproducibility is the agreement between the results of discontinuous measurements of INR carried out under changing measuring conditions over time [15,16]. The analytical variability deals with the apparatus used for INR measurements, and it has been suggested that this should have a CV of < 3% .
Accuracy is the level of agreement between the result of one measurement and the true value. Accuracy can be divided into analytical accuracy (‘does the test give the same results estimated purely numerically as the gold standard?’) and diagnostic accuracy (‘does the test provide accurate information about diagnosis, prognosis, risk of disease, and other clinical issues as the gold standard?’) . The analytical accuracy should ideally be below ± 0.2 INR within the therapeutic INR target range (for most purposes, between 2 and 3) [3,17,18]. The diagnostic accuracy can be defined as relevant (i.e. not resulting in a change in coumarin dosage), using different degrees of agreement [19,20], or according to the definition by Poller et al. , where a deviation of ≥ 15% was defined as clinically important.
Components of variability regarding the INR measurements
The variability of the INR measurements depends on the combination of biological, preanalytic and analytic variables. The analytic variability is displayed and discussed in the paragraphs regarding POCT coagulometer performance (see below).The preanalytic variability depends on various factors – does the patient or healthcare provider take the blood sample, handle the test strips, take the temperature, etc. – but the influence of these factors has not yet been quantified.
The biological variability comprises factors within the patient that influence the INR (e.g. vitamin K-dependent factors, calcium, and magnesium), and also the interaction between patient factors and the PT measurement system (e.g. type of thromboplastin and reagent used) . The biological variability is therefore the ‘within-patient’ variability, and it is important to estimate and isolate the influence of this factor as a basis for evaluating a new method/apparatus for INR measurements. The biological variability has been found to have a CV of approximately 9% [18,22]. However, the preanalytic and biological variabilities are often considered as one parameter . With the POCT coagulometers, the concomitant application of low molecular weight heparin, hematocrit, fibrinogen level, the presence of antiphospholipid antibodies and, potentially, polyglobulia, hypertriglyceredemia, hemolysis, sensitivity to factor V and hyperbilirubinemia has a relatively large impact on the INR result [24–27].
Working principles of the POCT coagulometers
All POCT coagulometers work in basically the same way; a drop (10–30 μL) of capillary whole blood is released by a finger puncture device applied to a test strip, and inserted into the POCT coagulometer. A clotting process is initiated by thromboplastin, and clot formation is subsequently detected. However, the method of clot detection is different from one POCT coagulometer to another, and there are minor differences in terms of function. All POCT coagulometers are calibrated with lot-specific code chips or stored lot-specific conversion equations. Furthermore, most of them allow for both internal and external quality control.
Available types of POCT coagulometer
A brief overview of currently available POCT coagulometers is given in Table 1. More in-depth discussions of the functionality and technical specifications of the POCT coagulometers are given in other papers, e.g. [28–30]. Some POCT coagulometers are not shown in Table 1, for the following reasons. There are simplified versions (Hemochrom Jr and GEM PCL) of ProTime/ProTime 3 Correct [31,32], the predecessors of CoaguChek in terms of the Coumatrack (Dupont Pharmaceutical Co., Wilmington, DE, USA) [33–37] and Biotrack 512 (Ciba-Corning, East Walpole, MA, USA) [31,38,39]. Furthermore, AvoSure PT-Pro (Avocet Medical, San Jose, CA, USA) and Harmony (LifeScan, Johnson & Johnson, Milpitas, CA, USA) have been withdrawn from the market . The Thrombolytic Assessment System/RapidpointCoag (Pharmanetics, Morrisville, NC, USA) is not included, because it does not analyze capillary blood [31,41–44]. The i-STAT System (Abbott Point of Care, Princeton, NJ, USA) is a semiprofessional system, and is thus not shown . Furthermore, since 1 January 2011, Roche Diagnostics has not manufactured any more strips for CoaguChek and CoaguChek S, and accordingly the results for these POCT coagulometers are not presented. Finally, only POCT coagulometers displaying the result in INR are included.
Table 1. Overview of currently available coagulometers
|Manufacturer||Roche Diagnostics||HemoSense||International Technidyne Corporation||Unipath|
|Methodology||Electrochemical detection of thrombin activity||Electrochemical detection of changes in impedance in sample||Cessation of blood flow through capillary channel||Oscillation of a metal disk (magnetic field)|
|Thromboplastin||Human recombinant (ISI = 1.0)||Human recombinant (ISI = 1.0)||Human recombinant (ISI = 1.0)||Rabbit brain (ISI value is not available)|
Fourteen studies have included the CoaguChek XS (Table 2) [45–58]. The precision in terms of imprecision varied from a CV% of 1.4 to a CV% of 5.9. Regarding the accuracy, the correlation coefficient varied from 0.91 to 0.95 and the concordance was approximately 95%, although, again, it was highly dependent on the definition of boundaries.
Table 2. Precision and accuracy of the CoaguChek XS coagulometer
|Bauman et al. ||CV = 5.0||Mean difference = 0.13; 95% limits of agreement, − 0.22–0.48 |
Concordance = 98% (narrow*)
Concordance = 98% (expanded†)
Median deviation = 5.7%
|Bereznicki et al. ||–||Correlation coefficient = 0.91 |
Concordance = 100% (narrow*)
Concordance = 100% (expanded†)
5.1% of measurements deviated by ≥ 15% from the laboratory measurements
|Braun et al. ||CV = 5.92 and CV = 5.16||Correlation coefficient = 0.84 (CCS vs. CCXS) |
MRD = 12.1%
|Christensen et al. ||CV = 2.3||0.40 INR lower than laboratory measurement |
40% of measurements deviated by ≥ 15% from the laboratory measurements
|Greenway et al. ||–||Correlation coefficient = 0.81 (as compared with standard laboratory) |
Correlation coeficient = 0.90 (as compared with WHO reference method)
|Karon et al. ||–||Median = − 0.2 INR|
|Kong et al. ||–||Correlation coefficient = 0.95|
|Nam et al. ||CV = 1.4||Concordance = 0.96 (mean difference)|
|Paioni et al. ||–||Thromboplastin 1: |
Mean difference = − 013; 95% limits of agreement, − 0.58–0.32
Concordance = 100% (narrow*)
Concordance = 100% (expanded†)
Mean difference = − 014; 95% limits of agreement, − 0.58–0.31
Concordance = 93% (narrow*)
Concordance = 100% (expanded†)
|Plesch et al. ||CV = 2.9–4.0||Correlation coefficient = 0.94–0.97 |
Mean bias: − 0.22–0.18
|Plesch et al. ||–||Correlation coefficient = 0.96–0.98 |
MRD = 6.4–9.6%
|Sobieraj-Teague et al. ||–||Correlation coefficient = 0.95 |
Concordance = 93.5% (within 0.5 INR)
Concordance = 99.4% (within 0.8 INR)
Concordance = 99.4% (narrow*)
Concordance = 99.4% (expanded†)
|Torreiro et al. ||–||Correlation coefficient = 0.95 |
MRD = 7%
|Williams and Griffiths ||–||96% (within 0.5 INR)|
INRatio, ProTime/ProTime 3, and SmartCheck INR System
Studies on these POCT coagulometers [41,59–65] are presented in Table 3. The studies are limited in number (N = 8), and hence firm conclusions are difficult to obtain. The precision in terms of imprecision varied from a CV% of 3.7 to to a CV% of 8.4. Regarding the accuracy, the correlation coefficient varied from 0.77 to 0.97, and the concordance was approximately 85%, although, again, this was highly dependent on the definition of boundaries.
Table 3. Precision and accuracy of the INRatio, ProTime/ProTime 3 and SmartCheck INR System
|INRatio||Hemkens et al. ||–||Concordance = 87%|
MRD = 11.7%
| ||Murray et al. ||–||Correlation coefficient = 0.921|
| ||Taborski et al. ||CV = 5.4–8.4||Correlation coefficient = 0.954 |
Concordance = 81%
MRD = 6.87%
|ProTime/ProTime 3||Andrew et al. 2001 ||CV = 5.4–8.4||Concordance = 81%|
MRD = 6.9%
| ||Biasiolo et al. ||–||Concordance (± 2 SD): − 0.72–0.76 INR|
| ||McBane et al. ||–||Correlation coefficient = 0.73 |
Concordance = 39%
Mean difference (SD) from plasma = 0.8 INR (0.68) higher
| ||Nowatzke et al. ||CV = 7.4||Correlation coefficient = 0.885|
|SmartCheck INR System||Gardiner et al. ||CV = 3.7–6.3||Correlation coefficient = 0.89–0.90 |
Concordance = 88–97% (± 0.5 INR)
Four studies were classified as high-quality studies (Table 4): three with CoaguChek XS and one with ProTime.
Table 4. Precision and accuracy of the coagulometers in high-quality studies
|Christensen ||CoaguChek XS||CV = 2.3||0.40 INR lower than laboratory value |
40% of measurements deviated by ≥ 15% from the laboratory measurements
|Plesch et al. ||CoaguChek XS||CV = 2.9–4.0||Correlation coefficient = 0.94–0.97 |
Mean bias: − 0.22–0.18
|Plesch et al. ||CoaguChek XS||–||Correlation coefficient = 0.96–0.98, |
MRD = 6.4–9.6%
|Biasiolo et al. ||ProTime||–||Concordance (± 2 SD): − 0.72–0.76 INR|
Quality control and calibration
Whereas laboratory-based INR measurements have been highly standardized and are subject to international quality control [2,66], the same is not true for POCT coagulometer measurements.
Internal quality control
Some manufacturers of POCT coagulometers provide control solutions (liquid) with a known INR value (internal quality control). The frequency of performance of this quality control varies between one and 12 times per year [67–71]. CoaguChek XS uses electronic ‘onboard’ quality control.
External quality control
External quality control can be accomplished with different methods [4,21,67,72–75]:
- 1 Comparing the INR obtained from venous samples analyzed in a laboratory with that of the POCT coagulometer;
- 2 Comparing the INR of a reference POCT coagulometer with that of the POCT coagulometer;
- 3 Comparing plasma (with a known INR value) sent from a central laboratory with the result of the POCT coagulometer;
- 4 Comparing INR measured on a certified (calibrated) POCT coagulometer with that of the patient’s POCT coagulometer, using five sets of plasma.
Method 1 is generally named the split-sample method . Method 2 uses a certified POCT coagulometer as comparator, but does not otherwise differ from method 1. Method 3 is the method developed by the UK National External Quality Assessment Scheme for Blood Coagulation (NEQAS), which provides lyophilized plasma for external quality control to a number of centers using POCT coagulometers . One sample containing two sets of plasma (e.g. sent out every third month) is provided. Method 4 is the method endorsed by the European Action on Anticoagulation (EAA) (formerly named the European Concerted Action on Anticoagulation), which recommends five samples at each time-setting [21,76]. The samples can be delivered by the External Quality Control of Diagnostic Assays and Tests Foundation.
In terms of calibration of the POCT coagulometers, only CoaguChek has been investigated. The manufacturer of CoaguChek calibrates each lot (batch) of test strips and the corresponding code chip. The code chips carry the lot-specific information of the calibration. It is the ISI value that can be calibrated. It is possible to calibrate CoaguChek with whole blood, fresh, citrated plasma [77–79], and even lyophilized plasma .
We found that the POCT coagulometers generally had adequate precision, but the results regarding accuracy were not consistent throughout the studies. In terms of precision, the CV should ideally be < 3% , but the CVs were found to range from 1.4% to 8.4% (Tables 2–4).
Estimating the precision of a POCT coagulometer presents problems, as it is not possible to perform ordinary within-run or between-run estimations. Between-run estimation does not apply to single test systems, because they are not designed for batch or random testing. Methods to circumvent this have been suggested [3,81], but none of these methods is comparable to ordinary within-run or between-run estimations. It is not possible to store the blood without interfering with the coagulation process, and it is not possible to reuse the test strip. A feasible method of estimating the within-run imprecision is to perform INR measurements in duplicate (e.g. using two different fingers), but no established method exists for between-run estimation. All other things being equal, the laboratory is therefore bound to perform better than the POCT coagulometers in terms of precision.
In terms of accuracy, the POCT coagulometers generally tend to overestimate the INR when INR measurements are high, especially above 4.0 [20,47,82]. A tendency to underestimate INR is found when the INR is within or below the therapeutic INR target range [49–51]. Christensen et al.  found that CoaguChek XS underestimated the INR as compared with laboratory measurements with 0.40 INR. Accordingly, no firm conclusions can be drawn.
Regarding clinical accuracy, the INR measurements deviated by ≥ 15% in 40% of all measurements obtained with CoaguChek XS .
When estimating variability, one has to be aware of the comparator used (laboratory). Laboratory measurements also display variability, especially as the original World Health Organization (WHO) method is used only in a very limited number of studies [21,83]. The INR varies with different reagent–coagulometer combinations. The laboratories have an equal interlaboratory variability (CV of approximately 10%) as the POCT coagulometers [84,85]. However, these studies are potentially flawed, as they predominantly estimate the variability of the laboratory from an overall median, and it is not clear whether certified plasma was used. If so, the CV of the interlaboratory variability could be reduced by 50%.
In the studies reporting the highest reduction in interlaboratory variability, it is still in the order of a CV of 5%. This figure should be compared with the inaccuracy found in the POCT coagulometers, and considered in respect of this. Even a CV of 5% of the laboratory measurement variability will result in less pronounced inaccuracy when the POCT coagulometer and the laboratory are compared. This has to be taken into account when a comparison of accuracy is made, as it cannot a priori be stated which of the two methods is the more accurate . Interpretation of the results regarding the potential inaccuracy of the POCT coagulometers has to be performed in this context. When a measurement on the patient’s POCT coagulometer is different from the that of local laboratory, which one is the ‘correct’ INR? This is not known.
Many of the studies shown in Tables 2 and 3 were not optimally designed; for example, most studies used correlation coefficients for estimating accuracy, and not mean vs. difference (Bland–Altman plot) for assessing the mean/median difference . Other parameters regarding the comparator (laboratory), including the ISI value, type of quality control performed in the laboratory, use of calibrated plasma, definition of analytic vs. clinical accuracy, number of included laboratories, and use of automated vs. manual detection of clotting time, vary significantly between studies. Many of the studies have also been supported by the manufacturer of the apparatus (POCT coagulometer), which could introduce potential bias. Finally, the precision of a method should be thoroughly estimated before methods are compared in terms of accuracy . Accordingly, we differentiated studies according to quality, and presented data from the high-quality studies in a separate table (Table 4).
The parameters applied for defining high-quality studies were used in order to allow a more realistic and applicable quantitative assessment of the precision and accuracy. We chose these parameters for the following reasons: (i) a high number of measurements (> 100) provide a more precise estimate ; (ii) a correct comparator (either manual tilt tube determination, a laboratory using certified plasma [defined as plasma where the INR has been assigned to a laboratory by a reference center] or explicit use of external quality control) gives a more correct result [5,66,86–88]; and (iii) application of Bland–Altman plots or equivalents, as the use of linear interpolation is misleading . In the high-quality studies, the performance of the POCT coagulometers was also found to be adequate for clinical use.
In order to increase applicability and comparison between studies, future research should focus on establishing uniform guidelines for estimating the performance of POCT coagulometers.
Regarding internal quality control, one has to differentiate between electronic keys supplied with some POCT coagulometers, and liquid or lyophilized internal quality control samples available from the manufacturer to which both a target value and a confidence interval are assigned. The first estimates the functionality of the individual/single device, but does not test the accuracy . Therefore, as a single quality control method, it is not sufficient.
External quality control for checking the accuracy has been recommended by some groups [75,90,91], but has not yet been widely adopted. The different methods have some drawbacks. Method 1 (the split-sample method)  is dependent on the quality of the INR measured in the laboratory, and does not account optimally for imprecision or accuracy . Method 2 has the same drawbacks as method 1. Method 3 is the method developed by the UK NEQAS, and does estimate imprecision by using two samples, but does not estimate inaccuracy, as the deviation is based on a deviation from an overall performance. Furthermore, as the reporting of results is performed centrally, there is a time delay. The method proposed by the EAA (method 4) is, from a theoretical point of view, superior, as it takes imprecision and inaccuracy into account [14,93], and the result is instantly available to the patient, so there is no time delay. However, it has the drawback of requiring a considerable logistical set-up and substantial economic resources. Furthermore, it can be criticized for using plasma, and the results obtained with this do not necessarily reflect the performance of the test result on the native sample (i.e. whole blood) .
The use of external quality control costs money and is voluntary, and this involves obvious problems in terms of selection and uniformity. Obviously, the quality control has to be independent of the manufacturers of the POCT coagulometers. There have been no studies proving a clinical impact of external quality control (e.g. a reduction in clinical complications [bleeding and thromboembolic events]). The recommendations are therefore primarily based on general assumptions from different groups (e.g. the EAA).
In Germany, where more than 100 000 patients perform PSM, there is no program for external quality control . The implementation of external quality control would require considerable resources, and the benefit would be doubtful. However, from a theoretical viewpoint, external quality control seems relevant [14,54,83,86]. All of the methods are potentially applicable, and each method has advantages and limitations. Hopefully, further trials will compare methods and evaluate the potential effect, optimally by using clinical endpoints.
Another option is a control system that could probably detect unforeseen incidents or irregularities in the test system or in the patient. If a patient suddenly changes in dose without notice, for example by more than 20%, this should always lead to a control visit and a parallel test with a laboratory-based INR. This will detect changes in either the patient or in the test system. However, it cannot completely be regarded as external quality control.
It has been suggested that calibration of the POCT coagulometers should and could be performed [4,78,79,96] in order to provide accurate INR measurements . Although possible, this does not seem feasible. It requires an ISI calibration for parallel conventional manual PT testing with the local PT test system (instrument–thromboplastin combination) and an International Reference Preparation on plasma from the same whole blood samples used in tests on the POCT coagulometer in order to comply with WHO guidelines . Calibration of each individual apparatus is therefore not a feasible option . Poller et al.  have consequently recommended calibration with lyophilized plasma for the manufacturers of the POCT coagulometers and for centers involved in regulatory controls or clinical trials. Leichsenring et al.  performed an ISI calibration on a master lot. This is an ‘overall’ calibration of the thromboplastin generally used in CoaguChek XS and not of the individual POCT coagulometers. Others have argued that CoaguChek, with its test strips and code chips, is a local PT system, and calibration can therefore be safely performed by the manufacturer, and no calibration should take place afterwards .
There are several limitations of this review. Some studies may be missing, especially because of the exclusion of studies not written in English. Furthermore, the definition of high-quality studies was performed according to our own, and not validated, criteria. However, the identification of high-quality studies enabled us to provide a more correct outcome assessment. We could also have applied the criterion that the measurements should be taken and analyzed by the patients themselves, in order to include this potential preanalytic variability, but this would have excluded most of the high-quality studies found.
In conclusion, the precision of the POCT coagulometers was generally adequate for clinical use. Their performance in terms of accuracy has to be viewed in the context of the inherent inaccuracies in INR measurements. The accuracy of POCT coagulometers seems, in this respect, to be generally acceptable, and they can be used in a clinical setting.