Oral anticoagulants are widely used for the treatment and prophylaxis of thromboembolic disorders (Hirsh et al, 1998). Although no accurate information is presently available on the number of patients treated with oral anticoagulants, a rough estimate made in Italy from the consumption of coumarin drugs put the figure at 300 000 in 1993, with an average increase of 10% per year (unpublished observation by the Italian Federation of Oral Anticoagulant Clinics). This figure brings the numbers of patients currently on oral anticoagulants up to 600 000, which represent roughly 1% of the Italian population. These numbers are probably similar in other European countries and North America, and may increase further because of the increasing numbers of elderly patients put on oral anticoagulants for atrial fibrillation (Connolly et al, 1991; Stroke Prevention in Atrial Fibrillation Investigators, 1994) and because of the recent suggestion of low-dose warfarin in the primary prevention of ischaemic heart disease (The Medical Research Council's General Practice Research Framework Investigators, 1998). This situation will probably create organizational problems in the near future and the way of dealing with clinical and laboratory monitoring might need to be reshaped. A shift from central laboratory monitoring, by means of conventional prothrombin time (PT) systems, to small peripheral units and/or patient self-testing might be the solution. Peripheral units could be equipped with near-patient testing devices run by nurses or patients themselves and linked via modem to a prescription centre equipped with software capable of automated prescription (Poller et al, 1993, 1998) and run by an expert physician. Essential to this organizational scheme is the provision of reliable, easy-to-handle devices for PT measurements. Currently, there are many commercial devices using thromboplastin-containing test cartridges for PT testing that might serve the purpose. This article will review the present state-of-the-art of such devices with respect to their suitability to fulfil the requirements set by the World Health Organization (WHO) for the expression of results with the universal scale [International Normalized ratio (INR)] to report results for the laboratory control of oral anticoagulant treatment (WHO Expert Committee on Biological Standardization, 1999).
There are several types of commercial devices and the way they work is often not disclosed by manufacturers. Therefore, a detailed description is beyond the scope of the present review. In general, they are small portable instruments able to detect clot formation upon addition of an unmeasured drop of native whole blood (citrated blood or plasma may be suitable for some of the devices) to a reaction chamber built in a small cartridge that incorporates freeze-dried thromboplastin reagent with or without calcium chloride. It is important to emphasize that some of the devices may require the presence of red blood cells in order to record clot formation. They cannot work with plasma. Addition of the test material within the reactive area of the cartridge prewarmed at 37°C starts the reaction by reconstituting the reagent. In some of the devices, the mixture of test material and thromboplastin flows through capillary channels and stops as the clot forms. The time elapsed from the beginning of the reaction is recorded and displayed by the monitor as plasma-equivalent clotting time. In other devices, paramagnetic iron oxide particles are mixed with the thromboplastin reagent within the reaction chamber. The addition of blood starts the reaction and paramagnetic particles are free to move within the reaction chamber as an electromagnet turns on and off. Movement stops as soon as the clot forms and this is recorded by the device as plasma-equivalent clotting time. Clotting times may be automatically converted into PT ratio (patient-to-normal) and eventually into the INR by means of the mean normal PT (MNPT) and the International Sensitivity Index (ISI) encoded in the test cartridge or in a batch-specific code chip.
Requirements for expression of results as inr
The prerequisite for a reliable measure of the INR is the reliability of the parameters needed for its calculation, such as the PT for patient blood, the ISI of the system and the MNPT of the population (WHO Expert Committee on Biological Standardization, 1999) (see equation below):
PT for patient blood
Currently available devices use native venous or capillary (finger stick) blood. Blood flow from finger stick, contamination of blood with tissues, haematocrit and platelets are important preanalytical variables that might affect PT results. In this respect, near-patient testing devices are more prone to artefacts than conventional systems because the finger stick method is more difficult to standardize than venepuncture. On the other hand, it should be realized that preanalytical variability may be also high for conventional plasma PT measurement if one considers the effect of different blood collection systems (Chantarangkul et al, 1998; Van den Besselaar et al, 1999, 2000) and the time/condition of specimen storage before testing (Adcock et al, 1998). These variables do not affect near-patient testing.
International Sensitivity Index (ISI)
The ISI of conventional systems meant to measure the INR must be determined by calibration against the appropriate international reference preparation (IRP) using fresh plasmas from healthy subjects and patients on oral anticoagulant therapy (WHO Expert Committee on Biological Standardization, 1999). To comply with WHO recommendations, the calibration should meet the following criteria. First, paired PTs obtained with the system under calibration and with the IRP should be linearly related when plotted on a double log scale. Second, the overall regression line should adequately represent normal and patient data points. To this end, it should be tested whether the line drawn through patients-only data points passes through the mean of normals (Tomenson, 1984). An additional criterion to comply with WHO recommendations deals with the precision of the calibration. The within-laboratory precision of the estimated slope, expressed as the coefficient of variation (CV), should be smaller than 3%. To date, no specific recommendations have been issued for the calibration of near-patient testing devices and manufacturers do not disclose the protocol they use for that purpose. Attempts have been made in our and other laboratories to calibrate near-patient testing devices (Tripodi et al, 1993, 1997; Kitchen & Preston, 1997). We employed the same protocol used for conventional systems (WHO Expert Committee on Biological Standardization, 1999). PT testing with the near-patient testing devices has been performed on venous blood. Suitable aliquots of this blood were also collected into evacuated blood collection tubes containing trisodium citrate and centrifuged to obtain platelet-poor plasma which was then used for PT testing with the IRP and the manual (tilt-tube) technique. We also used the criteria described above to judge whether the calibration of near-patient testing devices complied with the WHO recommendations. Results for the two devices under calibration (TAS, Cardiovascular Diagnostics, Durham NC, USA and Coagucheck, Roche, Mannheim, Germany) have shown that they had an ISI value close to that encoded on the test cartridge by the manufacturers and that the calibration was in agreement with the WHO requirements because (i) the relationship between paired log-PTs with the IRP and with the devices for patients on oral anticoagulants and for healthy subjects were linearly related, (ii) the regression line drawn through patients' data points passed through the mean of normals, and (iii) the within-laboratory precision of the calibration, expressed as the CV of the slope, was less than 3% (Figs 1 and 2).
Mean normal prothrombin time (MNPT)
According to the WHO guidelines, the MNPT needed to calculate the PT ratio (patient-to-normal clotting time) must be the geometric mean clotting time of the normal population (WHO Expert Committee on Biological Standardization, 1999). The geometric mean value calculated from the PTs of a series of at least 20 healthy subjects may be a good approximation of the MNPT. Surrogate MNPT obtained from frozen or lyophilized pooled normal plasmas has been proposed (D'Angelo et al, 1997). However, it should be emphasized that freezing or lyophilization may introduce unpredictable changes that may significantly affect the PT of these plasmas (Poller et al, 1999a). Therefore, they may be suitable only if the clotting time is related to the MNPT and their stability is acceptable (WHO Expert Committee on Biological Standardization, 1999). Usually, near-patient testing devices are provided with MNPTs encoded in the test cartridge or in a batch-specific code chip. However, it might happen that the value encoded is significantly different from the value recorded locally by testing blood samples from 20 or more healthy subjects (Kitchen & Preston, 1997; Tripodi et al, 1997) and this may result in inaccurate INR determinations. In some devices, to compensate for local variation, the operator is able to change the MNPT encoded by the manufacturer.
Reliability of near-patient testing devices
Overall, the reliability of near-patient testing devices depends on the reproducibility of measurement and on the accuracy of encoded calibration parameters.
Little information is available on the reproducibility of near-patient testing devices. This is usually evaluated by repeated measurements of the same sample run on the same combination of meter and test cartridge but it is difficult to apply to those near-patient testing devices that require native whole blood. In fact, it would be difficult, if not impossible, to perform repeated measurements for the same non-citrated blood sample over time. Alternatively, it is possible to assess the reproducibility of near-patient testing devices by testing two native blood samples from two different finger sticks on each patient of the same series. The reproducibility assessed with the above procedure for one commercial near-patient testing device was acceptable (median CV = 4·18%), but poorer than that of the conventional laboratory INR measurement (median CV < 1·5%) (Kaatz et al, 1995). However, a more appropriate comparison between the reproducibility of the two systems would require not only duplicate finger sticks for near-patient testing devices but also duplicate venous blood collections for conventional laboratory methods. Reproducibility of measurement of different meters of the same make using the same lot of cartridges is also an important issue, which is still to be evaluated.
The accuracy of the INR measured with near-patient testing devices depends essentially on the calibration (see above). Presently, the responsibility of calibration rests entirely with manufacturers because access to the software to change encoded parameters is not possible for the majority of commercial devices. Furthermore, the calibration procedure would be too complex for the average user. Easier calibration procedures for near-patient testing devices are currently under investigation by the European Concerted Action on Anticoagulation (ECAA) working group (Poller et al, 1999b). The conclusions of such studies may form the basis for the preparation of official guidelines for calibration of near-patient testing devices. For the time being, users of these devices should check the performance by comparing the INRs measured using the devices with those measured using accurate methods. By definition, the true INR of a given plasma sample should be the one measured with the primary IRP for thromboplastin, called 67/40 (WHO Expert Committee on Biological Standardization, 1977), coupled with the manual (tilt-tube) technique to detect clot formation, which may be collectively defined as the standard method. However, the IRP 67/40 was discontinued many years ago and replaced by other IRPs, which were calibrated against their predecessors. Therefore, the true INR is not known. For practical purposes, it can be assumed that the INR measured with one of the established IRPs for thromboplastin coupled with the manual (tilt-tube) technique is a good approximation of the true value. The INR measured with other conventional systems may also be considered as a good approximation of the true value only if they have been calibrated against an IRP.
The agreement between paired INR measurements (i.e. those obtained with the near-patient testing device and those obtained with the standard method) can be assessed by statistically or clinically relevant criteria.
Statistically relevant criteria are concerned with the correlation analysis of paired INR measurements and/or comparison of mean values. Although they are widely used it should be realized that these statistical evaluations, if used alone, are not very informative. For instance, two methods might be highly correlated (high correlation coefficient) even though their results are systematically biased. Furthermore, two methods meant to measure the same analyte show data points that are usually very close to the regression line. Therefore, it would be difficult to assess any disagreement between paired measurements. It might be more useful to plot the differences of paired measurements against the average value (Bland & Altman, 1986, 1995). This enables an assessment of systematic differences over the whole range of measured INRs.
Clinically relevant criteria rests on the assumption that the INR values measured for the same patient by two systems are in agreement, if using either INR does not result in changes of dose prescription. Requirements for agreement of paired INR measurements have been developed (Anderson et al, 1993; Douketis et al, 1998) and may form the basis for near-patient testing assessment. Being more closely related to decision making on dose prescription, the agreement based on clinically relevant criteria should be considered more meaningful than that based on statistically relevant criteria.
Clinical studies using near-patient testing devices
In principle, the reliability of near-patient testing devices may also be assessed in prospective studies using appropriate end-points i.e. the time in the therapeutic interval and the occurrence of haemorrhagic or thrombotic complications. These studies can provide direct evidence not only on the reliability of near-patient testing devices but also on the efficacy of patients self-testing or self-management. A number of such studies has been performed on relatively small series of selected patients over the last few years. All of them used the time spent in the therapeutic interval or the deviation of the INR from the individual INR target interval as end-points (Ansell et al, 1995; Massicotte et al, 1995; Hasenkam et al, 1997; Sawicki, 1999; Cosmi et al, 2000; Cromheecke et al, 2000; Watzke et al, 2000). None of them used clinical end-points such as the occurrence of haemorrhagic or thrombotic complications. Invariably, all studies found the management with near-patient testing devices better or at least as effective as the management run by specialists or general practitioners in combination with conventional laboratory control. Similar conclusions were also reached in those studies in which the comparison of effectiveness has been made versus specialized anticoagulant clinics (Cosmi et al, 2000; Cromheecke et al, 2000; Watzke et al, 2000).
Quality assurance for near-patient testing devices
Daily checks of performance and successful participation in external quality assessment programmes are essential prerequisites for any measuring system meant to be used in clinical decision making. For near-patient testing devices, daily checks of performance can be achieved using electronic and liquid control samples provided by the manufacturers. These ensure that the device is operating properly before actual testing is performed. Participation in external quality assessment programmes, which is quite easy for conventional laboratory systems, is more complex and far reaching for near-patient testing devices. Typically, such programmes require that unknown test samples (usually freeze-dried, citrated plasma) are sent by a central control authority out to the participants in the programme who are asked to measure the INR and return their results for statistical evaluation. The complication with near-patient testing devices is that some of them require fresh native whole blood specimens and these cannot be easily shipped to distant locations. In principle, lyophilized citrated control plasmas can be used instead of native blood provided that they are recalcified before measurement. However, recalcification may add additional variability to the measurement owing to variations in the concentration of calcium chloride and pipetting used by participants, and to the variable time length from recalcification to sample application, which may influence the final result. In addition, some instruments cannot use plasma as they require red cells for clot detection. All these problems should be carefully considered in planning external quality assessment programmes for near-patient testing devices.
Finally, non-pathology operators (i.e. nurses, patients, etc.) should be trained specifically in order to avoid misuse of near-patient testing devices. Training sessions and checks of performance should be preferably run under the responsibility of competent professionals working with the laboratory in which patients refer for oral anticoagulant management.
The use of near-patient testing devices in combination with computerized anticoagulant dosage (Fitzmaurice et al, 2000) is opening new venues in oral anticoagulant monitoring. Decentralization of laboratory control and patient self-testing may help clinicians to cope with the increasing numbers of patients. Health services and patients may spare economic resources (Taborski et al, 1999; Lafata et al, 2000) and time for laboratory control respectively. Finally, patient self-management (i.e. self-testing and self-prescription), already proposed (Ansell et al, 1989) and now implemented in some countries of the European Union (Sawicki, 1999; Taborski & Muller-Berghaus, 1999) may be also feasible on a larger scale (Cosmi et al, 2000; Cromheecke et al, 2000; Watzke et al, 2000). Self management should be implemented in close cooperation with oral anticoagulant clinics, not only for training, but also for counselling and supporting patients in case of poor regulation of anticoagulation, occurrence of side-effects, management of invasive procedures, etc.
Currently, there are many different commercial manufacturers who prepare and distribute their devices, but there are no official protocols to guide in the calibration of such devices and individual manufacturers tend to solve the problem in a different way. As a consequence, commercial devices cannot be used interchangeably and generalization about their performance is not possible. Therefore, international standardization is urgently needed. Until then, it is recommended that manufacturers calibrate their devices against WHO IRPs following the guidelines issued for conventional systems and the criteria mentioned above. Users should assess their accuracy against standard methods (i.e. against the INR determined with IRPs or conventional systems calibrated against IRPs) (van den Besselaar, 2000). To avoid misuse, it is recommended that competent professionals supervise training sessions and performance of near-patients testing devices. Finally performance checks using internal and external quality assessment schemes are highly recommended.