https://orcid.org/0000-0002-2103-1661
Abstract
Our objective was to maintain low interlaboratory variation and bias in international normalized ratio (INR) results following a network change in instrumentation and reagents, using a process of ongoing standardization and harmonization.
Network-wide standardization to new common instrument and reagent platforms followed by network-wide application of a simple novel process of verification of international sensitive index and mean normal prothrombin time values for each new lot of prothrombin time (PT) reagent that does not require use of World Health Organization reference thromboplastin or INR calibration/certified plasma.
The network transitioned from mechanical hemostasis detection instruments with associated PT reagent (Diagnostica Stago; NeoPTimal) to optical detection (ACL TOPs) with associated PT reagent (Werfen; RecombiPlasTin 2G). Comparing 3 years of data for each situation, the network (n = 27 laboratories) maintained low INR variability and bias relative to general mechanical and optical groups and other laboratories.
Harmonized support for patient management of vitamin K antagonists such as warfarin was continuously maintained in our geography, with potentially positive implications for other coagulation laboratories and geographies. For the United States in particular, paucity of US Food and Drug Administration–cleared INR certified plasmas potentially compromises INR test accuracy; our novel approach may provide workable alternatives for other laboratories/networks.
Warfarin therapy requires laboratory monitoring using the international normalized ratio (INR), but interlaboratory variation in reported INR values remains a concern.
We show continued low interlaboratory variation in reported INR values and low bias in a large 27-laboratory network compared with external quality assessment data reported median values as reference.
Low variation and bias were maintained even after a complete change of reagent and instrumentation from mechanical (Stago) to optical clot detection (Werfen).
INTRODUCTION
Anticoagulation therapy is applied for a variety of clinical reasons but primarily to treat or prevent thrombosis. Notably, oral vitamin K antagonist (VKA) anticoagulants such as warfarin have been used clinically for over 50 years.1 Although recent attention has transitioned to the so-called new, direct, or non-VKA oral anticoagulants (DOACs or NOACs), VKAs remain in clinical use for many patients for a variety of reasons.2,3 For example, VKAs appear to be safer and more efficacious for patients with mechanical valves or those who are triple positive for antiphospholipid antibodies.4-7 However, several shortcomings to VKA therapy, including high interpatient variability, and numerous drug and food interactions mandate continued need to monitor efficacy of therapy.2,3,8,9 Depending on the indication, patients are maintained within narrow therapeutic ranges of the laboratory test called the international normalized ratio (INR). The INR is actually derived from another routine coagulation test called the prothrombin time (PT), using mathematical adjustments (correction factors) according to both instrument and test reagent VKA therapy sensitivity, using the following formula: INR = (PT/MNPT)ISI, where the MNPT is the mean normal PT, and the ISI is the international sensitivity index.10
Maintaining patients within prescribed narrow therapeutic ranges (typically 2.0-3.0 or 2.5-3.5) remains an ongoing challenge for their clinicians. In published clinical trials for the DOACs, where it can be reasoned that patients are perhaps more closely monitored than in routine clinical practice, time within the therapeutic range (TTR) was reported as being 55% to 68%,11 with results outside these ranges potentially reflecting risk for bleeding (INR too high) or thrombosis (INR too low). High INR variability can also represent an additional risk factor (in addition to TTR) for clinically adverse events.12
The PT/INR testing can entail a large variety of methods, a large variety of coagulation analyzers and reagents, or near-patient point-of-care testing.13 Laboratory-based INR testing has been available for decades, with experienced laboratory personnel performing these tests and with test results being controlled by the use of internal quality control and external quality assessment (EQA); despite this, there exists large interlaboratory variation, as demonstrated by various EQA studies.14-17 Analytically, modern coagulation instrumentation has high precision, suggesting that the major reasons for continued high interlaboratory variation in INR testing could be due to differences in ISI and MNPT values assigned to various reagents as used by different laboratories. Different ISI and MNPT values are anticipated for different reagents, but depending on methodology used for local assignment of ISI and MNPT values, different values for these can also be applied to the same reagent/instruments by different laboratories.18 This high variability in assigned ISI and MNPT values infers that some might be incorrectly applied by a proportion of laboratories, and thus there is scope for improvement in assigned values to improve accuracy in INR testing.
The ISI values are often assigned by individual reagent/instrument manufacturers by using a World Health Organization (WHO)–recommended procedure and/or WHO-assigned thromboplastin reference plasmas.19-22 Nevertheless, it remains mandatory for laboratories to locally verify any manufacturer-assigned ISI.19 Where an ISI value is not available for a particular reagent/instrument combination (eg, if these derive from different manufacturers), the laboratory has to assign and then verify their local ISI value. Given the complexity of the WHO-recommended procedure for defining or verifying ISI values, requiring at least 60 VKA-stabilized patient samples, 20 normal samples, a WHO reference thromboplastin, and a manual tilt method for PT testing,20,21 INR testing laboratories cannot apply such an onerous process. Accordingly, the Clinical and Laboratory Standards Institute (CLSI) recommends use of commercial reference plasma calibration sets (alternatively called certified plasmas) for local ISI assignment.19 Still, this remains difficult in the United States since US Food and Drug Administration (FDA)–cleared material is currently limited to only 2 manufacturer products to our knowledge.23 Irrespective, wider use of generic certified plasmas may still produce different ISI and MNPT values when a laboratory uses different commercial products for the same reagent/instrument combination,24 leading to ambiguity regarding the correctness of any ISI value, as well as leading to continued interlaboratory variability in INR values. While a historical reduction of US-based interlaboratory INR variation was achieved following increased use of thromboplastin reagents with lower ISI values,14 this progress may be undone by continued application of inaccurate ISI and MNPT values.
The MNPT has to be typically locally defined, based on the assessed population, again using a WHO- and CLSI-recommended procedure (requiring 20 normal individuals), or with some certified plasma sets.10,19 Similar to ISI, evidence indicates the generation of different MNPT values using different methods, different commercial certified plasma products, and using different sets of normal individuals,24 leading to similar uncertainty regarding accuracy of any given MNPT value, as well as further promoting high interlaboratory INR test variation.
Finally, although fewer patients are now on VKAs, being increasingly migrated across to DOACs,25 and consequently leading to potentially decreasing INR testing, this does not change the requirements for laboratories to verify ISI and MNPT values. Essentially, whether a laboratory is performing 400 INRs per day (eg, such as the Westmead laboratory) or 20 INRs per day, laboratories still need to manage INR testing and attempt to ensure test accuracy.
The Westmead laboratory has previously reported a sequence of publications related to a novel approach to verification of local ISI and MNPT values, which does not require ongoing use of certified plasmas or WHO reference thromboplastin reagent.24,26-31 Using this approach, the Westmead-based laboratory has evidenced ongoing comparative INRs that very closely match our national EQA peer group median values for around 15 years.24,26-28 In our most recent published evaluation, we expanded this approach to a network of 27 laboratories; when combined with standardization of INR reagent and coagulation instrumentation, a substantive decrease in interlaboratory variation in INRs was achieved.28 In the current report, we again demonstrate that continued use of this approach over the past 7 years provides evidence of ongoing low interlaboratory coefficients of variation (CVs) and bias in these same laboratories, despite a transition to completely different instrumentation and reagents in all 27 laboratories.
MATERIALS AND METHODS
Setting
The immediate pathology network being reported in this study was previously named Pathology West28,29 and has since been split into separate functional units of West and Rural/Regional. Nevertheless, this identical group of 27 laboratories is spread across a wide geographic area of our state of New South Wales (NSW), Australia FIGURE 1. The Westmead-based site, located at the Institute of Clinical Pathology and Medical Research (ICPMR), services Westmead Hospital, one of the largest tertiary-level academic teaching hospitals in Australia, as well as provides all specialized hemostasis testing for Westmead Hospital and the entire 27-laboratory network. The 27-laboratory network is also a part of NSW Health Pathology, the largest publicly funded pathology health organization in Australia, currently comprising 65 laboratories. The growth and transition of our laboratory networks has been described in prior publications.28,29 In total, the NSW Health Pathology network comprises the West and Rural/Regional network of 27 laboratories, plus several other functional laboratory networks in different localities of NSW (ie, South, South West, East, Central, North).29
Setting for current study. The Institute of Clinical Pathology and Medical Research (ICPMR) laboratory, based at Westmead Hospital, was established in 1975 as a stand-alone laboratory to service the hemostasis pathology needs of Westmead Hospital. The ICPMR laboratory became part of New South Wales (NSW) Health Pathology when it was founded in 2012 and later became part of the Pathology West network of laboratories (n = 27) (still part of NSW Health Pathology). Subsequently, the Pathology West network was split into 2 functional units (named West and Rural/Regional) but together still comprised the same 27 laboratories of the former Pathology West. We report on interlaboratory variability in international normalized ratio (INR) testing across all 27 laboratories (35 instruments) during 3 years while using mechanical clot detection analyzers and reagents from Stago (2017-2019) compared to 3 years while using optical clot detection analyzers and reagents from Werfen (2021-2023) (see also TABLE 1).
Summary of Methodologies in Place for International Normalized Ratio Testing in New South Wales Health Pathology “West” and “Rural/Regional” Laboratories
| Characteristic | Instruments in use | PT reagents in use | Total different combinations (reagent/instrument) |
|---|---|---|---|
| Prechange over (2017-2019) | Stago: STA-Satellite ×23; STA-Compact Max ×10; STA-R Evolution ×2 | Stago: Neoplastine CI Plus, then NeoPTimal, ×27 laboratories/35 instruments | Instrument manufacturers: n = 1 (Stago) Different instrument types: n = 3 (all Stago) Total instruments: n = 35 Different reagent types: n = 1 (all Neoplastine CI Plus then NeoPTimal) Different instrument/reagent combinations: n = 3 |
| Postchange over (2021-2023) | Werfen: ACL TOP CTS 350 × 29; ACL TOP CTS 550 × 4; ACLTOP 750 CTS ×2 | Werfen: HemosIL RecombiPlasTin 2G ×27 laboratories/35 instruments | Instrument manufacturers: n = 1 (Werfen) Different instrument types: n = 3 (all Werfen) Total instruments: n = 35 Different reagent types: n = 1 (all HemosIL RecombiPlasTin 2G) Different instrument/reagent combinations: n = 3 |
| Characteristic | Instruments in use | PT reagents in use | Total different combinations (reagent/instrument) |
|---|---|---|---|
| Prechange over (2017-2019) | Stago: STA-Satellite ×23; STA-Compact Max ×10; STA-R Evolution ×2 | Stago: Neoplastine CI Plus, then NeoPTimal, ×27 laboratories/35 instruments | Instrument manufacturers: n = 1 (Stago) Different instrument types: n = 3 (all Stago) Total instruments: n = 35 Different reagent types: n = 1 (all Neoplastine CI Plus then NeoPTimal) Different instrument/reagent combinations: n = 3 |
| Postchange over (2021-2023) | Werfen: ACL TOP CTS 350 × 29; ACL TOP CTS 550 × 4; ACLTOP 750 CTS ×2 | Werfen: HemosIL RecombiPlasTin 2G ×27 laboratories/35 instruments | Instrument manufacturers: n = 1 (Werfen) Different instrument types: n = 3 (all Werfen) Total instruments: n = 35 Different reagent types: n = 1 (all HemosIL RecombiPlasTin 2G) Different instrument/reagent combinations: n = 3 |
PT, prothrombin time.
Summary of Methodologies in Place for International Normalized Ratio Testing in New South Wales Health Pathology “West” and “Rural/Regional” Laboratories
| Characteristic | Instruments in use | PT reagents in use | Total different combinations (reagent/instrument) |
|---|---|---|---|
| Prechange over (2017-2019) | Stago: STA-Satellite ×23; STA-Compact Max ×10; STA-R Evolution ×2 | Stago: Neoplastine CI Plus, then NeoPTimal, ×27 laboratories/35 instruments | Instrument manufacturers: n = 1 (Stago) Different instrument types: n = 3 (all Stago) Total instruments: n = 35 Different reagent types: n = 1 (all Neoplastine CI Plus then NeoPTimal) Different instrument/reagent combinations: n = 3 |
| Postchange over (2021-2023) | Werfen: ACL TOP CTS 350 × 29; ACL TOP CTS 550 × 4; ACLTOP 750 CTS ×2 | Werfen: HemosIL RecombiPlasTin 2G ×27 laboratories/35 instruments | Instrument manufacturers: n = 1 (Werfen) Different instrument types: n = 3 (all Werfen) Total instruments: n = 35 Different reagent types: n = 1 (all HemosIL RecombiPlasTin 2G) Different instrument/reagent combinations: n = 3 |
| Characteristic | Instruments in use | PT reagents in use | Total different combinations (reagent/instrument) |
|---|---|---|---|
| Prechange over (2017-2019) | Stago: STA-Satellite ×23; STA-Compact Max ×10; STA-R Evolution ×2 | Stago: Neoplastine CI Plus, then NeoPTimal, ×27 laboratories/35 instruments | Instrument manufacturers: n = 1 (Stago) Different instrument types: n = 3 (all Stago) Total instruments: n = 35 Different reagent types: n = 1 (all Neoplastine CI Plus then NeoPTimal) Different instrument/reagent combinations: n = 3 |
| Postchange over (2021-2023) | Werfen: ACL TOP CTS 350 × 29; ACL TOP CTS 550 × 4; ACLTOP 750 CTS ×2 | Werfen: HemosIL RecombiPlasTin 2G ×27 laboratories/35 instruments | Instrument manufacturers: n = 1 (Werfen) Different instrument types: n = 3 (all Werfen) Total instruments: n = 35 Different reagent types: n = 1 (all HemosIL RecombiPlasTin 2G) Different instrument/reagent combinations: n = 3 |
PT, prothrombin time.
NSW Health Pathology (ie, all 65 laboratories) recently transitioned from prior mechanical clot detection instrumentation and reagents (Diagnostica Stago S.A.S.) to optical clot detection instrumentation and reagents (Werfen), subsequent to a public tender.29,30 The pre- and postequipment and reagent combinations, as used by the 27-laboratory network (previously Pathology West, now West and Rural/Regional), is shown in TABLE 1. The new instrument installation and combined reagent/instrument evaluation and harmonization process was coordinated and finalized in the 27-laboratory network within 2020/early 2021, as previously published.30
The Royal College of Pathologists of Australasia Quality Assurance Program (RCPAQAP) Haematology provides EQA services within our geographic region for INR testing (as well as many other hemostasis analytes). Participants for the General Haemostasis program, which includes the laboratory INR test module, mainly comprise Australasian laboratories, although about a third of enrolled participants are from overseas countries. Participants (currently n = 658 laboratories, but this represents 880 reported data sets given some multiple instruments per laboratory) include 375 laboratories located in Australia, 175 in Malaysia, 32 in Hong Kong, and 27 in New Zealand. Participants currently employ at least 32 different instrument groups and 10 different reagents, with at least 58 different instrument/reagent combinations. In addition, 26 results are submitted for manual INR testing, with 3 different reagent combinations. Lyophilized plasma samples (n = 16/year, tested as 2 samples × 8 test periods/year) that cover the INR range of 1.0 to 5.0 are sent to participants for INR testing, with results currently submitted online and then analyzed using robust statistics. The median of the laboratory-specific reagent group is used for comparative statistical analysis, providing there are (1) greater than 10 users in the group (process used until 2021) or (2) at least 6 users (process used from 2022). Otherwise, results are compared with overall median values.
A Novel Approach to Estimation and Verification of ISI and MNPT Values
The Westmead-based ICPMR laboratory has previously reported on this novel approach to estimation and verification of ISI and MNPT values, a process that does not require any ongoing use of commercial (certified) plasma sets or WHO reference thromboplastin, over the past 16 years, in a series of chronological publications.24,26-28 Initiated at the Westmead laboratory,24 the process was subsequently expanded to a small number of other Sydney metropolitan laboratories to evaluate its effectiveness in a broader setting26,27 and then to the entire Pathology West Network of 27 laboratories, in order to harmonize INR processes and to reduce interlaboratory test variation in the group at that time.28 In brief, a process of comparative simple regression analysis is undertaken using the existing (but previously validated) thromboplastin/instrument combination, with previously validated MNPT and ISI values as the reference method and a proposed new (replacement) thromboplastin/instrument combination as the evaluation method.10,24,26-28,31,32 Using this procedure, the Westmead site has managed to maintain evidence of very close comparability of local INR values to RCPAQAP median INR values, reflecting a surrogate gold standard for accuracy for the past 16 years,24,26-28 despite initially using a disparate reagent manufacturer (Thromborel S; Siemens) and instrument (STAR analyzer; Diagnostica Stago) combination without any manufacturer-assigned ISI or MNPT values from 2008 to 201224 and eventually expanding this approach to the entire network of 27 laboratories when they all transitioned to Stago instruments and reagents.27 In the current report, we follow a subsequent transition in the same 27-laboratory network from Stago instruments and reagents to Werfen instruments and reagents, as partially reported earlier29,30 (see also TABLE 1).
Statistical and Comparative Analysis
The current report largely uses descriptive statistics such as medians and CV to assess variability. A 3-year period of Stago instrument/reagent usage (2017-2019 inclusive) is compared with an equivalent 3-year period of Werfen instrument/reagent usage (2021-2023 inclusive). The year 2020 is excluded from our analysis since this year marked the major period of transition from Stago to Werfen instrument/reagent usage in the network. Data were obtained from RCPAQAP EQA reports, initially from end-of-year (EOY) reports that summarized INR data for each sample tested in each separate year and including data on RCPAQAP median INRs, method median INRs, and participant-reported INR values. The RCPAQAP ceased providing EOY summary reports in 2022, and from these years, we accessed the last report in each year, which included a summary of past survey data. All these documents are archived for the 27-laboratory network in a network (Q-Pulse) quality system and thus available for our assessment. The data from individual network laboratory reports (n = 27 laboratories, for n = 35 instruments) were extracted for each year and then analyzed using Prism Software (GraphPad Software; version 10.2.1 for MacOS). For regression analysis, the simple linear regression tool from this software was used.
RESULTS
Some regression analysis data illustrating the process employed by the Westmead laboratory to initially estimate ISI and MNPT values for replacement reagents compared with existing reagents are shown in FIGURE 2 and FIGURE 3. For FIGURE 2, we show the evaluation of PT/INR data from the evaluation of ACL TOPs that we previously reported in 2021.30 In that prior publication, we identified that all the ACL TOPs, comprising 350, 550, and 750 models, could be considered a single family or class of analyzers. The Westmead laboratory coevaluated ACL TOPs 350 and 750 and Werfen reagent against the then existing Stago analyzer (STA-R Evolution) and Stago reagents TABLE 1. FIGURE 2 shows the regression analysis as applicable to the transition period, reflecting a comparison of the then current Stago analyzer (STA-R Evolution) and reagent (NeoPTimal) against the evaluation analyzers (ACL TOP 350 and 750) and reagent (HemosIL RecombiPlasTin 2G). As noted, the existing test system for INR was already validated according to the Westmead-led regression analysis procedure, with good ongoing performance in EQA FIGURE 4. Using the established Stago instrument/reagent system as reference, regression analysis can be used to estimate MNPT and ISI values for the evaluation instruments (ACL TOPs 350 and 750) and Werfen reagent (RecombiPlasTin 2G) FIGURE 2. For the ACL TOP 350, the estimated ISI for the evaluation PT reagent was 0.98, which compared well to the manufacturer-assigned ISI of 0.99 FIGURE 2A. The MNPT could be estimated as 11.9. The INR values generated with this data set on ACL TOP 350/RecombiPlasTin could be plotted against the Stago INR values (STAR/NeoPTimal) FIGURE 2B, and differences assessed by Bland-Altman analysis FIGURE 2C, and these data show little bias. Since it is preferred to use the manufacturer’s assigned ISI, a second analysis can be done holding the ISI at the manufacturer ISI (0.99 for this example) and then determining the optimal MNPT required to provide similar outcomes, which for this example was 12.2. Thus, using an ISI of 0.99 with MNPT of 12.2 yielded nearly overlapping outcomes to use of the estimated values of 0.98/11.9 FIGURE 2B-D. The second analysis with the Stago analyzer/NeoPTimal reagent and ACL TOP 750/RecombiPlasTin gave similar outcomes to that of the ACL TOP 350/RecombiPlasTin FIGURE 2E-H. In summary, the estimated ISI of 0.98 was identical and the estimated MNPT of 12.0 very similar (to 11.9).
Transition from Stago instruments and prothrombin time (PT) reagent to Werfen Instruments and PT reagent. This figure shows the application of the linear regression procedure to estimate mean normal prothrombin time (MNPT) and international sensitivity index (ISI) values across this changeover, using data generated at Westmead Hospital, being the lead laboratory of the West and Rural/Regional Network of New South Wales Health Pathology. For the ACL TOP 350, 162 samples sequentially received into the laboratory were coassessed with the Stago analyzer. For the ACL TOP 750, 190 samples sequentially received into the laboratory were coassessed with the Stago analyzer. There are no other specific inclusion or exclusion criteria for sample selection, as the intention is to generate equivalence to the current reference method using a variety of test samples. See the main text for further explanation of the process. A, Estimation of ISI and MNPT based on existing reagent—Westmead ACL TOP 350. B, Validation of ISI and MNPT for new reagent based on existing reagent—Westmead ACL TOP 350. C, Bland-Altman: international normalized ratio (INR) (new) Cal ISI 0.98/MNPT 11.9 vs INR current—Westmead ACL TOP 350. D, Bland-Altman: INR (new) IL ISI 0.99/calc. MNPT 12.2 vs INR current—Westmead ACL TOP 350. E, Estimation of ISI and MNPT based on existing reagent—Westmead ACL TOP 750. F, Validation of ISI and MNPT for new reagent based on existing reagent—Westmead ACL TOP 750. G, Bland-Altman: INR (new) Cal ISI 0.98/MNPT 11.9 vs INR current—Westmead ACL TOP 750. H, Bland-Altman: INR (new) IL ISI 0.99/calculated MNPT 12.0 vs INR current—Westmead ACL TOP 750.
The most recent example of the use of the linear regression analysis process to estimate/verify mean normal prothrombin time (MNPT) and international sensitivity index (ISI) values from the 27-laboratory network of West/Rural/Regional. For this evaluation, a total of 1,378 samples sequentially received into various laboratories were coassessed with the existing and replacement prothrombin time (PT) reagents (see also TABLE 3). There are no other specific inclusion or exclusion criteria for sample selection, as the intention is to generate equivalence to the current reference method using a variety of test samples. See the main text for further explanation of the process. A, PT current vs PT new—composite ACL TOPs. B, Estimation of ISI and MNPT based on existing reagent—composite ACL TOPs. C, Validation of ISI and MNPT for new reagent based on existing reagent—composite ACL TOPs. Bland-Altman: international normalized ratio (INR) new IL ISI 1.02/calc MNPT 11.4 vs INR current—composite ACL TOPs. E, Bland-Altman: INR (new) Cal ISI 1.00/MNPT 11.1 vs INR current—composite ACL TOPs.
Linear regression lines for all Royal College of Pathologists of Australasia Quality Assurance Program (RCPAQAP) participant data available for each year of data analysis for international normalized ratio (INR) testing by all methods: (A) 2017 (777 lines), (B) 2018 (777 lines), (C) 2019 (797 lines), (D) 2021 (860 lines), (E) 2022 (891 lines), and (F) 2023 (903 lines). Westmead-based Institute of Clinical Pathology and Medical Research (ICPMR) lab line shown as solid red in each figure. The line of identity (or equivalence) is shown in each figure as the black line from bottom left to top right of each figure. The RCPAQAP changed its analysis and report software in 2022, hence the different formats shown for 2022 and 2023. The green regression lines in E and F indicate lines from all participants with identical methodology to the ICPMR, being Werfen ACL TOP family (350, 550, or 750) and HemosIL RecombiPlasTin 2G reagent. Note that nearly a third of these green regression lines belong to participants deriving from New South Wales Health (NSW) Pathology laboratory sites (n = 65 laboratories with 84 instruments). None of the Pathology West/Rural/Regional network group showed any major deviations from the line of equivalence (see also FIGURE 5). The dotted green lines represent the RCPAQAP’s Allowable Performance Specifications (APS) and median based on the chosen method. The dotted black lines represent the overall APS and median. Occasional regression lines (indicated by blue arrows in each figure) bear little similarity to the majority of the lines or to the line of identity; these may reflect either poor selection of international sensitivity index (ISI) or mean normal prothrombin time, or large transcription or random errors reported during that year. The regression lines expand out over the INR range, being closer to the line of identity at INRs close to 1.0 and further from this line at high INRs, and we assert predominantly due to differences in ISI values used in each case.
Subsequent to implementation of the ACL TOPs and Werfen PT reagent, ongoing evaluation of reagent lot changes was implemented. FIGURE 3 summarizes the last such evaluation, which reflects our current practice. First, a large number of samples reflecting a wide range of INR values are tested using the existing PT reagent that has already had ISI and MNPT estimates previously validated by the above procedure and prior EQA performance (ie, this is used as the reference) and also tested with a new potential replacement reagent or reagent lot (this is the evaluation reagent). This testing is performed by several laboratory sites (n = 12 instruments in this example) and data initially analyzed separately to identify any discordance. Subsequently, the composite of comparative PT values is plotted to get an idea of potential bias FIGURE 3A across the network should the evaluation reagent be subsequently accepted for implementation; bias was not evident in this case comparing existing RecombiPlasTin reagent lot N0421610 with the planned replacement lot N0632741. Then, the log of the derived PT values for the evaluation reagent is plotted on the y-axis against the log of the reference reagent INR values on the x-axis FIGURE 3B. The ISI can be estimated from the data as the value corresponding to 1/slope, which for this example yielded a value of 1.00. The manufacturer-assigned ISI for this reagent lot was similar to our estimate at 1.02. The MNPT can also be estimated with this process and is equivalent to the antilog of the y-intercept at X = 0.0 (ie, equivalent to INR = 1.0) and corresponding to an MNPT of 11.1 for this example. As a third step, the INRs generated using the existing prevalidated reagent lot (and preexisting prevalidated ISI and MNPT values) are plotted against the INRs using the evaluation reagent as calculated using the ISI and MNPT values estimated in the second step (ie, as per FIGURE 3B, ISI = 1.00, MNPT = 11.1 FIGURE 3C). As shown, the regression lines essentially overlap, basically indicating equivalence with the reference INRs. Since there is a preference to use the manufacturer-assigned ISI value rather than the estimated ISI, given ease of application with barcoded reagents across a large laboratory and instrument network, a fourth step is taken, which uses the manufacturer-assigned ISI values and calculates the equivalent optimal MNPT that would permit maintenance of this equivalence to the reference. In this example, using manufacturer’s assigned ISI value of 1.02, application of an MNPT value of 11.4 generated a regression line that also essentially overlaps the reference regression line, as well as overlapping the line generated using the estimated ISI and MNPT values FIGURE 3C. Finally, Bland-Altman analysis can be used to show the differences in INR values between reference values vs those generated using the evaluation (new) reagent and the estimated ISI (1.00) and MNPT (11.1) FIGURE 3E, or the manufacturer-provided ISI (1.02) and the recalculated (optimal) MNPT value (11.4 in this example) FIGURE 3D, with these data confirming similar outcomes (ie, slight variation in INR values around the average difference of 0.0 indicating random noise, plus no evidence of systemic bias). In this example, the broader network of NSW Health Pathology using this new lot of PT reagent was recommended to adopt the manufacturer’s instrument-specific ISI value (1.02) and an MNPT value of 11.4.
FIGURE 4 shows a summary of overall participant-submitted data for RCPAQAP INR test results from all RCPAQAP participant laboratories for 2017 to 2019 and then 2021 to 2023, representing the years of data analysis for the current study. Data are shown as regression analysis of laboratory-submitted INR values vs RCPAQAP median INR results. These figures identify the laboratory-reported INR values for samples (n = 16/year) distributed by the RCPAQAP Haematology for each year vs the RCPAQAP group median values, which essentially are representative of the overall group performance (n = 777-880 depending on year), with these medians acting as overall surrogate reference or gold standard INR values.
Several issues can be emphasized. First, there is a sizable spread of laboratory-reported INR values around the line of identity (or equivalence), with the spread increasing as INR increases. This pattern is representative of variable ISI and/or MNPT effects and resultant INR calculations, which increases the relative INR as the ISI increases/MNPT decreases. The outcome is similar in each survey year. The Westmead-based ICPMR data are shown in each figure as a red line, indicating ongoing good concordance of ICPMR (main campus) INR results with median RCPAQAP values. Occasional participant lines (identified by blue arrows in FIGURE 4) show little resemblance to RCPAQAP medians values, and regression lines are quite disparate to other participants; these usually indicate the existence of large analytical or potentially typographical (transcription) error(s) during that year by that participant. Data lines that generally follow, but diverge from, the line of identity may designate too low or too high ISI and/or MNPT values relative to the overall group.
Low relative variability for the local network of 27 laboratories using mechanical clot detection instruments and reagent from Stago was reported in a previous 2016 study,28 with more recent data (2017-2019, but still using Stago instruments and reagent) now shown in FIGURE 5A-C. Data following the changeover to optical clot detection instruments and reagent from Werfen (2021-2023) are shown in FIGURE 5D-F. As presented, the entire network of 27 laboratories (for a total of 35 instruments) comprises a small group of reasonably tight regression lines (n = 35 in each figure), as compared to data from all participants (ie, FIGURE 4). The Westmead-based ICPMR data (n = 2 instruments in each figure) are identified by the red regression lines.
![Linear regression lines for each year of data analysis for international normalized ratio (INR) testing by New South Wales (NSW) Health Pathology (Pathology West, West, or Rural/Regional) network laboratory participants (27 laboratories; 35 regression lines in each case) of the Royal College of Pathologists of Australasia Quality Assurance Program: (A) 2017 Stago, (B) 2018 Stago, (C) 2019 Stago, (D) 2021 Werfen, (E) 2022 Werfen, and (F) 2023 Werfen. NSW Health Pathology network laboratory participants used Stago instruments (STA-R Evolution, STA-Compact Max, or STA-Satellite) and Neoplastine CI Plus then NeoPTimal PT reagent in 2017 to 2019 inclusive, as well as Werfen instruments (ACL TOP family [350, 550, or 750]) and HemosIL RecombiPlasTin 2G PT reagent in 2021 to 2023 inclusive. The line of identity (or equivalence) is shown in each figure as the black line from bottom left to top right of each figure. Westmead-based Institute of Clinical Pathology and Medical Research laboratory data (2 instruments, so 2 lines) shown as solid red lines in each figure. Occasional data (blue circles) reflect random errors (1 example in 2017, 3 examples in 2019) or transcription errors (reversal of data entry; 4 examples in 2021). The effect of a large random error on the regression line is highlighted with a blue arrow in C (2019).](https://oup.silverchair-cdn.com/oup/backfile/Content_public/Journal/ajcp/163/1/10.1093_ajcp_aqae090/1/m_aqae090_fig5.jpeg?Expires=1749334453&Signature=NtmMw0BsasVdwvieWTcE0xNw~Zm~7F3O-o9yTWSLQITNOCL5O6bjgsRx29ILLAyif8VgVqJ-RONSBBdmI15AnD8lf3gFL39lpfafkZvqKFrYT6L0i2mZEsCXT79U3hKMOpigo2XbroZbz8z8dnZRgmmfdA~h7vckIbpQzH~X1mVwmmmccNQW82QJ2c4sN-t5EwcBUOUl8GRrNb18g-AItuL2IkoWpTBjN4hYwsfIuqPFBaRSYzDZy849UvA~ejy4pdVHlEHceySi3G0h-2FYtxOlYtDyhPfNPQDfQILVQEYyGxomB9UnLKuU-8z4NtlPfxR8DgUN1ZQ3MgrrxOzp1g__&Key-Pair-Id=APKAIE5G5CRDK6RD3PGA)
Linear regression lines for each year of data analysis for international normalized ratio (INR) testing by New South Wales (NSW) Health Pathology (Pathology West, West, or Rural/Regional) network laboratory participants (27 laboratories; 35 regression lines in each case) of the Royal College of Pathologists of Australasia Quality Assurance Program: (A) 2017 Stago, (B) 2018 Stago, (C) 2019 Stago, (D) 2021 Werfen, (E) 2022 Werfen, and (F) 2023 Werfen. NSW Health Pathology network laboratory participants used Stago instruments (STA-R Evolution, STA-Compact Max, or STA-Satellite) and Neoplastine CI Plus then NeoPTimal PT reagent in 2017 to 2019 inclusive, as well as Werfen instruments (ACL TOP family [350, 550, or 750]) and HemosIL RecombiPlasTin 2G PT reagent in 2021 to 2023 inclusive. The line of identity (or equivalence) is shown in each figure as the black line from bottom left to top right of each figure. Westmead-based Institute of Clinical Pathology and Medical Research laboratory data (2 instruments, so 2 lines) shown as solid red lines in each figure. Occasional data (blue circles) reflect random errors (1 example in 2017, 3 examples in 2019) or transcription errors (reversal of data entry; 4 examples in 2021). The effect of a large random error on the regression line is highlighted with a blue arrow in C (2019).
Assessment of INR interlaboratory variability pre- and posttransition to Werfen instruments/reagents is shown in FIGURE 6, using interlaboratory CVs for each INR sample tested in the comparative periods, as compared with all method or all mechanical clot detection or all optical clot detection CVs. The 27-laboratory network-based CVs (generally <5%) were similar across all years, indicating that despite the entire 27-laboratory network transitioning to new instruments (n = 35) and reagents, a low relative interlaboratory variation was maintained. Network CVs (generally <5%) were similar across all INR values. In contrast, RCPAQAP interlaboratory all-method INR CVs ranged from 5.6% to 14.7% for mechanical clot detection, 2.0% to 13.7% for optical clot detection, and 5.2% to 14.6% for all methods over the period of analysis. Moreover, increasing CVs were observed for higher INR values for mechanical clot detection, optical clot detection, and all methods combined, consistent with expectations that this is where the greatest differences in INR would be expected due to ISI variation (and consistent with regression line variance in FIGURE 4).
Plots of interlaboratory international normalized ratio (INR) coefficients of variation (CVs) vs median Royal College of Pathologists of Australasia Quality Assurance Program (RCPAQAP) INR values. Data shown separately for years 2017 to 2019 Stago and 2021 to 2023 Werfen, and for all mechanical clot detection instruments (including Stago analyzers), all optical clot detection instruments (including Werfen analyzers), all methods, New South Wales (NSW) Health Pathology network laboratories (n = 27 for 35 instruments). Resulting linear regression lines for each data set also shown. There is near-horizontal linearity for NSW Health Pathology network laboratories, with CVs in general less than 5%, suggesting little variation in CVs according to INR value. In contrast, upward trajectories are generally evident for all methods or for all mechanical or optical test systems (ie, CVs range upward with increasing INR). The range of CVs was as follows: (1) all mechanical clot detection instruments years 2017-2019: 5.6% to 12.6% and for years 2020-2023: 5.6% to 14.7%; (2) all optical clot detection instruments years 2017-2019: 5.1% to 12.2% and for years 2020-2023: 2.0% to 13.7%; (3) all methods years 2017-2019: 5.1% to 12.2% and for years 2020-2023: 2.0% to 13.7%; (4) NSW Health Pathology Network (Pathology West, West, or Rural/Regional) laboratories years 2017-2019 (mechanical): 3.2% to 7.4% and for years 2020-2023 (optical): 0% to 5.9%. The 0% CV (arrow) was achieved on a sample reported by the network as yielding an INR of 1.0 by all laboratories/instruments submitting INR data.
TABLE 2 summarizes the outcome of the linear regression analysis for Werfen reagents evaluated during the past 3 years (2021-2023 inclusive). In each case, the final outcome of composite data is shown. In each case, the ISI estimated using the linear regression procedure (fourth column) was similar to the ISI assigned by the manufacturer (fifth column). The MNPT estimated by the regression procedure is shown in the sixth column. Due to a preference to use the manufacturer-assigned ISI, the analysis detailed previously (eg, FIGURES 2 and 3), and holding the ISI to that assigned by the manufacturer, we can calculate the optimal MNPT (seventh column) to maintain outcomes shown in FIGURE 4. Of further interest, one of our partner network sites based at John Hunter Hospital, as part of the additional evaluations taken in the NSW Health Pathology North Network, undertakes an MNPT estimation step using the Werfen IL calibration plasma. As shown in TABLE 2 (eighth column), these MNPT values were identical to the Westmead site’s independently determined optimal MNPT values using the linear regression procedure.
Summary of Linear Regression Analysis Performed Over 2021 to 2023 Inclusivea
| Year | Existing reagent (8 mL Recombiplastin) lot | Replacement (evaluation) reagent (8 mL Recombiplastin) lot | ISI estimated via linear regression procedure | Manufacturer-assigned ISI | MNPT estimated via linear regression procedure | MNPT implemented following regression procedure | MNPT estimated using Werfen IL calibration kit (JHH laboratory) |
|---|---|---|---|---|---|---|---|
| 2021 | N0202259 | N0815649 | 0.94 | 0.97 | 10.7 | 11.0 | 11.0 |
| 2022 | N0815649 | N0421610 | 1.01 | 0.99 | 11.1 | 10.9 | 10.9 |
| 2023 | N0421610 | N0632741 | 1.00 | 1.02 | 11.1 | 11.4 | 11.4 |
| Year | Existing reagent (8 mL Recombiplastin) lot | Replacement (evaluation) reagent (8 mL Recombiplastin) lot | ISI estimated via linear regression procedure | Manufacturer-assigned ISI | MNPT estimated via linear regression procedure | MNPT implemented following regression procedure | MNPT estimated using Werfen IL calibration kit (JHH laboratory) |
|---|---|---|---|---|---|---|---|
| 2021 | N0202259 | N0815649 | 0.94 | 0.97 | 10.7 | 11.0 | 11.0 |
| 2022 | N0815649 | N0421610 | 1.01 | 0.99 | 11.1 | 10.9 | 10.9 |
| 2023 | N0421610 | N0632741 | 1.00 | 1.02 | 11.1 | 11.4 | 11.4 |
ISI, international sensitivity index; JHH, John Hunter Hospital; MNPT, mean normal prothrombin time.
aData reflect the composite data set; see FIGURE 3 for the latest (2023) assessment as an example. JHH is part of the North Network of New South Wales (NSW) Health Pathology laboratories.
Summary of Linear Regression Analysis Performed Over 2021 to 2023 Inclusivea
| Year | Existing reagent (8 mL Recombiplastin) lot | Replacement (evaluation) reagent (8 mL Recombiplastin) lot | ISI estimated via linear regression procedure | Manufacturer-assigned ISI | MNPT estimated via linear regression procedure | MNPT implemented following regression procedure | MNPT estimated using Werfen IL calibration kit (JHH laboratory) |
|---|---|---|---|---|---|---|---|
| 2021 | N0202259 | N0815649 | 0.94 | 0.97 | 10.7 | 11.0 | 11.0 |
| 2022 | N0815649 | N0421610 | 1.01 | 0.99 | 11.1 | 10.9 | 10.9 |
| 2023 | N0421610 | N0632741 | 1.00 | 1.02 | 11.1 | 11.4 | 11.4 |
| Year | Existing reagent (8 mL Recombiplastin) lot | Replacement (evaluation) reagent (8 mL Recombiplastin) lot | ISI estimated via linear regression procedure | Manufacturer-assigned ISI | MNPT estimated via linear regression procedure | MNPT implemented following regression procedure | MNPT estimated using Werfen IL calibration kit (JHH laboratory) |
|---|---|---|---|---|---|---|---|
| 2021 | N0202259 | N0815649 | 0.94 | 0.97 | 10.7 | 11.0 | 11.0 |
| 2022 | N0815649 | N0421610 | 1.01 | 0.99 | 11.1 | 10.9 | 10.9 |
| 2023 | N0421610 | N0632741 | 1.00 | 1.02 | 11.1 | 11.4 | 11.4 |
ISI, international sensitivity index; JHH, John Hunter Hospital; MNPT, mean normal prothrombin time.
aData reflect the composite data set; see FIGURE 3 for the latest (2023) assessment as an example. JHH is part of the North Network of New South Wales (NSW) Health Pathology laboratories.
Finally, TABLE 3 shows further detail around the outcome of the last (2023) evaluation. As previously noted, the composite of data from several sites using different instrument types (ACL TOP 350, 550, 750) is used to generate composite data for implementation across the 27-laboratory network FIGURE 3. On this occasion, there was general concordance in ISI and MNPT values estimated using the linear regression procedure, and so no data were excluded. The pooled data were used to generate the final values for ISI and MNPT, which were then adopted by all laboratories using the 8-mL volume reagents, as summarized in TABLE 2.
Outcomes From Individual Sites Participating in the Last Evaluation Process (2023)
| Participating site (ACL TOP instrument) (network) | No. of samples | Part A (estimation phase) | Part B (application phase)a | ||
|---|---|---|---|---|---|
| ISI | MNPT | ISI | MNPT | ||
| Westmead (750) (West) | 221 | 0.99 | 11.1 | 1.02 | 11.4 |
| John Hunter Hospital (550) (North) | 241 | 1.01 | 11.2 | 1.02 | 11.4 |
| Royal North Shore Hospital (750A) (North) | 86 | 1.03 | 11.3 | 1.02 | 11.3 |
| Royal North Shore Hospital (750B) (North) | 106 | 1.00 | 11.1 | 1.02 | 11.3 |
| Royal North Shore Hospital (350) (North) | 136 | 1.00 | 11.1 | 1.02 | 11.4 |
| St George Hospital (550) (East) | 180 | 1.00 | 11.3 | 1.02 | 11.4 |
| Sutherland Hospital (350) (East) | 61 | 0.98 | 11.1 | 1.02 | 11.4 |
| Nepean (550) (West) | 58 | 0.99 | 11.2 | 1.02 | 11.4 |
| Blacktown (550) (West) | 73 | 0.99 | 10.9 | 1.02 | 11.4 |
| Dubbo (350) (Rural/Regional) | 49 | 0.99 | 11.2 | 1.02 | 11.4 |
| Orange (350) (Rural/Regional) | 107 | 0.98 | 11.0 | 1.02 | 11.4 |
| Wagga (350) (Rural/Regional) | 61 | 1.00 | 11.4 | 1.02 | 11.5 |
| Manufacturer | 1.02 | NA | 1.02 | NA | |
| Composite (all)b | 1,378 | 1.00 | 11.1 | 1.02 | 11.4 |
| Participating site (ACL TOP instrument) (network) | No. of samples | Part A (estimation phase) | Part B (application phase)a | ||
|---|---|---|---|---|---|
| ISI | MNPT | ISI | MNPT | ||
| Westmead (750) (West) | 221 | 0.99 | 11.1 | 1.02 | 11.4 |
| John Hunter Hospital (550) (North) | 241 | 1.01 | 11.2 | 1.02 | 11.4 |
| Royal North Shore Hospital (750A) (North) | 86 | 1.03 | 11.3 | 1.02 | 11.3 |
| Royal North Shore Hospital (750B) (North) | 106 | 1.00 | 11.1 | 1.02 | 11.3 |
| Royal North Shore Hospital (350) (North) | 136 | 1.00 | 11.1 | 1.02 | 11.4 |
| St George Hospital (550) (East) | 180 | 1.00 | 11.3 | 1.02 | 11.4 |
| Sutherland Hospital (350) (East) | 61 | 0.98 | 11.1 | 1.02 | 11.4 |
| Nepean (550) (West) | 58 | 0.99 | 11.2 | 1.02 | 11.4 |
| Blacktown (550) (West) | 73 | 0.99 | 10.9 | 1.02 | 11.4 |
| Dubbo (350) (Rural/Regional) | 49 | 0.99 | 11.2 | 1.02 | 11.4 |
| Orange (350) (Rural/Regional) | 107 | 0.98 | 11.0 | 1.02 | 11.4 |
| Wagga (350) (Rural/Regional) | 61 | 1.00 | 11.4 | 1.02 | 11.5 |
| Manufacturer | 1.02 | NA | 1.02 | NA | |
| Composite (all)b | 1,378 | 1.00 | 11.1 | 1.02 | 11.4 |
ISI, international sensitivity index; MNPT, mean normal prothrombin time; NA, not applicable.
aHolding ISI to the manufacturer-assigned ISI, then generation of optimal MNPT to apply to yield findings that permit continuation of optimized international normalized ratios, as evidenced by external quality assessment data.
bUsing composite data.
Outcomes From Individual Sites Participating in the Last Evaluation Process (2023)
| Participating site (ACL TOP instrument) (network) | No. of samples | Part A (estimation phase) | Part B (application phase)a | ||
|---|---|---|---|---|---|
| ISI | MNPT | ISI | MNPT | ||
| Westmead (750) (West) | 221 | 0.99 | 11.1 | 1.02 | 11.4 |
| John Hunter Hospital (550) (North) | 241 | 1.01 | 11.2 | 1.02 | 11.4 |
| Royal North Shore Hospital (750A) (North) | 86 | 1.03 | 11.3 | 1.02 | 11.3 |
| Royal North Shore Hospital (750B) (North) | 106 | 1.00 | 11.1 | 1.02 | 11.3 |
| Royal North Shore Hospital (350) (North) | 136 | 1.00 | 11.1 | 1.02 | 11.4 |
| St George Hospital (550) (East) | 180 | 1.00 | 11.3 | 1.02 | 11.4 |
| Sutherland Hospital (350) (East) | 61 | 0.98 | 11.1 | 1.02 | 11.4 |
| Nepean (550) (West) | 58 | 0.99 | 11.2 | 1.02 | 11.4 |
| Blacktown (550) (West) | 73 | 0.99 | 10.9 | 1.02 | 11.4 |
| Dubbo (350) (Rural/Regional) | 49 | 0.99 | 11.2 | 1.02 | 11.4 |
| Orange (350) (Rural/Regional) | 107 | 0.98 | 11.0 | 1.02 | 11.4 |
| Wagga (350) (Rural/Regional) | 61 | 1.00 | 11.4 | 1.02 | 11.5 |
| Manufacturer | 1.02 | NA | 1.02 | NA | |
| Composite (all)b | 1,378 | 1.00 | 11.1 | 1.02 | 11.4 |
| Participating site (ACL TOP instrument) (network) | No. of samples | Part A (estimation phase) | Part B (application phase)a | ||
|---|---|---|---|---|---|
| ISI | MNPT | ISI | MNPT | ||
| Westmead (750) (West) | 221 | 0.99 | 11.1 | 1.02 | 11.4 |
| John Hunter Hospital (550) (North) | 241 | 1.01 | 11.2 | 1.02 | 11.4 |
| Royal North Shore Hospital (750A) (North) | 86 | 1.03 | 11.3 | 1.02 | 11.3 |
| Royal North Shore Hospital (750B) (North) | 106 | 1.00 | 11.1 | 1.02 | 11.3 |
| Royal North Shore Hospital (350) (North) | 136 | 1.00 | 11.1 | 1.02 | 11.4 |
| St George Hospital (550) (East) | 180 | 1.00 | 11.3 | 1.02 | 11.4 |
| Sutherland Hospital (350) (East) | 61 | 0.98 | 11.1 | 1.02 | 11.4 |
| Nepean (550) (West) | 58 | 0.99 | 11.2 | 1.02 | 11.4 |
| Blacktown (550) (West) | 73 | 0.99 | 10.9 | 1.02 | 11.4 |
| Dubbo (350) (Rural/Regional) | 49 | 0.99 | 11.2 | 1.02 | 11.4 |
| Orange (350) (Rural/Regional) | 107 | 0.98 | 11.0 | 1.02 | 11.4 |
| Wagga (350) (Rural/Regional) | 61 | 1.00 | 11.4 | 1.02 | 11.5 |
| Manufacturer | 1.02 | NA | 1.02 | NA | |
| Composite (all)b | 1,378 | 1.00 | 11.1 | 1.02 | 11.4 |
ISI, international sensitivity index; MNPT, mean normal prothrombin time; NA, not applicable.
aHolding ISI to the manufacturer-assigned ISI, then generation of optimal MNPT to apply to yield findings that permit continuation of optimized international normalized ratios, as evidenced by external quality assessment data.
bUsing composite data.
DISCUSSION
In this report, we have shown continuance of low interlaboratory variation in our network of 27 laboratories (35 instruments), as well as low bias from EQA median values (representing a surrogate of reference or true INR values), even after a recent and complete change of instruments (from mechanical to optical clot detection) and a change of reagents (first from Stago Neoplastine CI Plus to NeoPTimal, then to HemosIL RecombiPlasTin 2G), subsequent to a government tender.30 Our 27-laboratory network covers nearly 85% of our state of NSW by accessible land mass FIGURE 1 and comprises a full spread of laboratory types, from supporting a tertiary-level academic teaching hospital (such as Westmead Hospital), to midsize regional centers, to small remote laboratories.29 Importantly, the process of MNPT/ISI estimation/verification does not require the use of a WHO thromboplastin reagent or any ongoing use of INR calibration/certified plasma sets. Once applied, the verified ISI and MNPT values are simply adopted by the different sites and used for their specific instrumentation. The process does require, however, a validated thromboplastin/instrument combination with prevalidated ISI and MNPT values. In our study, the originating reference thromboplastin/instrument combination (with prevalidated ISI and MNPT values) was historically and sequentially derived as previously reported,10,24 initially at the Westmead-based ICPMR laboratory using a combination of different manufacturer INR calibration/certified plasmas and a set of nearly 80 normal plasmas.24 Subsequently, the procedure was expanded to a small set of Sydney metropolitan laboratories within our local network26,27 and, in our last 2016 report, to the entire network of 27 laboratories,28 with this same set of 27 laboratories evaluated in the current study. One additional new improvement lies with the current use of a single ISI value and a single MNPT value across all instruments/laboratories in our network, be they ACL TOP 350, 550, or 750 analyzers. This is because we have already established these instruments as a single class of instruments, with no noted variation in test results between them.30 In contrast, prior usage28 as applied to the Stago analyzers was for potentially different ISI and/or MNPT values depending on which analyzer (STAR, Compact, or Satellite) was used by a given laboratory. This new improvement thus further facilitates more uniform harmonization and standardization across our large network. The process has since also been adopted for application within the majority of laboratories within the broader NSW Health Pathology Network.29 Although we do not feel it is necessary to perform calibration checks using calibration plasmas, since Werfen offers such material, it has been used regularly by one of our networks (North), and data shown in TABLE 3 indicate the equivalence on MNPT estimated using these to the MNPT values independently derived from the simple regression analysis.
In total, there are many areas of potential error in INR results, including preanalytical, analytical, and postanalytical issues.10,13,18,19,24,26-34 In a real-world testing, sample integrity is important for accurate test results, and thus preanalytical issues (sample collection and processing) are very important. In an EQA setting, the preanalytical environment is better controlled and ensured because of homogeneity and stability testing of individual proficiency test material, but it may remain a potential source of error. As lyophilized material is sent to laboratories, one relevant existing preanalytical issue will be sample reconstitution, and thus occasional random error events of higher or lower than expected INR could reflect a laboratory issue with sample reconstitution (eg, faulty pipet, poor quality water, water temperature, inappropriate technique). Analytically, hemostasis instruments, as used to provide INR test results, can in general be recognized as very robust, but an occasional analytical instrument or reagent issue also cannot be completely discounted. Nevertheless, evident significant random error events were noted to be very low in the current data analysis (only a few events for our network evident in FIGURE 5). Also, given samples are sent as pairs per EQA challenge, the possibility for transcription error exists, where participants reverse the paired sample results and incorrectly report these in the online portal. Again, a few examples are shown in FIGURE 5 (ie, but again, only a few such events for our network in 2021-2023). It is likely that the composite of these events led to an interlaboratory CV of ~5%, as evidenced for our network FIGURE 6.
Nevertheless, we propose that the main problem with laboratory INR reporting is systematic error associated with use of variable MNPT and ISI values in INR calculations, highlighting the divergence of regression lines in FIGURE 4. In a prior report, the RCPAQAP could also identify the varied use of different ISI and MNPT values by different laboratories even when using the same reagent/instrument combinations.18 Although the INR system was developed to reduce interlaboratory variability when monitoring VKA therapy compared with PT variation, which we agree has been achieved, substantial variation in INRs is still evident in 2024. The CVs for overall interlaboratory INR data from the RCPAQAP over the current period of analysis range from ~5% at INR 1.0 to ~15% at INR 4.0. This relationship reflects a combination of all the elements identified above, including the variety of methods used by manufacturers or laboratories for establishment and application of specific reagent ISI and MNPT values, as well as noncommutability of INR values across different reagent/instrument combinations. However, we judge that substantial elements of these variations continue to be the outcome of incorrect assignment of ISI and MNPT values by many laboratories. As noted in the Materials and Methods section, RCPAQAP INR participants currently comprise 658, with at least 58 different instrument/reagent combinations. These variations are likely to be much higher when considering global usage.
Being an Australian authorship team, the situation in the United States is less clear to us, although concern has been reported that variation in INR-reported values, once reduced due to increasing adoption of thromboplastin reagents with ISI values closer to 1.0,14 may again be increasing. Importantly, there are only 2 FDA-cleared certified plasma sets available in the United States to our knowledge, and these may only be validated for use for specific manufacturer/instruments.23 Therefore, US laboratories may need different strategies when they use other, or disparate, manufactured products. As an example, the US Mayo Network reported the strategy that they employed,33 but that process is different from that reported here. Their objective was to develop a process to verify/assign ISI and MNPT of a single thromboplastin reagent from 1 manufacturer across multiple instruments, including several from another manufacturer and across several campuses of a single institution, the Mayo Clinic. In this study, RecombiPlasTin 2G (R2G) was evaluated on the ACL TOP 700 (IL), STA-R Evolution, STA Compact, and STA Satellite. Random normal donor samples (n = 25) were used to verify/assign MNPT. A subset of the normal donors (n = 8) and 13 warfarin pools (INR range, 1.3-3.9), created from stable warfarin patient plasma, were used for ISI verification/assignment. The manufacturer’s assigned ISI was first verified on the ACL TOP 700 (reference method), then assigned on 3 unsupported instruments using orthogonal regression analysis. The MNPT and manufacturer-assigned ISI (11.0, 0.95) were verified on the ACL TOP 700 and subsequently assigned on the STA-R Evolution (11.6, 1.04), STA Compact (11.5, 1.02), and STA Satellite (10.9, 0.99). Linear correlations of the INR results from all the 4 instruments demonstrated an r2 > 0.99. The authors concluded that the process provided a reproducible approach to assigning ISIs on unsupported reagent/instrument combinations. Their data also confirmed that ISIs of the same PT reagent differ significantly on different instruments, thus confirming the requirement for evaluations and validation of ISIs for different reagent/instrument combinations. It is not known if this network continues this process in 2024.
Although the use of VKA therapy is diminishing, this does not mean that laboratories no longer need to worry about INR testing. In our network, since we service mainly hospital sites, the number of routine coagulation tests performed has not substantially fallen over the past 10 to 15 years. Nevertheless, even if performance of INR tests for VKA monitoring is generally falling, any laboratory required to provide an INR test service will still be required to provide that service, irrespective of INR test numbers. In other words, although the volume of INR testing is likely to have fallen after the emergence of DOACs, the number of laboratories providing INR testing has likely not fallen, as highlighted, for example, in our geography using RCPAQAP data FIGURE 4, which actually shows an increasing participation rate over the years. All laboratories providing INR testing for VKA therapy will still need to estimate/verify ISI and MNPT values for some time to come, and we herein provide an option that continues to evidence the accuracy of our ISI and MNPT determinations, using EQA median data as reference or truth, now for over 15 years.
CONCLUSIONS
We report continued maintenance of low interlaboratory variation and low bias from RCPAQAP median values, representative of a surrogate reference or true INR, in our network of 27 laboratories, for 35 instruments, even subsequent to a complete changeover in instrumentation/reagents and due to continued application of a novel procedure for ISI and MNPT validation. This low interlaboratory variation in INR reporting should translate to low variability of reported INR values in patients on VKA therapy, as potentially tested at different sites in our network, either because they may be traveling or else should samples be transferred between facilities. We believe that this continues to provide better clinical utility for INR testing for our clients. Finally, these outcomes may also have implications with respect to improvements in maintaining patients within TTR, as well as reducing patient variability in INR-reported values, which in turn may improve patient outcomes.12
Acknowledgments
We thank the RCPAQAP Haematology for their general collaborative participation, as well as for making available some data reported in this article via their online portal. We also thank various staff within our NSW Health Pathology Network who may have contributed technical assistance with testing of INR samples during reagent/instrument verification exercises as ultimately leading to the findings reported in this article and as shown, for example, in FIGURES 2 and 3. The opinions expressed in this article are those of the authors and not necessarily those of NSW Health Pathology, the RCPAQAP, or other institutions to which we are affiliated.
Conflict of interest disclosure: The authors have nothing to disclose.
