Ian Jennings, UK NEQAS for Blood Coagulation, Rutledge Mews, 3 Southbourne Road, Sheffield S10 2QN, UK. E-mail: email@example.com
Summary. Laboratory screening for lupus anticoagulant (LA) has been shown to be suboptimal in several studies. Guidelines have recently been published by an expert group for the British Committee for Standards in Haematology, in an attempt to standardize and improve screening procedures. The value of using screening tests conforming with these guidelines was investigated in a United Kingdom National External Quality Assessment Scheme (UK NEQAS) proficiency testing exercise. The correct diagnosis was achieved by 97% of laboratories for a LA-negative sample. However, 18·3% of centres reported a false-negative result for a sample from a LA-positive subject. A significantly higher proportion of centres that used methods conforming with the published guidelines achieved the correct diagnosis for this sample (P < 0·002, chi-square test). A wide variety of screening tests were used by laboratories in this study. Within-method agreement could be improved by the use of a common normal pooled plasma to determine ratios. However, between-method agreement was not improved by this procedure. We conclude that adoption of methods compliant with national guidelines may improve the diagnosis of LA. There is a need, however, for reference and standardization materials to ensure further improvement in the accuracy of LA methods.
The antiphospholipid syndrome (APS) is defined by the presence of antiphospholipid antibodies (APA) in association with thrombotic disease and/or recurrent, otherwise unexplained, pregnancy failure, and identification of APS is of diagnostic and prognostic importance. To ensure that APA are detected, it is necessary to us both solid-phase immunoassays (e.g. anticardiolipin) and coagulation tests for lupus anticoagulant (LA) (Greaves et al, 2000). There is evidence that the association of thrombosis is more closely linked to LA than to anticardiolipin, and therefore specific and sensitive assays for LA are of particular importance (Arnout, 2001). However, LA are heterogeneous in nature and previous studies have shown laboratory screening to be suboptimal (Roussi et al, 1996; Jennings et al, 1997) and identified inaccuracies in the detection of LA as a result.
We report here results from a United Kingdom National Quality Assessment Scheme (UK NEQAS) survey in which samples with and without LA were distributed for investigation by routine hospital laboratories. The accuracy of diagnosis was determined and the value of using methods conforming to recently published guidelines by the relevant expert group [the British Committee for Standards in Haematology (BCSH) (Greaves et al, 2000)] was explored. The use of a common normal pooled plasma (NPP) in the standardization of the dilute Russell's Viper Venom Time (DRVVT) was also evaluated.
Materials and methods
All laboratories that were registered with the quality assurance scheme (n = 238) received two UK NEQAS LA quality assessment exercise plasmas, one classified as positive and one negative by the Scheme laboratory using national guidelines. One plasma was obtained by plasmapheresis of a patient with a history of venous thromboembolic disease and a previously identified persistent LA, detected by DRVVT and Kaolin clotting time (KCT); the second was prepared from a pool of plasmas from donations to the National Blood Authority. The donors involved are screened routinely for absence of disease and the plasmas tested negative for LA using the DRVVT. Plasma was collected into citrate phosphate dextrose (CPD), centrifuged twice (2000 g, 10 min) to ensure a platelet count below 10 × 109/l and buffered with 0·8% HEPES prior to lyophilization, as previously described (Kitchen et al, 1994). Vials (1 ml) of each sample were distributed to the 238 participants, who were asked to carry out their routine LA screening procedure and report their findings. Laboratories were offered the option of reporting positive LA screens as ‘weak’, ‘moderate’ or ‘strong’, as some DRVVT kits grade response in this way.
Data were analysed by specific test and reagent groups. The following reagents were used by 10 or more centres in this exercise: Dade-Behring Actin FS & Actin FSL activated partial thromboplastin time (APTT) reagents (Sysmex, Milton Keynes, UK); IL APTT lyophilized silica, IL APTT-SP liquid, IL LA Screen & LA Confirm, Hemoliance Synthasil APTT reagent [Instrumentation Laboratories (IL), Warrington, UK]; MDA Platelin LS (Organon Teknika, Cambridge, UK), Gradipore LA Screen, LA confirm and Kaoclot (Gradipore, North Ryde, Australia); DVV Test & Confirm (American Diagnostica, Greenwich, CT, USA); Manchester DRVVT kit (Thrombosis Reference Centre, Manchester, UK). Results for the LA-positive sample were re-analysed using the LA-negative sample results in place of the local normal pooled plasma data.
Results were obtained from 233 of the 238 registered centres in this exercise. The interpretations offered by the participants for the UK NEQAS LA-positive and LA-negative samples are shown in Table I. A total of 43 different combinations of LA screening methods, with and without mixing and phospholipid correction studies, were used. Table II shows interpretations provided by participants for the UK NEQAS LA-positive sample by combination of tests used; those shown in bold conform to the criteria in the BCSH guidelines for lupus anticoagulant testing (Greaves et al, 2000):
Table I. Laboratory interpretations for LA-positive and LA-negative samples distributed in this exercise.
Lupus screen negative
Table II. Interpretations related to testing strategy based on published guidelines.
Patterns of testing in lupus anticoagulant screening
Number of centres
Interpretations (% of users of each combination) UK NEQAS LA-positive sample
Bold text indicates a testing strategy compliant with published guidelines.
DRVVT + Confirm/phospholipid correction, + mixing studies for APTT/KCT
DRVVT + mixing studies + phospholipid correction of other tests
DRVVT + mixing studies, no confirm/phospholipid correction
DRVVT + Confirm/phospholipid correction, no mixing studies
DRVVT + APTT
APTT only or APTT + KCT
Overall, compliant with guidelines
Overall, not compliant with guidelines
1Prolongation of a phospholipid-dependent coagulation test
2Evidence of an inhibitor demonstrated by mixing studies
3Confirmation of the phospholipid-dependent nature of the inhibitor.
APTT were performed by 197 centres (85%). A total of 24 different APTT reagents were used by participants; results for the UK NEQAS plasmas obtained with reagents used by 10 or more centres are shown in Fig 1. Median APTT ratios ranged from 1·23 to 2·26 for the LA-positive sample.
The DRVVT was used by 228 (98%) centres. Nine different commercial kits were used, in addition to in-house methods. A variety of different test algorithms were also used by participants. Table III shows results obtained for the LA-positive sample with methods used by more than 10 centres. The following ratios and algorithms were calculated using local pooled normal plasma; DRVVT ratio (patient time/control time), test/confirm ratio (clotting time with screening reagent/time with confirm reagent), and percentage correction of ratio (% correction of clotting time for Manchester kit users). These algorithms were determined as previously described (Greaves et al, 2000). The normalized test/confirm ratio was determined as (DRVVT test/normal)/(confirm test/normal).
Table III. DRVVT performance: comparison of median results and within-method agreement for ratios determined, using normal plasma in local use and a common normal pooled plasma (see text).
NPP used by individual centres
For Manchester and in-house methods, ‘confirm’ includes concentrated phospholipid and platelet neutralizing procedures. For the Manchester method, the percentage correction of clotting time, widely used by users of this kit, has been shown.
P < 0·005.
P = 0·50.
P < 0·02.
No significant difference between medians was obtained with the two NPP sources for any other algorithm.
The test data for the LA-positive plasma were also recalculated against the values obtained by participants for the LA-negative plasma. Median DRVVT ratios for the LA-positive and -negative samples ranged from 1·18 to 1·34 and from 0·94 to 1·00 respectively. For the LA-positive samples, when test/normal DRVVT ratios were calculated for each laboratory using the normal plasma in local use, the overall coefficient of variation (CV) was 11·4%, compared with 11·2% when results obtained with the LA-negative plasma (common NPP) were used as the normal plasma value. However, using the common NPP values, within-method CVs were reduced, by an average 1% for test/normal ratios, 7·5% for normalized test/confirm ratios, 52% for percentage correction of ratio values and, for the Manchester reagent, 48·6% to 21·9% (see Table III).
The kaolin clotting time (KCT) was used by 79 centres (34%). Twenty-four centres used the Gradipore Kaoclot (colloidal kaolin) kit, on seven different coagulometers; the median test/normal and 1:4 test/normal ratios were 1·50 and 1·20, respectively, for the LA-positive sample. The median ratios obtained by 44 in-house method users (eight different coagulometers) were 1·32 and 1·02 respectively. A higher proportion (23·2%) of centres using in-house methods reported a negative LA screen for this sample compared with centres using the Kaoclot method (4·3%). Median test/normal ratios for the LA-negative sample were 0·8 for both Gradipore and in-house methods.
Recalculation of test/normal ratios with the LA-negative plasma (common NPP) value in place of the local normal plasma value resulted in only marginal improvement in between-centre agreement (Kaoclot users, CV 17·5% with local NPP, 15·4% with common NPP; in-house methods, CV 41·5% and 40·2% respectively).
Dilute prothrombin time (dPT)
Twelve centres used the dPT. Dilutions used ranged from 1/1 to 1/1000; seven centres using Dade-Behring Innovin reported test/normal dPT ratios ranging from 1·25 to 2·75 for the LA-positive sample; three centres using Manchester reagent, one using Diagnostic Reagents thromboplastin and one using an unspecified reagent reported test/normal dPT ratios ranging from 0·78 to 1·07. For the LA-negative sample, ratios ranged from 0·81 to 1·17.
Nine centres reported analysis using a combination of LA-sensitive and LA-insensitive APTT reagents; four centres used the Immuno LA kit, all reporting a positive LA screen for the LA-positive sample; three centres used the Stago Staclot LA kit, two of which reported a positive LA screen for the LA-positive sample.
In view of the high risk of recurrent thrombosis in subjects with primary APS and the apparent efficacy of antithrombotic prophylaxis in recurrent pregnancy failure (Backos et al, 1999), accurate identification of the presence or absence of LA is of potential therapeutic importance. Failure to recognize APS can result in anticoagulant therapy being inappropriately withheld. In contrast a false-positive test may lead to inadvisable exposure to anticoagulant therapy with its inevitable risk of bleeding. Suboptimal accuracy in screening for lupus anticoagulant has been reported previously (Jennings et al, 1997). Several guideline documents have been published in an attempt to improve accuracy (Machin et al, 1991; Brandt et al, 1995). They all recommend a similar approach, in which the identification of a LA is determined by prolongation of phospholipid-dependent coagulation tests, the inhibitor demonstrated by repeating the test after mixing the test plasma with normal plasma and phospholipid dependence of the inhibitor confirmed in order to exclude inhibitors to specific clotting factors. The most recent guidelines from the British Committee for Standards in Haematology was published in June 2000 (Greaves et al, 2000), following the demonstration of continuing suboptimal performance in the relevant NEQAS exercises (Jennings et al, 1997). The criteria for diagnosis of LA were reaffirmed by the authors, and the following recommendations are summarized: attention should be paid to preanalytical variables; a coagulation screen is used to exclude factor deficiency or anticoagulant effect; the APTT may be used in screening, but a sensitive reagent should be used; mixing tests with normal plasma and/or pooled normal plasma is informative; a second test, preferably KCT or DRVVT, should be used, together with a correction procedure; further tests may be of help for equivocal results; immunoassays (IgG, IgM anticardiolipin, β2 glycoprotein I) should be performed; presence of antiphospholipid antibodies should be confirmed with separate blood samples collected at least 6 weeks apart.
The proficiency testing exercise described here was carried out in November of the year in which these guidelines were published. The results highlight the enormous heterogeneity in reagents and methods used, and ongoing problems in the accuracy of testing for LA, but are suggestive of an improvement (Table II) in relation to previous surveys. This could be attributed to adherence to national guidelines, thus demonstrating the potential importance of such documents.
For the LA-negative sample, prepared from a pool of normal donors, 97·5% of centres reported a negative LA screen. Two centres reporting a moderate-positive LA were shown to have transposed the results for this and the LA-positive sample. Only one centre reported a weak LA. A previous UK NEQAS proficiency testing exercise identified a high proportion of false-positive results for a ‘complicated’ LA-negative sample, from a patient with factor IX coagulant deficiency. However, the specificity for this straightforward LA-negative sample was high.
For the LA-positive sample, 43 centres (18·5%) reported a negative screen. Previous studies have also shown a high proportion of false-negative interpretations for LA-positive samples (Jennings et al, 1997; Arnout et al, 1999). The interpretations were evaluated by comparison with the pattern of testing used by participants. The largest proportion (46%) of centres used both the APTT and DRVVT, with correction of the DRVVT with a high concentration phospholipid reagent, and mixing studies with either the APTT or KCT. Sixty-four per cent of centres used a combination of methods, complying with criteria published in the BCSH guidelines. Seventy-three per cent of these reported a positive LA screen, with 13% reporting a negative screen. Thirty-six per cent of centres used a combination of methods which did not fulfil the BCSH criteria. Among these, a significantly larger proportion of centres (32%) reported a negative screen (P < 0·002, chi-square test). These observations further imply that observance of recommended guidelines may improve diagnostic accuracy.
Of those centres reporting a positive LA screen, considerable variability in grading of results (weak, moderate or strong) was observed. The clinical significance of strength of LA positivity is not yet clear however (Le Querrec et al, 2001).
Among APTT reagents, variable sensitivity to the LA-positive sample was noted; median APTT ratios ranged from 1·23 for Dade-Behring Actin FS to 2·26 for Instrumentation Laboratories SP-liquid reagent. This variable sensitivity has been previously described (Arnout et al, 1999; Jacobsen et al, 2000) and is the basis for the LA screening test approach adopted by some laboratories (Brancaccio et al, 1997). However, it is important that centres are aware of the LA sensitivity of their APTT reagent; identification of possible LA-positive subjects during routine screening may be compromised with some reagents.
The DRVVT is now the most widely used test for LA screening, used by almost all centres; three fifths of centres use the DRVVT exclusively or with the APTT alone. Differences in DRVVT ratios obtained with different kits and different instrumentation have been reported (Mackie et al, 1998; Lawrie et al, 1999; Gardiner et al, 2000). In this exercise, the large number of reagent/instrument combinations used (39) precluded any valid statistical analysis of instrumentation effects. For the LA-positive sample in this exercise, median test/normal ratios obtained with kits used by more than 10 centres ranged from 1·18 for the Thrombosis Reference Centre DRVVT kit to 1·34 for the American Diagnostica kit. A higher proportion of users of the former (19%) reported a negative screen compared with users of the latter (11%). This difference in sensitivity was not observed in a single-centre study (data not shown) or a multicentre study (Mackie et al, 1998).
One of the possible sources of discrepancy between different methods is the source of NPP used in the assay. As a further exercise, data from participants were re-analysed using the DRVVT results obtained with a common NPP, the lupus-negative sample. For the majority of methods, median values were not significantly different from those obtained using locally prepared NPP, with the exception of the American Diagnostica kit, for which the test/normal ratio, normalized test/confirm ratio and percentage correction of ratio were all significantly higher when the common NPP was used. However the variations between different kits were still apparent, indicating they may be related to reagent rather than NPP source. The overall CV for test/normal ratios was practically identical when calculated using local normal plasma and common NPP values. However, within reagent/kit groups, the CV was reduced, by 7·5% for normalized test/confirm ratios and 52% for percentage correction of ratio values. The reduction was not uniform; with the Thrombosis Reference Centre kit, the CV for the test/normal ratio was higher when the common NPP was used; this kit includes a lyophilized NPP preparation, so it is likely that users already us a common NPP. However, improved CVs were seen with this kit for the normalized test/confirm ratio and percentage correction of clotting time algorithms. This analysis suggests that use of a common NPP within a reagent or method group may help improve between-laboratory agreement, but inherent differences between kits will not be reduced by this method. A further improvement may be obtained by the use of local and instrument-specific normal reference ranges. It was unclear, from the information returned in this survey, how many laboratories determined local cut-off values.
Differences in sensitivity were observed for KCT and dPT methods. The median KCT ratio obtained by centres using the Gradipore method was significantly greater than for in-house methods. Wide variation was seen between individual centres; test/normal ratios ranged from 1·23 to 2·16 for Gradipore kit users and 0·84–4·3 for in-house method users. CVs were only marginally improved when data were recalculated using the common NPP. Thus, a common method/reagent reduced variability. In the in-house tests, it is likely that some laboratories used a reagent that was not entirely suitable for their automate, e.g. poor stirring of kaolin, which sediments out rapidly, or a turbid reagent on a photo-optical device.
Only 12 centres used the dPT, so caution should be exercised in the interpretation of results. However, marked differences in sensitivity were seen between reagents; over a range of dilutions from 1/5 to 1/1000, raised test/normal ratios were obtained for the LA-positive plasma with Dade-Behring Innovin (ratios from 1·25 to 2·75). The sensitivity of this reagent to LA has previously been reported (Arnout et al, 1994). With Manchester and Diagnostic Reagents' thromboplastins, ratios between 0·78 and 1·07 were reported. If this test is to be used, careful selection of an appropriate reagent is required.
In conclusion, the results of this exercise confirmed previous observations on the extraordinary range of reagent and test combinations currently used by laboratories testing for LA, as well as an unacceptable rate of false-negative results of potential clinical significance. However, there was also an indication that conforming with national guidelines may improve diagnostic accuracy. Finally, there is a need for the development of readily accessible reference and standardization materials in order to facilitate essential improvements in the accuracy of LA test methods.