Diagnostic accuracy study of a factor VIII ELISA for ...

7 downloads 0 Views 145KB Size Report
... Gary W. Moore2; Sean Platton3; James C. Maloney2; Ben Palmer4; Louise Bowles3; K. John Pasi1; ...... Verbruggen B, van HW, Novakova I, Lillicrap D, et al.
1

New Technologies, Diagnostic Tools and Drugs

Diagnostic accuracy study of a factor VIII ELISA for detection of factor VIII antibodies in congenital and acquired haemophilia A Paul Batty1; Gary W. Moore2; Sean Platton3; James C. Maloney2; Ben Palmer4; Louise Bowles3; K. John Pasi1; Savita Rangarajan5; Daniel P. Hart1 1The

Royal London Hospital Haemophilia Centre, Barts and The London School of Medicine & Dentistry, QMUL, London, UK; 2Haemostasis and Thrombosis, Viapath Guy’s and St Thomas’ NHS Foundation Trust, London, UK; 3The Royal London Hospital Haemophilia Centre, Barts Health NHS Trust, London, UK; 4The United Kingdom National Haemophilia Database, Manchester, United Kingdom, 5Centre for Haemostasis and Thrombosis, Guy’s and St Thomas’ NHS Foundation Trust, London, UK

Summary Antibody formation to factor VIII (FVIII) remains the greatest clinical and diagnostic challenge to the haemophilia-treating physician. Current guidance for testing for inhibitory FVIII antibodies (inhibitors) recommends the functional Nijmegen-Bethesda assay (NBA). A FVIII ELISA offers a complementary, immunological approach for FVIII antibody testing. It was the aim of this study to retrospectively evaluate the performance of a FVIII ELISA (index) for detection of FVIII antibodies, compared with the NBA (reference). All samples sent for routine FVIII antibody testing at two haemophilia Comprehensive Care Centres, were tested in parallel using the NBA and a solid-phase, indirect FVIII ELISA kit (Immucor). A total of 497 samples from 239 patients (severe haemophilia A=140, non-severe haemophilia A=85, acquired haemophilia A=14) were available for analysis. Sixty-three samples tested positive by the NBA (prevalence 12.7 %, 95 % confiCorrespondence to: Dr. Daniel Hart Senior Clinical Lecturer and Honorary Consultant in Haemostasis The Royal London Hospital Haemophilia Centre The Royal London Hospital, London, E1 1BB, UK Tel.: +44 203 594 1869, Fax: +44 203 594 1859 E-mail: [email protected]

Introduction Antibodies to factor VIII (FVIII) are currently one of the greatest clinical and diagnostic challenges faced by the haemophilia-treating physician. These antibodies may develop in patients with congenital haemophilia A in response to FVIII replacement therapy or rarely as an autoimmune phenomenon (acquired haemophilia A). In patients with severe haemophilia A, laboratories are required to provide urgent inhibitor surveillance assay results during early treatment of previously untreated patients (PUPs) (1, 2) or where there is impaired treatment efficacy or pharmacokinetic profile following infusion of a replacement FVIII concentrate. Anti-FVIII antibodies may, however, arise at any stage of life in a person with any severity of haemophilia A (3–5). Similarly, there is urgency for diagnostic assays to confirm the presence of auto-immune, inhibitory activity in acquired haemophilia A (AHA), which is then followed by more routine treatment surveillance assays (6, 7). Aside from these notable exceptions, which comprise a minority of tests, most anti-FVIII antibody assays performed within the com-

dence interval [CI], 9.9–15.9 %), with a median inhibitor titre of 1.2 BU/ml (range 0.7–978.0). The FVIII ELISA demonstrated a specificity of 94.0 % (95 %CI, 91.3–96.0), sensitivity of 77.8 % (95 %CI, 65.5–87.3), negative predictive value of 96.7 % (95 %CI, 94.5–98.2), positive predictive value 65.3 % (95 %CI, 53.5–76.0), negative likelihood ratio 0.2 (95 %CI, 0.1–0.4), positive likelihood ratio 13.0 (95 %CI, 8.7–19.3) and a diagnostic odds ratio of 54.9 (95 %CI, 27.0–112.0). Strong positive correlation (r=0.77, p< 0.001) was seen between the results of the NBA (log adjusted) and FVIII ELISA optical density. In conclusion, FVIII ELISA offers a simple, specific, surveillance method enabling batch testing of non-urgent samples for the presence of FVIII antibodies.

Keywords Factor VIII inhibitors, haemophilia A, ELISA, Bethesda, sensitivity Financial support: Paul Batty has received an unrestricted research grant from Octapharma. Received: December 19, 2014 Accepted after major revision: May 19, 2015 Epub ahead of print: June 11, 2015 http://dx.doi.org/TH14-12-1062 Thromb Haemost 2015; 114: ■■■■

prehensive care haemostasis laboratory are performed for routine surveillance or screening. The optimal test for this purpose would be a test that could be performed simply which is specific with a high negative predictive value. The most widely used assay for detection of antibodies to FVIII is the Bethesda assay (8, 9). This assay and subsequent modifications (i. e. Nijmegen modification) are functional assays that provide quantitative assessment of antibodies’ inhibitory capacity to FVIII (8, 9). The Bethesda assay was initially proposed to help standardise diagnostic practice in the description of inhibitory antibodies in patients with congenital severe haemophilia A to facilitate inter-centre and study comparison (8). This assay is associated with significant inter-laboratory variability (10–12) and affected by numerous pre-analytical variables (13, 14). It is also limited in the detection of FVIII antibodies with weak or no direct inhibitory capacity. Despite these limitations the NBA is considered the “gold standard” in the diagnosis of FVIII antibodies (2). A number of “immunological” assays have been described for the diagnosis of FVIII antibodies in research settings, including

© Schattauer 2015

Thrombosis and Haemostasis 114.4/2015 Downloaded from www.thrombosis-online.com on 2015-06-14 | ID: 1001077309 | IP: 138.37.211.113 Note: Uncorrected proof, epub ahead of print online For personal or educational use only. No other uses without permission. All rights reserved.

2

Batty et al. Diagnostic test evaluation of FVIII ELISA

enzyme linked immunosorbent assay (ELISA) (15–20), fluorescence based immunoassays (21–23) and immunoprecipitation (24, 25). These assays detect antibodies to FVIII with both inhibitory and non-inhibitory (non-neutralising) capacity. Non-neutralising antibodies have been reported in approximately 18 % of patients with congenital haemophilia (23), although their clinical significance remains unclear (20). A commercially available, solidphase FVIII ELISA kit has been previously described (15, 19). This offers a simple approach to detection of FVIII antibodies and is potentially less affected by pre-analytical variables affecting functional inhibitor assays. We reviewed our experience of usage of this assay following its introduction as part of routine laboratory care in two haemophilia Comprehensive Care Centres.

Materials and methods Sample selection and inclusion criteria All samples sent for routine FVIII inhibitor testing from patients with congenital and acquired haemophilia A were tested in parallel using the Nijmegen-Bethesda assay (NBA) and a FVIII ELISA as part of routine clinical care. Samples were only tested in the local treatment centre laboratory and not repeated by the other centre. At centre 1 (The Royal London Hospital, London, UK), samples were tested between 9/9/2010–23/4/2012 and at centre 2 (St Thomas’ Hospital, London, UK) between 11/12/2009–28/09/2011. Testing was performed within the specialist haemostasis laboratories and the results of both assays were available to the treating physician, with neither laboratory nor clinical staff being blinded to results of these tests. An assessment of the performance of the ELISA was made based on the results of anonymised samples. A sub-group analysis was performed based on diagnostic labels provided at the time of venepuncture. No additional demographic or clinical data were retrieved to maintain patient anonymity. Samples were included if they had a diagnostic header of either congenital (severe or nonsevere) or acquired haemophilia A at the time of testing and where results from both assays were available. Samples were excluded if the diagnostic header indicated a non-haemophilia A (congenital or acquired) diagnosis or if one (or more) of the assays had not been performed. Severe haemophilia was defined as a baseline FVIII:C < 1 IU/dl and non-severe haemophilia 1–40 IU/dl (26). Acquired haemophilia A was defined as the presence of bleeding (acute or recent onset) with an unexplained prolongation of the activated partial thromboplastin time (aPTT) that did not correct on mixing studies and associated with low FVIII:C (6, 7). Local institutional approval was granted at both centres for the evaluation of the efficacy of the introduction of the ELISA in routine laboratory practice.

Inhibitor testing – Nijmegen Bethesda Assay (reference test) Inhibitor testing was performed within the specialist haemostasis laboratories of both centres. Both centres used either the

Nijmegen-Bethesda assay or modifications to this as previously described (9, 27, 28). Briefly, FVIII deficient and patient plasma were incubated at 37 °C for 2 hours in equal volumes of control plasma. Residual FVIII coagulant activity (FVIII:C) was then measured with 1 Bethesda Unit/ml (BU/ml) representing inhibitory activity resulting in 50 % residual FVIII:C (8). Centre 1 used the Nijmegen-Bethesda assay with a one-stage FVIII:C measurement. Centre 2 used a modified Nijmegen-Bethesda assay, with substitution of 4 % bovine serum albumin (BSA) for FVIII deficient plasma (27) and a chromogenic FVIII:C (28). The cut-off for positivity was defined locally as ≥ 0.6 BU/ml for centre 1 based on consensus recommendations (29). The cut-off for centre 2 was ≥ 0.7 BU/ml, defined in-house using dilution studies of a commercially available lyophilised FVIII inhibitor plasma (30). Antibodies were defined as being of low titre if the NBA was < 5 BU/ml and high titre if the NBA was ≥ 5 BU/ml (26).

Factor VIII indirect ELISA (index test) A commercially available solid phase indirect FVIII ELISA (Immucor, Norcross, GA, USA) was used as per the manufacturer’s instruction. This kit contains 96 microwells pre-coated with recombinant full-length recombinant H1 haplotype FVIII (rFLFVIII) (Kogenate®, Bayer Corp., Pittsburg, PA, USA) (31, 32). Briefly, samples and controls diluted 1:4 with diluent buffer were added to the microwells and incubated at 37 °C for 30–35 minutes (min). Following washing, 15 µl conjugate (alkaline phosphatase conjugated goat antibody to human IgG, 0.1 % sodium azide diluted 1:100 in diluent) was added to all wells and incubated at 37 °C for 30–35 min. Following washing, 50 µl substrate (P-nitrophenyl phosphate, reconstituted in de-ionised water and diluted 1:100) was added to all wells and incubated in the dark at room temperature (22–25 °C) for 30–35 min. The reaction was stopped using 50 µl Stopping Solution (3M Sodium Hydroxide) and the absorbance (optical density [OD]) was read at 405 or 410 nm using a GEN5 plate reader (BioTek, Winooski, VT, USA). This kit has three controls: negative; positive; kit controls (KC). The negative control is derived from a normal (non-haemophilia) human donor, with no demonstrable anti-FVIII antibody presence. The positive and KC are derived from human serum containing antibodies to human FVIII. The KC is lot specific, defined by dilution studies of a known positive sample and is tested by the manufacturer to ensure that the threshold results in the expected reportable results in over 90 test samples (positive/negative Bethesda activity). All samples in our study were tested in duplicate and defined as being positive if they displayed an OD greater than that of the KC.

Statistical analyses Evaluation of diagnostic accuracy was performed for the FVIII ELISA (index test), assuming the NBA to be the “gold standard” comparator (reference test). Analysis was performed on the anonymised grouped data as a post-hoc analysis. Presented data represent a whole case analysis as it is likely that missing data was

Thrombosis and Haemostasis 114.4/2015

© Schattauer 2015

Downloaded from www.thrombosis-online.com on 2015-06-14 | ID: 1001077309 | IP: 138.37.211.113 Note: Uncorrected proof, epub ahead of print online For personal or educational use only. No other uses without permission. All rights reserved.

Batty et al. Diagnostic test evaluation of FVIII ELISA

missing not at random (MNAR); for example a test may not have been performed due to high or low clinical suspicion on the basis of previous clinical/laboratory information. The performance of the index test (ELISA) was summarised using sensitivity, specificity and predictive values (negative and positive). Confidence intervals (binomial exact) were calculated from the standard error taking these values as proportions (33). Adjustment for inter-individual clustering of data was made for these measures, modelled using a generalised estimating equation approach as described previously (34). Further post-hoc analysis of the unadjusted data is presented in the form of likelihood ratios (negative and positive) and diagnostic odds ratios (DORs) (35). Confidence intervals (CIs) were calculated from the standard error for the likelihood ratio as previously described (36) and for the DOR using the formula log DOR ± 1.96SE(log DOR) (35). An assessment of association of the results of both assays was performed using correlation (Pearson) of the log-adjusted NBA results in comparison to the ELISA OD. Adjustment was made for inter-individual clustering within these analyses, through usage of the first recorded NBA positive database result. Statistical analysis of continuous variables was performed using a one-way ANOVA, with post-hoc comparison of groups using the post-hoc Turkey test (37). All tests are two sided with a p value < 0.05 taken as significant. Statistical analyses were performed within IBM SPSS version 21 (IBM Corporation, Armonk, NY, USA) and Stata Statistical Software: Release 12 (College Station, TX, USA: StataCorp LP). The findings are presented as per the recommendations of the Standards for the Reporting of Diagnostic Accuracy Studies (STARD) Statement (38). The STARD checklist is included with the accompanying Suppl. Material (available online at www.thrombosis-online.com).

Results A total of 569 samples from 273 patients were sent for inhibitor testing within the observation period. Of these samples, 17 (14 patients) were excluded for a non-haemophilia A diagnosis and 55 (45 patients) due to incomplete testing (▶ Figure 1). The reasons for exclusion due to incomplete testing were absence of NBA (n=51), lack of ELISA testing (n=2) or neither test being performed (n=2). Following exclusions, data were available for 497 samples from 239 patients. Of these samples 291 (59 %) samples were from patients with severe haemophilia A (140 patients), 129 (26 %) from patients with non-severe haemophilia A (86 patients) and 77 (15 %) from patients with acquired haemophilia A (14 patients) (▶ Figure 1). The mean number of samples tested per patient was 2.1 (range 1–20), with a single sample tested in 145 patients (60.7 %). Patients with AHA had significantly more samples tested per patient (5.5 ± 6.0) than those with severe (mean 2.1 ± SD 2.0, p< 0.005) or non-severe HA (mean 1.5 ± SD 1.9, p< 0.005)

hibitor sample and patient prevalence of 12.7 % and 11.8 % respectively. Of samples testing positive by the NBA (n=63), the median inhibitor titre was 1.2 BU/ml (range 0.7–978.0) with 49 being low-titre (median 1.1 BU/ml; range 0.7–4.8) and 14 hightitre (median 63.7 BU/ml; range 11.8–978.0). Of the 63 NBA positive samples, 49 (18 patients) were also positive by ELISA and 14 (13 patients) negative by ELISA (▶ Figure 1). All samples that tested negative by the ELISA and positive by NBA (n=14) were of low inhibitory titre with a mean inhibitor titre of 0.8 BU/ml (range 0.7–1.0). The majority of these samples were tested in centre 2 (n=13). In samples classified as low-titre by the NBA and ELISA positive (n=35), the mean ELISA OD was 0.82 (range 0.25–1.98). Twenty nine of these samples (29/35, 82.9 %) had an ELISA OD ≤ 1.0, 5/35 (14.3 %) an OD > 1 and < 2 and a single sample had an OD of 2.0. For samples classified as high titre (n=14), the mean ELISA OD was 2.03 (range 0.59–3.36). Within this group 12/14 (85.7 %) samples had an OD ≥ 1.0 and 9/14 (64.3 %) had an OD ≥ 2.0. Two samples, from one patient, with high titre antibodies by the NBA, were recorded as having an OD ≤ 1.0. There were 434 samples (226 patients) that tested negative by the NBA of which 408 also tested negative by the ELISA (218 patients), whilst 26 tested positive (15 patients) by the ELISA. Of the 26 ELISA positive/NBA negative samples, 14 (5 patients) were tested by centre 1 and 12 (10 patients) by centre 2. Additional samples were available for 8/15 patients, of which 6/8 were positive by the NBA and/or ELISA at another point within the observational period. The mean ELISA OD for these samples was 0.63 (range 0.20–1.59). The mean FVIII:C was 62.4IU/dL (range < 1–362.4), with 18 (69 %) having a FVIII:C ≥ 5 IU/dl. Calculation of the performance of the ELISA for all samples, demonstrated a specificity 94.0 % (95 % CI, 91.3–96.0), sensitivity 77.8 % (95 % CI, 65.5–87.3), NPV 96.7 % (95 % CI, 94.5–98.2), PPV 65.3 % (95 % CI, 53.5–76.0), LR+ 13.0 (95 % CI, 8.7–19.3), LR- 0.2 (95 % CI, 0.1–0.4) and a DOR of 54.9 (95 % CI, 27.0–112.0) (▶ Table 1). In the subgroup analysis, similar findings are seen with high specificity, negative predictive values, positive likelihood ratios for use of the ELISA in patients with congenital HA (severe and non-severe). High sensitivity and negative predictive value are seen in patients with AHA, although these values may be skewed due to the smaller sample size (n=77, 14 patients) and no “falsenegative” results seen (▶ Table 1). In the subgroup of samples from patients with AHA, 12 (56 %) tested positive by ELISA, but were negative by the NBA. Finally, the performance of the ELISA was modelled using generalised estimating equations, to adjust for potential clustering of results within individual patients. This analysis gave a specificity of 95.3 % (95 % CI, 92.0–97.3 %), sensitivity 69.4 % (95 % CI, 52.8–82.2 %), NPV 96.7 % (95 % CI, 94.4–98.0 %) and a PPV 65.3 % (95 % CI, 41.0–74.0 %).

Assessment of association of ELISA OD with NBA

Diagnostic accuracy assessment There were 63 samples positive by the NBA, from 28 patients (severe=17, non-severe=6, acquired=5), representing a positive in-

An assessment of association was made for the results for the ELISA OD and NBA inhibitor titre for samples positive by the NBA (n=63). To account for potential inter-individual clustering

© Schattauer 2015

Thrombosis and Haemostasis 114.4/2015 Downloaded from www.thrombosis-online.com on 2015-06-14 | ID: 1001077309 | IP: 138.37.211.113 Note: Uncorrected proof, epub ahead of print online For personal or educational use only. No other uses without permission. All rights reserved.

3

4

Batty et al. Diagnostic test evaluation of FVIII ELISA

Figure 1 : Sample classification by test result and diagnostic header.

of data, the results of the first recorded database entry for each patient was used for this analysis, resulting in a total of 28 samples. There was strong correlation (r=0.774, p< 0.0005) between the log adjusted result of the NBA and the ELISA optical density (▶ Figure 2). Finally, the effect of variability in methodology between centres on measures of association was assessed by repeating the correlation analysis by centre (centre 1 = 7 samples, centre 2 = 21 samples). Correlation of samples for centre 1, gave an r value of 0.670 (p=0.10) and 0.894 (p< 0.0005) for centre 2.

Discussion We present data on a large cohort of samples from patients with congenital and acquired haemophilia A tested as part of routine clinical care using the NBA and a commercially available FVIII ELISA in parallel. Use of the ELISA in both centres has demonstrated high specificity, negative predictive value and diagnostic odds ratio suggesting a potential role as a screening test for FVIII antibody presence. This is the first study in the setting of haemophilia reported following the recommendations of the STARD statement (38), enabling future use in meta-analyses.

There have been two previously published studies comparing performance of an earlier version of this FVIII ELISA kit to a functional inhibitor assay (15, 19). The first compared samples from 131 samples (93 patients – 92 congenital HA and 1 AHA) with the New Oxford Inhibitor Assay (15). This study demonstrated an unadjusted specificity of 78.4 %, sensitivity of 97.7 %, NPV 98.6 %, PPV 68.9 % with moderate correlation (r=0.68) between the results for the two assays. The second study evaluated the ELISA in 246 plasma samples (176 patients) with known inhibitor activity by the Bethesda (classical and Nijmegen modification) assay (19). Of these samples, 235 were positive by ELISA (unadjusted sensitivity 95.5 %) with strong correlation (r=0.82) seen between the logtransformed results of the Bethesda assay and FVIII ELISA OD. Since these two studies, changes have been made to the FVIII plate antigen and method of determining positivity for this ELISA kit. The earlier version of this FVIII ELISA used a plate antigen of Recombinate® (Baxter Healthcare Corp, Deerfield, IL, USA), a H2 haplotype rFL-FVIII (31, 32), with positivity derived from an absorbance value twice the mean of the negative control (15, 19). The ELISA kit used in our study used a plate antigen of Kogenate®, an H1 haplotype rFL-FVIII, and positivity was defined using a kit control consisting of human serum containing FVIII antibodies in

Thrombosis and Haemostasis 114.4/2015

© Schattauer 2015

Downloaded from www.thrombosis-online.com on 2015-06-14 | ID: 1001077309 | IP: 138.37.211.113 Note: Uncorrected proof, epub ahead of print online For personal or educational use only. No other uses without permission. All rights reserved.

Batty et al. Diagnostic test evaluation of FVIII ELISA

Table 1: Diagnostic Accuracy Evaluation with comparison of the FVIII ELISA to the Nijmegen Bethesda assay. HA = haemophilia A; CI = confidence interval. *No false negative results seen (specificity/negative predictive value = 100%).

Severe HA

Non-Severe HA

Acquired HA

Total

Samples

291

129

77

497

Patients

140

85

14

239

Antibody prevalence (95%CI)

15.1% (11.0 – 19.8)

9.3% (4.9 – 15.7)

9.1% (3.7 – 17.8)

12.7% (9.9 – 15.9)

Sensitivity (95%CI)

77.3% (62.2 – 88.5)

66.7% (34.9 – 90.1)

100.0% (59.0 – 100.0)

77.8% (65.5 – 87.3)

Specificity (95%CI)

96.8% (93.7 – 98.6)

94.9% (89.2 – 98.1)

82.9% (72.0 – 90.8)

94.0% (91.3 – 96.0)

Positive predictive value (95%CI)

81.0% (65.9 – 91.4)

57.1% (28.9 – 82.3)

36.8% (16.3 – 61.6)

65.3% (53.5 – 76.0)

Negative predictive value (95%CI)

96.0% (92.7 – 98.1)

96.5% (91.3 – 99.0)

100.0% (93.8 – 100.0)

96.7% (94.5 – 98.2)

Positive likelihood ratio (95%CI)

23.9 (11.8 – 48.1)

13.0 (5.4 – 31.2)

5.8 (3.5 – 9.8)

13.0 (8.7 – 19.3)

Negative likelihood ratio (95%CI)

0.2 (0.1 – 0.4)

0.4 (0.2 – 0.8)

*

0.2 (0.1 – 0.4)

Diagnostic odds ratio (95%CI)

102.0 (37.9 – 272.0)

37.0 (9.1 – 152.0)

*

54.9 (27.0 – 112.0)

bovine albumin and 0.1 % NaN3 (31). The main difference in primary sequence between H1 and H2 haplotype FVIII is a single nucleotide polymorphism (Asp1241Glu) within the B-domain (31). Although good correlation has been demonstrated between the results for the two assays in an evaluation of the effect of antiphospholipd antibodies (32), our study is the first directly comparing this modified ELISA kit with current methodology for inhibitor testing. Our findings of a high NPV for the ELISA are in keeping with those previously described (15). With our usage of this ELISA we have, however, seen greater specificity (94.0 % vs 78.4 %) (15), albeit with a lower sensitivity (77.8 % vs 97.7 % vs 95.5 %) (15, 19). These differences may be in part due to differences in methodology in comparison with previous studies which compared samples that were either known to be positive (19) or “at risk of inhibitor development” (15), rather than usage in routine laboratory practice, a role which is closer to being a screening tool. An interesting observation from the sub-group analysis of patients with AHA was that a substantial proportion of samples (n=12, 15.6 %, three patients) were positive by ELISA (mean OD 0.583, range 0.246–1.585) but negative by the NBA. Of these, 7/12 (58.3 %) had an OD value ≥ 2-fold that of the KC. All of these samples had substantial quantities of FVIII:C present (mean 122.8, range 40.8–362.4) prior to testing. The presence of FVIII:C in samples, when using a functional inhibitor assay such as NBA, may impair detection of potentially clinically relevant FVIII antibodies (39). Modifications for the presence of FVIII:C in samples have been described previously which include adjustment, either mathematical, in vitro standardisation of assay FVIII:C or pre-analytical heat-treatment (39). Data from our group and others has previously demonstrated that pre-analytical heat treatment of samples from AHA patients with residual FVIII:C may reveal per-

sistent antibody detected by ELISA (40) and/or NBA (40, 41). Given the difficulties in the diagnosis and accurate quantification of FVIII inhibitors in patients with type II antibodies (complex kinetics), who often have residual FVIII:C, the ELISA technique offers an attractive platform for surveillance. The clinical relevance of persistently detectable antibody by ELISA upon normalisation of FVIII:C in this AHA group of patients requires further prospective study. There are a number of difficulties in assessing new diagnostic tests for clinical practice. The first of these relates to the intended role for this test, which may be either to act as a triage (screening), replacement or an add-on (42). For FVIII antibodies, the ideal test from a clinical, immunological and translational perspective would allow detection of both inhibitory and non-inhibitory antibodies and give a quantitative result that correlates with the functional inhibitory capacity. Ideally this test should be reproducible and insensitive to the pre-analytical variables affecting current laboratory methods. The second limitation is where there is an imperfect reference standard or where the true relevance of the spectrum of pathological findings is unknown, which is the current situation in testing for FVIII antibodies. Methodology for assessment of diagnostic tests is also limited in the situation where a new assay may detect potentially relevant findings that are not detected on current assays, as may be the case for apparent “non-inhibitory” antibodies. Despite recent studies that have attempted to follow up patients with non-inhibitory antibodies in the context of severe HA longitudinally, the physiological relevance of these remains uncertain in severe, and unexplored in non-severe HA (20). It is not known whether the presence of non-inhibitory antibodies may have prognostic significance, for example predicting the future presence of inhibitory antibodies (epitope spreading) or re-

© Schattauer 2015

Thrombosis and Haemostasis 114.4/2015 Downloaded from www.thrombosis-online.com on 2015-06-14 | ID: 1001077309 | IP: 138.37.211.113 Note: Uncorrected proof, epub ahead of print online For personal or educational use only. No other uses without permission. All rights reserved.

5

6

Batty et al. Diagnostic test evaluation of FVIII ELISA

What is known about this topic?

• • •

Enzyme linked immunosorbent assay (ELISA) is a routine laboratory assay used in most clinical laboratories. Research usage of a commercial FVIII ELISA has previously suggested this assay to have a high sensitivity. The role of FVIII ELISA in routine laboratory inhibitor surveillance is not well characterised.

What does this paper add?

• •

Usage of a commercially available FVIII ELISA offers a highly specific method of testing for FVIII antibodies in clinical samples. This is the first study in haemophilia to conform to the STAndards for the Reporting of Diagnostic accuracy studies statement (STARD) enabling future meta-analysis inclusion.

lapse following immune tolerance induction (congenital HA) or immunosuppression (AHA). One of the major limitations of this study relates to the retrospective design to evaluate the performance of the FVIII ELISA. Usage of anonymised samples may result in patients being miscategorised or may not take into account other important clinical variables present at the time of testing. This approach also does not allow assessment of the reproducibility of results. This study suffers from one of the limitations affecting many diagnostic accuracy studies in that there are often multiple samples taken on the same patient (43), resulting in the formation of clustered data. Without adequate adjustment this may occasionally skew the size of effect or result in inappropriately narrow CIs. For example, a patient may always test negative by one assay and positive by the other assay, resulting in false elevation of either the rate of false positives or negatives, resulting in data from that patient being over-weighted. Within our analysis we have attempted to adjust for this in the assessment of diagnostic accuracy and correlation. We acknowledge there were differences in the methodology used for performing the NBA at the two centres, although these have been well described within the literature (28, 44, 45). Usage of 4 % BSA in place of FVIII deficient plasma has been described in two studies (44, 45). The first of these, which initially described this approach, first used a spiking experiment of purified anti-FVIII:C IgG combined with plasma from a single patient with severe haemophilia A, titrated to an inhibitor titre of 1.5 BU/ml and performed the assay in quadruplicate (44). This showed good agreement between the results of the inhibitor assay using congenital FVIII:C deficiency plasma (FDP) or 4 % BSA, with or without the addition of VWF. This preliminary analysis was then confirmed in samples from six patients with known inhibitor presence (44). A validation of this modification has been recently published, comparing usage of FDP with 4 % BSA, in 59 samples (35 patients) with known FVIII inhibitors. This study demonstrated good agreement between the results of the two assays, especially in those of low-inhibitory capacity (< 2 BU/ml) (45). Comparison has also been made between usage of a one stage and chromogenic FVIII

assay as part of the Bethesda assay (28). This study compared 1,005 samples (702 patients) using the NBA and Chromogenic Bethesda assay (CBA). Within this, 880/883 samples negative by the NBA were also negative by CBA and 43/80 specimens (53.8 %) with NBU 0.5–1.9 were also positive by the CBA. In samples positive by the NBA with a titre ≥ 2.0 BU/ml, 42/42 were positive by CBA with excellent correlation (r=0.98, p< 0.0001) between the assay results. This variability in methodology will be a common limitation in any study that uses local inhibitor testing which may result in inter-laboratory variability described previously (10–12). To assess whether this might impact on these analyses, the correlation analyses were repeated using data from each centre showing similar findings. Differences in the correlation seen for the centres most likely relates to the smaller number of NBA positive samples analysed for centre 1 (n=7). An alternative explanation may relate to a higher proportion of samples from centre 1 being from patients with acquired or non-severe haemophilia (centre 1: 4/7, 57.1 %; centre 2: 7/21, 33.3 %) which could display more complex inactivation kinetics (46). Finally, a potential limitation to this commercially available ELISA is the use of a single plate antigen (Kogenate®). Difference in specificity to different recombinant antigens was first described for functional inhibitor assays (47). More recent work has confirmed these findings of difference in antigen specificity when tested by validated, in-house research-laboratory ELISAs (48–50). These findings are of particular interest clinically given the small differences seen in primary sequence between full-length recombinant FVIII products. It is possible that some samples that tested negative by ELISA in our study may have demonstrated antigen specificity to an alternative product. However, further evaluation of this is beyond the scope of this study, which aimed to describe experience of introducing a standardised, commercial ELISA for FVIII antibody testing into routine practice. In summary, we present analysis of clinical usage of a commercially available FVIII ELISA kit in comparison to the NijmegenBethesda assay in two large, UK comprehensive care haemophilia centres. Our experience has shown good specificity and negative predictive value for this assay and good correlation between the log-adjusted inhibitor titre and the results of the ELISA optical density. The ELISA technique is a routine assay platform, available in most service laboratories. This offers the potential for efficient batching of FVIII antibody surveillance for non-urgent clinical samples, representing the majority of inhibitor screen requests. Conflicts of interest

None declared.

References 1. Gouw SC, van den Berg HM, Fischer K, et al. Intensity of factor VIII treatment and inhibitor development in children with severe hemophilia A: the RODIN study. Blood 2013; 121: 4046–4055. 2. Collins PW, Chalmers E, Hart DP, et al. Diagnosis and treatment of factor VIII and IX inhibitors in congenital haemophilia: (4th edition). UK Haemophilia Centre Doctors Organization. Br J Haematol 2013; 160: 153–170.

Thrombosis and Haemostasis 114.4/2015

© Schattauer 2015

Downloaded from www.thrombosis-online.com on 2015-06-14 | ID: 1001077309 | IP: 138.37.211.113 Note: Uncorrected proof, epub ahead of print online For personal or educational use only. No other uses without permission. All rights reserved.

Batty et al. Diagnostic test evaluation of FVIII ELISA 3. Hay CR, Palmer B, Chalmers E, et al. Incidence of factor VIII inhibitors throughout life in severe hemophilia A in the United Kingdom. Blood 2011; 117: 6367–6370. 4. Mauser-Bunschoten EP, Den Uijl IE, Schutgens RE, et al. Risk of inhibitor development in mild haemophilia A increases with age. Haemophilia 2012; 18: 263–267. 5. Eckhardt CL, van Velzen AS, Peters M, et al. Factor VIII gene (F8) mutation and risk of inhibitor development in nonsevere hemophilia A. Blood 2013; 122: 1954–1962. 6. Collins W, Chalmers E, Hart D, et al. Diagnosis and management of acquired coagulation inhibitors: a guideline from UKHCDO. Br J Haematol 2013; 162: 758–773. 7. Huth-Kuhne A, Baudo F, Collins P, et al. International recommendations on the diagnosis and treatment of patients with acquired hemophilia A. Haematologica 2009; 94: 566–575. 8. Kasper CK, Aledort L, Aronson D, et al. Proceedings: A more uniform measurement of factor VIII inhibitors. Thromb Diath Haemorrh 1975; 34: 612. 9. Verbruggen B, Novakova I, Wessels H, et al. The Nijmegen modification of the Bethesda assay for factor VIII:C inhibitors: improved specificity and reliability. Thromb Haemost 1995; 73: 247–251. 10. Meijer P, Verbruggen B. The between-laboratory variation of factor VIII inhibitor testing: the experience of the external quality assessment program of the ECAT foundation. Semin Thromb Hemost 2009; 35: 786–793. 11. Kitchen S, Jennings I, Preston FE, et al. Interlaboratory variation in factor VIII:C inhibitor assay results is sufficient to influence patient management: data from the UK national quality external assessment scheme for blood coagulation. Semin Thromb Hemost 2009; 35: 778–785. 12. Peerschke EI, Castellone DD, Ledford-Kraemer M, et al. Laboratory assessment of factor VIII inhibitor titer: the North American Specialized Coagulation Laboratory Association experience. Am J Clin Pathol 2009; 131: 552–558. 13. Favaloro EJ, Bonar R, Duncan E, et al. Mis-identification of factor inhibitors by diagnostic haemostasis laboratories: recognition of pitfalls and elucidation of strategies. A follow up to a large multicentre evaluation. Pathology 2007; 39: 504–511. 14. Favaloro EJ, Bonar R, Duncan E, et al. Identification of factor inhibitors by diagnostic haemostasis laboratories: a large multi-centre evaluation. Thromb Haemost 2006; 96: 73–78. 15. Martin PG, Sukhu K, Chambers E, et al. Evaluation of a novel ELISA screening test for detection of factor VIII inhibitory antibodies in haemophiliacs. Clin Lab Haematol 1999; 21: 125–128. 16. Shetty S, Ghosh K, Mohanty D. An ELISA assay for the detection of factor VIII antibodies – comparison with the conventional Bethesda assay in a large cohort of haemophilia samples. Acta Haematol 2003; 109: 18–22. 17. Ling M, Duncan EM, Rodgers SE, et al. Low detection rate of antibodies to nonfunctional epitopes on factor VIII in patients with hemophilia A and negative for inhibitors by Bethesda assay. J Thromb Haemost 2003; 1: 2548–2553. 18. Towfighi F, Gharagozlou S, Sharifian RA, et al. Comparative measurement of anti-factor VIII antibody by Bethesda assay and ELISA reveals restricted isotype profile and epitope specificity. Acta Haematol 2005; 114: 84–90. 19. Sahud MA, Pratt KP, Zhukov O, et al. ELISA system for detection of immune responses to FVIII: a study of 246 samples and correlation with the Bethesda assay. Haemophilia 2007; 13: 317–322. 20. Klintman J, Hillarp A, Berntorp E, et al. Long-term anti-FVIII antibody response in Bethesda-negative haemophilia A patients receiving continuous replacement therapy. Br J Haematol 2013; 163: 385–392. 21. Krudysz-Amblo J, Parhami-Seren B, Butenas S, et al. Quantitation of anti-factor VIII antibodies in human plasma. Blood 2009; 113: 2587–2594. 22. Zakarija A, Harris S, Rademaker AW, et al. Alloantibodies to factor VIII in haemophilia. Haemophilia 2011; 17: 636–640. 23. Lebreton A, Lapalud P, Chambost H, et al. Prevalence and epitope specificity of non-neutralising antibodies in a large cohort of haemophilia A patients without inhibitors. Thromb Haemost 2011; 105: 954–961. 24. Mondorf W, Klinge J, Luban NL, et al. Low factor VIII recovery in haemophilia A patients without inhibitor titre is not due to the presence of anti-factor VIII antibodies undetectable by the Bethesda assay. Haemophilia 2001; 7: 13–19. 25. Scandella D, Mondorf W, Klinge J. The natural history of the immune response to exogenous factor VIII in severe haemophilia A. Haemophilia 1998; 4: 546–551.

26. White GC, Rosendaal F, Aledort LM, et al. Definitions in hemophilia. Recommendation of the scientific subcommittee on factor VIII and factor IX of the scientific and standardization committee of the International Society on Thrombosis and Haemostasis. Thromb Haemost 2001; 85: 560. 27. Verbruggen B, van HW, Novakova I, Lillicrap D, et al. A 4 % solution of bovine serum albumin may be used in place of factor VIII:C deficient plasma in the control sample in the Nijmegen Modification of the Bethesda factor VIII:C inhibitor assay. Thromb Haemost 2002; 88: 362–364. 28. Miller CH, Rice AS, Boylan B, et al. Comparison of clot-based, chromogenic and fluorescence assays for measurement of factor VIII inhibitors in the US Hemophilia Inhibitor Research Study. J Thromb Haemost 2013; 11: 1300–1309. 29. European Medicines Agency. Report of Expert Meeting on Factor VIII Products and Inhibitor Development. 2007. London, European Medicines Agency. 30. Moore GW, Maloney JC, Ngundu CJ, et al. Assessing lowest limits of detection of FVIII inhibitor assays. EAHAD 2010. 31. Viel KR, Ameri A, Abshire TC, et al. Inhibitors of factor VIII in black patients with hemophilia. N Engl J Med 2009; 360: 1618–1627. 32. Sahud M, Zhukov O, Mo K, et al. False-positive results in ELISA-based anti FVIII antibody assay may occur with lupus anticoagulant and phospholipid antibodies. Haemophilia 2012; 18: 777–781. 33. Clopper CJ, Pearson ES. The use of confidence or fiducial limits illustrated in the case of the binomial. Biometrika 1934; 26: 404–413. 34. Genders TS, Spronk S, Stijnen T, et al. Methods for calculating sensitivity and specificity of clustered data: a tutorial. Radiology 2012; 265: 910–916. 35. Glas AS, Lijmer JG, Prins MH, et al. The diagnostic odds ratio: a single indicator of test performance. J Clin Epidemiol 2003; 56: 1129–1135. 36. Deeks JJ, Altman DG. Diagnostic tests 4: likelihood ratios. Br Med J 2004; 329: 168–169. 37. Turkey JW. Comparing Individual Means in the Analysis of Variance. Biometrics 1949; 5: 99–114. 38. Bossuyt PM, Reitsma JB, Bruns DE, et al. Towards complete and accurate reporting of studies of diagnostic accuracy: the STARD initiative. Br Med J 2003; 326: 41–44. 39. Kitchen S, McCraw A, Echenagucia M. Diagnosis of Hemophilia and Other Bleeding Disorders : A Laboratory Manual. Second Edition ed. Montréal: World Federation of Hemophilia (WFH); 2010. 40. Batty P, Platton S, Bowles L, et al. Pre-analytical heat treatment and a FVIII ELISA improve Factor VIII antibody detection in acquired haemophilia A. Br J Haematol 2014; 166: 953–956. 41. Sahud M, Ruden S, Soriano C, et al. Characterization of Factor VIII Autoantibodies with Complex Reaction Kinetics is Improved by Heat Inactivation of the Test Plasma to Remove Residual Factor Activity. J Thromb Haemost 2005; 3: P0651. 42. Bossuyt PM, Irwig L, Craig J, et al. Comparative accuracy: assessing new tests against existing diagnostic pathways. Br Med J 2006; 332: 1089–1092. 43. Pepe MS. The Statistical Evaluation of Medical Tests for Classification and Prediction. Oxford: Oxford University Press; 2003. 44. Verbruggen B, van Heerde W, Novakova I, et al. A 4 % solution of bovine serum albumin may be used in place of factor VIII:C deficient plasma in the control sample in the Nijmegen Modification of the Bethesda factor VIII:C inhibitor assay. Thromb Haemost 2002; 88: 362–364. 45. Kershaw GW, Chen LS, Jayakodi D, et al. Validation of 4 % albumin as a diluent in the Bethesda Assay for FVIII inhibitors. Thromb Res 2013; 132: 735–741. 46. Gawryl MS, Hoyer LW. Inactivation of factor VIII coagulant activity by two different types of human antibodies. Blood 1982; 60: 1103–1119. 47. Astermark J, Voorberg J, Lenk H, DiMichele D, Shapiro A, Tjonnfjord G, et al. Impact of inhibitor epitope profile on the neutralizing effect against plasma-derived and recombinant factor VIII concentrates in vitro. Haemophilia 2003; 9: 567–572. 48. Vincent AM, Lillicrap D, Boulanger A, et al. Non-neutralizing anti-FVIII antibodies: different binding specificity to different recombinant FVIII concentrates. Haemophilia 2009;15:374–6. 49. Klintman J, Hillarp A, Donfield S, et al. Antibody formation and specificity in Bethesda-negative brother pairs with haemophilia A. Haemophilia 2013; 19: 106–112. 50. Butenas S, Krudysz-Amblo J, Rivard GE, Mann G. Product-dependent anti-factor VIII antibodies. Haemophilia 2013; 19: 619–625.

© Schattauer 2015

Thrombosis and Haemostasis 114.4/2015 Downloaded from www.thrombosis-online.com on 2015-06-14 | ID: 1001077309 | IP: 138.37.211.113 Note: Uncorrected proof, epub ahead of print online For personal or educational use only. No other uses without permission. All rights reserved.

7