Measurement of factor VIII (FVIII) inhibitors in the U.S. was standardized in 1975 at a meeting in Bethesda, Maryland, which produced a method bearing the name of the conference site [1]. The Bethesda assay (BA) calls for a two-hour incubation of a mixture of patient plasma with normal pool plasma (NPP) and comparison of the FVIII activity remaining with a similarly treated control mixture containing buffer and NPP. The percent residual activity in the patient mix is converted to Bethesda units (BU). One BU is defined as the amount of inhibitor producing a residual activity of 50%. This method has persisted virtually unchanged in the majority of laboratories in the U.S. despite the introduction of modifications to the method in Europe in 1995 [2]. The Nijmegen modification [3], or Nijmegen-Bethesda assay (NBA), calls for two changes in the standard BA: 1) buffering of the NPP used in patient and control mixtures for incubation, and 2) use of FVIII-deficient plasma (VIIIDP) instead of buffer in the control mixture and for dilution. A subsequent modification allowed substitution of 4% bovine serum albumin for the FVIIIDP [4], thereby reducing cost. While studies have documented that these modifications have improved the performance of the BA for measurement of FVIII inhibitors [5-6], both the BA and the NBA have shown poor inter-laboratory correlation in four different proficiency testing programs [2, 7-9]. For the U.S. inhibitor surveillance pilot project conducted at the Centers for Disease Control and Prevention (CDC) in 2006-2009 [10], centralized inhibitor measurement was instituted using a method modified to facilitate national inhibitor surveillance by allowing simplified sample handling and valid testing of treated patients. This paper presents the details of the adopted method and its validation.

Results on 1353 specimens from 710 patients with hemophilia A (HA) and 289 specimens from 160 patients with hemophilia B (HB) were examined. Characteristics of enrolled patients are shown in Table 1. As determined by the enrolling sites, 122 HA patients and 2 HB patients had a history of a clinically significant inhibitor (HI) at the time of study enrollment. One HB patient and 63 of 122 HA patients were reported to have previously had an inhibitor titer >5 BU.

Modifications of the published NBA were validated during the implementation of the project. Two shipping methods were compared on split samples. Fifty specimens with inhibitor titers ranging from 0 to 900 were split and shipped to CDC both on cold packs and frozen on dry ice; their NBUs showed excellent correlation (r=0.998). The cold pack method was chosen to simplify specimen handling.

Initial tests of 228 frozen specimens from severe HA patients showed that126 (55%) had measurable VIII:C. All were from patients reported to have been treated with FVIII-containing products within 72 hours of blood collection. In the unmodified NBA, these specimens showed a residual activity of ≥100% and an inhibitor titer of 0. In order to eliminate the VIII:C while preserving the inhibitory antibodies, plasmas were heated to 56°C for 30 minutes and centrifuged. Inhibitor results on 202 patient specimens were compared with and without heat treatment (Table 2). Of 159 specimens with negative HI, 120 had unheated NBU of 0. After heating, 45 (37.5%) of these remained 0, 74 (61.7%) increased from 0 to 0.1-0.2 NBU, and one increased from 0 to 0.7 NBU (Figure 1a). Among 37 specimens with unheated NBU greater than 0 but below 0.5, 31 (83.8%) remained constant, 6 (16.2%) decreased below 0.1, and none increased. Of 30 specimens with positive HI but with inhibitor results below 0.5 NBU at enrollment, 5 (16.7%) rose to ≥0.5 NBU after heating, significantly more than the 1 of 159 (0.6%) with negative HI (P=0.0004); one positive specimen decreased from 0.9 to 0.4 NBU (Figure 1b). In a series of 21 specimens with results ≥0.5 NBU (Figure 2), the correlation of the inhibitor titers in heated and unheated specimens was 0.94 (P<0.0001). Plasma from 30 individuals without hemophilia was also heat-treated and tested for inhibitor. Of those, 20 had 0 NBU, 9 had 0.1 NBU, and 1 had 0.2 NBU (Table 2).

VIII:C and VIII:Ag were measured in heated and unheated specimens from patients who had been infused with FVIII products within 24 hours of blood draw. Among 15 severe patients, all had measurable VIII:C and VIII:CAg before heating and <1 IU/dL after heating. In 7 patients with moderate HA, VIII:C and VIII:Ag also decreased to <1 IU/dL after heating.

The steps of the method adopted for the CDC surveillance are shown in Table 3. Coefficient of variation (CV) was 9.8% for VIII:C of the negative control (n=117) and 10.3% for NBU of a 1 NBU positive control (n=114). Within-runs variation of the negative control was 5.3% (n=10) and of the positive control 4.8% (n=10). Substitution of 4% bovine serum albumin buffer (BSA) for the VIIIDP in the incubation mixture was also tested but was not adopted as part of the standard procedure. In 71 specimens, the correlation of BSA and VIIIDP was 0.990. There were no differences in results if BSA was used for the incubation mixture and for dilution of specimens with inhibitor titers ranging from 0 to 56.5. CV for the negative control was 10.6%.

In order to establish a reference range for the assay, results at enrollment were compared for the 588 patients with negative HI and 86 of the 122 patients with positive HI. Thirty-six patients who had undergone successful ITI were excluded. Figure 3a shows those with NBU below 1.0 with lines plotted at NBU of 0.4, 0.5, and 0.6. No specimen from the 588 negative HI patients (left) had NBU between 0.4 and 0.7. Among 56 specimens from patients with positive HI (right), none showed NBU of 0.4. A cut-off for a positive inhibitor of ≥0.5 misclassified 4 fewer specimens with positive HI than a cut-off of ≥0.6, showing a higher sensitivity. A positive CDC inhibitor was defined as ≥0.5. The results in this range for all inhibitor tests completed on these subjects (Figure 3b) were similar to those at enrollment, supporting the cut-off assignment.

Table 4 shows all of the results at enrollment. Of 588 patients with negative HI reported by the sites, 581 (98.8%) had NBU <0.5. Six positive specimens ranged from 0.5 to 1.1, and one was 10.2. All were negative for DRVVT and were drawn peripherally. Three, including the high titer inhibitor, were confirmed on repeat specimens; one was not confirmed in a specimen drawn 3 months later but reappeared after 2 years of surveillance; three have not been retested. Of 122 patients with a positive HI, 51 (41.8%) had a positive CDC inhibitor at enrollment, including 18 (14.8%) with high titer antibodies (>5.0 NBU). Thirty-six (29.5%) had undergone successful ITI, and 35 others (28.7%) were negative at the time of enrollment. A maximum inhibitor titer >5 BU was reported for 65 of 122 patients, and 3 more had a high titer when tested at CDC, for a total of 68 (55.7%) with high titer inhibitors.

Some characteristics of the adopted assay method allow evaluation of its limitations. If a true negative specimen would be expected to have a result within 2 standard deviations of the mean of the normal control (30.55±2.98), the lowest %RA expected would be 67, producing an inhibitor titer of 0.6 NBU in a specimen that is actually negative. Additionally, using the regression equation for the NBU curve, the change in the VIII:C of the incubation mix with change in NBU was calculated to be 2.9 IU/dL per 0.1 NBU. A change from a titer of 0 to one of 0.5 would result from a change in the incubation mix of only 14.5 IU/dL.

Factor IX inhibitor assays were performed with and without heat treatment on 17 specimens. All had NBU of 0 pre heating. One specimen from a patient with positive HI rose from 0 to 0.6 NBU after heating. Of 160 HB patients enrolled, 2 were reported to have a positive HI. Both were positive at enrollment at 0.6 and 5.6 NBU using the modified method. All other patients (98.9%) had NBU ≤0.2 at enrollment, including 115 (61.5%) with 0 NBU (Figure 4a). Results for all 289 specimens tested are shown in Figure 4b. CV was 6.8 % for the negative control. Within runs CV was 5.8% for the negative control (n=10). No positive control was used.

Because of the increased use of prophylaxis and ITI, sampling a hemophilia patient in the baseline state, without circulating FVIII, has become difficult; however, the presence of even a minimum amount of patient FVIII violates an assay principle of the BA, that the patient and control mixtures be equivalent before incubation. While it is possible to arithmetically account for the additional FVIII in the mix, the kinetics of the antigen-antibody reaction may be altered. There may be competition between the infused FVIII and that provided in the assay, which are likely to be different, or the antibody may be bound to the infused factor, leading to a false-negative result. If the additional FVIII is not accounted for, a residual activity of 100% or higher will be measured leading to an NBU result of 0.

To avoid this problem, we added a heat treatment step to eliminate FVIII from the sample prior to testing. As antibodies, inhibitors are more stable than many coagulation proteins, a characteristic that allows conditions under which the inhibitor remains but the FVIII, both activity and antigen, is eliminated. Heat treatment of plasmas has been reported by others [11, 12], but few data on the efficacy of this procedure have been provided. Verbruggen and colleagues [12] reported no change with heat treatment in inhibitor positive specimens. We, too, found little change in inhibitor positive plasmas, however, heating did appear to cause increase in some specimens that were inhibitor negative before heating. Five of 30 specimens from patients with positive HI which were <0.5 NBU rose into the positive range after heating. Only one of 159 patients with negative HI and none of the normal subjects increased to that range. This suggests that the modified method increases sensitivity to low titer inhibitors.

In the clinical setting, this modified assay allows testing of patients receiving replacement therapy when an urgent inhibitor titer is needed during emergency treatment or hospitalization, as well as inhibitor screening at routine clinic visits in patients receiving prophylaxis, ITI, or episodic care, when washout is not feasible or might place the patient at risk of breakthrough bleeding. It cannot, however, detect all inhibitors; some are not detected in clot-based assays, and some may be cleared in vivo [13]. The use of appropriate washout and pharmacokinetic studies is warranted when there is clinical suspicion of an inhibitor and a negative test. Alternative measurement methods, such as ELISA and fluorescence assays, may also prove useful.

Our data support the use of ≥0.5 NBU to define a positive inhibitor when the CDC method is used. A similarly large study of the unaltered Nijmegen method also provides data using that cutoff [5]. Assigning significance to inhibitor titers in the 0.1-0.4 range stretches the capability of the clot-based methodology. We and others [14] have demonstrated that some normal subjects show such low titers with both the BA and NBA. These assays assume that the curve of log %RA versus dilution is linear between 25 and 75 %RA, allowing quantitation of inhibitors in that range, which covers 0.4-2.0 BU. It is generally assumed in practice that the curve is also linear between 75 and 100 %RA, i.e., between 0 and 0.4 BU, however, it has been suggested that the sensitivity of the inhibitor assay does not extend below 0.4 BU (75 %RA), the limit of the range used for calculation in the original Bethesda assay, and that any inhibitor titer <0.4 BU should be considered negative [2, 11]. Our data are consistent with that concept. More than 98% of our patients without clinical history fall into that range. From a practical standpoint, there is sufficient variation in the VIII:C assay to make low titer inhibitor readings difficult to interpret. We have demonstrated mathematically that an inhibitor titer as high as 0.6 NBU is possible in a specimen that is actually negative. Using the regression equation for the NBU curve, it can be calculated that the difference in the VIII:C of the incubated mix varies by only 2.9 IU/dL per 0.1 NBU. The difference between a titer of 0 and one of 0.5 is only 14.5 IU/dL. Maintaining sufficient control of the test to achieve reproducible results in this range is challenging in many clinical laboratories. Other methods, such as chromogenic, ELISA, or fluorescence assays, may prove to be more sensitive and specific.

Evaluation of inhibitor test methods is difficult, because there is no gold standard against which to compare them. In practice, both laboratory and clinical evidence is used to determine whether a patient has an inhibitor. The inhibitor-negative population is defined largely by test, although non-neutralizing antibodies that are not detected in clot-based assays are suspected in some patients [13]. A significant portion of patients with past inhibitors test negative because they have been successfully treated, either by ITI or use of alternatives to factor replacement. A limitation of this study is the difficulty in defining the inhibitor and non-inhibitor populations, for which we have relied on clinical judgment. Another drawback is the possibility that heating and slow centrifugation of the patient specimens might have introduced some inhibitory substance causing prolonged clotting times in the patient mix but not in the control mix, which was not heated, leading to higher inhibitor titers. The data suggest that that is not the case. Only one patient with negative HI and negative titer and no normal subject developed a positive titer after heating, as compared to 16.7% of patients with positive HI (p=0.0004) in the direct comparison study, and only 7 positive titers (1.2%) occurred among a total of 588 tests on patients with negative HI at enrollment when the described method was used.

Our experience with inhibitor measurement in large numbers of specimens by a central laboratory demonstrates that the CDC modification of the NBA can be standardized to be within acceptable limits for clinical tests and is suitable for large-scale surveillance without regard for treatment status. Harmonization of regional or local laboratories for the same purpose could conceivably be accomplished by adoption of a detailed standard method with controls and uniform reagents.