Proposals for the measurement of anti-β2-glycoprotein I antibodies. Standardization Group of the European Forum on Antiphospholipid Antibodies

Authors


Guido Reber, Haemostasis Unit, Geneva University Hospital, 1211 Geneva 14, Switzerland.
Tel.: +41 22 372 9755; e-mail: guido.reber@hcuge.ch

In 1990, β2-glycoprotein I (β2-GPI) was shown to be the main protein cofactor involved in the binding of antiphospholipid antibodies to anionic phospholipids [1–3]. The presence of β2-GPI was mandatory for the measurement of antibodies of autoimmune origin in the widely used anticardiolipin antibody ELISA (aCL), because these antibodies are mainly directed to a β2-GPI–cardiolipin complex rather than to cardiolipin alone.

In 1995, antibodies directed to β2-GPI coated in the absence of phospholipids were reported [4]. These antibodies were associated with thrombotic events [4,5]. Since then, numerous studies have confirmed this association and this prompted many laboratories to measure anti-β2-GPI antibodies. In the late 1990s, commercial kits became available.

At the Sapporo consensus meeting [6], anti-β2-GPI antibodies were not included in the list of biological criteria defining the antiphospholipid syndrome (APS). However, at the standardization subcommittee meeting on lupus anticoagulants of the International Society on Thrombosis and Haemostasis (Boston, July 2002), it was proposed to include anti-β2-GPI antibodies in the list of APS biological criteria.

Several studies have shown poor agreement in aCL assay results obtained with either commercial [7] or home-made [8] ELISAs despite previous recommendations for assay design [8,9] and the use of unique reference calibrators available worldwide (University of Louisville calibrators). This was confirmed by external quality control surveys [10]. For anti-β2-GPI antibody assays, few studies have evaluated the agreement in assay results between commercial and also between home-made ELISAs.

In the framework of the European Forum on Antiphospholipid Antibodies, our standardization group addressed this issue in a multicenter study involving laboratories using commercial and home-made ELISAs [11]. Because of the poor agreement observed in low and medium positive samples, two additional multicenter studies were conducted by our group. In the first, samples, six calibrators [pool of patient samples, Forum Calibrators (FC)], humanized monoclonal antibodies (Mab), gift of T. Koike (Hokkaido University, Sapporo, Japan)[12] as well as a lyophilized human β2-GPI preparation were sent to laboratories using home-made assays. Using their local assay protocol, centers were asked to test the samples, the FC and six dilutions of the Mab in two microplates, one coated with their routine β2-GPI and the other one coated with the β2-GPI we sent. The calibration curves obtained with the FC and with the Mabs allowed the expression of the antibody content of the samples in common units (Forum Units and Mab equivalents in ng mL−1). The second study investigated the agreement between commercial kits. Fourteen samples, with the FC and eight Mab dilutions, were assayed with 10 commercial kits in three centers. The results of these studies were presented at the subcommittee meeting on lupus anticoagulants of the International Society on Thrombosis and Haemostasis (Birmingham, July 2003).

In summary, the results of our studies showed that the agreement was good for high positive samples, fair to poor for medium positive and poor for low positive. There was agreement in sample classification (positive or negative) between the 10 commercial kits in 12/22 samples for IgG isotype and in 5/22 for IgM isotype. Because no international reference calibrator for anti-β2-GPI assays is available, each kit expressed the results in its own arbitrary units, which made direct comparisons of the data impossible. Therefore, the optical density (OD) of the samples and the OD of the cut-off values of the assays were expressed in Forum Units using the calibration curve obtained with FC or in ng mL−1 equivalents of Mab using the calibration curve obtained with Mab dilutions. The cut-off values differed widely (up to five times) between assays. For the majority of commercial kits, no indication was available in the insert about the way to calculate the cut-off value. For the other kits, as well as for the majority of home-made assays, the mean value plus several standard deviations was employed.

Based on these results, our group propose the following recommendations to the manufacturers of kits for anti-β2-GPI antibody measurement, to laboratories using commercial kits and to laboratories performing home-made assays.

(1) To perform duplicates

As suggested by USA and European authorities and as stated in all kits tested so far, assays have to be performed in duplicate (calibrators, controls and samples). Although this may appear obvious, the relatively high coefficient of variation compared with the usual clinical chemistry automated assays has to be taken into account. This is of importance for low positive and borderline samples [11].

(2) Control group for the determination of the upper limit of the reference range (cut-off)

The minimal requirement to check the stated cut-off value is to test at least 50 (preferably 100) healthy normal individuals. The control group should include mainly women, because these antibodies are found more frequently in women. Each laboratory, even those using commercial kits, should determine its local cut-off value.

(3) Calculation of the cut-off value

Because the distribution of the values is not Gaussian, the cut-off, as for aCL antibodies [7], has to be calculated by the method of percentiles and not by adding standard deviations to the mean value [11]. When analyzing the distribution of the values of the control group, it must be kept in mind that anti-β2-GPI antibodies are found in 2–5% of healthy individuals, especially in individuals aged more than 60 [13]. Kit manufacturers are asked to indicate the units values corresponding to the 95%, 97.5% and 99% of the distribution of the control group in addition to the cut-off value they recommend. To fulfil regulation authority requirements, the proposed cut-off takes into account sensitivity and specificity data obtained from clinical studies.

(4) Humanized monoclonal antibodies as reference calibrators

Because there are still no worldwide accepted anti-β2-GPI calibrators, we propose to use Mab dilutions as calibrators to allow comparisons between assay systems.

Our group is aware that a Mab, because it is directed to one single epitope, cannot mirror the diversity of the subsets of patients' autoimmune polyclonal antibodies with different binding affinities, generally lower than Mabs. Mabs may not detect microheterogeneities in β2-GPI preparations because they possibly do not affect their binding, as would be the case for subsets of patient antibodies. However, the probability that a given patient pool contains all the antibodies present in all patients is low. To achieve this goal, the number of patients included in the pool should be very high.

Unlike the Louisville University calibrators used in all aCL assays, Mabs are not affected by batch-to-batch variation because their affinity does not change notably with time. This ensures that the stability of assay systems can be checked over many years. Mabs may be used to compare assay systems (as performed in our study), local and kit calibrators as well as day-to-day β2-GPI preparations. In addition, Mabs could be useful as positive controls and particularly as cut-off point markers. We propose that kit manufacturers agree to select one set of humanized Mabs and that they establish the conditions concerning their production and availability.

We ask manufacturers to compare their calibrators with Mab dilutions, to express the cut-off of their assay not only in their own arbitrary units but also in ng mL−1 Mab equivalents and to include in the insert of the kits a graph showing the calibration curves obtained with their own and the Mab calibrators.

In conclusion, although we are aware that the proposed recommendations will not resolve entirely the disagreement in assay results and sample classification observed in our studies, we think that they may improve the comparison between assay systems. In particular, we have to keep in mind that perfect standardization is an unrealistic goal, which implies that we have to try to improve, step by step, the agreement between assay results. As long as a better standardization has not been achieved, the incorporation of anti-β2-GPI assay into the APS criteria is questionable.

Ancillary