In patients with thrombosis or pregnancy complications, a persistently positive antiphospholipid antibody (aPL) (lupus anticoagulant test (LA), anticardiolipin antibodies (aCL), and anti-β₂ glycoprotein I antibodies (anti-β₂GPI)) establishes a classification of definite antiphospholipid syndrome (APS).¹ In a clinical setting, physicians judging aPL tests face several challenges. First, aPL in individual patients may vary over time, though how much spontaneous variation occurs is unknown. Second, it is also unknown whether any variation that does occur reflects autoimmune disease activity, drug treatment, or interlaboratory differences. Third, interlaboratory correlation among aCL results is not well established—a point particularly important in the USA where, because of rapidly changing insurance coverage systems, physicians can no longer specify the laboratories in which their patients’ specimens are tested.

The primary objective of this study was to determine the stability and the degree of variation of aPL results over time in a large cohort of well evaluated aPL positive patients. Second, we analysed factors we considered likely to contribute to aPL variation, and the validity of aPL in a real world setting in which aPL tests are done in multiple laboratories. The results should offer guidance to practising physicians and researchers in the management of aPL positive patients.

METHODS

We identified aPL positive patients from those entered into two databases: a national antiphospholipid syndrome collaborative registry (APSCORE)²; and an asymptomatic (no history of vascular or pregnancy events) aPL positive registry (APLASA).³ Inclusion criteria were: positive aPL (aCL or LA test or both, on two occasions six weeks apart), with or without APS classification (based on the Sapporo criteria) for APSCORE; and positive aPL without APS classification for APLASA. We reviewed medical and registry records for all available aPL tests (LA test, aCL, and anti-β₂GPI) including testing dates and laboratories. In addition, we also reviewed demographic variables, definite and possible aPL related clinical manifestations (venous and arterial thrombosis, pregnancy morbidity, livedo reticularis, thrombocytopenia, and migraine), coexisting autoimmune diseases, comorbidities (hypertension requiring antihypertensive agents, diabetes mellitus requiring antidiabetic agents, hypercholesterolaemia requiring cholesterol lowering agents, and current smoking), and drug treatments.

Lupus anticoagulant tests were grouped as positive and negative, based on the guidelines of the International Society on Thrombosis and Hemostasis.⁴ The aCL and anti-β₂GPI results were expressed as their immunoglobulin subclasses (aCL: IgG, IU or GPL U/ml; IgM, IU or MPL U/ml; and IgA, IU or APL U/ml; anti-β₂GPI: IgG, U/ml; IgM, U/ml; and IgA, U/ml). They were divided into four groups: 0–19 U (negative); 20–39 U (low positive); 40–80 U (moderate positive); and >80 U (high positive). We classified aPL positive patients into three groups: vascular events with or without pregnancy events (Sapporo APS classification criteria were met); pregnancy events only (Sapporo APS classification criteria were met); and asymptomatic (Sapporo APS classification criteria were not met; patients with solely non-Sapporo aPL manifestations such as livedo reticularis or cardiac valve disease were also included in this group).

We arbitrarily selected the initial aPL result as the anchor. Stability of aPL was defined as the percentage of subsequent results in the same group; variation of aPL was defined as the percentage of subsequent results in different groups. The relations between clinically relevant variation (positive v negative for LA test, negative to low v moderate to high titre for aCL and anti-β₂GPI tests) and aPL related clinical manifestations or drug treatment (aspirin, warfarin, hydroxychloroquine) were analysed (χ² test) independently of other manifestations of autoimmune disease activity.

To examine interlaboratory variation, we identified aPL positive patients who had same specimen aCL testing from different commercial suppliers and from our immunology laboratory. Same specimen aCL results from different sources were analysed for: consistency (percentage of results within the same group); agreement between aCL groups (Cohen’s κ test); and correlation between aCL results (Spearman rank correlation test to test the direction and strength of the relation).

RESULTS

Two hundred and four patients had positive low-medium-high titre aCL or LA test or both; table 1 gives their demographic and clinical characteristics. Fifty seven had had vascular events (21 venous only, 30 arterial only, and six venous and arterial). Table 2 gives the clinical characteristics and drug treatment of aPL positive patients who had vascular or only pregnancy event, or were asymptomatic at the time of the study entry. Patients with vascular events more often had livedo reticularis, migraine, hypertension, hypercholesterolaemia, and concomitant LA and aCL positivity (p<0.05).

We identified 1652 aPL tests between 1984 and 2004: 387 LA tests (61% done at our institution and 39% at other laboratories); 1097 aCL IgG/M/A tests (58% done using our in house assay, 24% at Quest Diagnostics, 12% at other laboratories, and 6% by the APSCORE assay); and 168 anti-β₂GPI IgG/M/A tests (54% done at Quest Diagnostics, 37% by the APSCORE assay, and 9% at other laboratories) (table 3).

Of 159 patients tested for LA, 96 (60%) had more than one LA test (total / mean (SD) / median number of LA tests in 96 patients: 324 / 3.5 (1.8) / 3; mean follow up time, 2.4 years). Fifty one initial tests were positive, of which 37 (73%) remained positive based on the final test (mean follow up, 2.1 years; mean number of repeat tests, 2.2); 39 of 51 patients (77%) with an initial positive LA test had persistent LA positivity and 82 of 94 subsequent tests (87%) from 51 patients were positive. Forty five tests were initially negative, of which 37 (82%) remained negative based on the final test (mean follow up, 2.7 years; mean number of repeat tests, 2.7); 40 of 45 patients (89%) with an initially negative LA test had persistent LA negativity, and 121 of 134 subsequent tests (90%) from 45 patients were negative.

One hundred and seventy nine of 204 patients (88%) had more than one aCL test (total / mean (SD) / median number of aCL tests in 179 patients: 1072 / 6.2 (4.7) / 4; mean follow up time, 3.5 years). The highest isotype of the initial test was IgG, IgM, and IgA for 106, 54, and 19 persons, respectively (IgG and IgM for three (analysed as IgG), and IgM and IgA for two (analysed as IgM)). Nine patients had isolated aCL IgA positivity (LA status unknown in four). Table 4 shows the distribution of patients based on the highest initial aCL test; the change in the aCL groups based on the final aCL test; and the stability of aCL tests over time. The combined stability of aCL was 88% for negative to low positive and 75% for moderate to high positive groups. Furthermore, 71% of patients with an initial negative to low positive aCL result and 62% of patients with an initial medium to high positive aCL result had all of the subsequent aCL results in the same range.

One hundred and nine persons had anti-β₂GPI tests but only 27 (13%) had more than one (total / mean (SD) / median number of anti-β₂GPI tests in 27 patients: 87 / 3.2 (1.5) / 2; mean follow up time, 1.0 years). The highest isotype of the initial test was IgG, IgM, and IgA for 11, 9, and 1 persons, respectively (six were reported as negative for all the isotypes). Table 5 shows the change in the anti-β₂GPI for patients with more than one test. Ninety six per cent of subsequent tests in patients with an initial anti-β₂GPI titre of <40 U/ml and 76% of subsequent tests in patients with an initial anti-β₂GPI titre of ⩾40 U/ml remained in the same group. Furthermore, 83% of patients with an initial negative to low positive anti-β₂GPI result and 67% of patients with an initial medium to high positive anti-β₂GPI result had all of the subsequent anti-β₂GPI results in the same range.

Excluding asymptomatic patients participating in a trial testing low dose aspirin v placebo, aspirin, warfarin, and hydroxychloroquine use did not differ between persons whose aCL titres decreased from medium to high to negative to low (n = 26), increased from negative to low to medium to high (n = 13), or remained stable (n = 65) (p = 0.93, p = 0.13, and p = 0.83, respectively). Similarly, definite and possible aPL related clinical manifestations were not statistically different between the patients whose aCL titres decreased, increased, or remained stable. The diagnosis of SLE was present in 35%, 62%, and 41% of patients whose aCL titres decreased, increased, or remained stable, respectively (p = 0.2). When 91 SLE patients were compared with 51 patients with no connective tissue disorders, there was no statistical difference between the number of patients whose aCL titres decreased (12% v 22%, p = 0.25), increased (12% v 6%, p = 0.34), or remained stable (76% v 72%, p = 0.72).

Thirty one patients had same specimen aCL tests by our institution’s in house assay and APSCORE; 75 by our in house assay and Diamedix Diagnostics aCL kit; and 77 by our in house assay and Inova Diagnostics aCL kit. Table 6 shows the consistency, agreement, and correlation between our institution’s in house assay and other aCL kits. The consistency ranged from 64% to 88% and the correlation ranged from 0.5 to 0.8. Agreement was moderate for aCL IgG and aCL IgM; agreement for aCL IgA was marginal.

DISCUSSION

Our data show that aPL results remain stable for at least three quarters of subsequent tests during a mean follow up of 2.4 years for the LA test, 3.5 years for the aCL test, and 1.0 year for the anti-β₂GPI test. The small amount of variation that occurs does not appear to result from aspirin, warfarin, or hydroxychloroquine use.

The Sapporo criteria require, without definition, medium to high titre aCL is to make the diagnosis of APS.¹ Different clinical laboratories define “medium to high titre aCL” as more than 15–20 international units (IU), 2.0–2.5 times the median, or the 99th centile of normal population titres. In the absence of consensus among laboratories⁵ or APS experts⁶ about what constitutes medium to high titre aCL, several studies concluded that more than 40 IU or GPL U/ml is more predictive of thrombotic events^7,8,9,10,11 than lower titres. We used a cut off aCL limit of 40 U for “medium to high titre aCL” and found that, regardless of the laboratory carrying out the test, if the initial aCL result was less than 40 U, the probability of obtaining a repeat study within the same range in the next 3.5 years was 88%. If the initial aCL result was equal to or more than 40 U, the probability that a second test will also be in the same range was 75%. Antiphospholipid antibodies bind primarily to the negatively charged phospholipids through the phospholipid binding plasma protein β₂GPI, and some APS patients may have only anti-β₂GPI antibodies.^12,13 Despite the small number of patients, we also found that anti-β₂GPI results of less than or more than 40 U/ml were stable in at least three quarters of subsequent tests.

Similarly, at least three quarters of LA tests stayed either positive or negative. Although our study was not primarily designed to test this question, we found that concurrent positivity of LA and aCL tests, traditional thrombosis risk factors, livedo reticularis, and migraine were more common in APS patients with vascular events than in aPL positive asymptomatic persons. These results are consistent with other studies showing that a positive LA test is more commonly associated with thrombosis than is aCL,¹⁴ and that dual positivity of aCL and LA test is more commonly associated with thrombosis than is single positivity of aCL or LA test.¹⁵

Infection induced transient aPL positivity is common in the general population and usually not pathogenic. Vila et al analysed 552 healthy donors for the prevalence of aPL and found that 86% of positive IgG aCL and 97% of positive IgM aCL tests were negative in one year.¹⁶ As our study population derives from registries that require two positive aPL results as inclusion criteria, our results do not apply to patients with single aPL positivity.

Although McCarty et al reported that aPL decreased in patients treated with hydroxychloroquine 200 mg twice daily and aspirin 81 mg once daily,¹⁷ we did not observe any relation between aspirin, warfarin, or hydroxychloroquine treatment and change in aCL titre. The retrospective nature of the study did not allow us to analyse corticosteroid or immunosuppressive drug use accurately.

In order to determine the possible contribution of interlaboratory differences to aPL variation, we analysed the degree of variation between aCL kits. Based on same day specimens, the consistency of aCL results among different suppliers ranged from 64% to 88%, with moderate agreement for IgG and IgM. The agreement for aCL IgA was marginal. Although interlaboratory agreement is better when positive aCL results are compared in semiquantitative measures (ranges of positivity),¹⁸ others also found medium agreement between kits especially for IgG aCL.¹⁹ We did not analyse interlaboratory variations between anti-β₂GPI antibody kits; however, Reber et al showed that anti-β₂GPI antibody kits are also poorly standardised, especially at lower titres, and the agreement is better with medium to high titre IgG anti-β₂GPI.²⁰ Thus physicians should be conservative in their interpretation of aCL and anti-β₂GPI levels in individual patients when results from multiple laboratories are available, a common scenario in the USA. The standardisation of aCL and anti-β₂GPI assays can still be improved to provide better agreement, a point also noted by others.²¹ Nonetheless, although interlaboratory differences might have contributed to the degree of aPL variation over time, we found that at least three quarters of aPL results were stable in the range of negative to low and moderate to high over time.

Our study has several limitations. It could not evaluate the effect of SLE activity on aPL variation, a point argued by some investigators²² but contradicted by others.²³ Second, our titre groupings were arbitrary but consistent with clinical experience. Within these ranges, aPL variation was low. However, the clinical significance of our cut off points requires further verification. Third, aPL tests from multiple laboratories were included in our analysis; however, this was intentional in order to simulate a real world experience in which multiple and visit to visit comparisons must rely on studies done in different laboratories owing to rapidly changing insurance coverage systems. Lastly, the study did not evaluate aPL variation beyond three to four years, and it is possible that more variation may be observed with longer follow up.

Conclusions

Repeat aPL results remain stable for at least three quarters of subsequent tests regardless of the laboratory carrying out the test. The variation that occurs does not appear to result from aspirin, warfarin, or hydroxychloroquine use. Although it is unknown if and what level of aPL variation has clinical significance, our findings should offer guidance to practising physicians and researchers in the management of aPL positive patients.