Antiphospholipid-associated thrombocytopenia or autoimmune hemolytic anemia in patients with or without definite primary antiphospholipid syndrome according to the Sapporo revised classification criteria: a 6-year follow-up study

Thrombocytopenia as a manifestation of the primary antiphospholipid syndrome (APS) has been reported with a prevalence between 30% and 46%.^1-3  Conversely, autoimmune hemolytic anemia (AIHA) and Evans syndrome have a frequency of 4% and 10%, respectively.^2,3  Recently, researchers from the Euro-Phospholipid project, which gathered 1000 patients with either primary or secondary APS, reported a prevalence of 29.6% of thrombocytopenia and 9.7% of AIHA.⁴  In a more recent follow-up study, Gómez-Puerta et al⁵  found that 48 of 128 patients with PAPS had thrombocytopenia (38%). Authors also reported that 11 (8%) of them evolved into systemic lupus erythematosus (SLE), 3 with thrombocytopenia, after a median disease duration of 8.2 years. In this study, hemolytic anemia and Coombs positivity conferred a significant risk for the subsequent development of SLE.⁵  Some authors have not found any clinical association between thrombocytopenia and other APS manifestation,⁶  whereas others have shown it is related with cardiac valve thickening, epilepsy, chorea, arthritis, and livedo reticularis.⁷ 

In 1989, in a study of 500 patients with SLE (with or without hematologic manifestations) Delezé and collaborators reported that patients with thrombocytopenia had greater levels and positivity of IgG anticardiolipin (aCL), whereas patients with AIHA were more frequently positive and had greater levels of the immunoglobulin M (IgM) isotype.⁸  These findings were later confirmed by several groups (most recently in a review by Uthman et al⁹ ). Our group also reported that the IgM aCL from SLE patients with AIHA cross-reacted extensively with phosphatidylcholine, the most abundant phospholipid of the outer leaflet of the erythrocyte membrane.¹⁰ 

The pathogenesis of thrombocytopenia related to antiphospholipid (aPL) antibodies is still unclear. It seems that these autoantibodies bind to activated platelets^11,12  via β₂-glycoprotein-I (β₂GP-I).¹³  Thrombocytopenia in APS also has been associated with antiplatelet glycoprotein antibodies that at times correlate more closely than either aPL or clinical manifestations.^14-16  Spontaneous and induced murine models of APS develop thrombocytopenia as one of their relevant disease features.^17-19  Interestingly, one of these models (NZW × NXSB) F1 has been proposed as an experimental model of idiopathic thrombocytopenic purpura.²⁰  In the case of human and mice autoimmune hemolytic anemia, IgM antiphosphatidylcholine antibodies have been linked to complement-dependent erythrocyte destruction,^21-23  being the NZB mice the spontaneous model of AIHA related to IgM anti-phosphatidylcholine autoantibodies.²⁴ 

In 1992, Alarcón-Segovia and collaborators²⁵  proposed that AIHA and thrombocytopenia could be considered part of a set of preliminary classification criteria of APS in patients with SLE. However, the Sapporo APS criteria recently revised in Sydney only include thrombosis or pregnancy morbidity as clinical criteria for the classification of APS.²⁶  Thus, a patient with aPL antibodies and a hemocytopenia without a history of thrombosis or pregnancy morbidity is not classified as having APS. As is known, these hematologic manifestations may antedate thrombosis or pregnancy morbidity. Even more, some authors have proposed that this subgroup of patients might indeed represent a prethrombotic state preceding the onset of APS.²⁷  In agreement with this notion, Diz-Küçükkaya et al²⁸  reported that 60% of aPL-positive patients diagnosed as having immune thrombocytopenic purpura (ITP) developed thrombosis in a 5-year follow-up study.

To better understand these notions, we studied patients with thrombocytopenia, autoimmune hemolytic anemia, or both with positive aPL according to the updated laboratory criteria for APS and compared them with hemocytopenic patients and definite APS.²⁶  The analysis of these findings is the objective of the current report.

We retrospectively reviewed the medical records of 187 consecutive, unselected patients with primary APS who from January 1986 to December 2008 visited the Department of Immunology of Rheumatology of the Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán, a tertiary referral care center. This registry includes patients who fulfilled the revised classification criteria for APS as well as patients with positive aPL and noncriteria clinical manifestations of APS.²⁶  We included patients with the diagnosis of thrombocytopenia defined as a platelet count of less than 100 × 10⁹ platelets/L confirmed in a least one occasion. Patients also were included if they had AIHA, which was defined as the presence of low hemoglobin, direct positive Coombs test, corrected reticulocyte count greater than 2%, elevated indirect bilirrubin, elevated lactate dehydrogenase, and low haptoglobin levels. If patients had both thrombocytopenia and AIHA, they were diagnosed as having Evans syndrome. We included patients who also fulfilled one or more of the laboratory criteria for APS according to the updated classification for APS.²⁶  In brief, patients with 2 or more positive determinations 12 weeks apart of immunoglobulin G (IgG) or IgM antibodies to cardiolipin (ie, aCL), or IgG or IgM antibodies to β₂-glycoprotein-I (anti–β₂GP-I) or positive lupus anticoagulant test (LAC). Patients were excluded from our study if they had positive anti-dsDNA antibodies as shown by enzyme-linked immunosorbent assay (ELISA), positive antinuclear antibodies with a peripheral pattern by immunofluorescence, or if they had 4 or more criteria for SLE according to the American College of Rheumatology classification criteria.²⁹ 

Patients' clinical records were carefully reviewed according to a preestablished protocol. We retrospectively collected demographic features, age at onset of the disease, age at the hematologic manifestation, and the presence and time of onset of other clinical manifestations, such as obstetric morbidity, arterial or venous thrombosis (confirmed by image diagnostic methods or biopsy), livedo reticularis, and skin ulceration. We also registered the time of follow-up until the last medical appointment, the use of corticosteroids, immunosupressors, aspirin, and splenectomy. Relapse was considered when after achieving response to treatment (any point at follow-up) it could not be maintained.

The group of patients was further subdivided into 2 subgroups. The first subgroup included patients who had both the serologic and clinical APS criteria (clinical APS).²⁶  The other subgroup was comprised by patients with the serologic criteria but without any obstetric or thrombotic event during follow-up (nonclinical APS).²⁶  All research reported was approved by the Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán institutional review board.

Response was defined if: (1) the platelet count was ≥ 100 00 μL at the last clinic visit when the initial platelet count was > 50 000 μL or (2) a platelet count increased of at least 20 000 from baseline to a final value greater than 50 000. We considered response to treatment in patients with hemolytic anemia when the hemoglobin levels reached 10 g/dL in the absence of hemolysis (normal corrected reticulocyte count, normal lactate dehydrogenase, and haptoglobin levels) at the last clinic visit.

aCL (IgG and IgM) was determined by ELISA according to published methods.³⁰  IgG and IgM antibodies to phospholipid-free human β2GP-I were determined by ELISA according to Cabiedes et al.³⁰  LAC was determined by LAC/1 screening reactant and a confirmatory test LAC/2 according to published guidelines.³¹  Cutoff points for aCL or anti–β₂GP-I at time of study were considered positive according to reference values of the immunology and rheumatology laboratory of our institution. These correspond to values greater than the 90th percentile of 100 normal control patients.³⁰ 

Categorical variables were compared using either the χ² or Fisher exact test when appropriate, and continuous variables were compared by the use of the Student t test or Mann-Whitney U test when non-normally distributed. Odds ratios (ORs) were reported with a 95% confidence intervals (CIs) for each antibody. A 2-tailed P < .05 was considered significant. All analyses were performed by the use of SPSS for Windows 17.0 (SPSS Inc).

We found 55 patients (44 women, 80%) with hemocytopenia who fulfilled 1 or more of the updated laboratory criteria for APS.²⁶  Patients had a mean age of 40 ± 12 years at the time of the hematologic manifestation and an overall median follow-up of 6.8 years (range, 0.14-37 years). Thirty-five patients (64%) had thrombocytopenia, 14 (25%) had AIHA, and 6 (11%) had Evans syndrome. In Table 1 we show that 25 patients (22 women, 88%) also fulfilled 1 clinical criteria for APS (13 thrombotic episodes and 12 pregnancy morbidities; 1 patient had both) after a median follow-up of 13.2 years (range, 1.45-37 years; clinical APS group) whereas 30 patients (22 women, 73%) did not develop a thrombotic event nor had pregnancy morbidities after a median follow-up of 5.4 years (range, 0.12-24 years, nonclinical APS group). The difference in follow-up of clinical versus nonclinical APS is statistically significant (P < .001). Table 1 also shows that the hematologic manifestations were equally distributed between the clinical APS and nonclinical APS groups. The response to treatment at the last clinic visit, the frequency of hematologic relapses, prednisone use, immunosuppressive treatment, aspirin use, and the frequency of splenectomy did not differ in patients regardless of their APS status. The basal platelet count and the basal and final hemoglobin levels were similar in both groups. Finally, platelets from clinical APS patients were slightly greater than those from nonclinical APS patients at their last medical appointment (244 500 vs 136 133 μL; P = .01; Table 1).

The time elapsed between the hematologic manifestation and the first APS clinical criterion is shown in Figure 1. As seen, the hemocytopenia preceded the clinical criterion in 6 patients (24%) with a median of 3 years (range, 0.5-10 years), whereas in 15 patients (60%), the hematologic manifestation appeared 4.4 years (range, 0-32 years) after they fulfilled the clinical APS criterion (Figure 1). In 4 patients (16%) the clinical APS and the hemocytopenia started simultaneously (data not shown).

The prevalence of aCL, anti–β₂GP-I, and LAC in patients with or without clinical APS is shown in Table 2. As per inclusion criteria, both groups of patients had high prevalence of all tested aPL; it ranged from 56% to 66% in the case of IgG aCL, whereas it was 84% to 96% for IgM aCL. Table 2 also shows that patients from the clinical APS group were more frequently positive for LAC (14/22, 63.7%) than patients with nonclinical APS (9/27, 30%). This difference is statistically significant (OR, 3.5; 95% CI, 1.07-11.4, P < .02). Because LAC-positive patients were concurrently positive for any aCL isotype in both groups of patients, the results are therefore the same than those from LAC-positive alone (Table 2). In the same table, we show that 8 of 21 (38%) patients from the clinical APS group and 7 of 20 (26%) from the nonclinical APS group were positive for IgG aCL, IgG anti–β₂GP-I, and LAC; the difference was not statistically significant. We also analyzed the frequencies of LAC in patients with or without definite APS by hematologic manifestation. These results are shown in Table 3. As shown, patients with thrombocytopenia, including 2 with Evans syndrome, from the clinical APS group were 2-fold more frequently positive for LAC (71%) than LAC-positive patients from the nonclinical APS group (35%; (OR, 4.46; 95% CI, 1.11-17.90; P = .05). In contrast, the frequencies of IgG anti–β₂GP-I (or any isotype) in patients with thrombocytopenia plus Evans syndrome or with AIHA, including 4 with Evans syndrome, regardless of their clinical APS status, were not significantly different (Table 3).

Finally, we also analyzed the frequencies of aPL in patients with hematologic manifestations regardless of their APS status. We found that patients with thrombocytopenia plus Evans syndrome had approximately 2-fold greater frequency of IgG anti–β₂GP-I (25/38, 66%) than patients without it (5/14, 36%). The difference was statistically significant (OR, 3.46; 95% CI, 1.00-12.48; P = .05). In contrast, the frequencies of aCL (IgG/IgM), anti–β₂GP-I (IgG/IgM), and the titers of these antibodies were not statistically significant among the studied groups (data not shown).

Since the early reports of APS, it has became clear that thrombocytopenia and hemolytic anemia are frequently found in these patients.⁹  For instance, Harris et al³²  found that high titers of IgG aCL have a 78% and 77% predictive value for thrombosis and thrombocytopenia, respectively, and only 44% for recurrent fetal loss. As mentioned previously, thrombocytopenia has been reported with prevalences between 30% and 46% in patients with primary APS.^1-3  Conversely, AIHA and Evans syndrome have a frequency of 4% and 10%, respectively.^2,3  Despite this finding, it is still controversial whether these hematologic manifestations are part of the APS.²⁶  Along these lines, our group has considered thrombocytopenia and/or AIHA as APS clinical manifestations since 1992.²⁵  We took advantage of this argument and hypothesized that there could be a group of patients who, despite fulfilling the revised Sydney laboratory criteria for APS,²⁶  would continue having thrombocytopenia and/or AIHA as their main and only clinical manifestation. Here, we report this group of patients and compare them with patients who, in addition to the cytopenias and the revised Sydney laboratory criteria, also met one accepted clinical APS criterion.²⁶  We found that the latter group had a longer follow-up (13.2 years) than the former (5.4 years). We do not have an explanation for this difference in follow-up, but we acknowledge that it could be a limitation of our study; in addition, if the follow-up of patients with clinical APS is reduced to ≤ 5.4 years (n = 20), only 5 of them (25%) would meet the corresponding clinical APS criterion (data not shown). It is unknown whether patients from the nonclinical APS group will fulfill a clinical criterion for APS after a follow-up time similar to their clinical APS counterparts. Should this happen, it would nevertheless support our contention that thrombocytopenia and/or AIHA in patients with the appropriate serologic setting comprise a distinct subgroup of APS clinical manifestations.

As per inclusion criteria, we studied a highly homogenous group of patients with high frequencies of aCL, anti–β₂GP-I, and LAC. This is underscored by the finding that the frequency of the 3 serologic markers was similar throughout our studied patients despite their APS status. Despite this finding, patients with definite APS were more frequently positive for LAC than those without it, regardless of their hematologic manifestation. Interestingly, when patients with thrombocytopenia were compared with those without, the strength of the association of LAC for clinical APS was even stronger (Table 3). This finding is in agreement with the notion that LAC is a strong risk factor for thrombosis.^33,34  Our study also demonstrated that except for thrombosis or pregnancy morbidity, hemocytopenic patients with or without clinical APS had the same global response to treatment, number of relapses, prednisone, aspirin or use immunosupressive use, and frequency of splenectomy. As stated previously, the only difference for which we were able to account between clinical and nonclinical APS patients was their aPL profile.

Here, we showed that a group of patients with hematologic manifestations and positive aPL who did not develop thrombosis or any clinical and/or serologic features of SLE during follow-up. Noteworthy, this thrombosis-free state occurred in patients with IgG anti–β₂GP-I antibodies, which, as shown previously, not only confer high risk for thrombosis³⁵  but also are frequently responsible for the in vitro lupus anticoagulant activity.³⁶  Taken together, our data thus suggest that IgG anti–β₂GP-I comprise heterogeneous prothrombotic and nonprothrombotic (thrombocytolytic?) subpopulations of antibodies. Whether the heterogeneity of anti–β₂GP-I is the result of differences in epitope recognition, as has been clearly shown by de Laat et al^37-39  and by our group,⁴⁰  is currently unknown.

Here, we could not confirm the report by Pullarkat et al⁴¹  that patients with AIHA and positive LAC are more prone to developing thrombosis than LAC-negative patients. Whether this is attributable to the small number of patients with AIHA included in our study also is not known.

A study of patients with isolated immune thrombocytopenia may shed some light regarding the pathogenesis of aPL in this autoimmune manifestation. In 742 patients with ITP from 9 published series, the frequency of aPL varied from 25% to 67% (mean, 43%).^9,32,42  In only 4 of these reports did the authors address the issue of whether aPL confers a risk for thrombosis in patients with ITP.^28,42-44  In a prospective study of 149 patients with ITP, Stasi et al⁴³  reported that 69 (46%) were positive for aPL. These authors found that after a mean follow-up of approximately 2.5 years, none of their patients developed thrombosis despite the unusually high frequency of LAC (44%) compared with normal healthy control patients (3/174, 1.7%).⁴³  These results are in sharp contrast with the findings of Diz-Küçükkaya et al, who after a 5-year prospective study of 82 patients with ITP, found that 14 of 31 patients (45%) with aPL developed APS.²⁸  These researchers also reported that after follow-up, only 33% of LAC-positive patients remained thrombosis-free compared with 96% of LAC-negative patients (P < .0001).²⁸  In another retrospective study of 216 patients with ITP, Pierrot-Deseilligny et al⁴⁴  found that 14 of 55 (25%) aPL-positive patients developed thrombosis after 2.5 years of disease duration. These investigators also reported that LAC was an independent risk factor for thrombosis with a hazard ratio of 3.1 and 9.9 in the univariate and multivariate analysis, respectively.⁴⁴  Finally, in an earlier study of 27 patients, Funauchi et al⁴²  reported that 5 of 7 (71%) ITP patients with aPL had a tendency for recurrent fetal loss and thrombosis. From these studies totaling 623 patients with ITP, we can conclude that 33 of 48 (69%) patients with thrombosis had serum aPL with a sensitivity of 0.69 (95% CI, 0.54-0.81) and a specificity of 0.72 (95% CI, 0.68-0.76) of aPL for thrombosis in ITP (OR, 5.71; 95% CI, 2.90-11.34; P < .0001). As stated previously, in the current work approximately one-half of our aPL-positive patients with thrombocytopenia developed APS, whereas one-half of them remained free of thrombosis and/or pregnancy morbidity. It appears then that depending upon their aPL profile, patients with hematologic manifestations belong to a subset of patients with APS, some develop thrombosis during follow-up whereas some continue having thrombocytopenia or hemolytic anemia as isolated clinical manifestations. These patients could be dubbed as having “hematologic APS,” more so in the absence of clinical and serologic features of SLE.

It has been long known that LAC confers a 4.5% person/year for thrombosis.^33,36,45  Our study and those mentioned previously were not designed to address whether or not patients with hemocytopenia and who are positive LAC should receive anticoagulant therapy to prevent or reduce the risk of thrombosis. A prospective, multicentric, multidisciplinary study is needed to settle this important issue.

We dedicate this work to the memory of Dr Javier Cabiedes, whose death leaves a great void for those of us who had the privilege of knowing him as a collaborator and friend.

Contribution: L.C.-K. performed research and collected, analyzed, and interpreted data; G.H.-M. designed research, analyzed and interpreted data, performed statistical analysis, and wrote the manuscript; and A.R.C. designed research, analyzed and interpreted data, performed statistical analysis, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Antonio R. Cabral, Department of Immunology and Rheumatology, Vasco de Quiroga 15, Delegación Tlalpan, Mexico DF 014000; e-mail: arcabral1952@yahoo.com.mx.