To investigate which serologic and clinical findings predict adverse pregnancy outcome in patients with antiphospholipid antibody (aPL) and to test the hypothesis that a pattern of clinical and serologic variables can identify women at highest risk of adverse pregnancy outcome.
Women enrolled in a multicenter prospective observational study of risk factors for adverse pregnancy outcome in patients with aPL (lupus anticoagulant [LAC], anticardiolipin antibody [aCL], and/or antibody to β2-glycoprotein I [anti-β2GPI]) and/or systemic lupus erythematosus (SLE) were recruited for the present prospective study. Demographic, clinical, serologic, and treatment data were recorded at the time of the first study visit. The relationship between individual and combined variables and adverse pregnancy outcome was assessed by bivariate and multivariate analysis.
Between 2003 and 2011 we enrolled 144 pregnant patients, of whom 28 had adverse pregnancy outcome. Thirty-nine percent of the patients with LAC had adverse pregnancy outcome, compared to 3% of those who did not have LAC (P < 0.0001). Among women with IgG aCL at a level of ≥40 units/ml, only 8% of those who were LAC negative had adverse pregnancy outcome, compared to 43% of those who were LAC positive (P = 0.002). IgM aCL, IgG anti-β2GPI, and IgM anti-β2GPI did not predict adverse pregnancy outcome. In bivariate analysis, adverse pregnancy outcome occurred in 52% of patients with and 13% of patients without prior thrombosis (P = 0.00005), and in 23% with SLE versus 17% without SLE (not significant); SLE was a predictor in multivariate analysis. Prior pregnancy loss did not predict adverse pregnancy outcome. Simultaneous positivity for aCL, anti-β2GPI, and LAC did not predict adverse pregnancy outcome better than did positivity for LAC alone.
LAC is the primary predictor of adverse pregnancy outcome after 12 weeks' gestation in aPL-associated pregnancies. Anticardiolipin antibody and anti-β2GPI, if LAC is not also present, do not predict adverse pregnancy outcome.
Antiphospholipid antibodies (aPL, which include lupus anticoagulant [LAC], IgG and IgM anticardiolipin antibody [aCL], and IgG and IgM anti–β2-glycoprotein I antibody [anti-β2GPI]) (1), are closely associated with pregnancy complications, but many women with aPL have normal pregnancies. Whether there are specific serologic or clinical findings, such as associated systemic lupus erythematosus (SLE), that can predict which patients with aPL are most likely to experience adverse pregnancy outcome is controversial. Different retrospective studies have suggested different antibody profiles, e.g., simultaneous presence of LAC, aCL, and anti-β2GPI, presence of anti-β2GPI alone, and presence of other antibody combinations, as identifiers of at-risk patients; other studies suggest that clinical characteristics such as SLE or prior pregnancy loss are predictive, and there is no consensus (2–14). Many physicians test all pregnant patients for aPL and treat those who test positive regardless of how positivity is defined. Identification of specific predictors of high risk of adverse pregnancy outcome would allow targeting of therapy to those most likely to benefit. This study was undertaken to determine whether there is a factor or combination of factors that is associated with increased risk of adverse pregnancy outcome in this patient population.
PATIENTS AND METHODS
Definitions, patient selection, and monitoring.
The PROMISSE study (Predictors of Pregnancy Outcome: Biomarkers in Antiphospholipid Antibody Syndrome and Systemic Lupus Erythematosus study) is an ongoing multicenter, National Institutes of Health (NIH)–funded prospective observational study of pregnancies in women with aPL, SLE, or both; healthy pregnant controls are included as well. Patients are monitored monthly throughout their pregnancies and data and samples collected. For the present study we enrolled the subset of PROMISSE study participants who were positive for aPL at any titer and who delivered between September 2003 and March 2011.
At 7 study sites, consecutive pregnant women referred because of suspected aPL positivity, SLE, or both, as well as healthy controls, were recruited. Gestations of eligible patients had to be 12 weeks or earlier (SLE and normal) or 18 weeks or earlier (aPL patients, of whom 41 of 144 were enrolled between weeks 12 and 18). Healthy pregnant controls, selected among women with no known illness, no prior fetal loss, no more than one embryonic loss, and at least one successful pregnancy, were tested in parallel with the patient groups. One hundred sixteen women declined to participate.
The choice to exclude patients whose pregnancies ended before 12 weeks was based on the difficulty of identifying and qualifying patients so shortly after diagnosis of pregnancy and the high frequency of early miscarriage due to chromosome errors or unidentifiable causes among normal women (8, 11), which would markedly increase the resources needed to screen such candidates. The choice to include aPL-positive patients up to 18 weeks of gestation was made at the request of the NIH-appointed Observational Study Monitoring Board after the first interim analysis revealed slow recruitment due in part to exclusion of pregnancies after 12 weeks and because very few adverse pregnancy outcomes occurred in this interval. SLE was diagnosed according to the American College of Rheumatology criteria for classification of the disease (15). Derivation of the study groups is depicted in Figure 1.
Inclusion criteria were as follows: confirmed positive aPL at 2 or more separate time points and at least 1 positive result from testing in the core laboratories (see below) during pregnancy, live intrauterine pregnancy confirmed by ultrasonography, age 18–45 years, ability to provide informed consent, and hematocrit level >26%. Exclusion criteria (chosen so that other non-aPL causes of adverse pregnancy outcome would not confound the findings) were treatment with prednisone >20 mg/day, urine protein ≥1,000 mg in 24 hours or protein:creatinine ratio ≥1,000 mg protein/gm creatinine on spot urine collection, erythrocyte casts seen on urinalysis, serum creatinine >1.2 mg/dl, type 1 or 2 diabetes mellitus antedating pregnancy, blood pressure ≥140/90 mm Hg at the screening visit, and multifetal pregnancy. Patients were evaluated monthly by an obstetrician and each trimester by a rheumatologist through 3 months postpartum, with physician examination, questionnaires, obstetric ultrasonography, and laboratory testing.
At the time this study was initiated, the 1999 (Sapporo) criteria for the classification of antiphospholipid syndrome (16) required that positive test results be replicable at intervals of ≥6 weeks and that aCL and anti-β2GPI be present in moderate to high titer (≥40 units). In a 2006 revised consensus conference report, the interval was changed to 12 weeks, and it was suggested that anti-β2GPI be reported as percentiles (1); however, because the study was ongoing, we retained the 1999 definitions.
The Institutional Review Board at each of the PROMISSE study sites approved participation of patients. All participants provided written informed consent.
Adverse pregnancy outcomes.
Adverse pregnancy outcome was defined as an otherwise unexplained fetal death after 12 weeks' gestation; neonatal death prior to discharge, associated with complications of prematurity; preterm delivery prior to 34 weeks because of gestational hypertension, preeclampsia, or placental insufficiency; and small for gestational age (birth weight below the fifth percentile, as determined using the smoothed percentiles of birth weight by gestational age described by Alexander et al) (17–20). Obstetrician members of the PROMISSE team adjudicated causes of fetal death in equivocal cases.
Antiphospholipid antibody tests.
In this report we present only data from the core laboratories that were obtained at the screening visit; analyses from other time points did not affect the conclusions. Blood samples were processed immediately, shipped on dry ice, and maintained frozen at −80°C until tested.
The LAC core laboratory in Toronto is directed by one of the authors (CAL). LAC was detected using a panel of 3 tests (21), i.e., the dilute Russell's viper venom time (dRVVT) test, an LAC-sensitive test for activated partial thromboplastin time (APTT), and the dilute prothrombin time (dPT) test, according to previously published methods that fulfill criteria proposed by the International Society on Thrombosis and Haemostasis (22). Based on mixing studies and phospholipid dependence, LAC is determined to be present or absent. LAC values measured at the core laboratory throughout pregnancy and postpartum did not change after initiation of either low molecular weight or unfractionated heparin therapy, and addition of heparinase to plasma from patients receiving heparin confirmed the lack of effect of heparin on LAC. Testing of plasma samples after addition of appropriate amounts of heparin further demonstrated that heparin treatment did not cause false-positive LAC results, and no patient developed LAC upon introduction of either low molecular weight or unfractionated heparin. The reagents used to screen and confirm LAC contain Polybrene, which inactivates heparin at 1–2 units/ml, which is higher than the typical therapeutic range of 0.3–0.7 units/ml. No patient was receiving warfarin at the time of testing.
Anticardiolipin antibody was tested, as previously described (23) and validated in international standardization workshops (24), by cardiolipin enzyme-linked immunosorbent assay at the laboratory of one of the authors (JES) in New York. Normal ranges for the individual tests are based on a maximum of 3 SD above the mean value in a panel of normal controls and are as follows: IgG aCL 0–20 units/ml, IgM aCL 0–10 units/ml.
Anti-β2GPI was tested at the laboratory of one of the authors (JM) in Oklahoma City. A Nunc MaxiSorp microtiter plate was coated with β2GPI antigen diluted in borate buffered saline (BBS; pH 9.6) and incubated overnight at 4°C. A 6-point standard curve was prepared using an in-house calibrator initially diluted 1:100 in 0.5% bovine serum albumin/BBS/1.2% Tween 80. The plate was then washed twice with BBS. Test samples at 1:100, quality control samples (known positive and negative) at 1:100, and standard curve serial dilution serum samples were added (final volume for each sample 200 μl/well). All samples were tested in duplicate. The plate was incubated for 2 hours at room temperature. At the end of incubation the plate was washed a further 3 times, after which 100 μl of working-strength detection antibody (alkaline phosphatase–conjugated goat anti-human IgG or IgM; Sigma) was added to each well, with incubation at room temperature for a further 90 minutes. Three more washes were performed, and the reaction was visualized by the addition of diethanolamine substrate buffer (75 μl/well) containing p-nitrophenyl phosphate (Sigma). Absorbance at 405 nm was read on a Bio-Tek ELx808 automated microplate reader (25). Normal ranges are based on a maximum of 2 SD above the mean value in 60 normal controls and are as follows: IgG anti-β2GPI 0–25 units/ml, IgM anti-β2GPI 0–25 units/ml. Concordance between results of anti-β2GPI determinations at the core laboratory, simultaneously performed assays at the Hospital for Special Surgery, and commercially performed assays by Quest Diagnostics was >80%. In accordance with international criteria for both IgG and IgM aCL and IgG and IgM anti-β2GPI aPL (1), levels of ≥40 units/ml were considered high-titer positive, and levels that were positive but <40 units/ml were considered low-titer positive
The patients' physicians made all treatment decisions without knowledge of results obtained at the core laboratories. Of the 75 patients who were being treated with heparin at the first study visit, 56 were receiving low molecular weight heparin and 19 were receiving unfractionated heparin. The heparin was prescribed at therapeutic doses in 36 cases and at prophylactic doses in 39. Of the 97 patients who regularly took aspirin (Table 1), all but 1 took 81 mg per day; the other took 325 mg per day.
Table 1. Adverse pregnancy outcome in patients with aPL, according to demographic, clinical, serologic, and treatment characteristics at the first study visit*
Fisher's exact test was used to evaluate bivariate associations between categorical variables. Multivariate analyses were performed by fitting a Poisson regression model with a robust error variance to estimate the relative risk posed by an exposure variable adjusted for the effects of other covariates, and the corresponding 95% confidence interval (95% CI) (26). Variables that were predictive of adverse pregnancy outcome in bivariate analyses (P < 0.20) as well as those deemed to be associated with adverse pregnancy outcome a priori based on clinical factors were considered for inclusion in the model. The final model was determined using a backward selection approach and included only those covariates that remained significant at the P< 0.05 level. An internal validation of the predictive model was also performed using a 5-fold cross-validation procedure. The model was re-fit at each step of the validation process using the training data, and evaluated on the corresponding test set. Results of the validation procedure are reported as the sensitivity, specificity, and accuracy of the model averaged over all test sets.
Eight hundred six pregnant women were recruited for the PROMISSE study (610 referred for pregnancies associated with aPL, SLE, or both, and 196 referred as normal controls) (Figure 1). Ninety-two of the women with aPL and/or SLE and 36 of the controls were found at the initial screening to meet exclusion criteria. Positivity for aPL was confirmed at the core laboratories in 144 patients, of whom 87 had aPL only and 57 had aPL and SLE (Figure 1 and Table 1). Six screened patients with confirmed aPL and 3 in whom aPL positivity was suspected but not confirmed at the core laboratories experienced pregnancy loss before 12 weeks and were excluded from analysis. Two of the women recruited as normal controls repeatedly had positive results on aCL testing and were added to the aPL group.
Antiphospholipid antibody results and adverse pregnancy outcome.
Twenty-eight (19%) of the 144 patients who were positive for aPL at screening had adverse pregnancy outcome, compared to 3 (2%) of the 158 normal controls (P < 0.0001). Sixty-four aPL-positive patients had LAC, (Table 1), of whom 25 (39%) had adverse pregnancy outcome, compared with 2 (3%) of 76 women who were LAC negative (P < 0.0001) (LAC data on 1 patient were missing at screening). Of the 66 women with IgG aCL values ≥40 units/ml, 19 (29%) had adverse pregnancy outcome, whereas only 9 of 77 (12%) with IgG aCL values <40 units/ml had adverse pregnancy outcome (P = 0.01) (Table 1). All 9 women with adverse pregnancy outcome whose IgG aCL levels were <40 units/ml were LAC positive; among those with IgG aCL levels ≥40 units/ml, the adverse pregnancy outcome rate was 8% in the LAC-negative group, compared to 43% in the LAC-positive group (P = 0.002). IgM aCL levels ≥40 units/ml and IgG or IgM anti-β2GPI levels ≥40 units/ml did not predict adverse pregnancy outcome, nor did positivity for a combination of serologic features have better predictive power than did positivity for LAC alone (Figure 2). Analysis for aPL positivity at other times in the pregnancy, or for consistent versus transient positivity, did not alter the results. APTT, dRVVT, and dPT were similar in their abilities to predict adverse pregnancy outcome (Table 2). In 23 patients, the dPT was the only one of these tests to yield a positive result; 6 of these 23 patients had an adverse pregnancy outcome.
Table 2. Association of LAC positivity on different screening tests with adverse pregnancy outcome*
Clinical and demographic features and adverse pregnancy outcome.
Twenty-five of the aPL-positive patients had prior thromboses (4 arterial only, 14 venous only, 5 arterial and venous, and 2 unknown). Thirteen of these 25 patients (52%) had adverse pregnancy outcome, compared with 15 (13%) of 119 without prior thromboses (P = 0.00005). In bivariate analyses, prior fetal loss and nonwhite race did not distinguish between patients with and those without adverse pregnancy outcome (Table 1). All 5 patients who had LAC, SLE, and prior thrombosis had adverse pregnancy outcome, compared with 1 of 38 patients who lacked all 3 of these characteristics. SLE did not affect the adverse pregnancy outcome rate in bivariate analyses, nor did treatment with corticosteroids or hydroxychloroquine. The rate of adverse pregnancy outcome was lower among patients who were regularly receiving aspirin at the time of the first study visit, and higher among those who were receiving heparin (Table 1).
Among the 28 patients with adverse pregnancy outcome, the outcomes were as follows: fetal death (n = 14), neonatal death (n = 1), delivery before 34 weeks (n = 16), preeclampsia or gestational hypertension leading to delivery before 34 weeks (n = 12), placental insufficiency (n = 5), and fetal growth restriction (n = 5) (events are double-counted). There was no evident pattern indicating a relationship between the type of adverse pregnancy outcome and the serologic results; however, the numbers were small.
Results of multivariate analysis.
Multivariate analyses were performed by fitting Poisson regression models. The final model included LAC status at screening (relative risk [RR] 12.15, 95% CI 2.92–50.54, P = 0.0006), history of thrombosis (RR 1.90, 95% CI 1.14–3.17, P = 0.01), SLE (RR 2.16, 95% CI 1.27–3.68, P = 0.005), age (RR 1.56 for every 5-year decrease in age, 95% CI 1.18–2.08, P = 0.002), and race (RR 3.24 for white versus nonwhite, 95% CI 1.16–9.07, P = 0.03). The area under the receiver operating characteristic (ROC) curve for this model was 0.90, indicating a high ability to discriminate between patients who would and those who would not have adverse pregnancy outcome. The accuracy of the model averaged across the test sets in the 5-fold cross-validation procedure was 78% (sensitivity 71%, specificity 80%) using the cut point on the ROC curve closest to (0,1), i.e., perfect sensitivity and specificity, as the threshold for classification. Heparin treatment was not associated with risk of adverse pregnancy outcome after adjustment for the same predictors (RR 0.94, 95% CI 0.42–2.13, P = 0.88), nor were levels of IgG or IgM aCL or anti-β2GPI. The protective effect of aspirin was also no longer significant after adjustment for the predictors in the final model (RR 0.64, 95% CI 0.36–1.12, P = 0.12). Similar multivariate results were obtained when adverse pregnancy outcome included mild events (delivery after 34 weeks because of preeclampsia and growth restriction) as well as severe events as described above (RR for LAC positivity at screening 2.76, 95% CI 1.55–4.91, P = 0.0006).
In 1980, Soulier and Boffa described an association between LAC and recurrent pregnancy loss (27), which was confirmed initially by case reports and later by clinical series associating aPL with (generally second-trimester) pregnancy loss, fetal growth restriction, and severe preeclampsia (28). There is no consensus concerning which, if any, autoantibody profiles predict pregnancy outcome in women with aPL (4–14); in a recent retrospective analysis of 38 patients who were tested for LAC, aCL, and anti-β2GPI it was argued that triple positivity is the best predictor (4, 5), whereas investigators from Saudi Arabia argued for (broadly defined) LAC positivity (14).
Our prospective study, with rigorous definitions and with all laboratory analyses performed at central core laboratories, demonstrates that women who have LAC at screening are at highest risk for adverse pregnancy outcome, that LAC-negative women, regardless of IgG or IgM aCL or anti-β2GPI status, are not at high risk, and that LAC is the only component of triple aPL positivity that has predictive power. Although some have suggested that anti-β2GPI should be the most predictive antibody (29, 30), our findings do not support this.
Strengths of the present study are its prospective design, large sample size, rigorous clinical definitions, consideration of multiple clinical variables, and use of core laboratories. A weakness is that the study was referral center–based and treatment was directed by the patients' physicians. In these respects, however, our study does not differ from previously reported studies on this topic. The use of multivariate analysis allowed us to stratify patients by serologic and clinical features and by treatment, and our conclusions regarding LAC still hold after these stratifications. In order to avoid confounding aPL-associated causes of adverse pregnancy outcome with other causes, such as preexisting hypertension, diabetes, renal failure, or multifetal pregnancies, we excluded patients with such conditions, and we were unable to study patients with early pregnancy loss. Our conclusions therefore apply to aPL-positive women after the twelfth week of pregnancy who are in relatively good health.
Unexpectedly, we did not see a beneficial effect of heparin treatment on adverse pregnancy outcome. Our study was not designed or powered as a treatment trial, patients were not randomly allocated to treatment arms, and numbers are small, so our findings with regard to heparin are at best hypothesis-generating. These results possibly reflect physician choice to prescribe heparin selectively in patients at higher risk (for reasons not identified by our criteria) or that heparin is less effective than is currently assumed.
Our definition of LAC positivity required a positive result obtained using at least 1 of 3 specific assays. Consensus recommendations regarding diagnosis of LAC positivity are not widely used at commercial laboratories or at our university hospital laboratories (31, 32), nor do they account well for reagent variation or for quantitative and qualitative differences among LACs diagnosed with different assays (33, 34). Collectively, our 3 screening tests identified 93% of the patients who had adverse pregnancy outcome. Individually, dRVVT identified 52%, APTT 63%, and dPT 78%. Thus, in clinical use, a negative result on LAC testing performed by only 1 method is not fully reassuring. Since insurers often will not pay for more than a single test and laboratories rarely offer an option to perform multiple tests, clinicians must retain a high level of concern. On the other hand, any positive test result for LAC is clear evidence for pregnancy risk.
Measurement of aCL and anti-β2GPI at commercial laboratories is also problematic, as shown by our finding that 40% of referred patients with other laboratory–identified aPL had negative results at our core laboratories. We have previously commented on the inconsistencies of laboratories in measuring aPL (35, 36). Because positivity for aPL in all of the patients categorized as aPL positive in the present study was confirmed at our core laboratories, and because we cross-checked core laboratory results with findings at other laboratories known to be reliable, we believe the results reported here are as close to true determinations of aPL as possible.
The findings described herein demonstrate that, among aPL-positive patients with or without SLE, LAC positivity, with or without other aPL antibodies, is the primary predictor of adverse pregnancy outcome. Prior thrombosis, SLE, age, and race are lesser contributors to risk. These risk factors should be taken into consideration in future treatment trials, and clinicians should treat only those patients who are truly at high risk (37).
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Lockshin had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Lockshin, Kim, Laskin, Guerra, Branch, Merrill, Buyon, Salmon.
Acquisition of data. Lockshin, Kim, Laskin, Guerra, Branch, Merrill, Petri, Porter, Sammaritano, Stephenson, Buyon, Salmon.
Analysis and interpretation of data. Lockshin, Kim, Laskin, Guerra, Branch, Merrill, Petri, Porter, Sammaritano, Stephenson, Buyon, Salmon.
The authors wish to thank Christine Clark for performing and interpreting LAC assays, and Emily Reeves, Vamsy Bobba, and Karen Spitzer for assistance in recruiting patients.