Adverse pregnancy outcomes in women with systemic lupus erythematosus: can we improve predictions with machine learning?

OBJECTIVES: Nearly 20% of pregnancies in patients with SLE result in an adverse pregnancy outcome (APO). We previously developed an APO prediction model using logistic regression and data from Predictors of pRegnancy Outcome: bioMarkers In Antiphospholipid Antibody Syndrome and Systemic Lupus Erythematosus (PROMISSE), a large multicentre study of pregnant women with mild/moderate SLE and/or antiphospholipid antibodies. Our goal was to determine whether machine learning (ML) approaches improve APO prediction and identify other risk factors. METHODS: The PROMISSE data included 41 predictors from 385 subjects; 18.4% had APO (preterm delivery due to placental insufficiency/pre-eclampsia, fetal/neonatal death, fetal growth restriction). Logistic regression with stepwise selection (LR-S), least absolute shrinkage and selection operator (LASSO), random forest (RF), neural network (NN), support vector machines (SVM-RBF), gradient boosting (GB) and SuperLearner (SL) were compared by cross-validated area under the ROC curve (AUC) and calibration. RESULTS: Previously identified APO risk factors, antihypertensive medication use, low platelets, SLE disease activity and lupus anticoagulant (LAC), were confirmed as important with each algorithm. LASSO additionally revealed potential interactions between LAC and anticardiolipin IgG, among others. SL performed the best (AUC=0.78), but was statistically indistinguishable from LASSO, SVM-RBF and RF (AUC=0.77 for all). LR-S, NN and GB had worse AUC (0.71-0.74) and calibration scores. CONCLUSIONS: We predicted APO with reasonable accuracy using variables routinely assessed prior to the 12th week of pregnancy. LASSO and some ML methods performed better than a standard logistic regression approach. Substantial improvement in APO prediction will likely be realised, not with increasingly complex algorithms but by the discovery of new biomarkers and APO risk factors.

Low frequency of flares during pregnancy and post-partum in stable lupus patients.

BACKGROUND: Lupus patients are at risk for pregnancy loss, and it has been generally accepted that women with SLE should have low disease activity prior to conception. However, there are conflicting results regarding the effect of pregnancy on SLE flares. This study aims to identify predictors of flares during and after pregnancy in SLE patients with inactive or stable disease activity during the first trimester and to characterize and estimate the frequency of post-partum flares in these patients. METHODS: SLE patients in the multicenter, prospective PROMISSE (Predictors of Pregnancy Outcome: Biomarkers in Antiphospholipid Antibody Syndrome and Systemic Lupus Erythematosus) study were evaluated for flares during and after pregnancy using the SELENA-SLEDAI Flare Index. Flares during pregnancy were assessed in all 384 patients and post-partum flares in 234 patients with study visits 2-6 months post-partum. Logistic regression models were fit to the data to identify independent risk factors for flare. RESULTS: During pregnancy, 20.8% of patients had mild/moderate flares and 6.25% had severe. Post-partum, 27.7% of patients had mild/moderate flares and 1.7% had severe. The mild flares rarely required treatment. Younger age, low C4 and higher PGA at baseline were independently associated with higher risk of having at least one mild/moderate or severe flare during pregnancy. Older patients were at decreased risk of flare, as well as those with quiescent disease at baseline. No variables evaluated at baseline or the visit most proximal to delivery was significantly associated with risk of flare post-partum. Medications were not associated with flare during or after pregnancy. CONCLUSION: In patients with inactive or stable mild disease activity at the time of conception, lupus disease flares during and after pregnancy are typically mild and occur at similar rates. Flares during pregnancy are predicted by the patients' age and clinical and serological activity at baseline.

Longitudinal profiling of human blood transcriptome in healthy and lupus pregnancy.

Systemic lupus erythematosus carries an increased risk of pregnancy complications, including preeclampsia and fetal adverse outcomes. To identify the underlying molecular mechanisms, we longitudinally profiled the blood transcriptome of 92 lupus patients and 43 healthy women during pregnancy and postpartum and performed multicolor flow cytometry in a subset of them. We also profiled 25 healthy women undergoing assisted reproductive technology to monitor transcriptional changes around embryo implantation. Sustained down-regulation of multiple immune signatures, including interferon and plasma cells, was observed during healthy pregnancy. These changes appeared early after embryo implantation and were mirrored in uncomplicated lupus pregnancies. Patients with preeclampsia displayed early up-regulation of neutrophil signatures that correlated with expansion of immature neutrophils. Lupus pregnancies with fetal complications carried the highest interferon and plasma cell signatures as well as activated CD4+ T cell counts. Thus, blood immunomonitoring reveals that both healthy and uncomplicated lupus pregnancies exhibit early and sustained transcriptional modulation of lupus-related signatures, and a lack thereof associates with adverse outcomes.

Performance Evaluation and Clinical Associations of Immunoassays That Detect Antibodies to Negatively Charged Phospholipids Other Than Cardiolipin.

OBJECTIVES: We evaluate the performance characteristics of antiphosphatidylserine (anti-PS), antiphosphatidylinositol (anti-PI), and antiphospholipid mixture (APhL) enzyme-linked immunosorbent assays (ELISAs) compared with anticardiolipin (aCL) and anti-ß2 glycoprotein I (anti-ß2GPI) in a large group of patients with antiphospholipid (aPL)-related diseases. METHODS: Serum samples from 548 patients from the Hopkins and Jamaican systemic lupus erythematosus cohorts, the PROMISSE cohort, and the Antiphospholipid Standardization Laboratory were examined for immunoglobulin G (IgG)/immunoglobulin M (IgM) positivity in aCL, anti-ß2GPI, anti-PS, anti-PI, and APhL ELISA assays. RESULTS: All IgG assays were associated with one or more thrombotic and/or obstetric manifestations, with an increased risk associated with higher antibody titers. Analytical performance was similar among assays, but IgG assays performed better than IgM counterparts. CONCLUSIONS: Increasing titers of APhL, anti-PS, and anti-PI antibodies could indicate an increased risk of thrombotic and/or obstetric aPL-related manifestations. These assays may be promising biomarkers for particular APS manifestations.

Complement activation predicts adverse pregnancy outcome in patients with systemic lupus erythematosus and/or antiphospholipid antibodies.

OBJECTIVE: Studies in mouse models implicate complement activation as a causative factor in adverse pregnancy outcomes (APOs). We investigated whether activation of complement early in pregnancy predicts APOs in women with systemic lupus erythematosus (SLE) and/or antiphospholipid (aPL) antibodies. METHODS: The PROMISSE Study enrolled pregnant women with SLE and/or aPL antibodies (n=487) and pregnant healthy controls (n=204) at <12 weeks gestation and evaluated them monthly. APOs were: fetal/neonatal death, preterm delivery <36 weeks because of placental insufficiency or preeclampsia and/or growth restriction <5th percentile. Complement activation products were measured on serial blood samples obtained at each monthly visit. RESULTS: APO occurred in 20.5% of SLE and/or aPL pregnancies. As early as 12-15 weeks, levels of Bb and sC5b-9 were significantly higher in patients with APOs and remained elevated through 31 weeks compared with those with normal outcomes. Moreover, Bb and sC5b-9 were significantly higher in patients with SLE and/or aPL without APOs compared with healthy controls. In logistic regression analyses, Bb and sC5b-9 at 12-15 weeks remained significantly associated with APO (ORadj=1.41 per SD increase; 95% CI 1.06 to 1.89; P=0.019 and ORadj=1.37 per SD increase; 95% CI 1.05 to 1.80; P=0.022, respectively) after controlling for demographic and clinical risk factors for APOs in PROMISSE. When analyses were restricted to patients with aPL (n=161), associations between Bb at 12-15 weeks and APOs became stronger (ORadj=2.01 per SD increase; 95% CI 1.16 to 3.49; P=0.013). CONCLUSION: In pregnant patients with SLE and/or aPL, increased Bb and sC5b-9 detectable early in pregnancy are strongly predictive of APOs and support activation of complement, particularly the alternative pathway, as a contributor to APOs.

Kidney Outcomes and Risk Factors for Nephritis (Flare/De Novo) in a Multiethnic Cohort of Pregnant Patients with Lupus.

BACKGROUND AND OBJECTIVES: Kidney disease is a critical concern in counseling patients with lupus considering pregnancy. This study sought to assess the risk of renal flares during pregnancy in women with previous lupus nephritis in partial or complete remission, particularly in those with antidouble-stranded DNA antibodies and low complement levels, and the risk of new-onset nephritis in patients with stable/mildly active SLE. DESIGN, SETTING, PARTICIPANTS, & MEASUREMENTS: We assessed active nephritis (renal flares and de novo kidney disease) and associated predictors during pregnancy in patients with lupus with urine protein ≤1000 mg and serum creatinine <1.2 mg/dl at baseline; 373 patients (52% ethnic/racial minorities) enrolled between 2003 and 2012 were prospectively followed in the Predictors of Pregnancy Outcome: Biomarkers in Antiphospholipid Syndrome and Systemic Lupus Erythematosus Study. Active nephritis was defined by proteinuria increase of >500 mg and/or red blood cell casts. RESULTS: Of 118 patients with previous kidney disease, 13 renal flares (11%) occurred (seven of 89 in complete remission and six of 29 in partial remission) compared with four with de novo kidney involvement (2%) in 255 patients without past kidney disease (P<0.001). Active nephritis was not associated with ethnicity, race, age, creatinine, BP, or antihypertensive and other medications. In multivariable logistic regression analyses, patients with past kidney disease in complete or partial remission more often experienced active nephritis (adjusted odds ratio, 6.88; 95% confidence interval, 1.84 to 25.71; P=0.004 and adjusted odds ratio, 20.98; 95% confidence interval, 4.69 to 93.98; P<0.001, respectively) than those without past kidney disease. Low C4 was associated with renal flares/de novo disease (adjusted odds ratio, 5.59; 95% confidence interval, 1.64 to 19.13; P<0.01) but not low C3 or positive anti-dsDNA alone. CONCLUSIONS: De novo kidney involvement in SLE, even in ethnic/racial minorities, is uncommon during pregnancy. Past kidney disease and low C4 at baseline independently associate with higher risk of developing active nephritis. Antibodies to dsDNA alone should not raise concern, even in patients with past kidney disease, if in remission.

Brief Report: Changes in Antiphospholipid Antibody Titers During Pregnancy: Effects on Pregnancy Outcomes.

OBJECTIVE: To measure variance in antiphospholipid antibody (aPL) levels during pregnancy and to determine if variation affects pregnancy outcomes. METHODS: We used data from the PROMISSE (Predictors of Pregnancy Outcome: Biomarkers in Antiphospholipid Antibody Syndrome and Systemic Lupus Erythematosus) study, a multicenter prospective study of pregnant women with aPL and/or systemic lupus erythematosus (SLE). Antiphospholipid antibodies were considered present if any of the following were positive: anticardiolipin (aCL), anti-ß2 -glycoprotein I (anti-ß2 GPI) titers ≥40 IgG phospholipid (GPL) or IgM phospholipid (MPL) units, and/or lupus anticoagulant (LAC). Antiphospholipid antibodies were measured every trimester and postpartum. Adverse pregnancy outcomes were defined as fetal/neonatal death, preterm delivery (<36 weeks) due to preeclampsia or placental insufficiency, or growth restriction. RESULTS: One hundred fifty-two aPL-positive patients were studied. Fifty-seven percent had clinical antiphospholipid syndrome (APS) and 36% had SLE. IgG aPL levels were significantly lower during the second and third trimesters compared to initial screening, but IgG aCL and anti-ß2 GPI remained high-positive through pregnancy in 93% of patients during the second trimester, and in 85% of patients during the third trimester. IgM aPL titers were negative in the majority of patients and decreased modestly during pregnancy among patients who were positive. LAC frequency also decreased, but 75% of patients remained positive through the second trimester. Only 4% of patients with aPL at baseline did not have aPL in either the second or third trimesters. Changes in aPL levels or aPL status were not associated with adverse pregnancy outcomes. LAC was the only aPL associated with adverse pregnancy outcomes. CONCLUSION: The aPL in the cohort decreased marginally during pregnancy, and changes were not associated with pregnancy outcomes. Our results suggest that, among women with aPL and/or SLE, measuring aPL early in pregnancy is sufficient to assess risk. Repeat aPL testing through pregnancy is unnecessary.

Lupus anticoagulant is the main predictor of adverse pregnancy outcomes in aPL-positive patients: validation of PROMISSE study results.

OBJECTIVE: We previously reported that lupus anticoagulant (LAC) is the main predictor of poor pregnancy outcome in antiphospholipid antibody (aPL)-positive patients. We sought to confirm this finding in an independent group of patients who were subsequently recruited into the PROMISSE study. METHODS: The PROMISSE study is a multicentre, prospective, observational study of pregnancy outcomes in women with aPL and/or systemic lupus erythematosus (SLE) that enrolled patients from 2003 to 2015. All consecutive, aPL-positive patients from the PROMISSE study who completed their pregnancy between April 2011 and January 2015 (after the previous PROMISSE report) are included in the current report. Patients were followed monthly until delivery, and aPL was tested at first, second and third trimesters of pregnancy and at 12 weeks post partum. Adverse pregnancy outcomes (APOs) were defined as fetal death after 12 weeks of gestation, neonatal death, delivery prior to 36 weeks of gestation due to pre-eclampsia or placental insufficiency or small-for-gestational age (birth weight <5th percentile). RESULTS: Forty-four aPL-positive patients are included in this paper. Thirteen patients had APOs, which occurred in 80% of cases during the second trimester of pregnancy. LAC was present in 69% of patients with APOs compared with 27% of patients without APOs (p=0.01). No association was found between anticardiolipin antibodies (aCL) or anti-ß2 glycoprotein I antibodies (aß2GPI) IgG or IgM positivity and APOs. Definite antiphospholipid syndrome (history of thrombosis and/or pregnancy morbidity and aPL) was found in 92% of patients with any APOs compared with 45% of patients without APOs (p=0.004). Conversely, the frequency of SLE was not statistically different between those with and without APOs (30% vs 39%). CONCLUSIONS: Our findings, in an independent group of aPL-positive patients from the PROMISSE study, confirm that LAC, but not aCL and aß2GPI, is predictive of poor pregnancy outcomes after 12 weeks of pregnancy. TRIAL REGISTRATION NUMBER: NCT00198068.

Angiogenic factor imbalance early in pregnancy predicts adverse outcomes in patients with lupus and antiphospholipid antibodies: results of the PROMISSE study.

BACKGROUND: Over 20% of pregnancies in patients with systemic lupus erythematosus (SLE) and/or antiphospholipid antibodies (APL) result in an adverse pregnancy outcome (APO) related to abnormal placentation. The ability to identify, early in pregnancy, patients who are destined for poor outcomes would significantly impact care of this high-risk population. In nonautoimmune patients, circulating angiogenic factors are dysregulated in disorders of placentation, such as preeclampsia (PE) and fetal growth restriction. OBJECTIVE: We sought to determine whether early dysregulation of circulating angiogenic factors can predict APO in high-risk SLE and/or APL pregnancies. STUDY DESIGN: We used data and samples from the Predictors of Pregnancy Outcome: Biomarkers in APL Syndrome and SLE (PROMISSE), a multicenter prospective study that enrolled 492 pregnant women with SLE and/or APL from September 2003 through August 2013. Patients were followed through pregnancy from <12 weeks gestation. Circulating levels of soluble fms-like tyrosine kinase-1 (sFlt1), placental growth factor (PlGF), and soluble endoglin were measured monthly and subjects followed up for APO, classified as severe (PE <34 weeks, fetal/neonatal death, indicated preterm delivery <30 weeks) or moderate (PE ≥34 weeks, indicated preterm delivery 30-36 weeks, growth restriction without PE). RESULTS: Severe APOs occurred in 12% and moderate APOs in 10% of patients. By 12-15 weeks, sFlt1, PlGF, and soluble endoglin levels were markedly altered in women who developed severe APO. After adjusting for clinical risk factors, sFlt1 was the strongest predictor of severe APO among 12-15 week measures (odds ratio, 17.3 comparing highest and lowest quartiles; 95% confidence interval [CI], 3.5-84.8; positive predictive value [PPV], 61%; negative predictive value [NPV], 93%). At 16-19 weeks, the combination of sFlt1 and PlGF was most predictive of severe APO, with risk greatest for subjects with both PlGF in lowest quartile (<70.3 pg/mL) and sFlt1 in highest quartile (>1872 pg/mL; odds ratio, 31.1; 95% CI, 8.0-121.9; PPV, 58%; NPV, 95%). Severe APO rate in this high-risk subgroup was 94% (95% CI, 70-99.8%), if lupus anticoagulant or history of high blood pressure was additionally present. In contrast, among patients with both sFlt1 <1872 pg/mL and PlGF >70.3 pg/mL, rate of severe APO was only 4.6% (95% CI, 2.1-8.6%). CONCLUSION: Circulating angiogenic factors measured during early gestation have a high NPV in ruling out the development of severe adverse outcomes among patients with SLE and/or APL syndrome. Timely risk stratification of patients is important for effective clinical care and optimal allocation of health care resources.

Predictors of Pregnancy Outcomes in Patients With Lupus: A Cohort Study.

BACKGROUND: Because systemic lupus erythematosus (SLE) affects women of reproductive age, pregnancy is a major concern. OBJECTIVE: To identify predictors of adverse pregnancy outcomes (APOs) in patients with inactive or stable active SLE. DESIGN: Prospective cohort. SETTING: Multicenter. PATIENTS: 385 patients (49% non-Hispanic white; 31% with prior nephritis) with SLE in the PROMISSE (Predictors of Pregnancy Outcome: Biomarkers in Antiphospholipid Antibody Syndrome and Systemic Lupus Erythematosus) study. Exclusion criteria were urinary protein-creatinine ratio greater than 1000 mg/g, creatinine level greater than 1.2 mg/dL, prednisone use greater than 20 mg/d, and multifetal pregnancy. MEASUREMENTS: APOs included fetal or neonatal death; birth before 36 weeks due to placental insufficiency, hypertension, or preeclampsia; and small-for-gestational-age (SGA) neonate (birthweight below the fifth percentile). Disease activity was assessed with the Systemic Lupus Erythematosus Pregnancy Disease Activity Index and the Physician's Global Assessment (PGA). RESULTS: APOs occurred in 19.0% (95% CI, 15.2% to 23.2%) of pregnancies; fetal death occurred in 4%, neonatal death occurred in 1%, preterm delivery occurred in 9%, and SGA neonate occurred in 10%. Severe flares in the second and third trimesters occurred in 2.5% and 3.0%, respectively. Baseline predictors of APOs included presence of lupus anticoagulant (LAC) (odds ratio [OR], 8.32 [CI, 3.59 to 19.26]), antihypertensive use (OR, 7.05 [CI, 3.05 to 16.31]), PGA score greater than 1 (OR, 4.02 [CI, 1.84 to 8.82]), and low platelet count (OR, 1.33 [CI, 1.09 to 1.63] per decrease of 50 × 109 cells/L). Non-Hispanic white race was protective (OR, 0.45 [CI, 0.24 to 0.84]). Maternal flares, higher disease activity, and smaller increases in C3 level later in pregnancy also predicted APOs. Among women without baseline risk factors, the APO rate was 7.8%. For those who either were LAC-positive or were LAC-negative but nonwhite or Hispanic and using antihypertensives, the APO rate was 58.0% and fetal or neonatal mortality was 22.0%. LIMITATION: Patients with high disease activity were excluded. CONCLUSION: In pregnant patients with inactive or stable mild/moderate SLE, severe flares are infrequent and, absent specific risk factors, outcomes are favorable. PRIMARY FUNDING SOURCE: National Institutes of Health.

Evaluation of the risk of anti-SSA/Ro-SSB/La antibody-associated cardiac manifestations of neonatal lupus in fetuses of mothers with systemic lupus erythematosus exposed to hydroxychloroquine.

BACKGROUND: Based on the potential involvement of Toll-like receptor (TLR) signalling in the pathogenesis of neonatal lupus (NL), it was hypothesised that fetal exposure to hydroxychloroquine (HCQ), a TLR inhibitor, might reduce the risk of anti-SSA/Ro/SSB/La antibody-associated cardiac manifestations of NL (cardiac-NL). METHODS: Cardiac-NL children (N=50) and controls (N=151) were drawn from the following overlapping pregnancy studies: Research Registry for NL; PR Interval and Dexamethasone Evaluation in Cardiac-NL; and Predictors of Pregnancy Outcomes: Biomarkers in Antiphospholipid Syndrome and Systemic Lupus Erythematosus (SLE). Pregnancies met the following inclusion criteria: documentation of maternal anti-SSA/Ro/SSB/La antibodies at pregnancy, confirmation of medication use and child's outcome, a diagnosis of SLE before pregnancy and birth by 31 December 2007. RESULTS: Seven (14%) of the cardiac-NL children were exposed to HCQ compared with 56 (37%) of the controls (p=0.002; OR 0.28; 95% CI 0.12 to 0.63). Cases and controls were similar with respect to demographic and antibody status. Multivariable analysis adjusting for birth year, maternal race/ethnicity, antibody status, non-fluorinated steroid use and prior cardiac-NL risk yielded an OR associated with HCQ use of 0.46 (95% CI 0.18 to 1.18; p=0.10). CONCLUSION: This case-control study suggests that, in mothers with SLE with anti-SSA/Ro/SSB/La antibodies, exposure to HCQ during pregnancy may decrease the risk of fetal development of cardiac-NL. Prospective studies are needed for confirmation.