Lupus Anticoagulants, Anticardiolipin Antibodies, and Fetal Loss: A Case–Control Study

Identification of Case Patients and Controls

The case patients were women hospitalized with a medically confirmed diagnosis of spontaneous abortion or fetal death at Sainte-Justine Hospital, a university-affiliated institution offering pediatric and obstetrical care. The study period lasted from May 1987 to November 1989. Case patients who reported previous spontaneous pregnancy loss were not eligible for the study; those admitted at night and discharged before the next morning and those admitted during the weekend or on legal holidays were also excluded. A total of 1011 women were admitted because of fetal loss during the study period. Of these, 357 were excluded because they were admitted outside working hours, and 138 because they reported previous spontaneous fetal loss. Of the 516 eligible women, 331 (64 percent) were enrolled. Among the eligible women who were not enrolled, 30 percent (n = 55) declined to participate because of reluctance to allow further blood sampling after the routine sample had been taken. Therefore, early in the course of the study, arrangements were made for blood to be drawn only once for both the study and routine testing. Other reasons for nonparticipation were that the patient left the hospital before the interviewer could complete data collection (n = 108), an insufficient quantity of blood was taken (n = 9), the patient was not fluent in French or English (n = 6), and the patient declined for unknown reasons (n = 7).

Three controls were matched to each case patient (total, 993 controls) according to the period of pregnancy of the case patient as based on the time since the last menstrual period. Women who reported that they had had spontaneous fetal loss were not eligible to be controls. The period of gestation was categorized as 16 weeks of pregnancy or less, 17 to 20 weeks, 21 to 27 weeks, and 28 weeks or more. The controls were recruited from among pregnant women who presented at the hospital test center for routine blood analyses ordered by their physicians. From the outset of recruitment, blood was obtained only once for both routine analyses and testing for lupus anticoagulants and anticardiolipin antibodies. Fifteen eligible controls declined to participate in the study, four others were not fluent in French or in English, and two were not enrolled for unknown reasons.

Laboratory Procedures

Blood Sampling

Blood was obtained by clean venipuncture and collected directly into silicone-coated Vacutainer tubes (Becton Dickinson, Rutherford, N.J.) containing 105 mM buffered citrate solution, for a final ratio of one part anticoagulant to nine parts blood. After the sample was centrifuged twice at 12,000×g for 20 minutes at 4°C, platelet-poor plasma was fast-frozen in small aliquots and kept at —70°C until testing. Normal plasma was obtained under the same conditions from 20 adult male volunteers, pooled, and frozen in small aliquots to provide a single lot of normal pooled plasma containing fewer than 3×10⁹ platelets per liter. The plasma from the study subjects and the male volunteers was thawed at 37°C for 15 minutes and then kept on melting ice before it was used in assays. Standards for the anticardiolipin assay were obtained from Dr. E.N. Harris (University of Louisville, Louisville, Ky.); the samples used were numbers 19, 87, 09, 62, and 61.

Coagulation Tests

Figure 1. Figure 1. Stepwise Screening for Lupus Anticoagulant in Women with Fetal Loss (Case Patients) and Controls.

APTT denotes the activated partial-thromboplastin time determined in plasma samples containing one part normal pooled plasma (from healthy men) to four parts patient plasma (from the case patients and controls); dRVVT the dilute Russell's viper—venom time determined in samples containing one part normal pooled plasma to one part patient plasma; PNP the platelet neutralization procedure; FDP the assay for fibrin—fibrinogen degradation products; and LA lupus anticoagulant.

Plasma samples containing lupus anticoagulants were identified in a stepwise fashion (Fig. 1). The activated partial-thromboplastin time (APTT) was determined with a commercial thromboplastin (Actin FSL, American Dade, Aguada, Puerto Rico) in 1:4 mixtures of pooled:patient plasma. Specimens were considered abnormal when the APTT exceeded 37.1 seconds (mean +2 SD for the 20 normal plasma samples). The dilute Russell's viper—venom time (dRVVT) was determined as described by Thiagarajan et al.⁹ in 1:1 mixtures of pooled:patient plasma. The venom and the phospholipid (Thrombofax, Ortho Diagnostic Systems, Raritan, N.J.) were diluted to concentrations of 1:200 and 1:32, respectively. Specimens were considered abnormal when the dRVVT exceeded 27.5 seconds (mean +2 SD for the 20 normal plasma samples). Specimens shown to be abnormal by either test were also tested according to the platelet neutralization procedure,¹⁰ with use of a reagent from Bio Data (Hatboro, Pa.). The results of this test were considered abnormal if the clotting time determined with the reagent was five or more seconds shorter than the time determined in parallel in the same sample with the use of 150 mM sodium chloride. Specimens shown to be abnormal by the platelet neutralization procedure were tested with thrombin (final concentration, 1 unit per milliliter) to rule out the presence of heparin. Specimens with a thrombin time of 22.7 seconds or less (mean +2 SD for the 20 normal plasma samples) were retained for further testing. Specimens that contained more than 10 μg of fibrin—fibrinogen degradation products per milliliter (Thrombo-Wellcotest, Laboratoires Wellcome, Paris) were excluded.

In summary, plasma samples shown to have an abnormal APTT or dRVVT on testing of standardized mixtures of pooled:patient plasma, as well as an abnormal clotting time on testing with the platelet neutralization procedure and no evidence of heparin or fibrin—fibrinogen degradation products, were considered positive for lupus anticoagulants.

Assay for Anticardiolipin Antibodies

The enzyme-linked immunosorbent assay for IgG anticardiolipin antibodies was performed as previously described,¹¹ with minor modifications to make it consistent with recommendations of the Kingston Antiphospholipid Study Group.² Optical density was measured at 405 nm with a Microwell System Reader 510 (Organon Teknika, Turnhout, Belgium) programmed to generate a linear regression—curve fit. Results were reported in immunoglobulin G antiphospholipid (GPL) units. As then recommended,² a specimen containing less than 5 GPL units was considered negative. In our laboratory, a specimen with 5 GPL units has an optical density comparable to the mean +5 SD in normal controls.

Interview

A face-to-face interview with all case patients and controls was carried out at the hospital. The following potential confounding factors were measured^{12 , 13}: the mother's age, race, and level of education; the number of previous pregnancies and induced abortions; smoking, alcohol use, and coffee drinking during pregnancy; and medical conditions diagnosed by a physician (uterine and cervical anatomical abnormalities, endometriosis, hypertension present before pregnancy, and ppregnancy-induced hypertension). Finally, factors related to work during pregnancy were ascertained: work requiring physical effort; exposure to chemical products, radiation, or video-display terminals; and work schedule.

Statistical Analysis

A conditional logistic-regression model¹⁴ was used to analyze the matched sets of subjects — i.e., three controls matched to each case patient (EGRET package, Statistics and Epidemiology Research, Seattle). Odds ratios and 95 percent confidence intervals were estimated. All statistical tests were two-sided.

All variables were defined as categorized indicator variables so that the relation between a factor and the risk of fetal loss was not forced to be linear. Categories were defined a priori. Variables were included in the model by backward selection. For each exposure variable (presence or absence of lupus anticoagulant, and presence or absence of IgG anticardiolipin antibody), a full model was considered first, including the exposure variable and all potential confounding variables. The change-in-estimate method¹⁵ was used to determine whether a variable should be excluded from the model.

Two-factor interaction terms were defined as the product of each variable in the final model and exposure. These terms were tested with the score test,¹⁴ which tests the null hypothesis that the coefficients for the interaction terms are zero. None of the score tests gave results statistically significant at a probability level of 0.05; therefore, no interaction terms were kept in the model.

In this study design, controls were selected from among women still at risk for fetal loss when it was diagnosed in a case patient. A woman originally selected as a control might therefore be identified as a case patient at a later date.¹⁶ In such instances, the blood tests were repeated but the interview was not. Seventeen controls became case patients during the same pregnancy. A woman who had a second pregnancy during the study period was not eligible for the study the second time.

The mean age (±SD) of the case patients was 28.7±5.45 years, and that of the controls 27.4±4.89 years. At entry, 157 case patients (47.4 percent) were in their first pregnancy, 104 (31.4 percent) in their second pregnancy, and 70 (21.1 percent) in a later pregnancy; the corresponding figures in the controls were 469 (47.2 percent), 321 (32.3 percent), and 203 (20.4 percent). Women with previous spontaneous fetal loss were excluded from the study, but 67 case patients (20.2 percent) and 173 controls (17.4 percent) reported that they had had an induced abortion. The distribution of case patients and controls with respect to sociodemographic, obstetrical, medical, and occupational characteristics was similar. However, uterine abnormalities and cervical abnormalities were more frequent among the case patients than the controls (2.0 percent and 3.7 percent, respectively, vs. 0.5 percent and 1.5 percent, respectively), and more cases than controls had irregular work schedules (19.3 percent vs. 11.8 percent).

Table 1. Table 1. Case Patients and Controls with Lupus Anticoagulants (LA) and IgG Anticardiolipin Antibodies (ACA), According to the Period of Pregnancy.

The distribution of the subjects according to their week of gestation is shown in Table 1, as well as the numbers who had lupus anticoagulants and IgG anticardiolipin antibodies. Figure 1 shows the results of the different coagulation tests used in a stepwise fashion to identify specimens that contained lupus anticoagulants. Lupus anticoagulants were present in 17 case patients (5.1 percent) and in 38 controls (3.8 percent). The odds ratio for the association between lupus anticoagulants and fetal loss, adjusted only for the period of pregnancy (the variable used for matching), was 1.36 (95 percent confidence interval, 0.75 to 2.43; P = 0.30). Four case patients (1.2 percent) had levels of IgG anticardiolipin antibody ranging from 6 to 66 GPL units (mean, 26), and 15 controls (1.5 percent) had levels of 5 to 94 GPL units (mean, 20). The odds ratio for the association between IgG anticardiolipin antibody and fetal loss, adjusted for the period of pregnancy, was 0.80 (95 percent confidence interval, 0.26 to 2.41; P = 0.69).

Table 2. Table 2. Final Conditional Logistic-Regression Model for the Analysis of Fetal Loss and Lupus Anticoagulants.

In the final model (Table 2), the odds ratio for fetal loss among women with lupus anticoagulants was 1.42 (95 percent confidence interval, 0.72 to 2.80; P = 0.31). When a similar analysis was performed in which IgG anticardiolipin antibodies served as the exposure variable, the odds ratio was 1.28 (95 percent confidence interval, 0.38 to 4.21; P = 0.61); this analysis included the same variables as those in the final model, with lupus anticoagulants as the exposure variable.

The results of this study suggest that women with a spontaneous abortion or fetal death are unlikely to have either lupus anticoagulants or IgG anticardiolipin antibodies more frequently than women in the same period of pregnancy whose pregnancy is normal. These conclusions apply only to women without previous spontaneous fetal loss.

Our study design, limiting eligibility as case patients to women who presented during the day on working weekdays and excluding eligible case patients who were discharged overnight or who declined to have blood drawn, was unlikely to have introduced bias, since there is no indication that these factors may be associated with the antibodies studied. Women presenting at the same hospital as the case patients, during the same period of pregnancy and with a similar absence of previous spontaneous fetal loss were considered a reasonable choice for controls since they were thought to be representative of the population from which the case patients were drawn.

Differential misclassification of subjects is unlikely since the laboratory measurements were performed blindly with regard to whether a subject was a case patient or a control. If residual platelets were present in the plasma because of weak centrifugation, the sensitivity of the assay for lupus anticoagulants might have been decreased.¹⁷ However, this is very unlikely to have occurred, since our specimens were centrifuged twice at 12,000×g for 20 minutes, which is clearly in excess of the common practice of centrifuging at 1500×g for 10 minutes.¹⁸ The determination of the APTT in a 1:4 mixture of normal pooled:patient plasma and the dRVVT in a 1:1 mixture may not be as sensitive as other tests,¹¹ but it agrees with current recommendations.¹⁹ Several types of enzyme-linked immunosorbent assay have been used; these were reviewed by Harris,² whose recommendations for performing tests and reporting results we followed. In an international evaluation of quality control during the course of this study, our laboratory achieved a 96 percent agreement with other laboratories.²⁰

Lockwood et al.²¹ measured antibody levels in 737 women recruited at an obstetrical clinic, excluding women with three or more consecutive spontaneous fetal losses; the mean age of the group was 22.7 years. The outcome of pregnancy in 2 women with lupus anticoagulants was compared with that in 735 women without them; the outcome in 16 women who were positive for IgG anticardiolipin antibodies was also compared with that in 718 who were negative. Both women with lupus anticoagulants had a spontaneous abortion (infinite relative risk); 4 of the 16 women who were positive for anticardiolipin antibodies had a spontaneous abortion, as compared with 36 of the 718 antibody-negative women (relative risk = 5; 95 percent confidence interval, 2.10 to 12.34).

Most other reports in the literature involve case series of women with recurrent miscarriages (or women with habitual abortions). In five studies, which included 58 to 190 subjects, women with miscarriages were compared with women with no history of fetal loss^{22 23 24 25 26}; the presence of lupus anticoagulant was indicated by an abnormal APTT or dRVVT. In three studies, the test for anticardiolipin antibodies was considered positive if its result was more than 5 SD above the control value. In all five studies, odds ratios for the coagulation test were infinite (none of the controls had a positive test) and were increased for anticardiolipin antibodies. These results contrast with ours. If the true odds ratio in our study population was high, our study should have been able to detect it with considerable power. None of the previous studies addressed the practical problems related to the application of the concept of habitual abortion.^{27 , 28} Gravidity and maternal age, which are independent risk factors for spontaneous abortion,²⁹ were not controlled for in any of the previous studies, and this could have greatly biased their results. Case patients defined as "habitual aborters" were probably older than controls, who, in some studies, had never been pregnant.

By defining case patients in this way, previous investigators have postulated that lupus anticoagulants or anticardiolipin antibodies become risk factors only after a number of fetal losses. The plausibility of such a hypothesis is not supported by any data and is questionable: if these antibodies are associated with a risk of miscarriage either directly or indirectly, as indicators of some underlying immunologic disorder, nothing leads us to believe that the risk should be manifested only after at least two fetal losses rather than at any time during reproductive life. On the other hand, if the prevalence of these antibodies increases with the number of spontaneous abortions, that might imply an increase with gravidity and maternal age, in which case choosing younger controls — probably with fewer pregnancies — and not adjusting for age in the analysis would grossly overestimate the risk. The apparent clinical association in some women between lupus anticoagulants or anticardiolipin antibodies (or both) and repeated fetal loss, a problem that this study did not address, remains challenging and should be investigated further.