Literature search results are shown in . The initial search identified 374 articles. Three potentially relevant non-English publications were identified. An article written in Polish and one of the two Russian articles were excluded based on review of the English abstract. The second Russian article could not be found in English and was also excluded. Seventeen articles remained after inclusion and exclusion criteria were applied and were evaluated for the association of FVL with thrombosis in SLE patients. A total of 2090 SLE and/or aPL positive patients were included in this preliminary meta-analysis. Characteristics of included studies are shown in . Eight studies recruited subjects based on aPL status (many of whom had SLE), 6 studies recruited subjects based on SLE status (many of whom were aPL positive), and 3 studies recruited both aPL positive and SLE subjects. All studies were retrospective. Studies were performed in a wide variety of countries including the United States (3 studies)12-14
, Canada (3 studies)10,15,16
, the Netherlands (2 studies) 17,18
, Italy (2 studies) 8,19
, Hungary (2 studies) 20,21
, and one study each from Sweden 22
, England 23
, Spain 24
, Turkey 25
, and Argentina 26
. Five studies included small proportions of non-European subjects 10,12-14,16
(which we were able to exclude from the individual patient data meta-analysis).
Literature search strategy for meta-analysis of the factor V Leiden polymorphism and thrombosis among subjects with SLE and/or aPL positivity
Summary of 17 studies included in meta-analysis of the factor V Leiden polymorphism and thrombosis among subjects with SLE and/or aPL positivity.
Using a standardized form, we abstracted data on study design, study setting, sample size, enrollment criteria, number of SLE and/or aPL positive subjects, ethnicity, covariates considered, type of thromboses considered, number of subjects with lupus anticoagulant and/or anti-cardiolipin antibody, number of subjects with at least one thrombotic event, and number of subjects with multiple events. In order to calculate unadjusted odds ratios (OR) for the association between FVL and thrombosis, we also abstracted the number of patients with FVL (both with and without thrombotic events) and the number of patients who were FVL negative (both with and without thrombotic events).
Eight hundred twenty-one subjects (39%) had a thrombotic event, 1099 (53%) had SLE and 1245 (60%) were aPL positive. One hundred thirty-four (6.4%) were FVL (risk allele) positive. All but two studies8,13
clearly stated that they confirmed thrombotic events using medical records and x-ray reports. Types of thromboses assessed in these studies varied. Some studies considered broad categories of venous and arterial thromboses12,13,23
. Most studies considered many different types of thromboses including DVT, PE, CVA, MI, retinal vein thrombosis, recurrent miscarriages in the first trimester, and at least one miscarriage in the second or third trimester. Most patients who tested positive for FVL were heterozygous and only two studies included a patient homozygous for FVL.8,21
Most studies measured both anti-cardiolipin and lupus anticoagulants. Three recent studies also measured ß2glycoprotein. 10,16,21
Publication bias assessment
To determine if publication bias existed among our seventeen included studies, we first generated a funnel plot (). This plot did not reveal substantial asymmetry (there were similar numbers of negative and positive results). We then performed Egger's regression asymmetry test, a statistical test to assess for publication bias. The results indicated no evidence for publication bias (p = 0.46).
Funnel plot of 17 studies included in meta-analysis of the factor V Leiden polymorphism and thrombosis among subjects with SLE and/or aPL positivity
A Mantel-Haenszel test statistic was used to assess homogeneity. There was no evidence to reject the null hypothesis that the studies were homogeneous (p = 0.72) therefore, data were combined using the Mantel-Haenszel method (fixed effects model) to obtain a summary estimate and these results are displayed in a Forrest plot in . ORs for association of FVL with thrombosis in individual studies range from 1.5 to 20 with one study obtaining an OR of 0.4 (Sasso et. al.), indicating an inverse association (although this was not a statistically significant finding). Most confidence intervals had wide ranges and included 1.0, and therefore were not statistically significant. The pooled OR for association of FVL with thrombosis was 2.88 (95% C.I. 1.98 - 4.20). ORs were also calculated separately for studies that recruited patients based on aPL positivity and studies that recruited patients based on an SLE diagnosis to determine whether results of these studies differed. The pooled OR for studies that recruited patients based on aPL status was 2.87 (95% C.I. 1.60-5.16) and for those that recruited subjects based on SLE status was 2.69 (95% C.I. 1.56-4.62). Because these ORs were similar, all studies were pooled in the final analysis.
Forrest plot of 17 studies included in meta-analysis of the factor V Leiden polymorphism and thrombosis among subjects with SLE and/or aPL positivity
Individual Patient Data Results
For our secondary meta-analysis, we queried all authors of the original manuscripts to request individual level patient data in order to calculate the association between FVL and thrombosis while also adjusting for relevant covariates such as age, gender, smoking status and aPL. This data also allowed for subgroup analysis, for example, analysis of subjects who specifically had a DVT. Seven of seventeen authors responded to our query and five contributed individual patient level data. In addition, we included 1210 SLE subjects from the UCSF Lupus Genetics Collection genotyped for FVL and with well-described covariates and thrombosis outcomes. Thus, a total of 2030 subjects were available for analysis in this dataset.
Among these subjects, the average age was 43 (S.D. 13.8), 1691 (83%) subjects were female, 1359 (67%) were of European-derived ancestry, 104 (5%) tested positive for the FVL polymorphism, 621 (31%) experienced a thrombotic event, 675 (33%) were ever-smokers, and 1046 (51%) had been treated with hydroxychloroquine. Eight hundred ten (40%) subjects were aCL (anti-cardiolipin) or LAC (lupus anticoagulant) positive. One hundred seventy-two (8.5%) subjects had a DVT, 56 (2.8%) had a PE, 106 (5.2%) had a CVA, 52 had an (2.6%) MI, and 6 (0.3%) had a retinal vein thrombosis. Four hundred thirty-one (21%) had more than one thrombotic event.
Associations between FVL and thrombosis were first calculated separately for SLE patients and for aPL (anti-phospholipid antibody) positive patients. Calculations were restricted to European-derived subjects (n=1447) since FVL is present exclusively in European-derived populations. Bivariate analyses were performed in which each predictor was tested for association with thrombosis (). Bivariate analysis in European-derived subjects with SLE demonstrated statistically significant associations between thrombosis and the following predictors: FVL (OR 1.81, 95% C.I. 1.05-3.06), aPL positivity (LAC or aPL) (OR 2.84, 95% C.I. 2.11-3.82), and age (OR 1.19 per 10 years, 95% C.I. 1.06 - 1.32) but not with gender or smoking status. Bivariate analyses in European-derived subjects who were aPL positive demonstrated statistically significant associations between thrombosis and FVL (OR 2.34, 95% C.I. 1.13-5.02), gender (OR 0.34 for female gender, 95% C.I. 0.21 - 0.57), and smoking (OR 1.62, 95% C.I. 1.08-2.42) but not with age or SLE status.
Individual Patient Data Meta-Analysis Bivariate Results for Association with Thrombosis in European-derived Subjects
In a logistic regression model examining European-derived patients (), age (OR 1.20 per 10 years, 95% C.I. 1.09 - 1.33), gender (OR 0.44 for female gender, 95% C.I. 0.30 - 0.63), FVL (OR 1.91, 95% C.I. 1.16 - 3.16), smoking status (OR 1.38, 95% C.I. 1.05 - 1.82) and aPL positive status (OR 4.54, 95% C.I. 3.45 - 5.97) were significantly associated with thrombosis. This model evaluated the association of our primary predictor (FVL) with thrombosis while adjusting for the other covariates. When this logistic regression analysis was restricted to SLE patients, similar results were obtained with age (OR 1.29, 95% C.I. 1.14 - 1.46), and aPL positivity (OR 4.05, 95% C.I. 2.92 - 5.61) significantly associated with thrombosis and borderline significant results supporting an increased thrombosis risk associated with FVL (OR 1.76, 95% C.I. 0.99 - 3.12). Therefore, even when adjusting for other covariates, FVL was still associated with thrombosis, confirming the results of the original meta-analysis with a slightly lower OR for the association of FVL with thrombosis. Interaction terms were also evaluated between FVL and smoking and FVL and aPL. No evidence of interaction was found.
Individual Patient Data Meta-Analysis Multivariate Results for Association with Thrombosis in European-derived Subjects
Sensitivity analyses were performed in order to determine whether the results would be substantially different when certain model parameters were changed. If the results differ substantially when a small change is made in a parameter then the results are sensitive to that particular parameter. For example, several individual studies considered both venous and arterial thromboses in their outcome definition, however, FVL has been shown to be a risk factor specifically for venous thromboses. To address this point, we first did a sensitivity analysis on 936 European-derived SLE patients for whom venous thrombosis information was available (). FVL (OR 3.19, 95% C.I. 1.60-6.35) and aPL (OR 4.49, 95% C.I. 2.71-7.43) were risk factors for venous thrombosis while female gender was protective (OR 0.46, 95% C.I. 0.23-0.91).
Sensitivity analyses for individual patient data for association of the factor V Leiden polymorphism with venous thrombosis and deep venous thrombosis (DVT) in European-derived Subjects with SLE
We also performed a sensitivity analysis on a group of 942 European-derived SLE subjects for whom DVT information was specifically available (a subset of the subjects with venous thrombosis). Results were similar with gender (OR 0.47 for female, 95% C.I. 0.23 - 0.96), FVL (OR 2.33, 95% C.I. 1.09 - 4.92), and aPL (OR 4.19, 95% C.I. 2.54 - 6.91) being statistically significant risk factors. Of note, the OR for FVL was higher when these venous outcomes were considered individually.
Subgroup analyses were also performed in European SLE patients for whom the explanatory variable aPL was considered separately as LAC and aCL (IgG or IgM). In these smaller analyses (n=329), FVL (OR 4.84, 95% C.I. 1.30-17.96) and LAC (OR 8.84, 95% C.I. 3.35-23.29) were statistically significant risk factors for thrombosis.
To assess representativeness of the included datasets, we compared included studies to those which did not contribute individual patient-level data. Included studies spanned the publication dates of all studies so results were not biased toward recent publications. Included studies were all from English-speaking nations including the USA, England, or Canada. While most studies included SLE subjects, five of the six studies included in the individual patient data meta-analysis recruited based on aPL positive status, which may have biased towards subjects likely to have thrombosis. The addition of 1210 SLE subjects from the UCSF Lupus Genetics Collection recruited based on SLE status, however, should help to reduce this bias.