In a systematic review and meta-analysis, we aimed to summarize data of all observational studies prospectively assessing the association of heterozygous FVL mutation and recurrent VTE. We aimed to calculate relative risks (RR) overall and in various subgroups of patients. To set this into context, we observed the frequency of testing in Switzerland using a large claim-based dataset.

A comprehensive search strategy for MEDLINE (1946 to February 03, 2022) and EMBASE (1974 to 2022 February 03) databases was developed, and the Ovid interface used (Supplementary List 1). The search strategy was based on three elements: heterozygous FVL mutation (patients); recurrent VTE (outcome); and prospective cohort study (study design). The search strategy was improved using keywords found in key publications and no limits were applied. The sensitivity was tested in eight key publications (100%). The literature search was completed by hand search using reference lists of articles retrieved. All included studies were checked for published errata. The last search run was done on the fourth of February 2022. All records were carefully assessed for eligibility by screening of title, abstract and full text by two reviewers in duplicate (D.E., M.N.).

The following inclusion criteria were applied: (a) prospective cohort studies, (b) patients tested for FVL mutation/ activated protein C resistance (APCR) at baseline, (c) objectively confirmed VTE, (d) recurrent VTE defined as primary outcome, and (e) numbers of recurrences or recurrence rates reported separately in patients with and without FVL mutation. Exclusion criteria were (1) retrospective studies, (2) case-control studies, case reports, and (3) studies conducted in close subgroups (e.g., children, perioperative VTE, upper extremity deep venous thrombosis, and homozygous FVL mutation). Articles based on the same cohort were compiled and the publication with the (a) highest number of patients, and (b) most complete clinical data were selected for meta-analysis.

Recurrent VTE was defined as objectively confirmed VTE. For deep venous thrombosis (DVT), one of the following imaging techniques must have been used: venography, duplex sonography, or compression ultrasonography. For pulmonary embolism (PE), ventilation-perfusion scan, spiral computed tomography, or pulmonary angiography should have been used (56–58).

First, several characteristics were retrieved to summarize each cohort: name of cohort, country, setting (type of health care institution), time period of patient recruitment, inclusion criteria and all publications. Secondly, detailed data were extracted out of the selected publication for meta-analysis: first author, year of publication, age of patients (mean or median), total number of patients, number of female patients, number of FVL mutation patients (at baseline), location of initial VTE (isolated distal DVT, proximal DVT/PE or mixed DVT/PE), triggering factor first VTE (unprovoked, provoked, mixed), duration of anticoagulation (months), type of anticoagulation [Vitamin-K antagonist (VKA), direct oral anticoagulants (DOAC)], absolute number of patients with unprovoked VTE, number of cancer patients, observation period (months), total number of patients with recurrence, number of FVL mutation patients with recurrence, number of non-FVL mutation patients with recurrence and recurrence rate of FVL mutation patients.

The methodological quality of the primary studies was assessed using the Newcastle-Ottawa Scale (NOS) for cohort studies (59). The following three domains were applied: (a) selection of patients, (b) comparability of study groups, and (c) outcome of interest. The questions were modified to fit the present research question (Supplementary List 2). The assessment was done in duplicate (D.E., M.N.) and discrepancies were resolved by discussion.

To set this analysis into context, we assessed the frequency and trends of testing for FVL mutation in the Swiss health care system. Health care claims data of Helsana, one of the largest Swiss health insurance companies were used. Approximately 15% of the Swiss population are insured with Helsana for obligatory basic insurance, and the population is considered representative (60, 61). All invoices submitted for reimbursement for FVL mutation (#6200.64) and APCR (#1086.00) of the list of analyses from the Federal Office of Public Health were retrieved between 2014 and 2020 (62).

Using the extracted data, the relative risks (RR) and their 95% confidence intervals (CI) were calculated for each primary study. The RRs were then calculated using a random-effects model based on the Mantel-Haenszel estimator, and the corresponding 95% CI were computed. Heterogeneity between studies was assessed using Higgins' I². All analyses were performed using the “meta,” “etaphor,” and “dmetar” packages for R. As a first sensitivity analysis, a leave-one-out analysis was performed to check for outliers. Studentized residuals and Cook's distance were calculated, and studies with studentized residuals outside of −1 and 1, and Cook's distances >50% of the lower tail of a Chi-square distribution with p (p = number of model coefficients) degrees of freedom were flagged as potentially influential outliers. These studies were excluded from the overall analysis. Furthermore, subgroup analysis was performed for the following subgroups: Year of publication (<2000, 2000–2010, >2010), location of the initial VTE (mixed, proximal DVT/PE), presence of triggering risk factors for the initial VTE (unprovoked, provoked, and mixed), the anticoagulation drug used (VKA, DOAC), and the presence of cancer (no cancer, mixed). A funnel plot was additionally created to assess publication bias.

The literature search retrieved a total of 2,581 publications, 2,573 accessed in MEDLINE and EMBASE databases, and eight identified by manual review (Figure 1). After removing duplicates, the title and abstract of the 2,259 remaining publications were screened, giving 131 publications for full-text screening (including 100 journal articles and 21 conference abstracts). A total of 67 publications were excluded with reasons. Eventually, 31 different prospective cohort studies were identified (Table 1; Figure 1). Per cohort, the publication with the most complete clinical data was selected for further analysis. No publication with sufficient data were identified in seven cohorts (44, 45, 47, 54, 80, 101, 104). Twenty-four publications were finally considered for meta-analysis (Figure 1) (17, 36–41, 46, 48, 63–65, 67, 69–71, 73, 75, 78, 82–84, 86, 103).

Thirty-one prospective cohort studies conducted in Europe (n = 23), North America (n = 3), Europe and North America (n = 2), and other areas (n = 3) were identified. The number of publications per cohort ranged from 1 (17, 38–41, 45–48, 54, 63, 64, 69, 70, 73, 80, 82, 83, 103, 104) to 15 (86). Twenty-three cohorts included patients with a first VTE (36, 39–41, 44–46, 48, 54, 64, 65, 67, 69, 70, 75, 78, 80, 82, 84, 86, 101, 103, 104), and eight cohorts included patients with any VTE (17, 37, 38, 47, 63, 71, 73, 83). Detailed cohort characteristics are reported in Table 1.

Details of the primary studies included in the meta-analysis are reported in Table 2, summarizing data of 13,571 patients, including 2,840 patients with FVL mutation (21%). The number of patients varied between 72 (83) and 1,267 (37). The prevalence of FVL mutation ranged between 8.4% (36) and 28% (86). The mean or median age varied between 37 years (40) and 76 years (48). The observation periods varied from six (83) to 88 (63, 65) months. VKA were used in most studies (36, 37, 41, 46, 48, 64, 65, 67, 69, 71, 73, 75, 78, 82, 84, 86, 103), summarizing 8,654 patients (64%). DOAC were used as anticoagulant in one study (17), and the type of anticoagulant was not specified in six studies (38–40, 63, 70, 83). The inclusion criteria and the type and location of the primary event is reported in Table 1. Eight studies included patients with a first unprovoked VTE only (36, 41, 48, 69, 73, 84, 86, 103) and one study provided separate data (provoked/unprovoked) (67). Both provoked and unprovoked VTE were included in 15 primary studies (17, 37–40, 46, 63–65, 70, 71, 75, 78, 82, 83). Patients with cancer were excluded in 16 studies (36, 39–41, 46, 48, 64, 65, 67, 69, 73, 78, 82, 84, 86, 103) and not reported in two studies (63, 83). Overall, 341 cancer patients were reported in six studies (17, 37, 38, 70, 71, 75). A funnel plot is given in Supplementary Figure S1.

A summary of the methodological quality according to the NOS tool is given in Figure 2; detailed results for all studies are reported in the Supplementary Table S1. With at least six NOS criteria fulfilled in twenty studies, the overall methodological quality was high (17, 36–41, 48, 63–65, 67, 69, 71, 73, 75, 78, 82, 84, 86). Three to five criteria were fulfilled by four studies (46, 70, 83, 103). The three domains most frequently not met were (1) method reported for distinguishing the initial and recurrent VTE (37, 39–41, 46, 48, 63, 65, 67, 70, 71, 83, 103), (2) follow-up longer than 2 years (17, 69, 75, 83, 84, 103), and (3) follow-up rate ≥90% of patients (17, 40, 41, 46, 48, 67, 83, 103).

Assessing all studies for potential influential outliers using statistical criteria (105), we identified the study from Franco Moreno et al. (103) (Supplementary Table S2; Supplementary Figure S2). Thus, this study was excluded for the purpose of the overall analysis. A recurrent event was recorded in 1,867 individuals (14%). Recurrent events were observed in 18% of the FVL mutation patients and in 13% of the non-FVL mutation patients. Details are reported in Table 2. The relative risk was 1.46 (95% CI: 1.31, 1.64, I² = 0.17; 95% prediction interval 1.10, 1.94) (Figure 3).

In several sensitivity analyses, we assessed the risk in specific subgroups. Among the primary studies, the RR varied between 0.45 (95% CI: 0.11, 1.79) (69) and 4.77 (95% CI: 1.55, 14.68) (67). A RR smaller than one was calculated in four primary studies (48, 69, 70, 73). Focusing on different anticoagulants, the RR was 1.65 (95% CI: 1.33, 2.04) in patients treated with VKA, and 1.28 (95% CI: 0.27, 6.08) in patients treated with DOAC (Supplementary Figure S3). Pooling studies with unprovoked VTE only, the RR was 1.53 (95% CI: 0.99, 2.35) (Supplementary Figure S4). It was 1.47 (95% CI: 1.27, 1.71) in studies including both, patients with provoked and unprovoked VTE. In one study group (67), patients with a first provoked VTE only were analyzed, resulting in a RR of 4.77 (95% CI: 1.55, 14.68). Considering different localizations of the initial event, the RR was 1.29 (95% CI: 0.28, 6.08) in patients with PE (Supplementary Figure S5), and 1.52 (95% CI: 1.2, 1.93) in patients with proximal DVT or PE. It was 1.6 (95% CI: 1.32, 1.95) in patients with proximal DVT/PE or distal DVT. Excluding patients with cancer, the RR was 1.59 (95% CI: 1.27, 1.99) (Supplementary Figure S6). The RR was 1.69 (95% CI: 1.14, 2.51) in studies published after 2011, 1.52 (95% CI: 1.33, 1.75) in studies published between 2001 and 2011, and 1.44 (95% CI: 0.77, 2.68) in studies published before 2001 (Supplementary Figure S7).

Analysis of Helsana health care claims data in Switzerland showed that 46,522 APCR tests and 49,625 polymerase-chain reaction (PCR) for FVL mutation were recorded between 2014 and 2020 (Supplementary Table S3; Supplementary Figure S8). For APCR, the frequency of testing varied between 6,206 (0.1% of the population, 2014) and 7,206 (0.1%, 2016). For PCR, the frequency ranged between 6,793 (0.1%, 2017) and 7,614 (0.1%, 2019). Considering patients having any test, the total number of patients with APCR and/or PCR varied between 9,661 (0.2%, 2017) and 10,614 (0.2%, 2016). The frequency of testing was stable between 2014 and 2020.