Classic Thrombophilias and Thrombotic Risk Among Middle-Aged and Older Adults: A Population-Based Cohort Study.

Background Five classic thrombophilias have been recognized: factor V Leiden (rs6025), the prothrombin G20210A variant (rs1799963), and protein C, protein S, and antithrombin deficiencies. This study aimed to determine the thrombotic risk of classic thrombophilias in a cohort of middle-aged and older adults. Methods and Results Factor V Leiden, prothrombin G20210A and protein-coding variants in the PROC (protein C), PROS1 (protein S), and SERPINC1 (antithrombin) anticoagulant genes were determined in 29 387 subjects (born 1923-1950, 60% women) who participated in the Malmö Diet and Cancer study (1991-1996). The Human Gene Mutation Database was used to define 68 disease-causing mutations. Patients were followed up from baseline until the first event of venous thromboembolism (VTE), death, or Dec 31, 2018. Carriership (n=908, 3.1%) for disease-causing mutations in the PROC, PROS1, and SERPINC1 genes was associated with incident VTE: Hazard ratio (HR) was 1.6 (95% CI, 1.3-1.9). Variants not in Human Gene Mutation Database were not linked to VTE (HR, 1.1; 95% CI, 0.8-1.5). Heterozygosity for rs6025 and rs1799963 was associated with incident VTE: HR, 1.8 (95% CI, 1.6-2.0) and HR, 1.6 (95% CI, 1.3-2.0), respectively. The HR for carrying 1 classical thrombophilia variant was 1.7 (95% CI, 1.6-1.9). HR was 3.9 (95% CI, 3.1-5.0) for carriers of ≥2 thrombophilia variants. Conclusions The 5 classic thrombophilias are associated with a dose-graded risk of VTE in middle-aged and older adults. Disease-causing variants in the PROC, PROS1, and SERPINC1 genes were more common than the rs1799963 variant but the conferred genetic risk was comparable with the rs6025 and rs1799963 variants.

American College of Medical Genetics and Genomics

Human Gene Mutation Database

Malmö Diet and Cancer cohort

principal component analysis

What Is New?

To our knowledge, this is the first population‐based genetic study of all 5 classical thrombophilias, factor V Leiden (rs6025), the prothrombin G20210A variant (rs1799963), and PROC (protein C), PROS1 (protein S), and SERPINC1 (antithrombin) deficiencies, in middle‐aged and older adults.

The 5 classical thrombophilias are associated with a dose‐graded risk of incident VTE in middle‐aged and older adults.

The conferred genetic risk of disease‐causing variants in the PROC, PROS1, and SERPINC1 genes is comparable with the rs6025 and rs1799963 variants.

What Are the Clinical Implications?

The 5 classical thrombophilias are common risk factors for incident venous thromboembolism in a general population of middle‐aged and older individuals.

Although new risk variants for venous thromboembolism (VTE) have been discovered, only 5 classic genetic risk factors are widely recognized for thrombophilia. , , , , , , The classic thrombotic risk factors are factor V Leiden (rs6025), the prothrombin G20210A (rs1799963) variant, and inherited deficiencies of the natural anticoagulants antithrombin (coded for by SERPINC1), protein C (PROC), and protein S (PROS1). , , , , The risk of VTE attributed to the common rs6025 and rs1799963 variants has been confirmed in large cohort studies. , , The carrier frequency among White people for the rs6025 variant is around 5% to 10% and around 2% for the rs1799963 variant. , , , , , , The relative VTE risk for the rs6025 variant was 2.7 (95% CI, 1.3 to 5.6) and 1.7 (95% CI, 0.9–3.1) for the rs1799963 variant in 2 cohort studies by Ridker et al. , In a large Danish study the relative risk for the rs6025 variant was 2.2 (95% CI, 2.0–2.5) and 1.5 (95% CI, 1.2–1.9) for the rs1799963 variant. These risk estimates for the common rs6025 and rs1799963 variants are lower than initially estimated among selected patients in case‐control studies and family studies. , , , ,

Inherited deficiencies of the natural coagulation inhibitors antithrombin, protein C, and protein S are all deemed to be strong but rare (<1%) risk factors for VTE. , , , , , , Several hundreds of causative mutations have been described that are responsible for these deficiencies. Due to the rareness of these deficiencies, the estimated VTE risk of inherited deficiencies of the natural anticoagulants antithrombin, protein C, and protein S has been determined mainly in studies of selected thrombosis‐prone families. The increased risk of VTE in patients with inherited deficiencies of these 3 anticoagulant proteins was estimated to be 10‐fold in these studies. , , , , However, the MEGA case‐control study could not confirm that protein S deficiency was associated with VTE, and deficiencies of antithrombin, protein C, and protein S have also been detected in healthy individuals without personal and/or family history of VTE, thereby questioning their importance. , , , ,

Variants causing such deficiencies (ie, low antigen and/or functional levels) are described in the Human Gene Mutation Database (HGMD). Deficiencies of the 3 natural anticoagulants are either quantitative (type I or III) or qualitative (type II). , , , , The importance of inherited deficiencies of antithrombin, protein C, and protein S remains to be determined in large population‐based cohort studies of the general population. Moreover, classic thrombophilias are believed to be of importance mainly in young people aged <50 years. The importance of classic thrombophilias among the general population of middle‐aged and older adults remains to be determined, although VTE is most common among middle‐aged and older adults.

In the present study the prevalence and the VTE‐associated risk of disease‐causing variants in the SERPINC1, PROS1 and PROC genes were analyzed in addition to factor V Leiden and the prothrombin G20210A variants in the Malmö Diet and Cancer cohort (MDC). The MDC is a follow‐up study of middle‐aged and older adults.

Methods

Study Population

Because of ethical and legal restrictions related to the Swedish Biobanks in Medical Care Act (2002:297) and the Personal Data Act (1998:204), data are available upon request from the data access group of MDC Study by contacting Anders Dahlin (anders.dahlin@med.lu.se). The MDC is a population‐based prospective cohort study from the city of Malmö in the south of Sweden. Sample characteristics, data collection, and clinical definitions for the MDC have been described previously. A total of 30 446 individuals, men (n=12 120, born 1923–1945) and women (n=18 326, born 1923–1950), out of an eligible population of ≈74 000 individuals attended a baseline examination between March 1991 and September 1996. DNA testing was available for 29 387 subjects sampled at baseline. DNA was extracted from peripheral blood cells and assigned to batches without regard to disease status or personal identity. The ethics committee at Lund University Lund, Sweden approved the study (LU 51/90) and all participants provided informed written consent.

Clinical Examination

Participants underwent a medical history, physical examination, and laboratory assessment at baseline. Blood pressure was measured using a mercury‐column sphygmomanometer after 10 minutes of rest in the supine position. Cigarette smoking was determined by a self‐administered questionnaire. Subjects were categorized as current smokers (ie, those who smoked regularly or occasionally) or non‐smokers (ie, former smokers and never smokers). High alcohol consumption was defined as >40 g alcohol per day for men and >30 g per day for women. Weight and height were measured to the nearest 0.1 kg and 0.5 cm, respectively, with subjects wearing light clothing and no shoes. Current body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared (kg/m2). One primary outcome, VTE, and 2 secondary outcomes, deep venous thrombosis (DVT) and pulmonary embolism (PE) were examined. VTE, DVT, and PE were defined based on the International Classification of Diseases, Seventh to Tenth Revisions (ICD‐7 to ICD‐10) (Table S1). VTE events (main and secondary diagnoses) were identified through linkage of the 10‐digit personal identification number of each Swedish citizen with the Swedish National Patient Register. The Swedish National Patient Register had a 100% coverage for inpatients in Malmö from 1970 until end of follow‐up (Dec 31, 2018) and for outpatients from 2001. The diagnosis of VTE in the Swedish National Patient Register has been shown to have an accuracy of 95%, whereas the overall validity of the Swedish National Patient Register is 87%. Events were determined both before baseline (prevalent events) and after baseline during follow‐up (incident events). Patients with VTE in Malmö University Hospital are diagnosed with objective methods such as phlebography, compression ultrasound, ventilation/perfusion lung scan (V/Q lung scan), computed tomography (CT), or magnetic resonance tomography.

Exome Sequencing and Genetic Analysis

Whole‐exome sequencing was performed by Regeneron Genetics Center (Tarrytown, NY). Targeted exonic regions were captured using a slightly modified version of the xGen probe library (Integrated DNA Technologies). Captured DNA was PCR amplified and sequenced with v4 chemistry using 75 bp paired‐end reads on the Illumina HiSeq 2500. Whole‐exome sequencing was performed such that >85% of targeted bases are covered at a read depth of >20×. We used previously established bioinformatics procedures to process and analyze exome sequence data. ANNOVAR (2019‐10‐24) was used to aggregate variant annotation, allele frequencies, and in silico predictions of deleteriousness. The rs6025 and the rs1799963 variants were determined from whole‐exome sequencing data. The HGMD database (http://www.hgmd.cf.ac.uk/ac/index.php) was used to define high‐risk variants, ie, disease causing, in the SERPINC1, PROC, and PROS1 genes. In June 2020, the HGMD database contained >289 000 different gene lesions identified in >11 100 genes manually curated from 72 987 articles published in >3100 peer‐reviewed journals. There is generally a good agreement between the variant classification in HGMD and the American College of Medical Genetics and Genomics (ACMG) guidelines and the ClinVar database. When variants in HGMD were classified based on the ACMG guidelines, misclassification was observed in only 3.47% (2289/65 896) of variants. The overall concordance between HGMD and the ClinVar database was 97.62% (52 499/53 780) of variants studied. However, we also read the cited articles in the HGMD database and searched PubMed and could confirm that the HGMD variants were correctly curated. All non‐synonymous variants present in the HGMD or affecting the same codon as a variant already occurring in the HGMD were defined as high‐risk variants. This is in accordance with the PM5 criteria of ACMG (moderate evidence of pathogenicity): Novel missense change at an amino acid residue where a different missense change determined to be pathogenic has been seen before. Moreover, we analyzed the twelve‐missense mutations occurring at an amino acid that HGMD has reported to be pathogenic with meta‐analytic support vector machine. Nine out of 12 (75%) missense variant were pathogenic in meta‐analytic support vector machine. We have therefore chosen to designate all as high‐risk variants as these results supported the PM5 criteria of ACMG.

Principal component analysis (PCA) was performed as described. The reference genomes were obtained from the 1000 Genomes Project server. , The PCA was performed with independent (R2 measure of linkage disequilibrium <0.2) common (minor allele frequency ≥5%) autosomal biallelic variants that were detected in both the reference genomes and the MDC exomes. To avoid extended linkage disequilibrium and high variability regions, such as the major histocompatibility complex, these regions were omitted from the PCA. The principal components were first obtained from the reference genomes and then projected individuals from the MDC onto the principal‐component space via PLINK2. ,

Statistical Analysis

R version 4.0.0 was used for all statistical analyses. The contribution to VTE risk was assessed for sex, age, BMI, smoking status, systolic blood pressure, high alcohol consumption, rs6025, rs1799963 and high‐risk variants in PROC, PROS1, and SERPINC1. Logistic regression was conducted to calculate crude odds ratios (ORs) for 4 dependent variables (outcomes): all VTE (prevalent and/or incident VTE), prevalent VTE (VTE between 1970 and baseline), incident VTE (VTE after baseline with no prevalent VTE), and recurrent VTE (≥2 incident VTE during follow‐up without prevalent VTE) compared with those without a VTE. Prevalent VTE cases were excluded for determination of OR for incident and recurrent VTE. The Martingale residuals were used for checking linearity of the predictor with logit‐transformed outcome for continuous variables (age, BMI and systolic blood pressure). No important deviations from linearity were observed. Prevalent VTE cases were excluded for determination of incident VTE. Incidence rates were calculated as a measure for absolute risk and incidence rate ratios were also presented as a robust measure of relative risk. Cox proportional hazards regression was used to examine the crude and adjusted association between genotype and incident VTE, DVT, and PE. Time was given as the number of years from baseline examination until death, emigration, incident VTE, or end of follow‐up, whichever occurred first. Age and sex were included as covariates in the sex‐ and age‐adjusted model. The multivariable model also included BMI, smoking status, blood pressure (systolic), and high alcohol consumption (>30 g/d women, >40 g/d men) as covariates. We also included the top 2 eigenvectors from the PCA analysis as covariates in the Cox proportional hazard regression models to control for population stratification. The fit of the proportional hazards model was checked visually by plotting the incidence rates over time and by calculating Schoenfeld (partial) residuals. Schoenfeld residuals were used as a dependent variable and time as an independent variable, to assess the proportional hazards assumption. No violation against the proportional hazards assumption was observed. Possible interactions between gene variants and age, sex, and risk factors (BMI, smoking status, systolic blood pressure, high alcohol consumption) on VTE was explored by introducing interaction terms in the multivariable models. No interactions were observed. The subjects were categorized according to genotype and Kaplan–Meier plots were calculated for VTE. For curve comparisons, the log‐rank test was used.

Results

Cohort Characteristics

A total of 29 387 individuals were available for analysis. The baseline characteristics of the study population are presented in Table 1. A total of 3177 (10.8%) individuals (1869 women, 1308 men) were affected by VTE (prevalent and/or incident VTE). Of these individuals 593 (2.0%) (315 women, 278 men) were affected by VTE between 1970 and baseline, and 2584 (8.8%) individuals were only affected after baseline, ie, incident VTE during follow‐up without previous VTE until December 31, 2018. In total, 1491 individuals were affected by recurrent incident VTE during follow‐up.

Table 1

Cohort Characteristics for Complete Cohort, No VTE (ie, Individuals With No VTE Between 1970–2018), all VTE (ie, VTE Between 1970–2018), Prevalent VTE (ie, VTE between 1970 and Baseline), Incident VTE During Follow‐Up Without Prevalent VTE, and Recurrent VTE During Follow‐Up (ie, at Least Two VTE During Follow‐Up Without Prevalent VTE)

Independent variables (predictors)	Complete cohort n=29 387	No VTE n=26 210	All VTE n=3177	Prevalent VTE n=593	Incident VTE n=2584	Recurrent VTE n=1491
Female sex (%)	17 687 (60.2%)	15 818 (60.4%)	1869 (58.8%)	315 (53.1%)	1554 (60.1%)	890 (59.7%)
Age at baseline, y (mean±SD)	58.0±7.6	57.8±7.6	59.7±7.4	60.7±7.7	59.4±7.3	59.1±7.2
Body mass index (mean±SD)	25.8±4.0	25.7±4.0	26.8±4.4	27.7±4.7	26.6±4.3	26.8±4.3
Current smoking (%)	7797 (26.5%)	6975 (26.6%)	822 (25.9%)	146 (24.6%)	676 (26.2%)	380 (25.5%)
Systolic blood pressure (mean±SD)	141±20	141±20	142.5±20.1	143.9±20.6	142.2±20.0	141.7±20.2
High alcohol consumption (%)	1177 (4.0%)	1057 (4%)	120 (3.8%)	23 (3.9%)	97 (3.8%)	51 (3.4%)
rs6025 heterozygotes (%)	3282 (11.2%)	2679 (10.2%)	603 (19.0%)	153 (25.8%)	450 (17.4%)	292 (19.6%)
rs6025 homozygotes (%)	119 (0.4%)	75 (0.3%)	44 (1.4%)	16 (2.7%)	28 (1.1%)	21 (1.4%)
rs1799963 heterozygotes (%)	542 (1.8%)	452 (1.7%)	90 (2.8%)	18 (3.0%)	72 (2.8%)	53 (3.6%)
rs1799963 homozygotes (%)	2 (0%)	1 (0%)	1 (0%)	0 (0%)	1 (0%)	1 (0%)
≥1 high‐risk variant* (%)	908 (3.1%)	755 (2.9%)	153 (4.8%)	34 (5.7%)	119 (4.6%)	74 (5.0%)
≥1 low‐risk variant † (%)	483 (1.6%)	428 (1.6%)	55 (1.7%)	10 (1.7%)	45 (1.7%)	28 (1.9%)

VTE indicates venous thromboembolism.

High‐risk variant=non‐synonymous Human Gene Mutation Database variants in the 3 anticoagulant genes SERPINC1, PROC, and PROS1 (see Methods section).

Low‐risk variant=Low‐risk non‐synonymous non‐Human Gene Mutation Database variants in the 3 anticoagulant genes SERPINC1, PROC, and PROS1 (see Methods section).

Genetic Findings

The prevalence of heterozygotes for rs6025 was 11.2% (n=3282), whereas it was 1.8% (n=542) for rs1799963. The prevalence of homozygotes for rs6025 was 0.4% (n=119), in comparison with 0.007% (n=2) for rs1799963. Excluding synonymous variants, a total of 171 different coding variants were identified in the PROC (55 variants), PROS1 (69), and SERPINC1 (47) genes (Tables S2 through S5). In total, 68 variants were identified as high risk or disease causing (see Methods for definition), whereas 113 (66%) variants were present in the gnomAD database and all were rare (minor allele frequency <0.01 among the non‐Finnish European population in gnomAD). Carriership for any of the 68 different high‐risk variants in any of the SERPINC1 (n=14), PROC (n=29) or PROS1 (n=25) genes was 3.1% (n=908). In a few cases (n=12) the obtained HGMD accession number referred to another amino acid shift in the same codon. Among variants known to be associated with deficiency (decreased antigen and/or functional activity), some were associated with type II deficiency (functional deficiency with normal antigen levels): 9 out of 24 (37.5%) PROC variants, 6 out of 23 (26%) PROS1 variants, and 7 out of 9 (78%) SERPINC1 variants. Three individuals were heterozygotes for Factor V Cambridge (rs118203906). One was affected by incident VTE and 2 were unaffected. No other rare variants associated with VTE in the F5 or F2 genes were observed.

The MDC population has only 12% admixture from foreign‐born individuals. Among foreign‐born individuals only 1% were non‐European. To evaluate the population structure and admixture of the MDC population on the basis of the genetic data, we constructed principal components from 3 ancestral super populations (European, East Asian, and African ancestries) from the 1000 Genomes Project and projected MDC study subjects onto the principal‐component space. When all results from individuals from the MDC study were plotted, most study participants were found to cluster together (Figure S1). These results confirm that the MDC population is mainly comprised of individuals with complete or partial European White ancestry.

Risk of VTE

The crude ORs for all VTE, prevalent VTE, incident VTE, and recurrent incident VTE were increased for rs6025, rs1799963, and high‐risk variants but not low‐risk variants (Table 2 and Table S6). Higher ORs for VTE for the rs6025, rs1799963, and high‐risk variants were found for those with prevalent VTE compared with incident VTE, although 95% CI overlapped. The genotype ORs were also higher for recurrent incident VTE than for those with only 1 incident VTE event, although 95% CIs overlapped.

Table 2

Crude Odd Ratios for All VTE (ie, VTE Between 1970 and Baseline and/or During Follow‐Up), Prevalent VTE (ie, VTE Between 1970 and Baseline), Incident VTE During Follow‐Up (Without Prevalent VTE), and Recurrent VTE During Follow‐Up (Without Prevalent VTE)

Independent variables (predictors)	All VTE n=3177 OR (95% CI)	P value	Prevalent VTE n=593 OR (95% CI)	P value	Incident VTE n=2584 OR (95% CI)	P value	Recurrent VTE n=1491 OR (95% CI)	P value
Female sex	0.94 (0.87–1.04)	0.10	0.74 (0.63–0.88)	4.0e‐4	0.99 (0.91–1.08)	0.81	0.97 (0.87–1.08)	0.61
Age at baseline	1.03 (1.03–1.04)	2.1e‐38	1.05 (1.04–1.06)	4.5e‐20	1.03 (1.02–1.03)	1.1e‐24	1.02 (1.01–1.03)	4.2e‐10
Body mass index	1.07 (1.06–1.07)	3.2e‐49	1.10 (1.08–1.12)	1.0e‐29	1.06 (1.05–1.07)	4.0e‐29	1.06 (1.05–1.08)	6.0e‐25
Current smoking	0.96 (0.88–1.05)	0.37	0.90 (0.74–1.08)	0.28	0.98 (0.89–1.07)	0.62	0.94 (0.84–1.06)	0.33
Systolic blood pressure	1.00 (1.00–1.01)	3.8e‐05	1.01 (1.00–1.01)	8.7e‐04	1.00 (1.00–1.01)	2.5e‐03	1.00 (1.00–1.00)	0.17
High alcohol consumption	0.93 (0.77–1.13)	0.49	0.96 (0.61–1.43)	0.85	0.93 (0.75–1.14)	0.49	0.84 (0.63–1.10)	0.24
rs6025 heterozygotes	2.06 (1.87–2.27)	6.6e‐48	3.05 (2.52–3.68)	3.1e‐31	1.85 (1.66–2.06)	2.0e‐28	2.14 (1.87–2.44)	9.8e‐29
rs6025 homozygotes	4.89 (3.34–7.08)	8.7e‐17	9.66 (5.41–16.23)	3.9e‐16	3.82 (2.43–5.83)	1.7e‐09	4.98 (2.99–7.95)	1.0e‐10
rs1799963 heterozygotes	1.66 (1.31–2.08)	1.4e‐05	1.78 (1.07–2.79)	0.018	1.63 (1.26–2.09)	1.3e‐04	2.10 (1.56–2.78)	5.0e‐7
≥1 high‐risk variant*	1.71 (1.43–2.03)	1.4e‐09	2.00 (1.37–2.82)	1.5e‐04	1.63 (1.34–1.98)	7.5e‐07	1.76 (1.37–2.22)	4.2e‐6
≥1 low‐risk variant * (%)	1.05 (0.79–1.38)	0.73	1.02 (0.51–1.81)	0.94	1.06 (0.77–1.42)	0.72	1.14 (0.76–1.64)	0.50

For the dependent variables all outcomes were compared with no venous thromboembolism (1970–2018).

Prevalent cases (n=593) were excluded when calculating odds ratio for incident venous thromboembolism and recurrent venous thromboembolism. The independent variables age, body mass index, and systolic blood pressure were included as continuous variables. OR indicates odds ratio; and VTE, venous thromboembolism.

High‐risk variant=High‐risk non‐synonymous Human Gene Mutation Database variants in the 3 anticoagulant genes SERPINC1, PROC, and PROS1 (see method section).

Low‐risk variant=Low‐risk non‐synonymous non‐Human Gene Mutation Database variants in the 3 anticoagulant genes SERPINC1, PROC, and PROS1 (see Methods section).

During a median follow‐up of 23.1 years (interquartile range, 16.8–24.9 years), a total of 2584 incident VTE events occurred (1030 men, 1554 women) among individuals without prevalent VTE (Table 3). The sum of the follow‐up time was 587 992.7 years, corresponding to a VTE incidence rate of 4.4 (95% CI, 4.2–4.6) per 1000 person‐years. The thrombosis‐free survival curves using Kaplan‒Meier analysis are presented in Figure for rs6025, rs1799963, and high‐risk variants in PROC, PROS1, and SERPINC1. Although there were low numbers of individuals carrying 2 alleles, a clear dose‐response is observed in relationship to numbers of affected alleles, see Figure – Panel D where all 5 classic thrombophilic gene variants are included.

Table 3

Hazard Ratios (HRs) For Incident Venous Thromboembolism Adjusted for Either Age, Sex, and Ancestry*, or Multivariable HRs Adjusted for Age, Sex, Body Mass Index, Smoking, High Alcohol Consumption, and Ancestry*

Independent variables (predictors)	Participants at risk	VTE events	Mean age at first VTE	Crude IR	Crude IRR	Age‐ and sex‐adjusted HR		Multivariable HR
	n	n	Years (SD)	IR (95% CI)	IRR (95% CI)	HR (95% CI)	P value	HR (95% CI)	P value
Complete cohort	28 794	2584	73.7 (8.6)	4.4 (4.2–4.6)	1	…	…	…	…
Model with factor V Leiden (rs6025)
Reference no rs6025	25 562	2106	74.1 (8.6)	4.0 (3.9–4.2)	1	1		1
rs6025 heterozygotes	3129	450	72.3 (8.5)	7.2 (6.6–7.9)	1.8 (1.6–2.0)	1.8 (1.6–2.0)	3.3e‐30	1.8 (1.6–2.0)	9.1e‐31
rs6025 homozygotes	103	28	69.5 (9.5)	14.5 (9.6–20.9)	3.6 (2.5–5.2)	3.8 (2.6–5.6)	1.6e‐12	4.0 (2.7–5.8)	3.9e‐13
Model with prothrombin variant (rs1799963)
Reference no rs1799963	28 268	2511	73.8 (8.6)	4.3 (4.2–4.5)	1	1		1
rs1799963 heterozygotes	524	72	71.5 (8.1)	6.9 (5.4–8.7)	1.6 (1.3–2.0)	1.6 (1.3–2.1)	4.4e‐05	1.6 (1.3–2.0)	6.7e‐05
Model with non‐synonymous variants in SERPINC1, PROC, and PROS1 genes
Reference no non‐synonymous variants	27 447	2420	73.7 (8.6)	4.3 (4.1–4.5)	1	1		1
1 low‐risk variant *	469	45	72.4 (8.6)	4.7 (3.5–6.4)	1.1 (0.8–1.5)	1.1 (0.8–1.5)	0.54	1.1 (0.8–1.5)	0.56
1 high‐risk variant †	868	117	74.0 (8.2)	6.7 (5.5–8.0)	1.6 (1.3–1.85)	1.6 (1.3–1.9)	2.9e‐06	1.6 (1.3–1.9)	2.7e‐07
Model with all 5 classical thrombophilia variants: rs6025, rs1799963, and high‐risk variants in SERPINC1, PROC, and PROS1 genes
No thrombophilia ‡	24 312	1954	74.2 (8.6)	3.9 (3.7–4.1)	1	1		1
1 thrombophilia § variant	4225	561	72.4 (8.4)	6.6 (6.1–7.2)	1.7 (1.5–1.9)	1.7 (1.6–1.9)	1.1e‐28	1.7 (1.6–1.9)	4.3e‐29
≥2 thrombophilia § variants	257	69	71.1 (9.0)	14.3 (11.1–18.1)	3.7 (2.9–4.6)	3.9 (3.0–4.9)	2.2e‐28	3.9 (3.1–5.0)	1.1e‐28

Number of participants at risk, number of venous thromboembolism cases during follow‐up, mean age at first venous thromboembolism event (SD), incidence rates, and incidence rate ratios are also presented. Prevalent cases of venous thromboembolism were excluded from the study population. HR indicates hazard ratio; IR, incidence rates; IRR, incidence rate ratios; and VTE, venous thromboembolism.

Ancestry was controlled for by including the top 2 eigenvectors from the principal component analysis as covariates in Cox proportional hazard regression models.

Low‐risk variant=Low‐risk non‐synonymous non‐Human Gene Mutation Database variants in the 3 anticoagulant genes SERPINC1, PROC, and PROS1 (see Methods section).

High‐risk variant=High‐risk non‐synonymous Human Gene Mutation Database variants in the 3 anticoagulant genes SERPINC1, PROC, and PROS1 (see Methods section).

No thrombophilia=No rs6025 allele, no rs1799963 allele, and no high‐risk variant‡.

Thrombophilia=presence of rs6025 allele, rs1799963 allele, or high‐risk variants‡ in the 3 anticoagulant genes SERPINC1, PROC, and PROS1 (see Methods section).

Figure 1

Kaplan‒Meier survival curves showing the proportion of the population remaining free from an venous thromboembolism event stratified on number of factor V Leiden alleles (A), number of rs1799963 alleles (B), number of high‐risk variants in SERPINC1, PROC, or PROS1 (C) and number of any thrombophilic variants (D).

The log‐rank test for curve comparisons, was highly significant for all A through D panels (P<0.0001). VTE indicates venous thromboembolism.

Participants at risk, VTE events, incidence rates, incidence rate ratios, and hazard ratios are presented in Table 3. The multivariable Cox regression models were adjusted for age, sex, BMI, smoking status, systolic blood pressure, high alcohol consumption, and the top 2 eigenvectors from the PCA analysis. The multivariable HRs were 1.8 (95% CI, 1.6–2.0) for heterozygous rs6025 carriers and 1.6 (95% CI, 1.3–2.0) for heterozygous rs1799963 carriers. The multivariable HR was 1.6 (95% CI, 1.3–1.9) for heterozygous high‐risk variant carriers. The HR for LOF variants was 1.6 (95% CI, 1.3–1.9) and the HR for HGMD risk variants excluding LOF was 1.6 (95% CI, 1.3–1.9). Carriers of non‐synonymous variants not in HGMD (low‐risk variants) had no increased risk of VTE (adjusted HR, 1.1, 95% CI, 0.8–1.5). The HRs were higher for those with more than one thrombophilia allele (Table 3). The multivariable HR was 4.0 (95%CI 2.7–5.8) for homozygous rs6025 carriers. The adjusted HR was 11.3 (95% CI, 1.6–80.8) for homozygous rs1799963 carriers and 5.0 (95% CI, 1.2–20.1) for carriers of >1 high‐risk variant allele in the SERPINC1, PROC, and PROS1 genes. The 95% CI were wide, and these data are therefore not presented in a table. One of 2 homozygous rs1799963 carriers and 2 of 6 carriers of >1 high‐risk variant allele had VTE. The HR for carrying 1 classical thrombophilia variant was 1.7 (95% CI, 1.6–1.9). HR was 3.9 (95% CI, 3.1–5.0) for carriers of ≥2 thrombophilia variants. Thus, the 95% CI did not overlap for those with 1 thrombophilic variant compared with those with ≥2 thrombophilic variants. Adjustment variables had no important effect on genotype risks indicating no confounding from these variables.

The multivariable HRs for VTE were significantly increased for high‐risk variants in each of the 3 anticoagulant genes: SERPINC1 (OR, 1.5; 95% CI, 1.1–1.9), PROC (OR, 2.4; 95% CI, 1.5–3.7), and PROS1 genes (OR, 1.5; 95% CI, 1.1–2.0) (Table S7). Protein S Heerlen and antithrombin Dublin were both significantly associated with incident and/or prevalent VTE (Table S8): protein S Heerlen (OR, 1.93; 95% CI, 1.22–2.94) and antithrombin Dublin (OR, 1.66; 95% CI, 1.21–2.25). Antithrombin Basel was not significantly associated with incident and/or prevalent VTE (OR 1.52; 95% CI, 0.70–2.94). In a sensitivity analysis, high‐risk variants of PROC, PROS1, and SERPINC1 genes were associated with incident VTE even if these 3 variants were excluded from the analysis (OR, 1.68; 95% CI, 1.30–2.15).

Risk of DVT and PE

Both for heterozygous and homozygous carriers of factor V Leiden (rs6025), the HR for DVT of the legs (HR, 2.3; 95% CI, 2.0–2.7, and HR, 5.4; 95% CI, 3.4–8.6) was higher than for PE (HR, 1.5; 95% CI, 1.2–1.7 and HR, 1.7; 95% CI, 0.8–3.9) (Tables S9 and S10). Otherwise, there were no clear difference for heterozygotes for the rs1799963 variant and the high‐risk PROC, PROS1, and SERPINC1 variants on DVT of the legs (HR, 1.5; 95% CI, 1.0–2.1 and HR, 1.7; 95% CI, 1.3–2.2) and PE risk (HR, 1.7; 95% CI, 1.2–2.3, and HR, 1.4; 95% CI, 1.0–1.9).

Sensitivity Analysis

The ACMG criteria were determined with Varsome (https://varsome.com/). However, the HR for carrying 1 ACMG positive variant in SERPINC1, PROC, or PROS1 was 1.4 (95% CI, 1.1–1.9). HR was 1.4 (95% CI, 0.2–10.0) for carriers of ≥2 thrombophilia variants (Table S11). A sensitivity analysis was also performed with exclusion of malignancy that occurred before cancer. No major differences were observed although the HRs tended to be slightly higher when malignancy was excluded (Table S12). For instance, the multivariable HRs were 1.9 (95% CI, 1.7–2.2) for heterozygous rs6025 carriers and 1.7 (95% CI, 1.3–2.3) for heterozygous rs1799963 carriers (Table S12). In a sensitivity analysis all related people were excluded (ie, up to second‐degree cousins). No major difference in effect size was observed (Table S13). The HR for carrying 1 classical thrombophilia variant was 1.7 (95% CI, 1.5–1.9). HR was 4.1 (95% CI, 3.1–5.3) for carriers of ≥2 thrombophilia variants (Table S13).

Additional Analysis

Compound heterozygosity of rs6025 with a high‐risk variant in SERPINC1, PROC, or PROS1 increased the VTE risk (Table S14). The crude incidence rate was 12.1 (7.6–18.4) per 1000 person‐years with an adjusted HR of 2.6 (95% CI, 1.8–3.7). No significant interaction was observed. The HR for the interaction term was 1.02 (not included in the Table). Compound heterozygosity of rs6025 or rs1799963 with a low‐risk variant in the SERPINC1, PROC, or PROS1 did not significantly increase the VTE risk but the number of individuals were few (Table S15). No significant interaction was observed between genetic risk factors and other included variables in the models (Figure S2).

Discussion

In the present study of elderly and middle‐aged adults, classic thrombophilias were found to be associated with a dose‐graded risk of incident VTE. Present recommendations , suggest thrombophilia screening before the age of 50 years among patients with VTE, but classic thrombophilias also inflict an increased risk of VTE in middle‐aged and older adults. Although the relative risk is moderate, this corresponds to a high incidence rate of VTE for classic thrombophilias in elderly and middle‐aged adults, especially for those with ≥2 risk alleles. To the best of our knowledge, this is the first large population‐based cohort study determining the prevalence and VTE risk of disease‐causing PROC, PROS1, and SERPINC1 variants among middle‐aged and older adults. PROS1, PROC, and SERPINC1 disease‐causing variants are present in the Swedish population with a frequency higher than previously anticipated from population studies based on plasma analysis of protein levels. , , , However, a recent gene‐based study suggests higher prevalence. Risk estimates were weaker than previous estimates in family studies of thrombosis‐prone families. , , , , Antithrombin type I deficiency was, however, rare, and it was not possible to estimate its VTE risk. Thus, there might exist rare variants that could cause more severe phenotypes, but it was not possible to estimate them in the present study. Most SERPINC1 variants were heparin‐binding site defects (type II antithrombin deficiency), which are known to be associated with lower VTE risk compared with type I deficiency. , , , , Deficiencies of the 3 natural anticoagulants have previously been considered to be autosomal dominant disorders with varying degrees of penetrance on VTE risk. The present study instead suggests that many variants causing deficiency of the natural anticoagulants are risk variants, such as the rs6025 and rs1799963 variants. This was already suggested in 1995 in families with segregation of both protein S deficiency and rs6025. In these families, protein S deficiency and rs6025 was of equal importance on VTE risk. Still, homozygosity for severe variants in the natural anticoagulant genes (PROC and PROS1) cause a lethal condition called purpura fulminans. , , , , Homozygosity for severe antithrombin deficiency has not been described and is probably lethal shortly after conception.

The lower‐than‐expected VTE risk observed not only for high‐risk variants of SERPINC1, PROC, and PROS1 genes but also for rs6025 and rs1799963, is partly related to the elderly and middle‐age status of the included population. Thus, this study includes an age group with high incidence of VTE, even in the absence of any known genetic predisposition, which attenuates the genetic relative risk. It has been established that the genetic contribution of complex traits such as VTE is age‐dependent. In the present study OR was generally higher for prevalent VTE than for incident VTE (Table 2), which may be explained by lower age of the individuals with prevalent VTE (VTE before baseline) than incident VTE (VTE after study baseline during follow‐up). The population‐based cohort design of the study with unbiased sampling of cases may also decrease the OR/HR because of reduction of selection bias. Compared with only 1 incident VTE event, ORs were higher for recurrent incident VTE suggesting enrichment of thrombophilic variants in patients with recurrent disease. Divergent results exist for the importance of heterozygosity for the rs6025 and rs1799963 variants in recurrent VTE. , , , ,

Protein S Heerlen (Ser501Pro), antithrombin Dublin (Val30Glu) and antithrombin Basel (Pro73Leu) are all individually reported to be associated with VTE. , , In the present study protein S Heerlen and antithrombin Dublin were significantly associated with VTE. The OR was not significant for antithrombin Basel. Antithrombin Basel is associated with decreased heparin cofactor activity, ie, type II heparin‐binding site defects. Antithrombin Dublin is complex. It has been associated with transient antithrombin deficiency.

The higher than expected prevalence (3.1%) of disease‐causing mutations in protein S, protein C, and antithrombin genes based on next‐generation sequencing in the Swedish population is in line with the finding of patients deficient in the 3 natural anticoagulants that are present in healthy adults. , , , , Familial thrombophilia is often the result of >1 segregating in a family. , , PROS1, PROC, or SERPINC1 variants translating into low protein levels identified in families with a high prevalence of thrombotic events may occur together with other thrombophilias, such as factor V Leiden and the prothrombin G20210A variant, which was also shown in the present study (Table 3). , , , However, although there was a high incidence rate, the HR for combined genetic variants was lower than previously estimated from studies of selected thrombosis‐prone families. It is possible that families with high inheritance of VTE have not only 1 or 2 risk variants, but even more variants might be present.

This study has clinical implications. The prevalence of disease‐causing variants is higher than expected and targeted sequencing in high‐risk individuals for variants in SERPINC1, PROC, and PROS1 even among middle‐aged and older patients with a VTE might be indicated. Identified variants would then need to be curated and plasma‐based studies of unknown non‐synonymous mutations might be worthwhile. The finding of enrichment of classical thrombophilia among cases with recurrent incident VTE suggests that classical thrombophilia is involved in recurrent VTE, and further studies could be warranted. The present study may not give a conclusive answer as to whether thromboprophylaxis is warranted in asymptomatic individuals, but the present study at least suggests that recommendations for younger individuals with classical thrombophilia should also be valid for middle‐aged and older individuals with classical thrombophilia.

Limitations of the study are that no family data exist, and no plasma samples are available for measurement of the individual proteins, ie, antithrombin, protein C, and protein S. We therefore do not know which of the de novo mutations are associated with a low protein level and/or anticoagulant activity of the respective protein. However, all identified variants were rare, and many had a positive prediction score for pathogenicity (Tables S2 through S4). Still, variants that are absent in the HGMD were not associated with thrombosis. A strength of the study is the identification of high‐risk variants using HGMD variants previously associated with deficiencies of the 3 anticoagulant proteins. Another strength is that the prevalence of the factor V Leiden and the prothrombin G20210A thrombophilic variants is similar to the SweGen project (https://swefreq.nbis.se/) and previous studies from Malmö. , , A further strength is also the high coverage of VTE diagnosis in the Malmö registers and that VTE diagnosis is confirmed by objective methods in Malmö and Sweden, and the high quality of Swedish registers. , , The present study also confirmed the factor V paradox, ie, rs6025 is associated with a higher risk of DVT than PE, which further strengthens the validity of our data. A minor limitation is that Swedes who had a fatal thrombosis prior to the recruitment (which occurred at a mean age of 58 years) would not have been included in the study cohort. This phenomenon, called depletion of the susceptible, tends to attenuate the HR toward the null. Another limitation is that we were unable to study provoked and unprovoked cases separately. Higher relative genotype risks are likely to be found for unprovoked cases. A limitation is also the classification of novel missense changes at amino acid residues where different missense changes determined to be pathogenic have been seen before in the HGMD database. However, using ACMG criteria gave weaker association with VTE than our definition of high and low risk variants.

Limited information about ethnicity exists. However, only those foreigners or immigrants who could speak Swedish adequately were included in the study. In total, 88% (not shown) of the total MDC cohort were born in Sweden (around 99% of those born outside Sweden were from European countries). Still, we included the top 2 eigenvectors from the PCA analysis as covariates in Cox proportional hazard regression models to control for population stratification. We therefore believe our results are of general interest and valid for many individuals of White origin in Europe and the United States. , The Swedish population is closely related to German and British people. , , Moreover, Malmö, and Skåne were an integral part of Denmark until AD 1658. Malmö and several cities in Skåne were important towns in the Hanseatic league. Analysis of genetic differentiation indicating that the population of Sweden’s southernmost counties are genetically closer to samples of Northern European ancestry than to the populations of Sweden’s northernmost counties. , Moreover, there is a high degree of correlation in minor allele frequency for variants in genes involved in VTE between gnomAD (non‐Finnish Europeans) and SweGen giving further support for the generalizability of our results. ,

In conclusion, the present study shows that classic thrombophilias are important even in elderly and middle‐aged people. The relative genetic risk of classic thrombophilia was dose‐graded. Moreover, PROC, PROS1, SERPINC1 disease causing variants were more common than the rs1799963 variant, but the genetic risk conferred by them was lower than previously anticipated and comparable with the rs6025 and rs1799963 variants. A reservation for certain rare type I deficiency variants causing a severe phenotype should be made. However, these are rare variants, and they were not possible to risk‐estimate in the present population‐based cohort study.

Appendix

Regeneron Genetics Center

RGC Management and Leadership Team

Goncalo Abecasis, PhD, Aris Baras, MD, Michael Cantor, MD, Giovanni Coppola, MD, Aris Economides, PhD, Luca A. Lotta, MD, PhD, John D. Overton, PhD, Jeffrey G. Reid, PhD, Alan Shuldiner, MD

Contribution: All authors contributed to securing funding, study design and oversight. All authors reviewed the final version of the manuscript.

Sequencing and Laboratory Operations

Christina Beechert, Caitlin Forsythe, M.S., Erin D. Fuller, Zhenhua Gu, M.S., Michael Lattari, Alexander Lopez, M.S., John D. Overton, PhD, Thomas D. Schleicher, M.S., Maria Sotiropoulos Padilla, M.S., Louis Widom, Sarah E. Wolf, M.S., Manasi Pradhan, M.S., Kia Manoochehri, Ricardo H. Ulloa.

Contribution: C.B., C.F., A.L., and J.D.O. performed and are responsible for sample genotyping. C.B, C.F., E.D.F., M.L., M.S.P., L.W., S.E.W., A.L., and J.D.O. performed and are responsible for exome sequencing. T.D.S., Z.G., A.L., and J.D.O. conceived and are responsible for laboratory automation. M.P., K.M., R.U., and J.D.O are responsible for sample tracking and the library information management system.

Genome Informatics

Xiaodong Bai, PhD, Suganthi Balasubramanian, PhD, Andrew Blumenfeld, Boris Boutkov, PhD, Gisu Eom, Lukas Habegger, PhD, Alicia Hawes, B.S., Shareef Khalid, Olga Krasheninina, M.S., Rouel Lanche, Adam J. Mansfield, B.A., Evan K. Maxwell, PhD, Mrunali Nafde, Sean O’Keeffe, M.S., Max Orelus, Razvan Panea, PhD, Tommy Polanco, B.A., Ayesha Rasool, M.S., Jeffrey G. Reid, PhD, William Salerno, PhD, Jeffrey C. Staples, PhD

Contribution: X.B., A.H., O.K., A.M., S.O., R.P., T.P., A.R., W.S. and J.G.R. performed and are responsible for the compute logistics, analysis, and infrastructure needed to produce exome and genotype data. G.E., M.O., M.N. and J.G.R. provided compute infrastructure development and operational support. S.B., S.K., and J.G.R. provide variant and gene annotations and their functional interpretation of variants. E.M., J.S., R.L., B.B., A.B., L.H., J.G.R. conceived and are responsible for creating, developing, and deploying analysis platforms and computational methods for analyzing genomic data.

Research Program Management

Marcus B. Jones, PhD, Jason Mighty, PhD, Lyndon J. Mitnaul, PhD

Contribution: All authors contributed to the management and coordination of all research activities, planning, and execution. All authors contributed to the review process for the final version of the manuscript.

Sources of Funding

This work was supported by a grant awarded to Dr Bengt Zöller by ALF–funding (the Medical Training and Research Agreement—Avtal om Läkarutbildning och Forskning) from Region Skåne, Sparbanken Skåne, and by the Swedish Research Council.

Disclosures

None.

Tables S1–S15

Figures S1–S2

Click here for additional data file.