Low borderline plasma levels of antithrombin, protein C and protein S are risk factors for venous thromboembolism

Authors

  • P. BUCCIARELLI,

    1. A. Bianchi Bonomi Hemophilia and Thrombosis Center, Department of Medicine and Medical Specialties, Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico and University of Milan, Milan;
    Search for more papers by this author
  • S. M. PASSAMONTI,

    1. A. Bianchi Bonomi Hemophilia and Thrombosis Center, Department of Medicine and Medical Specialties, Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico and University of Milan, Milan;
    Search for more papers by this author
  • E. BIGUZZI,

    1. A. Bianchi Bonomi Hemophilia and Thrombosis Center, Department of Medicine and Medical Specialties, Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico and University of Milan, Milan;
    Search for more papers by this author
  • F. GIANNIELLO,

    1. A. Bianchi Bonomi Hemophilia and Thrombosis Center, Department of Medicine and Medical Specialties, Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico and University of Milan, Milan;
    Search for more papers by this author
  • F. FRANCHI,

    1. A. Bianchi Bonomi Hemophilia and Thrombosis Center, Department of Medicine and Medical Specialties, Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico and University of Milan, Milan;
    Search for more papers by this author
  • P. M. MANNUCCI,

    1. Scientific Direction, Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policlinico, Milan, Italy
    Search for more papers by this author
  • I. MARTINELLI

    1. A. Bianchi Bonomi Hemophilia and Thrombosis Center, Department of Medicine and Medical Specialties, Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico and University of Milan, Milan;
    Search for more papers by this author

Paolo Bucciarelli, A. Bianchi Bonomi Hemophilia and Thrombosis Center, Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policlinico, Via Pace, 9 - 20122 Milan, Italy.
Tel.: +39 02 55035274; fax: +39 02 50320723.
E-mail: bucciarelli@policlinico.mi.it

Abstract

Summary.  Background:  Inherited deficiencies of antithrombin (AT), protein C (PC) and protein S (PS) are risk factors for venous thromboembolism (VTE). They are usually defined by laboratory cut-offs (in our setting 81, 70 and 63 IU dL−1, respectively), which give only a rough idea of the VTE risk associated with plasma levels of these proteins.

Objectives:  We investigated whether the risk of VTE associated with the plasma deficiencies of AT, PC or PS has a dose–response effect, and whether low borderline levels of these proteins are associated with an increased risk of VTE, both in the whole study population and separately in carriers of either factor V Leiden or G20210A prothrombin gene mutation.

Patients/Methods:  A case–control study of 1401 patients with a first objectively-documented VTE and 1847 healthy controls has been carried out.

Results:  A dose–response effect on the VTE risk was observed for all the three anticoagulant proteins. Compared with individuals with AT, PC or PS levels > 100 IU/dL, the adjusted odds ratio (95%CI) of VTE was 2.00 (1.44–2.78) for AT levels between 76 and 85 IUdL−1, 2.21 (1.54–3.18) and 1.84 (1.31–2.59) for PC and PS levels between 61 and 75 IUdL−1. The risk of unprovoked VTE in factor V Leiden or prothrombin G20210A carriers appears 2 to 3-fold increased when levels of AT or PS are low borderline.

Conclusions:  Low borderline plasma levels of AT, PC and PS are associated with a 2-fold increased risk of VTE and should be considered in the assessment of the individual VTE risk.

Introduction

Venous thromboembolism (VTE) is a multifactorial disease with an incidence in the general population of one to two cases per 1000 person-years [1], which develops with no apparent cause in about 40% of cases [2]. Antithrombin (AT), protein C (PC) and protein S (PS) are naturally occurring anticoagulant proteins that play an important role in the control of thrombus formation and propagation [3,4]. Inherited deficiencies of these proteins are rare in the general population (< 1% altogether), and represent a strong risk factor for VTE, with relative risks ranging from 5 to 50 according to the type of the coagulation defect, the design of the study and patient selection [5–9]. The diagnosis of AT, PC or PS deficiency is based on laboratory cut-offs, which represent an unrealistic way to describe the risk of VTE associated with plasma levels of these inhibitors. This is particularly true for PS deficiency, whose diagnosis is notoriously difficult [10]. Depending on the type of test used (functional, total or free antigenic PS) and study populations (pedigrees with familial thrombophilia or population-based studies), PS deficiency has been claimed to be a strong, mild or even absent risk factor for VTE [11]. As for all continuous variables, dichotomization is inappropriate to estimate a relationship between continuous predictors and an outcome variable [12,13].

The 1691 guanine to adenine substitution in coagulation factor (F) V gene (factor V Leiden [FVL]) and the 20210 guanine to adenine substitution in the 3′-untranslated region of the prothrombin gene (FII G20210A) are the two most common genetic risk factors for VTE [14,15]. These gain-of-function mutations that are common in the general population (approximately 3.0% and 4.0% in the Italian population), are associated with a smaller risk of VTE than inherited deficiencies of the naturally occurring anticoagulants [8,9]. Carriers of FVL or FII G20210A do not have the same degree of VTE risk throughout life. Some of them suffer from VTE early in life; others may develop VTE at an older age or remain free from thrombosis [16]. In symptomatic individuals, the co-segregation of either mutation with AT, PC or PS deficiency, the frequency of exposure to transient risk factors for VTE (such as oral contraceptive use, surgery, immobilization and pregnancy/puerperium), and the co-existence of other unknown thrombophilic abnormalities are thought to modulate the thrombotic risk, thus explaining at least in part the aforementioned different clinical phenotypes [17–22].

We hypothesized that varying plasma levels of the naturally occurring anticoagulant proteins may explain at least in part the variable thrombotic risk in carriers of FVL or FII G20210A. This case–control study was carried out to assess the risk of VTE associated with varying plasma levels of AT, PC and PS (dose–response effect), and to establish whether or not plasma levels of these anticoagulants around the cut-off (low borderline levels) are associated with an increased risk of VTE, both in the whole study population and separately in carriers of FVL or FII G20210A.

Methods

Study sample

One thousand five hundred and sixty-eight patients consecutively referred to our Thrombosis Center between January 1999 and July 2009 for a thrombophilia work-up after a first episode of VTE were included in the study. Their demographic data, medical history and exposure to risk factors for VTE were collected. One hundred and sixty-seven patients were excluded because VTE was not objectively confirmed (= 124) or was provoked by cancer (n = 34), a strong and independent risk factor for VTE, or because they had liver disease (n = 8) or nephrotic syndrome (n = 1), conditions that may affect plasma levels of AT, PC or PS. VTE included lower-limb deep vein thrombosis (diagnosed by compression ultrasound or venography) and/or pulmonary embolism (V/Q lung scan, CT scan or pulmonary angiography). Surgery, trauma, prolonged bed rest (> 1 week), pregnancy/puerperium and combined oral contraceptive use were considered transient risk factors for VTE. Events that had occurred in the absence of the aforementioned risk factors were considered unprovoked. Hence, 1401 patients with a first, objectively-documented episode of VTE were included in the analysis. Healthy individuals, partners or friends of the patients who agreed to be tested for thrombophilia during the same time-frame as the patients, formed the control group (n = 1847). Previous episodes of thrombosis were excluded by means of a validated questionnaire [23]. The study was approved by the Hospital Institutional Review Board and all patients and controls gave written informed consent to participate to the study.

Laboratory tests

In all patients and controls blood was collected into vacuum tubes containing 3.2% sodium citrate (Vacuette Premium tubes; Greiner Bio-One, Kremsmünster, Austria), and centrifuged within 15 min at 20 °C for 20 min at 2880 × g. The plasma obtained was aliquoted and snap-frozen in liquid nitrogen, and then stored at −80 °C until analysis. Patients with surgery- or pregnancy-related VTE were tested for AT, PC and PS at least 3 months after operation or delivery, in order to avoid changes in plasma levels of the naturally occurring anticoagulants related to these conditions [11,24].

Functional plasma levels of AT were measured with a chromogenic assay based on Xa inactivation (Antithrombin; Instrumentation Laboratories, Bedford, MA, USA), following the manufacturer’s instructions. In the case of concomitant therapy with heparin, the test was repeated at least 1 week after heparin discontinuation, and this value was considered for the purpose of this study. Functional PC plasma levels were measured with an automated APTT-based clotting assay (ProClot, Instrumentation Laboratories), following the manufacturer’s instructions. Because FVL, lupus anticoagulants and high levels of coagulation FVIII (above 200 IU dL−1) can interfere with APTT-based clotting assays for PC measurement [24,25], when such conditions were diagnosed functional PC was measured with either a chromogenic method (Chromogenic Protein C, Instrumentation Laboratories; n = 206) or a clotting assay after absorption with BaCl2 and subsequent elution (n = 48). When low functional levels of either AT or PC were found, the corresponding levels of immunoreactive proteins were measured by radial immunodiffusion for AT (NOR-Partigen Immunodiffusionplatte Antithrombin, Dade Behring, Marburg, Germany) or home made ELISA assay using specific polyclonal antibodies for PC (Dako A370 and P374, Glostrup, Denmark), in order to distinguish the more common type I (quantitative) from type II (qualitative) deficiency [26].

PS was measured with functional assays based on the PS-mediated prolongation of the prothrombin time. A functional method using bovine brain thromboplastin and endogenous Protac-mediated activated PC (Instrumentation Laboratory, Milan, Italy) [27] was employed from 1999 to 2005, whereas from 2006 to 2009 another functional method using recombinant rabbit brain thromboplastin and exogenous activated PC (Hemosil ProS; Instrumentation Laboratory, Orangeburg, NY, USA) was adopted [28]. Because the presence of FVL or antiphospholipid antibodies often may interfere with the measurement of PS functional levels [24,29], in the presence of such conditions free PS antigen instead of functional levels was measured, as recently recommended [11,24]. Total PS antigen was first measured with two polyclonal antibodies (Dako A384 and P419) [30], then free PS antigen levels were measured after precipitation of PS-C4b binding protein complexes with polyethyleneglycol 6000 (3.5% final concentration) using the same couple of polyclonal antibodies [30]. Free PS antigen levels were also measured in the case of low or low borderline functional levels, allowing us to distinguish between type I or III (quantitative) and the rare type II (qualitative) PS deficiency [26]. In case of low or low borderline levels of AT, PC or PS, individuals were retested to confirm the result and the mean value between the two measurements was chosen for the purpose of this study.

The lower limit of the laboratory reference of the naturally anticoagulant proteins (cut-off point) was set for this study at the 2.5th percentile of the distribution of levels in the control group (i.e. 81 IU dL−1 for AT, 70 IU dL−1 for PC and 63 IU dL−1 for both functional and free antigenic PS). These cut-off points are used to diagnose potential congenital deficiencies that are usually in the heterozygous state, as the very rare and more clinically severe homozygous deficiencies are diagnosed by levels lower than 5–10 IU dL−1. Because PC and PS are vitamin K-dependent proteins and their plasma levels decrease by the concomitant use of vitamin K antagonists, PC and PS were measured at least 1 month after withdrawal of these drugs. In 203 patients who required lifelong therapy with vitamin K antagonists, PC and PS were not measured. Patients and controls tested for AT, PC and PS are shown in Fig. 1.

Figure 1.

 Study flow-chart. AT, antithrombin; PC, protein C; PS, protein S; VKA, vitamin K antagonists.

DNA analysis for FVL and FII G20210A was carried out as previously described [14,15]. Antiphospholipid antibodies (lupus anticoagulant and anticardiolipin antibodies) were searched for in patients only according to previously described methods [31]. Total plasma homocysteine levels after overnight fasting and 4 h after an oral methionine load (3.8 g m−2 body surface area) were measured as previously described [32].

Statistical analysis

Continuous variables were expressed as median with minimum and maximum values, and categorical variables as counts. To assess the risk of VTE associated with varying plasma levels of AT, PC and PS, a univariable logistic regression model was first fitted taking these variables as continuous and using restricted cubic spline functions (with 3 knots) to search for possible non-linear relationships in the log odds. In this case, the odds ratio and its 95% confidence intervals (95%CI) were calculated as a measure of the relative risk of VTE for every 20 IU dL−1 decrease of each naturally occurring anticoagulant protein. Then these variables were divided into six categories (> 100, 86–100, 76–85, 61–75, 45–60 and < 45 IU dL−1), chosen with the goal of allowing the estimation of the risk of VTE also in categories containing the cut-off point of the normal laboratory range (low borderline values). The odds ratio and its 95%CI were calculated as a measure of the relative risk of VTE for individuals in each category of AT, PC or PS, taking the highest category as reference. Because qualitative (type II) AT and PC deficiencies carry a lower risk of VTE than quantitative (type I) deficiencies under the same functional plasma levels [24], patients with type II deficiency were excluded from the analysis. The same was done for the very rare type II PS deficiency, associated with an uncertain risk of VTE [10]. Because women with a VTE episode that had occurred during oral contraceptive intake were no longer taking the pill at the time of blood sampling, and owing to the influence of oral contraceptives on plasma levels of AT, PC and PS [24], the 182 control women tested on oral contraceptives were excluded from the analysis of VTE risk. Their AT, PC and PS median plasma levels were significantly different to those in control women not using oral contraceptives [for AT: 95 IU dL−1 (71–123) vs. 99 IU dL−1 (50–130), Mann–Whitney U-test, P < 0.001; for PC: 100 IU dL−1 (61–165) vs. 97 IU dL−1 (60–199), P = 0.034; for PS: 89 IU dL−1 (51–159) vs. 94 IU dL−1 (20–175), P < 0.001].

Unconditional multivariable logistic regression analysis was performed to control for such possible confounders as sex (0 = female, 1 = male), age (continuous variable), body mass index (continuous variable), and presence of thrombophilic defects other than that under examination (0 = no, 1 = yes). The analysis was also carried out separately for unprovoked VTE and VTE secondary to transient risk factors, and in the subgroup of carriers of FVL or FII G20210A, the two most common gain-of-function mutations of the coagulation pathway. Due to the relatively small size of the latter subgroup, fewer categories of AT, PC and PS levels were evaluated than in the total study sample (for AT: > 100, 86–100, 76–85 and ≤ 75 IU dL−1; for PC and PS: > 100, 76–100, 61–75 and ≤ 60 IU dL−1). The odds ratio obtained in the analysis of this subsample represents the relative risk of VTE in the lowest categories of AT, PC and PS levels compared with the reference category (i.e. AT, PC and PS levels > 100 IU dL−1). This relative risk must be multiplied by the corresponding risk of VTE for individuals with FVL or FII G20210A relative to those without mutations, and represents the additional VTE risk that in these individuals is associated with relatively low levels of AT, PC or PS. P ≤ 0.05 was chosen as the cut-off level for statistical significance. All analyses were performed with the statistical software spss (release 17.0; SPSS Inc, Chicago, IL, USA), except the analysis of VTE risk using spline functions, for which the statistical software r was used (release 2.9.1; R Project for Statistical Computing, Vienna, Austria).

Results

Demographic and clinical characteristics of the study population are shown in Table 1. Patients and controls were similar for sex distribution and age at the first visit. The median time (min-max) between VTE and blood sampling was 10 months (1–130 months). Body mass index was higher and thrombophilic abnormalities were more frequent in patients than in controls. VTE was unprovoked in 642 patients (46%). A family history of venous thrombosis (defined as at least one first or second-degree family member with venous thrombosis) was positive in 238 of the 642 patients with unprovoked VTE (37%) and in 220 of the 759 patients with VTE secondary to transient risk factors (29%; chi-squared test, P = 0.002). A diagnosis of inherited AT, PC or PS deficiency, confirmed by finding at least one relative with the same deficiency, was made in 6% of patients and 1% of controls. One patient and six controls with type II AT deficiency, five patients with type II PC deficiency and one patient and one control with PS type II deficiency were excluded from the analysis of VTE risk (Statistical analysis).

Table 1.   Demographic and clinical characteristics of the study population
 Patients (n = 1401)Controls (n = 1847)
  1. Continuous variables are expressed as median with minimum and maximum values between brackets. NA, not applicable; VTE, venous thromboembolism; DVT, deep vein thrombosis; PE, pulmonary embolism; AT, antithrombin; PC, protein C; PS, protein S. *Some patients had more than one risk factor. Percentage calculated in women of reproductive age (527 patients) after exclusion of those who were taking oral contraceptives. Percentage calculated in women of reproductive age (527 patients and 736 controls) after exclusion of pregnant women. §Some individuals carried more than one defect. Twenty-eight patients and three controls were carriers of both factor V Leiden and prothrombin G20210A.

Sex (M/F)655/746736/1111
Age at visit (years)44 (7–89)43 (11–84)
Age at thrombosis (years)42 (7–88)NA
Body mass index (Kg m−2)25.2 (13.3–53.3)23.5 (15.6–45.2)
Types of VTE event, n (%)
 DVT887 (63)NA
 DVT + PE262 (19)
 PE252 (18)
Transient risk factors for VTE, n (%)*
 None642 (46)
 Surgery184 (13)
 Trauma/bed rest281 (20)
 Pregnancy/puerperium74 (38)
 Oral contraceptive use330 (73)182 (25)
Thrombophilia, n (%)§
 None748 (53)1592 (86)
 AT, PC or PS inherited deficiency86 (6)19 (1)
 Factor V Leiden203 (15)55 (3)
 Prothrombin G20210A170 (12)64 (3.5)
 Antiphospholipid antibodies73 (5)
 Hyperhomocysteinemia227 (16)119 (6)

For all the three naturally occurring anticoagulant proteins, at plasma levels below 100 IU dL−1 the risk of VTE (expressed as log odds) linearly increased with the decrease of their plasma levels. The linear rise on a log odds scale corresponds to an exponential rise on a non-logarithmic scale. For every 20 IU dL−1 decrease of plasma levels, the VTE risk was 40% increased for AT (odds ratio, 1.43; 95%CI, 1.23–1.68) and 13% increased for both PC (1.13; 1.04–1.22) and PS (1.13; 1.09–1.17). Table 2 shows the odds ratios of VTE for different categories of plasma levels of the naturally occurring anticoagulant proteins in the whole study population. The risk of VTE in the lowest category (< 45 IU dL−1) could be estimated only for PS, because no control had AT or PC plasma levels below 45 IU dL−1. For all the naturally occurring anticoagulant proteins, the pattern of increasing risk according to the decrease of their plasma levels was confirmed, and was already statistically significant in the categories of levels around the cut-off point (i.e. 76–85 IU dL−1 for AT, and 61–75 IU dL−1 for PC and PS), and also after adjustment for potential confounders. For AT, even plasma levels considered normal (86–100 IU dL−1) increased significantly the risk of VTE. The VTE risk estimates were higher for unprovoked VTE than for VTE secondary to transient risk factors (Table 3), and showed a pattern of increased risk similar to that observed in the whole study group, being already statistically significant for low borderline levels of the anticoagulants. The risk was lower for PS than for AT and PC deficiencies. Individuals in the category of low plasma levels between 45 and 60 IU dL−1 had a 23-fold (for AT), 13-fold (for PC) and 7-fold (for PS) increased risk of unprovoked VTE, compared with individuals with levels > 100 IU dL−1.

Table 2.   Risk of VTE associated with different categories of plasma levels of AT, PC or PS
 No. (%)OR (95% CI)ORadj (95% CI)*
PatientsControls
  1. One patient and six controls with antithrombin type II deficiency, five patients with protein C type II deficiency, one patient and one control with protein S type II deficiency and 182 control women taking oral contraceptives were excluded. *Odds ratio adjusted for sex, age, body mass index and thrombophilia (all defects except that under study).

Antithrombin (IU dL−1)
 > 100569 (41)807 (49)1 (reference)1 (reference)
 86–100648 (47)739 (45)1.24 (1.07–1.45)1.31 (1.11–1.54)
 76–85119 (9)87 (5)1.94 (1.44–2.61)2.00 (1.44–2.78)
 61–7522 (2)7 (0.4)4.46 (1.89–10.51)4.77 (1.93–11.83)
 45–609 (0.7)2 (0.1)6.38 (1.37–29.65)7.74 (1.59–37.73)
 < 454 (0.3)0 (–)
Protein C (IU dL−1)
 > 100537 (51)929 (56)1 (reference)1 (reference)
 86–100281 (27)444 (27)1.10 (0.91–1.32)1.13 (0.92–1.38)
 76–85129 (12)197 (12)1.13 (0.89–1.45)1.21 (0.93–1.59)
 61–7580 (8)75 (5)1.85 (1.32–2.57)2.21 (1.54–3.18)
 45–6010 (1)3 (0.2)5.77 (1.58–21.05)5.35 (1.35–21.15)
 < 456 (0.6)0 (–)
Protein S (IU dL−1)
 > 100586 (56)985 (60)1 (reference)1 (reference)
 86–100245 (23)400 (24)1.03 (0.85–1.24)1.17 (0.95–1.45)
 76–8597 (9)150 (9)1.09 (0.83–1.43)1.22 (0.90–1.67)
 61–7594 (9)101 (6)1.56 (1.16–2.11)1.84 (1.31–2.59)
 45–6012 (1)4 (0.2)5.04 (1.62–15.71)6.14 (1.82–20.70)
 < 4516 (1.5)4 (0.2)6.72 (2.24–20.21)7.43 (2.36–23.38)
Table 3.   Risk of VTE associated with different categories of plasma levels of AT, PC or PS after stratification of VTE episodes as unprovoked and secondary to transient risk factors
 Unprovoked VTESecondary VTE
No. (%)ORadj (95% CI)*No. (%)ORadj (95% CI) *
PatientsControlsPatientsControls
  1. Patients and controls with antithrombin, protein C or protein S type II deficiency and control women taking oral contraceptives were excluded. *Odds ratio adjusted for sex, age, body mass index and thrombophilia (all defects except that under study).

Antithrombin (IU dL−1)
 > 100263 (42)807 (49)1 (reference)306 (41)807 (49)1 (reference)
 86–100275 (44)739 (45)1.10 (0.88–1.38)373 (50)739 (45)1.45 (1.18-1.77)
 76–8567 (11)87 (5)1.93 (1.29–2.88)52 (7)87 (5)1.78 (1.17-2.72)
 61–7514 (2)7 (0.4)5.88 (2.14–16.14)8 (1)7 (0.4)2.67 (0.74-9.61)
 45–607 (1)2 (0.1)23.10 (4.39–122)2 (0.3)2 (0.1)3.12 (0.42-23.13)
 < 454 (0.6)0 (–)0 (–)0 (–)
Protein C (IU dL−1)
 > 100253 (58)929 (56)1 (reference)284 (47)929 (56)1 (reference)
 86–10097 (22)444 (27)0.85 (0.63–1.13)184 (31)444 (27)1.25 (0.98-1.59)
 76–8548 (11)197 (12)1.14 (0.78–1.67)81 (13)197 (12)1.17 (0.85-1.62)
 61–7531 (7)75 (5)2.30 (1.38–3.89)49 (8)75 (5)2.02 (1.32-3.11)
 45–608 (2)3 (0.2)13.47 (3.18–57.06)2 (0.3)3 (0.2)1.86 (0.29-12.02)
 < 453 (1)0 (–)3 (0.5)0 (–)
Protein S (IU dL−1)
 > 100280 (63)985 (60)1 (reference)306 (50)985 (60)1 (reference)
 86–10089 (20)400 (24)1.14 (0.85–1.54)156 (26)400 (24)1.14 (0.88-1.48)
 76–8533 (7)150 (9)1.39 (0.88–2.19)64 (11)150 (9)1.18 (0.82-1.69)
 61–7527 (6)101 (6)1.70 (1.03–2.83)67 (11)101 (6)1.85 (1.25-2.72)
 45–603 (1)4 (0.2)6.83 (1.30–35.99)9 (1.5)4 (0.2)5.25 (1.39-19.86)
 < 4510 (2)4 (0.2)15.49 (4.43–54.09)6 (1)4 (0.2)4.57 (1.18-17.74)

Three hundred and forty-five VTE patients were carriers of FVL (193 heterozygotes and 10 homozygotes) or FII G20210A (159 heterozygotes and 11 homozygotes), and among them 28 carried both mutations. Among controls, 116 were carriers of FVL (55 heterozygotes) or FII G20210A (63 heterozygotes and one homozygote), three of whom carried both mutations. Table 4 shows the risk of VTE (all episodes and unprovoked VTE only) according to different categories of AT, PC and PS plasma levels in carriers of FVL or FII G20210A. Taking as reference individuals with plasma levels of anticoagulant proteins > 100 IU dL−1, the risk of all VTE was 2-fold higher in individuals with AT levels ranging between 76 and 85 IU dL−1, 1.1-fold higher in individuals with PC levels between 61 and 75 IU dL−1 and 1.5-fold higher in individuals with PS levels between 61 and 75 IU dL−1. Limiting the analysis to the 142 patients with unprovoked VTE, an increased risk of VTE for low borderline levels of the anticoagulants was found only for AT and PS. Carriers of FVL or FII G20210A with AT levels ranging between 76 and 85 IU dL−1 had an approximately 3-fold increased risk of unprovoked VTE compared with those with AT > 100 IU dL−1 (adjusted odds ratio, 2.93; 95%CI, 0.93–9.28). A 2.5-fold increased risk of unprovoked VTE was also found for carriers of FVL or FII G20210A with PS levels ranging between 61 and 75 IU dL−1 compared with those with PS > 100 IU dL−1 (adjusted odds ratio, 2.55; 95%CI, 0.74–8.85). This risk should be multiplied for the baseline risk of VTE associated with the carriership of either FVL or FII G20210A. For instance, the adjusted odds ratio for unprovoked VTE in carriers of FVL or FII G20210A and AT plasma levels > 100 IU dL−1 (vs. individuals with no mutation and AT > 100 IU dL−1) was 3.77 (95%CI, 2.46–5.77). Hence, the risk is 11-fold higher (3.77 × 2.93) if carriers of FVL or FII G20210A also have low borderline AT plasma levels (76–85 IU dL−1).

Table 4.   Risk of VTE according to different categories of plasma levels of AT, PC and PS in carriers of factor V Leiden or FII G20210A
Factor V Leiden or FII G20210A andAll VTEUnprovoked VTE
No. (%)ORadj (95% CI)*No. (%)ORadj (95% CI) *
PatientsControlsPatientsControls
  1. One patient with antithrombin, one with protein C type II deficiency and 12 control women taking oral contraceptives were excluded. *Odds ratio adjusted for sex, age, body mass index and thrombophilia (all defects except those under study). In 52 patients protein C and protein S were not measured because they were on lifelong VKA therapy.

Antithrombin (IU dL−1) N = 336 N = 103  N = 139 N = 103 
 > 100153 (45)60 (59)1 (reference)62 (45)60 (59)1 (reference)
 86–100141 (42)38 (37)1.40 (0.86–2.28)56 (40)38 (37)1.38 (0.75-2.55)
 76–8535 (10)5 (5)2.39 (0.86–6.63)18 (13)5 (5)2.93 (0.93-9.28)
 ≤ 757 (2)0 (–)3 (2)0 (–)
Protein C (IU dL−1) N = 245 N = 103  N = 89 N = 103 
 > 100124 (51)55 (53)1 (reference)49 (55)55 (53)1 (reference)
 76–100105 (43)43 (42)0.82 (0.49–1.39)36 (40)43 (42)0.92 (0.48-1.75)
 61–7513 (5)5 (5)1.14 (0.37–3.55)2 (2)5 (5)0.90 (0.14-5.55)
 ≤ 603 (1)0 (--)2 (2)0 (--)
Protein S (IU dL−1) N = 246 N = 102  N = 90 N = 102 
 > 100128 (52)58 (57)1 (reference)49 (54)58 (57)1 (reference)
 76–10088 (36)37 (36)0.99 (0.57–1.70)29 (32)37 (36)1.36 (0.66-2.80)
 61–7522 (9)6 (6)1.47 (0.53–4.09)8 (9)6 (6)2.55 (0.74-8.85)
 ≤ 608 (3)1 (1)2.34 (0.26–21.19)4 (4)1 (1)5.84 (0.54-62.98)

Discussion

This case–control study shows that the risk of VTE associated with plasma levels of the naturally occurring anticoagulant proteins AT, PC and PS is not limited to some predefined cut-off values that define patients with these deficiencies, but should be considered a continuum, being progressively higher with decreasing plasma levels of these proteins and also present for low borderline levels (i.e. levels around the cut-off point). Indeed, we found that the risk of VTE, which sharply increases at levels below 60 IU dL−1, starts to increase already at low borderline plasma levels, particularly for AT and PC. AT plasma levels between 86 and 100 IU dL−1 were associated with a small but statistically significant increased risk of VTE, which became 2-fold higher at levels between 76 and 85 IU dL−1 that usually include the laboratory cut-off used to discriminate individuals with and without AT deficiency. Similarly, PC levels between 61 and 75 IU dL−1 were associated with a statistically significant 2-fold increased risk of VTE. The management of patients with plasma levels of the naturally occurring anticoagulants around the lower limit of the normal range is often challenging, because the impact on the risk of VTE of having levels of such proteins in this grey area is not clear [5,26]. Our data suggest that these patients should be considered to be at a significantly higher risk of VTE, albeit lower than that of patients with levels as low as 60 IU dL−1 or lower, which are typical of heterozygous deficiencies of these naturally occurring anticoagulant proteins. The VTE risk estimates were much higher when only unprovoked VTE episodes were considered, perhaps because of the co-inheritance of some unknown genetic risk factors that may increase the risk of unprovoked VTE, as underlined also by the slightly higher prevalence of a positive family history for venous thrombosis in patients with unprovoked VTE than in those with VTE secondary to transient risk factors.

Before the present study, only the Leiden Thrombophilia Study [5] has evaluated in detail whether or not there is a gradient between plasma levels of the naturally occurring anticoagulants and the magnitude of the risk of VTE. In the Leiden Thrombophilia Study, a dose–response effect on the risk of VTE was demonstrated only for PC, with a 3 to 4-fold increased risk for levels in the lowest categories (i.e. 55–65 and < 55 IU dL−1), smaller than ours and statistically more uncertain [5]. For AT a 2.2-fold increased risk of VTE was observed after dichotomization of this variable at the lower cut-off of the normal laboratory range (i.e. 80 IU dL−1). For PS (measured as both total and free antigen), also dichotomized at the lower cut-off of laboratory normality (i.e. 67 IU dL−1 for total and 57 IU dL−1 for free antigen), no increased risk of VTE for total antigenic PS, and only a marginally not statistically significant increased risk for free antigenic PS, were found [5]. The smaller sample size of the Leiden study compared with this study is perhaps the main explanation for the statistical uncertainty of the VTE risk estimates associated with low levels of PC and AT. Pertaining to PS, the absence in the Leiden study of a clear association between low PS plasma levels and VTE risk was not confirmed in our own study, which showed a dose–response effect on the risk of VTE also for PS levels, albeit lower than that of AT and PC. The different methodological approaches used to measure PS plasma levels in these two studies (only total or free PS antigen in the Leiden study, a functional test followed by a free antigenic PS test in selected cases in the present study) may also in part explain the different VTE risk estimates. Another possible explanation may be the different selection of patients. The Leiden study was a population-based case–control study, whereas we selected patients referred to a tertiary care unit, thus leading to the possibility of an overestimation of the true risk of VTE. In a subsequent reanalysis of the Leiden Thrombophilia Study, in which sex and hormonal status were taken into account to set different cut-off values [33], low free antigenic PS was associated with a 2-fold increased risk of VTE in women but not in men. Finally, the heterogeneity of the VTE risk in relation to the different genotypes associated with PS deficiency may be considered as another possible explanation for the difference in the VTE risk estimates between studies, because genetic variants causing severe effects on secretion and function of the protein are associated with the highest risk of VTE [34]. The interpretation of results pertaining to PS in this study needs further details. The 7-fold higher risk of unprovoked VTE found for PS plasma levels between 45 and 60 IU dL−1 was lower than that found in the same level category for AT and PC (23-fold and 13-fold higher risk, respectively), for whom the cut-off point was set at a higher level (i.e. 81 IU dL−1 for AT and 70 IU dL−1 for PC) than for PS (63 IU dL−1 for both functional and free antigenic tests). The expected sharp increase in VTE risk found for PS levels < 45 IU dL−1 could not be calculated for AT and PC, because no control had such low plasma levels of these two anticoagulant proteins.

This study also shows that in carriers of FVL or FII G20210A gain-of-function gene mutations the risk of VTE is modulated by plasma levels of the naturally occurring anticoagulant proteins. Previous studies (mainly carried out in the frame of pedigrees of families with inherited deficiencies) reported an increased risk of VTE when a deficiency of AT, PC or PS was co-inherited with one of these gain-of-function mutations [6,7,17–19]. At variance with those studies (which defined AT, PC or PS deficiency on the basis of very low plasma levels), our study found an increased risk of VTE also in carriers of FVL or FII G20210A mutations with low borderline levels of anticoagulant proteins. This was particularly true for AT and PS, because a 2 to 3-fold increased risk of unprovoked VTE was observed for carriers of FVL or FII G20210A who also had low borderline AT (between 76 and 85 IU dL−1) or PS (between 61 and 75 IU dL−1) levels. This finding may in part explain the variable thrombotic risk associated with FVL or FII G20210A, besides the co-inheritance of other known or unknown thrombophilic mutations. However, owing to the reduced size of this subsample, the statistical uncertainty of the risk estimates hinders us from making firm conclusions.

The strength of this study is its large sample size, which allowed us to clearly demonstrate a dose–response effect of the association between plasma levels of the naturally occurring anticoagulants and the risk of VTE. The best way to analyze the effect of continuous predictors on the outcome is to take them as they are. Some studies indeed demonstrated that categorization of a continuous predictor always results in loss of power (i.e. a reduction of the probability of finding a positive effect of the predictor on the outcome when it is truly present) and in the reduction of its predictive capabilities [35–37]. This is particularly true when dichotomization is carried out, especially at extreme values of the distribution [38]. Dividing AT, PC and PS levels into several categories is the best way to approach the analysis of these predictors as continuous. The large sample size reduced the possibility of power reduction due to over-parametrization and high variance of the risk estimates within categories. The choice to measure PS first with a functional assay and then to use free antigenic PS tests either to confirm low or low borderline results or to directly measure PS in individuals with FVL or antiphospholipid antibodies is in line with the most recent recommendations [11,24] and should have limited the possibility of false positive results. This was also true for AT and PC, as all individuals with low or low borderline plasma levels of these proteins were retested to confirm the result.

Among limitations of this study there is the use of different functional or antigenic assays to measure PS over the 10-year time-frame of this study. This may have led to a misclassification of patients and controls in different categories of PS plasma levels, thus potentially influencing the VTE risk estimates. However, due to the design of the study, any effect due to the analytical variability for PS should be equally distributed among patients and controls, an important prerequisite to draw meaningful conclusions from a case–control study, therefore limiting the possible consequences of the misclassification for risk estimates. The possibility of misclassification can be reasonably ruled out for AT and PC, because the functional assays used for their measurement did not change over the entire course of the study period. Another limitation of this study is the selection of patients from a tertiary center where they were referred for thrombophilia screening, because they do not represent VTE patients from the general population, and hence an overestimation of the VTE risk may exist. This limitation may have affected the magnitude of the risk estimates, but is unlikely to have influenced the dose–response effect on the VTE risk associated with varying plasma levels of AT, PC and PS.

In conclusion, low borderline plasma levels of the naturally occurring anticoagulant proteins AT, PC and PS are associated with an increased risk of VTE. Because of the demonstrated dose–response effect on the risk of VTE, measurement of the plasma levels of these inhibitors should be considered for refining the individual risk of VTE. The duration and intensity of anticoagulant therapy should be prescribed according to individual plasma levels of these inhibitors rather than on the basis of a simple cut-off point.

Addendum

P. Bucciarelli and I. Martinelli were responsible for study design and coordination; P. Bucciarelli, I. Martinelli, S. M. Passamonti and F. Gianniello were involved in data collection of patients and controls; E. Biguzzi and F. Franchi were responsible for laboratory analysis; P. Bucciarelli was responsible for statistical analysis and wrote the initial draft of the manuscript; P. Bucciarelli, I. Martinelli, S. M. Passamonti, E. Biguzzi and P. M. Mannucci were responsible for revisions of the manuscript; all authors were responsible for approval of the final manuscript.

Disclosures of Conflict of Interests

The authors state that they have no conflict of interest.

Ancillary