Clarification of the Risk for Venous Thrombosis Associated with Hereditary Protein S Deficiency by Investigation of a Large Kindred with a Characterized Gene Defect

Methods

Participants

The manifestations of thrombosis in the 122-member family under investigation (most family members live in northern Sweden) were attributed to protein S deficiency in 1993 [23]. This kindred has also been part of a larger study that involved 18 families with phenotypic protein S deficiency [16, 24] and provided an explanation for phenotypic variation in familial protein S deficiency [14]. All participants gave informed consent, and the medical ethics committee at the University of Lund approved all of these studies, including the present one. Study participants answered a questionnaire about their medical history, with emphasis on manifestations of deep venous thrombosis, pulmonary embolism, superficial thrombophlebitis, and arterial thrombosis. Symptomatic family members were also interviewed by a physician or had their medical records reviewed.

The term deep venous thrombosis includes deep venous thrombosis of the leg and thrombosis in such unusual locations as the axillary, mesenteric, and cerebral veins. Thrombotic event refers to deep venous thrombosis, pulmonary embolism, or superficial thrombophlebitis diagnosed by a physician on the basis of physical examination.

Laboratory Methods

Blood sampling and routine coagulation were performed as described elsewhere [25]. Total and free protein S antigen levels were measured by doing radioimmunoassay [26]. Protein S levels were compared with laboratory reference ranges for levels of both free (reference range, 56 to 182 nmol/L) and total (reference range, 219 to 407 nmol/L) antigen. Persons receiving anticoagulation were compared with an anticoagulated control group. The control groups that we used have been described elsewhere [24]. Of the 122 family members, 44 had free protein S antigen levels below the lower limit of the reference range; 13 of the 44 were receiving oral anticoagulants at the time of sampling.

Molecular Genetic Investigation

The methods used to identify and detect the novel protein S gene mutation, Gly295→Val, in 122 genomic DNA samples have been reported elsewhere [14, 27]. All 44 family members with reduced free protein S antigen levels were heterozygous for the mutation; the remaining 78 relatives who had normal free protein S antigen levels were normal at this site. This finding confirmed that the Gly295→Val mutation was the cause of protein S deficiency in this family. A single asymptomatic family member with normal free protein S antigen levels had previously been found to be heterozygous for the factor V Arg506→Gln mutation.

Statistical Analysis

Thrombosis-free survival curves were constructed according to the method of Kaplan and Meier [28]. Two curves were compared by using the log-rank test, which results in a test statistic with chi-squared distribution and one degree of freedom [29]. This analysis was performed by using Statistica software (Statsoft, Inc., Tulsa, Oklahoma). Univariate and multivariate Cox regression analyses [30] were performed by using Statistica software or SPS (SPS, Inc., Chicago, Illinois). All 122 family members were included in the analysis for risk for thrombosis.

Results

Demographic and Clinical Data

Samples of plasma and genomic DNA were available for 122 germline family members (60 men and 62 women; mean age ±SD, 36 ± 17 years [range, 7 to 82 years]) spanning five generations. The distribution of patient samples was 1, 8, 41, 57, and 15 from the first, second, third, fourth, and fifth generations, respectively. A histogram of the current ages of the study participants is shown in Figure 1. Twenty-five (57%) of the 44 family members with the Gly295→Val mutation had one or more venous thrombotic events (mean age at first event, 31 years [range, 11 to 71 years]) compared with 5 (6%) of the 78 family members who lacked the mutation (mean age at first event, 29 years [range, 16 to 43 years]). The clinical manifestations in symptomatic family members are summarized in Table 1. First thrombotic events were associated with one or more circumstantial risk factors in 12 symptomatic relatives (48%) with the Gly295→Val mutation and 2 family members (40%) who lacked the mutation (Table 2). Three carriers of the mutation (6.8%) and none of the normal family members had arterial thrombotic events (postoperative bilateral arterial thrombosis requiring bilateral above-the-knee amputation, myocardial infarction, and embolization requiring below-the-knee amputation).

View larger version:

• In this window
• In a new window

Figure 1. The total number of persons in each age range is shown. Striped bars represent persons who carried the Gly295→Val mutation; white bars indicate persons who were normal at this site. Histogram of the current ages of the investigated persons in the study family.

According to Kaplan-Meier analysis of thrombosis-free survival, the probability that a family member who carries the Gly295→Val mutation would remain free of venous thrombosis at 30 years of age is 0.5 (95% CI, 0.33 to 0.66) compared with 0.97 (CI, 0.93 to 1.0) for normal family members (Figure 2). The difference between the two curves is highly significant (P < 0.001). In a univariate Cox analysis that considered both age and the presence or absence of the Gly295→Val mutation, the risk associated with the mutation was 10.2 (CI, 3.89 to 26.7; P < 0.001). This risk was essentially unchanged when two other potential risk factors, smoking and obesity (body mass index > 25 kg/m²), were considered (hazard ratio, 11.5 [CI, 4.33 to 30.6]; P < 0.001) (Table 3). Smoking was a significant risk factor in the multivariate model (hazard ratio, 3.07 [CI, 1.32 to 7.18]; P = 0.0095) (Table 3). The risk for arterial thrombosis associated with the Gly295→Val mutation was not formally examined.

View larger version:

• In this window
• In a new window

Figure 2. The lower panel indicates the number of family members who remain at risk at 10-year intervals. Data for 78 family members who lacked the Gly295→Val mutation and 44 heterozygous Gly295→Val carriers ( ) are shown. The difference between the curves is statistically significant ( < 0.001). The probability that a carrier of the mutation would remain free of venous thrombosis at 30 years was 0.5 (95% CI, 0.33 to 0.66) compared with 0.97 (CI, 0.93 to 1.0) for normal family members. The curve for affected family members reaches zero because no tested family member with the Gly295→Val mutation older than 71 years of age remained free of thrombosis. Kaplan-Meier analysis of thrombosis-free survival showing the probability of remaining free of thrombosis at a certain age.asteriskP

Univariate Cox regression analysis was also performed to investigate the risk associated with the presence of the Gly295→Val mutation for thrombotic events that were treated with oral anticoagulants (20 family members) or confirmed by use of an objective technique (phlebography, ultrasonography, lung scintigraphy, or surgery [16 family members]). Among the family members who did not carry the Gly295→Val mutation, 1 had a thrombotic event that was treated with oral anticoagulants and was confirmed objectively. In both cases (in which the diagnostic criteria were increasingly exacting), the risk for venous thrombosis did not decrease and remained statistically significant (results not shown).

Phenotypic Data

Levels of free and total protein S antigen in family members with and without the Gly295→Val mutation are shown in Figure 3. Only data from persons who did not receive anticoagulation are shown. Of the 31 family members with the mutation who did not receive anticoagulation, 8 had total protein S antigen levels within the normal reference range; several additional family members had borderline levels. Therefore, assessment of total protein S antigen levels had a low specificity for predicting the presence of the mutation that causes protein S deficiency. However, as Figure 3 clearly shows, the assay for free protein S antigen completely separated persons with and persons without the Gly295→Val mutation. In addition, no carriers of the mutation had borderline free protein S antigen levels; therefore, this assay predicted the presence of the mutation with certainty.

View larger version:

• In this window
• In a new window

Figure 3. Dashed lines represent the lower limits of the normal ranges. Column A represents family members who lack the Gly295→Val mutation; column B represents family members with the Gly295→Val mutation; and column C represents an unrelated, normal control group. It is clear that measurement of free rather than total protein S antigen is fully predictive of the presence of the mutation that causes protein S deficiency. Phenotypic data, including total and free protein S antigen levels, in family members who did not receive anticoagulation and in an unrelated control group.

Discussion

By using univariate and multivariate statistical analyses, we found that a missense mutation in the gene that encodes protein S (Gly295→Val) was a strong risk factor for thrombosis in a clinically affected family. Kaplan-Meier analysis of thrombosis free-survival (Figure 2) that considered only the Gly295→Val mutation showed the relation between onset of thrombosis and age in carriers of the mutation. In our current study, little thrombotic risk was found before the age of 15 years. Thus, age itself is a risk factor, a finding that is well established for venous thromboembolism [31]. By 30 years of age, the probability that family members who carried the Gly295→Val mutation would remain free of venous thrombosis was 0.5; this number is similar to previous estimates, although Engesser and colleagues [11] assessed protein S deficiency on the basis of total protein S antigen only.

In carriers of the mutation, the crude odds ratio for developing venous thrombosis was 19.2; this odds ratio decreased to 10.2 in the univariate Cox analysis. When two other potential risk factors were included in a proportional hazards model, the odds ratio remained essentially unchanged (11.5) (Table 3). These risk estimates were based on thrombotic events that were diagnosed by a clinician who did a physical examination. A potential difficulty with this analysis is that diagnosis may not have been reliable for thrombotic events that occurred many years ago. However, when only thrombotic events that were treated or confirmed by an objective diagnostic test were examined by using univariate Cox analysis, the relative risk did not decrease, and the mutation remained a highly significant risk factor for thrombosis. Under these more stringent conditions, both analyses excluded family members in whom the only manifestation of thrombosis was superficial thrombophlebitis. This thrombotic inflammatory condition is commonly associated with protein S deficiency [11]. Even so, these additional analyses strongly support the high risk estimate for the Gly295→Val substitution.

The relative risk of 11.5 associated with the Gly295→Val substitution is high compared with published estimates for genetically determined protein C deficiency (univariate relative risk in a case–control study, 6.5 [17]; multivariate relative risk in a family-based study, 8.8 [32]). Only one previous family-based study [16] examined the risk associated with phenotypic protein S deficiency by using multivariate analysis. Again, the risk for thrombosis (multivariate relative risk, 6.8) was found to be much lower than the one that we found in our study.

In large family studies, survival bias may influence outcome of risk analysis. No evidence suggests that the Gly295→Val substitution extends life span, but acute thrombotic events, such as pulmonary embolism, may increase mortality rates. However, formal studies investigating mortality rates in Dutch carriers of other genetic mutations involved in the regulation of coagulation (antithrombin and factor V) showed no effect on standardized mortality rates or cause of death [33-36].

A recent population-based case–control study [17] has contributed to the uncertainty surrounding the risk for venous thrombosis associated with protein S deficiency because this risk was found to be negligible. Because of the lack of information about the presence of causative gene mutations and the prevalence of protein S deficiency in the population examined (which is likely to be very low), this estimate must be treated with caution. However, that study indicated that alternate assays for protein S have different values in predicting thrombotic risk. When diagnosis was determined by low levels of free protein S antigen alone or low levels of free and total protein S antigen, the univariate relative risk (1.6) was higher than the risk determined by low levels of total protein S (0.7).

In our current study, phenotypic data indicated that assessment of total protein S antigen levels was not fully predictive of the presence or absence of the Gly295→Val mutation and, therefore, risk for thrombosis. Several members of the study family who carried the mutation had levels of total protein S antigen that were in the normal range, which would have resulted in misdiagnosis if free protein S antigen had not been measured or the presence of a causative mutation had not been determined. Thus, studies that measure total protein S antigen alone [11] may underestimate the risk for thrombosis associated with protein S deficiency. Conversely, we found that measurement of free protein S antigen levels was wholly predictive of the mutation and deficiency. Measurement of free protein S antigen levels must therefore be considered to have greater value than assessment of total protein S antigen levels in identifying persons at increased risk for venous thrombosis. However, the relative predictive value of assessing levels of free protein S compared with total levels of protein S has not been formally quantified.

In addition to the high relative risk associated with the Gly295→Val mutation, the incidence of thrombosis in normal family members was higher than expected (5 of 78) and the mean age at first thrombotic event was similar in family members with and without the mutation (29 years and 31 years, respectively). These findings may suggest the presence of an additional genetic risk factor that predisposes to thrombosis, although any possible influence of resistance to activated protein C (a common risk factor for thrombosis in Europeans [37]) was excluded by molecular investigation. Unknown genetic risk factors may contribute to the thrombotic complications in the study family. Such potential risk factors may have only a mild effect themselves but may have more severe clinical implications when found in combination with a mutation that causes protein S deficiency. However, a similar proportion of first thrombotic events was associated with circumstantial risk factors in family members without and with the mutation (40% and 48%, respectively); this finding supports the idea that the Gly295→Val mutation is the predominant cause of venous thrombosis in this family. Furthermore, any potential contributing genetic risk factor would have to be situated at a locus closely linked to the protein S gene mutation to explain the high relative risk associated with the Gly295→Val mutation throughout five generations of such a large family.

When considered as crude odds ratios, both smoking and obesity were found to be acquired risk factors for venous thrombosis (relative risk, approximately 2 to 3) (Table 3). In the proportional risk model, the risk associated with smoking remained statistically significant. This result was surprising because smoking is thought to be a risk factor for arterial rather than venous thrombosis. In our study, all three carriers of the mutation who had arterial thrombosis were smokers. In this context, it is interesting that smoking is reported to reduce the plasma concentration of protein S [38].

Protein S deficiency is now known to be heterogeneous with respect to its genetic basis. The recently published database of protein S gene mutations [22] lists missense mutations, stop codons, and frame-shift mutations. Most gene defects were associated with reduced levels of free protein S antigen. These mutations were reported in conjunction with normal or reduced total protein S antigen levels (or both). This result supports the concept that the important effect of such mutations is on free protein S levels. In the family that we investigated, the missense mutation resulted in decreased free protein S antigen levels, suggesting that our conclusions about risk for thrombosis are generalizable to other protein S-deficient families. Three reservations against general applicability, however, must be considered. First, in the affected family members, free protein S antigen levels were unambiguous and were well below the lower limit of the laboratory reference range. If a protein S gene mutation resulted in only an intermediate reduction of free protein S levels, the risk for thrombosis may be lower. Second, families with expressed dysfunctional variant protein S may have reduced risk for thrombosis, depending on the nature of the defect. Finally, as discussed above, the calculated risks often differ between family studies (such as ours) and population-based case–control studies. A potential genetic factor that could explain such differences, the factor V Arg506→Gln mutation (an additional risk factor for venous thrombosis in protein S deficiency [16]), does not confound the analysis in the family that we studied. Even so, the magnitude of the risk associated with protein S deficiency defined in this study cannot necessarily be applied to persons with reduced protein S levels outside the context of familial thrombophilia.

In summary, we provide evidence that in a clinically affected family, carriers of a protein S gene mutation are at greatly increased risk for venous thrombosis. In other families with protein S deficiency, identification of persons at risk is important so that prophylaxis can be offered in such high-risk situations as immobilization, pregnancy, and surgery. In the absence of molecular investigations, we believe that identification would be achieved most efficiently by assessing free protein S antigen levels.

Dr. Zoller: Department of Internal Medicine, University of Lund, University Hospital, S-22185 Lund, Sweden.

Drs. Garcia de Frutos and Dahlback: Department of Clinical Chemistry, University of Lund, S-20502 Malmo, Sweden.