Discussion

This large nationwide familial study of hospitalized cases of VTE in Sweden highlights the value of sibling history as a predictor of the risk of VTE. The family history of siblings appears to be an important risk factor for VTE with a higher relative risk than for many identified genetic risk factors.^{19–21,32–35} For many complex diseases, the average familial risks in first-degree relatives is ≈2, which is in agreement with the present study.³⁶ Although a familial risk of ≈2 might appear modest, it suggests that uncovering genetic and nongenetic sources of familial aggregation might be worthwhile.³⁶ Moreover, the dissected underlying familial risk factors (genetic or nongenetic) are likely to have a magnitude that is stronger than the familial risk itself.³⁶ The 5 major identified genetic risk factors contribute to only ≈30% of the family history.¹⁵ The newly identified common, but much weaker, alleles probably contribute to only a small fraction of the familial risks.^{19–21,32–35} It has been estimated that 10 additive alleles, each with a genotype relative risk of 2.0 and an allele frequency of 0.1, will only explain a familial risk of 1.06.³⁷ If multiplicative gene-gene interactions are present, the same 10 alleles will still only explain a familial risk of 1.2.³⁷ Many newly identified alleles have an estimated genotype relative risk of <1.5.^{19–21,22–35} An additive allele with an allele frequency of 0.1 and a genotype risk of 1.5 explains only a familial risk of 1.01.³⁷ Thus, the overall familial risk of 2.45 in our study is most likely explained in <50% by the alleles known today to be associated with VTE. Thus, in clinical practice, the family history may be more useful for risk assessment than testing for the presently identified alleles associated with VTE.

Our study is the first nationwide one demonstrating an increased familial risk among subjects aged 10 to 69 years but not in those aged <10 years. No significant difference in the magnitude of familial SIRs was observed for males and females, indicating that familial factors are important in both sexes, although the incidence rates for both familial and nonfamilial cases were higher for younger females and higher for older males. The effect of age and gender on the SIRs of siblings has not been determined before in a nationwide study. Thus, no previous studies similar to ours were available for comparison. However, the odds ratios for the family history (any parents, brothers, or sisters affected) as a risk factor for deep venous thrombosis have been estimated in a case-control study.¹⁵ The odds ratios decreased with age from 3.2 at age 18 to 29 years to 2.2 at age 60 to 69 years, in accord with our results. Divergent results regarding the influence of gender on the overall incidence rates in nonfamilial cases have been cited in the literature.^1–3,18 The incidence rate for familial cases has not been determined. Our study is in agreement with a previous report from Silverstein et al¹ that found an increased incidence rate among young women and a modest increase in men at older ages. Another interesting finding is that the familial SIRs decreased in magnitude with age but remained significant. Thus, familial factors are also important at older ages.

The highest familial risks of 51.87 and 53.69 were observed for siblings who had 2 and ≥3 affected probands, respectively. Together, these patients accounted for only 0.46% of all siblings. The high risks are in the order of magnitude of observed risks among individuals with combined prothrombotic genetic defects, such as homozygous activated protein C resistance mutation or the combination of heterozygous activated protein C resistance mutation and protein C, protein S, or antithrombin deficiency.^4,9 However, whether there are any other rare thrombotic genetic defects other than deficiencies of the natural inhibitors antithrombin, protein S, and protein C remains to be determined. Genome scanning of large thrombophilic families may be an important option for identifying such risk factors.³⁸

The low incidence rate in children aged <10 years in our study was similar to those in previous reports.³⁹ The lack of familial cases among children aged <10 years could be due to the small number of cases, but it could also reflect the fact that the risk is very low even among familial cases and that in most children aged <10 years, 1 or more acquired risk factors, such as cancer, infection, congenital heart disease, and a central venous catheter, are necessary to provoke VTE.⁴⁰ Our results are in agreement with family studies of already identified genetic risk factors such as deficiencies of protein C and protein S, in which thrombosis is seldom manifested before age 10 years.^41,42

The slightly increased risk for VTE among spouses, with a SIR of 1.07 (95% CI, 1.04 to 1.10), suggests that most of the familial risk for VTE is explained by genetic rather than family environmental factors. Moreover, age difference between siblings did not affect the familial SIR risks, which further supports the finding of a lack of any major familial environmental effect. Thus, our data suggest the view that most of the familial risks observed in our study are genetic, in agreement with a published Danish twin study.⁸

The present design with focusing of the study on hospitalized cases of VTE has potential advantages and disadvantages. However, many of the limitations in the study involve a nondifferential bias. The disadvantages include the reduction in the number of cases and the possibility that some selective factors operate in the process of hospitalization to favor siblings being hospitalized. Affordability of healthcare is probably not a selective factor in Sweden, nor is the likelihood of seeking medical advice important because of equal access to primary and hospital care. Hospitalizations for VTE and other conditions often require a physician's referral from primary care. Thus, each hospitalized patient is often seen by 2 physicians, of which the physician in the hospital is often a specialist. The low spouse correlation does not support any strong selection bias for hospitalization of certain families.

Because the follow-up period was 21 years, some sibling pairs were lost because one of the pairs was hospitalized before the start of the follow-up in 1987. This can lead to an underestimation or overestimation of the familial SIR if more or fewer familial cases are lost than nonfamilial cases and they are not in proportion to the estimated familial SIRs. Our total estimates of familial SIRs for males and females of 2.52 and 2.37 are similar to the reported odds ratios of 2.2, 2.3, and 2.7 for the risk of venous thrombosis due to a family history of VTE,^15–17 suggesting that our study gives a valid estimate of the familial risks in siblings. One explanation could be the high risk of recurrent VTE. Thus, many lost siblings will eventually be affected by a thrombotic event during the follow-up period between 1987 and 2007 and classified as thrombotic cases. Another potential nondifferential bias is the introduction of low-molecular-weight heparin and the treatment of many deep venous thrombosis patients as outpatients. However, age-specific incidence rates were calculated for the whole follow-up period, divided into five 5-year periods, and an adjustment was done for the possibility of changes in the incidence rate over time. A number of less severe thrombotic cases are treated as outpatient cases. If familial cases tend to be more severe than nonfamilial ones, this could overestimate the VTE risk, but there is no evidence supporting this hypothesis. However, under any circumstances, our data will reflect the risk for severe hospitalized cases.

The advantages include complete nationwide coverage in a country with a high standard in the medical diagnosis of siblings, often by a specialist during extended examinations in the clinic. Importantly, the diagnoses are those with which the patients are discharged and are thus accurate. The Swedish Hospitalization Discharge Register contains no information about diagnostic procedures, which is a limitation, but it would be a nondifferential bias. However, with respect to VTE, the overall diagnostic validity of the Swedish Inpatient Register is close to 90%.⁴³ Moreover, the diagnosis of VTE in the Swedish Inpatient Register has been validated and found to be correct in 95% of VTE diagnoses.⁴⁴ In addition, in 91% of cases, VTE is diagnosed by an objective method.⁴⁴ This is in agreement with a recent Swedish study demonstrating that VTE is seldom diagnosed without objective investigations.⁴⁵ Because it is possible that the diagnostic methods might have varied between geographic regions, we adjusted for geographic region to minimize this possible bias. The high quality of the Swedish Inpatient Register for VTE is in agreement with studies that have found a validity of ≈95% for stroke and myocardial infarction.^46–48 Only a main diagnosis of VTE was used in the present study to further increase the accuracy.

An important advantage of our study is that it is nationwide and based on all hospitalized patients without the risk of selection or recall bias. The Swedish family data set (ie, the Multigeneration Register) is also a validated source that has been proven to be reliable in the study of many familial diseases.^22–26 Data in the MigMed Database are almost complete. In 2001, personal code numbers were missing in only 0.4% and the main diagnosis in 0.9% of hospitalizations.^22–26 Information on occupational status, retrieved from the national census records in the database, was 99.2% complete.^22–26 Because environmental risk factors were not available, we adjusted for socioeconomic status (occupation) in the models as a proxy for socioeconomic environment.

The lack of major risk factors is a potential confounder, but it is most likely a nondifferential bias. Moreover, major risk factors do not appear to affect the overall familial risk¹⁵ and are not usually included when familial risks are estimated.^15–17

In our study, siblings were not excluded because of cancer or any other disease, in agreement with the study by Bezemer et al,¹⁵ a practice that is otherwise common when familial risks are studied. Thus, our data reflect the total impact of a familial history of VTE in the whole population of Swedish siblings. Another possible strength of our study is that we included not only cases of deep venous thrombosis but also superficial thrombophlebitis, pulmonary embolism, and rare manifestations of venous thrombosis such as cerebral sinus thrombosis, representing the whole spectrum of thrombotic manifestations.⁴ This definition of VTE could also be a potential limitation if the genetic background is different for different manifestations. However, this definition of VTE has been used by Souto et al⁶ and is common in studies of familial thrombophilia.^41,42 We tested for heterogeneity of familial risks, but no differences were observed for the different manifestations of VTE. Our total estimate of familial SIRs is similar to those reported for risk of venous thrombosis and pulmonary embolism,^15–17 suggesting that our definition of a VTE study provides a valid estimation of the familial risks.

In conclusion, familial factors are important for the development of VTE, at least from the age of 10 to 69 years. The relative contribution of familial factors decreases with age, although the absolute risk increases. Although some sex-related differences exist regarding the age-specific familial incidence rates, familial factors are highly important in both females and males. Our data support the view that VTE is a multifactorial disease with familial risks comparable to those of many other complex disorders. However, in a small fraction of siblings, the familial risk is very high, suggesting a segregation of rare but strong genetic risk factors. Epidemiological studies such as ours may be helpful in designing future genetic studies of VTE.