Subjects, Materials, and Methods

Subjects All subjects who participated in the 1991–1994 Copenhagen City Heart Study were included in our population-based study. Subjects ⩾20 years old were selected randomly after age stratification from among residents of Copenhagen. Of the 17,180 individuals invited, 10,135 participated, 9259 provided blood samples, and 9253 were genotyped for factor V Leiden. Details of study procedures have been described elsewhere [15, 16]. More than 99% of subjects were whites of Danish decent. All subjects provided written, informed consent, and the ethics committee for Copenhagen and Frederiksberg approved the study. Participants of the Copenhagen City Heart Study completed a self-administered questionnaire, which was validated by the subject and an investigator on the day of attendance

The Danish civil registration system contains a personal identifier for all individuals living in Denmark [17]. The personal identifier was used as a key to retrieve and merge individual data from different databases

On the basis of the World Health Organization’s International Classification of Diseases (8th and 10th editions; available at: http://www.who.int.classifications/icd/en), data on morbidity were drawn from the Danish National Hospital Discharge Register (data from study inclusion until 31 December 2000 were available), whereas data on mortality were drawn from the Danish National Register of Causes of Death (data from study inclusion until 31 December 2001 were available). Infectious diseases were classified as pneumonia, 480–486, A48.1, and J12–J18; cystitis, 595 and N30; pyelonephritis, 590 and N10–N12; HIV/AIDS, 137 and B20–B24; tuberculosis, 010–018, A15–A19, and B90; skin infection, 680–686, A46, and L00–L08; diarrheal diseases, 000–001, 004–006, 008–009, A00–A01, A03A04, and A06–A09; sepsis, 036.1, 038, A32.7, A39.2–4, A40–41, and A48.3; hepatitis, 070 and B15–B19; parasitic infection, B50–B64; other viral infection, 040–061, 075, 079, B00–B09, and B25–B34; and meningitis, 320, A39.0, and G00–G03. Furthermore, we gathered information on cases of deep venous thrombosis (451.00, 451.08, 451.09, 451.90, 451.92, 671.01–671.09, I80.1, I80.2, I80.3, O22.3, and O87.1) and pulmonary embolism (450.99, 673.99, I26.0, I26.9, and O88.2)

DNA analysis The Arg506Gln mutation was identified by restriction-fragment length polymorphism polymerase chain reaction (PCR) [11, 18]. To avoid misclassification of factor V Leiden genotypes, all subjects had their genotypes determined on agarose gel by 2 independent investigators, and data entry was compared with a hard copy by 2 independent investigators

Statistical analyses All subjects were censored at death, emigration, or on 31 December 2000, whichever came first. Separate analyses were conducted from study inclusion to a first event for each hospitalization outcome (i.e., the total number of hospitalization events is larger than the number of subjects). Except for the analysis of genotype distribution, heterozygous and homozygous individuals were combined (“carriers”) and compared with noncarriers. The Kruskal-Wallis and χ² tests were used for univariate analyses. Cox regression analysis with forced entry was used to examine time until disease or death using hazard ratios (relative risks [RRs]) with 95% confidence intervals (CIs). The model was adjusted for sex, age at inclusion or diagnosis, year of study inclusion, smoking habits (current, former, or never), alcohol consumption (<21 vs. ⩾22 drinks/week), level of education (<9 years vs. ⩾9 years), and income (<200,000 vs. ⩾200,000 Danish kroner). In the analysis of the risk of skin infection, self-reported diabetes at baseline was included in the model. In the analysis of the risk of pneumonia, chronic lung disease at baseline was included in the model. Measurements of lung function (forced expiratory volume in 1 s [FEV₁] and forced vital capacity [FVC]) were done as described elsewhere [19], and chronic pulmonary disease at baseline was defined as FEV₁/FVC<70% and FEV₁ <80% of predicted values [20]

Comorbidity was assessed on the basis of a complete hospital-discharge history by the Charlson index score (0, low; 1–2, medium; and >2, high) [21]. Mortality associated with an infectious disease was defined as any death that occurred within 28 days of the primary diagnosis [5, 7, 13]

The likelihood-ratio test was used to test for interaction. There were no significant interactions. The adequacy of the model was checked by testing the proportional-hazards assumption in different ways: by conducting the traditional graphics check based on the log of the cumulative hazard and by performing a formal test of proportionality based on Schoenfeld residuals according to the method of Lemeshow and Hosmer [22]. Statistical analysis was performed with SPSS (version 11.5; SPSS) and Stata (version 9.0; Stata) statistical software. P<.05 on a 2-sided test was considered to be significant

Results

In this adult Danish general population cohort, 8534 persons were factor V Leiden noncarriers, 699 were heterozygotes, and 20 were homozygotes. This genotype distribution did not differ from that predicted by Hardy-Weinberg equilibrium (P = .16). Characteristics of subjects according to their genotype are shown in table 1. In all further analysis, heterozygotes and homozygotes were combined (carriers) and compared with noncarriers. Among all subjects, 204 had a history of thromboembolism during the study period; of these, 52 received a diagnosis of infection that led to hospitalization

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Table 1

Baseline characteristics of subjects, according to factor V Leiden genotype

Morbidity The cohort was monitored for a total of 66,789 person-years, during which a total of 1383 infections in 1093 individuals led to a first hospitalization in each disease category. The incidence of any hospitalization for infectious disease was 167/10,000 person-years (95% CI, 132–203) in factor V Leiden carriers, compared with 158/10,000 person-years (95% CI, 148–168) in noncarriers. This corresponded to an RR of infection of 1.08 (95% CI, 0.87–1.35; P=.49) in carriers, compared with noncarriers, after adjustment for age, sex, year of study inclusion, smoking, alcohol consumption, income, and level of education

Infections were grouped into 15 categories: meningitis, tuberculosis, diarrheal disease, HIV/AIDS, influenza, mycosis, other viral infections, other infections, parasitic infections, pneumonia, sepsis, skin infection, viral hepatitis, upper respiratory-tract infection, and urinary-tract infection. Meningitis (n=10), viral hepatitis (n=12), and influenza (n=10) all occurred in noncarriers. Of 14 persons with HIV/AIDS, only 2 were carriers; there was 1 carrier each among 35 persons with other infections, 19 persons with tuberculosis, and 27 persons with mycosis. Thus, the following analyses of infection risk according to factor V Leiden genotype included diarrheal disease, other viral infections, parasitic infections, pneumonia, sepsis, skin infection, upper respiratory-tract infection, and urinary-tract infection. With the exception of individuals with skin infection (n=1), pneumonia (n=2), and upper respiratory-infection (n=1), none of the 20 homozygous carriers were hospitalized for infections

The risk of urinary-tract infection was decreased in carriers, compared with that in noncarriers (RR, 0.55 [95% CI, 0.31–0.99]), whereas the risk of skin infection was increased (RR, 1.73 [95% CI, 1.11–2.71]) (table 2). There was no association between carrier status and risk of diarrheal disease, other viral infections, parasitic infections, pneumonia, sepsis, and upper-respiratory tract infection according to the results of univariate analysis

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Table 2

Morbidity leading to hospitalization, according to factor V Leiden genotype, during 8 years of follow-up of 9253 individuals from the general population

Carriers remained at a decreased risk of urinary-tract infection during follow-up (RR, 0.55 [95% CI, 0.31–0.99]), after adjustment for age, sex, year of study inclusion, smoking, alcohol consumption, income, and level of education. Other independent risk factors included age (RR, 1.02/year increment [95% CI, 1.02–1.03/year increment]), female sex (RR, 1.35 [95% CI, 1.03–1.78]), lower level of education (RR, 1.47 [95% CI, 1.13–1.92]), and low income (RR, 2.11 [95% CI, 1.56–2.86]). Subgroup analysis of subjects with either cystitis or pyelonephritis showed that both risks were nonsignificantly decreased in factor V Leiden carriers (RR, 0.59 [95% CI, 0.32–1.07] and RR, 0.32 [95% CI, 0.04–2.36], respectively). After the exclusion of individuals with a thromboembolic event before a diagnosis of urinary-tract infection, the adjusted RR was 0.50 (95% CI, 0.26–0.94)

Carriers remained at an increased risk of skin infection during follow-up (RR, 1.72 [95% CI, 1.09–2.72]), after adjustment for age, sex, year of study inclusion, smoking, alcohol consumption, income, and level of education. Other independent risk factors included age (RR, 1.02/year increment [95% CI, 1.01–1.04/year increment]), male sex (RR, 1.87 [95% CI, 1.35–2.59]), low income (RR, 1.43 [95% CI, 1.01–2.03]), and self-reported diabetes (RR, 3.07 [95% CI, 1.89–4.98]). Subgroup analysis showed that the RR of cellulitis in factor V Leiden carriers was higher (RR, 1.77 [95% CI, 1.01–3.11]), whereas the RR of boils/carbuncles in factor V Leiden carriers was 1.66 (95% CI, 0.79–3.48). After the exclusion of individuals with a thromboembolic event before a diagnosis of skin infection, the adjusted RR was 1.64 (95% CI, 1.01–2.69)

Mortality Mortality was defined as death within 28 days after admission to the hospital. Among 1261 cases of meningitis, tuberculosis, diarrheal disease, other infections, pneumonia, sepsis, skin infection, and urinary-tract infection, there were 127 deaths. Mortality rates ranged from 1.6% (diarrheal disease) to 37.3% (sepsis) (table 3). There were no deaths within 28 days among subjects with HIV/AIDS, influenza, mycosis, parasitic infections, other viral infections, viral hepatitis, or upper respiratory-tract infection

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Table 3

Twenty-eight–day mortality, according to factor V Leiden genotype, after hospitalization for infectious disease

Unadjusted Cox regression analysis showed that factor V Leiden carriers had an RR of death from sepsis at 28 days of 3.35 (95% CI, 1.29–8.76). After we controlled for age at diagnosis, sex, comorbidity, alcohol consumption, smoking, level of education, and income, factor V Leiden remained associated with disease-associated mortality (RR, 4.95 [95% CI, 1.50–16.33]). After the exclusion of individuals with a thromboembolic event before a diagnosis of sepsis, the adjusted RR was 4.81 (95% CI, 1.45–16.01). We did not find any association between deaths from pneumonia, urinary-tract infection, skin infection, diarrheal disease, tuberculosis, meningitis, or parasitic infections according to factor V Leiden genotype. After subgrouping and reanalysis according to known or unknown etiology, neither pneumonia, urinary-tract infection, nor skin infection was associated with mortality according to genotype

Discussion

In the present large, community-based study, subjects with the factor V Leiden mutation had an increased risk of hospitalization for skin infection, a decreased risk of hospitalization for urinary-tract infection, and an increased risk of short-term death from sepsis. A genetic association between the severity of sepsis and the presence of the factor V Leiden mutation has previously been shown, but the existing data have been conflicting [13, 14]. Kondaveeti et al. [14] showed that children carrying the factor V Leiden mutation who had meningococcal disease were at an increased risk of fulminant purpura. This may be biologically plausible, because intravascular thrombosis is a pathophysiological feature of meningococcal sepsis and because factor V Leiden increases the risk of venous thromboembolism by 3–18-fold in factor V Leiden heterozygotes and homozygotes [11]. The study by Kondaveeti et al. was underpowered to detect a difference in outcome. In contrast, in a large study of subjects with severe sepsis, factor V Leiden carriers had improved survival rates, compared with noncarriers [13]. That study suggested that the difference in survival resulted from an increased level of thrombin that subsequently increased levels of activated protein C but that the increased levels of thrombin did not increase the risk of excessive thrombosis. With regard to survival, our results support the opposite conclusion. There are differences between the studies that may have influenced the opposing results. Kerlin et al. [13] studied subjects in a randomized controlled trial in which individuals met specific inclusion and exclusion criteria (e.g., subjects with some chronic diseases, recent deep-vein thrombosis, pulmonary embolism, or thrombocytopenia were excluded), which thus possibly introduced bias. Ours was a population-based study in which all subjects who had received a diagnosis of sepsis were included. Both studies were adjusted for age and comorbidity at the time of sepsis, but our study could not adjust for disease severity, because such information was not gathered. Genetic-association studies of complex diseases notoriously come to conflicting results [23]. Further studies and meta-analysis will be needed to determine the real role that factor V Leiden plays in sepsis

Several single-nucleotide polymorphisms (SNPs) have been associated with an altered risk of infection [24], and others, mainly proinflammatory cytokines—such as tumor necrosis factor–α, interleukin (IL)–1β, IL-6, and plasminogen activator inhibitor–1—have been shown to affect the outcome of infectious diseases [25]. To our knowledge, no other SNP in the coagulative pathway has been implicated in susceptibility to infectious disease. Factor V Leiden increased the risk of hospitalization for skin infections. The reasons for this are not immediately clear. It may be that thrombosis at the site of an infection impairs leukocyte trafficking and other components of innate immunity by reducing tissue perfusion [6]. Certainly, subjects with decreased vascular perfusion associated with atherosclerosis and diabetes have an increased risk of skin infections [26, 27]. However, the risk of urinary-tract infection was reduced in factor V Leiden carriers, which suggests that a procoagulative state protects against urinary-tract infection. We speculate that mucosal shedding caused by thrombosis may occur and, thus, prevent infection

Infectious agents evoke the release of numerous inflammatory mediators that interconnect and interact with coagulation. With ongoing infection, coagulation homeostasis may fail and lead to a procoagulant and antifibrinolytic state termed “disseminated intravascular coagulation” [8]. Genetic variants of genes in the coagulation pathway, such as in the protein C gene, that cause thrombosis and decreased fibrinolysis have been shown to promote, sustain, or aggravate disseminated intravascular coagulation [28]. This is compatible with our findings, given that individuals with sepsis, which is commonly associated with disseminated intravascular coagulation, had an increased risk of death if they also carried the factor V Leiden mutation. SNPs in other genes involved in coagulation (e.g., activated protein C, thrombomodulin, and endothelial protein C receptor) have effects on coagulation, but their role in susceptibility to and the progression of infectious disease remains to be determined [28, 29]

The present study had several limitations. Subjects were genotyped only if they attended the 1991–1994 third session of the Copenhagen City Heart Study. Selection bias may have occurred if death or morbidity prevented certain individuals from being genotyped. However, 2 observations make substantial selection bias against genotype less likely: (1) age percentiles for noncarriers and carriers displayed a linear relationship, as would be expected if no selection occurred against carriers, and (2) the genotype distribution did not differ from Hardy-Weinberg equilibrium. Nevertheless, if such a bias exists, we may have underestimated the effect of the factor V Leiden mutation on the risk of infection and outcome. If a correction for multiple comparisons was performed, none of the associations would have remained statistically significant. This fact underscores the need for additional studies of a potential association between genetic polymorphisms of genes in the coagulative pathway and the risk of infection

Information on anticoagulant therapy is not available for the Copenhagen City Heart Study. If a risk of infection and death is conferred by the increased risk of thromboembolism caused by the factor V Leiden mutation, then the use of anticoagulants could bias the RRs that we have shown. However, when subjects with a known prior episode of venous thromboembolism, who may have received anticoagulant therapy, were excluded from the analyses, the associations remained largely unchanged

A lack of the identification of all infectious events in the present study could have introduced misclassification bias and tended to bias the association toward the null. However, because all subjects were treated in public hospitals in Denmark and all hospitals report to the Danish Hospital Discharge Register and the Danish National Register of Causes of Death, any underestimation of disease incidence were unlikely to have seriously affected our conclusions. Misclassification of end points probably cannot explain our findings, because such misclassification is probably independent of factor V Leiden genotype

The present study was based on hospitalization diagnosis codes; thus, milder episodes of, for example, urinary-tract and skin infections that did not require hospitalization were not captured. Because the subjects’ genotypes were unknown before hospitalization, this should not have biased our results, but it is possible that our findings do not apply to individuals who are treated for an infection on an ambulatory basis. Subjects with a known factor V Leiden mutation or a previous thromboembolic event may have had a lower threshold for admission to hospital, leading to differential selection bias. In Denmark, factor V Leiden genotyping is unlikely to have been performed unless a thromboembolic event had occurred; in fact, when all thromboembolic events were excluded from the regression analyses, the RRs remained largely unchanged

The presence of comorbidity is likely to influence the risk of hospitalization for an infection. Information was only ascertained at baseline for a select few comorbidities, but, when these were included in the regression analysis, they did not alter the associations that were detected, although they were independent risk factors. In the analysis of the risk of death, we used the Charlson score. This score estimates comorbidity on the basis of hospitalizations codes; thus, the presence of comorbidity may be underestimated. Only a few events occurred in some categories, and the findings should therefore be interpreted cautiously. In conclusion, we found that the factor V Leiden mutation may significantly affect susceptibility to urinary-tract and skin infections and increase mortality from sepsis. Further genetic association studies of the factor V Leiden mutation and the susceptibility to and outcome of infectious disease are warranted