While ischemic heart disease (IHD) remains the most common cause of sudden death overall, arrhythmias and conduction disorders become more predominant as causes of death in the young (1). Although a proportion of premature deaths remain unexplained after routine autopsy (1), molecular autopsies have recently identified mutations associated with QT-interval prolongation and more subtle abnormalities on electrocardiography, as well as long QT and Brugada syndromes, in some of the deceased (2). Furthermore, cardiac evaluations in relatives of the victims have found a variety of cardiac conditions (6) and associated mutations (7). A family history of sudden death has been shown to increase the risk of ventricular arrhythmia (8), while sudden (9) and premature death (10) in relatives is associated with an increased risk of a similar death. These findings suggest that fatal cardiovascular events (e.g., lethal arrhythmias, myocardial infarctions) and less severe cardiovascular diseases may co-occur in families. Thus, a family history of premature cardiovascular death may indicate a familial cardiac frailty that predisposes to early cardiovascular disease.

Despite these suggestive studies and the idea of familial cardiac vulnerability at which they hint, the risk of early cardiovascular disease given a family history of premature cardiovascular death has never been examined in a large population-based study. Earlier studies have lacked power and/or control groups, and have been limited by selection of heavily affected families (6) and self-reported family history of cardiovascular events (e.g., “heart attack” [(13)] or “sudden cardiac death” [8]). We aimed to address these limitations by studying family history of premature death and subsequent risk of early cardiovascular disease in a cohort study based on all residents of Denmark. Our aim was to evaluate the effects of premature cardiovascular and noncardiovascular deaths in relatives on the risk of early cardiovascular disease overall, and in particular on the risk of IHD and arrhythmias.

Free, universally accessible, high-quality public health care and registration of health information in Denmark permit the conduct of sound register-based epidemiological studies. Since 1968, the Danish Civil Registration System (CRS) has assigned to each Danish resident a unique personal identification number used in all Danish national registers, permitting linkage of individual-level information from multiple registers (14).

The Danish Family Relations Database (DFRD), a virtual pedigree tool with kinship information from parent–child links registered in the CRS, allows for the identification of family members for most persons residing in Denmark. For most individuals born after 1949, first-degree relatives (parents, children, and siblings) and half-siblings can be identified. Second-degree relatives (grandparents, grandchildren, aunts/uncles, nieces/nephews) and third-degree relatives (full cousins) can be identified for 90% of individuals born after 1984. (However, the analysis of second- and third-degree relatives was not restricted to individuals born after 1984.)

The Danish National Patient Register (NPR) contains information on inpatient diagnoses assigned since 1978 and outpatient diagnoses assigned from 1995 onward (15). Diagnoses are registered using International Classification of Diseases (ICD) codes, with ICD-8 codes used until 1993 and ICD-10 codes used thereafter. Register quality and diagnosis validity are considered good (15).

The Danish Causes of Death Register contains death certificate information, with registration required by law; virtually all deaths have been registered since 1970 (18).

The National Diabetes Register is based on hospital diabetes diagnoses, filled prescriptions for diabetes treatment, primary care measurements of blood glucose, and diabetic chiropody. The register contains nationwide information on diabetes diagnoses from 1991 onward. Validation of diabetes diagnoses in this register found a sensitivity of 96% and a positive predictive value of 89% (19).

The Danish Register of Medicinal Product Statistics allows tracking of individual prescription medication histories from 1994 onward (20). Dispensed medications are identified by Anatomical Therapeutic Chemical classification codes. As a proxy for elevated blood lipid levels, we identified persons with hyperlipidemia, defined as having filled 2 or more prescriptions for any lipid-modifying drug (Anatomical Therapeutic Chemical code C01).

The study cohort included all residents of Denmark born after 1949 and alive on January 1, 1978, or born thereafter. Cohort members were followed up from January 1, 1978, or birth, whichever came later until the first of the following events: 1) cardiovascular disease at <50 years of age; 2) 50th birthday; 3) death; 4) emigration; 5) designated “missing” in the CRS; or 6) December 31, 2008 (end of follow-up). In cardiovascular disease subgroup analyses, other types of cardiovascular disease were ignored. A person was considered to have developed cardiovascular disease upon first registration in the NPR with any of the following: ICD-8 codes 390.00 to 429.9; ICD-10 codes I00 to I51. We also considered the following cardiovascular disease subtypes: IHD (410 to 414.99, I20 to I25.9); hypertension (400.09 to 99, I10 to I15); arrhythmia (427.20 to 97, I44 to I49.9); supraventricular arrhythmias (427.90, 427.93 to 4, I47.1, I47.8 to 9, I48); ventricular arrhythmia (427.91, 427.97, I46, I47.0, I47.2, I49.0); non–first-degree atrioventricular blocks (427.21, 427.23, and I44.1,2,3); Wolff-Parkinson-White syndrome (I45.6; no corresponding ICD-8 code). Possible undiagnosed or unregistered arrhythmias were captured by identifying persons registered with >2 episodes of syncope (427.24, 782.5, I45.9, R55), hereafter referred to as “recurrent syncope.”

Using the DFRD, we identified known relatives for each cohort member; to be included in the cohort, a person had to have at least 1 identifiable relative in the DFRD. Due to the nature of CRS (upon which the DFRD is based), the proportion of identifiable relatives varied with degree of kinship (see Data Sources) (Table 1). Using the CRS, we then identified those relatives who had died before age 60 years. Using information from the Causes of Death Register, we grouped premature deaths into: 1) cardiovascular deaths (deaths due to cardiovascular disease as defined above [see Study Cohort and Outcomes]); 2) possible cardiovascular deaths (a heterogeneous group of deaths likely to include misclassified cardiac deaths, including deaths due to stroke and cerebral hemorrhage [ICD-8 codes 430 to 435.99, ICD-10 codes I60 to I67], unexplained deaths [795.99, 796.39, 795 to 796, R95 to R99], asthma deaths [493, J46], and epilepsy [345, G40 to G41.9]); and 3) noncardiovascular deaths, all deaths other than the cardiovascular and possible cardiovascular deaths.

Family history of premature death was considered as a time-dependent variable. A cohort member was classified as exposed from the date of death of the first relative (if any) to die before age 60 years. If the premature death occurred before the cohort member's birth, he or she was considered exposed from birth. If there were no premature deaths in a cohort member's family, all of his follow-up time was classified as unexposed.

We estimated the relative risk of early cardiovascular disease by family history of premature death. Using log-linear Poisson regression, we calculated incidence rate ratios (IRRs) comparing the rate of early cardiovascular disease among persons whose relatives died prematurely with that among persons without such family history. To reduce possible bias from incomplete identification of family members, in relative-specific analyses we compared only those individuals who had a relative of that type. All IRRs were adjusted for attained age (1-year categories), sex, and calendar period (5-year categories). When evaluating risk of early IHD, additional IRRs adjusted for hypertension, diabetes, and hyperlipidemia were estimated in subanalyses with follow-up from 1994 onward. Homogeneity of IRRs was tested using Wald chi-square tests. All analyses were conducted using PROC GENMOD in SAS version 9.2 (SAS Institute Inc., Cary, North Carolina).

In the main analysis with any early cardiovascular disease as the end point, the cohort consisted of 3,985,301 persons born in 1950 or later and followed up for 89,294,258 person-years; 119,432 persons were censored at some point due to emigration and 2,063 were designated missing in CRS. We identified 257,711 premature deaths among relatives; 34,362 were cardiovascular deaths, 27,162 of which were from IHD. Among the cohort members, 129,825 were diagnosed with an early cardiovascular disease during the follow-up period. (Table 1) shows the basic characteristics of cohort members with early cardiovascular disease. (Table 2) shows age and sex distributions for those diagnosed with early IHD and arrhythmias.

Given a premature cardiovascular death in any first-degree relative, the risk of developing an early cardiovascular disease was increased 72% (95% confidence interval [CI]: 68% to 77%) compared with the risk in persons with no history of premature cardiovascular death among first-degree relatives. In contrast, having a first-degree relative who died prematurely of noncardiovascular causes increased the risk of early cardiovascular disease only 12% (95% CI: 10% to 14%). The difference between these estimates was statistically significant (p_{cardiac vs. noncardiac} < 0.0001). (Table 3) provides estimates of the associations between a family history of premature death (by type of death and type of relative) and early cardiovascular disease. Given premature deaths from IHD in first- or second-degree relatives, the IRRs for early cardiovascular disease were 1.70 (95% CI: 1.65 to 1.7) and 1.14 (95% CI: 1.07 to 1.23), respectively, and IRRs for IHD were 2.27 (95% CI: 2.17 to 2.39) and 1.37 (95% CI: 1.10 to 1.71), respectively.

(Table 4) shows IRRs for early cardiovascular disease by number of premature cardiovascular deaths among first- and second-degree relatives. Two or more premature cardiovascular deaths among either first- or second-degree relatives significantly increased the risk of early cardiovascular disease, by 2-fold or more compared with 1 premature death among the relatives in question (p_{≥2 vs. 1 first-degree relative} < 0.0001; p_{≥2 vs. 1 second-degree relative} = 0.072). Risks conferred by premature cardiovascular deaths were consistently lower for deaths among grandparents (second-degree) than for parental deaths (first-degree) (p_{parents vs. grandparents} < 0.0001). We found that risk estimates for full siblings (first-degree) and half-siblings (second-degree) were comparable (p_{full siblings vs.half-siblings} = 0.70), and risks associated with premature deaths in full siblings and half-siblings were greater than risks associated with parental and grandparental deaths, respectively (p_{full siblings vs. parents} < 0.0001; p_{half-siblings vs. grandparents} = 0.0004) (Table 4).

We also stratified our analyses by age and relative's age at death (age <35 years vs. ≥35 years) (Figure 18_gr1). The increases in risk of any early cardiovascular disease and ventricular arrhythmia associated with cardiovascular death in a relative were greatest early in life and when the relative died at a young age.

Analyses adjusted for sex, age, and calendar period (Table 3) showed that risk of IHD increased 2- to 5-fold given premature cardiovascular deaths in first-degree relatives and up to 2-fold given premature cardiovascular deaths in second-degree relatives. In subanalyses with follow-up from 1994 onward, we were also able to adjust IRRs for diabetes, hypertension, and hyperlipidemia. (Table 5) compares IRRs adjusted only for sex, age, and calendar period with the more fully adjusted IRRs (follow-up in both instances from 1994 onward). Although most IRRs decreased somewhat, the pattern seen in (Table 4) was unchanged and the additional adjustment did not affect our conclusions.

We observed moderately strong, statistically significant associations between premature cardiovascular deaths in first-degree relatives and early arrhythmias for most types of arrhythmias, whereas for second-degree relatives only the association with ventricular arrhythmia was significant (Table 6). Associations between premature cardiovascular deaths in first- and second-degree relatives and recurrent syncope were weaker but still significant (IRR: 1.21 [95% CI: 1.16 to 1.28] and IRR: 1.10 [95% CI: 1.04 to 1.16], respectively). When excluding persons with pre-existing IHD, IRRs for any arrhythmia and ventricular arrhythmia decreased by up to 25%. However, most estimates remained significantly elevated, with higher risks associated with deaths in first-degree relatives and ≥2 deaths (Table 6).

Unexplained premature deaths in relatives increased the risks of ventricular arrhythmias by as much as 80% (Table 3). For supraventricular arrhythmias, the IRRs given unexplained deaths in first- and second-degree relatives were 1.19 (95% CI: 1.03 to 1.38) and 1.12 (95% CI: 0.90 to 1.38), respectively. For recurrent syncope, the corresponding estimates were 1.44 (95% CI: 1.32 to 1.58) and 1.16 (95% CI: 1.05 to 1.28). IRRs for atrioventricular blocks and Wolff-Parkinson-White syndrome were not significantly different from 1.

In a cohort of almost 4 million persons and 89 million person-years of follow-up, we found strong associations between a family history of premature cardiovascular death and risk of early cardiovascular disease. Risks increased with young age, closer kinship, number of premature deaths among relatives, and younger age of relative at death. Our findings suggest that a family history of premature cardiovascular death, and possibly unexplained death, signals a familial predisposition to cardiovascular disease, ventricular arrhythmia in particular. Our results are clinically relevant because, beyond specific arrhythmic syndromes and the cardiomyopathies, known associations between genetic variants and cardiovascular disease risk are modest and not useful for general cardiovascular disease risk prediction.

Associations were stronger for cardiovascular deaths than for noncardiovascular deaths, suggesting that premature cardiovascular deaths in the family indicate a predisposition to early cardiovascular disease. Similar to previous reports, most premature cardiovascular deaths were due to IHD, and a substantial proportion of the association with early cardiovascular disease probably reflects this finding (1). However, analyses excluding persons with IHD also yielded strong associations between premature deaths in the family and arrhythmias. Several categories of possible cardiovascular deaths, most interestingly the unexplained deaths, were also significantly associated with early cardiovascular disease, most notably arrhythmias and recurrent syncope, supporting the claim that unexplained deaths suggest a familial predisposition to arrhythmias and other cardiac diseases (6). Explanations for the modest increase in risk conferred by noncardiovascular deaths are: 1) misclassification of cardiovascular deaths as noncardiovascular (minimized by grouping possible cardiovascular deaths separately); 2) ascertainment bias due to increased medical attention to persons with a premature death in the family; 3) environmental factors (e.g., smoking or air pollution) common to family members that cause both premature noncardiovascular death and early cardiovascular disease (unlikely, as candidate factors clustering in families and strong enough to produce both outcomes early enough in life are difficult to identify); or 4) shared genetic conditions causing both premature death and early cardiovascular disease, such as familial cardiac frailty conferred by a nonideal ion-channel profile resulting in arrhythmia (early cardiovascular disease) in some family members and premature death with noncardiovascular causes (e.g., infections complicated by a lethal arrhythmia) in others.

Early disease onset is often a marker for a genetic contribution to disease. Results of analyses examining the effect of the relative's age at death and cohort members attained age reinforce the idea of an underlying genetic mechanism; the risk of early cardiovascular disease was greatest in the young and with deaths at an early age, with relative risks as high as 11.

The Paris Prospective Study reported a relative risk of 9 for sudden death given the sudden death of both parents (9), and the INTERHEART study recently reported that a history of myocardial infarction in both parents increased an individual's risk of myocardial infarction more than 2-fold (21). Our study expands on these findings by showing that ≥2 premature cardiovascular deaths in the family greatly increase the risk of several types of early cardiovascular disease. Furthermore, we evaluated the effects of premature cardiovascular death in specific types of relatives. Stronger associations with premature cardiovascular deaths in parents as opposed to in grandparents support the idea of a complex, polygenic familial cardiac vulnerability predisposing to early cardiovascular disease. However, the associations with premature cardiovascular deaths in full siblings and half-siblings were both stronger than those with premature cardiovascular deaths in parents and grandparents, suggesting that while shared genes confer some degree of risk, the interaction of shared environment and shared genetic predisposition might also be important. On the other hand, surveillance bias in terms of increased medical attention to surviving siblings may have led to an overestimation of sibling effects.

IHD has a complicated genetic component (22); previous studies of myocardial infarction have identified patterns similar to those we observed (11). Although our conclusions regarding the effect of a family history of premature cardiovascular death on risk of early IHD were unaffected by adjustment for diabetes, hypertension, and hyperlipidemia, the effects of sibling and half-sibling deaths decreased substantially, indicating that a large component of these associations is explained by the additional risk factors. However, because IHD risk evaluation does not currently systematically incorporate family history (22) and genetic mechanisms underlying IHD are largely unexplained, our findings are of clinical value in primary prevention and risk-evaluation settings (13).

Premature cardiovascular deaths in relatives were significantly associated with the risks of recurrent syncope and most types of arrhythmia, with the risk of ventricular arrhythmia increased most dramatically. That these associations persisted when only nonischemic arrhythmias were considered implies that the increased risks were not simply a reflection of familial IHD and that arrhythmias, ventricular arrhythmia in particular, have a genetic component (a conclusion reinforced by our findings of very strong relative risks for ventricular arrhythmia in the young).

A family history of premature noncardiovascular death was associated with a 20% to 40% increased relative risk of arrhythmia and increases in ventricular arrhythmia risk given a premature death in a first-degree relative were comparable for unexplained and cardiovascular deaths (81% vs. 94%). Unexplained deaths were not significantly associated with the nonventricular arrhythmias but were associated with recurrent syncope. Previous studies have associated unexplained deaths in relatives with early cardiovascular disease, especially arrhythmias (4). Our results support the notion that arrhythmias with a heritable component contribute to a proportion of premature cardiovascular deaths and possibly also to some unexplained and even noncardiovascular deaths (1).

Overall, the NPR is regarded as a valid data source (15), but incomplete or inaccurate registration of cardiovascular diseases may have resulted in misclassification and potentially inflated IRRs. Diagnoses of myocardial infarction have been validated; with a sensitivity of 97%, a false-positive rate of 6.5%, and a positive predictive value of 93% (17), for heart failure the sensitivity is 29% and the positive predictive value is 81% (16). The validations are reassuring, especially regarding myocardial infarction, which is a major contributor to both the group of any early cardiovascular disease and IHD. The low sensitivity of heart failure could only inflate IRRs for any early cardiovascular disease and only if the rate of heart failure diagnoses increased in cohort members after a relative suffered a premature death due to increased medical awareness. Notably, in analysis not including heart failure, but examining IHD and arrhythmias, associations are stronger than for any early cardiovascular disease. This finding suggests that misclassification of outcomes due to possible limitations of the NPR was not a major source of bias.

More than 99% of Danish death certificates are registered in the Causes of Death register (18). Although the validity of myocardial infarction as a cause of death is good (positive predictive value: 97%; sensitivity: 89%) (17), studies have found poorer agreement between death certificates and autopsy findings for other causes (3). However, segregating possible cardiovascular causes of death in a separate group decreased the likelihood of misclassification of cardiovascular deaths as noncardiovascular.

Conventional risk factors for IHD (environmental and behavioral) explain a significant proportion of IHD, especially after age 50 years. However, we examined IHD risk in persons <50 years of age, and our associations were particularly strong in persons <35 years. Any risk factor would have to cluster strongly within families and be strongly associated with both premature death and early cardiovascular disease to produce the familial clustering we observed (29). Smoking and body mass index are not registered at a population level, but information on diabetes, hypertension, and hyperlipidemia were available. Adjustment for these risk factors only attenuated IRRs for IHD, suggesting that conventional risk factors cannot explain our findings; this is consistent with the findings from the INTERHEART study regarding parental history and risk of myocardial infarction (21). However, without adjustment for other risk factors, we cannot determine the causality behind the effect of family history. Future studies with evaluation of family history in combination with more risk factors might allow for direct clinical application of results regarding IHD.

We had no information on early cardiovascular disease diagnosed between 1950 and 1977, but those diagnosed with early cardiovascular disease before 1977 are likely to have come to medical attention again later, such that very few persons born before registration of cardiovascular disease diagnoses began were likely to have been misclassified as free of early cardiovascular disease. Only 3% of the cohort members were censored, primarily due to emigration. Bias due to loss to follow-up is unlikely as this would require that a family history of premature cardiovascular death, early cardiovascular disease, and emigration were somehow related. Ascertainment bias may theoretically have occurred if persons with a family history of premature death were more likely to have early cardiovascular disease diagnosed (as a consequence of increased medical attention to the survivors) than were persons without a death in the family. However, this is unlikely to have been a major problem as many outcomes were serious enough to have been brought to medical attention regardless of family history (e.g., myocardial infarction and ventricular arrhythmia).

Use of the Danish national registers allowed us to avoid differential misclassification of exposure and outcome. We obtained pedigree information from the Danish Family Relations Database without requiring any individual-level contact, allowing for the tremendous size of our study; our cohort included almost 4 million persons, with more than 89 million person-years of follow-up, making this, to our knowledge, the largest and most comprehensive study of the family history and risk of early cardiovascular disease.

Premature cardiovascular death in relatives was consistently associated with elevated risks of early cardiovascular disease, with significant differences in strength of association depending on relative type, degree of kinship, and age. These findings support the hypothesis that a family history of premature cardiovascular death signals a familial predisposition to early cardiovascular disease and emphasizes the contribution that a detailed family history may make in evaluations of early cardiovascular disease risk.