Overall, our results indicate that HCV+ is associated with a significant, moderately increased risk of death in a prevalent cohort of blood donors who were presumably unaware of their HCV infection and in good apparent health at the time of their blood donation. Plausibly, the association between HCV and mortality was strongest for liver-related causes, but the absolute risk of liver death was still lower than that reported in hospital-based studies. However, deaths due to drug and alcohol abuse and trauma/suicide were also significantly associated with HCV+ and, together, accounted for more excess mortality than liver-related causes. We also noted an unexpected association between HCV and cardiovascular deaths, an intriguing finding in light of recent literature suggesting links between HCV and coronary or carotid artery atherosclerosis (16
We were somewhat surprised to find an approximately threefold hazard ratio estimate for all-cause mortality. On the basis of the two studies described previously and several other prospective studies of HCV mortality in the literature, we had originally favored the null hypothesis that HCV would not be associated with mortality in an initially healthy blood donor population. Two other studies of transfusion recipients followed for 15–20 years failed to demonstrate any increased mortality in HCV-infected patients, including one with 924 HCV+ patients in the United Kingdom with a hazard ratio = 1.17 (95 percent CI: 0.92, 1.49) (20
) and another with 222 HCV+ patients from several US cohorts (21
). Studies of predominantly injection drug user populations have yielded consistent estimates of increased mortality ranging from 1.4 to 1.7, which failed to reach statistical significance (22
). A clinic-based study of 838 German patients with HCV infection, including 271 transfusion recipients and 134 injection drug users, was the one study to find a significant, albeit moderately increased all-cause mortality after more than 4 years of follow-up (standardized mortality ratio = 1.6, 95 percent CI: 1.3, 2.0) (24
One possible explanation for our higher HCV hazard ratio than those of other epidemiologic studies is suggested by our calculation of standardized mortality ratios relative to the US population. Our comparison group of HCV− blood donors had only 70 percent of the population mortality after standardizing for age and sex, no doubt because of exclusion of unhealthy persons from blood donation. Our HCV+ subjects had a standardized mortality ratio of 2.13, indicating twofold increased mortality relative to the general population, instead of the threefold hazard ratio relative to our blood donor comparison group. In other words, it is possible that a “healthy blood donor” effect may have slightly inflated our mortality estimate. On the other hand, because our HCV+ subjects did pass blood donor medical selection criteria albeit with likely underreporting of HCV risk factors such as injection drug use, the general population may not be the appropriate comparison, so the “true” effect size may be between two and three.
On the other hand, our study found lower rates of liver-related mortality than did previous hospital- and clinic-based studies that may have been subject to selection bias (6
). The annual incidence of liver-related death was only about 1/1,000, or 2 percent over 20 years, in the HCV+ group, with liver cancer accounting for one fifth of these deaths. At least 10–15 years of HCV infection may precede serious disease and death, but we do not expect a substantial increase in liver mortality with follow-up of the cohort beyond 7.7 years post donation. From our earlier studies of blood donors, we found that most were infected during the epidemic of injection drug use in the 1960s and 1970s (2
), so our subjects had probably been infected for 20–30 years at the time of their index blood donation. Also, our cohort included a broad range of ages at enrollment, and we saw no trend toward increasing liver deaths with a longer interval between donation and death.
How did our HCV+ blood donors compare with their counterparts in the general population? We suspect that about half of our HCV+ blood donors were probably exposed during past injection drug use and that the remainder were exposed via blood transfusion or other parenteral exposures (2
). A survey of risky behavior among randomly selected blood donors found 0.5 percent lifetime injection drug use but only 0.04 percent injection drug use within the past 3 months (28
). That we still saw large excesses of deaths due to drug and alcohol abuse in a cohort with more remote injection drug use speaks to the chronic and recurrent nature of these addictions. Our finding of a threefold higher mortality due to trauma and suicide among HCV+ subjects could also be related to current or past substance abuse, which is known to be associated with trauma and psychiatric disease (5
). These premature deaths probably accounted for the association of younger age at donation with mortality in our multivariate model. After controlling for age at donation, we saw no evidence that suicide deaths clustered around the time of donation, suggesting that notification of HCV infection was not responsible for this excess. Without person-specific data on treatment for HCV infection, we could not test the hypothesis that interferon treatment for HCV may have contributed to the excess in suicide deaths, although another study found this to be unlikely (23
). Whatever the reason for the HCV association with deaths due to substance abuse, trauma, and suicide, some of these deaths may be preventable if primary care physicians address these issues in the counseling and treatment of HCV+ patients.
Our unanticipated finding of an association between HCV seropositivity and cardiovascular mortality may have at least two explanations and deserves additional research. First, there may have been potential confounding by unmeasured or inadequately controlled variables. We had incomplete racial data and did not have individual data for socioeconomic status, smoking, cholesterol, diet, or other risk factors among study subjects, which could have confounded our cardiovascular death association. Chronic cocaine injection has been linked to an increased risk of coronary artery calcification and could also confound our association with HCV (32
). Second, there may be a causal link between HCV infection and coronary disease. Accelerated coronary stenosis has been reported among patients who received HCV-infected cardiac allografts (16
) and among HCV-infected kidney transplant recipients (33
). Two other epidemiologic studies found associations between HCV and documented carotid artery (18
) and coronary (19
) atherosclerosis, while a third found positive associations with myocardial infection, stroke, and carotid atherosclerosis that were no longer significant after adjustment for age and other risk factors (34
). There have also been negative studies: HCV antibody status was not associated with first myocardial infarction in active duty military men (35
) nor was it among several chronic infections associated with the onset of carotid artery sclerosis (36
). HCV infection is often accompanied by an inflammatory response including cryoglobulinemia and vasculitis (37
), which could affect the coronary arteries, and has also been associated with a metabolic syndrome consisting of insulin resistance and type 2 diabetes, which may in turn predispose patients to accelerated coronary disease (39
The major strength of this study is that our data represent mortality in a relatively healthy population of blood donors among whom ongoing injection drug use is probably uncommon (28
). We had access to excellent data on standardized HCV antibody testing, name, address, and Social Security number. Finally, we believe that our ascertainment of vital status was accurate and complete in most cases, since National Death Index matching has been found to be highly sensitive and specific (40
There are several potential limitations of our study. First, our approach to controlling for confounding by race/ethnicity and socioeconomic status (i.e., matching by ZIP Code) is supported by other studies but is not a substitute for individual-level data (41
). Second, we could not analyze HCV viremia, because nucleic acid testing was done only on the fraction of donors enrolled since 1999. Third, high rates of competing mortality due to traumatic and substance abuse causes may have obscured some cases of liver disease that were present but unrecognized at death, or that would have developed if the HCV-infected deceased had lived longer. Conversely, unmeasured antiviral treatment or liver transplantation could have lowered liver disease mortality, especially in the later years of our study. Finally, our outcome data were derived from International Classification of
Diseases-coded death certificate data reported to the National Death Index, and we did not have access to original death certificates or medical records of subjects to better determine pathologic diagnoses.
In conclusion, these data suggest that overall mortality is moderately increased among HCV+ individuals and that much of the increase is due to trauma/suicide and drug/alcohol deaths, underscoring the need for evaluation and treatment for drug and alcohol abuse, depression, and trauma/suicide risk in such patients. Future research should measure morbidity, disability, and medical care associated with HCV infection by surveying a sample of living cohort members. We also hope that other cohorts of HCV-seropositive individuals will be studied in an effort to replicate our findings, particularly the cardiovascular mortality association.