Risk Factors for Hepatitis C Virus Infection in United States Blood Donors

Transcription

1 Risk Factors for Hepatitis C Virus Infection in United States Blood Donors EDWARD L. MURPHY, 1 STEPHEN M. BRYZMAN, 2 SIMONE A. GLYNN, 3 DANNIE I. AMETI, 3 RUTH A. THOMSON, 3 ALAN E. WILLIAMS, 4 CATHARIE C. NASS, 5 HELEN E. OWNBY, 6 GEORGE B. SCHREIBER, 3 FANHUI KONG, 3 KEITH R. NEAL, 7 AND GEORGE J. NEMO, 8 FOR THE NHLBI RETROVIRUS EPIDEMIOLOGY DONOR STUDY (REDS*) SEE EDITORIAL ON PAGE 790 Injection drug use (IDU) is a known risk factor for hepatitis C virus (HCV) infection, but the strength of other parenteral and sexual risk factors is unclear. In 1997, we performed a case-control study of 2,316 HCV-seropositive blood donors and 2,316 seronegative donors matched on age, sex, race/ethnicity, blood center, and first-time versus repeat-donor status. Odds ratios (OR) and 95% confidence intervals (CIs) were calculated using conditional logistic regression. Questionnaires were returned by 758 (33%) HCV and 1,039 (45%) control subjects (P.001). The final multivariate model included only the following independent HCV risk factors: IDU (OR 49.6; 95% CI: ), blood transfusion in non-idu (OR 10.9; 95% CI: ), sex with an IDU (OR 6.3; 95% CI: ), having been in jail more than 3 days (OR 2.9; 95% CI: ), religious scarification (OR 2.8; 95% CI: ), having been stuck or cut with a bloody object (OR 2.1; 95% CI: ), pierced ears or body parts (OR 2.0; 95% CI: ), and immunoglobulin injection (OR 1.6; 95% CI: ). Although drug inhalation and a high number of lifetime sex partners were significantly more common among HCV seropositives, they were not associated with HCV after controlling for IDU and other risk factors. IDU, blood transfusion among non-idu, and sex with an IDU are strong risk factors for HCV among United States blood donors. Weaker associations with incarceration, religious scarification, being stuck or cut with a bloody object, pierced ears or body parts, and immunoglobulin injection must be interpreted with caution. (HEPATOLOGY 2000;31: ) Since the discovery of hepatitis C virus (HCV) was reported in 1989, much has been learned about its epidemiology and pathogenesis. HCV seroprevalence in the general population ranges from 1% to 2% in a number of countries including the United States 1-3 to 12.6% in parts of Italy, 4 and 14.1% in areas of Japan. 5 As a result of selection for those at low risk of infectious disease, HCV prevalence is lower among blood donors, ranging from 0.06% to 1.3% in several countries, and 0.4% in the United States It is hyperendemic among injection drug users (IDUs) in industrialized countries, with infection rates of up to 90%, 11,12 consistent with the high frequency of parenteral blood exposures in this subgroup. Most HCV seropositives have persistent viremia, more than half have chronic hepatitis, and cirrhosis may occur in up to 20%. 13 Other investigators have implied that up to 40% of HCV seropositives do not have recognized parenteral risk factors, 14 leading to speculation that other as-yet-undiscovered modes of transmission may exist. Whether heterosexual transmission occurs at more than a negligible rate is also controversial Furthermore, a recent study reported that intranasal inhalation of cocaine appeared to be a risk factor for HCV infection in United States blood donors. 19 Clarification of the risk factors for and transmission modes of HCV is essential in developing better strategies to exclude persons at increased risk for HCV infection from donating blood, and in the prevention of secondary transmission from the estimated 4 million HCV seropositives in the United States. The Retrovirus Epidemiology Donor Study (REDS) has developed surveillance systems and survey methodologies to gather infectious-disease and demographic data at 5 U.S. blood centers since 1990, and has recently completed a large study on HCV seroprevalence. 20 We therefore undertook a large case-control study to evaluate known and putative risk factors for HCV infection. Abbreviations: HCV, hepatitis C virus; IDU, injection drug use(r); OR, odds ratio. From the 1 University of California San Francisco, San Francisco, CA; 2 East Bay Family Practice, Oakland, CA; 3 Westat, Inc., Rockville, MD; 4 American Red Cross Jerome H. Holland Laboratory, Rockville, MD; 5 American Red Cross Blood Services Chesapeake Region, Baltimore, MD; 6 American Red Cross Blood Services SE Michigan Region, Detroit, MI; 7 University of Nottingham, UK; and the 8 National Heart, Lung and Blood Institute, Bethesda, MD. Received August 11, 1999; accepted December 15, *See Appendix for a list of REDS investigators. The Retrovirus Epidemiology Donor Study (REDS) is funded by the following research contracts from the National Heart Lung and Blood Institute: N01-HB (superceded by ), , , , , and Address reprint requests to: Edward L. Murphy, M.D., M.P.H., Associate Professor, Laboratory Medicine, UCSF, Box 0884, San Francisco, CA murphy@labmed.ucsf.edu; fax: Copyright 2000 by the American Association for the Study of Liver Diseases /00/ $3.00/0 756 PATIENTS AND METHODS A case-control study design was followed. From surveillance data collected by REDS, we identified all 2,316 HCV-seropositive blood donors who had given blood in 1994 and 1995 at 5 large U.S. blood centers (see Appendix for details on the participating blood centers). Autologous blood donors were excluded, as were HCV positives from January through August 1994 at the American Red Cross Chesapeake/Potomac blood center, because these donors were involved in another HCV study. HCV cases had positive reactions on both an enzyme immunoassay and recombinant immunoblot, as described in detail previously. 20 A total of 2,316 HCV-seronegative control donors from the same database were matched to the HCV seropositives on age, sex, race/ethnicity, center, and first-time versus repeat-donor status. We also attempted to choose controls with donation dates similar to cases; the maximum difference in donation

2 HEPATOLOGY Vol. 31, No. 3, 2000 MURPHY ET AL. 757 dates between matched cases and controls was approximately 7 months. In March and April 1997, anonymous questionnaires were mailed by the blood centers to all 4,632 HCV seropositives and controls. Questionnaires designed specifically for this study (available upon request) were identical for cases and controls except for a final section with several questions on hepatitis symptoms and diagnoses for the HCV seropositives. Survey procedures have been described previously, 21 with the exception that follow-up mailings of the questionnaire were sent only to subjects who did not return a postcard to the blood center indicating completion of the survey. Survey procedures were approved by institutional review boards at the participating centers. Odds ratios (OR) and 95% confidence intervals (CIs) were calculated using conditional logistic regression (PROC PHREG on the SAS system 22 ). First, to assess univariate associations between HCV status and hypothesized risk factors, models were constructed that included only the risk factor in question as the independent variable and HCV status as the outcome variable. Strata corresponding to our matching variables were defined as follows: age at donation (less than 20, 20-29, 30-39, 40-49, 50-59, 60 years or more); sex (male, female); race/ethnicity (white, black, Hispanic, Asian, other); blood center (5 centers); and first-time versus repeat-donation status. Subjects with missing data were excluded from the models, but this did not usually exceed 10% of the total population. Because initial analysis indicated that IDU was a strong potential confounding variable, adjustment for IDU was performed for most other risk factors by including IDU and the variable in separate conditional logistic regression models. Linear trends were evaluated when appropriate, with numeric trend variables taking on increasing ordinal values for each level of the data they represented. The final set of independent HCV risk factors was derived from a stepwise conditional logistic regression model. A stepwise selection procedure was used to build a multivariate model that contained all the risk factors that were independently associated with HCV infection. In doing so, IDU and all risk factors that were significant (P.05) after adjusting for IDU were considered. These variables were introduced either as indicator variables ( yes vs. no, and if relevant, don t know vs. no for risk factors; predefined levels for education; and greater than the control median of the usual number of alcoholic drinks per week) or as a trend variable (e.g., number of sexual partners). Interactions between IDU and other significant risk factors were also tested. Only variables associated with HCV status at a significance level of P.05 were retained in the final model. RESULTS The questionnaire was mailed to 2,316 HCV donors and 2,316-seronegative controls, and was returned by 758 (33%) HCV and 1,039 (45%) control subjects (P.001, 2 test). Eighty-one percent of responders in both groups returned the questionnaire following the first mailing, and the rest after the second mailing. Response rates also differed (P.001, 2 test) by type of blood donation (apheresis 74%, whole blood 39%, directed 32%), first time (36%) versus repeat donor (46%), white or Asian (45%) versus black (25%) or Hispanic (31%) race/ethnicity, male (33%) versus female (48%), and age (21% for age less than 20 years versus 56% for age 60 years and higher). Response rates did not differ by blood center (P.3) or by date of last donation in a 6-month period (P.6). The demographic characteristics of the participating HCV and control donors are given in Table 1. Despite the differential response rates described above, the case and control groups had similar distributions of the matching variables (age, sex, race/ethnicity, blood center, and first-time versus repeat-donation status), date of donation, type of donation, marital status, and birthplace. However, HCV seropositives TABLE 1. Demographic Characteristics of HCV and Control Donors Who Returned Questionnaires HCV-Seropositive n (% of group) (N 758) Controls n (% of group) (N 1,039) Age* (yr) (2) 11 (1) (7) 70 (7) (40) 426 (41) (42) 389 (38) (7) 94 (9) (3) 37 (4) Sex* Male 400 (53) 568 (55) Female 351 (47) 463 (45) Race/ethnicity* White, non-hispanic 521 (69) 782 (75) Black, non-hispanic 78 (10) 100 (10) Hispanic 77 (10) 75 (7) Asian 16 (2) 20 (2) Other 13 (2) 20 (2) Marital status Married or living together 446 (60) 726 (71) Never married 118 (16) 136 (13) Divorced 128 (17) 125 (12) Separated 39 (5) 27 (3) Widowed 13 (2) 12 (1) Educational attainment Less than high school 89 (12) 33 (3) High school diploma 129 (17) 120 (12) Some college 328 (44) 280 (27) Associate/Bachelor or higher 201 (27) 594 (58) Alcohol intake Nondrinker 38 (5) 75 (7) Some 168 (24) 461 (46) Moderate-heavy 486 (70) 469 (47) Blood Center* Baltimore/Washington 122 (16) 180 (17) Southeast Michigan 155 (20) 221 (21) Oklahoma City 158 (21) 225 (22) San Francisco 79 (10) 109 (10) Los Angeles 244 (32) 304 (29) Donation date Jan-June (22) 261 (25) July-Dec (25) 253 (24) Jan-June (29) 264 (25) July-Dec (24) 261 (25) Donation history* First-time donor 529 (70) 702 (68) Repeat donor 229 (30) 337 (32) Donation type Whole blood 707 (93) 955 (92) Directed 46 (6) 69 (7) Apheresis/other 5 (1) 15 (1) Place of birth U.S. 684 (92) 943 (92) Non-U.S. 62 (8) 84 (8) NOTE. Numbers may not add up to total number of subjects because of missing values. Percentage may not exactly total 100 because of rounding error. *Matching variable. Sixty-six HCV and 34 controls did not answer this question. Alcohol categories for drinkers were based on median number of drinks/week reported by controls who drank alcohol. Some alcohol intake was defined as drinking 0, 1.6 drinks/week, with moderate to heavy drinking being defined as drinking more than 1.6 drinks/week.

3 758 MURPHY ET AL. HEPATOLOGY March 2000 had lower educational status (OR 0.36, 95% CI: for high school diploma; OR 0.38, 95% CI: for some college, and OR 0.10, 95% CI: for associate college degree or higher, with a reference group of less than high school diploma) and were heavier drinkers (OR 2.91, 95% CI: comparing alcohol intake of 2 or more drinks per week to nondrinkers). HCV seropositivity was most strongly associated with ever having injected illicit drugs, with 387(51%) HCV seropositives versus only 9 (1%) controls admitting to this risk factor (OR 134; 95% CI: ). Because of potential confounding by such a strong risk factor, OR for other potential HCV risk factors are reported both unadjusted and adjusted for IDU (Table 2). Among the HCV seropositives, 299 (84% of IDU) reported using needles that had been used previously by another person, compared with only 4 (44% of IDU) among controls (OR 2.8 for needle-sharing IDU; 95% CI: ). Living with an IDU and inhaling illicit drugs were also associated with HCV seropositivity, but OR decreased substantially after controlling for IDU, suggesting that these potential risk factors were subject to confounding. A history of blood transfusion was a clear risk factor for HCV in non-idu only, with an OR of 8.3 (95% CI: ) (Table 2). Only 10 cases and 1 control received blood after May 1990 when blood banks introduced HCV serological screening, so we were unable to analyze the risk of transfusion pre- and postscreening. Gamma globulin injection was associated with HCV in both IDU and non-idu (interaction term P.22), albeit with an IDU-adjusted OR of only 1.9 (95% CI: ). A history of having been stuck with a bloody needle or sharp instrument (not IDU-associated) was also a risk factor, with an adjusted OR of 3.2, whereas surgery, getting sutures, and nonspecific occupational blood exposure were weaker but significant risk factors, with an adjusted OR of 1.7. None of the other medical-related exposures had IDU-adjusted OR that were significant at P.05. Contrary to our a priori hypothesis, subjects who reported more lifetime episodes of teeth cleaning appeared to have a lower risk of HCV seropositivity, with a nonsignificant trend after adjustment for IDU. We also investigated minor parenteral and miscellaneous exposures in an effort to explain cases of HCV infection without major parenteral risk factors. Most of these variables had OR that decreased after adjustment for IDU, indicating potential confounding. Surprisingly, the strongest association with HCV was seen for a history of being in jail for more than 3 days (IDU-adjusted OR 5.0; 95% CI: ). The risk was stronger with shorter jail stays, with an adjusted OR of 3.9 (95% CI: ) for less than 1 year and 1.0 (95% CI: ) for jail stays of 1 year or more. The risk associated with tattooing (IDU-adjusted OR 3.9) was greater for multiple tattoos (IDU-adjusted OR 2.5 for 1 tattoo, OR 6.5 for 1 tattoo), and higher when tattooing was performed by a nonprofessional (IDU-adjusted OR 2.1 for paid a professional every time, OR 7.9 for sometimes, and OR 7.7 for never ). Piercing of ears or other body parts (OR 2.7), a history of taking part in any religious or ceremonial practices involving blood, needles, or knives (including blood-brother or -sister rituals) (OR 3.8), and a history of sharing a toothbrush or razor (OR 1.6) were associated with HCV after adjustment for IDU; acupuncture, electrolysis for hair removal, and ever having lived outside the United States were not significant after such adjustment. Although the unadjusted OR for ear piercing was higher for males than for females, there was no gender difference after adjusting for IDU. Sexual-behavior risk factors were also considered, with sexual contact defined as vaginal or anal penetration (Table 2). A history of sex with an IDU was a strong risk factor for HCV infection, even after adjustment for IDU (OR 9.7; 95% CI: ). Sex with a hepatitis case, sex with a transfusion recipient, and a history of a sexually transmitted disease were all associated with increased risk of HCV seropositivity after IDU adjustment. However, neither giving nor receiving money for sex was associated with HCV after adjustment for IDU. Among men, there was increased risk of HCV seropositivity with increasing numbers of female sexual partners (P trend.02), but not with an increasing number of male sexual partners (P trend.92) after adjustment for IDU. Men who admitted to sex with at least one other man were not at increased risk of HCV after adjustment for IDU (adjusted OR 1.0; 95% CI: ). Among women, there was a strong trend of increasing HCV risk with a higher number of lifetime male sexual partners (P trend.0001) and a slightly increased risk with more lifetime female sexual partners (P.02) after adjustment for IDU. Women who admitted to sex with at least one other woman were twice as likely to be HCV-seropositive than exclusively heterosexual women after controlling for IDU. Because of the potential for confounding by IDU and other strong risk factors, we obtained the ORs for all risk factors that were significantly associated with HCV after adjustment for IDU and other risk factors in a final conditional logistic regression model (Table 3). IDU remained the strongest independent risk for HCV seropositivity, with an OR Furthermore, clearly associated with HCV were previous blood transfusion in non-idu only (OR 10.9) and sex with an IDU (OR 6.3). Still significant, but with lower OR, were having been in jail for more than 3 days (OR 2.9), religious scarification or blood-brother ritual (OR 2.8), having been stuck with a bloody needle or sharp instrument (OR 2.1), having pierced ears or other body parts (OR 2.0), and ever receiving immunoglobulin injection (OR 1.6). Lower educational attainment and higher alcohol intake were significantly associated with HCV, and were therefore included in the final model. Other variables that had significant unadjusted or IDU-adjusted OR in Table 2, including inhaling drugs, living with an IDU, and sexual behaviors other than sex with an IDU, did not remain in the final model. Transfusion was the only variable found to be a risk factor only for non-idu. However, we were not able to properly test some interaction terms (e.g., IDU/religious scarification, IDU/pierced body parts) because of low subgroup sizes. It is therefore possible that some of the other risk factors may have different strengths in IDU and non-idu. Ninety percent of HCV seropositives had at least 1 of the 8 independent risk factors for HCV in our final model (Fig. 1), as compared with only 55% of controls. Thus, only 10% of HCV seropositive blood donors did not have at least 1 of these risk factors for HCV infection. DISCUSSION This large, well-controlled case-control study of risk factors for HCV seropositivity has demonstrated that IDU is the strongest risk factor for HCV seropositivity, even among persons donating blood. In contrast to another study, 19 we did

4 HEPATOLOGY Vol. 31, No. 3, 2000 MURPHY ET AL. 759 TABLE 2. Risk Factors for HCV Seropositivity, Both Unadjusted and Adjusted for IDU, Among United States Blood Donors at 5 REDS Blood Centers, Percent of Cases (N 758) Percent of Controls (N 1,039) Unadjusted IDU-Adjusted Drug-related Injected drugs (68-268) Lived with an IDU ( ) 5.1 ( ) Inhaled drugs ( ) 2.2 ( ) Transfusion/Medical Received blood transfusion ( ) non-idu: 8.3 ( ) IDU: 0.9 ( ) Gamma globulin injection ( ) 1.9 ( ) Bloody needle stick injury ( ) 3.2 ( ) Had surgery ( ) 1.7 ( ) Had sutures ( ) 1.7 ( ) Tooth extraction ( ) 1.3 ( ) Teeth cleaned Never times ( ) 0.7 ( ) times ( ) 0.5 ( ) 20 times ( ) 0.6 ( ) (P trend.0003) (P trend.06) Occupational blood exposure ( ) 1.7 ( ) Lived with hepatitis case ( ) 1.4 ( ) Relative with hepatitis ( ) 0.9 ( ) Lived with transfusion recipient ( ) 1.5 ( ) Parenteral/miscellaneous exposures Stuck/cut with bloody object ( ) 2.5 ( ) Bloody religious ritual ( ) 3.8 ( ) Shared toothbrush/razor ( ) 1.6 ( ) Been tattooed ( ) 3.9 ( ) Pierced ears/body parts ( ) 2.7 ( ) Acupuncture ( ) 1.4 ( ) Electrolysis hair removal ( ) 1.2 ( ) In jail more than 3 days ( ) 5.0 ( ) Lived outside the U.S ( ) 1.0 ( ) Sexual exposures (both sexes) Sex with an IDU ( ) 9.7 ( ) Sex with hepatitis case ( ) 2.2 ( ) Sex with transfusion recipient ( ) 2.5 ( ) Gave money for sex ( ) 1.5 ( ) Received money for sex ( ) 3.0 ( ) Sexually transmitted disease ( ) 2.5 ( ) Men only* Number of lifetime female partners: None ( ) 0.8 ( ) ( ) 1.5 ( ) 11 or more ( ) 1.7 ( ) (P trend.0001) (P trend.02) Number of lifetime male partners: None ( ) 0.9 ( ) 2 or more ( ) 1.1 ( ) (P trend.0001) (P trend.92) Sexual orientation Heterosexual Bisexual/homosexual ( ) 1.0 ( ) Not sexually active ( ) 0.7 ( ) Women only Number of lifetime male partners: None ( ) 1.2 ( ) ( ) 1.9 ( ) ( ) 3.2 ( ) 50 or more ( ) 8.8 ( ) (P trend.0001) (P trend.0001)

5 760 MURPHY ET AL. HEPATOLOGY March 2000 TABLE 2. Continued Percent of Cases (N 758) Percent of Controls (N 1,039) Unadjusted IDU-Adjusted Number of lifetime female partners: None ( ) 1.3 ( ) ( ) 3.7 ( ) (P trend.0001) (P trend.02) Sexual orientation Heterosexual Bisexual/homosexual ( ) 2.3 ( ) Not sexually active ( ) 0.9 ( ) NOTE. The variables are derived from questions asking about lifetime history for these risk factors. Percentages of HCV cases and seronegative controls responding affirmatively to each question are given; percentages may not add to 100 as a result of missing responses. not find that drug inhalation was a risk factor. Our data also provide evidence for 4 less-recognized HCV risk factors, including incarceration, being stuck with a bloody instrument (non-idu related), pierced ears and other body parts, and immunoglobulin injection, but did not find that a number of routine medical and dental procedures carried a risk of HCV infection in the United States. With the exception of sex with an IDU, the data suggest that HCV transmission may be limited to parenteral risk behaviors. Although the presence of a risk factor does not indicate that it was the mechanism of infection, 90% of HCV seropositives had 1 of our 8 independent risk factors, suggesting that the proportion of occult HCV infection may be as low as 10%. The finding that IDU was a strong risk factor for HCV infection among United States blood donors was not unexpected. First, up to 90% of American IDU are HCVseropositive, and the incidence of HCV infection is approximately 10 per 100 person-years among IDU in Baltimore. 11,12,23 Second, although persons who admit a history of drug injection are excluded from blood donation, one study showed that up to 1.1% of those accepted for blood donation admitted a history of IDU on a postdonation anonymous questionnaire. 21 Finally, smaller studies have implicated IDU as a risk factor for HCV seropositivity among blood donors in Australia, 24 the United Kingdom, 25 and the United States. 19 However, in contrast to one of these studies, 19 we did not find that intranasal drug inhalation was an independent risk factor for HCV. Confounding by unreported IDU and 1 or more of our independent risk factors may have inflated the previous study s multivariate OR of 8.0, because our unadjusted OR of 9.1 for drug inhalation fell to 2.2 after adjustment for IDU, and was no longer significant (OR 0.8) after adjusting for IDU and the other variables in our final model. Our data would suggest that excluding potential blood donors who only admit to intranasal inhalation of drugs may not be warranted. Our findings are also of relevance to potential iatrogenic sources of HCV infection. The association of blood transfusion with HCV seropositivity in non-idu was not surprising given previous literature implicating HCV as the primary cause of non-a, non-b transfusion-transmitted hepatitis. 26,27 Although intravenous immunoglobulin has previously been implicated in the transmission of HCV, 28 intramuscular immunoglobulin products have been thought to be free from infectious HCV. 29 Therefore, our finding that gamma globulin injection was a risk factor for HCV infection must be interpreted carefully and confirmed by other studies. That injuries from bloody needles or sharp instruments may transmit HCV is both biologically plausible and consistent with studies of HCV among medical personnel with such injuries. 30 Finally, the lack of an independent association with previous surgery, suturing, or dental procedures in this U.S. study is reassuring given reports of iatrogenic HCV infection TABLE 3. Final Multivariable Logistic Regression Model of Independent Risk Factors for HCV Seropositivity Among United States Blood Donors at the 5 REDS Blood Centers, * IDU 49.6 ( ) Blood transfusion Non-IDU 10.9 ( ) IDU 1.0 ( ) Sex with IDU 6.3 ( ) In jail more than 3 days 2.9 ( ) Religious scarification 2.8 ( ) Stuck with a bloody object 2.1 ( ) Pierced ears/body parts 2.0 ( ) Gamma globulin injection 1.6 ( ) NOTE. The variables are derived from questions asking lifetime history of these risk factors. *Adjusted for education, alcohol intake, and sex with someone with hepatitis. FIG. 1. The proportion of HCV-seropositive subjects with each of the independent risk factors found by this study. Subjects with more than 1 risk factor were classified as having the risk factor with the strongest OR, according to the hierarchy presented in Table 3. A total of 90% of HCV-seropositive subjects had at least 1 of these HCV risk factors.

6 HEPATOLOGY Vol. 31, No. 3, 2000 MURPHY ET AL. 761 in other countries. The association between HCV and a lower frequency of teeth cleaning is probably the result of unmeasured confounding by lower socioeconomic status. With the exception of sex with an IDU, these data do not support the hypothesis that sexual behavior is a strong risk factor for HCV seropositivity. Subjects giving money or receiving money for sex were not at increased risk, and there was no independent association of HCV seropositivity with higher numbers of male or female sexual partners, nor with being the sexual partner of a transfusion recipient or hepatitis case. Although sexual partners of IDU were at significantly increased risk of HCV regardless of their own IDU status (OR 2.4 for IDU and OR 7.4 for non-idu; interaction P value.2), such an association could be the result of actual sexual transmission or could possibly result from other shared parenteral risk factors. If HCV is poorly transmitted sexually, the low risk associated with multiple sexual partners from the general population might be missed in crosssectional or case-control studies. On the other hand, unreported IDU or minor parenteral exposures might also be more common among persons in sexual relationships with IDU, 31 which would tend to confound the hypothesis of sexual transmission. A definitive answer to the sexual-transmission question must therefore await prospective studies of discordant couples with good control for parenteral exposures. We also identified 3 less recognized risk factors for HCV seropositivity, including having been in jail, religious scarification, and having pierced ears or body parts. Even though this variable remained significant in the multivariate model, the association of HCV with having been in jail for more than 3 days is most likely to be a surrogate for other transmission modes, such as IDU or amateur tattooing among prisoners. Other studies have demonstrated incidence rates of 1 per 100 person-years in American prisoners 32 and 38 per 100 personyears in Australian prisoners, 33 and British IDU who had been in prison had almost twice the HCV seroprevalence of those who were never incarcerated. 34 Associations of HCV infection with ear and body-part piercing have been reported by some studies in males only, 19,35 but not by others. 36,37 We were able to find only one other study that found scarification or religious rituals involving blood to be associated with HCV infection. 38 We did not find an association between HCV and tattooing after controlling for IDU and other behaviors, although other studies have found tattooing to be a risk factor. 24,25,39,40 Although this was a large, well-matched case-control study, its conclusions must be tempered by its potential limitations. The low and differential (between cases and controls) response rate raises the possibility of response bias, which may have led to under- or overestimation of the magnitude of the associations observed. Nevertheless, analyzed cases and controls were similar in most characteristics, with the exception of educational level and alcohol intake. Although we controlled for these two variables in our analysis, it is possible that residual confounding remained. Whereas the large OR for IDU, previous blood transfusion, and sex with IDU are unlikely to be entirely caused by response bias or confounding, the weaker associations must be interpreted cautiously and replicated in other studies. Because of the number of potential risk factors considered, some of the associations we report may have been the result of chance, although most described parenteral infection routes that are biologically plausible. Finally, the blood-donor sampling frame of the study yielded data that are directly applicable to blood safety, but the conclusions must be extrapolated to the general population with caution. In conclusion, IDU is by far the strongest and most common risk factor for HCV seropositivity among prospective United States blood donors. This study adds further evidence that interview procedures designed to prevent blood donation by this risk group are not foolproof, and may be amenable to improvement based on behavioral science. 21 In addition, questions regarding incarceration, scarification, and body piercing may need to be developed or revised. With the exception of sex with an IDU, we did not find evidence to support the hypothesis that HCV is sexually transmitted. Our finding that 90% of the HCV-seropositive donors had at least 1 of the risk factors independently associated with HCV in our final model suggests that searches for additional minor modes of HCV transmission are likely to be unsuccessful. APPENDIX REDS is presently the responsibility of the following persons: Blood Centers: American Red Cross Blood Services Greater Chesapeake and Potomac Region: A. E. Williams (Holland Laboratory), C. C. Nass. American Red Cross Blood Services Southeastern Michigan Region: M. J. Higgins, J. Campbell. American Red Cross Blood Services Southern California Region: G. Garratty, S. Hutching. University of California San Francisco/Blood Centers of the Pacific: E. L. Murphy, M. P. Busch. Oklahoma Blood Institute: R. O. Gilcher, J. W. Smith. Medical Coordinating Center: Westat: G. B. Schreiber, R. A. Thomson. National Heart, Lung, and Blood Institute, NIH: G. J. Nemo Steering Committee Chairman: T. F. Zuck (Hoxworth Blood Center) REFERENCES 1. Benjelloun S, Bahbouhi B, Sekkat S, Bennani A, Hda N, Benslimane A. Anti-HCV seroprevalence and risk factors of hepatitis C virus infection in Moroccan population groups. Res Virol 1996;147: Robinson JW, Rosas M, Guzman F, Patarroyo ME, Moreno A. Comparison of prevalence of anti-hepatitis C virus antibodies in differing South American populations. J Med Virol 1996;50: Alter MJ, Mast EE, Moyer LA, Margolis HS. Hepatitis C. Infect Dis Clin North Am 1998;12: Guadagnino V, Stroffolini T, Rapicetta M, Costantino A, Kondili LA, Menniti-Ippolito F, Caroleo B, et al. Prevalence, risk factors, and genotype distribution of hepatitis C virus infection in the general population: a community-based survey in southern Italy. HEPATOLOGY 1997;26: Abdel-Wahab MF, Zakaria S, Kamel M, Abdel-Khaliq MK, Mabrouk MA, Salama H, Esmat G, et al. High seroprevalence of hepatitis C infection among risk groups in Egypt. Am J Trop Med Hyg 1994;51: Esteban JI, Lopez-Talavera JC, Genesca J, Madoz P, Viladomiu L, Muaniz E, Martin-Vega C, et al. High rate of infectivity and liver disease in blood donors with antibodies to hepatitis C virus [Comments]. Ann Intern Med 1991;115: Lee SD, Chan CY, Wang YJ, Wu JC, Lai KH, Tsai YT, Lo KJ. Seroepidemiology of hepatitis C virus infection in Taiwan. HEPATOLOGY 1991;13: Lai ME, Mazzoleni AP, Farci P, Melis A, Porru A, Orgiana G, Arnone M, et al. Markers of hepatitis C virus infection in Sardinian blood donors:

7 762 MURPHY ET AL. HEPATOLOGY March 2000 relationship with alanine aminotransferase levels. J Med Virol 1993;41: Archer GT, Buring ML, Clark B, Ismay SL, Kenrick KG, Purusothaman K, Kaldor JM, et al. Prevalence of hepatitis C virus antibodies in Sydney blood donors. Med J Aust 1992;157: Khan M, Husain M, Yano M, Hashizume K, Yousuf M, Tanaka E, Matsumotot A, et al. Comparison of seroepidemiology of hepatitis C in blood donors between Bangladesh and Japan. Gastroenterol Jpn 1993; 28(Suppl 5): Osmond DH, Padian NS, Sheppard HW, Glass S, Shiboski SC, Reingold A. Risk factors for hepatitis C virus seropositivity in heterosexual couples. JAMA 1993;269: Kelen GD, Green GB, Purcell RH, Chan DW, Qaquish BF, Sivertson KT, Quinn TC. Hepatitis B and hepatitis C in emergency department patients. N Engl J Med 1992;326: Alter HJ. To C or not to C: these are the questions. Blood 1995;85: Zeuzem S, Teuber G, Lee JH, Reuster B, Roth WK. Risk factors for the transmission of hepatitis C. J Hepatol 1996;24(Suppl 2): Osmond DH, Padian NS, Sheppard HW, Glass S, Shiboski SC, Reingold A. Risk factors for hepatitis C virus seropositivity in heterosexual couples. JAMA 1993;269: Nakashima K, Kashiwagi S, Hayashi J, Noguchi A, Hirata M, Kajiyama W, Urabe K, et al. Sexual transmission of hepatitis C virus among female prostitutes and patients with sexually transmitted diseases in Fukuoka, Kyushu, Japan. Am J Epidemiol 1992;136: Weinstock HS, Bolan G, Reingold AI, Polish LB. Hepatitis C virus infection among patients attending a clinic for sexually transmitted diseases. JAMA 1993;269: MacDonald M, Crofts N, Kaldor J. Transmission of hepatitis C virus: rates routes and cofactors. Am J Epidemiol 1996;18: Conry-Cantilena C, VanRaden M, Gibble J, Melpolder J, Shakil AO, Viladomiu L, Cheung L, et al. Routes of infection, viremia and liver disease in blood donors found to have hepatitis C virus infection. N Engl J Med 1996;334: Murphy EL, Bryzman S, Williams AE, Co-Chien H, Schreiber GB, Ownby HE, Gilcher RO, et al. Demographic determinants of hepatitis C virus seroprevalence among blood donors. JAMA 1996;275: Williams AE, Thomson R, Schreiber G, Watanabe K, Bethel J, Lo A, Kleinman SH, et al. Estimates of infectious disease risk factors in active U.S. blood donors. JAMA 1997;277: SAS Technical Report, SAS/STAT Software: the PHREG Procedure, version 6, Cary, NC, SAS Institute Inc., 1991: Donahue JG, Nelson KE, Munoz A, Vlahov D, Rennie LL, Taylor EL, Saah AJ, et al. Antibody to hepatitis C virus among cardiac surgery patients, homosexual men, and intravenous drug users in Baltimore, Maryland. Am J Epidemiol 1991;134: Kaldor JM, Archer GT, Buring ML, Ismay SL, Kenrick KG, Lien AS, Purusothaman K, et al. Risk factors for hepatitis C virus infection in blood donors: a case-control study. Med J Aust 1992;157: Neal KR, Jones DA, Killey D, James V. Risk factors for hepatitis C virus infection. A case-control study of blood donors in the Trent region (UK). Epidemiol Infect 1994;112: Alter HJ, Purcell RH, Shih JW, Melpolder JC, Houghton M, Choo QL, Kuo G. Detection of antibody to hepatitis C virus in prospectively followed transfusion recipients with acute and chronic non-a, non-b hepatitis. N Engl J Med 1989;321: Aach RD, Stevens CD, Hollinger FB, Mosley JW, Peterson DA, Taylor PE, Johnson RG, et al. Hepatitis C virus infection in post-transfusion hepatitis: an analysis with first and second generation assays. N Engl J Med 1991;325: Healey CJ, Sabharwal NK, Daub J, Davidson F, Yap PL, Fleming KA, Chapman RW, et al. Outbreak of acute hepatitis C following the use of anti-hepatitis C virus screened intravenous immunoglobulin therapy [Comments]. Gastroenterology 1996;110: Cristiano K, Pisani G, Wirz M, Pini C, Gentili G. More evidence on safety of intramuscular immune serum globulin produced from plasma unscreened for anti-hepatitis C virus antibodies [Letter]. Arch Intern Med 1998;158: Kiyosawa K, Sodeyama T, Tanaka E, Nakano Y, Furuta S, Nishioka K, Percell RH, et al. Hepatitis C in hospital employees with needle stick injuries. Ann Intern Med 1991;115: Tor J, Llibre JM, Carbonell M, Muga R, Ribera A, Soriano V, Clotet B, et al. Sexual transmission of hepatitis C virus and its relation with hepatitis B virus and HIV. Br Med J 1990;301: Vlahov D, Nelson KE, Quinn TC, Kendig N. Prevalence and incidence of hepatitis C infection among male prison inmates in Maryland. Eur J Epidemiol 1993;9: Crofts N, Stewart T, Hearne P, Ping XY, Breshkin AM, Locarnini SA. Spread of bloodborne viruses among Australian prison entrants. Br Med J 1995;310: McBride AJ, Ali IM, Clee W. Hepatitis C and injecting drug use in prisons. Br Med J 1994;309: MacLennan S, Moore MC, Hewitt PE, Nicholas S, Barbara JA. A study of anti-hepatitis C positive blood donors: the first year of screening. Transfusion Med 1994;4: Kim YS, Ahn YO, Kim DW. A case-control study on the risk factors of hepatitis C virus infection among Koreans. J Korean Med Sci 1996;11: Chen TZ, Wu JC, Yen FS, Sheng WY, Hwang SJ, Huo TI, Lee SD. Injection with nondisposable needles as an important route for transmission of acute community-acquired hepatitis C virus infection in Taiwan. J Med Virol 1995;46: Kiyosawa K, Tanaka E, Sodeyama T, Yoshizawa K, Yabu K, Furuta K, Imai H, et al. Transmission of hepatitis C in an isolated area in Japan: community-acquired infection. The South Kiso Hepatitis Study Group. Gastroenterology 1994;106: Abilgaard N, Peterslund NA. Hepatitis C transmitted by tattooing needle [Letter]. Lancet 1991;338: Ko YC, Ho MS, Chiang TA, Chang SJ, Chang PY. Tattooing as a risk of hepatitis C virus infection. J Med Virol 1992;38: