Article first published online: 27 AUG 2012
Copyright © 2012 American Association for the Study of Liver Diseases
Volume 56, Issue 5, pages 1699–1705, November 2012
How to Cite
Sarkar, M., Bacchetti, P., French, A. L., Tien, P., Glesby, M. J., Nowicki, M., Plankey, M., Gange, S., Sharp, G., Minkoff, H., Peters, M. G. and for the Women's Interagency HIV Study (WIHS) (2012), Lower liver-related death in African-American women with human immunodeficiency virus/hepatitis C virus coinfection, compared to Caucasian and Hispanic women. Hepatology, 56: 1699–1705. doi: 10.1002/hep.25859
Potential conflict of interest: Nothing to report.
The Women's Interagency HIV Study is funded by the National Institute of Allergy and Infectious Diseases (UO1-AI-35004, UO1-AI-31834, UO1-AI-34994, UO1-AI-34989, UO1-AI-34993, and UO1-AI-42590) and by the National Institute of Child Health and Human Development (UO1-HD-32632). The study is cofunded by the National Cancer Institute, the National Institute on Drug Abuse, and the National Institute on Deafness and Other Communication Disorders. Additional support was received through the National Institutes of Health (T32 DK060414; to M.S.).
- Issue published online: 31 OCT 2012
- Article first published online: 27 AUG 2012
- Accepted manuscript online: 22 MAY 2012 11:24AM EST
- Manuscript Accepted: 15 MAY 2012
- Manuscript Received: 14 MAR 2012
Among individuals with and without concurrent human immunodeficiency virus (HIV), racial/ethnic differences in the natural history of hepatitis C virus (HCV) have been described. African Americans have lower spontaneous HCV clearance than Caucasians, yet slower rates of liver fibrosis once chronically infected. It is not clear how these differences in the natural history of hepatitis C affect mortality, in either HIV-positive or -negative individuals. We conducted a cohort study of HIV/HCV coinfected women followed in the multicenter Women's Interagency HIV Study to determine the association of self-reported race/ethnicity with all-cause and liver-related mortality. Survival analyses were performed using Cox's proportional hazards models. The eligible cohort (n = 794) included 140 Caucasians, 159 Hispanics, and 495 African Americans. There were 438 deaths and 49 liver-related deaths during a median follow-up of 8.9 years and maximum follow-up of 16 years. African-American coinfected women had significantly lower liver-related mortality, compared to Caucasian (hazard ratio [HR], 0.41; 95% confidence interval [CI]: 0.19-0.88; P = 0.022) and Hispanic coinfected women (HR, 0.38; 95% CI: 0.19-0.76; P = 0.006). All-cause mortality was similar between racial/ethnic groups (HRs for all comparisons: 0.82-1.03; log-rank test: P = 0.8). Conclusions: African-American coinfected women were much less likely to die from liver disease, as compared to Caucasians and Hispanics, independent of other causes of death. Future studies are needed to investigate the reasons for this marked racial/ethnic discrepancy in liver-related mortality. (HEPATOLOGY 2012;56:1699–1705)
In the United States, at least 5 million people are infected with hepatitis C, 80% of whom are estimated to be viremic.1, 2 Because of shared modes of transmission, hepatitis C virus (HCV) infection in patients with human immunodeficiency virus (HIV) is common. Approximately one third of HIV-positive patients are coinfected with HCV, and this number approaches 80% among those with injection drug use (IDU).3 Among patients with HIV, hepatitis C–related liver disease remains the second leading cause of death.4 Data projecting liver-related mortality also indicate that deaths from hepatitis C are continuing to rise and remain an important cause of premature mortality.5-7
Racial/ethnic differences in the natural history of hepatitis C have been well described. Spontaneous HCV clearance is lower among African Americans, compared to Caucasians and Hispanics, yet African Americans appear to develop less fibrosis and inflammation once chronically infected, compared to other racial/ethnic groups.8-10 It remains unclear how these racial differences in the natural history of hepatitis C may affect liver-related death, in either HCV monoinfection or in HIV/HCV-coinfected individuals.
Data on race and mortality in patients with chronic HCV infection are conflicting. Several recent studies have reported higher death rates among Caucasians with chronic hepatitis C (CHC), compared to African Americans, although one large study observed opposite trends.11-14 Most studies have also focused on male predominant cohorts, with limited available data on race and mortality among women with HCV or HIV/HCV coinfection.
Given conflicting previous data on race and mortality, and the absence of robust data in either female or HIV-infected populations, we aimed to determine the association between race/ethnicity and liver-related death in a large cohort of HIV/HCV-coinfected women followed in the Women's Interagency HIV Study (WIHS). The rich ethnic diversity and long-term follow-up of WIHS participants has allowed us to address these important racial and ethnic distinctions.
Patients and Methods
We conducted a cohort study of women participating in the WIHS. The WIHS is a National Institutes of Health (NIH)-funded, prospective, multicenter cohort of women at risk for, or currently diagnosed with, HIV. Enrollment in the WIHS took place in two study cohorts: the first in 1994-1995 and the second in 2001-2002.15 Women in the WIHS are seen twice-yearly and undergo detailed histories, physical exams, structured interviews, and laboratory testing. This study was approved by the WIHS Executive Committee and the institutional review boards at the six participating WIHS study sites. Study eligibility included HIV/HCV coinfection at WIHS study entry, as defined by detectable HCV RNA, HCV antibody (Ab), and positive HIV western blotting. Because of the small number of women who self-reported as other than Caucasian, Hispanic, or African American, these individuals were excluded from the analysis.
Predictor and Outcome Measures.
The primary predictor was race/ethnicity determined by self-report at WIHS entry visit. We defined the African-American group as non-Hispanic African Americans. The Caucasian group was defined as non-Hispanic Caucasians. The Hispanic group was defined as Hispanic Caucasians, Hispanic African Americans, and other Hispanics. Our outcome of interest was primary liver-related death, as determined by death certificate verification. We also reported data on all-cause mortality. Primary death certificate data were reviewed by two clinicians to determine cause of death. In some cases, there was supplemental information from medical record review, communication with the primary clinician, or patient families. Primary liver-related deaths included those resulting from hepatic decompensation or hepatocellular carcinoma, although most death certificates simply recorded “hepatitis C” as the primary cause of liver death, without further specification.
The following covariates were included in our survival analysis: age; substance abuse history, including intravenous (IV) drugs, non-IV drugs, tobacco, and alcohol; HIV-related factors, such as HIV RNA levels, CD4 count, and highly active antiretroviral therapy (HAART); liver-related factors, including HCV RNA levels, HCV genotype, HCV treatment history, and chronic hepatitis B virus (HBV); and comorbid factors, such as diabetes mellitus (DM), hypertension (HTN), body mass index (BMI), glomerular filtration rate (GFR), and cancer history.
Plasma HIV RNA levels were measured using the NASBA/NuciSens HIV RNA assay (bioMerieux, Durham, NC) in laboratories certified by the NIH National Institute of Allergy and Infectious Diseases Virology Quality Assurance Certification Program. HCV and HBV serologies were performed using standard commercial assays and included hepatitis C Ab by EIA 3.0 (Ortho Clinical Diagnostics, Raritan, NJ) as well as hepatitis B surface antigen (HBsAg) (Abbott Laboratories, Abbott Park, IL). HCV RNA levels were measured by the COBAS Amplicor Monitor 2.0 assay (Roche Diagnostics, Branchburg, NJ), with a linear range of 600-700,000 IU/mL, or COBAS Taqman (Roche Diagnostics), with a linear range of 10-2.0 × 108 IU/mL.
Patient characteristics were compared using chi-square tests, t tests, and Kruskal-Wallis' tests, when appropriate. Cox's regression models were used to calculate the hazards ratios (HRs) and 95% confidence intervals (CIs) for factors associated with all-cause and liver-related mortality. All survival analyses used age as the time scale, with age at study entry treated as a left-truncation time to reflect the fact that only living women could be enrolled. This automatically accounts for the important influence of age on mortality risk and is a more biologically meaningful time scale than time since study enrolment. All variables that were measured repeatedly were analyzed as time-varying covariates, with the most recent value carried forward until a new measurement was made. The only non-time-varying covariates were race/ethnicity and HCV genotype, because these factors do not change over time. The final multivariate model was developed using forward selection of covariates, as well as inclusion of covariates with high biological plausibility of an association with death. A Fine-Gray competing-risks analysis,16 as well as a survival analysis of non-liver-related death, was performed to assess the possibility that racial/ethnic differences in liver-related death could be attributed to a differential risk of non-liver-related death among racial/ethnic groups. All analyses were performed using STATA software (version 11.0; StataCorp LP, College Station, TX).
We identified 794 women in the WIHS with confirmed CHC and HIV infection. Of these, 62.3% (495 of 794) were African American, 20% were Hispanic (159 of 794), and 17.7% (140 of 794) were Caucasian. Women were followed for up to 16 years, with a median follow-up of 8.9 years. The median follow-up for Caucasians, African Americans, and Hispanics was 9.0, 8.7, and 9.2 years, respectively. During this time, there were 438 deaths from all causes, including 49 liver-related deaths. Among primary causes of death, HIV/AIDS (36.9%) was the most common cause of death, followed by liver-related disease (11.2%) and homicides, suicides, and accidents (9.0%). Approximately 55.8% (276 of 495) of African Americans, 52.2% (83 of 159) of Hispanics, and 56.4% (79 of 140) of Caucasians died during follow-up. Liver disease was the primary cause of death in 7.6% (21 of 276) of African Americans, 20.5% (17 of 83) of Hispanics, and 13.9% (11 of 79) of Caucasians.
Compared to Hispanics and Caucasians, African-American women were older at study entry and time of death. They were also more likely to have heavy alcohol and tobacco use, to be diagnosed with hypertension, and to be infected with HCV genotype 1. Compared to Hispanics and African Americans, Caucasian women were more likely to use IV drugs. There were no significant differences in median CD4 count, median HIV viral load, or HAART use between racial/ethnic groups. HCV treatment was uncommon, with similar percentages between racial/ethnic groups (Table 1).
|Variable||N||All (n = 794)||Caucasian (n = 140)||Hispanic (n = 159)||African American (n = 495)||P Value|
|Mean age at entry (± SD)||794||39.7 (6.3)||37.4 (6.2)||37.5 (6.1)||41.0 (6.0)||<0.0001|
|Mean age at death (± SD)||438||46.1 (7.4)||43.6 (7.3)||44.5 (7.6)||47.3 (7.2)||<0.0001|
|Tobacco use at entry, %||794||90.8||86.4||88.1||92.9||0.0300|
|Ongoing tobacco use, %||794||85.3||85.0||81.1||86.7||0.2300|
|IDU at entry, %||794||84.1||88.6||82.4||83.4||0.2700|
|Ongoing IDU, %||794||34.4||45.7||32.1||31.9||0.0100|
|Non IDU at entry, %||792||89.9||95.0||88.7||88.9||0.0900|
|Ongoing non-IDU, %||794||68.6||71.4||66.0||68.7||0.6100|
|Alcohol use at entry, %||774||13.5||11.5||10.9||15.4||0.2400|
|Ongoing alcohol use, %||791||31.1||23.7||27.9||34.2||0.0400|
|Median CD4 at entry (IQR)||771||339 (179-545)||356 (208-584)||347 (191-551)||332 (172-516)||0.2100|
|Median nadir CD4 during study (IQR)||787||145 (42-257)||160 (65-271)||130 (42-289)||137 (32-245)||0.3900|
|Last available CD4 (IQR)||787||280 (96-509)||295 (103-499)||294 (93-502)||274 (94-510)||0.7600|
|Median log HIV RNA at entry (IQR)||773||4.3 (3.5-5.0)||4.4 (3.6-5.0)||4.0 (3.1-4.8)||4.3 (3.6-5.1)||0.3500|
|Median log last available HIV RNA (IQR)||793||3.4 (1.9-4.9)||3.0 (1.9-4.8)||3.2 (1.9-4.7)||3.5 (1.9-5.0)||0.6100|
|HAART use during study, %||794||69.9||71.4||71.7||68.9||0.7300|
|Median log HCV RNA at entry (IQR)||782||6.3 (5.8-6.7)||6.3 (5.8-6.7)||6.4 (5.8-6.7)||6.3 (5.8-6.6)||0.8700|
|Median log last available HCV RNA (IQR)||794||6.4 (5.9-6.8)||6.3 (5.7-6.7)||6.5 (6.0-6.8)||6.4 (5.9-6.9)||0.1900|
|HCV treatment during study, %||399||7.0||5.8||7.5||7.2||0.9100|
|HCV genotype 1, %||568||87.9||73.2||79.3||94.7||<0.0001|
|Chronic hepatitis B, %||794||2.4||2.9||2.5||2.2||0.9000|
|Diabetes history, %||794||19.9||13.6||18.9||22.0||0.0800|
|HTN history, %||794||49.8||42.9||39||55.2||<0.0001|
|Median BMI at entry (IQR)||745||25 (22-29)||24 (22-27)||25 (23-29)||25 (22-29)||0.2100|
|Median GFR at entry (IQR)||788||91.8 (75-108)||90.6 (73.3-106)||84.7 (74.5-104)||93.3 (77.4-109.3)||0.0300|
|Cancer history, %||794||3.3||4.3||3.1||3.0||0.7600|
We first analyzed racial/ethnic differences in all-cause mortality, which was similar between racial/ethnic groups on both uni- and multivariate models (HRs for all comparisons: 0.82-1.03; log-rank test: P = 0.8). These results contrasted markedly from our analysis of liver-related mortality. On univariate analysis, there was a trend toward lower risk of liver-related death among African Americans, compared to Caucasians (HR, 0.48; 95% CI: 0.23-1.03; P = 0.058) and a statistically significantly lower risk of liver-related death among African Americans, compared to Hispanics (HR, 0.44; 95% CI: 0.22-0.87; P = 0.019). On multivariate analysis, adjusted for CD4 count and HIV RNA levels, we observed even stronger racial/ethnic differences in liver-related mortality, with an HR comparing African-American to Caucasian women of 0.41 (95% CI: 0.19-0.88; P = 0.022) and an HR comparing African-American to Hispanic women of 0.38 (95% CI: 0.19-0.76; P = 0.006). Liver-related mortality was similar between Caucasian and Hispanic coinfected women on uni- and multivariate analyses (Table 2). These associations are demonstrated graphically in an age-adjusted survival curve, with a log-rank test of P = 0.032 (Fig. 1).
|Deaths From All Causes (n = 438)||Liver-Related Deaths (n = 49)|
|Racial/ethnic comparison||Unadjusteda HR (95% CI)||P Value||Adjusted HRb (95% CI)||P Value||Unadjusteda HR (95% CI)||P Value||Adjusted HRc (95% CI)||P Value|
|African American||0.92 (0.70-1.22)||0.58||0.82 (0.61-1.09)||0.17||0.48 (0.23-1.03)||0.058||0.41 (0.19-0.88)||0.022|
|Hispanic||0.89 (0.64-1.25)||0.51||0.90 (0.64-1.28)||0.56||1.11 (0.50-2.4)||0.79||1.09 (0.49-2.4)||0.83|
|Hispanic (Ref) versus African American||1.03 (0.79-1.35)||0.81||0.91 (0.69-1.19)||0.48||0.44 (0.22-0.87)||0.020||0.38 (0.19-0.76)||0.006|
Additional factors associated with liver-related mortality on univariate analysis included CD4 count (HR, 0.78 per 2-fold increase; 95% CI: 0.67-0.91; P = 0.002) and HIV RNA level (HR, 1.47 per 10-fold increase; 95% CI: 1.17.0-1.9; P = 0.001), although there was no significant association with HCV RNA levels (HR, 1.44; 95% CI: 0.94-2.2; P = 0.09). On multivariate analysis, HIV RNA levels were associated with liver-related death (HR, 1.44; 95% CI: 1.11-1.9; P = 0.006) (Table 3).
|Univariate Analysisa||Multivariate Analysisb|
|Variable||HR (95% CI)||P Value||HR (95% CI)||P Value|
|Tobacco use||1.15 (0.59-2.2)||0.680||1.0 (0.51-2.0)||0.990|
|IDU||1.0 (0.39-2.6)||0.990||0.72 (0.27-1.9)||0.510|
|Non-IDU||0.81 (0.43-1.6)||0.540||0.64 (0.33-1.26)||0.200|
|Alcohol use||0.64 (0.15-2.6)||0.530||0.53 (0.13-2.2)||0.380|
|CD4 count (per doubling)||0.78 (0.67-0.91)||0.002||0.88 (0.73-1.06)||0.190|
|Log HIV RNA||1.47 (1.17-1.9)||0.001||1.44 (1.11-1.9)||0.006|
|HAART||0.62 (0.34-1.13)||0.120||1.04 (0.53-2.0)||0.900|
|Log HCV RNA||1.44 (0.94-2.2)||0.090||1.35 (0.90-2.0)||0.150|
|HCV genotype 1||1.20 (0.43-3.4)||0.730||1.5 (0.53-4.5)||0.430|
|Chronic HBV||1.5 (0.21-11.1)||0.680||—||—|
|DM||0.58 (0.14-2.5)||0.460||0.65 (0.15-2.8)||0.560|
|HTN||0.85 (0.43-1.7)||0.640||1.04 (0.52-2.1)||0.920|
|BMI||0.95 (0.90-1.0)||0.100||0.98 (0.93-1.03)||0.480|
|GFR||0.99 (0.98-1.0)||0.270||0.99 (0.98-1.0)||0.350|
We also assessed liver-related mortality after adjusting for categories of factors that could affect risk of death. We continued to observe marked racial/ethnic differences in liver-related mortality after adjusting for cardiovascular risk factors, HIV immune control, and liver-related factors, suggesting that the observed racial/ethnic differences were not explained by these comorbid conditions (Table 4).
|Racial/ethnic comparison||Unadjusteda HR (95% CI)||P Value||Cardiovascular Risk Factors HRb (95% CI)||P Value||HIV-Related Factors HRc (95% CI)||P Value||Liver-Related Factors HR§ (95% CI)||P Value|
|African American||0.48 (0.23-1.03)||0.058||0.44 (0.20-0.95)||0.038||0.41 (0.19-0.88)||0.022||0.46 (0.22-0.98)||0.043|
|Hispanic||1.11 (0.50-2.4)||0.790||1.12 (0.51-2.5)||0.780||1.09 (0.46-2.4)||0.830||1.06 (0.48-2.3)||0.890|
|Hispanic (Ref) versus African American||0.44 (0.22-0.87)||0.019||0.39 (0.19-0.80)||0.010||0.38 (0.19-0.76)||0.006||0.43 (0.22-0.87)||0.019|
Importantly, we performed a competing-risks analysis to determine whether the lower risk of liver-related mortality among African-American women was the result of African Americans dying at a higher rate from non-liver-related diseases, as compared to other racial/ethnic groups. In addition, we performed a survival analysis of non-liver-related mortality. The competing-risks analysis estimated that African-American HIV/HCV-coinfected women had a similar risk of non-liver-related mortality as other racial/ethnic groups, as did the survival analysis of non-liver-related mortality (HR, 1.01; 95% CI: 0.75 -1.36; P = 0.95 versus Caucasians and HR, 1.18; 95% CI: 0.88-1.59; P = 0.27 versus Hispanics). Therefore, African-American women had a lower risk of liver-related mortality, as compared to Hispanic and Caucasian coinfected women, independent of other causes of death.
In this large cohort of HIV/HCV-coinfected women, we investigated liver-related and all-cause mortality. We observed marked differences in liver-related death between racial/ethnic groups, whereas all-cause mortality was similar between African-American, Hispanic, and Caucasian coinfected women. African Americans were approximately 60% less likely to die from liver disease, as compared to Caucasian and Hispanic women. This relationship persisted after adjusting for an extensive list of covariates that could possibly affect risk of death.
A previous investigation of HIV/HCV-coinfected veterans identified important racial differences in all-cause mortality, with significantly higher mortality among Caucasian coinfected males, compared to African Americans. Interestingly, these differences were not observed among HCV-monoinfected patients.14 Unlike our study, the Veteran study was predominantly male and limited by its cross-sectional design, small number of Hispanic patients, and inability to investigate liver-related death. Although the Veteran study did report important racial/ethnic differences in all-cause mortality among coinfected men, no previous data reflect liver-related mortality trends among coinfected populations or in HIV/HCV-coinfected women.
The reasons for the marked racial/ethnic differences in liver-related mortality in the current study certainly warrant further investigation. When we adjusted for many cofactors known to accelerate liver disease, such as HIV immune status, obesity, and alcohol use, we continued to observe significant differences between racial/ethnic groups. Although African Americans are less likely to spontaneously clear HCV, previous data do suggest that once chronically infected, African Americans tend to have less liver inflammation, as measured by alanine aminotransferase (ALT) levels and necroinflammatory scores on liver biopsy.10 Sugimoto et al. also investigated T-cell response in patients with chronic HCV infection and found that African Americans had a significantly more-robust T-cell response than Caucasians, as well as higher platelet counts, lower bilirubin, and lower ALT levels.17 These results were not due to alcohol use, gender differences, or HCV genotype, which were evenly distributed between African Americans and Caucasians. It is quite possible that differential immunologic response to HCV may account for differences in progression of HCV-related liver disease between racial/ethnic groups.
African Americans may also have slower rates of liver fibrosis than Caucasians or Hispanics once chronically infected with hepatitis C. However, most studies on this topic have been limited by small sample size, and these differences have generally not reached statistical significance.9, 10, 18, 19 Two studies have noted a significantly increased rate of liver fibrosis in Hispanics, compared to Caucasians and African Americans, after adjusting for comorbid conditions that could accelerate liver fibrosis.18, 20 Interestingly, these associations with fibrosis progression are similar to the racial/ethnic trends that we report for liver-related mortality, suggesting that differential rates of fibrosis progression may play an important role in explaining racial/ethnic differences in HCV-related mortality.
Recent data suggest that single nucleotide polymorphisms (SNPs) may play an important role in differential rates of liver fibrosis. A study by Barreiro et al. found that the CC genotype associated with the IL28B SNP rs12979860 was predictive of cirrhosis in HIV/HCV-coinfected patients.21 Because the frequency of the CC genotype is higher in Caucasians (39%) and Hispanics (35%), as compared to Africans Americans (16%),22 these data could, in part, explain why African Americans may have slower rates of liver fibrosis. However, other studies investigating the association between IL28B and fibrosis progression have been conflicting.23-25 A genome-wide association study investigating the role of IL28B is currently underway within the WIHS to help further address differential rates of liver fibrosis between racial/ethnic groups.
There were several limitations in the current study. First, we analyzed only primary cause of death and therefore may have underestimated the true prevalence of liver-related death. Many patients with decompensated liver disease may die from infectious complications, and these diagnoses may not have been coded as liver-related events. We also lacked data on the severity of liver disease, such as cirrhosis history or synthetic functions tests (albumin and international normalized ratio). However, liver-related deaths are often the result of hepatic decompensation or liver cancer, both of which were captured as liver-related events.4 Additional data on severity of liver disease or secondary causes of death would likely have strengthened our findings.
There were important strengths of the current study, most notably our ability to separately analyze liver-related and non-liver-related deaths. This showed that the lower risk of liver-related death among African-American women was not simply the result of African Americans dying at a higher rate from non-liver-related deaths, as compared to Hispanics and Caucasians. In addition, the primary cause of death was abstracted from death certificates, which are often subject to misclassification and miscoding. However, death data were ascertained in a similar fashion over time, and any misclassification would have unlikely varied by race. Finally, our survival analyses incorporated a time-varying analysis of many time-dependent factors that may have affected mortality risk. Alcohol use, for example, was analyzed over time, rather than use at a fixed time point in the study period. This statistical method allowed us to capture the complex and changing effect of many risk factors for death over the course of each participant's lifetime.
In conclusion, we observed novel, important racial/ethnic trends in this large cohort of HIV/HCV-coinfected women. Compared to Hispanic and Caucasian coinfected women, African-American coinfected women were much less likely to die from liver-related disease, which did not appear to be the result of differential classification of cause of death. Previous data have revealed racial/ethnic differences in immunologic response to HCV, which may affect rates of liver fibrosis. Future studies incorporating fibrosis progression may help us to better understand these marked racial/ethnic discrepancies in liver-related mortality.
- 20More advanced hepatic fibrosis in hispanics with chronic hepatitis C infection: role of patient demographics, hepatic necroinflammation, and steatosis. Am J Gastroenterol 2006; 101: 1817-1823., , , , , .Direct Link: