Potential conflict of interest: Nothing to report.
Supported by Investigator Initiated Research grant EPID-007-10S, Clinical Science Research and Development, Office of Research and Development, Veterans Affairs (to G. N. I.).
Cirrhosis is a leading cause of death among patients infected with human immunodeficiency virus (HIV). We sought to determine risk factors for and time trends in the prevalence of cirrhosis, decompensated cirrhosis, and hepatocellular carcinoma (HCC) among patients diagnosed with HIV who received care in the Veterans Affairs (VA) health care system nationally between 1996 and 2009 (n = 24,040 in 2009). Among patients coinfected with HIV and hepatitis C virus (HCV), there was a dramatic increase in the prevalence of cirrhosis (3.5%-13.2%), decompensated cirrhosis (1.9%-5.8%), and HCC (0.07%-1.6%). Little increase was observed among patients without HCV coinfection in the prevalence of cirrhosis (1.7%-2.2%), decompensated cirrhosis (1.1%-1.2%), and HCC (0.03%-0.13%). In 2009, HCV infection was present in the majority of patients with HIV who had cirrhosis (66%), decompensated cirrhosis (62%), and HCC (80%). Independent risk factors for cirrhosis included HCV infection (adjusted odds ratio [AOR], 5.82; 95% confidence interval [CI], 5.0-6.7), hepatitis B virus (HBV) infection (AOR, 2.40; 95% CI, 2.0-2.9), age (AOR, 1.03; 95% CI, 1.02-1.04), Hispanic ethnicity (AOR, 1.76; 95% CI, 1.4-2.2), diabetes (AOR, 1.79; 95% CI, 1.6-2.1), and alcohol abuse (AOR, 1.78; 95% CI, 1.5-2.1), whereas black race (AOR, 0.56; 95% CI, 0.48-0.64) and successful eradication of HCV (AOR, 0.61; 95% CI, 0.4-0.9) were protective. Independent risk factors for HCC included HCV infection (AOR, 10.0; 95% CI, 6.1-16.4), HBV infection (AOR, 2.82; 95% CI, 1.7-4.7), age (AOR, 1.05; 95% CI, 1.03-1.08), and low CD4+ cell count (AOR, 2.36; 95% CI, 1.3-4.2). Among 5999 HIV/HCV-coinfected patients, 994 (18%) had ever received HCV antiviral treatment, of whom 165 (17%) achieved sustained virologic response. Conclusion: The prevalence of cirrhosis and HCC has increased dramatically among HIV-infected patients driven primarily by the HCV epidemic. Potentially modifiable risk factors include HCV infection, HBV infection, diabetes, alcohol abuse, and low CD4+ cell count. (HEPATOLOGY 2013)
Cirrhosis and hepatocellular carcinoma (HCC) have emerged as leading causes of death among patients infected with human immunodeficiency virus (HIV), particularly in countries where widespread introduction of highly active antiretroviral treatment (HAART) since 1996 led to a dramatic reduction in deaths due to acquired immune deficiency syndrome (AIDS).1-3 Chronic hepatitis C virus (HCV) infection, hepatitis B virus (HBV) infection, and alcoholic liver disease, which are the most important causes of cirrhosis and HCC worldwide, are more common in HIV-infected than in non–HIV-infected persons.
Recent studies have suggested that the occurrence of both cirrhosis and HCC has been increasing rapidly in the United States over the last 10-15 years.4-6 It is likely that these increases are further magnified among HIV-infected patients due to the high prevalence of chronic liver disease, but recent population-based estimates are lacking. We sought to determine the proportion of HIV-infected patients in the Veterans Affairs (VA) health care system who had a diagnosis of cirrhosis or HCC each year between 1996 and 2009. We reasoned that the VA health care system was an appropriate setting for this study because it is the largest integrated health care system in the United States, providing long-term care for a large population of HIV-infected patients. Estimates of the proportion of HIV-infected patients who have cirrhosis or HCC are important to help us understand the burden of these conditions on HIV-infected patients and health care systems. Trends over time in these proportions can be helpful in making future projections and developing strategies to address these conditions. Finally, investigation of risk factors associated with cirrhosis and HCC in HIV-infected patients can help us understand the reasons for any trends in their prevalence and identify factors that can potentially be modified in order to reduce their prevalence.
AIDS, acquired immune deficiency syndrome; AOR, adjusted odds ratio; CCR, Clinical Case Registry; CI, confidence interval; HAART, highly active antiretroviral treatment; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; HIV, human immunodeficiency virus; ICD-9, International Classification of Diseases, Ninth Revision; SVR, sustained virologic response; VA, Veterans Affairs.
Patients and Methods
Study Design and Data Source
The VA health care system comprises 154 medical centers and 875 ambulatory care and community-based outpatient clinics throughout the United States. It is the largest integrated health care provider in the country and uses electronic medical records almost exclusively. The VA HIV Clinical Case Registry (CCR) is a national, continually updated database of all VA patients with diagnosed HIV infection.7, 8 It consists of a series of data sets extracted from the electronic medical records of all HIV-infected patients from every VA facility in the country. These data sets include demographics, inpatient and outpatient visits, problem lists, procedures, diagnostic tests, laboratory data, and pharmacy data. At each VA facility in the country, specifically designed CCR software identifies patients with laboratory test results (HIV antibody, western blotting, and HIV RNA) or one of 14 diagnostic International Classification of Diseases, Ninth Revision (ICD-9) codes (042, 042.0, 042.1, 042.2, 042.9, 043.0, 043.2, 043.3, 043.9, 044.0, 044.9, 079.53, 795.71, and 795.8) indicating probable HIV infection. Such patients are then individually reviewed by a local, specially trained CCR coordinator at each facility who confirms that they have HIV and enters them in the registry following a published protocol (Supporting Information, Appendix 1).9
We received HIV CCR data sets extending to December 31, 2009, and beginning on January 1, 1986, for the laboratory data set; January 1, 1987, for the inpatient admission data set; January 1, 1993, for the problem list data set; January 1, 1994, for the outpatient visit data set; and March 31, 1987, for the pharmacy data set.
Analytical Data Sets
We created 14 analytical data sets, each consisting of the HIV-infected patients who received care in any VA facility in each of 14 calendar years from 1996, the year that HAART was introduced, to 2009. This was defined by having at least one inpatient or outpatient visit during that calendar year, which is henceforth referred to as being “in-care” and is a standard definition used by the CCR and associated publications. In each data set, we identified the patients who had a diagnosis of cirrhosis or HCC recorded during or prior to that year. Additionally, we identified the age of each patient during that calendar year and the following diagnoses recorded during or prior to that year: diabetes, alcohol abuse or dependence, and drug abuse or dependence. Finally, we identified race, ethnicity, sex, and presence of chronic HCV and HBV infection.
Definition of Analytical Variables
Cirrhosis was defined by any of the following ICD-9 codes: 571.2 (cirrhosis with alcoholism) and 571.5 (cirrhosis with no mention of alcohol). Decompensated cirrhosis was defined by ICD-9 codes 456.0-456.21 (esophageal varices with or without bleeding), 789.5 (ascites), 567.23 (spontaneous bacterial peritonitis), 572.2 (hepatic encephalopathy), 572.4 (hepatorenal syndrome), 070.2x (viral hepatitis B with hepatic coma), 070.44 (chronic hepatitis C with hepatic coma), and 070.6 (unspecified viral hepatitis with hepatic coma). HCC was defined by the presence of ICD-9 code 155.0. We identified these ICD-9 codes recorded in any inpatient or outpatient medical record or in a patient problem list. The diagnoses of cirrhosis, decompensated cirrhosis, or HCC was based on the presence of these ICD-9 codes in two or more inpatient or outpatient encounters. These diagnostic definitions of cirrhosis and HCC, based on ICD-9 codes derived from VA electronic medical data sets, have been shown to have high agreement with review of the medical records by a physician (cirrhosis: positive predictive value, 88%; negative predictive value, 92%; HCC: positive predictive value, 94%; negative predictive value, 100%), including specifically in HIV-infected patients,10, 11 and have been extensively used in epidemiological studies.4, 10-14
HCV infection was defined by a positive HCV RNA viral load or a positive HCV RIBA and HCV antibody derived from the laboratory files. We defined receipt of HCV antiviral treatment by two or more prescriptions of pegylated or regular interferon, with or without ribavirin and sustained virologic response as a negative HCV viral load measured at least 12 weeks after the end of antiviral treatment.15, 16 HBV infection was defined by a positive hepatitis B surface antigen. All HIV viral loads and CD4+ cell counts were extracted from the laboratory files over a 2-year period starting 6 months prior to the outcomes of interest. Type 2 diabetes mellitus was defined by ICD-9 codes 250-250.92. Alcohol use disorders were identified by ICD-9 codes for alcohol abuse (305.00-305.03), dependence (303.90-303.93), and withdrawal (298.1). Substance use disorders were defined by the presence of codes for substance abuse (305.2-305.9), dependence (304.0-304.9), or drug withdrawal (292.0). We identified these ICD-9 codes for diabetes, alcohol and substance abuse disorders, and relevant HIV viral loads and CD4+ cell counts that were recorded at least 6 months prior to the earliest recorded diagnosis of cirrhosis, decompensated cirrhosis, or HCC to ensure that these potential exposures preceded the outcomes of interest.
Analyses were performed using STATA version 11 (College Station, TX). We calculated the proportion of in-care, HIV-infected patients in each calendar year who had a diagnosis of cirrhosis, decompensated cirrhosis, or HCC recorded that year or earlier, with or without adjusting by direct standardization using the age distribution of the entire population from all calendar years as the standard. Proportions were compared across calendar years using a linear regression model to perform a test of trends adjusting for age. Logistic regression was used to evaluate whether selected patient characteristics were associated with the presence of cirrhosis, decompensated cirrhosis, or HCC among the patients in VA care in calendar year 2009 (the most recent in this study). These characteristics (age, race, ethnicity, sex, HCV infection, HBV infection, diabetes, alcohol abuse, substance abuse, detectable HIV viral load, maximum CD4+ cell count <200/μL or maximum CD4 % <14%, and receipt and response to HCV antiviral treatment) were selected a priori and modeled simultaneously.
The number of HIV-infected patients in VA care increased substantially from 16,294 in 1996 to 23,974 in 2004 and remained approximately stable at ≈24,000 from 2004 to 2009 (Table 1). Between 1996 and 2009, there was a steady increase in the proportion of in-care patients with cirrhosis (2.3%-5%), decompensated cirrhosis (1.4%-2.4%), and HCC (0.04%-0.50%) as shown in Fig. 1 and in the Supporting Information, Appendix 2 (P < 0.001 for test of trends over time). Among HIV/HCV-coinfected patients, the increase in the proportion of patients with cirrhosis (3.5%-13.2%), decompensated cirrhosis (1.9%-5.8%), and HCC (0.07%-1.6%) during 1996-2009 was even more pronounced than among all HIV-infected patients (Fig. 1 and Supporting Information, Appendix 3; P < 0.001 for test of trends over time). In contrast, there was little change in the proportion of patients without HCV coinfection who had cirrhosis, decompensated cirrhosis, or HCC during 1996-2009 (Fig. 1 and Supporting Information, Appendix 4; P > 0.05 for test of trends over time). Figure 1 shows that the prevalence of decompensated cirrhosis reached a plateau in 2006, whereas the prevalence of cirrhosis and HCC continued to increase.
Table 1. Crude Prevalence of HCV Infection Among HIV-Infected Patients in VA Care, 1996-2009
HIV-Infected Patients in VA Care
HCV Status Unknown
Data are presented as no. (%).
As expected, we observed a dramatic reduction in mortality between 1996, when HAART was introduced, and 2009 (Fig. 1). However, this reduction was much less pronounced for patients with HCV coinfection than for patients without HCV, coincident with the dramatic increase in the prevalence of cirrhosis, decompensated cirrhosis, and HCC among patients with HCV. Age-adjusted mortality has been higher in patients with HCV than in those without HCV since 1998.
We calculated the median interval between the date of the earliest outpatient visit in the VA and the date of diagnosis of cirrhosis (2360 days [interquartile range, 901-3520]), decompensated cirrhosis (2135 days [interquartile range, 1081-2135]), or HCC (3351 days [interquartile range, 2360-4169]). These findings demonstrate that the majority of patients were diagnosed with cirrhosis or HCC several years after being seen in the VA for the first time, suggesting that they did not present to the VA because of symptoms of cirrhosis or HCC, or following a diagnosis of cirrhosis or HCC made elsewhere.
There was a decrease in the proportion of HIV-infected patients who also had diagnosed HCV coinfection from ≈35% in 1996 to ≈25% in 2009 (Table 1). Although a small proportion of patients did not have adequate virologic testing to confirm HCV status, it is unlikely that these missing data account for the decreasing prevalence of HCV infection. This is because the proportion with missing data was actually much lower in 2009 (6.4%) than in 1996 (13.8%). Therefore, if anything, the decline in actual HCV prevalence is likely to be even more pronounced than the decline in diagnosed HCV prevalence that we report (assuming that a similar proportion of patients with missing data had HCV across calendar years). There was a dramatic increase in the crude (6.5%-20.2%) and age-adjusted (8.5%-16.5%) prevalence of diabetes between 1996 and 2009, while the prevalence of a history of alcohol abuse/dependence was approximately constant around 30%-34% (data not shown).
During 2009, the most recent calendar year in our study, there were 24,040 HIV-infected patients in VA care, including 1190 (5%) patients with cirrhosis, 565 (2.4%) patients with decompensated cirrhosis, and 122 (0.5%) patients with HCC (Table 2). Among the 565 patients with decompensated cirrhosis, 205 (36%) had ascites, 237 (42%) had gastroesophageal varices (including 79 [14%] with bleeding varices), 184 (33%) had hepatic encephalopathy, 25 (4%) had spontaneous bacterial peritonitis, and 10 (2%) had hepatorenal syndrome. HCV infection was present in 66% of patients with cirrhosis, 62% of patients with decompensated cirrhosis, and 80% of patients with HCC (Table 2). HBV infection was present in 14% of patients with cirrhosis, 14% of patients with decompensated cirrhosis, and 17% of patients with HCC (although these may be underestimates, as 7%-13% of these patients did not have documented HBV status). Either HCV or HBV were present in 74% of patients with cirrhosis, 70% of patients with decompensated cirrhosis, and 92% of patients with HCC. Thus, HCV or HBV infection likely account for the majority of cases of cirrhosis and HCC among HIV-infected veterans.
Table 2. Characteristics of HIV-Infected Patients in VA Care During Calendar Year 2009, According to Presence or Absence of Cirrhosis, Decompensated Cirrhosis, and HCC
No cirrhosis or HCC (n = 22,802)
Compensated Cirrhosis (n = 1190)
Decompensated Cirrhosis (n = 565)
HCC (n = 122)
Limited to HCV-positive patients.
Limited to HCV-positive patients who receive antiviral treatment.
In multivariate logistic regression analyses, HCV infection, HBV infection, age, Hispanic ethnicity, diabetes, and alcohol abuse were associated (P < 0.05) with higher likelihood of cirrhosis and decompensated cirrhosis, whereas black race was associated with lower likelihood compared with white race (Table 3). Characteristics independently associated with HCC included HCV infection, HBV infection, age, and a low CD4+ cell count.
Table 3. Association of Selected Characteristics with the Development of Compensated Cirrhosis, Decompensated Cirrhosis, or HCC Among HIV-Infected Patients in VA Care During Calendar Year 2009
Independent Variables Included in the Model
Data are presented as AOR (95% CI) and were adjusted for the characteristics listed via multivariate logistic regression.
Limited to HCV-positive patients and additionally adjusted for HCV genotype.
Limited to patients who received HCV antiviral treatment and additionally adjusted for HCV genotype.
Age (per year)
HIV viral load, detectable
Maximum CD4+ cell count <200/μL or maximum CD4 % <14%
Among 5999 in-care patients in 2009 who were (or had been) infected with HCV, 994 (18%) received antiviral treatment, of whom 165 (17%) had documented sustained virologic response (SVR). Patients with cirrhosis or decompensated cirrhosis were more likely to have received antiviral treatment (Table 3). Among patients who received antiviral treatment, those who achieved SVR were less likely to develop cirrhosis (adjusted odds ratio [AOR], 0.61; 95% confidence interval [CI], 0.4-0.9) or decompensated cirrhosis (AOR, 0.39; 95% CI, 0.2-0.8) than those who did not achieve SVR (Table 3).
Our findings demonstrate that the burden of end-stage complications of liver disease has been rising dramatically in HIV-infected patients over the last 15 years. Among HIV/HCV coinfected patients, the prevalence of cirrhosis rose 3.7-fold (3.5% to 13.2%) and the prevalence of decompensated cirrhosis rose 3-fold (1.9% to 5.8%). The prevalence of HCC underwent the most dramatic change, increasing 23-fold from 0.07% to 1.62%. In contrast, the prevalence of cirrhosis and HCC was much lower in patients without HCV coinfection and underwent little increase. Thus, the epidemiology of cirrhosis and HCC in HIV-infected patients is dominated by the epidemiology of chronic HCV infection. The majority of persons with HCV in the United States were infected between 1965 and 1990, when the incidence of acute HCV infection peaked at 250,000-500,000 new infections annually before declining dramatically due to a reduction in high-risk behaviors and the availability of serological testing for HCV.17 Because HCV infection takes an average of 30-40 years to cause cirrhosis or HCC, the patients infected between 1965 and 1990 would be expected to develop cirrhosis and HCC between 1995 and 2030. These epidemiologic features of HCV in the United States, together with the decline in competing AIDS-related mortality, largely explain the trends we reported in cirrhosis and HCC among HIV/HCV-coinfected patients.
Among European patients with HIV/HCV coinfection, cirrhosis was diagnosed in 5%-14% by liver biopsy and in 19%-25% by transient elastography.18-21 These proportions are inflated relative to our study's results because elastography and biopsy diagnose preclinical cirrhosis and, more importantly, because patients with advanced liver disease were overrepresented in these studies. In contrast, our study included all patients in VA care throughout the country and defined cirrhosis clinically. Cirrhosis and HCC have accounted for a rapidly increasing proportion of deaths in HIV-infected persons: 2% in 1995 (France1), 14.5% in 1999-2004 (in Europe, the United States, and Australia in the D:A:D study)2, 17% in 2005 (France1), and 18% (including 3% from HCC) in the Swiss HIV Cohort Study during 2005-2009,3 in which liver disease was the overall leading cause of death. Taken together with these publications, our results highlight an evolving epidemic of HCV-related cirrhosis and HCC in HIV-infected patients.
We observed that the proportion of HIV/HCV-coinfected patients declined between 1996 and 2009. This decline is likely the result of an aging cohort of patients infected with HCV between 1965 and 1990 who died from complications of liver disease (as well as other causes). A decline in the number of patients with HCV in the entire United States (i.e., not restricted to HIV-coinfected patients) has also been described since approximately 2000.5
Our results highlight the importance of diagnosing and treating HCV infection in HIV-infected patients. The United States Public Health Service guidelines for adults and adolescents with HIV have mandated HCV testing since 1999, and this has been reiterated in subsequent recent national guidelines.22 Among patients in VA care in 2009, 94% had been tested for HCV, and among those positive for HCV, 18% had received antiviral treatment, of whom 17% achieved SVR. Unfortunately, low treatment rates (<7%) and SVR rates (21%-35%) are common in other HIV/HCV-coinfected cohorts.23, 24 Our findings suggest that patients who do achieve SVR are less likely to develop cirrhosis. Triple therapy regimens (including the protease inhibitor telaprevir or boceprevir in addition to pegylated interferon and ribavirin) have been shown to increase SVR by 30%-40% in HCV-monoinfected patients with genotype 1 infection. However, they are not yet approved for use in HIV/HCV-coinfected patients. Preliminary results of studies involving telaprevir25 or boceprevir26 suggested high end-of-treatment response rates, but coadministration of these new drugs may affect the effectiveness of ritonavir-boosted HIV protease inhibitors. The final results of these studies are eagerly awaited and could potentially transform HCV antiviral treatment in patients with HIV coinfection.
HCV coinfection was the most important risk factor for cirrhosis (AOR, 5.8; 95% CI, 5.0-6.7), decompensated cirrhosis (AOR, 4.4; 95% CI, 3.6-5.4), and HCC (AOR, 10; 95% CI, 6.1-16.4), being present in 66% of patients with cirrhosis, 62% of patients with decompensated cirrhosis, and 80% of patients with HCC in 2009. In addition to HCV infection, we identified HBV infection, alcohol abuse, age, diabetes, and Hispanic ethnicity as independent risk factors for cirrhosis and decompensated cirrhosis, whereas black persons were less likely to develop cirrhosis in multivariate models. These associations highlight the importance of treatment or prevention of alcohol abuse, HBV infection, and diabetes as additional strategies aimed at reducing the burden of cirrhosis in HIV-infected patients. Independent risk factors for HCC included HCV infection, HBV infection, age, and low CD4+ cell count. The association between HIV-related immune suppression (manifesting as a low CD4+ cell count) and HCC, as well as other non–AIDS-defining malignancies, has been described in recent studies and is receiving increasing attention.27, 28 Our finding that low CD4+ cell count was strongly associated with HCC but not with cirrhosis helps to exclude the possibility that the association between low CD4+ cell count and HCC was simply due to lowering of CD4+ cell count that may occur in cirrhosis, which often precedes HCC, secondary to portal hypertension and splenic sequestration of lymphocytes. Also, our finding that low CD4+ cell count but not plasma HIV RNA level was associated with HCC mirrors the results of a recent study,27 and suggests that immune suppression is more directly relevant to carcinogenesis than uncontrolled viral replication. The association between low CD4+ cell count and HCC supports recommendations that antiretroviral treatment should aim at reaching and maintaining a high CD4+ cell count to prevent malignancy.27
Patterns of antiretroviral medication use have changed substantially since 1996. There has been a reduction in use of thymidine analogue reverse-transcriptase inhibitors (stavudine and zidovudine) and didanosine, which have been associated with lipodystrophy hepatic steatosis, metabolic acidosis, and elevated serum liver enzymes,29, 30 and an increase in the use of nucleotide analogs such as tenofovir, lamivudine, and abacavir. “Older” protease inhibitors such as indinavir, ritonavir, and lopinavir have been shown to impair insulin sensitivity, an important cofactor in liver disease, whereas atazanavir and saquinavir may have minimal effects.31, 32 Thus, the increase in prevalence of cirrhosis that we report occurred despite increasing use of more “liver friendly” regimens and cannot readily be explained by known changes in the patterns of antiretroviral medication use.
Our study is limited by the fact that the definitions of cirrhosis, decompensated cirrhosis, and HCC were based on ICD-9 codes. However, these diagnostic definitions of cirrhosis and HCC based on ICD-9 codes derived from VA electronic medical data sets have been shown to have high agreement with physician medical record review, including specifically among HIV-infected patients,10, 11 and have been extensively used in epidemiological studies.4, 10-14 It would be practically impossible to do a representative study such as ours in which the diagnosis of cirrhosis, decompensated cirrhosis, or HCC is individually confirmed or excluded prospectively in each patient by a series of predetermined laboratory tests, imaging studies, physical examination findings, and histological examination of a liver biopsy. It is likely that cases of early cirrhosis with preserved liver function and no clinical manifestations were missed in our study, such that our estimates are, if anything, underestimates of the true prevalence of cirrhosis. Such cases of early cirrhosis are notoriously difficult to diagnose with certainty even by expert hepatologists in the absence of a liver biopsy, which is generally not performed in all patients simply to rule out occult cirrhosis. The increase in prevalence over time that we report might be due to increased diagnostic awareness—that is, it is possible the prevalence did not actually increase, but physicians were more likely to diagnose cirrhosis and HCC. If that were the case, the prevalence should have increased among the patients who were not coinfected with HCV, just as it did among the HCV-coinfected patients. Because we did not observe a substantial increase over time in the prevalence of cirrhosis and HCC among the patients who were not coinfected with HCV, we suspect that increased diagnostic awareness played little role in generating the trends that we described over time. In addition, the temporal trends that we observed in the diagnosis of decompensated cirrhosis also argue against increasing physician awareness playing a large part in the observed trends: the clinical presentation of decompensated cirrhosis is both dramatic and specific such that its diagnosis is much less subject to the level of physician awareness.
Our results apply directly to overwhelmingly male, HIV-infected patients in VA care throughout the United States—one of the largest samples of HIV-infected patients worldwide. Other populations of HIV-infected patients may have somewhat higher or lower absolute prevalence of cirrhosis, decompensated cirrhosis, or HCC depending on the relative inclusion of women and the prevalence of the major underlying risk factors for liver disease that we describe, especially viral hepatitis.
Strengths of this study include the fact that we studied a large population of HIV-infected patients who received care in the VA health care system, the largest integrated health care provider in the United States. We included all patients diagnosed with HIV who received care in any VA facility throughout the country between 1996 and 2009. These patients were identified using a carefully designed protocol to capture and confirm HIV diagnosis. Electronic records contained information extending backward in time for more than 2 decades. Because of the universal care provided by the VA, the integration of VA health care throughout the United States, and the fact that care is provided for free or for a nominal copay (for patients with documented financial means), patients who are in-care in the VA receive almost all their long-term care within the VA. This is in contrast to non-VA, HIV-infected patients who may receive care in multiple facilities and different geographical locations over time, thus having dispersed medical records that are not easily aggregated for research purposes. The fact that both cases (patients with cirrhosis and HCC) and controls were derived from the same, well-defined population of all diagnosed HIV patients in VA facilities enhances the internal validity of our study and reduces biases related to inappropriate selection of controls. HCV and HBV infection, the most important causes of liver disease in HIV-infected patients, were ascertained using gold standard serologic and virologic tests.
In conclusion, the prevalence of end-stage complications of liver disease (cirrhosis, decompensated cirrhosis and HCC) has been rising dramatically among HIV-infected patients, particularly those coinfected with HCV. These complications of end-stage liver disease are likely to constitute some of the most important clinical problems for HIV-infected patients and their physicians during the decade 2010-2020. Treatment or prevention of the modifiable risk factors for cirrhosis and HCC that we identified, especially HCV coinfection, may ameliorate the burden of cirrhosis and HCC and reverse their upward trends.