Potential conflict of interest: Nothing to report.
Hepatic steatosis (HS) is frequent in patients with hepatitis C virus (HCV) infection, occurring in 40%-80%, associating with metabolic and virus-related factors, namely, genotype 3 and viral load. Human immunodeficiency virus (HIV) infection and antiretroviral treatment seem to be risk factors for HS. Several studies addressed this issue in coinfected patients, with discrepant results. A meta-analysis was performed on the HS risk factors in coinfected patients. Eligible studies were identified through structured keywords including coinfection, HCV, HIV, and steatosis in relevant databases including PubMed. Pooled odds ratios (ORs) and confidence limits (CIs) were obtained with the random-effects model and the DerSimonian-Laird method. Twelve studies, including 1,989 coinfected patients, were selected. Twenty percent were infected with HCV genotype 3. The overall prevalence of HS was 50.8% (23%-72%). Four studies also included 1,540 HCV monoinfected patients, not showing an increased risk for HS in coinfected patients (OR 1.61, 95% CI 0.84-3.10, P = 0.151). In coinfected patients, HS was associated with higher body mass index (OR 1.13, 95% CI 1.07-1.19, P < 0.001), diabetes mellitus (OR 2.32, 95% CI 1.32-4.07, P = 0.003), elevated alanine aminotransferase levels (OR 1.28, 95% CI 1.02-1.61, P = 0.035), necroinflammatory activity (OR 1.72, 95% CI 1.11-2.67, P = 0.016), and fibrosis (OR 1.67, 95% CI 1.20-2.34, P = 0.003). No associations were found between HS and gender, other metabolic factors (dyslipidemia, glucose, metabolic syndrome), HCV-related factors (genotype, viral load), or HIV-related factors (viral load, CD4 count, antiretroviral therapy, and class of medication). Conclusion: In coinfected patients, HS does not seem to be more frequent than in HCV monoinfected patients and is mostly associated with metabolic factors, such as increased weight, diabetes mellitus, and more severe liver disease. The fact that no associations with HCV factors were found may be due to the small percentage of genotype 3-infected patients. HEPATOLOGY 2010
Human immunodeficiency virus (HIV) and hepatitis C virus (HCV) coinfection is a major health problem. Current estimations find that more than one-fourth of HIV-infected patients are coinfected with HCV,1 and HCV-related deaths are now the third cause of death in HIV-infected patients.2
When compared to the general population, HCV-infected patients have a 2.5-fold increased risk of developing hepatic steatosis (HS),3 often associated with faster fibrosis progression and more severe hepatic fibrosis,4 an increased risk of hepatocellular carcinoma development,5 and lower rates of sustained viral response to HCV antiviral therapy, mainly in genotype 3.6 There are two types of HS in HCV infection, a viral-related HS and a metabolic-related HS. The first is mostly associated with genotype 3 virus and is a consequence of the direct cytopathic effect on hepatocytes. There is a direct relation with viral load and it regresses after successful HCV therapy. The metabolic type, having allegedly a dismal effect in anti-HCV therapy, is more common amid other genotypes and is primarily related to metabolic factors like obesity, alcohol, diabetes mellitus (DM), or insulin resistance.7
Hepatic steatosis was present in about 30% of HIV-infected patients in the era before highly active antiretroviral therapy (HAART),8 comparable to the prevalence in the general population.9 In HIV-infected patients in the post-HAART era, the HS prevalence is indefinite.8 Several studies addressed the topic with a noninvasive approach and the majority did not exclude patients with HCV coinfection. Two well-designed studies, one assessing HS with ultrasonography10 and the other with computed tomography (CT) scans,11 found prevalences of 31% and 37%, respectively. Although the main risk factors found in these studies were metabolic, another study compared two groups of patients with HS and different HIV status; the results showed that HIV-infected patients had lower body mass index (BMI) and lower body fat percentage.12 In HIV, the HS etiology is most likely multifactorial13 relating to metabolic factors, e.g., insulin resistance/DM/dyslipidemia and lipodystrophy, all of which may be linked to antiretroviral therapy (ART). Protease inhibitors (PI) and nucleoside reverse transcriptase inhibitors (NRTI) have been related to insulin resistance14 and lipodystrophy15; NRTI may cause mitochondrial dysfunction with subsequent direct hepatotoxic effects.16 Moreover, HIV infection itself may facilitate DM by way of tumor necrosis factor-α stimulation and mitochondria damage.13 Also, other coexistent morbidities such as alcohol consumption or HCV infection may enhance steatogenesis.13
In the last 5 years several studies concentrated on the real prevalence of HS and its risk factors in HIV/HCV-coinfected patients, but with discrepant results. Therefore, we conducted an in-depth review and meta-analysis of available studies on the prevalence and risk factors for HS in HIV/HCV-coinfected patients and hereby report their results as well as the analysis by comparison with HCV monoinfected patients.
ALT, alanine aminotransferase; AST, aspartate aminotransferase; ART, antiretroviral therapy; BMI, body mass index; DM, diabetes mellitus; HAART; highly active antiretroviral therapy; HCV, hepatitis C virus; HIV, human immunodeficiency virus; HS, hepatic steatosis; NAFLD, nonalcoholic fatty liver disease; NNRTI, nonnucleoside reverse transcriptase inhibitors; NRTI, nucleoside reverse transcriptase inhibitors; PI, protease inhibitors.
Materials and Methods
This analysis was performed with the Meta-analysis of Observational Studies in Epidemiology (MOOSE) criteria.17 MEDLINE and Current Contents were searched by one investigator to identify relevant articles published until May 2009. In the electronic scrutiny, study references and relevant review articles on HIV/HCV coinfection were manually searched; eligible studies were identified through structured keywords: “steatosis,” “fatty liver,” “VHC,” “HIV,” and “coinfection.” All analyzed studies had to be published in English and to comprise a series of consecutive eligible patients, although allowing for exclusions due to other concomitant liver diseases and not necessarily excluding alcohol consumption.
The abstracts of all articles identified by the initial search were reviewed by another author and both authors reviewed the full text of all eligible studies reporting the prevalence of HS and its risk factors. Data elements sought from each included study were protocol-specified including steatosis prevalence, HCV monoinfection steatosis prevalence comparison data, study location, demographic data, HCV genotype, alcohol consumption, and other risk factors for HS.
For calculations, we computed the actual number of subjects from each study and performed a pooled analysis of the data. Pooled estimation of the prevalence of HS in HIV/HCV-coinfected patients and confidence limits were obtained with the random-effects model and the DerSimonian-Laird method.18 Analysis of the heterogeneity of prevalence across studies was done with λ2 tests. As all tests showed great heterogeneity, random effects models were preferred regarding fixed effects models.
Statistical analysis was conducted with STATA software (College Station, TX), version 10. Significance was established for P < 0.05.
Twelve studies were included19-30 addressing a total of 1,989 HIV/HCV-coinfected patients. In four of these studies20, 24, 25, 28 data regarding HCV-monoinfected patients were presented separately, with a total of 1,540 patients. Three studies had a retrospective recruitment21, 22, 25; the remaining were prospective. HS was always evaluated by liver biopsy. Their main characteristics are summarized in Tables 1 and 2. Studies with a noninvasive assessment of HS were not found.
Inclusion criteria were similar in all studies, which enrolled patients with known HIV/HCV coinfection, detectable HCV RNA assessed by way of sensitive polymerase chain reaction-based techniques, and liver biopsy performed to evaluate liver damage in HCV antiviral therapy candidates. Two studies23, 27 were subanalyses of major controlled randomized trials of HCV anti-viral treatment in coinfected populations, namely, the ANRS HC0231 and APRICOT trials.32 In these trials HIV infection was required to be stable, with good immune system status and under no ART or on a stable schedule for the 3 months preceding the study.
Exclusion criteria were similar but not identical across studies. All of them required patients to be HCV treatment-naives and negative for hepatitis B surface antigen. Except for significant alcohol consumption, greater than 40-50 g per day, which was an exclusion criterion in only five studies,23-25, 27, 28 other causes of liver disease were systematically excluded. Liver biopsies with less than 10 mm length were considered as an exclusion criterion in just two studies.25, 26 Decompensated liver cirrhosis was a transversal exclusion criterion.
Definitions and grading of HS were not homogeneous in all studies, although the majority19-24, 27-30 used Brunt's criteria or similar.33 Two studies25, 26 used the chronic hepatitis C validated METAVIR classification system,34 in which steatosis is considered mild if present in 1%-10% of the hepatocytes, moderate in 11%-30%, and severe when in more than 30%. Necroinflammation and fibrosis were graded or staged by METAVIR20, 21, 23, 25, 26 or Ishak classifications.35
In HIV/HCV-coinfected patients the overall prevalence of HS was 50.8%, ranging from 23%-72%. In the four studies that also evaluated patients with HCV monoinfection, the overall prevalence of HS was 48.6%, ranging from 33%-59%.
Evaluating all data together, HIV did not confer an increased risk for HS when compared to HCV monoinfection (pooled odds ratio [OR] 1.67, 95% confidence interval [CI] 0.84-2.10, P = 0.151) (Fig. 1).
There was no association between gender and the prevalence of HS in HIV/HCV-coinfected patients, either in individual studies or in the pooled data: pooled OR for female sex was 0.85, 95% CI 0.60-1.21, P = 0.365, in 10 studies.19-22, 24, 26, 27, 29, 30 Similarly, HS prevalence was not different in African-American patients: pooled OR was 0.76, 95% CI 0.24-2.48, P = 0.668, in two studies.21, 24 As verified in seven studies,19-22, 24, 29, 30 alcohol consumption was not associated with an increased risk of HS (pooled OR 0.95, 95% CI 0.70-1.30, P = 0.761). It should be noted that among these studies only Gaslightwala and Bini24 was restricted to patients who were drinking less than 2 beverages per day.
BMI, assessed in four studies,22, 23, 25, 26 was associated with an increased risk for HS; however, with small magnitude (pooled OR 1.13, 95% CI 1.07-1.19, P < 0.001) (Fig. 2).
Regarding the presence of DM, seven studies20-22, 24, 27, 29, 30 analyzed a possible association with increased risk for HS, although only two showed a significant association.24, 30 However, in the pooled data analysis there was a strong statistically significant association (pooled OR 2.32, 95% CI 1.32-4.07, P = 0.003) (Fig. 2). When analyzing the presence of increased fasting glucose levels, although there seemed to be an association in each of three studies,19, 22, 27 this was not confirmed in the pooled data (pooled OR 1.52, 95% CI 0.58-4.02, P = 0.396).
A possible association between HS and hypertriglyceridemia was evaluated in four studies19, 22, 24, 27; only two revealed a weak though statistically significant association.22, 24 The pooled data failed to show any association (pooled OR 1.00, 95% CI 0.97-1.05, P = 0.796).
The presence of metabolic syndrome was evaluated as a risk factor for HS in two studies,27, 30 and although one of the studies showed a positive association, the pooled data failed to confirm it (pooled OR 1.07, 95% CI 0.13-8.83, P = 0.949).
Lipodystrophy was studied only once,21 and a positive association with HS was found (OR 3.02, 95% CI 1.06-8.51).
Individually, just one27 out of seven studies19, 24, 25, 27, 29, 30 found a positive association between alanine aminotransferase (ALT) elevation and HS. Yet in the pooled data there was a weak but positive association (pooled OR 1.28, 95% CI 1.02-1.61, P = 0.035) (Fig. 2).
Two studies19, 26 presented conflicting results regarding aspartate aminotransferase (AST) elevation, and the pooled data failed to confirm an association (pooled OR 1.76, 95% CI 0.61-5.03, P = 0.292).
When nine studies were evaluated together,19-23, 26, 27, 29, 30 advanced fibrosis, i.e., bridging fibrosis or hepatic cirrhosis, presented a positive association with HS in the pooled data (pooled OR 1.67, 95% CI 2.34-3.01, P = 0.003) (Fig. 2). Severe necroinflammatory activity also presented a positive association with HS in the meta-analysis of seven studies19-22, 25-27 (pooled OR 1.72, 95% CI 1.11-2.67, P = 0.016), although that association was only statistically significant in three studies19, 25, 26 (Fig. 2).
HCV Factors Associations.
Regarding HCV infection, the pooled data did not corroborate any association with possible risk factors for HS. In fact, the analysis of nine studies20-27, 30 failed to demonstrate genotype 3 as a risk factor (pooled OR 2.00, 95% CI 0.82-4.87, P = 0.127). However, a subanalysis of the three European studies with a genotype 3 prevalence higher than 30%23, 25, 26 showed a strong association with HS (pooled OR 2.52, 95% CI 1.77-3.59, P < 0.001).
HCV viral load was assessed as a risk factor for HS in seven studies,19, 20, 22-24, 27, 29 with only two showing a positive association.23, 27 The Bani-Sadr et al.23 data suggested an increase of more than 50% in the risk for HS (OR 1.65, 95% CI 1.22-2.23) by each increase of 1 log10 IU/L in the HCV viral load. However, a meta-analysis of three studies20, 22, 24 evaluating the risk of high viral load did not confirm an association (pooled OR 0.92, 95% CI 0.92-1.07, P = 0.819). The definition of high viral load was not exactly the same between studies.
HIV Factor Associations.
Immune status, assessed by CD4 cells count, did not seem to relate to HS because four out of five studies that evaluated that association presented negative results.24-26, 29, 30 Only Gaslightwala and Bini24 found a higher prevalence of HS when CD4 cell counts were higher than 350 cells/mm3 (72% versus 51%, P = 0.004). A meta-analysis of two studies25, 30 that used the same cutoff of CD4 cells count higher than 200 cells /mm3 also failed to show a statistically significant association (pooled OR 0.83, 95% CI 0.39-1.79, P = 0.637).
The effect of the duration of HIV infection on the prevalence of HS was evaluated in three studies.20, 21, 30 Although Sterling et al.30 found a slightly lower duration of infection in patients with HS (18.5 ± 0.95 years versus 22.6 ± 0.75 years, P = 0.036), each year of HIV infection duration21 or being infected by HIV for more than 7 years20 were not associated with a different risk for HS.
Using different approaches, many studies evaluated a potential effect of the HIV viral load19-22, 24-26, 29, 30; seven, however, did not show a difference. Regarding having detectable versus nondetectable viral load, the McGovern et al.22 results suggested that patients with detectable viral load had less HS (OR 0.86, 95% CI 0.76-0.98). However, a meta-analysis of four studies20, 22, 24, 25 did not confirm an association (pooled OR 0.95, 95% CI 0.61-1.47, P = 0.815). Regarding the fact of having a viral load higher than 400 copies per mL, the pooled data of two studies19, 30 also did not find an association (pooled OR 0.29, 95% CI 0.80-2.09, P = 0.295).
Being under ART did not seem to affect the risk for HS in the pooled data of six studies20-22, 26, 29, 30 (pooled OR 1.04, 95% CI 0.56-1.95, P = 0.903). In fact, individually, only two of these studies22, 30 detected a marginal statistically significant difference, although in opposite directions.
Analyzing the class of medications, no differences in the use of PI, NRTI, or nonnucleoside reverse transcriptase inhibitors (NNRTI) were found in relation to HS. In fact, only one30 out of eight studies20-22, 24-26, 29, 30 that evaluated PI showed a weak association. The pooled data of seven studies21, 22, 24-26, 29, 30 was also negative (pooled OR 1.05, 95% CI 0.72-1.52, P = 0.799). In addition, out of six studies,22, 24-26, 29, 30 only McGovern et al.22 found a positive association between the use of NRTI and HS, with a more than 2-fold increased risk (OR 2.14, 95% CI 1.10-4.14). The pooled data of the remaining six studies also found no increased risk (pooled OR 0.89, 95% CI 0.35-2.24, P = 0.808). Lastly, concerning NNRTI therapy, out of seven studies21, 22, 24-26, 29, 30 only Neau et al.26 found it to have protective properties against HS (OR 0.40, 95% CI 0.20-0.90). The pooled data of these seven studies failed to find an association (pooled OR 0.86, 95% CI 0.61-1.19, P = 0.353).
The present review found a huge discrepancy in the prevalence of HS in HIV/HCV-coinfected patients, ranging from 23%-72%, which may be related to differences in the studied populations and comorbidities such as alcohol consumption, DM, obesity, and type and duration of HCV infection, HIV infection status, and others. Several studies' biases might be pointed out. In fact, two studies23, 27 evaluated patients from a clinical trial on HCV therapy, a selected population without significant alcohol or active drug consumption and stable HIV infection, which may not translate to the real population of HIV/HCV-coinfected patients. Further, in studies not including patients from clinical trials selection depended on having a liver biopsy performed before starting HCV antiviral treatment, and those may not be representative of the HIV/HCV-coinfected population. In fact, the threshold to perform a liver biopsy may be different in HIV/HCV-coinfected patients, in whom fibrosis progression is accelerated, and in HCV monoinfected patients. Furthermore, liver biopsies per se have sampling variability limitations36 and fat deposition is not homogenous throughout the liver. Only two studies specified that liver biopsies had to be higher than 10 mm in length25, 26; the amount of alcohol consumption allowed was different among studies, both of which further adds to the expected bias. Moreover, whenever heavy alcohol intake was an exclusion criterion no information could be drawn regarding the potential effect of mild alcohol consumption on HS.
Overall, HS was present in half of the patients, which was not different from the HCV monoinfected patients. That was not expected because HIV factors, and most important ART, have theoretical prosteatogenic effects. It can be hypothesized that HCV steatogenic effects are so strong that they supersede any HIV-related effects. Also, in the pre-HAART era, HS in patients with HIV was not different from the general population8 and those patients may represent some bias relating to HIV status and antiretroviral drugs exposition because they were selected from patients who were proposed for HCV treatment.
On the issue of demographic factors, HS prevalence did not show gender-related differences. Although nonalcoholic fatty liver disease (NAFLD) related to metabolic syndrome seems to associate with male gender,37 data in HCV4, 38 and HIV10, 11, 39 monoinfected patients are inconsistent. It is possible that when virus and drugs are added to metabolic factors, protection conferred by female sex is lost. The protective effect of African-American ethnicity described in NAFLD,40 probably related to lower visceral adipose tissue,41 was not found. Indeed, fat accumulates ectopically in response to HIV therapy, which may supersede those differences.
Concerning risk factors for HS in HIV/HCV-coinfected patients, metabolic factors were the most important, particularly DM, which doubled the risk of HS (OR 2.32), similar to NAFLD in the general population.42 Higher BMI was also a risk factor for HS in this review, although with a small effect (OR 1.13). This could be due to the fact that HIV infection after the introduction of HAART has transformed in a chronic illness increasingly accompanied by obesity,43 with HIV-infected individuals approaching weight levels seen in the general population, with a 14% obesity prevalence and 31% overweight.44 Even more important than obesity as a risk factor for HS is central obesity and ectopic fat, with a spillover of fatty acids that can then accumulate in the liver.45 One condition associated with ectopic fat is lipodystrophy, which affects 60%-80% of HIV patients under HAART.46 In fact, the only study evaluating the relation between lipodystrophy and HS found a strong association (OR 3.02).21 Dyslipidemia, particularly hypertriglyceridemia, was not associated with HS prevalence, although hypertriglyceridemia is considered a risk factor for fatty liver disease.47 In comparison to the general population, HIV-infected patients present a dismal lipid profile, with hypertriglyceridemia and hypercholesterolemia, particularly higher LDL and lower HDL cholesterol levels,48 which worsens in patients under ART.49 However, HCV infection, in particular with genotype 3, has a beneficial influence on antiretroviral-related lipid changes,50 as HCV disturbs the assembly and excretion of lipid complexes from the hepatocyte,51 thus potentially explaining the lack of association between dyslipidemia and HS in HIV/HCV-coinfected patients.
HCV-related factors, namely, genotype 3 or viral load, did not associate with HS, possibly due to the low overall prevalence of genotype 3 (20%), probably resulting from the fact that the majority of the cohorts come from the United States, where an 8% prevalence is estimated.52 In contrast, genotype 3 European prevalence varies from 10% to more than 50% in Eastern countries.53 In conformity, a subanalysis of the European studies with higher genotype 3 prevalence did in fact demonstrate a significant association with HS (OR 2.5).
Similarly, HIV-related factors, namely, viral load, CD4 cell count, ART, and class of medication, were not associated with HS. In the analyzed studies the same parameters were differently evaluated, with different cutoffs, allowing a meta-analysis only in a limited number of studies and decreasing its power to analyze that particular data.
In conclusion, HS was present in half of HIV/HCV-coinfected patients and does not seem to be more frequent when compared to HCV monoinfected patients. In coinfected patients, HS is associated with metabolic factors, namely, DM and BMI, and with more severe liver disease (necroinflammation and fibrosis). The fact that no associations with HCV factors were found may be due to the small percentage of patients with genotype 3 HCV infection, because a strong association was found in the studies with higher genotype 3 prevalence.
We thank Prof. Maria E. Camilo and Dr. Eunice Guerra for their excellent support in reviewing the article.