Baseline prevalence and predictors of liver fibrosis among HIV-positive individuals: a substudy of the INSIGHT Strategic Timing of AntiRetroviral Treatment trial

Abstract

Introduction:

Liver disease is increasingly recognised in HIV-positive individuals , even among those without viral hepatitis, partly due to the recent availability of noninvasive methods of liver fibrosis assessment. The role of antiretroviral therapy (ART) on liver fibrosis pathogenesis is unclear.

Methods:

Sites in the Strategic Timing of AntiRetroviral Treatment (START) study with access to FibroScan^® were invited to participate in the Liver Fibrosis Progression Substudy. All substudy participants underwent FibroScan^® at baseline, and two noninvasive serum algorithms, APRI and FIB-4, were calculated. Demographic and liver-related information were collected for all START participants at baseline.

Results:

230 participants were enrolled (11.5% viral hepatitis [B or C] coinfection) of whom 221 had a valid transient elastography (TE) result. TE score was 4.9 kPa (IQR 4.3–6.0). 17 patients (7.8%) (95% CI: 5.1–12.1) had a TE score of >7.2kPa indicating significant liver fibrosis. Baseline factors associated with higher TE scores in multivariate analysis were higher ALT per 10 U/L (p=0.045), higher log₁₀ HIV RNA (p<0.001) and Hispanic/Latino ethnicity (p=0.01). TE correlated weakly with noninvasive markers.

Discussion:

At baseline significant liver fibrosis was observed in approximately 8% of participants, with higher ALT and HIV RNA the only clinical factors associated with increasing TE score. TE will be used annually to monitor fibrosis and evaluate the role of ART in further fibrosis progression.

Introduction

Liver-related disease (LRD) is one of the most common causes of morbidity and mortality in HIV-positive individuals (1, 2). Rates of LRD show little evidence of decline since the introduction of widespread combination antiretroviral therapy (ART) (3, 4), although one recent study has suggested the rate may now be decreasing (5). Much of this burden is among individuals who are coinfected with either hepatitis B (HBV) or hepatitis C virus (HCV) and current guidelines recommend early initiation of ART in both HBV- and HCV-coinfected individuals (6–9); however, there is also concern for the potential for progressive liver disease in HIV-monoinfected individuals. Abnormal liver transaminases are commonly observed in HIV-monoinfected individuals, and the recent availability of noninvasive methods for detecting and monitoring liver fibrosis has resulted in the identification of asymptomatic liver damage in these individuals. The underlying causes have been variably linked to antiretroviral hepatotoxicity (including didanosine and abacavir use) (10, 11), nonalcoholic steatohepatitis (NASH) (12), protease inhibitor (PI)-induced metabolic syndrome, the effects of HIV infection itself (13) or immune activation resulting from microbial translocation (14, 15). The role of ART on liver enzyme elevation and the benefit of ART on limiting or preventing fibrosis progression or clinical liver disease is unclear (16).

The Strategic Timing of AntiRetroviral Treatment (START) study is the first randomised controlled trial to address the benefit of starting ART at CD4 cell counts >500 cells/μL. The aim of the Liver Fibrosis Progression Substudy is to compare the effect of early versus deferred ART on liver fibrosis progression in HIV-positive individuals. Transient elastography (TE) or FibroScan^® technology was used at baseline and on an annual basis to determine both the prevalence of significant fibrosis within the START population and to evaluate factors associated with progression of fibrosis. TE is a relatively novel means of determining liver fibrosis through the use of ultrasound-based technology. A noninvasive probe is used to generate a shear wave that travels through liver tissue. The speed of this shear wave is translated into a value measured in kilopascals (kPa), which has been shown to correlate with liver fibrosis as determined by the comparator test, liver biopsy (17). This baseline analysis describes the substudy participants, the baseline distribution of fibrosis as determined by TE and noninvasive serum markers, as well as the factors associated with increased TE scores.

Methods

Participants were enrolled into the START study as described elsewhere (18). Sites with access to FibroScan^® technology were identified and invited to participate in the Liver Fibrosis Progression Substudy. At the time the trial began, FibroScan^® devices were approved in the United States for research purposes only, so only one US-based site was included. Thirty-two sites in total from Asia (Thailand), Europe (Belgium, Denmark, Germany, Italy, Portugal, Spain, and Switzerland) and Australia had access to FibroScan^® and agreed to participate. All participants gave written informed consent to enrol in this substudy. FibroScan^® was performed by a trained operator at each site. Participants were only enrolled into the substudy if a valid FibroScan^® was obtained at baseline (19). A participant’s FibroScan^® measurements were considered valid and used in this analysis if 1) at least 10 readings were obtained; 2) the percent of successful readings was greater than 60%; and 3) the ratio of interquartile range and median value for the readings obtained was less than 0.3.

Demographic and clinical data obtained in START included serum markers, data on alcohol consumption, smoking, medical history and concomitant medication use.

Statistical Analysis

TE score was natural log transformed for all regression analyses due to the non-normal distribution. Univariate and multivariate linear regression models were built to examine the association between selected baseline variables and TE score. Variables examined included age, gender, race, region of enrolment, smoking history, body mass index (BMI), viral hepatitis B and C status, CD4 cell count, HIV RNA, alanine aminotransferase (ALT), alcohol use, glucose, diabetes mellitus and statin use. Variables with a p <0.20 in the univariate analysis were included in the multivariate model.

Two serum algorithms used to predict liver fibrosis (APRI and FIB-4) were calculated as follows: APRI: [(AST/ULN)/platelet counts (10⁹ /L)] × 100,where a score of >1.5 is indicative of bridging fibrosis (F3) and a score of >2 is indicative of cirrhosis (F4), and FIB-4: (Age x AST ) / ( Platelets x (sqr ( ALT ) ) ) where a score of >3.35 is indicative of stage F3/4.Spearman correlations were calculated to assess the relationship between TE scores and both APRI and FIB-4.

Results

Two-hundred-thirty (230) START participants were enrolled into the Liver Fibrosis Progression Substudy, of whom 229 had a TE report obtained. Baseline characteristics of the substudy participants are given in Table 1. While substudy participants were generally comparable to the overall START population, key differences included a lower proportion of women (12% vs 27%), a greater proportion of Asians (33% vs 8%) and a higher proportion of viral hepatitis (B or C) coinfection (11.5% vs 6.3%) in substudy participants. Laboratory variables including ALT, AST, and platelet and CD4 cell count were similar.

Table 1: Baseline Characteristics of Liver Fibrosis Progression Substudy Participants.

	LFP Substudy
No. participants enrolled	230
Demographics
Age (years; median, IQR)	35 [28, 42]
Gender (N, % female)	28 (12.2)
Race
Black (N, %)	14 (6.1)
Hispanic/Latino (N, %)	16 (7.0)
Asian (N, %)	76 (33.0)
White (N, %)	120 (52.2)
Other (N, %)	4 (1.7)
Hepatitis Coinfection
Hepatitis B1 (%)	12 (5.4)
Hepatitis C2 (%)	14 (6.1)
Laboratory Results
CD4 (cells/μl; median, IQR)	647 [583, 758]
Nadir CD4 (cells/ μl; median, IQR)	554 [471, 654]
HIV RNA (log10copies/mL; median, IQR)	4.4 [3.8, 4.8]
HIV RNA, undetectable (N, % with RNA ≤ 400 copies/mL)	9 (3.9)
AST/SGOT (IU/L; median, IQR)	24 [20, 31]
ALT/SGPT (IU/L; median, IQR)	24 [18, 35]
Platelets (log10cells/mm3; median, IQR)	5.4 [5.3, 5.4]

¹Hepatitis B defined as HBsAg positive based on a test within one year before randomisation

²Hepatitis C defined as HCV antibody positive from any time before randomisation

Distributions of characteristics known to be associated with chronic liver disease are given in Table 2 for substudy participants. Median BMI was in the “normal” range, the prevalence of diabetes mellitus was 2%. Current smoking was reported by 37% of substudy participants but the percent reporting heavy alcohol use was low, with the majority (75%) of participants drinking alcohol four or less drinks a day for three or less days per week. A history of alcoholism or substance dependence was reported in <1% and no participants had a known history of cirrhosis. Potentially hepatotoxic drugs (see Table 2 for list) were used in less than 3% of the cohort.

Table 2: Baseline characteristics associated with liver fibrosis in Liver Fibrosis Substudy participants.

BMI (kg/m 2 ; median, IQR)	22.9 [20.9, 25.2]
Smoking Status (N,%)
Current	85 (37.0)
Former	27 (11.7)
Never	118 (51.3)
Alcohol Consumption - Days Per Week (N,%)
0	77 (34.7)
< 1	78 (35.6)
1 – 3	44 (19.2)
4 – 7	23 (10.5)
Alcohol Consumption - Drinks at a Time (N,%)
0	51 (22.8)
1	41 (18.3)
2 – 4	93 (42.0)
5 – 7	24 (11.0)
>8	13 (5.9)
Medical History (N,%)
Cirrhosis	0 (0.0)
Chronic liver disease (excluding infectious hepatitis)	1 (0.4)
Hepatic steatosis	0 (0.0)
Diabetes mellitus	5 (2.2)
Alcoholism or substance dependence	2 (0.9)
Concomitant Treatments at Baseline (N,%)
Fibric acid	1 (0.4)
Statins	2 (0.9)
Anti-tuberculous drugs	0 (0.0)
Drug treatment for hepatitis C	0 (0.0)
Anabolic steroids	0 (0.0)
Systemic corticosteroids	1 (0.4)
Systemic antifungal drugs	1 (0.4)

Transient elastography results

TE results were obtained for 229 participants, however, 8 (3.5 %) participants did not have scans that met the criteria due to an IQR/median ratio of >0.3. Therefore 221 results were used to calculate fibrosis scores. Median TE score was 4.9 kPa (IQR 4.3–6.0) (Figure 1). Breakdown of TE score by category is given in Table 3. The majority had TE scores consistent with Metavir (20) stage F0-F1 (92.3%). In total 17 participants (7.8%) had TE of >7.2kPa indicating significant liver fibrosis (95% CI 5.1–12.1). Of these, 3 (1.4%) participants had TE scores consistent with stage F3 fibrosis (TE 9.6–10.8kPa) and 3 (1.4%) had scores consistent with cirrhosis (15kPa, 16kPa and 36 kPa).

Figure 1. Distribution of TE scores.

Table 3: Distribution of fibrosis by elastography and noninvasive markers.

TE Results
Median score (kPa; median, IQR)	4.9 [4.3, 6.0]
Median score (kPa; mean ± SD)	5.4 ± 2.6
Category 1: < 7.1 kPa (N,%)	204 (92.3)
Category 2: 7.2–9.4 kPa (N,%)	11 (5.0)
Category 3: 9.5–14.0 kPa (N,%)	3 (1.4)
Category 4: > 14.0 kPa (N,%)	3 (1.4
Log median score (Log kPa; mean ± SD)	1.6 ± 0.3
IQR of TE score (kPa; median, IQR)	0.7 [0.4, 1.0]
Noninvasive serum markers
FIB-4 (median, IQR)	0.72 [0.54, 1.00]
≤ 1.45 (N,%)	199 (89.6)
1.55–3.25 (N,%)	20 (9.0)
> 3.25 (N,%)	3 (1.4)
APRI score (median, IQR)	0.27 [0.20, 0.35]
≤ 0.5 (%)	204 (91.5)
0.5–1.5 (%)	17 (7.6)
> 1.5 (%)	2 (0.9)

Factors associated with higher TE score are displayed in Table 4. In univariate analysis, HCV coinfection, HIV RNA and ALT were associated with higher fibrosis score. In addition, Hispanic/Latino participants had on average higher TE scores than white participants. Age, known duration of HIV infection and trigylcerides were not associated with higher TE score. In multivariate analysis the only baseline factors associated with a higher fibrosis score were higher ALT per 10U/L (coefficient 0.01, p=0.045) a higher log₁₀ HIV RNA (coefficient 0.09, p=0.002), and Hispanic/Latino race (coefficient 0.21, p=0.01). When the association between HIV RNA and TE score was examined using HIV RNA as a categorical rather than continuous variable (approximate quartiles of <5000 c/ml, 5000–19999 c/ml, 20000–74,999, > 750,000c/ml) there appeared to be a threshold effect with 3.9% and 5.3% of individuals in the lowest two quartiles having significant fibrosis compared to 12.7% and 10.0% of those in the higher two quartiles.

Table 4. Factors associated with increasing transient elastography score in the LFP substudy.

	Univariate		Multivariate
Baseline Predictor	Coef (SE)	p-value	Coef (SE)	p-value
Race*	0.091			0.109
White (reference)	---	---	---	---
Black	0.046 (0.087)	0.600	0.001 (0.094)	0.994
Hispanic/Latino	0.209 (0.082)	0.011	0.208 (0.084)	0.014
Asian	−0.013 (0.043)	0.757	−0.028 (0.046)	0.544
Other	−0.105 (0.147)	0.474	−0.026 (0.167)	0.876
BMI (per 5 kg/m2 increase)	0.035 (0.025)	0.168	0.017 (0.026)	0.512
Hepatitis C coinfection	0.159 (0.080)	0.048	0.084 (0.087)	0.334
HIV RNA (log10 copies/mL)	0.072 (0.023)	0.002	0.077 (0.025)	0.002
ALT SGPT (per 10 U/L increase)	0.011 (0.005)	0.024	0.010 (0.005)	0.045
Alcohol intake
4–7 days a week with 2+ drinks a day versus less consumption	−0.096 (0.072)	0.187	−0.066 (0.072)	0.356
Glucose (per 10 mg/dL increase)	0.016 (0.009)	0.086	0.010 (0.011)	0.366
Diabetes	0.197 (0.131)	0.136	0.065 (0.162)	0.689
Use of statins	0.336 (0.206)	0.104	0.274 (0.214)	0.203

^*4df Chi-square p-value for race

Linear Regression of Natural log Transformed TE Score on baseline factors

Reduced Model - Including Predictors with p < 0.20 in Univariate Analysis

Serum fibrosis markers

Two serum algorithms commonly used to predict liver fibrosis (APRI and FIB-4) (21, 22) were calculated at baseline. As with TE, the median levels of both were in the normal range (median 0.27 APRI and median 0.72 FIB-4) and similar proportions of individuals were identified as having F2 stage fibrosis and above (8.5% by APRI and 10% by FIB-4). Concordance between the two serum algorithms was not always observed, 13.1% were identified as having ≥F2 by either APRI or FIB-4: 2.7% by APRI and not FIB4, 4.5% by FIB4 and not APRI and 5.9% by both. TE correlated weakly with APRI with a Spearman correlation of 0.139 (p=0.04, 95% CI 0.01, 0.27) but not with FIB-4, Spearman correlation 0.034 (p=0.63, 95% CI −0.11, 0.16). A number of significantly discordant outliers were observed (Figure 2). The participant with a TE score of 36 kPa was not characterised by either test (APRI 0.29, FIB4 0.56) as a cirrhotic. Removal of this outlier made no significant difference to the correlation (0.034, p=0.62 for FIB-4 and TE, 0.134, p=0.05 for APRI and TE). Additionally, in univariate analyses, HIV RNA was not associated with FIB-4 (p=0.37) or APRI (p=0.26).

Figure 2. A and B: Scatter plot of TE score with (A) APRI (B) FIB-4.

Discussion

The role and contribution of HIV viraemia in the pathogenesis of hepatic fibrosis has been postulated but remains unclear. HIV-positive individuals, particularly those on ART, may have multiple reasons to develop liver disease including comorbidities such as viral hepatitis coinfection and insulin resistance, lifestyle factors such as alcohol and drug use, and the use of potentially hepatotoxic medications including antiretroviral drugs themselves. In addition, liver fibrosis is usually insidious in onset and asymptomatic in nature and to date no simple test has been able to predict its presence accurately. Transaminase values are notoriously inaccurate for assessing presence of fibrosis, and liver biopsy (the previous gold standard) is invasive, unpopular with patients and often inaccurate due to sampling error. TE (using FibroScan^®) is now increasingly accepted as a simple, noninvasive method for assessing liver fibrosis with high patient acceptability, allowing longitudinal measurements (17).

In this analysis of liver fibrosis in untreated HIV-positive individuals we demonstrate firstly a low but not insignificant (approximately 8%) prevalence of significant fibrosis. Secondly, we demonstrate in multivariate analysis that the only baseline factors associated with higher TE score are higher ALT and higher HIV RNA level, along with Hispanic/Latino race. Finally we show a poor correlation between TE score and APRI and FIB-4, the two most commonly used serum noninvasive markers of liver fibrosis.

The prevalence of liver fibrosis in untreated HIV individuals has been infrequently assessed, particularly in the absence of viral hepatitis coinfection. However, with widespread availability of highly effective ART and extended life expectancy, attention is now shifting to the contribution of non-AIDS-related conditions to morbidity and mortality in HIV-positive individuals. Recently a number of studies have attempted to quantify the presence of liver disease in HIV-positive individuals, generally using noninvasive serum algorithms such as APRI or FIB-4 (23–25). In a cross-sectional study of 432 HIV-positive individuals (without HBV or HCV), significant fibrosis as determined by APRI was found in 8% of patients (24). In another study, again in HIV-monoinfected individuals but this time using FIB-4, a score of >1.45 (consistent with significant fibrosis) was found in 15.5% of individuals (23). In our study based on elastography assessment the prevalence of significant fibrosis (TE > 7.2 kPa indicating F2 and above) in untreated HIV-positive individuals was 7.8%.

In a general population study from France of 1358 “healthy” individuals significant fibrosis was identified in 8% (26). Many factors predisposing to liver disease exist within an otherwise healthy general population including alcohol excess and obesity linked to steatohepatitis. In fact, these two mechanisms were identified as the underlying cause of liver disease in 80% of the cases of fibrosis in the French study (26). In the START cohort, traditional factors associated with liver disease were generally infrequent; BMI was within normal range in the vast majority, rates of diabetes, alcohol use and substance dependence were very low as was the use of hepatotoxic drugs. Rates of coinfection with hepatitis B or C were higher than in the general population but much lower than would be expected in a typical HIV-positive population. This likely reflects the reluctance of clinicians to enrol coinfected participants in a randomised trial of deferral of ART when guidelines recommend early initiation for both HBV- and HCV-coinfected individuals.

The only two clinical factors in this substudy that were associated with a higher TE score were baseline ALT and baseline HIV RNA. The association with ALT is expected, but not causal as elevated transaminases are known to be a marker for liver injury and are commonly, although not always, abnormal in the setting of liver fibrosis. The association with HIV RNA is important as it underpins the biological notion that HIV itself may be associated with liver inflammation and that control of HIV viraemia may be beneficial for liver health. Various biological theories support a role for HIV-infection in liver injury including the role of HIV-related microbial translocation in promoting fibrogenesis via toll like receptor signalling (27), HIV activation of hepatic stellate cells that could play a primary role in fibrogenesis, direct invasion of the hepatocyte by HIV and the role of monocyte activation and cell death (14, 15, 28). Additionally, a number of studies have supported indirectly the role of HIV in promoting hepatic damage. In a recent large matched cohort study of over 20,000 HIV-positive individuals and 215,000 HIV-negative individuals, HIV-positive individuals had a significantly greater risk of both hepatic dysfunction (as measured by serious hepatic event or laboratory markers of dysfunction) and hepatic-related death (29). Factors associated with a higher risk were not only the predicted factors of viral hepatitis coinfection, alcohol and diabetes but also lower CD4 and higher HIV RNA. Other studies have confirmed the association between greater HIV RNA and increased hepatic fibrosis as measured using the serum algorithms of APRI and FIB-4 (24, 25, 30) but none have confirmed this association using elastography.

Serum algorithms based on biochemical and clinical markers have been widely used as markers of liver fibrosis as they are easily calculated in large cohorts with clinically available data. APRI utilises AST and platelet counts, whereas FIB-4 relies on age, ALT, AST and platelet count. Both have been validated in many liver disease settings and correlate moderately well with liver biopsy and other proprietary noninvasive markers such as Fibrotest^® (31). Both also correlate reasonably well in large studies with elastography although TE outperforms serum markers against the gold standard of histology (32, 33), and the relatively poor performance of APRI and FIB-4 against histology have been noted in a previous large meta-analysis (34). In our study APRI correlated weakly with TE whereas there was no significant correlation with FIB-4. It is not clear why this should be although it is likely related to the relatively small number of participants in the study and the ability of a few outliers to distort results (Figure 1). It is also likely that HIV infection impacts some of these laboratory data, such as platelet count, skewing the results for reasons unrelated to liver injury or fibrogenesis (35). The relatively small numbers of participants in the study not only limited the potential for an association between serum markers and TE score to be observed but is also the reason why some of the other expected predictors of fibrosis, in particular the presence of viral hepatitis coinfection, were not associated with liver fibrosis.

In summary, this baseline cross-sectional analysis of untreated HIV-positive participants undergoing elastography assessment demonstrates that overall rates of significant liver fibrosis are low in this generally liver-healthy population. However an association between TE scores and HIV RNA suggests that untreated HIV viraemia may promote liver fibrogenesis and may support a rationale for earlier ART initiation. Ongoing follow-up using TE assessment annually after initiation or during deferral of ART within the randomised arms of START may help answer this question.

The START study is registered at clinicaltrials.gov (NCT00867048).