Liver stiffness predicts clinical outcome in human immunodeficiency virus/hepatitis C virus-coinfected patients with compensated liver cirrhosis

Authors


  • Potential conflict of interest: Nothing to report.

Abstract

Our aim was to assess the predictive value of liver stiffness (LS), measured by transient elastography (TE), for clinical outcome in human immunodeficiency virus / hepatitis C virus (HIV/HCV)-coinfected patients with compensated liver cirrhosis. This was a prospective cohort study of 239 consecutive HIV/HCV-coinfected patients with a new diagnosis of cirrhosis, done by TE, and no previous decompensation of liver disease. The time from diagnosis to the first liver decompensation and death from liver disease, as well as the predictors of these outcomes, were evaluated. After a median (Q1-Q3) follow-up of 20 (9-34) months, 31 (13%, 95% confidence interval [CI]: 9%-17%) patients developed a decompensation. The incidence of decompensation was 6.7 cases per 100 person-years (95% CI, 4.7-9-6). Fourteen (8%) out of 181 patients with a baseline LS < 40 kPa developed a decompensation versus 17 (29%) out of 58 with LS ≥ 40 kPa (P = 0.001). Factors independently associated with decompensation were Child-Turcotte-Pugh (CTP) class B versus A (hazard ratio [HR] 7.7; 95% CI 3.3-18.5; P < 0.0001), log-plasma HCV RNA load (HR 2.1; 95% CI 1.2-3.6; P = 0.01), hepatitis B virus coinfection (HR, 10.3; 95% CI, 2.1-50.4; P = 0.004) and baseline LS (HR 1.03; 95% CI 1.01-1.05; P = 0.02). Fifteen (6%, 95% CI: 3.5%-9.9%) patients died, 10 of them due to liver disease, and one underwent liver transplantation. CTP class B (HR 16.5; 95% CI 3.4-68.2; P < 0.0001) and previous exposure to HCV therapy (HR 7.4; 95% CI 1.7-32.4, P = 0.007) were independently associated with liver-related death; baseline LS (HR 1.03; 95% CI 0.98-1.07; P = 0.08) was of borderline significance. Conclusion: LS predicts the development of hepatic decompensations and liver-related mortality in HIV/HCV-coinfection with compensated cirrhosis and provides additional prognostic information to that provided by the CTP score. (HEPATOLOGY 2012;56:228–238)

Endstage liver disease (ESLD) due to hepatitis C virus (HCV) chronic infection is a leading cause of mortality in human immunodeficiency virus (HIV)-infected individuals in Western countries.1-3 Previous cohort studies have shown that the natural history of HCV-related ESLD is accelerated in the setting of HIV infection4, 5 and that survival is dramatically low once decompensations occur.6 Given that liver transplantation is often the only therapeutic option at this stage, earlier recognition and optimal management of cirrhosis at initial stages are critical.

The assessment of liver stiffness (LS) by transient elastography (TE) has been added to routine daily clinical care of HIV/HCV-coinfected patients in some countries in Europe. TE accurately predicts the presence of fibrosis and cirrhosis in several clinical settings,7-12 including HIV/HCV-coinfection.13-15 Besides, LS provides additional information in the setting of ESLD. Thus, studies conducted in HCV-monoinfected16-19 and HIV/HCV-coinfected subjects20 have demonstrated that the presence of esophageal varices can be predicted by LS. Moreover, LS correlates with portal hypertension, the hallmark of the evolution of chronic liver disease, in patients with HCV-related cirrhosis,21, 22 including those infected by HIV.23 Due to this, it is reasonable to speculate that LS might be a predictor of clinical decompensations or mortality in HIV/HCV-coinfected patients with cirrhosis.

A single study has evaluated the impact of LS on survival in HIV/HCV-coinfected patients with cirrhosis.24 In that study, LS predicted overall mortality, but an analysis of the impact on specific liver-related mortality was not shown. Additionally, multivariate analyses of the predictors of overall mortality were not adjusted by some relevant factors such as Child-Turcotte-Pugh (CTP) or the model for endstage liver disease (MELD) scores. Thus, additional data regarding the impact of LS on survival in HIV/HCV-coinfected patients with ESLD are needed. Besides, LS should be evaluated as a surrogate marker of liver decompensations because it might add prognostic information to the one provided by CTP or MELD scores.

The objective of our study was to assess the predictive value of LS, measured by TE, for clinical outcome in HIV/HCV-coinfected patients with compensated liver cirrhosis.

Abbreviations

APRI, AST platelet ratio index; ART, antiretroviral therapy; AUROC, area under the receiver operating characteristic curve; CI, confidence interval; CTP, Child-Turcotte-Pugh; ESLD, endstage liver disease; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; HE, hepatic encephalopathy; HIV, human immunodeficiency virus; HR, hazard ratio; HRS, hepatorrenal syndrome; IDI, integrated discrimination improvement; LS, liver stiffness; MELD, model for endstage liver disease; NPV, negative predictive value; PHGB, portal hypertensive gastrointestinal bleeding; PPV, positive predictive value; ROC, receptor operating characteristic; SBP, spontaneous bacterial peritonitis; SVR, sustained virological response; TE, transient elastography.

Materials and Methods

Study Design and Patients.

This was a multicenter prospective cohort study conducted in seven hospitals in Andalusia, southern Spain. In February 2006, a prospective cohort of HIV/HCV-coinfected with compensated liver cirrhosis, diagnosed on the basis of LS, was created. From this date, all consecutive HIV-infected patients attending the participant hospitals were enrolled in this cohort if they met the following criteria: (1) HCV coinfection with detectable plasma HCV RNA at inclusion; (2) new diagnosis of liver cirrhosis based on the presence of LS > 14 kPa as measured by TE; (3) no evidence of metabolic or autoimmune liver disease according to clinical history, appropriate laboratory tests, and, when available, histological examination; and (4) No decompensation of liver disease before entering the cohort. Subjects who presented with a liver decompensation or hepatocellular carcinoma (HCC) at the time of cirrhosis diagnosis were excluded. HCC and decompensations of cirrhosis, which included portal hypertensive gastrointestinal bleeding (PHGB), ascites, hepatorrenal syndrome (HRS), spontaneous bacterial peritonitis (SBP), and hepatic encephalopathy (HE), were diagnosed according to criteria stated elsewhere.3, 4, 25

As stated above, cirrhosis was diagnosed by TE when LS ≥ 14 kPa was present. This threshold has been demonstrated to accurately predict the presence of cirrhosis in HIV/HCV-coinfected patients, with a reported area under the receiver operating characteristic curve (AUROC) and a positive predictive value (PPV) for the diagnosis of cirrhosis of 0.95% and 86%, respectively.14 TE examinations were performed by a single experienced operator in each center using the M-probe. In 12 (5%) patients who were overweight and an invalid LS measurement with the M-probe, the XL-probe was used.

Follow-up.

During follow-up, all individuals enrolled in the cohort were managed according to a specific protocol of care created by the investigator team. Thus, patients were evaluated at least every 6 months. In each visit, an assessment of symptoms and signs of HIV disease or hepatic decompensation was performed and routine hematological, immunological, virological, and biochemical examinations were done. Plasma HIV viral load was measured using a quantitative polymerase chain reaction (PCR) assay (Cobas AmpliPrep-Cobas TaqMan HIV-1 Test, v. 2.0; Roche Diagnostic Systems, Branchburg, NJ). Plasma HCV-RNA load measurements were performed using a quantitative PCR assay according to the available technique at each institution (Cobas AmpliPrep-Cobas TaqMan; Roche Diagnostic Systems, Meylan, France: detection limit of 50 IU/mL; Cobas TaqMan; Roche Diagnostic Systems, Pleasanton, CA: detection limit of 10 IU/mL). Antiretroviral therapy (ART) was prescribed along the follow-up according to the recommendations of international guidelines.

International guidelines, reported consensus, the recommendations of the Andalusian Society of Infectious Diseases for the management of cirrhosis in HIV/HCV-coinfected patients (www.saei.org) and some expert opinions25 were used to define the liver disease management and follow-up in the cohort protocol. Briefly, ultrasound abdominal examinations for HCC screening were performed every 6 months. CTP26 and MELD27 scores were computed at baseline and then every 6 months. All patients underwent an upper endoscopy at cohort entry for screening of esophageal varices. Varices were staged following the Japanese Research Society for Portal Hypertension staging system.28 From November 2009, the investigator team modified the initial protocol and allowed sparing endoscopy in patients showing an initial LS < 21 kPa, as the negative predictive value (NPV) of this cutoff value for the presence of esophageal varices requiring therapy in HIV/HCV-coinfected patients is 100%.20 Liver decompensations (PHGB, ascites, HRS, SBP, HE) and HCC were diagnosed and managed according to criteria stated elsewhere.3, 4, 25 Liver transplantation was considered according to the current recommendations in Spain.25 Finally, therapy against HCV was offered during follow-up according to the physician criteria and current guideline recommendations.29

Patients were prospectively seen until death, liver transplant, or the censoring date (January 31 2011). Vital status and causes of death were established from database and clinical records. Patients lost to the follow-up or their next of kin were contacted by way of telephone whenever possible.

Statistical Analysis.

Continuous variables are expressed as median (Q1-Q3) and survival times as mean (standard deviation [SD]). Categorical variables are presented as numbers (percentage; 95% confidence interval [CI]). Survival estimates at different timepoints are expressed as the cumulative proportion of survivors at the end of the period. Comparisons between continuous variables were made using Student's t test or Mann-Whitney U test, depending on the normality of distributions. Comparisons between categorical variables were made by the chi-square test or Fisher's test, when appropriate.

The primary endpoint of the study was the emergence of a first episode of hepatic decompensation and/or HCC. Secondary endpoints were death of any cause and liver-related death. The relationship between the time to the emergence of these endpoints and the following variables was analyzed: baseline age, sex, self-reported daily alcohol consumption, hepatitis B and D coinfections, HCV genotype, baseline plasma HCV viral load, previous therapy against HCV before enrolment, therapy against HCV and type of response during follow-up, Centers for Disease Control and Prevention clinical stage, CD4 cell counts, baseline plasma HIV viral load, percentage of follow-up with undetectable plasma HIV viral load, ART during follow-up, baseline CTP and MELD scores, baseline platelet count, baseline AST platelet ratio index (APRI),30 and FIB4 scores,31 presence at baseline of esophageal varices at risk of bleeding and baseline LS.

The baseline timepoint was considered to be the date of the first TE examination diagnostic of cirrhosis. The time-to-event was computed as the months elapsed from this timepoint to the different endpoints. Kaplan-Meier estimates of the cumulative probability of survival were built and survival curves were compared using the log-rank test. For these analyses continuous variables were categorized according to the median value or cutoff points considered clinically relevant. Namely, the CTP score was divided into class A (5-6 points), B (7-9 points), or C (10-15 points) and MELD score was categorized using a cutoff value of 14. For LS and HCV viral load, the cutoff point with the best sensitivity and specificity as predictor of the emergence of decompensations was selected using receptor operating characteristic (ROC) curves. Additionally, LS was also categorized by other clinically relevant cutoff points. Those variables with a P ≤ 0.1 on univariate analyses were entered in Cox proportional hazard models. Also, age, sex, and the achievement of sustained virological response (SVR) during follow-up were also included in these models. Finally, the presence of statistical interactions between LS, CTP, and MELD scores was evaluated by means of multivariate Cox regression analyses. Associations with P < 0.05 were considered significant. The hazard ratio (HR) and the respective 95% CIs were calculated. Comparisons between AUROC were performed using the Hanley and McNeil test. Also, the integrated discrimination improvement (IDI) was computed to compare the ability of the models to predict outcomes.32 Likewise, the sensitivity, specificity, PPV, and NPV were calculated. The statistical analysis was carried out using the SPSS 19 Statistical Software Package (Chicago, IL) and STATA v. 9 (StataCorp, College Station, TX).

Ethical Aspects.

The study was designed and conducted following the Helsinki Declaration. The Ethics Committee of the Hospital Universitario de Valme approved the study. All the participant subjects gave written consent to participate in the study.

Results

Patient Features.

In all, 239 patients were included in this study. The median (Q1-Q3) follow-up at the end of the study period was 20.7 (range: 9.5-34.5) months. Twelve (5%) patients were lost to the follow-up. Fifty-eight (24%) patients had previously undergone a liver biopsy, with a median (Q1-Q3) elapsed time before enrolment of 37 (range: 26-62) months. The median (Q1-Q3) elapsed time from last clinical visit with evidence of lack of cirrhosis before enrolment was 9 (range: 3-27) months. The main characteristics of the study population are summarized in Table 1. At baseline, 223 (93%) patients had clinical, ultrasound, or histological data supporting the diagnosis of cirrhosis established by TE. Thirty-nine (16%) patients had previously received therapy against HCV without achieving SVR. The median (Q1-Q3) elapsed time between previous therapy and enrolment was 48 (range: 32-91) months. Of these 39 patients, 21 were nonresponders and nine were relapsers, whereas the remaining nine patients had discontinued previous anti-HCV therapy voluntarily or due to adverse effects.

Table 1. Characteristics of the Study Population (n=239)
ParameterValue
  • CDC: Centers for Disease Control and Prevention; ART: antiretroviral therapy; MELD: model for endstage liver disease.

  • *

    Median (Q1-Q3)

  • available in 235 patients; available in

  • 224 and

  • §

    228 patients.

Age (years)*44 (41-48)
Male gender, no. (%)215 (90)
Daily alcohol intake > 50 g/day, no. (%)30 (12)
Previous intravenous drug users, no. (%)217 (91)
Previous therapy against HCV, no. (%)39 (16)
HCV genotype, no. (%) 
 1152 (64)
 24 (2)
 351 (22)
 428 (12)
HCV RNA load (log10 IU/mL)*6.16 (5.68-6.67)
Positive hepatitis B virus surface antigen, no. (%)11 (4.5)
Serum alanine aminotransferase (IU/L)*74 (48-109)
Serum aspartate aminotransferase (IU/L)*76 (53-108)
Total bilirubin (mg/dL)*0.8 (0.6-1.3)
Platelet count (/mm3)*119.000 (88.000-160.500)
HIV RNA load < 50 copies/mL, no. (%)166 (70)
CD4 cells/mL*406 (247-615)
CDC C stage, no. (%)17 (7)
ART, no. (%)216 (90)
MELD score*7 (6-9)
Child-Turcotte-Pugh class, no. (%) 
 A215 (90)
 B24 (10)
Liver stiffness (kPa)*26 (18-38)
Liver stiffness (kPa) 
 < 2193 (39)
 21-39.988 (37)
 > 4058 (24)
Esophageal varices, no. (%)§ 
 F0-F1213 (93)
 F2-F315 (7)

During follow-up, 93 (39%) patients started therapy against HCV with pegylated interferon plus ribavirin. At the end of study period, 11 out of these 93 patients were still receiving such therapy. SVR was achieved in 19 (23%) of the remaining 82 patients.

Hepatic Decompensations.

Thirty-one (13%, 95% CI: 9%-17%) patients developed a hepatic decompensation and/or HCC during the follow-up. The density of incidence of hepatic decompensation and/or HCC was 6.7 per 100 person-years (95% CI: 4.7%-9.6%). The probability of developing a decompensation and/or HCC at 1 year, 2 years, and 3 years was 7%, 10%, and 15%, respectively. Ascites was the most common type of first decompensation, appearing in 21 (68%) patients. PGHB occurred in four (13%) patients, HE in one (3%), SBP in one (3%), and HRS in one (3%). HCC was diagnosed in three (10%) patients. The median (Q1-Q3) time to development of liver decompensation and/or HCC was 19 (range: 3-37) months. Two patients of those who developed ascites as the initial decompensation presented a concomitant clinical event, PGHB in one case and HE in the other.

The factors associated with the emergence of a hepatic decompensation and/or HCC in univariate analyses are shown in Table 2. Higher baseline LS values were associated with developing a first hepatic event and/or HCC (Table 2, Fig. 1A). The probability of remaining free of decompensations and HCC at 1 year, 2 years, and 3 years was 97%, 93%, and 81%, respectively, for patients with a baseline LS < 40 kPa, whereas it was 80%, 74%, and 63% for patients with a baseline LS above or equal to 40 kPa. Figure 1B and Supporting Fig. 1A,B show the probability of remaining free of a decompensation and HCC according to baseline CTP stage, therapy against HCV during follow-up and MELD score, respectively. Figure 1C,D shows the probability of this event according to different categories of LS combined with baseline CTP stage and MELD score.

Table 2. Associations Between Emergence of an Episode of Hepatic Decompensation and/or HCC and Other Parameters (n=239)
ParameterCategoryNo. (%) with DecompensationP Bivariate (Log Rank Test)Multivariate Hazard Ratio (95% CI)P Multivariate
  • CDC: Centers for Diseases Control and Prevention; APRI: AST platelet ratio index; MELD: model for endstage liver disease.

  • *

    Considered as a continuous variable for multivariate analyses;

  • Eleven patients who were still receiving therapy at the end of follow-up were excluded from this univariate analysis.

  • Considered as a dichotomous variable (therapy vs. no therapy) for multivariate analyses;

  • §

    Considered as a dichotomous variable (>50.000 vs. ≥ 50.000) for multivariate analysis;

  • As an statistical interaction between Child-Turcotte-Pugh and MELD score was detected, the multivariate model was stratified according to baseline MELD score (< 14 vs. ≥ 14).

SexMale27 (12)0.472.1 (0.5-8.2)0.3
Female4 (17)   
Age< 44 years14 (13)0.841.06 (0.9-1.1)*0.06
≥ 44 years17 (13)   
Hepatitis B virus surface antigenNegative28 (12)   
Positive3 (33)0.110.3 (2.1-50.4)0.004
HCV genotype1-422 (12)   
2-38 (14)0.21 
HCV RNA load (log10 IU/mL)< 6.720 (11)   
≥ 6.711 (21)0.0012.1 (1.2-3.6)*0.01
Previous therapy against HCVNo25 (12)   
Yes6 (15)0.9 
Therapy against HCV during follow-upNo therapy18 (12)   
Therapy without SVR10 (16)0.41.8 (0.6-5.6)0.8
SVR3 (16)   
CDC stageA-B28 (13)   
C3 (18)0.13.8 (0.6-22.5)0.2
CD4 cell count (cells/mm3)< 2009 (18)0.061.1 (0.3-3.5)0.2
≥ 20020 (11)   
Percentage of follow-up with plasma HIV-RNA load below the detection level< 604 (16)0.13.7 (0.9-14.1)0.06
≥ 6021 (11)   
Esophageal varicesF0-F125 (12)0.9  
F2-F32 (12)  
Platelet count (/mm3)< 50.0005 (33)0.051.1 (0.2-4.4)§0.6
50.000-99.9996 (10)   
≥ 100.00019 (12)   
APRI score< 1.513 (12)   
≥ 1.516 (16)0.8 
FIB4 score< 3.2510 (8)   
≥ 3.2519 (16)0.2 
Child-Turcotte-Pugh classA17 (8)   
B14 (58)< 0.00017.7 (3.3-18.5)< 0.0001
MELD score< 1425 (11)   
≥ 144 (40)0.0004 
Liver stiffness (kPa)< 4014 (8)   
≥ 4017 (29)0.00071.03 (1.01-1.05)*0.02
Figure 1.

Probability of remaining free of developing a hepatic decompensation and/or hepatocellular carcinoma according to (A) baseline liver stiffness, (B) baseline CTP stage, (C) different categories of baseline liver stiffness and CTP stage, and (D) different categories of baseline liver stiffness and MELD score. LS, liver stiffness; CTP, Child-Turcotte-Pugh; MELD, model for endstage liver disease.

Multivariate Cox regression analyses demonstrated a statistical interaction between CTP and MELD scores. On the contrary, there were no interactions between CTP stage and LS or MELD score and LS. Thus, Cox regression analyses were stratified by the baseline MELD score. After multivariate analyses, baseline plasma HCV viral load, hepatitis B virus coinfection, CTP stage, and LS were independently associated with developing a liver decompensation and/or HCC (Table 2). Additionally, we compared the ability of LS, CTP stage, and MELD score to predict the primary outcome. Thus, multivariate models including either LS or CTP stage or MELD score were done and their respective AUROC computed. The AUROC (95% CI) for CTP stage including multivariate model was 0.76 (range: 0.65-0.88), whereas the respective figures for LS including multivariate model and MELD score including multivariate model were 0.72 (range: 0.62-0.82) and 0.71 (range: 0.61-0.81) (P > 0.2 for all comparisons). Finally, assessment of the IDI (i.e., the average improvement in the predicted probability of decompensation) indicated that the LS and the MELD including models yielded a better performance than the CTP one. Thus, the net improvement of the LS model versus the CTP one was 11% (P = 0.01), whereas the MELD model improved the CTP one by 9% (P = 0.02). LS and MELD models showed a similar performance, as the respective figure for the comparison between them was 1.8% (P = 0.4).

As the CTP score showed a strong impact on the emergence of decompensations, we performed analyses restricted to 215 patients harboring class A CTP stage at baseline. In these patients, a higher baseline LS tended to be associated with the occurrence of liver events. Namely, 10 (6%) out of 171 patients with an LS < 40 kPa developed a hepatic decompensation versus 7 (16%) out 44 with an LS ≥ 40 kPa (P = 0.1). Unfortunately, the relative low number of events precluded to perform reliable multivariate analyses.

Liver-Related Mortality.

Fifteen (6%, 95% CI: 3.5%-9.9%) patients died during follow-up. The mortality rate was 3.6 deaths per 100 person-years (95% CI: 2.2%-5.8%). In 10 patients, death was liver-related. Two patients died due to HE, two due to PHGB, two due to HRS, two due HCC, one due to SBP, and one due to liver failure following portal thrombosis. Five patients died to non liver-related causes: two patients due to cerebral bleeding, one patient due to a non-acquired immune deficiency syndrome (AIDS)-related neoplasm, one patient due to bacteremic pneumococcal pneumonia, and one patient suffered from sudden death. One patient underwent liver transplant. Thus, 11 patients, 10 with ESLD-related deaths and one undergoing a liver transplant, developed a liver-related death and/or transplantation. The rate of this outcome was 2.4 per 100 person-years (95% CI: 1.4%-4.4%).

Higher baseline LS values were associated with developing a liver-related death and/or transplantation (Table 3, Fig. 2). Liver-related mortality and/or transplantation tended to be lower in those patients who achieved SVR during follow-up (Table 3). Cox regression analyses did not yield statistical interactions between LS, CTP stage, and MELD score. Thus, these variables were included simultaneously into multivariate models. After multivariate analyses, baseline LS, CTP stage, and previous exposure to anti-HCV therapy before enrolment were independently associated with liver-related mortality and/or transplantation (Table 3).

Table 3. Associations Between Liver-Related Mortality and/or Liver Transplantation and Other Parameters (n=239)
ParameterCategoryNo. (%) with Liver-Related DeathsP Bivariate (Log Rank Test)Multivariate Hazard Ratio (95% CI)P Multivariate
  • CDC: Centers for Disease Control and Prevention; APRI: AST platelet ratio index; MELD: model for endstage liver disease.

  • *

    Considered as a continuous variable for multivariate analyses;

  • Eleven patients who were still receiving therapy at the end of follow-up were excluded from this univariate analysis;

  • Considered as a dichotomous variable (>50.000 vs. ≥ 50.000) for multivariate analysis.

SexMale10 (5)0.91.5 (0.2-12.8)0.7
Female1 (4)   
Age< 44 years5 (5)0.90.9 (0.8-1.1)*0.7
≥ 44 years6 (5)   
Hepatitis B virus surface antigenNegative10 (4)   
Positive1 (9)0.3 
HCV genotype1-49 (5)0.7 
2-32 (4)   
HCV RNA load (log10 IU/mL)< 6.76 (3)   
≥ 6.75 (10)0.011.4 (0.6-2.8)*0.4
Previous therapy against HCVNo7 (3)   
Yes4 (10)0.067.4 (1.7-32.4)0.007
Therapy against HCV during follow-upNo Therapy8 (5)   
Therapy without SVR3 (5)0.2 
SVR0 (0)   
Sustained virological response during follow-upNo11 (5)0.2  
Yes0 (0)  0.9
CDC stageA-B11 (5)0.4 
C0 (0)   
CD4 cell count (cells/mm3)< 2007 (4)   
≥ 2003 (6)0.4 
Percentage of follow-up with plasma HIV-RNA load below the detection level< 601 (4)0.5 
≥ 605 (3)   
Esophageal varicesF0-F19 (4)   
F2-F31 (7)0.8 
Platelet count (/mm3)< 50.0003 (20)0.02  
50.000-99.9992 (3) 1.9 (0.4-9.6)0.07
≥ 100.0006 (4)   
APRI score< 1.55 (8)0.9  
≥ 1.56 (5)  
FIB4 score< 3.254 (3)   
≥ 3.257 (6)0.4 
Child-Turcotte-Pugh classA6 (3)   
B5 (21)< 0.000116.5 (3.4-68.2)< 0.0001
MELD score< 149 (4)   
≥ 142 (20)0.0030.9 (0.7-1.4)*0.7
Liver stiffness (kPa)< 405 (3)   
≥ 406 (10)0.021.03 (0.98-1.07)*0.08
Figure 2.

Probability of liver-related death or liver transplantation according to baseline liver stiffness (LS).

Overall Mortality.

Baseline LS was associated with overall mortality (Supporting Fig. 2). Table 4 shows univariate and multivariate analyses of the predictors of overall mortality. After multivariate analyses, female sex, hepatitis B coinfection, previous therapy against HCV, therapy against HCV during follow-up, the percentage of follow-up with undetectable plasma HIV viral load, baseline platelet count, and MELD score were independently associated with death of any cause or liver transplantation.

Table 4. Associations Between Death of Any Cause or Liver Transplantation and Other Parameters (n=239)
ParameterCategoryNo. (%) with Liver-Related DeathsP Bivariate (Log Rank Test)Multivariate Hazard Ratio (95% CI)P Multivariate
  • CDC: Centers for Disease Control and Prevention; APRI: AST platelet ratio index; MELD: model for endstage liver disease.

  • *

    Considered as a continuous variable for multivariate analyses;

  • Eleven patients who were still receiving therapy at the end of follow-up were excluded from this univariate analysis;

  • Considered as a dichotomous variable (therapy vs. no therapy) for multivariate analyses;

  • §

    Considered as a dichotomous variable (>50.000 vs. ≥ 50.000) for multivariate analysis,

  • As an statistical interaction between Child-Turcotte-Pugh and MELD score was detected, the multivariate model was stratified according to baseline Child-Turcotte-Pugh score (A5 vs. ≥ A6).

SexMale13 (6)   
Female3 (12)0.23.9 (0.9-16.6)0.06
Age< 44 years7 (7)0.91.01 (0.9-1.1)*0.8
≥ 44 years9 (7)   
Hepatitis B virus surface antigenNegative14 (6)   
Positive2 (18)0.15.1 (0.9-26.4)0.05
HCV genotype1-413 (7)0.7  
2-33 (5)  
HCV RNA load (log10 IU/mL)< 6.78 (4)   
≥ 6.78 (16)0.0091.4 (0.7-2.3)*0.2
Previous therapy against HCVNo11 (5)   
Yes5 (13)0.15.7 (1.7-19.7)0.005
Therapy against HCV during follow-upNo Therapy14 (8)   
Therapy without SVR2 (4)0.070.18 (0.03-1.1)0.05
SVR0 (0)   
Sustained virological response during follow-upNo16 (7)0.6  
Yes0 (0)  
CDC stageA-B15 (7)1.0  
C1 (6)  
CD4 cell count (cells/mm3)< 2004 (8)   
≥ 20011 (6)0.5 
Percentage of follow-up with plasma HIV-RNA load below the detection level< 605 (16)0.044.8 (1.5-15.6)0.008
≥ 6011 (5)   
Esophageal varicesF0-F114 (6)   
F2-F31 (7)1.0 
Platelet count (/mm3))< 50.0004 (27)0.016.6 (1.9-23.1)§0.003
50.000-99.9993 (5)   
≥ 100.0009 (5)   
APRI score< 1.57 (7)0.9  
≥ 1.59 (7)  
FIB4 score< 3.256 (5)   
≥ 3.2510 (8)0.3 
Child-Turcotte-Pugh classA9 (4)   
B7 (29)< 0.0001
MELD score< 1412 (5)   
≥ 144 (40)0.0031.3 (1.1-1.5)*0.002
Liver stiffness (kPa)< 409 (5)   
≥ 407 (12)0.070.9 (0.8-1.1)*0.7

The performance of LS for predicting clinical outcomes are listed in Table 5.

Table 5. Performance of Different Categories of Baseline Liver Stiffness for Predicting Clinical Outcomes (n=239)
 AUROC (95% CI)Sensitivity (%)Specificity (%)PPV (%)NPV (%)
  1. AUROC: area under receiver operating characteristic; CI: confidence interval; PPV: positive predictive value; NPV: negative predictive value.

Hepatic decompensation and/or HCC0.718 (0.613-0.824)    
 LS < 21 kPa 84411793
 LS < 40 kPa 55813092
Liver-related mortality and/or liver transplantation0.728 (0.562-0.894)    
 LS < 21 kPa 8140698
 LS < 40 kPa 54781197
Overall mortality0.602 (0.426-0.779)    
 LS < 21 kPa 6239793
 LS < 40 kPa 44781395

Discussion

This is the first study, to our knowledge, that has demonstrated that LS predicts the emergence of clinical complications in patients with compensated liver cirrhosis due to HCV. Thus, whereas only 8% of patients with a baseline LS below 40 kPa developed a decompensation and/or HCC during follow-up, the respective figure for those individuals with a baseline LS above or equal to 40 kPa was 29%. Also, this association remained statistically significant after multivariate analyses controlling for other prognostic factors such as CTP or MELD scores. Importantly, an LS < 40 kPa accurately predicts a very low risk of decompensation or death in the mid-term. These results provided additional evidence that LS is more than a single estimation of liver fibrosis and may be a surrogate marker of liver function and portal hypertension. Previous studies have demonstrated that LS correlates well with portal hypertension in HCV-monoinfected patients,21, 22 including those coinfected by HIV.23 In addition, LS can predict the presence of esophageal varices.16-20 Finally, we have shown that LS predicts the emergence of clinical events and, consequently, may be used as a prognostic marker in HIV/HCV-coinfected patients with compensated cirrhosis. In fact, comparisons of the diagnostic performance of LS with other classical scores yielded a similar predictive ability of LS and MELD, that was slightly better than that provided by CTP. Studies addressing if composite scores using LS and classic prognostics scores, such as CTP and MELD, may improve the performance of the latter are required.

LS also predicted liver-related mortality in our study. The impact of LS on survival has been previously assessed in a single study.24 In that study, LS predicted overall mortality, but an analysis of the impact on specific liver-related mortality was not shown. Additionally, analyses of the predictors of overall mortality were not adjusted by some relevant factors such as CTP or MELD scores in that study. In our study, multivariate models yielded an association of LS with liver-related mortality that was very close to statistically significant, even when adjusting by CTP or MELD scores. These findings suggest that LS is an independent predictor of outcomes of ESLD in HIV/HCV-coinfected patients. Finally, although LS was associated with overall mortality in univariate analysis, this association did not remain statistically significant after adjusting by other parameters in multivariate analysis. The relatively low number of events, some of them not liver-related, may have precluded us to find an independent association of LS with death of any cause. However, it is reasonable not to expect that LS may be a predictor of overall mortality in the next years. In fact, liver-related deaths will probably decrease with new therapies against HCV as a consequence of higher rates of eradication of HCV. A larger follow-up of the cohort will definitely rule out or confirm if LS also predicts overall mortality.

Other prognostic factors found in this study were hepatitis B coinfection, high HCV RNA viral load, and CTP score. Also, MELD score predicted the development of liver events but its independent predictive value could not be assessed as a statistical interaction with CTP stage was found. Interestingly, MELD score was associated with overall mortality but not with liver-related mortality after multivariate analyses in our study. In fact, the predictive value of MELD score in HIV/HCV-coinfected patients remains controversial, with previous studies reporting no independent association with survival6, 33 and others finding such an association.34, 35 In our opinion, the lack of an independent association of MELD with liver-related mortality in our cohort could reflect a weaker predictive value in the long term, as is the case in this study, than in the short- and mid-term.

Achievement of SVR after treatment of hepatitis C is associated with a reduction in liver-related mortality in HIV-negative36 and HIV-positive patients.37 The impact of anti-HCV therapy on the survival or the risk of decompensations in HIV/HCV-coinfected patients with compensated cirrhosis has been only assessed in two previous cohort studies with apparent conflicting data.33, 38 In the present study, neither exposure to HCV therapy nor achieving SVR during follow-up were associated with a lower risk of developing decompensations. On the contrary, achieving SVR during follow-up tended to be associated with an improved survival in univariate analyses and exposure to therapy during follow-up was associated with a lower risk of death of any cause. However, these associations did not reach statistical significance, probably due to lack of power and insufficient follow-up. Finally, previous exposure to HCV therapy before enrolment was associated with increased mortality. In our opinion, this association probably reflects a longer time of evolution and an advanced stage of liver disease in previous nonresponder patients rather than a worrisome effect of therapy. Additionally, we cannot definitely exclude that selection bias of patients who received prior HCV therapy may have affected our results.

Our study may have some limitations. The follow-up period was somewhat short, and the number of some events, particularly liver-related deaths, was relatively low. This might have precluded identifying some potential predictors of mortality as the consecution of SVR. However, the follow-up was long enough to identify other stronger predictors of clinical outcomes such as CTP score or HBV coinfection. Even with the former limitation, this study has been able to prove that LS is an independent predictor of clinical events and liver-related mortality and that might provide additional prognostic information to the one provided by CTP or MELD scores. In our study, the diagnosis of cirrhosis was made by TE instead of biopsy. However, several studies have demonstrated that TE is highly reliable for the diagnosis of cirrhosis in HCV-monoinfected patients with or without HIV infection.13, 14, 16 Finally, we present data from a large, prospective cohort of HIV/HCV-coinfected patients with compensated cirrhosis diagnosed using the same method, TE, which avoids potential biases of previous cohorts, and prospectively followed and managed by a uniform management protocol. These are the strengths of our study.

In summary, LS, assessed by TE, is an independent predictor of the development of hepatic decompensations, HCC and liver-related mortality in HIV/HCV-coinfected patients with compensated cirrhosis, and provides additional prognostic information to that provided by CTP or MELD scores. In our opinion, this observation provides new evidence to consider incorporating sequential measurements of LS by TE to the routine daily clinical care of HIV/HCV-coinfected patients. In fact, the measurement of LS may help us to identify those patients at very high risk of decompensation and death. Also, HIV/HCV-coinfected patients bearing an LS ≥ 40 kPa should probably be seen more frequently. Finally, future studies should evaluate if LS is also an independent prognostic marker in patients with decompensated cirrhosis and if sequential assessment of LS in patients with cirrhosis results in a mortality benefit. If so, it may be added to CTP and MELD scores in the decision to consider patients to be referred to a liver transplantation program.

Acknowledgements

The authors thank Carmen Almeida from the Hospital Universitario de Valme, Seville, for helpful advice in statistical analyses.

Ancillary