Long-term Prognostic Value of the FibroTest in Patients With Non-alcoholic Fatty Liver Disease, Compared to Chronic Hepatitis C, B, and Alcoholic Liver Disease

Mona Munteanu; Raluca Pais; Valentina Peta; Olivier Deckmyn; Joseph Moussalli; Yen Ngo; Marika Rudler; Pascal Lebray; Frederic Charlotte; Vincent Thibault; Olivier Lucidarme; An Ngo; Françoise Imbert-Bismut; Chantal Housset; Dominique Thabut; Vlad Ratziu; Thierry Poynard

Disclosures

Aliment Pharmacol Ther. 2018;48(10):1117-1127. 

In This Article

Results

Characteristics of Included Subjects

After the exclusion of 434 patients mainly due to previous transplantation, absence of baseline fibrosis biomarker and absence of follow-up, and 2965 patients with mixed or other causes of chronic liver disease, a total of 7082 patients with NAFLD (n = 1079), CHC (n = 3449), CHB (n = 2051), and ALD (n = 503) were included (Figure 1). Besides the causes of liver disease there was no significant differences between the pre-included and included subset characteristics (Table 1).

Between patients with NAFLD patients and those of the comparator group with CHC there were two main significant differences associated with the prognosis, one negatively associated, an older age (median 56.7 years old; IQR 48.4–64.7) vs 48.0 (41.0–57.5; P < 0.001), and one positively associated, the lower severity of fibrosis presumed by FibroTest (median; IQR) 0.24; 0.13–0.43 vs 0.45 (0.21–0.71) including a lower prevalence of cirrhosis 6.8% vs 22.0% (P < 0.001) (Table 1).

Survivals

The median (range) follow-up was 6.0 years (0.1–19.3), with a 15–year-overall survival of 0.725 (0.704–0.746) among the 10 047 pre-included patients (Figure 2A).

Figure 2.

Survivals. All survivals were unadjusted Kaplan-Meier curves. A, 10–year-overall survival was 0.824 (0.813–0.835) among the 7082 included patients. B, 10–year-overall survival in NAFLD was 0.689 (0.647–0.730), lower than 0.832 (0.818–0.847) in CHC, and 0.891 (0.891–0.922) in CHB, but much higher than in ALD 0.408 (0.345–0.472). All P < 0.001. C, 10 year-liver-related survival was, in NAFLD 0.956 (0.940–0.971), higher than 0.921 (0.910–0.932; P = 0.004) in CHC, slightly lower than 0.969 (0.960–0.979; P = 0.04) in CHB, but much greater than in ALD 0.585 (0.515 0.655; P < 0.001).

The 10 year-overall-survivals according to the four chronic liver diseases included were all significantly different (P < 0.001), in NAFLD 0.689 (0.647–0.730), lower than 0.832 (0.818–0.847) in CHC, and 0.891 (0.891–0.922) in CHB, but much higher than in ALD 0.408 (0.345–0.472) (Figure 2B).

The 10 year-liver-related survivals according to the four chronic liver diseases included were, in NAFLD 0.956 (0.940–0.971), higher than 0.921 (0.910–0.932; P = 0.004) in CHC, slightly lower than 0.969 (0.960–0.979; P = 0.04) in CHB, but much higher than in ALD 0.585 (0.515–0.655; P < 0.001) (Figure 2C).

The overall survival of patients with NAFLD was lower than in patients with CHC since the fifth year of follow-up. This was not explained by a difference in cirrhosis prevalence at baseline (Figure 3A,B). This difference was explained by the older age in NAFLD vs CHC patients, and the associated non-liver-related deaths.

Figure 3.

Overall survivals according to presence or absence of cirrhosis at baseline. Patients with NAFLD had lower survivals whatever the absence or presence of cirrhosis. A, Patients without cirrhosis at baseline. B, Patients with cirrhosis at baseline

There was a dramatic difference in the survivals among NAFLD patients 50 years of age or older (Figure S1A), mainly due to the number of non-liver-related cancers (Figure S1B), and of the cardiovascular-related deaths (Figure S1C), despite less liver-related deaths in the CHC patients (Figure S1D).

Among patients younger than 50 years of age there was no difference between NAFLD vs CHC overall survivals (Figure S2A), including no difference between non-liver-related cancers deaths (Figure S2B), with an early difference in cardiovascular-related deaths (Figure S2C), but compensated by less liver-related deaths (Figure S2D).

FibroTest Prognostic Performance for Survival Without Liver-related Deaths (Primary Endpoint)

Liver-related deaths occurred in 38 cases (19 liver cancers, including 16 hepatocellular carcinoma and three intra-hepatic– cholangiocarcinoma in NAFLD group vs 226 (87 liver cancers, including 84 hepatocellular and 3 cholangiocarcinoma) in the CHC group.

The prognostic values, AUROC (mean; 95% CI) and Cox-risk-ratio (mean; 95% CI) of FibroTest for the prediction of 10 year-liver-related survivals in 1079 patients with NAFLD were highly significant (P < 0.001) 0.941 (0.905–0.978), 1224 (264–5613), vs random. AUROC in NAFLD was higher than in 3449 CHC 0.875 (0.849–0.901; P = 0.01), Cox-risk-ratio 1839 (721–4690), not different than in 2051 CHB 0.848 (0.723–0.974; P = 0.09), and then in 503 ALD, 0.695 (0.575–0.816; P = 0.06) (Figure 4).

Figure 4.

Prognostic performances (AUROCs) of FibroTest. AUROC of FibroTest 0.941 (0.905–0.978) (mean; 95% CI) purple line, for the prediction of 10 year-liver-related survivals in 1079 patients with NAFLD was significantly higher than in 2051 CHC 0.875 (0.849–0.901; P = 0.01) green line, and not different than in 2051 CHB 0.848 (0.723–0.974; P = 0.09) blue line, and then in 503 ALD, 0.695 (0.575–0.816; P = 0.06) red line

Proportionality assumption was respected for the Cox model (Figure S3).

All the five quantitative components of FibroTest were associated (AUROCs) with liver-related deaths both in NAFLD as in CHC (Table S2).

When post hoc analyses in the same patients compared FibroTest to FIB4 in 209 cases with NAFLD (29 liver related-deaths), there was no significant difference in AUROCs, FibroTest (0.935; 0.876–0.993) vs FIB–4 (0.866; 0.739–0.994; P = 0.32). There was also no significant difference in AUROCs, in 401 cases (7 liver related-deaths), FibroTest (0.939; 0.859–1.00) vs transient elastography (0.923; 0.790–1.00; P = 0.72). The applicability of TE-M was only 74.3% (436/587) (Table S3).

FibroTest Prognostic Value for Survival Without Non-liver-related Deaths

Deaths from any causes occurred in NAFLD 240 patients vs 449 in the CHC group. The AUROC for the prediction of overall-survival in patients with NAFLD was low, 0.507 (0.443–0.571) and much lower than its AUROC in patients with CHC, 0.738 (0.708–0.768; P < 0.001), as expected according to the higher prevalence of liver-related deaths in CHC (6.6%, n = 226) vs NAFLD (3.5%, n = 38) (Table 1).

Cardiovascular-related deaths occurred in 52 NAFLD vs 40 CHC patients. AUROC for the prediction of survival without cardiovascular deaths in patients with NAFLD was significant vs random, 0.584 (0.478–0.6921) and not different than in CHC, 0.614 (0.525–0.703; P = 0.73).

Non-liver-cancer-related deaths occurred in 69 NAFLD vs 63 CHC patients. FibroTest had no significant predictive value for survival without non-liver-cancer-related deaths in patients with NAFLD, AUROC = 0.370 (0.275–0.463), and was significant in CHC, 0.613 (0.535–0.690; P = 0.001). In the 69 patients with NAFLD and non-liver-cancer-related deaths, the prevalences of different type of cancers were expected in comparison with the general population prevalences, and without difference with those observed in CHC.

Other-related deaths occurred in 81 NAFLD vs120 CHC patients. FibroTest had no significant predictive value for survival without other-related deaths in patients with NAFLD 0.398 (0.307–0.489) and was significant in CHC, 0.580 (0.524–0.638; P = 0.002).

Number of Events and Cumulative Probability of Death According to Fibrotest Cutoffs

Most of the 10–yr mortality in cirrhotic patients was related to liver-related deaths both in NAFLD and CHC. In non-cirrhotic patients, whatever the stages of fibrosis presumed by FibroTest, there was a lower mortality related to liver deaths in NAFLD as compared to CHC (Table S4).

ApoA1 and Haptoglobin Prognostic Values for Survival Without Liver-related Deaths

Low ApoA1 was associated in patients with NAFLD, with cardiovascular-related deaths (P = 0.009) (Table 2A), and with all causes of deaths (P < 0.001) (Table 2), independently of age, gender, and fibrosis severity estimated by A2M. In CHC ApoA1 was associated with non-liver-related cancer death and all causes of death (Table 3).

High haptoglobin was associated in patients with NAFLD, with cardiovascular-related deaths (P = 0.03) (Table 2B), and with all causes of deaths (P = 0.003) (Table 2F), independently of age, gender and fibrosis severity estimated by A2M. In CHC high haptoglobin was associated with cardiovascular-related deaths, and with non-liver-related cancer death (Table 2D). In CHC, and contrarily to NAFLD low haptoglobin was associated with all causes of death, as expected by the high prevalence of liver-related deaths in CHC and the low level of haptoglobin in cirrhosis (Table 3).

Prognostic Performance of NASH or Steatosis Biomarkers

A total of 680 patients with NAFLD had FibroTest together with new NashTest–2 and SteatoTest–2 (Table S5). For liver-related deaths, in univariate analysis, despite the limited number of events (n = 17) the two biomarkers of NAFLD had significant predictive value, but only the significance of NashTest–2 persisted when adjusted on FibroTest (Table S5B). The other causes of deaths were not associated with elevated NashTest–2 or SteatoTest–2. For all causes of deaths, there was even a negative association with NashTest, significant both in uni and multivariate analyses (Table S5C,D).

Prediction of Deaths Among Patients With Cirrhosis

Staging of NAFLD into seven categories using FibroTest predicted decreasing 5–year survival without related-deaths in F4.3 vs F4.2 and F4.1 as validated previously in CHC and CHB (Figure S5).

Sensitivity Analyses

Exclusion of NAFLD patients with alcohol consumption missing (n = 250), or with rare alcohol consumption but under the standard definition of consumption at risk (daily alcohol consumption ≥30 g for men and ≥20 g for women) (n = 37) (Table S6), or with missing or positive HIV PCR (n = 18) (Table S7) does not change significantly the mains comparisons between survivals or the prognostic performances of FibroTest.

The multivariate analyses showed that the FibroTest significant prognostic value at 10–year persisted after adjustment by gender, BMI (cutoff = 27 kg/m2), and T2–diabetes, in NAFLD compared to CHC.

processing....