Patient characteristics
Table 1 summarizes the clinical characteristics of NAFL and NASH patients, showing body measurements, diagnosis of lifestyle related disease, biochemical data, diagnostic imaging marker, E, CAP measured with Fibroscan, liver/spleen (L/S) ratio measured with CT scan.
Statistical analysis showed that presence of type 2 DM, F-IRI and E value were significant variables. Type 2 DM could not be a marker, and F-IRI would be useful for NASH diagnosis. However, we did not select either type 2 DM or F-IRI in this study, since these markers would not change liver histology.
Selection of the predictors for NASH
We focused that inflammation and fibrosis markers should be selected as the predictors for NASH, since histological imaging was certainly affected by them according to natural history of NASH.
ALT is a marker of hepatitis level. ALT was not good enough for differential diagnosis of NASH from NAFL (p value 0.0186, Table1). However, we have experienced many NASH patients who showed persistent abnormality of ALT (more than 45 IU/L) for 3 months and more, while NAFL patients showed only a temporary higher value (Fig2). To evaluate the accuracy of persistent ALT (≧45 IU/L) in discriminating between NAFL and NASH, the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPT) were calculated as 79.6, 84.2, 92.9 and 61.5%, respectively. The ratio of persistent ALT abnormality was 3/19 (15.8%) for NAFL, and 39/49 (79.6%) for NASH. Statistical analysis showed it significant: NAFL vs all NASH, chi-square test <0.0001, shown in Table 2.
Type IV collagen 7S is known as a marker of liver fibrosis. But the results of this study about type IV collagen 7S did not show the significant t test (p value 0.0013) between NAFL vs NASH (Table 1). It is shown that type IV collagen 7S values of NASH S2-3 (5.9±1.1 ng/ml) and NASH s4(12.2±2.5 ng/ml) were higher than those of NAFL (3.7±0.4ng/ml) and NASHS0-1(4.4±1.4ng/ml) in Fig 3-a.
Receiver-operating characteristic (ROC) curves of the type IV collagen 7S were plotted in Fig 3-b. The first ROC curve (1) depicts the ability to discriminate between NAFL patients and NASH patients. At a cutoff value 3.9 ng/ml, the sensitivity, specificity, PPV and NPV were 72.3, 84.2, 91.9 and 55.2%, respectively. Area under the curve (AUROC) was 0.823. Strictly speaking, type Ⅳ collagen 7S could not be good enough to differentiate all NASH from NAFL. According to the second ROC curve (2) for differentiation between NAFL / NASHS0-1patients and NASHS2-3 / NASH s4 patients, the sensitivity, specificity, PPV and NPV were 92.9, 86.5, 65.0 and 97.8% at a cutoff value 5.1ng/ml and the AUROC was 0.896 (Fig 3-b). T test revealed a significant difference (p value <0.0001) between NAFL / NASHS0-1vs NASHS2-3 / NASH s4 (Table 2).
E value (kPa) was selected as the third marker. E value expresses the level of liver stiffness measurement with Fibroscan. Fig 4-a shows that the distribution of E values changed higher from NAFL, NASH S0-1, NASH S2-3 to NASHs4 together with the fibrosis and/or chronic inflammation. The median liver stiffness values were NAFL 4.9±1.1, NASH S0-1 8.8±3.0, NASH S2-3 13.3±6.2 and NASHs4 48.5±14.2 kPa. The AUROC of the E value were plotted in Fig.4-b. This curve depicts the ability to distinguish between NAFL patients and all NASH patients. The sensitivity, specificity, PPV and NPV were 93.6, 83.3, 91.7 and 82.4% respectively at a cutoff value 5.5 kPa. The AUROC was 0.926 (Fig.4-b). Furthermore, t test revealed a significant difference (p value <0.0001) between NAFL and NASH (Table 2).
From the results shown above, we selected three markers: (1) persistent abnormality of ALT as a hepatitis one, (2) type Ⅳ collagen 7S as a liver fibrosis one and (3) E value as a liver stiffness measurement including fibrosis and chronic inflammatory activity. These markers complementarily can detect the characters of NASH. We called the ACE system using three markers as a pattern recognition method.
Pathological conditions of NASH change from fatty liver to chronic hepatitis, liver cirrhosis and hepatoma along a natural history. The positive patterns using three markers must change along the different state, since the medical conditions of NAFL and every stage of NASH reflect the difference of hepatitis, fibrosis and stiffness.
Positive pattern combinations using three markers (ACE system)
The total number of combination would be eight: all positive 1, two positive one negative 3, one positive two negative 3, and all negative 1, shown in Fig. 5-a (①~⑧). Fourteen patients of NAFL showed all negative ⑧, 19 patients of NASH S0-1 showed positive for ALT and E value ③ and 11 patients of NASH S2-3 showed all positive ①. Almost none of the patients showed type Ⅳ collagen 7S positive and E value negative ② and ⑥. Since the sensitivity of E value was higher than that of type IV collagen 7S, both ② and ⑥ groups could be deleted from Fig. 5-b.
Fig.5-b summarizes the relationship between the positive patterns using three markers and pathological condition (NAFL, NASHs0-1, NASHs2-3 and NASH s4). Considering the natural history of NAFL and NASH, major conditions are following: NAFL pattern ⑧ all negative, NASH hepatitis pattern ③ (ALT(+),Ⅳ collagen(-), E(+)), NASH fibrosis pattern ① (ALT(+), Ⅳ collagen(+), E (+) all positive) and NASH liver cirrhosis pattern ⑤ (ALT(-), Ⅳ collagen(+), E (+)). Pattern ④ (ALT(+), Ⅳ collagen(-), E(-)) suggested early hepatitis pattern, which showed the stage when NAFL and NASH could not be distinguished. Pattern ⑦ (ALT(-), Ⅳ collagen(-), E(+)) was shown by some of NASHS0-1 patients. It is supposed that NASH patients had hepatitis before, but had not during liver function tests. These patients who were pathologically diagnosed as NASHS0-1, showed that only E value was positive. This condition suggests that E value measurement is a sensitive for liver stiffness remained in NASH patients even if the inflammation happened in the past. Described above, pattern ⑦ (NASH post-hepatitis pattern) would be situated between ④ and ③.
Frequency distributions of NAFL and NASH cases are shown in Fig.6. NAFL, NASH s0-1, NASH s2-3, NASH s4 cases had peaks at combination⑧, ③, ① and ⑤, respectively. It should be noted that these NAFLD cases lined up in order and properly distributed in the combination patterns.
Comparing the ACE score to previously reported scoring systems
FIB4 index12,13, NAFLD fibrosis score (NFS)14,15 and BARD score16,17were previously established and were utilized for predicting advanced fibrosis (stage3 and 4) versus others. On the other hand, NAFIC score18 used to be utilized to differentiate between NAFL and NASH patients. Formulas of these scoring systems were described in Table 3.
In order to compare the difference between our system and others for differentiating NAFL and NASH, each positive pattern combination was weighted 0 ~ 3points as ACE score.
ACE score was determined; combination⑧(NAFL main)was given 0 point,④ and ⑦ 0.5point, ③ (early NASH) 1point, ① (middle NASH) 2points, and ⑤ (late NASH) 3points respectively.
The AUROCs of these scoring systems were summarized in Table3 and Fig 7. The AUROC was greatest for ACE score (0.953), followed by NAFIC score (0.898), FIB4 index (0.633), NFS (0.592) and BARD score (0.542) (Table 3). According to these results, NAFIC score was useful for differentiating NASH from NAFL, but ACE score was superior to other scores to distinguish NAFLD patients.