Non-invasive predictors of prognosis of Asian patients with histopathologically-confirmed lean nonalcoholic fatty liver disease

Background The prognostic factors of morbidity and mortality in patients with lean NAFLD (body mass index < 25.0 kg/m2) are unknown. Methods In this retrospective study, 446 Japanese patients with histopathologically-confirmed NAFLD (lean NAFLD, n = 170) were followed for liver events, cardiovascular events, type 2 diabetes mellitus, and non-liver malignancies. The median observation period was 4.6 years. We also investigated the predictors of severe fibrosis (stage 3–4) and mortality in lean NAFLD patients. Results Glycolipid metabolic markers, liver function tests, NAFLD fibrosis score (NFS), and histological scoring were significantly lower in lean NAFLD patients than in non-lean NAFLD. The incidence of liver cancer was higher while that of T2DM was lower in lean NAFLD. Kaplan–Meier analysis showed no significant difference in overall survival between the lean and non-lean NAFLD. Multivariate analysis of data of lean NAFLD identified NFS ≥ − 1.455 as significant independent predictor of severe fibrosis, while history of liver cancer and NFS ≥ − 1.455 were predictors of overall survival. Conclusions Although patients with lean NAFLD have better histopathological and biochemical profile compared to patients with non-lean NAFLD, the prognosis is not different between the two groups. Lean NAFLD patients with NFS ≥ − 1.455 or history of liver cancer should be monitored carefully during follow-up.

as body mass index (BMI) < 25.0 kg/m 2 [9]. In Japan, we often encounter lean NAFLD patients in daily practice.
The prognosis of lean NAFLD is considered to be better than that of obese NAFLD [10]. However, a higher overall mortality was reported in patients with lean NAFLD compared with obese NAFLD patients in an 11-year follow-up study, despite presentation with a healthier metabolic profile, including low insulin resistance [11]. Moreover, lean NAFLD has been reported to be independently associated with increased risk of all-cause and cardiovascular mortality compared with lean individuals without NAFLD [12]. To date, the prognosis of lean NAFLD is still unknown in Japan.
The present study was designed to define the clinical features of histopathologically-confirmed lean NAFLD patients. Furthermore, by using non-invasive tests, we analyzed the data to obtain clinically meaningful prognostic factors and predictors of severe fibrosis and survival.

Patients
Patients admitted to our hospital with liver dysfunction and/or fatty liver, diagnosed by clinical examination, laboratory tests and abdominal ultrasonography, between 1976 and 2019, underwent liver biopsy. Histopathological examination confirmed the diagnosis of NAFLD in 446 of these patients. These patients included those in whom histopathological examination showed microscopic changes consistent with steatosis in at least 5% of hepatocytes and patients with history of alcohol intake of < 20 g/ day. The median duration of follow-up from diagnosis to death or last visit was 4.6 years (range 0.0-43.5 years). The clinical features of the patients at the time of histopathologic diagnosis of NAFLD are summarized in Table 1. We excluded patients with (1) underlying liver disease (e.g., viral hepatitis, autoimmune hepatitis, druginduced liver disease, or primary biliary cholangitis); (2) systemic autoimmune diseases (e.g., systemic lupus erythematosus or rheumatoid arthritis); and (3) metabolic diseases (e.g., hemochromatosis, α-1-antitrypsin deficiency, or Wilson's disease).
The study was conducted in compliance with the International Conference on Harmonization Guideline for Good Clinical Practice (E6) and the 2013 Declaration of Helsinki. The study protocol was approved by the Toranomon Hospital Institutional Review Board (#953). Written informed consent for liver biopsy was provided by all patients.

Diagnosis and follow-up
In this study, we selected the following liver-related events for study outcome: liver cancer, hepatic encephalopathy, esophagogastric varices with bleeding, ascites, and jaundice. Other outcomes included cardiovascular events (e.g., coronary artery disease, heart valve disease, arrhythmia, heart failure, hypertension, orthostatic hypotension, shock, endocarditis, diseases of the aorta and its branches, disorders of the peripheral vascular system, and stroke), type 2 diabetes mellitus (defined as high fasting blood glucose level ≥ 126 mg/dL, high hemoglobin A1c ≥ 6.5%, use of glucose-lowering agents, or self-reported history of clinical diagnosis), and nonliver malignancy. Mortality was evaluated for all patients. Hematologic and biochemical data were collected at least twice yearly after the diagnosis of NAFLD. Ultrasonography, computed tomography, and/or magnetic resonance imaging studies were conducted at least once annually during the follow-up.

Clinical parameters
We included in the analysis various clinicopathologic and genetic parameters that could affect NAFLD prognosis. The NAFLD fibrosis score (NFS), calculated as (− 1.675 , has been used as a parameter for progression of fibrosis [16]. By applying the low cutoff point (score less than − 1.455), 77% of the patients without significant fibrosis were correctly identified, whereas 22% of patients with a low cutoff point were incorrectly staged [16]. We used the Europeans definition of obesity (BMI ≥ 30 kg/m 2 ) [17], while lean NAFLD was defined as BMI < 25.0 kg/m 2 [9]. Patatin-like phospholipase domain containing protein 3 (PNPLA3) rs738409, transmembrane 6 superfamily member 2 (TM6SF2) rs58542926, and hydroxysteroid 17-beta dehydrogenase 13 (HSD17B13) rs6834314 were genotyped by the TaqMan single-nucleotide polymorphism (SNP) genotyping assay (Applied Biosystems, Foster City, CA).

Statistical analysis
The baseline characteristics were compared using the Mann-Whitney U test for continuous variables or Fisher's exact test for categorical variables. The incidence of each event was analyzed during the period from the time of histopathological diagnosis of NAFLD until the last visit or occurrence of event. Overall survival was estimated using the Kaplan-Meier method, and differences between curves were evaluated using the log-rank test. All parameters that showed strong correlation with others were considered confounding factors and excluded from the statistical analysis. The remaining parameters were entered into multivariate analysis using the logistic regression analysis and the Cox proportional hazards model. Statistical significance was set at P < 0.05. All statistical analyses were carried out using the EZR software [18].

Comparison of background factors stratified by BMI
The baseline characteristics and laboratory data stratified by BMI of the 446 patients are shown in Table 2. The prevalence of type 2 diabetes mellitus, hypertension and hyperuricemia were significantly lower in the lean NAFLD group (BMI < 25). Aspartate aminotransferase, alanine aminotransferase, triglyceride, fasting blood sugar, glycated hemoglobin, ferritin, and NAFLD fibrosis score (NFS) were significantly lower in the lean NAFLD group. Various histopathological findings (e.g., steatosis 3, lobular inflammation 2-3, NAFLD activity score 5-8, and fibrosis stage 3-4) were significantly lower in the lean NAFLD group. Table 3 lists the incidence of liver events, cardiovascular events, type 2 diabetes mellitus, and non-liver malignancies in patients with NAFLD. We analyzed the person-years method for patients with new onset during follow-up who did not have each disease before or at liver biopsy. In the lean NAFLD group, 6/155 (3.9%) patients developed liver cancer (rate per 1000 person years, 4.49). Furthermore, 4/148 (2.7%) patients confirmed to have no previous or current liver-related events at NAFLD diagnosis developed liver-related events (rate per 1000 person years, 3.08). Furthermore, 15/168 (8.9%) patients developed cardiovascular events, with a development rate per 1000 person years of 11.07. Further analysis showed 12/128 (9.4%) patients developed type 2 diabetes mellitus (rate per 1000 person years, 10.95) and 12/152 (7.9%) patients developed non-liver malignancies (rate per 1000 person years, 9.60). The liver cancer development rate per 1000 person years tended to be higher in the lean NAFLD group (4.49) than the non-lean group (1.76). On the other hand, the proportion of patients with T2DM was lower in the lean group than the non-lean group (10.95 vs 19.88). Table 4 lists the number and rate of mortality associated with liver-related events, non-liver cancer malignancies, and other events in patients with NAFLD. A person-year analysis was performed on patients who did not have the disease before or at the time of liver biopsy and died of the disease during follow-up. For the lean NAFLD group, . Deaths from liver-related diseases tended to be more common in the lean NAFLD group than the nonlean NAFLD group, but the difference was not statistically significant. Kaplan-Meier analysis also showed no significant difference in overall survival between the two groups (P = 0.39) (Fig. 1). Table 5 summarizes the baseline characteristics and laboratory data of the 170 patients with lean NAFLD (n = 132, with fibrosis stage 0-2, n = 38 for fibrosis stage 3-4). Age, history of liver cancer and prevalence of type 2 diabetes mellitus were significantly higher in patients with fibrosis stage 3-4. Furthermore, hemoglobin, platelet count, total cholesterol, and lowdensity lipoprotein cholesterol were significantly lower, while aspartate aminotransferase, fasting blood sugar, glycated hemoglobin, and NFS were significantly higher in patients with fibrosis stage 3-4. PNPLA3 GG was significantly more frequent in the group of fibrosis stage 3-4. For histopathological findings, ballooning 2 and lobular inflammation 2-3 were significantly more frequent in the fibrosis score 3-4 group. All parameters that correlated strongly with others were considered confounding factors and excluded from statistical analysis. Thus, 14 non-invasive potential predictive factors of severe fibrosis (stage 3-4) were analyzed (Table 6). Univariate analysis identified four significant parameters; history of liver cancer, hemoglobin, NFS, and PNPLA3. These parameters were entered into multivariate analysis using the logistic regression analysis. The results identified NFS as a significant and independent factor determinant of Table 3 Incidence of liver events, cardiovascular events, type 2 Table 6).

Non-invasive predictors of survival in patients with lean NAFLD
Parameters that correlated strongly with others were considered confounding factors and excluded from statistical analysis. Thus, 14 non-invasive potential predictive factors of prognosis were analyzed (Table 7). Univariate analysis showed that history of previous liver cancer and NFS correlated significantly with survival. These two factors were entered into multivariate analysis using the Cox proportional hazards model. The analysis identified both parameters as significant and independent prognostic factors for lean NAFLD (yes for history of liver cancer, P < 0.01, NFS: ≥ − 1.455, P = 0.026) ( Table 7). Furthermore, Kaplan-Meier analysis also showed that overall survival was significantly shorter in patients with previous liver cancer (P < 0.01) and in those with high NFS (≥ − 1.455, P < 0.014) (Fig. 2).

Discussion
The clinical characteristics and prognostic factors of lean NAFLD remains unclear. In this retrospective study, we investigated the clinical features of histopathologicallyconfirmed cases of lean NAFLD and analyzed clinically useful parameters obtained non-invasively for the prediction of severe fibrosis and prognosis.
Our study included 170 patients with lean NAFLD and 276 with non-lean NAFLD. Lean NAFLD was associated with a lower complication rate of metabolic syndrome and better general biochemical data and pathological findings than the non-lean NAFLD. However, the rates of liver-related events, cardiovascular events, and malignancies were not different between the two groups, though the incidence of liver cancer was particularly higher in lean NAFLD. The overall survival rate was not significantly different between the two groups. Previous studies suggested that the metabolic profile of lean NAFLD is similar or slightly better than that of the nonlean NAFLD [19,20], which was also noted in this study. The finding that patients with lean NAFLD are at higher risk of severe liver disease despite the lower prevalence of advanced fibrosis and NASH at baseline is paradoxical. Logically, this suggests that progression of liver fibrosis is faster in lean NAFLD than in NAFLD obese patients [9]. Further studies are needed to confirm this speculation.
Several studies have described age, diabetes, cirrhosis, low platelet count and low albumin levels as significant prognostic factors for NAFLD/NASH [21][22][23][24][25][26][27][28]. Although there are only a few reports on the prognostic factors of lean NAFLD among NAFLDs, one previous study identified fibrosis stage, hypertension, and age as independent prognostic factors [9]. As mentioned above, lean NAFLD patients have shorter history and less abnormal laboratory findings than those with non-lean NAFLD. Analysis of non-invasive predictors of fibrosis and prognosis of lean NAFLD is needed because early intervention is needed to improve prognosis. Such non-invasive markers should serve to: i) in primary care settings, identify the risk of NAFLD among individuals with increased metabolic risk; ii) in secondary and tertiary care settings, identify those with worse prognosis, e.g. severe NASH; iii) monitor disease progression; and iv) predict response to therapeutic interventions. Achieving these objectives could reduce the need for liver biopsy [8]. NFS or FIB-4 index are clinically useful tools for identifying NAFLD patients with higher likelihood of having bridging fibrosis (stage 3) or cirrhosis (stage 4) [29,30]. The NAFLD fibrosis score (NFS) has the advantage that no special test items are included. We identified NFS of ≥ − 1.455 as a non-invasively measured parameter for the prediction of severe fibrosis in lean NAFLD. More importantly, for the first time, we found that NFS of ≥ − 1.455 is also a non-invasive independent and significant predictor of prognosis.
The present study has certain limitations. First, the median observation period was 4.6 years, which is a relatively short in prognostic studies. Second, the subject of this study were patient admitted to the Department of Hepatology of our hospital for liver biopsy. All subjects were Asians and admitted for the purpose of scrutiny of liver disease. Thus, a selection bias cannot be excluded. Further studies are needed that include patients of

Conclusions
Although patients with lean NAFLD had better histologic and biochemical profile compared to patients with non-lean NAFLD, it may be risky to end those follow-ups based on the lack of differences in prognosis between the two groups. In lean NAFLD, patients with NAFLD fibrosis score of ≥ − 1.455 or history of liver cancer should be followed-up carefully.