The increase in metabolic syndrome due to obesity has become a global problem, and Japan is no exception [1
]. Nonalcoholic fatty liver disease (NAFLD) covers a spectrum of liver diseases that range from benign simple steatosis to the hepatic inflammation and fibrosis of nonalcoholic steatohepatitis, cirrhosis, and hepatocellular carcinoma. NAFLD is a hepatic manifestation of metabolic syndrome [4
]. Therefore, the prognosis of NAFLD is affected not only by liver-related diseases such as cirrhosis and hepatocarcinogenesis, but also by all diseases and conditions that have a common background with NAFLD. These diseases and conditions include obesity, insulin resistance, diabetes, and dyslipidemia [6
According to previous reports, the survival rate of NAFLD was significantly lower than that of the general population. The most common causes of death in NAFLD in the United States are cardiovascular diseases (CVDs) (25%), extrahepatic malignancies (28%), and liver diseases (13%). Factors related to mortality are age, impaired glucose tolerance, and cirrhosis [7
]. Because NAFLD is a liver disease against the background of lifestyle-related diseases, the prognosis for NAFLD is associated not only with the liver effects, but also with the effects of the progression of other lifestyle-related diseases. These diseases include CVD and chronic kidney disease (CKD) that result from visceral obesity, arteriosclerosis, and diabetes. Multiple large epidemiologic studies have shown that NAFLD is an independent risk factor for the development of CVD [10
]. In addition, CVD mortality in NAFLD was higher than in the general population [13
]. NAFLD has also been reported to be a risk factor for CKD independent of metabolic syndrome [14
]. Therefore, diagnosis of NAFLD is important. The prevalence of NAFLD in Japan is reported to be 29.7% and it is estimated that 37.4 million people have NAFLD [17
Imaging examinations including abdominal ultrasonography are generally used for diagnosing NAFLD; however, it is difficult from the viewpoint of medical economy to test the whole population. In addition, with the influence of COVID-19 [18
], which has recently spread throughout the entire world, abdominal ultrasound can be risky due to the concentrated contact. It has also been reported that NAFLD is a risk factor for severe COVID-19 infection [20
]. Therefore, it is currently required to predict the diagnosis of NAFLD by a noncontact procedure rather than ultrasound examination. By previous reports, there are three indexes for predicting NAFLD: the fatty liver index (FLI), Zhejiang University (ZJU) index, and hepatic steatosis index (HSI) [21
]. However, the results of a direct comparison of the diagnostic performance of these indexes remains unclear. Moreover, it is important to confirm whether the diagnostic performance of these indexes is reliable in patients with mortality risks such as liver fibrosis and diabetes [24
]. The aim of this study was to examine the diagnostic performance of prediction formulas to identify NAFLD in the general population who underwent health checkups in Japan.
Noninvasive prediction formulas (HSI, ZJU index, and FLI) were tested in the current study in the identification of NAFLD. Using these indexes, NAFLD could be diagnosed accurately without an imaging examination in Japanese subjects who received health checkups. Ultrasound is the gold standard and most common imaging examination to diagnose fatty liver; however, with the high prevalence of NAFLD [8
], it is impossible to recommend ultrasound for all in a general population. These indexes enable the identification of people with NAFLD in a large population who should receive an imaging examination.
The available guidelines, however, have never confirmed the actual screening procedure to identify NAFLD in the high-risk population, including diabetes patients, much less in the general population. This is because of uncertainties in diagnostic tests and treatment options, along with a lack of evidence related to the long-term benefits and cost-effectiveness of screening [35
]. The guideline by the European Association for the Study of the Liver, European Association for the Study of Diabetes, and European Association for the Study of Obesity, regarding the utility of nonimaging biomarkers, including the FLI, is referred to for the screening of a large population [36
]. Byrne and Targher recommended the use of the FLI as well as ultrasound for the screening of NAFLD in patients with type 2 diabetes [37
]. Taken together, in the global “pandemic” of NAFLD, easy and low-cost screening procedures such as prediction formulas are warranted and should be promoted. In addition, general and common parameters are preferred as components of prediction formulas. According to the availability of the parameters, the appropriate prediction formula and nonimaging indexes should be used. Indexes analyzed in the current study comprised only general and common parameters: BMI, FPG, TG, and ALT for the ZJU index; BMI, AST, ALT, and diabetes for the HSI; and TG, BMI, GGT, and WC for the FLI. In the primary care setting and health checkup sites, where these parameters could be measured, any of the indexes should be tested to identify potential NAFLD patients.
There were several differences among the HSI, ZJU index, and FLI in the current study. In comparing the ROC of individual indexes, the ZJU index and FLI showed a significantly greater AUROC than the HSI. A possible explanation for the difference is the diagnosis of diabetes required for the HSI. According to the original HSI study by Lee et al., the diagnosis of diabetes was based on the FPG, HbA1c, and antidiabetic medications [22
], whereas medication information was missing in our study. However, the sensitivity and specificity obtained in our current study (at a cut-off value of 30.0, sensitivity and specificity were 93.4% and 56.8%, respectively; at a cut-off value of 36.0, sensitivity and specificity were 49.4% and 94.4%, respectively) were comparable with the original study (at a cut-off value of 30.0, the sensitivity and specificity were 92.5% and 40.0%, respectively; at a cut-off value of 36.0, sensitivity and specificity were 46.0% and 92.4%, respectively). These results suggest that the diagnostic performance of the HSI was validated in the Japanese general population of the current study—as were the ZJU index and FLI. The distribution of the FLI was quite different from the other indexes (Figure 1
); the frequent range of the index (peak of the distribution curve) shifted to the negative direction and the number of the subjects with an intermediate probability of NAFLD and a high probability of NAFLD were fewer than in the HSI and ZJU index. This unique distribution of the FLI might have resulted in a higher specificity than the HSI and ZJU index in our current study: 98.4% in the overall subjects (Table 2
). However, the sensitivity of the FLI was lower than that of other indexes. According to the previous studies, the cut-off values of the FLI were optimized in Asia. Yang et al. reported from Taiwan that the optimal cut-off value to rule in an NALFD diagnosis by ultrasound was 35 for males and 20 for females, and the cut-off value to rule out the diagnosis was 25 for males and 10 for females [38
]. According to another report from Taiwan, for a sensitivity ≥ 90%, the cut-off value was 15 for males and 5 for females, and for a specificity ≥ 90%, the cut-off value was 50 for males and 25 for females [39
]. These cut-off values were lower than in the original study reported by Bedogni et al. from Italy [23
], suggesting that the cut-off value, especially the low cut-off value for higher sensitivity, should be optimized in the Asian cohort. Without optimization of the original cut-off values, our results suggest that the HSI and ZJU index represented ≥90% sensitivity in the Japanese cohort and would be better for screening in the general population in Asia.
Among the noninvasive tests to diagnose the liver fibrosis of NAFLD, the FIB-4 index is a common and easy-access procedure that is calculated using AST, ALT, platelet count, and age [29
]. Moreover, liver fibrosis is the most important finding to predict prognosis and to identify the treatment indication [24
]. Therefore, screening with the approach “FIB-4 index first,” ahead of the diagnosis of fatty liver could be a novel and upcoming strategy to identify greater-risk NALFD in the primary care setting and at health checkups [38
]. Hence, we tested the diagnostic performance of the HSI, ZJU index, and FLI in the subjects stratified by the FIB-4 index (Figure 3
and Table 3
). The diagnostic performances of the HSI, ZJU index, and FLI in the individual FIB-4 index categories were similar to those of the overall subjects. Moreover, the diagnostic performance of the FLI increased in the subjects with an intermediate or high risk of advanced fibrosis. Taken together, the combination of the FIB-4 index and any of the ZJU index, HSI, and FLI would be useful to simultaneously predict the NAFLD and fibrosis risk.
In the current study, the ZJU index and FLI, which showed a greater AUROC than the HSI in the subjects, overall were compared under several specific conditions (Figure 4
). Either in males or females, the ZJU index showed a greater AUROC than the FLI. The ZJU index reflects a gender difference in the formula, but the FLI does not [21
], whereby the ZJU index might show a greater AUROC than the FLI when the subjects are divided by gender. However, whereas the ZJU index reflects FPG, the AUROC in subjects with diabetes was smaller than for the FLI, which does not include FPG, HbA1c, and the diagnosis of diabetes. This feature suggests that the interaction between the specific condition and individual components of the formula could both positively and negatively affect the diagnostic performance of the formula. According to our results, at least the ZJU index could be recommended for male only/female only subjects, and the FLI could be suitable for subjects with diabetes.
Adequate diagnostic performance of the ZJU index and FLI raised the question about the characteristics of the subjects with NAFLD having both a negative ZJU index and FLI. These false negative subjects should be carefully managed to avoid missing imaging examinations. In the current study, 107 subjects had NAFLD with both a negative ZJU index and FLI (Table 4
). The median of the BMI, WC, liver enzymes, and metabolic parameters were all within the reference range but significantly higher than in the subjects without NAFLD. By the multivariate analysis, male sex, BMI > 22 kg/m2
, and abnormal WC were independent factors associated with NAFLD (Table 5
). Interestingly, these variables are not the results of laboratory tests, suggesting that physical findings are important for male subjects with blood test values within the reference range, and for those who have abdominal obesity and/or a BMI higher than 22 kg/m2
, imaging examination could be recommended.
There are several limitations in the current study. Since the subjects were health checkup examinees and they were 20–65 years of age, the diagnostic performance of the HSI, ZJU index, and FLI should be confirmed in adolescent or younger subjects, as well as in older subjects. Because NAFLD was diagnosed by ultrasound and the subjects never underwent liver biopsy or an imaging examination such as MRI, evaluation for fatty liver was not quantitative and inter-/intra-observer error could be present in the ultrasound diagnosis. According to the recent development of an ultrasound-based technique, the attenuation of ultrasound in the liver parenchyma can be measured and the severity of fatty liver is quantitatively represented; FibroScan (Echosens, Paris, France) equips a controlled attenuation parameter (CAP) [42
], and attenuation coefficient (Hitachi, Tokyo, Japan) [44
] and attenuation imaging (Canon Medical Systems Corporation, Otawara, Tochigi, Japan) [45
] are installed on the B mode ultrasound. Using these ultrasound-based techniques for the diagnosis of fatty liver as the standard, the diagnostic performance of nonimaging indexes, including HSI, ZJU index and FLI, should be tested in further study. On the other hand, these ultrasound-based techniques are relatively new and need to be validated further. There is evidence that skin capsular distance, BMI, and several other factors affect the reliability of CAP [43
]. The accessibility of these ultrasound-based techniques is limited in the general population and primary care settings. Therefore, nonimaging indexes should be developed and individual features of nonimaging indexes should be well known.
In conclusion, the HSI, ZJU index, and FLI are useful to diagnose NAFLD in Japanese health checkup examinees. According to the availability of the parameters and characteristics of the cohort, the appropriate index should be used for screening for NAFLD.