The Fatty liver Index (FLI) 15 years later: a reappraisal

The Fatty Liver Index (FLI) is a non-invasive biomarker proposed, in 2006, by Bedogni’s group, to aid in identifying patients with suspected nonalcoholic fatty liver disease (NAFLD) to be submitted to liver ultrasonography to confirm steatosis. Criteria of Assessment of Narrative Review Articles, a scale for the assessment of quality of narrative review articles, inspired our review article, which aims at evaluating the scope of published articles on FLI issued over the last 15-year period. The analysis of retrieved data identified the following conclusions. First, given that FLI and NAFLD share the same risk factors, FLI can be used to identify NAFLD among populations at risk to be submitted to screening. Second, FLI is able to identify the hazard of atherosclerosis, both at a subclinical stage and as an overt disease. Third, FLI detects incident diabetes and chronic kidney disease. However, evidence supporting the notion that FLI also predicts the metabolic syndrome, some endocrine disorders, certain tumor types, and overall and cause-specific mortality appears to be more limited. In conclusion, 15 years after its first publication, FLI has been validated as a robust biomarker of both steatosis and NAFLD. Moreover, the scope of FLI has been expanded to previously unexpected areas. Finally, we discuss FLI limitations and a research agenda aimed at further improving the accuracy of FLI scores in predicting liver-related outcomes, endocrine-metabolic disorders, cancer risk, and survival. Page 2 of Lonardo et al. Metab Target Organ Damage 2021;1:10 https://dx.doi.org/10.20517/mtod.2021.08 20


Nonalcoholic fatty liver disease: definition and natural course of hepatic and extra-hepatic involvement
Nonalcoholic fatty liver disease (NAFLD) describes hepatic fatty changes which are bi-directionally associated with the metabolic syndrome (MetS) and its individual components [1] . By definition, NAFLD requires the exclusion of competing causes of liver disease.
As a systemic disorder, NAFLD has a "hepatic" as well as an "extra-hepatic" natural history. The former includes manifestations such as simple steatosis, nonalcoholic steatohepatitis (NASH), fibrosis, cirrhosis, and hepatocellular carcinoma [2,3] . The latter comprises involvement of the cardiovascular and endocrine systems, chronic respiratory disorders, the musculoskeletal system, the skin, and extra-hepatic tumors [4] .
Recently, NAFLD has also been clearly identified as a strong risk factor for incident chronic kidney disease (CKD) [5] .
According to the European Association for the Study of the Liver, further to ultrasonography, assessment of steatosis can also be accomplished with biomarkers such as the Fatty Liver Index (FLI), SteatoTest, and NAFLD Fat score [6] .

A brief history of FLI development and original aims
In 2006, Bedogni et al. [7] published an article entitled: "The Fatty Liver Index: a simple and accurate predictor of hepatic steatosis in the general population". This study validated an algorithm, based on body mass index (BMI), waist circumference, triglycerides, and gamma-glutamyltransferase (GGT), originally developed based on the analysis of 216 individuals with suspected liver disease compared to 280 controls without liver disease belonging to the general population of the Dionysos Nutrition & Liver Study (DNLS) [8] .
FLI, with a range between 0 and 100, can be calculated using the following two steps.
In the DNLS, values of FLI < 30 ruled out (negative likelihood ratio = 0.2) and values ≥ 60 ruled in (positive likelihood ratio = 4.3) fatty liver with a discrimination of 0.84 (95%CI: 0.81-0.87) as detected by the area under the receiver operating characteristic curve [7] . It is important to cite the precise words used by Bedogni et al. [7] to explain what role they attributed to FLI: "FLI may help physicians to select subjects for liver ultrasonography and intensified lifestyle counseling, and researchers to select patients for epidemiologic studies". As reported in this article, however, this prediction has largely been overcome by an extraordinary number of publications using FLI for purposes quite different from those originally suggested by the authors.

How the components of the FLI algorithm fit in the natural course of hepatic and extra-hepatic NAFLD
Bedogni et al. [8] used data from the DNLS to develop the FLI algorithm for the prediction of fatty liver. Age, sex, and alcohol were not associated with steatosis in any of the multivariable models leading to the development of FLI. The fact that BMI, waist circumference, GGT, and triglycerides were the independent predictors chosen for the final prediction model is in full agreement with our current understanding of the role of overall and regional adiposity [9] , lipid phenotype [10] , and GGT as an accurate surrogate index of insulin resistance in NAFLD [11] . Unfortunately, the role of GGT as a predictor of fatty liver seems to have generally been neglected by American and British physicians, possibly because GGT is not considered a primary liver test in these countries.

Scale for the quality Assessment of Narrative Review Articles
The present review article adheres to the Scale for the quality Assessment of Narrative Review Articles (SANRA) [12] . Without covering originality, topicality, conflicts of interest, or plausibility, and although not designed to provide a precise estimate of the quality of all theoretically possible manuscripts, this scale is based on formal criteria intended to help editors, reviewers, and readers in assessing the quality of a given narrative review article [12] . To this end, SANRA utilizes a simple sum scoring system based on quite limited scoring options (0, 1, and 2) [12] . In short, SANRA supports six qualifying points including: (1) justification of the article's importance for each journal's readership; (2) statement of concrete (single or multiple) aims or formulation of questions; (3) transparency about the sources of information on which the text is based and accurate description of search history; (4) extensive backing key statements with adequate referencing to key statements; (5) introducing appropriate arguments underlying scientific reasoning; and (6) appropriate presentation of data [12] .

Bibliographic research strategy and aims
The PubMed database was searched using those articles, without any language restrictions, exhibiting "Fatty Liver Index" in their titles. The search was completed on 1 June 2021. Overall, 93 articles were retrieved, 43 of which were retained based on the agreement of the authors and 50 were deemed as out of the aims of the present study. The reasons for exclusion were studies either based on limited case series or not relevant to illustrate the main topic of the present study, which aims to report on the scope of FLI use in contemporary medical literature.

RISK FACTORS FOR FLI
Five studies published thus far, as summarized in Table 1, have addressed the risk factors predicting FLI, considered to be a surrogate marker of NAFLD [13][14][15][16][17] .
Collectively, the data suggest that, in European and Japanese populations, age, sex, and lifestyle habits, including dietary and sedentary behavior, modulate total and visceral obesity and affect insulin resistance thereby contributing to determining the risk of NAFLD as assessed by FLI [ Table 1]. Integrating the conclusions of the original FLI paper and in agreement with general and recent notions pertaining to the epidemiology of NAFLD [18,19] , age and sex were identified as important modifiers of FLI variability in these populations [14] . The finding that consumption of sweetened beverages predisposes to while eating fruit protects from FLI [13] is also consistent with studies conducted on NAFLD [20] . Finally, the study by Klisic et al. [15] also confirmed that the range of "normal" transaminases must be updated, as originally suggested by Prati et al. [21] , and that risk factors for the development of NAFLD vary in adult men and Weber et al. [13] 2018 161 individuals with T2D and 62 T2D-free controls were extracted from the GDS Peripheral (M-value) and hepatic IR were assessed by hyperinsulinemic-euglycemic clamps with stable isotope dilution A doubling of SSB-derived sucrose plus non-sucrose bound as well as of non-sucrose bound fructose intake associated with a reduction of the M-value by -2.6% (-4.9; -0.2) and -2.7% (-5.2; -0.1) among T2D, respectively, with an increase in the odds of fatty liver by 16% and 17%, respectively, among T2D (all P < 0.05) Doubling fruit-derived sucrose plus non-sucrose bound fructose intake was independently associated with a reduced risk of fatty liver by 13% (P = 0.033) among those with T2D In this German study, peripheral insulin sensitivity was impaired by moderate intake of sugar sweetened beverages In contrast, fruit-derived fructose intake was beneficial for liver fat content assessed with FLI Leone et al. [14] 2019 Cross-sectional analysis of 8103 Italian overweight and obese adults volunteering for participation in a structured weight loss program Anthropometric measurements were taken and biochemical parameters measured VAT and SAT were measured by ultrasonography FLI was higher in men and increased with increasing age, VAT, and SAT The sex*VAT, age*VAT, sex*SAT, and age*SAT interactions negatively contributed to FLI, indicating a lower VAT and SAT contribution to FLI in men and in the elderly for every 1 cm of increment Deposits of abdominal adipose tissue are associated with FLI. However, their contribution is sex and age dependent Klisic et al. [15]  The material includes a nationally representative age-and gender-stratified sample, which was drawn from the population register according to an international protocol The following criteria for exclusion were applied: clinically manifest liver disease, diabetes or abnormal oral glucose tolerance, ischemic heart or brain disease, chronic inflammatory diseases, malignancy, or active infection Lifestyle was estimated with a total score Nivukoski et al. [16] 2020 The occurrence of FLI ≥ 60% indicating fatty liver increased from 2.4% in men with zero risk factors to 81.9% in those with a total risk score of 7-8 (P < 0.0005 for linear trend) and in women from 0% to 73.5% (P < 0.0005). The most striking individual impacts on the likelihood for FLI above 60% were observed for physical inactivity (P < 0.0005 for both genders) and alcohol consumption (P < 0.0005 for men). Interestingly, coffee consumption was also found to increase with increasing risk factor scores (P < 0.0005 for linear trend in both sexes) women and in boys and girls [19] .

FLI and NAFLD
The methodological criteria useful to evaluate the accuracy of FLI at identifying fatty liver and more specifically NAFLD have been discussed elsewhere [22] . Seven published studies thus far have evaluated the capacity of FLI to detect NAFLD in various epidemiological scenarios [ Table 2] [23][24][25][26][27][28][29] .
Collectively, the studies summarized in Table 2 indicate that, while being a useful tool for evaluating NAFLD in high-risk populations [e.g., type 2 diabetes (T2D) and obstructive sleep apnea], in the individual patient, FLI has a limited liability of ruling in or out NAFLD [29] . This conclusion is in agreement with the original report by Bedogni et al. [7] and with the known limitations of prediction algorithms employed at the individual level [30] . Moreover, it suggests that FLI should be used for epidemiological rather than clinical purposes. In this connection, Fedchuk et al. [31] compared the performance and limitations of various biomarkers, FLI included, as related to the accepted NAFLD diagnostic standard, i.e., liver histology. It was found that, albeit being able to identify steatosis and insulin resistance, all non-invasive biomarkers had a limited clinical utility given that they are confounded by fibrosis and inflammation and do not accurately quantify fatty changes. It should be noted, however, that FLI was developed in the general population, and this should be taken into account when it is used for prediction purposes [22] . Klisic et al. [23] 2018 139 T2D patients (50.1% men) were cross-sectionally evaluated Anthropometric, blood pressure, and biochemical parameters were recorded Multivariate LRA showed HDL-c and MDA independently predicted higher FLI scores (OR = 0.056 and P = 0.029 and OR = 1.105 and P = 0.016, respectively) ROC curve analysis showed that the addition of fatty liver risk factors* to each analyzed biochemical parameter (HDL-c, non-HDL-c, hsCRP, MDA, and AOPP) in Model 1 increased the ability to discriminate patients with and without FL (AUC = 0.832, AUC = 0.808, AUC = 0.798, AUC = 0.824, and AUC = 0.743, respectively) Model 2 (which included all five predictors listed above) improved discrimination abilities for fatty liver status (AUC = 0.909) Additionally, Model 2 had both higher sensitivity and higher specificity (89.3% and 87.5%, respectively) than each individual predictor in Model 1 T2D patients at a high risk of fatty liver disease may be identified through a structured approach, including biomarkers of oxidative stress, dyslipidemia, and inflammation Chen et al. [24] 2019 326 consecutive adults with and 105 without NAFLD were recruited All were newly diagnosed with OSAHS Steatosis was diagnosed with US Accuracy and cutoffs of the FLI and HSI in detecting NAFLD were assessed with AUROC curve and the maximum Youden index analysis, respectively Both FLI and HSI values were significantly higher in patients with NAFLD than in controls The AUROC of FLI and HSI for predicting NAFLD was 0.802 (95%CI: 0.762-0.839) and 0.753 (95%CI: 0.710-0.793), respectively FLI had a significantly higher AUROC than HSI (P = 0.0383) The optimal cutoff value of FLI and HSI was 60 (sensitivity 66% and specificity 80%) and 35 (sensitivity 81% and specificity 60%) Both FLI and HSI can serve as screening tools for NAFLD in adults with OSAHS FLI performs better than HSI to this end Hsu et al. [25] 2019 From 9293 examinees who underwent routine health checkups, 4000 were enrolled, aged ≥ 20 years, with a BMI < 24 kg/m 2 in our lean-NAFLD study population. NAFLD diagnoses were made according to the patients' histories, laboratory values, and US criteria. Clinical variables, FPG, lipid, and liver profiles were evaluated using multiple LRA. The predictive ability and optimal cutoff values for NAFLD were determined according to the area under the ROC curve Overall, 18.5% (n = 740) of the lean population had NAFLD. Male sex, BMI, body fat mass, FPG, SUA, ALT, TG, and FLI values were associated with NAFLD. FLI had the best discriminative ability to predict lean-NAFLD compared to the other biochemical markers. Using the Youden index test, an optimum cutoff value for FLI of 15 was found to have the highest discriminant ability The prevalence of lean-NAFLD was not low. FLI was superior to other predictors including sex, liver function, and other metabolic factors, in the prediction of lean-NAFLD. FLI may be considered an easy to use, noninvasive marker to screen for lean-NAFLD Rabbitt et al. [26] 2020 Patients attending the AMU over a 3-month period were invited to participate. Those with excess alcohol consumption or pre-existing liver disease were excluded Using established FLI cutoffs, 414 participants were grouped into low (FLI ≤ 30), medium (30 < FLI ≤ 60), and high (FLI > 60) risk of NAFLD High-risk patients were offered review including LSM and CAP score In total, 134 patients were at low risk, 96 at medium risk, and 184 at high risk of NAFLD. Male sex (P < 0.0001) and increasing age (P < 0.0001) were associated with higher risk. Of the 120 high-risk patients who attended follow up, 13 participants had LSM > 7 kPa. Higher FLI scores were associated with higher CAP scores (P < 0.0001) but did not predict higher LSMs. FGP and HbA1c were found to be associated with higher LSM

FLI and the metabolic syndrome
It should be preliminarily noted that measurement of waist circumference and triglyceride serum concentration are included both in FLI and in MetS. Therefore, an agreement between FLI and MetS, in principle, has to be expected. Two publications have evaluated the association of FLI with the MetS.
In the first study, Khang et al. [32] evaluated the association between FLI and metabolic disorders and determined the cutoff value of FLI to screen for MetS. To this end, 10,107 adults aged ≥ 19 years from the Korean National Health and Nutrition Examination Surveys were selected. NAFLD, which was identified based on an increased FLI (≥ 60), after the exclusion of alcohol or viral liver disease, had an agestandardized prevalence = 10.0%. Individuals with the higher FLI scores had a higher prevalence of arterial hypertension, T2D, and MetS. At multivariate analysis, the group with higher FLI scores had a significantly higher risk for hypertension (

FLI vs. US-FLI
In 2012, by combining the ultrasonographic features of steatosis into a simple semi-quantitative index, Ballestri et al. [34] proposed the so-called "Ultrasonographic Fatty Liver Index" (US-FLI), which was shown to be significantly correlated with metabolic derangements and individual criteria for the histological diagnosis of NASH. Visual examples of the elementary components of ultrasonographic semeiotics of US-FLI have been published elsewhere. In short, the US-FLI scoring system ranges 2-8 based on the intensity of liver/kidney contrast, posterior attenuation of ultrasound beam, vessel blurring, difficult visualization of gallbladder wall, difficult visualization of the diaphragm, and areas of focal sparing. NAFLD is diagnosed by the minimum score 2. US-FLI, initially proposed to select which NAFLD patients should be submitted to liver biopsy, has been validated (reviewed in [35,36] ). Therefore, it is logical to ascertain whether FLI and US-FLI provide comparable clinical information. To answer this research question, Xavier et al. [37] enrolled 96 NAFLD patients in whom transient elastography was performed. They demonstrated that US-FLI was significantly superior to the FLI scores in discriminating between different grades of steatosis, but that the two scores should be applied together to obtain a more precise diagnosis of fatty liver and NAFLD.
The data reported in Table 3 consistently show that high FLI scores predict both subclinical atherosclerosis (intracranial vertebrobasilar stenosis, arterial stiffness, and left ventricle mass) and overt disease (incident cardiovascular disease, cardiometabolic disease, heart failure, and adverse major cardiovascular events). These studies are in full agreement with common notions on NAFLD being associated with subclinical atherosclerosis and cardiovascular events [1,[47][48][49][50] . Collectively, studies suggest that, at least for epidemiological purposes, FLI is a reliable marker of the full pre-clinical and clinical spectrum of atherosclerosis at various anatomic sites. Olubamwo et al. [41] 2019 Iwasaki et al. [44] 2021 FLI score was estimated among 2437 Japanese men. Employees of a single construction company submitted to mandatory annual health checkups and who were free of any history of CVD. baPWV was also measured at the beginning of the study and after a 3-year follow-up FLI was significantly correlated with the baPWV (r = 0.24, P < 0.01) Furthermore, the delta change of the FLI was significantly correlated with the delta change of the baPWV during the study period (r = 0.11, P = 0.01) FLI may be a marker of AS among Japanese men without any history of CVD Significant and positive associations between FLI and LVM (BHS: β = 0.59, P < 0.001; YFS: β = 0.41, P < 0.001) and LVMI (BHS: β = 0.14, P < 0.001; YFS: β = 0.09, P < 0.001) were found in both study cohorts The association of FLI with LVMI was stronger in women than men (BHS: Pinteraction = 0.01; YFS: P-interaction < 0.01), and the relationship between FLI and LVM/LVMI was stronger in black than white individuals (LVM: Pinteraction = 0.02; LVMI: P-interaction = 0.04) Li et al. [45]

FLI, PREDIABETES, AND DIABETES
Robust evidence supports the notion that NAFLD is not only an effect of pre-existent impaired glucose tolerance and T2D but also a precursor of incident T2D and MetS [51][52][53] .
The studies summarized in Table 4 support the conclusion that there is a direct dose-response association between FLI scores and risk of incident T2D [59] , as FLI ≥ 60 specifically predicted T2D among men without MetS [57] . Consistently, FLI < 30 predicts prediabetes reversal, particularly among individuals with a healthy lifestyle [61] . Therefore, in a primary care setting, FLI may screen individuals to be submitted to aggressive intervention to prevent the progression of prediabetes to overt T2D [56] . This is an originally unexpected but logical utilization of FLI based on the pathophysiology of the NAFLD-T2D association [62][63][64] . If FLI is indeed associated with incident T2D, we can postulate that FLI is also able to identify such a link between NAFLD and CKD.

FLI AND CHRONIC KIDNEY DISEASE
Recently, research and clinical interest has been raised on the independent association of NAFLD with CKD [5] . Two published studies thus far have used FLI to All IFG groups were significantly associated with incident T2D irrespective of FLI scores FLI is associated with the development of T2D regardless of sex and the presence or absence of IFG, and it may be a useful predictor of future risk of incident T2D even in individuals without IFG Franch-Nadal et al. [56] 2018 FLI was calculated at baseline for 1142 adult subjects with prediabetes attending primary care centers and classified into three categories: no steatosis (FLI < 30), intermediate (FLI: 30-60) and hepatic steatosis (FLI ≥ 60) The incidence rate of T2D in each FLI category was assessed at 3 years of follow-up and calculated using fully adjusted* Cox regression models The proportion of subjects with prediabetes and hepatic steatosis (FLI ≥ 60) at baseline was 55.7% The incidence rate of T2D at 3 years follow-up was 1.3, 2.9, and 6.0 per 100 personyears for FLI < 30, FLI 30 to < 60, and FLI ≥ 60, respectively The most significant variables increasing the risk of developing T2D were MetS (HR = 3.02; 95%CI: 2.14-4.26) and FLI ≥ 60 (HR = 4.52; 95%CI: 2.10-9.72). Moreover, FLI ≥ 60 was independently associated with T2D incidence: the HR was 4. Wargny et al. [58] 2019 The IT-DIAB study, a 5-year, prospective, observational study carried out in occupational centers based in three French cities, included 389 individuals with prediabetes, defined as FPG ≥ 100 and ≤ 125 mg/dL. NOD conversion was defined as a first FPG value ≥ 126 mg/dL and prediabetes reversion as a first FPG value < 110 mg/dL The associations of both events with baseline FLI were studied separately using multivariate Cox models After a median follow-up of 3.9 years (range: 0.  [66] evaluated the risk of CKD (defined by either estimated glomerular filtration rate < 60 mL/min/1.73 m 2 or positive for urinary protein during a 10-year followup) in 14,163 subjects (male/female: 9077/5086) subjects submitted to annual health examinations. Multivariable Cox regression with restricted cubic spines adjusting for confounders showed that hazard ratios (HRs) of CKD development increased with increasing FLI at baseline in both men and women, and adding FLI to conventional CKD risk factors resulted in a significant improvement in predicting CKD, suggesting that, in a general population cohort study, high FLI scores predict incident CKD in either sex.

FLI AND ENDOCRINE DERANGEMENTS
NAFLD has also been associated with a variety of endocrine derangements [67][68][69] , some of which predisposing to secondary NAFLD forms, whereas other endocrinopathies probably result from pre-existent NAFLD [70] . Does FLI have a role in this setting? Two studies seem to suggest so, although this is a scarcely explored area.
Liu et al. [71] , by studying 552 Taiwanese aging men, found that FLI scores were associated with the risk of testosterone deficiency, especially in those without MetS.
Ahn et al. [72] , in their study on 4264 Koreans, found a novel nexus linking liver and bone that increases the risk of osteoporosis in men with NAFLD.
Clearly, much research remains to be conducted to ascertain which other endocrinopathies may be associated with FLI scores.

FLI AND TUMORS
Various pathomechanisms potentially link NAFLD and various types of tumors, colon adenoma and carcinoma in particular [73] . Three studies regarding FLI and tumors have been published thus far, two of them focusing on colorectal adenoma and carcinoma.
In the first study, Ze et al. [74] , based on a retrospective observational study on 2976 consecutive > 40-year-old subjects undergoing routine checkups, found that a high FLI may be useful in predicting colorectal adenoma in relatively healthy Asian populations.
A second study, by Choi et al. [75] , was conducted in Korea on data from the National Health Insurance Corporation 2009 to 2012. Although FLI ≥ 60 was associated with colorectal cancer (CRC) regardless of BMI, the association was more prominent among individuals with a normal BMI. In particular, NAFLD was more closely associated with CRC in the absence of T2D, hypertension, or dyslipidemia than when (one or more of) these conditions were present.
The third study regards FLI and breast cancer. Park et al. [76] , using the Korean National Health Insurance Corporation, found that FLI scores of 30-60 and ≥ 60 were significantly associated with increased breast cancer risk in post-menopausal women hazard ratio (HR = 1.07, 95%CI: 1.04-1.11; and HR = 1.11, 95%CI: 1.05-1.17, respectively), while no such an association was found in pre-menopausal women.

DOES FLI PREDICT MORTALITY?
Whether FLI is able to assess the risk of death has to be answered cautiously because of the many methodological issues associated with the identification of independent risk factors for mortality [49] . Given that NAFLD carries an excess of mortality owing to cardiovascular, cancer, and liver-related causes [77] , it is plausible that FLI may be a good marker of increased risk of mortality. Three studies addressed this research question.
Lerchbaum et al. [78] , by calculating FLI scores among 3270 subjects submitted to coronary angiography, found that, following a median follow-up time of 7.7 years, patients with high FLI scores compared to those with the lowest FLI scores were independently associated with increased mortality owing to all-causes, cardiovascular causes, and non-cardiovascular causes. The excess risk owing to fatal cancer was of borderline significance.
Based on a median 29-year follow-up of a cohort of 1552 middle-aged men from the Kuopio Ischemic Disease Risk Factor Study, Setti et al. [79] found that those men who had both renal hyperfiltration (RHF)which was associated with smoking -and fatty liver evaluated with FLI scores -which was associated with obesity -had the highest risk of mortality owing to all causes (HR = 1.96, 95%CI: 1.27-3.01). Conversely, having fatty liver associated with normal estimated glomerular filtration rate modestly increased the risk of all-cause mortality (HR = 1.35, 95%CI: 1.09-1.66). Finally, intermediate-risk profiles of all-cause mortality were found among those men who had RHF associated with normal FLI scores. The risk of mortality owing to cardiovascular causes was associated with RHF, rather than with FLI scores. Collectively, the data suggest that RHF and FLI scores are strongly associated with mortality owing to all causes as well as due to cardiovascular causes.
Using a study population of about 3 million individuals submitted to repeated evaluation for health screening purposes over four years, Lee et al. [80] evaluated whether FLI measurements repeated over time could predict incident myocardial infarction (MI), stroke and mortality owing to all causes. They defined "FLI points" as the number of times, ranging from zero to four, participants exhibited FLI scores ≥ 60. This study found that that the higher are the FLI points, the higher is the risk of mortality owing to all causes, MI, and stroke (P for trend < 0.001, all). After adjustment for demographic confounders, metabolic cofactor, lifestyle habits, and income, those individuals with four FLI points had a higher risk of mortality owing to all causes (aHR = 1.86, 95%CI: 1.75-1.98, P < 0.001), incident MI (aHR = 1.3, 95%CI: 1.21-1.40, P < 0.001), and incident stroke (aHR = 1.27, 95%CI: 1.19-1.37, P < 0.001). By comparing the first to the last FLI points, the group of individuals with "incident NAFLD" exhibited an increased hazard of mortality compared to the "no NAFLD" group (aHR = 1.46, 95%CI: 1.37-1.55). Consistently, the "regression of NAFLD" group compared to the group with "persistent NAFLD" showed a decreased mortality risk (aHR = 0.83, 95%CI: 0.77-0.89). This study supports the notion that repeating evaluations of FLI scores over time may allow a better profiling of the risks of mortality, MI, and stroke. Moreover, changes of FLI scores over time may help clinicians in evaluating the efficacy of NAFLD treatment and re-modulating prognosis of these patients.

CONCLUSIONS AND RESEARCH AGENDA
Historically, FLI was proposed in the epidemiological arena as a surrogate index of NAFLD to be used for the identification of cases with suspected NAFLD to be submitted to further ultrasonographic assessment. The data presented in the present SANRA review demonstrate that this primary aim of FLI scores has now been largely overcome by a plethora of other indications. These span all aspects from diagnosis of NAFLD to its (mainly extra-hepatic) manifestations and complications such as atherosclerosis, diabetes, CKD, and tumors.
Importantly, repeating FLI scores over time may allow a non-invasive prediction of overall mortality and serve as a surrogate marker of NAFLD treatment response [81,82] and a useful tool for selecting T2D patients to submit to liver biopsy [83] . We expect that the future of FLI will see a further growth in the use of this simple biomarker in several metabolic diseases, NAFLD among them. Conversely, little has been published regarding the ability of FLI to predict liver-related outcomes such as cirrhosis and HCC.
While being based on robust markers of NAFLD pathophysiology, FLI should also be improved by incorporating major modifiers of NAFLD epidemiology, namely age, sex, and reproductive status [84,85] , which were originally left out from the FLI algorithm. Of course, the fact that age and sex were left out from the multivariable model which gave birth to the FLI in a single population does not imply that they could not be predictors of NAFLD or other NAFLD-associated outcomes in different populations. The effect of age and sex and other variables of interest can be studied by using them as predictors of a given outcome together with FLI [30] .
There are two additional limitations to the use of FLI in clinical practice. The first is the measurement of waist circumference, which, regrettably, tends to be disregarded in the general practice. The second limitation is the "grey zone" of indeterminate FLI scores, which is sex and age dependent and averaged 27.5% in a recent study [14] . The best diagnostic strategy to follow among this substantial proportion of cases remains to identified.
In conclusion, additional studies are eagerly awaited given the importance of FLI as a non-invasive biomarker of NAFLD both in clinical practice and in the research arena.