Gastroenterology
Volume 137, Issue 3 , Pages 772-775, September 2009

Nonalcoholic Fatty Liver Disease Liver Fat Score and Fat Equation to Predict and Quantitate Hepatic Steatosis: Promising But Not Prime Time!

  • Naga Chalasani

      Affiliations

    • Corresponding Author InformationReprint requests Address requests for reprints to: Naga Chalasani, MD, Professor and Director, Division of Gastroenterology and Hepatology, Indiana University School of Medicine, RG 4100; 1050 Wishard Boulevard, Indiana, Indiana 46202. phone: (317) 278-0414; fax (317) 278-1949

published online 27 July 2009.

Article Outline

 

See “Prediction of non-alcoholic fatty liver disease and liver fat using metabolic and genetic factors” by Kotronen A, Peltonen M, Hakkarainen A, et al, on page 865.

Nonalcoholic fatty liver disease (NAFLD) is a very common clinical condition that is frequently encountered by primary care physicians, gastroenterologists, endocrinologists, and gynecologists.1 It encompasses a spectrum of liver histologic abnormalities ranging from simple steatosis to steatohepatitis, advanced fibrosis, and cirrhosis. Recent studies have demonstrated that NAFLD is a component of the metabolic syndrome and is an independent predictor of cardiovascular disease.1 Although a definitive diagnosis of NAFLD requires liver biopsy, it is often diagnosed based on elevated liver biochemistries along with evidence of steatosis by ultrasound in an appropriate clinical setting (eg, no significant alcohol consumption). However, routine liver biochemistries and ultrasound imaging lack sufficient sensitivity, and moreover, they do not have the ability to quantify the degree of steatosis. Several recent studies have shown that proton magnetic resonance spectroscopy (MRS) is very sensitive in identifying steatosis and can reproducibly quantify hepatic steatosis.2, 3

In this issue of Gastroenterology,4 Kotronen et al take an important step forward and report 2 mathematical equations that predict the presence of hepatic steatosis and its quantity as identified by proton MRS in a relatively large cohort of Finnish adults. Five variables (presence of metabolic syndrome, type 2 diabetes mellitus, fasting insulin, fasting aspartate aminotransferase [AST], and the AST/alanine aminotransferase [ALT] ratio) were associated independently with NAFLD as diagnosed by proton MRS.4 Based on these 5 variables, the authors constructed 2 mathematical equations, namely the “NAFLD liver fat score” and the “liver fat equation,” to diagnose NAFLD and estimate liver fat content respectively with very good sensitivity and specificity. In addition, they validated the observation made elsewhere that the single nucleotide polymorphism rs 738409 in PNPLA3 gene (coding for adiponutrin gene) is a predictor of hepatic fat content,5 but its addition did not improve the performance of the NAFLD liver fat score.4

Kotronen et al should be congratulated for this and a number of other important contributions that have enhanced our understanding of hepatic fat in nonalcoholic individuals with metabolic risks such as obesity and type 2 diabetes. Because we live in a society of unbearable health care expenditure, studies that attempt to develop less expensive surrogates for advanced but expensive technology are highly desirable. However, several aspects of the study by Kotronen et al require further discussion and these prediction models need to be validated externally in independent cohorts before incorporating them into clinical practice or research studies.

This study defined the presence of NAFLD as liver fat ≥55.6 mg triglyceride (TG)/g liver tissue or ≥5.56% of liver tissue weight based on proton MRS.4 This definition was derived from the study by Szczepaniak et al,3 who analyzed the MRS data from 2287 participants of the Dallas Heart Study. Of these, 345 subjects were presumed to be “normal” because of the lack of obvious risk factors for steatosis, and hepatic TG content >5.56% (or 55.6 mg/g liver tissue) represented the 95th percentile in the distribution of hepatic TG content, and thus was defined as “abnormal.”3 It is unclear if a criterion developed on “normal” multiethnic individuals from Dallas can be generalized to Finnish adults. It would be optimal if Kotronen et al validated this definition in Finnish adults without risk factors for steatosis.

Nearly 15 years ago, Longo et al6 demonstrated that proton MR spectrometry can be used to predict hepatic steatosis in humans. In this study, 29 patients with likely NAFLD underwent percutaneous liver biopsy for histomorphometry and in vivo proton MRS, and in vitro MRS was additionally done in a subset on liver biopsy specimens (n = 6). The authors showed a statistically significant relationship between fat percentage estimated by proton MRS and by histomorphometry (r = 0.70). In a more recent study comprising 38 human liver biopsy samples, we have shown that fat content quantified by in vitro MRS correlated significantly with hepatic TG content (r = 0.63; P = .004) and histologic grading by pathologists (r = 0.61; P = .006).7 Taken together, these studies indicate that proton MRS can quantify hepatic steatosis in a reliable fashion.

The presence of hepatic steatosis not only identifies those at risk to have (or possibly develop) steatohepatitis and hepatic fibrosis, but is associated also with hepatic insulin resistance and is an independent predictor of cardiovascular disease.1 MRS is by far the most sensitive imaging tool available to identify hepatic steatosis, but it is expensive and is entirely a research tool available at selected institutions. Therefore, Kotronen et al developed the “NAFLD liver fat score” and the “liver fat equation” to predict the presence and quantity of hepatic fat as assessed by proton MRS.4 In their study, a liver fat score above −0.640 predicted the presence of steatosis with 86% sensitivity and 71% specificity. This is good for a prediction model that is based on easily available blood tests and clinical data. Before one accepts these prediction models as sensitive tools for identifying and quantifying steatosis, however, they need to be validated externally in non-Finnish individuals. In addition, the “NAFLD liver fat score” needs to be compared against a range of serum ALT values. For example, ALT values >31 U/L in men and >19 U/L in women are very sensitive in identifying those with steatosis, but may lack adequate specificity.8 Kotronen et al did not find a significant relationship previously between ALT slope and liver fat (measured by MRS) in 70 nondiabetic obese subjects.9

Interestingly, there seems to be redundancy in the variables included in these equations, and thus, one wonders if they can be simplified. Although diabetes mellitus is part of the metabolic syndrome and insulin levels are so intricately related to diabetes, they both are included in the equations in addition to the metabolic syndrome. Similarly, in addition to the AST/ALT ratio, fasting AST is included as a separate variable. One should explore if a prediction model with comparable performance can be constructed just with 2 variables—the metabolic syndrome and the AST/ALT ratio.

What is the utility of quantifying hepatic steatosis in daily clinical practice? The relevance of the severity of steatosis in the pathogenesis or progression of steatohepatitis continues to be unclear. In a cross-sectional study, we have shown that patients with severe steatosis were more likely to have steatohepatitis, lobular inflammation, and zone 3 fibrosis, but not ballooning or cirrhosis.10 But it has been argued recently that accumulation of TG is perhaps a protective mechanism by which potentially toxic free fatty acids are converted into neutral fats.11 We have all encountered patients who have severe steatosis by imaging, but when biopsied liver histology shows very little in the form of hepatocyte ballooning or fibrosis. Assessing the severity of steatosis by either MRS or the Finnish hepatic fat equation does not seem to be advantageous clinically at this time. It is not the severity of steatosis, but the presence of nonalcoholic steatohepatitis (NASH) and fibrosis that dictates long-term hepatic prognosis. In clinical practice, in a patient with NAFLD, 2 questions that must be answered include the following: (1) Does he or she have steatohepatitis or advanced fibrosis? and (2) Does he or she have multiple metabolic comorbidities that significantly elevate the risk for cardiovascular events? However, quantifying hepatic steatosis might be of value in drug development in terms of identifying therapeutic and toxicity signals.

Currently, phase II therapeutic studies of NASH generally use ALT as an end point, but noninvasively assessed steatosis grade might be a better end point (liver histology is impractical in phase II studies). Improvement in steatohepatitis (or ballooning or fibrosis) is almost always accompanied by improvement in steatosis grade. Therefore, an improved steatosis grade might serve as a potential therapeutic signal. Similarly, there is significant interest among several pharmaceutical companies to monitor steatosis as a side effect to various experimental therapies (eg, MTP inhibitors, anti-psychosis agents). Currently, these companies are using MRS for monitoring this potential side effect, but a validated clinical model that predicts and quantitates steatosis would be highly desirable.

In this author's opinion, if externally validated, the NAFLD liver fat score can serve an important role in identifying those without hepatic steatosis for case-control genetic and nongenetic research studies. For example, it has been difficult to design and conduct genome-wide association studies in NAFLD because it is not easy to exclude steatosis in matched controls. Until now, unexplained elevation in ALT has been used as a surrogate for NAFLD in population-based epidemiological studies (eg, NHANES III). But, if externally validated, the liver fat score would be a better surrogate for NAFLD itself than unexplained elevation in serum ALT levels. Although there have been numerous clinical models to predict advanced fibrosis in patients with NAFLD,1 clinical decision aids developed by Kotronen et al are the first to predict NAFLD and, upon validation, they might be useful in selected situations as noted.

In a previous study, Romeo et al reported the relationship between genome-wide survey of 9229 nonsynonymous SNPs and hepatic fat detected by MRS in 2111 multiethnic adults residing in Dallas county.12 A single variant in PNPLA3 (rs738409) was associated very strongly with hepatic fat content (P = 5.9 × 10−10). This variant is a cytosine to guanine substitution that changes codon 148 from isoleucine to methionine (PNPLA3-I148M). The effect of this variant seems to be quite striking in Hispanics, who have the greatest risk of NAFLD. For example, Hispanic-Americans who are homozygous for PNPLA3-I148M allele had ∼8% hepatic fat in comparison with ∼3.5% hepatic fat in those who are not carriers. This observation has now been confirmed by 2 additional studies, one from Argentina13 and another by Kotronen et al14 (Table 1).

Table 1. Studies Evaluating the Relationship Between PNPLA3 Polymorphism and NAFLD
AuthorStudy designEnd point(s)Key resultsComments
Romeo et al11GWAS of 2,111 multiethnic adults residing in Dallas countyLiver fat by MRSStrong relationship between PNPLA3 GG genotype (I148M) and hepatic steatosis (P = 5.0 × 10−10). This relationship persisted even after controlling for diabetes and alcohol.Majority of patients had at least moderate alcohol consumption16; histology not available.
Kotronen et al13Genotyping for PNPLA3 variability in 291 Finnish adultsLiver fat by MRS and hepatic PNPLA3 expression in a subset of 32 patientsPNPLA3 GG genotype (I148M) had significantly higher hepatic fat compared with CG or CC genotypes, even after controlling for age, gender, and BMI. PNLPA3 genotypes did not influence insulin levels or hepatic insulin resistance.Hepatic PNPLA3 expression increased with obesity, but no significant relationship between hepatic PNPLA3 mRNA expression and genotype was reported; histologic correlates not made.
Sookoian et al12Genotyping for PNPLA3 polymorphism in 266 individuals (172 NAFLD and 94 controls) from ArgentinaNAFLD Liver histologyPNPLA3 GG genotype (I148M) was associated with 2-fold increased risk of NAFLD, even after controlling for age, gender, BMI, and HOMA; NAFLD patients with PNPLA3 GG genotype had significantly higher NAFLD severity score compared with GC or CC genotypes (P < .00001), even after controlling for covariates.Controls are not BMI matched and thus one could not completely exclude an interaction with BMI; it is unclear how NAFLD severity was calculated.
Yuan et al17GWAS in 7715 individuals in development cohort and 4704 in validation cohortPlasma levels of liver enzymesPNPLA3 GG genotype was associated with increased levels of ALT and ASTThis is a population-based GWAS.

BMI, body mass index; GWAS, genome-wide association studies.

Interestingly, in a discrete trait analysis of 103 NAFLD patients with liver histology (40 with simple steatosis and 63 with NASH), individuals with NASH had significantly greater prevalence of PNPLA3-I148M homozygosity compared with those with simple steatosis alone (52% vs 30%; χ2 P = .04).13 However, the functional significance of PNPLA3 is not known. It encodes a transmembrane protein (adiponutrin) and is predominantly expressed in white and brown adipose tissue. Unlike adipose TG lipase (ATGL or PNPLA2), another closely related patatin-like phospholipase that is involved actively in TG lipolysis, PNPLA3 may be more related to lipid storage in adipose tissue. Adipose tissue PNLPA3 expression correlates significantly with body mass index in obese children and its expression was very elevated in obese and diabetic rats.14, 15

Where do we go from here? Other investigators with access to large MRS datasets (ie, Dallas Heart Study investigators) should consider examining the generalizability of NAFLD fat score in a different patient population and should also explore whether the equation can be simplified without significantly sacrificing its performance. Similarly, some clinical trialists might have an opportunity to explore their datasets and examine the relevance of these models in the drug development arena. PNPLA3 seems to be a very strong candidate for further investigations in terms of its relationship to pathogenesis and progression of steatohepatitis. Several ongoing NAFLD genome-wide association studies with histology end points might provide additional insight into its role in the pathogenesis of NASH.

Back to Article Outline

References 

  1. Vuppalanchi R, Chalasani N. Nonalcoholic fatty liver disease and non-alcoholic steatohepatitis: selected practical issues in their management. Hepatology. 2009;49:306–317
  2. Cassidy FH, Yokoo T, Aganovic L, et al. Fatty liver disease: MR imaging techniques for the detection and quantification of liver steatosis. Radiographics. 2009;29:231–260
  3. Szczepaniak LS, Babcock EE, Schick F, et al. Measurement of intracellular triglyceride stores by 1H spectroscopy: validation in vivo. Am J Physiol Endocrinol Metab. 1999;276:977–989
  4. Kotronen A, Peltonen M, Hakkarainen A, et al. Prediction of non-alcoholic fatty liver disease and liver fat using metabolic and genetic factors. Gastroenterology. 2009;137:865–872
  5. Romeo S, Kozlitina J, Xing C, et al. Genetic variation in PNLPA3 confers susceptibility to nonalcoholic fatty liver disease. Nat Genet. 2008;40:1461–1465
  6. Longo R, Pollesello P, Ricci C, et al. Proton MR spectroscopy in quantitative in vivo determination of fat content in human liver steatosis. J Magn Reson Imag. 1995;5:281–285
  7. Vuppalanchi R, Cummings OW, Ulbright T, et al. Hepatic steatosis (Relationship among histological, radiological and biochemical assessments). J Clin Gastroenterol. 2007;41:206–210
  8. Prati D, Taioli E, Zanella A, et al. Updated definitions of healthy ranges for serum alanine aminotransferase levels. Ann Intern Med. 2002;137:1–10
  9. Kotronen A, Juurinen L, Hakkarainen A, et al. Liver fat is increased in type 2 diabetic patients and underestimated by serum alanine aminotransferase compared with equally obese nondiabetic subjects. Diabetes Care. 2008;31:165–169
  10. Chalasani N, Wilson L, Kleiner DE, et al. NASH Clinical Research Network Relationship of steatosis grade and zonal location to histological features of steatohepatitis in adult patients with nonalcoholic fatty liver disease. J Hepatol. 2008;48:829–834
  11. Choi SS, Diehl AM. Hepatic triglyceride synthesis and nonalcoholic fatty liver disease. Curr Opin Lipidol. 2008;19:295–300
  12. Sookoian S, Castano GO, Burgueno AL, et al. A nonsynonymous gene variant in the adiponutrin gene is associated with nonalcoholic fatty liver disease severity. J Lipid Res. 2009 May 20;[Epub ahead of print]
  13. Kotronen A, Johansson LE, Johansson LM, et al. A common variant in PNPLA3, which encodes adiponutrin, is associated with liver fat content in humans. Diabetolgia. 2009;52:1056–1060
  14. Steinberg GR, Kem BE, Watt MJ, et al. Adipocyte triglyceride lipase expression in human obesity. Am J Physiol Endocrinol Metab. 2007;293:E958–E964
  15. Baulande S, Lasnier F, Lucas M, et al. Adiponutrin, a transmembrane protein corresponding to a novel dietary and obesity linked mRNA specifically expressed in the adipose lineage. J Biol Chem. 2001;276:33336–33344
  16. Browning JD, Szczepaniak LS, Dobbins R, et al. Prevalence of hepatic steatosis in an urban population in the United States: Impact of ethnicity. Hepatology. 2004;40:1387–1395
  17. Yuan X, Waterworth D, Perry JRB, et al. Population-based genome-wide association studies reveal six loci influencing plasma levels of liver enzymes. Am J Human Genetics. 2008;83:520–528

 Conflicts of interest The author discloses the following: Dr Chalasani served within the past 12 months as a paid consultant to KariBio, Metabasis, Atherogenics, Advanced Life Sciences, Johnson and Johnson, Debiovision, Eli Lilly, and Amylin. Within the past 12 months, he has participated in clinical studies with Gilead and Monarch LifeSciences.

 Funding Supported in part by K24 DK 069290.

PII: S0016-5085(09)01188-3

doi:10.1053/j.gastro.2009.07.032

Refers to article:

  • Editorial Accompanies ArticleAdditional Online Content Available Prediction of Non-Alcoholic Fatty Liver Disease and Liver Fat Using Metabolic and Genetic Factors , 15 June 2009

    Anna Kotronen, Markku Peltonen, Antti Hakkarainen, Ksenia Sevastianova, Robert Bergholm, Lina M. Johansson, Nina Lundbom, Aila Rissanen, Martin Ridderstråle, Leif Groop, Marju Orho–Melander, Hannele Yki–Järvinen
    Gastroenterology September 2009 (Vol. 137, Issue 3, Pages 865-872)

Gastroenterology
Volume 137, Issue 3 , Pages 772-775, September 2009