The natural history of non-alcoholic fatty liver disease in children: a follow-up study for up to 20 years

The main demographic and clinical features are summarized in Table 1 . There was a slightly higher proportion of boys than girls, and two thirds were obese. Most patients had symptoms or signs at presentation. The features of the metabolic syndrome were common with more than half having a BMI >97 th percentile. Fifty-five (83.3%) children presented with at least one feature of the metabolic syndrome whereas overt metabolic syndrome (i.e., ≥3 features) was present in 19 (28.8%) children. Other features that worse the cardiovascular risk profile such as hypercholesterolemia and high LDL-cholesterol were also common. The main laboratory data gathered at the time of NAFLD diagnosis are summarized in Table 2 . ALT and AST levels were each within the normal range in few patients. The AST/ALT ratio was greater than 1 in 29% of children. GGT was elevated in 88% of patients while serum alkaline phosphatase was above the normal value for age and sex in few patients. Serum total bilirubin, albumin and prothrombin time were essentially within the normal range in all patients. Positive autoantibodies in low titers were found in 20% of patients including ANA in 15.4%, and SMA in 10% of patients. The main liver biopsy features are summarized in Table 3 . Some degree of fibrosis was present in 59% of children including mild fibrosis (stage 1-2) in 11, septal/bridging fibrosis (stage 3) in 4, and cirrhotic-stage disease in 2. The mean NAFLD activity score was 3.5 ± 1.02. Portal based injury was seen in 9 (31%) children but associated with zone 3 injury in most. Interface hepatitis or other features suggestive of autoimmune hepatitis were not present in any case. As summarized in table 4 , patients undergoing liver biopsy had significantly higher levels of ALT and lower levels of total cholesterol and triglycerides; otherwise patients undergoing liver biopsy were similar to those who were not biopsied. The mean follow-up of the total cohort was 6.4 ± 4.5 (range 0.05 to 20) years, for a total of 409.6 person-years. During this time, treatment recommendations included lifestyle modifications consisting of an exercise program along with diet modifications tailored to individual need and preference alone or in 8 (12.1%) patients in combination with either ursodeoxycholic acid or vitamin E. One-year after initiation of the prescribed lifestyle modification, 49% of children were able to lose at least 10% of their baseline weight, and 86% of them showed significant improvement or normalization of aminostrasferases. Neither ursodeoxycholic acid nor vitamin E treatment appeared to impact further the liver enzymes levels although their effect independent of weight loss could not be assessed due to the very small number of patients on either treatment. At the time of last follow-up, however, most patients (76%) had re-gained weight, and in 46% of them aminotransferases returned to baseline values. Interestingly, 4 children developed type II diabetes 4, 6, 7 and 11 years after the diagnosis of NAFLD. Other complications that occurred during follow-up were cholecystitis requiring cholecystectomy (6 patients), morbid obesity requiring bariatric surgery (2 patients), contraceptive induced liver injury (1 patient) and bilateral oophorectomy and hysterectomy for endometriosis (1 patient). A total of 13 liver biopsies were obtained from 5 patients over a mean period of 41.4 ± 28.8 months ( Table 5 ). Grade of steatosis and lobular inflammation either worsened or remained the same in all patients. Progression of fibrosis stage was documented in 4 cases. One patient without fibrosis at presentation developed stage 1 fibrosis at 19 months, and cirrhosis (stage 4 fibrosis) at 57 months. Another patient presented without fibrosis, but progressed to stage 1 fibrosis at 39 months, and to stage 3 fibrosis at 82 months. Two other patients without fibrosis at presentation progressed both to stage 1 fibrosis, one at 28 months and the other at 7 months. There was no improvement in any of those histological features in any case. During follow-up, 2 patients underwent liver transplantation, and 2 additional patients died. The observed number of events in the total person-years of follow up was 4/409.6 = 9.8 per thousand. The observed number of events in the NAFLD cohort was significantly higher than the expected number of events in the United States population of same age and sex (4 vs. 0.29416, p<0.00001) with a SMR of 13.6 (95% confidence intervals, 3.8, 34.8). The observed survival free of liver transplantation in the NAFLD cohort as compared to the expected survival of the general United States population of the same age and sex is illustrated in Figure 1 . The 2 patients who underwent liver transplantation were those 2 who presented with cirrhosis on liver biopsy. The first case was a Hispanic female diagnosed with cirrhotic-stage NASH at 11 years of age, when she presented with a BMI of 26.9 kg/m 2 , hypercholesterolemia and hypertriglyceridemia. She was found with grade 3 esophageal varices and developed recurrent variceal bleeding requiring variceal band ligation in multiple occasions. She underwent liver transplantation at 20 years of age due to end-stage liver disease and hepatopulmonary syndrome. In the post-transplant period, she was diagnosed with recurrent NASH at 9 months, stage1 fibrosis at 2 years and 3 months, and stage 2 fibrosis at 3 years and 3 months post liver transplantation. She is currently alive. The second case was a white female diagnosed with cirrhotic-stage NASH at 18.9 years of age when presented with a BMI of 33.6 kg/m 2 , and low HDL-cholesterol. She developed severe hypoxemia from hepatopulmonary syndrome without any other liver complication requiring liver transplantation at 25 years of age. She was found with macrovesicular steatosis on protocol liver biopsy as early as 14 days post liver transplantation, with well-established NASH at 6 weeks post liver transplantation, and with bridging fibrosis at 1 year. She was diagnosed with cirrhotic stage NASH in the graft and hepatopulmonary syndrome 2 years post liver transplantation, requiring re-transplantation 2.3 years after the first liver transplant procedure. Finally, she died from multiple organ failure at age 27 years. The 2 deaths recorded were both non-liver related, and none of these two cases had liver biopsy performed at any time.

The study is the first to describe the long-term survival of children with NAFLD who underwent a follow-up of up to 20-years. The study demonstrates that NAFLD in children is a disease of progressive potential. Some children presented with cirrhosis, other progressed to advanced fibrosis or cirrhosis during follow-up, and some developed end-stage liver disease with the consequent need of liver transplantation. The study shows that NAFLD in children is associated with a significantly shorter long-term survival as compared to the expected survival of the general population of the same age and sex; our children with NAFLD had a 13.8-fold higher risk of dying or requiring liver transplantation than the general population of same age and sex. The two deaths recorded were not liver-related, but the inclusion of these two cases among the 4 cases reaching the outcome of death or liver transplantation is appropriate as the comparison was done to overall mortality in the general population of same age and sex regardless of the causes of death. The study also provides interesting data regarding the progressive potential of NAFLD to more advanced disease. Four of the five children with repeated liver biopsy did not have fibrosis on diagnosis liver biopsy, but two developed mild (stage 1) fibrosis, and the other two developed advanced (stage 3-4) fibrosis. The progression of liver damage in these patients over a relatively short period of time highlights the importance of identifying those children with NAFLD who are at risk of having a more progressive liver disease. In some recent series, presence and severity of fibrosis was consistently associated with a higher BMI or larger waist circumference. 10 , 12 , 14 Older age and higher levels of AST and insulin have been found associated with fibrosis in some series 10 , 12 However, further studies are needed to accurately identify those children who are more likely to progress to end stage liver disease. Interestingly, the two patients in our cohort who underwent liver transplantation had hepatopulmonary syndrome as the main indication for transplant. However, whether or not there is an association between progression to cirrhosis and development of severe hepatopulmonary syndrome requiring liver transplantation in pediatric NAFLD remains uncertain, and further studies in this area are needed. It is also intriguing that both cases undergoing liver transplantation in our series developed recurrent NASH, with cirrhotic stage disease in one patient who required re-transplantation. Recurrence of NASH post liver transplantation in children has been documented in two isolated cases 34 , 35 both male patients of age 13 and 16 years who developed decompensated liver disease from NAFLD. Both cases had a history of hypothalamic/pituitary dysfunction in one case associated with hepatopulmonary syndrome. These two cases 34 , 35 extended prior observations of the development of severe liver disease from NAFLD in patients with hypothalamic/pituitary dysfunction. 36 Similar to other pediatric series of NAFLD, most of our children were diagnosed in the second decade of life, girls and boys were affected almost equally with a slight male predominance, and most were symptomatic at presentation. Alike adults with NAFLD, a high proportion of our children were obese and had associated several features of the metabolic syndrome. Unlike adults, almost a third of our children had portal-based injury on liver biopsy, but most of them had pericentral/perisinusoidal injury as well. Two of our children had type II diabetes prior to the diagnosis of NAFLD, whereas four patients developed type II diabetes within 11 years after NAFLD was diagnosed. Therefore, children with NAFLD should be closely monitored for development of type II diabetes later in life. We found a high proportion of children (27.3%) with HDL-cholesterol below the 5 th percentile for their age and sex which has not been reported in pediatric NAFLD before. Similar to adults with NAFLD, 37 20% of our children tested positive for low titer of ANA and/or SMA. Interestingly, the vast majority (88%) of our children had elevated GGT with alkaline phosphatase levels within the normal range in most of them. To our knowledge, this high proportion of children with elevated GGT levels has not been described in any other series of pediatric NAFLD. Recently, higher serum levels of GGT have been associated with several cardiovascular disease risk factors or components of the metabolic syndrome. 38 - 41 GGT is located on the external surface of most cells and mediates the uptake of glutathione, an important component of intracellular antioxidant defenses. GGT could be informative in children with NAFLD because its expression is enhanced by oxidative stress and it could be released by several conditions inducing cellular stress and insulin resistance; both insulin resistance and oxidative stress are key components in the development of NAFLD. 42 The main strengths of our study are the inclusion of children with the whole spectrum of NAFLD from simple steatosis to cirrhosis along with the long-term follow-up of up to 20 years. The cases were well documented with all children having the diagnosis of NAFLD confirmed by radiological findings, and in almost half of them with liver histology. However, our study has some limitations. First, our patients were seen in a referral tertiary care medical center, and although the results may be extrapolated to other similar medical centers, the results most likely may not apply to children with NAFLD from the community. In this regard, larger community- or population-based studies are necessary to determine the prognosis of NAFLD in children from the general population. Second, most of our children (80%) were white, and thus, whether or not the long-term prognosis of pediatric NAFLD is any different among the different ethnic groups needs to be investigated. Finally, since liver biopsy is not part of the standard of care to confirm the diagnosis of NAFLD, only about a half of our children underwent liver biopsy, and thus, we were not able to determine the prognostic significance of the individual histological features. In summary, our study demonstrates that NAFLD in children is associated with a significantly shorter survival as compared to survival of the general population of same age and sex. NAFLD in children may progress to cirrhosis and end-stage liver disease with the consequent need for liver transplantation, but NAFLD with severe NASH may recur in the allograft. Further studies are needed to identify those children with NAFLD who are at a higher risk for disease progression who would be expected to benefit the most from medical therapy.

This was a retrospective longitudinal hospital-based cohort study. The study was approved by the Mayo Institutional Review Board and all patients or responsible guardian gave written informed consent for participation in medical research. Pediatric patients with NAFLD were identified using our Mayo computerized master diagnosis index which is a database of medical records of every single patient seen at Mayo Clinic. Each unit medical record contains all inpatient and outpatient medical information for each individual patient seen at Mayo Clinic since 1907. This has led to the creation of a unified medical index system, the Rochester Epidemiology Project (REP) by which the details of the medical care provided to Mayo patients can be studied. 23 The REP Mayo computerized master diagnosis indexes all medical diagnoses made at each health encounter by health care providers at Mayo. 23 All diagnoses made in outpatient office or clinic visits, hospitalizations, emergency room visits, nursing home care, surgical procedures, autopsies, and death certificates are recorded in the database. Thus, the REP Mayo diagnosis index makes it possible to identify a group of patients with certain characteristics and follow them longitudinally assessing long-term outcomes such as mortality and causes of death. Patients with a diagnosis of NAFLD were identified by searching the REP master diagnostic index using Hospital Adaptation of International Classification of Diseases (HICDA) codes for fatty liver, hepatic steatosis or steatohepatitis (5710-42-42, 5710-43-1, 5710-43-0, 5470-42-0 to 4, 2790-44-1). For the purpose of this study we used the definition of child as an individual under the age of 21 years as proposed by the National Institute of Health ( http://grants.nih.gov/grants/funding/children/children.htm ). Patients had their first medical evaluation for their liver disease at our institution during a 15-year period from 1/1/1985 to 12/31/1999. The date 1/1/1985 was chosen since the first case of pediatric NAFLD was reported in the mid 1980's; 20 the date 12/31/1999 was chosen to have a 15-year ascertainment period and a follow-up of more than 5 years for the last patient enrolled. The diagnosis of NAFLD required 1) confirmation of diffuse fatty infiltration of the liver in imaging studies regardless of aminotransferases levels; 2) average daily ethanol consumption of less than 10 grams; and 3) appropriate exclusion of other liver diseases based on standard clinical, laboratory, imaging and/or liver biopsy features. Laboratory tests to rule out other liver diseases included viral hepatitis panel (for hepatitis A, B and C performed either at the time of first evaluation or during the follow up), ceruloplasmin levels, alpha-1-antitrypsin levels and phenotype, autoantibodies (nuclear antibody [ANA], smooth muscle antibody [SMA], antibody to the liver/kidney microsome type 1, and antimitochondrial antibody), and standard metabolic/inborn error panel (lactate/piruvate ratio, urine and serum organic acids and aminoacids). The REP master diagnostic index identified a total of 130 cases. After an extensive review of the medical records of these patients, a total of 66 children with unequivocal NAFLD as defined by our diagnostic criteria detailed above were identified. A complete medical history and physical examination, and a complete laboratory evaluation were performed in all patients at the time of first medical evaluation in our institution and repeated at regular intervals thereafter. Laboratory evaluation included liver biochemistries (serum aspartate aminotransferase [AST], alanine aminotransferase [ALT], alkaline phosphatase activity, γ-glutamyl transferase [GGT], total bilirubin, albumin levels, and prothrombin time), fasting blood glucose, fasting lipid profiles (triglyceride, total cholesterol, HDL-cholesterol, and LDL-cholesterol levels), and specific laboratory tests to ruled other liver diseases as described before. All patients underwent abdominal imaging with ultrasonography, CT scan, and/or magnetic resonance imaging confirming the presence of fatty infiltration of the liver. The body mass index (BMI) based on body weight (kilograms) divided by the square of height (meters) was calculated in every case. BMI percentile was determined according to age and sex based on data from the Center for Disease Control and Prevention. 24 Obesity was defined by a BMI >95 th percentile for age and sex. 24 Abnormalities in the fasting levels of triglycerides and HDL-cholesterol were adjusted according to age, sex, and race or ethnic group (>95 th percentile for triglycerides; <5 th percentile for HDL-cholesterol) as recommended. 25 Diabetes mellitus was diagnosed based on standard criteria as recommended by the American Diabetes Association. 26 Hypertension was defined as a systolic or diastolic value that exceeded the 95 th percentile for age, sex and height. 27 Hypercholesterolemia was defined as a fasting total cholesterol level ≥ 200 mg/dL. 28 High LDL-cholesterol was defined by a LDL-cholesterol level ≥ 130 mg/dL. 28 In addition, patients were classified as having the metabolic syndrome if they met three or more of the following five criteria for age and sex as proposed: 29 BMI above the 97 th percentile (which correspond to a z -score of 2.0 or more); triglyceride level above the 95 th percentile; HDL cholesterol level below the 5 th percentile; systolic or diastolic blood pressure above the 95 th percentile; and impaired glucose tolerance. As oral glucose tolerance test was not performed, we used a fasting glucose value of at least 100 mg/dl to replace impaired glucose tolerance as recently proposed by the International Diabetes Federation. 30 A baseline liver biopsy was performed in 29 patients at time of diagnosis and follow-up liver biopsies in 5 of these patients. Since there are no established guidelines of when to perform a liver biopsy in patients with NAFLD, the decision to perform a baseline liver biopsy in our patient population was made on an individual basis by the treating Gastroenterologist, and in most (82%) cases was performed due to persistently abnormal liver enzymes. Liver biopsy features including grade of steatosis, inflammatory infiltrate, and ballooning, presence of Mallory hyaline, and stage of fibrosis were graded according to the scoring system proposed by Kleiner et al. 31 Continuous variables are presented as mean ± standard deviation (SD), and discrete variables are expressed as the number (percentage) of patients with a condition. Comparisons between patients with and without liver biopsy were performed with 2-sample t-tests for continuous variables and chi-square tests for categorical variables. Survival curves were created using the Kaplan-Meier method. The starting point for survival analysis was date of diagnosis of NAFLD. Patient follow-up was extended up to April 2008. The end-points for survival analysis were death or liver transplantation. For survival comparison we calculated the expected number of deaths for a cohort with the same age and sex distribution and the same amount of observation time (exposure to death) as the 66 children with NAFLD. The estimates were made using mortality data for United States from the U.S. Center for Health Statistics as previously detailed. 32 We used the relationship between the log-rank test and the Poisson distribution. The p-value calculated (from the one sample logrank test) depends on the assumption that the number of deaths follows a Poisson distribution with an expected value equal to the expected number of deaths. 32 The standardized mortality ratio (SMR) was calculated using the Ederer method based on age and sex to derive the expected number of events. 33