Incidence of Nonalcoholic Fatty Liver Disease in Children: 2009–2018

We conducted a multicenter retrospective cohort study using electronic health record data from Kaiser Permanente Southern California (KPSC), which serves 4.8 million members through 12 medical centers with 300 satellite clinics. We identified members aged 5 to 18 years in each calendar year between January 1, 2009, and December 31, 2018. The Kaiser Permanente Institutional Review Board approved this study.

Patients diagnosed with NAFLD by a physician were identified by the International Classification of Diseases, Ninth Revision (571.8) and International Classification of Diseases, 10th Revision (K76.0 and K75.81) codes for NAFLD or NASH. The date of diagnosis was recorded. Patients were excluded if pregnant within 1 year of the diagnosis of NAFLD.

Data on patient demographics, comorbidities, laboratory, imaging, and vital signs were ascertained from the electronic health record and health plan administrative sources by using diagnoses and procedure codes. Values for alanine aminotransferase (ALT) and aspartate aminotransferase were collected at the time point closest to the date of diagnosis of NAFLD. Other data, including imaging and comorbidities, were captured for 12 months before and after the diagnosis. The method used by the treating physician to diagnose NAFLD was documented (laboratories, imaging, and histology). The North American Society for Pediatric Gastroenterology, Hepatology, and Nutrition guidelines recommend using ALT to screen children with obesity or overweight with additional risk factors for NAFLD beginning at age 9 years. Therefore, we recorded whether screening with ALT was performed at well-child visits for children aged 9 to 18 with overweight or obesity. The KPSC laboratory upper limit of normal (ULN) for ALT was 54 U/L. The screening result was considered positive at 2 different levels: one result was having an ALT level greater than the institutional ULN, and the other was an ALT level >30 U/L because this has been shown to be a clinically relevant cutoff for detecting NAFLD.^10,11  In patients with NAFLD, the completion of screening for diabetes (fasting glucose), dyslipidemia (lipid panel), and hypertension (blood pressure) was recorded.

We performed descriptive statistics. Continuous data were reported as mean and SD or median and range. Means of continuous data were compared by using the Student’s t test. Categorical data were reported in absolute values and percentages. Rates of screening for children with obesity or overweight were calculated, and rates of positive test results were evaluated for an ALT level >54 and an ALT level >30 U/L. A logistic regression model was developed to identify factors associated with children who were screened for NAFLD having additional workup for NAFLD after screening. A χ² test was used to test the association of demographic data between patients with NAFLD and the KPSC population. ALT was reported as both a continuous variable and a categorical variable by bins of multiples of the ULN. Analysis of variance was used to compare ALT distribution across the groups. The Cochran-Armitage test for trend was used to test the trend of NAFLD incidence rates over time. Nominal 2-sided P values <.05 were considered to be statistically significant. Statistical analyses were performed by using SAS statistical software version 9.4 (SAS Institute, Inc, Cary, NC).

We identified all members aged 5 to 18 years on an annual basis from 2009 to 2018. The year-by-year numbers are shown in Table 1 and represent a total of 7 884 844 patient-years.

From 2009 to 2018, there were 206 117 children aged 9 to 18 with obesity who had a well-child visit, and an ALT screening was performed in 54.0% (111 207 of 206 117). The rate of a screening test with positive results was 8.9% when using an ALT level >54 U/L (n = 9891) and 28.4% when using an ALT level >30 U/L (n = 31 547). There were 212 710 children aged 9 to 18 with overweight that had a well-child visit, and an ALT screening was performed in 24.0% (50 951 of 212 710). The rate of a screening test with positive results was 2.2% when using an ALT level >54 U/L (n = 1105) and 10.0% when using an ALT level >30 U/L (n = 5111). Among children aged 9 to 18 with overweight or obesity and an ALT level >30 U/L, 12.3% received further workup for NAFLD. Logistic regression was performed to assess what factors were associated with having further evaluation for NAFLD (Table 2). Among children ages 9 to 18 with overweight or obesity who were screened for NAFLD, older age, higher BMI, and higher ALT level were all associated with greater odds of having further evaluation for NAFLD. In addition, among children screened for NAFLD, female patients had twice the odds of having further evaluation for NAFLD (odds ratio: 2.057; 95% confidence interval [CI]: 1.880–2.251). In addition, the odds were 38% greater of having further evaluation for NAFLD in the years 2012–2018 than during the years 2009–2011.

There were 3608 children diagnosed with NAFLD from 2009 through 2018. As shown in Table 1, the incidence of NAFLD significantly increased over time (P < .0001). The incidence rate was 36.0 per 100 000 in 2009 and 58.2 per 100 000 in 2018.

Demographics and clinical parameters are shown in Table 3. The mean age at diagnosis was 14.1 ± 3.2 years; the distribution by age is shown in Fig 1. Patients with NAFLD were significantly more likely to be male than the general population of patients (59.9% vs 50.8%; P < .0001). The distribution of race and ethnicity was significantly different from the general population, most notable for a higher frequency of children of Hispanic ethnicity (78.3% vs 43.3%; P < .0001). The distribution of BMI classification was underweight (0.4%), normal weight (6.2%), overweight (8.7%), and obesity (84.8%). The distribution of ALT by multiples of the ULN is shown in Fig 2. Children with NAFLD had a mean ALT level of 99 ± 99 U/L. Among children with NAFLD, 24.8% had an ALT level that was less than the laboratory ULN (ie, ≤54 U/L) and, as shown in Table 3, were significantly older (14.2 ± 3.1 vs 13.8 ± 3.2 years; P < .001) and more likely to be female (53.1% vs 35.8%; P < .001). The distribution of race and ethnicity was also significantly different between these 2 groups.

Patients were mainly diagnosed with NAFLD by imaging (61.1%; 2205 of 3608). Among those with an imaging-based diagnosis, ultrasonography was most common at 79.2% (1747 of 2205), followed by computed tomography (CT) scan at 17.4% (385 of 2208) and MRI at 3.3% (73 of 2205). A diagnosis was made by liver histology in 1.4% of patients. In the remaining 37.5% (1352 of 2205) of patients, NAFLD was diagnosed by the presence of elevated liver chemistry in the appropriate clinical context, with the exclusion of other potential etiologies. The distribution of ALT was significantly (P < .0001) different by the modality used to diagnose NAFLD. Children diagnosed with NAFLD by liver histology had the highest ALT level, and children diagnosed with NAFLD by CT scan had the lowest ALT level (median [interquartile range]: histology 141 U/L [80–227], ultrasound 84 U/L [51–129], laboratories only 73 U/L [41–109], MRI 55 U/L [30–107], and CT scan 38 U/L [24–72]). The distribution of age, sex, and BMI z score did not differ by diagnostic modality.

Among children with NAFLD, screening rates for diabetes, dyslipidemia, and hypertension were 47.3%, 14.4%, and 100%, respectively. Among those screened for diabetes, 10.9% (186 of 1706) had glucose ≥126 mg/dL, and 67.2% of these were diagnosed with diabetes. The overall rate of diabetes was 3.5% in children with NAFLD (125 of 3608). Of those children with diabetes, 19% (24 of 125) were classified as type 1 diabetes, and 81% (101 of 125) were classified as type 2 diabetes by their treating pediatric endocrinologist. Among those who were screened for dyslipidemia, clinically actionable, elevated low-density lipoprotein cholesterol was present in 15.4% (79 of 512) of patients, and clinically actionable hypertriglyceridemia was present in 73.5% (381 of 518). Systolic blood pressure of 130 to 139 mm Hg was observed in 13.3% (481 of 3608) of patients and ≥140 mm Hg in 6.2% (224 of 3608). Diastolic blood pressure of ≥80 mm Hg was observed in 10.3% (373 of 3608). Medication to treat hypertension was prescribed for 7.2% (260 of 3608) of patients.

We evaluated the screening practices and annual incidence of NAFLD in children within a large health maintenance organization. The incidence of NAFLD increased significantly over the 10-year observation period. Despite this, a majority of children with NAFLD likely went undiagnosed. We performed screening in more than half of the children with obesity. However, most patients with a positive screening test result did not have further evaluation for NAFLD, perhaps because of the complexity of evaluating elevated ALT levels. Because of this, the approach to diagnosis was not standardized. In this study, liver imaging was the most common diagnostic modality used to confirm NAFLD. These children diagnosed with NAFLD in a community setting had high rates of associated comorbidities of diabetes, dyslipidemia, and hypertension.

In our study, the incidence rate of NAFLD in children increased by 62% from 36.0 per 100 000 in 2009 to 58.2 per 100 000 in 2018. Although authors of many studies have assessed the prevalence of NAFLD, to our knowledge, there have been no studies in which authors evaluate the annual incidence. We, therefore, cannot compare our findings of incidence directly to other studies. Although the absolute number of patients in our study is much larger than that of any other clinical series of pediatric NAFLD, the number of patients who would be expected on the basis of available prevalence data is higher. Prevalence estimates in children with NAFLD in the United States range from 4.5% to 9.6%.^4,12,13  On the basis of these rates, if all children were diagnosed with NAFLD, we would have expected 40 000 to 90 000 cases of NAFLD. In our study, we observed a higher rate for screening than previously reported; screening was done in more than half of the children with obesity and a quarter of the children with overweight. Although many children went unscreened, the number of children identified with an elevated ALT level >30 U/L (36 658) approached the predicted prevalence estimate range. Thus, although the incidence rate of NAFLD in children increased, we conclude that screening was underused, and it is likely that many children with NAFLD went undetected.

In addition to underscreening, a barrier to diagnosing NAFLD is the complexity of interpreting ALT values once obtained. There was a substantial gap between the number of positive screening test results and the number of children who received a diagnosis of NAFLD. This may be due to confusion in ALT interpretation based on differences between laboratory normal reference ranges and biologically based normal values for ALT. The median ALT ULN used for children nationally is 53 U/L, which is similar to what is used by KPSC (54 U/L). However, this is approximately twice the biological ULN for ALT, as determined by the SAFETY study: 26 U/L for boys and 22 U/L for girls.⁵  Notably, in the later portion of our study, the odds of a child with overweight or obesity and an elevated ALT level receiving a full workup and being diagnosed with NAFLD increased. Additionally, there may be certain patient characteristics that influence interpretation of ALT values, because children with an older age, higher BMI, and higher degree of ALT elevation were more likely to receive an NAFLD diagnosis. The interpretation of ALT screening results by primary care pediatricians and resulting diagnostic decision-making were a large determinant as to whether NAFLD was ultimately diagnosed.

The evaluation necessary to confirm a diagnosis of NAFLD in a child who has an elevated ALT level is controversial and not straightforward.¹⁴  The definitive diagnosis of NAFLD requires exclusion of other causes of steatosis by clinical history, laboratory studies, and histology. In clinical practice, the extent of the evaluation based on the degree and duration of an elevated ALT level and the decision of when to perform a liver biopsy is not standardized.^15,16  In the current study, a majority of children were diagnosed with NAFLD via abdominal ultrasound, and this is likely the most common method used to support a diagnosis of NAFLD at other institutions as well. However, ultrasound does not measure hepatic steatosis directly. The relationship between changes on ultrasound-derived images and liver fat is inferred, thus making it subjective and nonquantitative.¹⁷  How to balance these limitations of ultrasound with the limitations of more expensive or invasive diagnostic tools remains a work in progress for the field. The current data from a large, real-world integrated health system reveal that there is a gap between guidelines that is based largely on expert opinion and pragmatic decisions in patient care that needs to be addressed to better understand the true scope of this problem and inform optimal diagnostic and management strategies.

Screening of children with NAFLD for comorbidities was high for hypertension, moderate for type 2 diabetes, and low for dyslipidemia. The North American Society for Pediatric Gastroenterology, Hepatology, and Nutrition recommends that all children with NAFLD should be screened for type 2 diabetes, dyslipidemia, and hypertension.¹⁸  Nationally, pediatric gastroenterologists report that 80% screen for type 2 diabetes, 72% screen for dyslipidemia, and 56% screen for hypertension.¹⁹ In our study, the rate of screening for hypertension was much higher. This shows the value of universal blood pressure measurement within a large health care system. In previous studies, authors have reported that up to one-third of children with NAFLD have elevated blood pressure.^7,20  In the current study, the rate of overt hypertension requiring treatment with medication was 7.2%, which is much higher than the rate of 2.4% for the general pediatric population.²¹  Unlike hypertension, less than half of the children with NAFLD were screened for diabetes. In our cohort, if all children with NAFLD had been screened for diabetes, the prevalence of type 2 diabetes would have been close to 6%. This is similar to the rate in a national multicenter study of children with biopsy-proven NAFLD, which reported that the prevalence of type 2 diabetes was 6.5%.⁸  Although the rate of screening for dyslipidemia was low at 14.4%, the rate of detection of clinically actionable low-density lipoprotein cholesterol observed was once again similar to that observed in national multicenter data: 15.4% at KPSC and 14% in the NASH Clinical Research Network.⁹  The comorbidities of hypertension, type 2 diabetes, and dyslipidemia are an important component of the morbidity and mortality associated with NAFLD. These data reinforce the high prevalence of these conditions in children with NAFLD and support the importance of screening as recommended in clinical guidelines.

To our knowledge, this study is the first to evaluate the incidence of NAFLD in children. Our study was strengthened by evaluating the incidence annually for a decade. An additional strength was the use of the KPSC database. Using the database, we studied nearly 8 million patient-years that were broadly representative of the general population. The prevalence of patients of Hispanic ethnicity within KPSC was similar to that of southern California and the southwestern United States but higher than in other areas of the nation. Therefore, the rates of NAFLD observed may differ elsewhere. The database also allowed us to carefully document comorbidities. The study was limited by the realities of clinical practice. We were able to document children who were diagnosed with NAFLD but observed that many children may have remained undiagnosed. We further observed that the reference standard diagnosis of NAFLD to include liver histology was infrequently used; thus, some diagnoses may have been inaccurate, and we are not able to comment on what fraction of the population with NAFLD had NASH and/or fibrosis.

The number of children diagnosed with NAFLD each year increased over a decade, and the annual rate rose faster in the more recent years. Health care systems and physicians will need to prepare for these growing numbers of children with chronic liver disease. The identification of all children with NAFLD would require improvement in screening practices as well as the evaluation of elevated ALT levels uncovered by screening. Notably, children diagnosed with NAFLD in a large, community-based health maintenance organization setting have similar rates of important comorbidities, as previously reported in tertiary care centers. Physicians caring for children with NAFLD should be prepared to identify and address type 2 diabetes, dyslipidemia, and hypertension. Optimal recognition and management of NAFLD and its associated comorbidities will require systematic coordination between primary care pediatrics and subspecialty pediatrics, including pediatric gastroenterology.

ALT

alanine aminotransferase

CI

confidence interval

CT

computed tomography

KPSC

Kaiser Permanente Southern California

NAFLD

nonalcoholic fatty liver disease

NASH

nonalcoholic steatohepatitis

ULN

upper limit of normal