Hepatic Steatosis Is Highly Prevalent Across the Paediatric Age Spectrum, Including in Pre-school Age Children

J. Dhaliwal; G. B. Chavhan; E. Lurz; A. Shalabi; N. Yuen; B. Williams; I. Martincevic; A. Amirabadi; P. W. Wales; W. Lee; S. C. Ling; M. Mouzaki


Aliment Pharmacol Ther. 2018;48(5):556-563. 

In This Article


This was a retrospective cohort study at The Hospital for Sick Children (SickKids) in Toronto. We included previously healthy children, ages 1–17 years of age, who had undergone an abdominal CT scan as a part of routine trauma assessment from 2004 to 2015. Children were excluded if they had an associated spleen and/or liver injury, known underlying metabolic or other disorder known to contribute to hepatic steatosis (eg mitochondrial disorder), concurrent medication use that could predispose to hepatic steatosis (eg corticosteroids, methotrexate) or abdominal CT scans that were not of adequate quality to assess for hepatic steatosis. The local Ethical Review Boards had approved the study prior to data collection.

Data Collection

All data were collected from the SickKids Trauma Database. The Hospital for Sick Children is a level 1 paediatric trauma centre and receives referrals from across Ontario. A trauma data analyst populates the SickKids Trauma Registry prospectively, with each admission. These data are subsequently submitted to the Ministry of Health.

Anthropometrics and baseline demographics were collected. The data included patient age, gender, weight z-score, length/height z-score (CDC z-score), medication use at the time of CT scan and past medical or surgical history. Serum transaminase levels obtained on the day of the CT scan were also collected. Elevated alanine aminotransferase (ALT) was defined as >26 U/L in males and >22 U/L in females.[16]

Abdominal Computer Tomography Assessment

Abdominal contrast CT scans were used for non-invasive assessments of hepatic steatosis and measurements of anthropometry. Volume data were acquired using helical scanning and images were reconstructed into 5 mm slice thickness for viewing. Attenuation (density) of liver and spleen, and area (%) of abdominal subcutaneous tissue and visceral adipose tissue were measured on axial images. Liver parenchymal density in Hounsfield units (HU) was measured using region of interests (ROI) of approximately 200 mm2 in both the right and left hepatic lobes, avoiding major vessels. Similarly, splenic attenuation was measured with ROI of approximately 200 mm2 (Figure 1A). Hepatic steatosis was defined as a difference between liver and spleen attenuation index of less than −25 HU.[17] Subcutaneous adipose tissue was measured as distance from skin to the anterior abdominal wall musculature, at mid-clavicular line at the level of the umbilicus.[15,17] An average of two measurements was recorded and used in the analyses (Figure 1B). A single reviewer blinded to participant characteristics performed all the radiological measurements including hepatic attenuation, splenic attenuation and subcutaneous adipose tissue. An independent radiologist analysed a subset (8%) of the same set of CT scans to assess inter-observer reliability for these measurements. The inter-observer reliability agreement between the reviewers, with respect to the difference in hepatic and spleen attenuation (HU) and subcutaneous fat measurements (mm), was excellent, with intra-class correlation (ICC) values of 0.92 (95% CI: 0.85–0.96) and 0.99 (95% CI: 0.98–0.99), respectively.

Figure 1.

A and B, CT image with measurement of liver and splenic parenchymal attenuation at the level of the umbilicus, with measurement of subcutaneous adipose tissue in the right and left mid-clavicular line

A software modification was performed in MATLAB programming language (Mathworks, 1 Apple Hill Drive, Natick, MA, USA) to quantify the total adipose tissue region based on the voxels having a value within −200 and −20 HU. A connected region-clustering algorithm allowed the separation of abdominal subcutaneous fat from visceral fat (Figure 2A). A user created mask (Figure 2B) was used to correct for mislabeled voxels within the HU region, as well as, the usage of erosion and dilation (skin, partial volume effects). One slice at the umbilicus level was selected for segmentation. The percentage of the total area occupied by abdominal subcutaneous adipose and visceral adipose tissue was measured on each slice. A different reviewer, who was blinded to participant characteristics, performed the measurements. We have predetermined excellent inter-observer reliability (ICC = 0.99) of this technique that will be published in a separate technical study.

Figure 2.

A and B, CT image at level of umbilicus with area of abdominal subcutaneous (red), visceral adipose tissue (blue) and visceral adipose tissue masked (orange)

Statistical Analyses

Normally distributed continuous variables were described as mean and standard deviation (±SD), and nonnormally distributed continuous variable as median and interquartile range (IQR). Categorical variables were expressed as frequency and proportions. Continuous variables were compared with analysis of variance [ANOVA], independent sample t test, or Mann-Whitney test, as appropriate. Categorical variables were compared with the Pearson Chi-square test. Inter-observer for difference in hepatic and splenic attenuation, and subcutaneous adipose tissue measurements in mm2 was assessed using interclass coefficient (ICC). Statistical significance was defined as two-tailed P < 0.05. All analyses were performed using IBM SPSS Statistics Version 24 (IBM Corp., Armonk, NY, USA).