In the last decade, transient elastography (TE) has significantly enhanced clinical monitoring of adults with chronic liver disease, changing the role of and need for liver biopsy. This noninvasive technique, which measures liver stiffness, is particularly useful in differentiating advanced fibrosis on liver biopsy from no or minimal fibrosis in adults.(1) TE is not as well studied in pediatrics, although experience is growing. A meta‐analysis of transient and shear wave elastography reports in children with chronic liver disease published before 2017 revealed sensitivity 90%, specificity 79%, and receiver operating characteristic 0.92 for varying definitions of portal hypertension.(2) One of these studies assessed 249 children with cystic fibrosis, while the pediatric studies in biliary atresia (BA) were single center, including up to 73 children.(3, 4) However, there is a pressing need for high‐quality multicenter data for TE in children of varying ages. Specific attention should be accorded to the wide spectrum of liver conditions seen in childhood, which are remarkably different from those seen in adulthood.
While the most common liver conditions in adults are fatty liver disease and chronic hepatitis C infection, infants and children often suffer from a range of congenital cholestatic disorders, including BA, alpha‐1 antitrypsin deficiency (A1ATD), and Alagille syndrome (ALGS). BA occurs in one in 8,000 to 18,000 live births and is manifest by uniquely rapid progression to advanced fibrosis and subsequent cirrhosis in the first months to years of life.(5) BA is also the most common indication for liver transplantation during early childhood. Although most individuals with A1ATD are asymptomatic, 15% may present with neonatal cholestasis or cirrhosis in childhood or adulthood.(6, 7) ALGS is an autosomal‐dominant multisystem disorder with paucity of bile ducts and progressive liver disease; 40% to 75% or more require liver transplantation before adulthood.(8, 9) Noninvasive measurements, such as TE, are an unmet need to assess the progression of liver disease in each of these distinct disorders.
Hepatic fibrosis, cirrhosis, and portal hypertension are common final pathways of a number of pediatric liver diseases. Portal hypertension may occur in the first months of life in BA and presents during early childhood in patients with ALGS and A1ATD. While splenomegaly on physical examination and thrombocytopenia are crude measures of portal hypertension, more precise data would be helpful in assessing the progression of pediatric liver disease over time and in predicting the risk of complications of portal hypertension, including variceal hemorrhage and ascites.(10)
The National Institute of Diabetes and Digestive and Kidney Diseases/National Institutes of Health‐sponsored Childhood Liver Disease Research Network (ChiLDReN) is a multicenter consortium that longitudinally follows infants and children with cholestatic liver diseases, including BA, A1ATD, ALGS, progressive familial intrahepatic cholestasis, bile acid synthetic defects, and mitochondrial disorders. Clinical history, physical examination, and laboratory findings are collected annually on participants in ChiLDReN research protocols. Since 2016, the FibroScan in Pediatric Cholestatic Liver Disease (FORCE) study has investigated the use of TE to assess liver stiffness in children with BA, A1ATD, and ALGS. This study provides a unique opportunity to assess liver stiffness in the context of clinical and laboratory markers of liver disease and portal hypertension in a prospective, longitudinal, multicenter approach in both a variety of cholestatic diseases and in a wide range of pediatric age groups.
Participants and Methods
Children and young adults 21 years of age or younger were approached and enrolled in FORCE if they had a diagnosis of BA and were an active participant in one of two longitudinal observational studies (Prospective Database of Infants with Cholestasis [PROBE; NCT00061828] or Biliary Atresia Study in Infants and Children [BASIC; NCT00345553]), or had ALGS or A1ATD and were an active participant in the Longitudinal Observational Study of Genetic Causes of Intrahepatic Cholestasis (LOGIC; NCT00571272). Children with known polysplenia/asplenia, situs inversus, clinically significant ascites, an implantable active medical device (such as a pacemaker or defibrillator), an open wound near the FibroScan site, current pregnancy, or who had undergone liver transplantation were not eligible for FORCE.
Written informed consent was obtained from caregivers or the participant, and assent was obtained from the child when appropriate according to local Institutional Review Board (IRB) rules. This study was approved by local IRBs and complied with the Declaration of Helsinki and Good Clinical Practice Guidelines.
Liver stiffness (reported in kPa) was measured by vibration‐controlled TE in nonfasted and nonsedated participants, using FibroScan according to the manufacturer’s instructions (Echosens, Waltham, MA). The time since last food or nonclear liquid intake was recorded. A valid scan included at least 10 valid measurements using the appropriate probe (S or M) and examination type (S1 or S2) (according to the manufacturer’s instructions) with an interquartile kPa range (IQR)/median value of <30%. The XL probe was not used for FORCE, and as such, participants with a skin to capsule distance >2.5 cm were excluded from continued participation in the study. Spleen excursion below the left costal margin was assessed and measured by physical examination at the time of scanning. Laboratory studies were obtained as part of routine clinical care. Thrombocytopenia was defined as a platelet count <150,000 μL. Aspartate aminotransferase (AST) to platelet ratio index (APRI) and gamma‐glutamyl transpeptidase (GGT) to platelet ratio were calculated.(11) Clinically evident portal hypertension (CEPH) was categorized as definite (dCEPH), possible (pCEPH), or absent (aCEPH), using a described research definition.(10)
Sample Size Calculation
Enrollment in FORCE was powered to achieve two specific aims. The first aim was to detect a significant difference in liver stiffness in children with BA and CEPH versus those with aCEPH. It was estimated that 49% of participants with BA would have dCEPH while 34% would have aCEPH. Data from Chongrisawat et al.(4) were used to estimate liver stiffness measurements (LSMs) for dCEPH and aCEPH; BA with splenomegaly was equated with dCEPH, while BA without splenomegaly was equated with aCEPH. A sample size of 192 participants with BA was estimated to have >99% power to detect a 12‐kPa difference and >80% power to detect a 6.2‐kPa difference in LSMs between dCEPH and aCEPH. An additional consideration for sample size determination was based on future investigations directed at detecting an increase between 4 and 5.3 kPa over 2 years in children with BA. Given potential attrition over time, a decision was made to enroll 250 participants with BA who had undergone valid baseline TE. The number of participants with A1ATD and ALGS was a convenience sample based on enrollees in LOGIC.
Summary statistics for demographics and conventional laboratory determinants of liver disease were reported for the three diseases. To evaluate the feasibility of performing TE in children with BA, ALGS, and A1ATD, we calculated the proportion and 95% confidence intervals (CIs) of participants with valid (as defined above) LSMs among all participants for whom FibroScan was attempted. Feasibility analyses were based on all enrolled participants. Subsequent analyses were performed only on participants with valid baseline LSMs.
The association of conventional laboratory determinants of liver disease at enrollment was assessed using Spearman correlation coefficients. We then used scatter plots to graphically explore the relationships between LSM and these laboratory values with penalized splines. Multiple linear regression models, including conventional laboratory tests, were further used to assess how much variance in LSM can be explained by these conventional laboratory tests. LSM, GGT, AST, and total bilirubin (TB) were modeled log transformed due to non‐normality of the distributions. All analyses were conducted in the three diseases separately.
Additional linear regression models using all combined data were used to study differences in associations between conventional laboratory determinants and LSM among the three different disease types. We studied these conventional laboratory determinants individually. In each linear regression model, the conventional laboratory determinant under investigation, disease types, and the interaction terms between disease types and the conventional laboratory determinant were included as covariates. The differences in associations between conventional laboratory determinants and LSM among disease types were assessed by testing these interaction terms.
We compared the distribution of LSMs at enrollment between participants with dCEPH and aCEPH in each of the three disease groups. Summary statistics (mean, SD, median, quartiles) and boxplots of LSMs were calculated to inspect the relationships between the distribution of LSMs of the CEPH status groups. The t test was employed for testing the difference in log‐transformed LSMs at enrollment between subjects with dCEPHT and those with aCEPHT, as described in our a priori‐developed statistical analysis plan. If the primary comparison described above was significant, we compared dCEPH versus pCEPH and pCEPH versus aCEPH using post‐hoc comparison in the analysis of variance (ANOVA) test. The intent of this analysis sequence was to reduce type I error. Furthermore, we compared conventional laboratory tests of liver disease among the three diseases by CEPH status using ANOVA and post‐hoc comparisons. We produced radar plots (median values of the parameters for each group along the spokes of the circles) to provide a visual representation of the distribution of LSMs and other clinical parameters by diagnosis and CEPH status. Values were scaled so that the center of each radar plot represents more favorable values for the clinical parameters observed among all the groups. The specific values used are shown in a table below the radar plots.
Between November 2016 and August 2019, 811 of 1,010 active participants with BA in PROBE or BASIC BA, and with ALGS and A1ATD in LOGIC were potentially eligible for FORCE (Fig. 1). A total of 550 participants consented to the study, 458 of whom had a valid baseline TE measurement (BA, 254; A1ATD, 104; ALGS, 100). Demographics of potentially eligible but not enrolled participants were generally similar to those enrolled in FORCE, although enrollment of participants less than 1 year of age was diminished (Supporting Table S1).
Feasibility of FibroScan
A total of 458 (83.3%; 95% CI, 80.2%‐86.4%) of the 550 participants who enrolled in FORCE had a valid TE measurement. Among the 92 participants without a valid scan (Supporting Table S2), 45 studies were invalid due to incorrect probe or examination type selection while 22 were invalid due to an IQR/median >30%. Behavioral issues accounted for 27% of the invalid studies, and these were more frequent in younger participants. The proportion of valid scans by age was 62.3% (95% CI, 49.2%‐75.3%), 80.5% (95% CI, 73.4%‐87.7%), 85.8% (95% CI, 80.6%‐91.0%), and 88.2% (95% CI, 83.7%‐92.6%) for <2, 2‐5, 5‐10, and 10+ years of age, respectively. All subsequent analyses were performed on participants with valid baseline LSMs.
The mean age of participants with a valid LSM was 8.8 (SD, 4.9) years, 48% were female participants, 67% were white, and 18% were Hispanic (Supporting Table S3). Biochemical testing as part of routine clinical practice was performed in >90% of participants (Table 1). The cohort overall had preserved hepatic synthetic function with only a small portion with serum albumin <3.0 g/dL; international normalized ratio (INR) was primarily normal, and pediatric end‐stage liver disease (PELD) scores were typically <0. TB, GGT, ALT, and AST were highest in ALGS participants, consistent with the more profound cholestasis in this group. There were greater numbers of participants with thrombocytopenia in the BA cohort. APRI distribution was similar in ALGS and BA, while it was less remarkable in A1ATD. GGT to platelet ratio was highest in ALGS.
|Characteristic||BA (n = 254)||A1ATD (n = 104)||ALGS (n = 100)||All diagnoses (n = 458)|
|Mean, mg/dL (SD)||1.0 (1.3)||0.6 (0.7)||3.0 (4.7)||1.3 (2.6)|
|Median, mg/dL (Q1, Q3)||0.6 (0.4, 1.0)||0.4 (0.3, 0.6)||1.1 (0.6, 3.0)||0.6 (0.4, 1.1)|
|Mean, mg/dL (SD)||140.5 (190.5)||65.5 (102.0)||451.0 (400.7)||189.6 (275.1)|
|Median, mg/dL (Q1, Q3)||74.0 (26.0, 172.0)||27.0 (17.0, 53.0)||326.5 (171.5, 609.0)||77.0 (26.0, 234.0)|
|Mean, mg/dL (SD)||78 (71)||60 (42)||155 (120)||91 (87)|
|Median, mg/dL (Q1, Q3)||51 (34, 92)||48 (31, 76)||111 (74, 189)||59 (36, 109)|
|Mean, mg/dL (SD)||81 (84)||73 (57)||174 (125)||99 (98)|
|Median, mg/dL (Q1, Q3)||51 (30, 94)||59 (36, 85)||143 (89, 248)||68 (36, 122)|
|Mean, g/dL (SD)||4.2 (0.5)||4.4 (0.4)||4.2 (0.5)||4.2 (0.5)|
|n (%) <3.0||4 (2%)||1 (1%)||3 (3%)||8 (2%)|
|n (%) <2.5||0 (0%)||0 (0%)||1 (1%)||1 (0%)|
|n (%) <2.0||0 (0%)||0 (0%)||0 (0%)||0 (0%)|
|Mean, g/dL (SD)||1.1 (0.1)||1.1 (0.1)||1.1 (0.1)||1.1 (0.1)|
|Mean, g/dL (SD)||2.6 (4.4)||0.9 (1.9)||4.6 (7.9)||2.6 (5.2)|
|Median, g/dL (Q1, Q3)||1.0 (0.3, 2.9)||0.2 (0.1, 0.5)||2.7 (1.3, 5.6)||0.9 (0.2, 3.0)|
|Platelet count, n||247||97||95||439|
|Mean (SD)||173 (104)||273 (102)||262 (100)||214 (113)|
|n (%) <150||118 (48%)||10 (10%)||12 (13%)||140 (32%)|
|n (%) <100||84 (34%)||6 (6%)||5 (5%)||95 (22%)|
|n (%) <50||18 (7%)||3 (3%)||1 (1%)||22 (5%)|
|Mean (SD)||1.9 (2.5)||0.8 (1.1)||2.2 (4.4)||1.7 (2.9)|
|Median (Q1, Q3)||0.9 (0.4, 2.3)||0.4 (0.3, 0.8)||1.2 (0.7, 2.6)||0.8 (0.4, 1.9)|
|n (%) >1.0||109 (46%)||18 (19%)||54 (57%)||181 (42%)|
|n (%) >1.5||86 (36%)||13 (13%)||35 (37%)||134 (31%)|
|n (%) >2.0||68 (29%)||8 (8%)||27 (28%)||103 (24%)|
|PELD score, n||193||75||83||351|
|Mean (SD)||−9.3 (5.7)||−12.1 (5.4)||−4.3 (8.4)||−8.7 (6.9)|
|Median (Q1, Q3)||−10.5 (−13.0, −5.7)||−13.1 (−15.8, −10.3)||−6.6 (−11.1, 0.8)||−10.4 (−13.4, −5.2)|
|n (%) >0||13 (7%)||3 (4%)||24 (29%)||40 (11%)|
|n (%) >10||1 (1%)||0 (0%)||6 (7%)||7 (2%)|
|n (%) >20||0 (0%)||0 (0%)||0 (0%)||0 (0%)|
Correlation of LSM with Biochemical Characteristics of Liver Disease
Each parameter was examined for its correlation with LSM in the context of the three diseases. In BA, there was a positive correlation between LSM and TB, INR, AST, ALT, GGT, GPR, PELD score, APRI, and spleen size and a negative correlation with albumin and platelet count (Fig. 2; Supporting Table S4). The negative relationship between platelet count and LSM in BA persisted even with platelet counts in the “normal” range (i.e., >150,000; P = 0.025) (Supporting Fig. S1). Similar correlations existed for A1ATD (except AST, ALT, and albumin) and for ALGS (except for INR, GGT, GPR, and ALT).
Association of LSM with CEPH
The relationship between LSM and portal hypertension was investigated using the recently developed research definition of CEPH.(10) dCEPH was more common in BA (48%) than in A1ATD (8%) or ALGS (15%) (Fig. 3; Supporting Tables S5 and S6). LSM was greater in dCEPH compared to aCEPH for all three diseases (dCEPH vs. aCEPH LSM kPa: mean (+SD) BA, 20.0 (13.6) vs. 10.8 (13.5), P < 0.001; mean A1ATD, 23.8 (12.5) vs. 7.8 (7.7), P < 0.001; mean ALGS, 16.1 (13.6) vs. 9.2 (5.3), P = 0.003). In BA LSM, pCEPH was lower than in dCEPH but not different from aCEPH. In contrast, in ALGS LSM, pCEPH was higher than aCEPH but not different from dCEPH. In A1ATD LSM, pCEPH was lower than dCEPH and higher than aCEPH. Among participants with the same CEPH status, the clinical, biochemical, and LSM varied by disease type. In aCEPH, LSM was higher in BA (Table 2). Participants with ALGS with aCEPH had higher TB, GGT, ALT, and AST levels. Participants with A1ATD with aCEPH had the lowest LSM, GGT, AST, and PELD score. These parameters were similar for BA and A1ATD with pCEPH (Supporting Table S7). Measurements of TB, GGT, AST, ALT, APRI, and PELD score were also higher in participants with ALGS with pCEPH. A complex picture emerged in comparing these parameters for dCEPH (Table 3). Spleen excursion below the costal margin was similar for all three diseases. LSM was higher and platelet count lower in BA and A1ATD compared to ALGS. However, TB, GGT, AST, ALT, and PELD score in ALGS were higher than in BA and A1ATD. The differences in parameters by disease type and CEPH status are summarized in the radar plot in Fig. 4. The number of hours of fasting did not impact the LSM among participants with the same disease type and CEPH status (Supporting Fig. S2; Supporting Table S8).
|Parameter||Mean (SD)||P Value for Differences in Means|
|BA (n = 71)||A1ATD (n = 86)||ALGS (n = 61)||Overall Difference||BA vs. A1ATD||BA vs. ALGS||A1ATD vs. ALGS|
|FibroScan LSM, kPa||10.8 (13.5)||7.8 (7.7)||9.2 (5.3)||0.02||0.01||0.95||0.02|
|TB, mg/dL||0.6 (0.6)||0.6 (0.7)||1.8 (2.5)||<0.001||0.66||<0.001||<0.001|
|GGT, mg/dL||149 (250)||57 (105)||418 (423)||<0.001||0.001||<0.001||<0.001|
|AST, mg/dL||66 (60)||59 (44)||142 (132)||<0.001||0.64||<0.001||<0.001|
|ALT, mg/dL||72 (89)||73 (59)||161 (140)||<0.001||0.21||<0.001||<0.001|
|Albumin, g/dL||4.3 (0.5)||4.4 (0.4)||4.3 (0.5)||0.33|
|INR, g/dL||1.1 (0.1)||1.1 (0.1)||1.0 (0.1)||0.05|
|GPR, g/dL||1.2 (2.2)||0.4 (0.9)||3.0 (3.1)||<0.001||<0.001||<0.001||<0.001|
|Platelet count||279 (90)||298 (84)||296 (85)||0.34|
|APRI||0.6 (0.6)||0.6 (0.6)||1.2 (1.0)||<0.001||0.33||<0.001||<0.001|
|PELD score||−12.1 (4.8)||−13.1 (4.8)||−7.0 (7.0)||<0.001||0.36||<0.001||<0.001|
|Spleen size (cm below costal margin)||0.1 (0.3)||0.1 (0.4)||0.0 (0.2)||0.49|
|Parameter||Mean (SD)||P Value for Differences in Means|
|BA (n = 123)||A1ATD (n = 8)||ALGS (n = 15)||Overall Difference||BA vs. A1ATD||BA vs. ALGS||A1ATD vs. ALGS|
|FibroScan LSM, kPa||20.0 (13.6)||23.8 (12.5)||16.1 (13.6)||0.14|
|TB, mg/dL||1.3 (1.5)||0.8 (0.4)||6.4 (8.6)||<0.001||0.59||<0.001||<0.001|
|GGT, mg/dL||150 (176)||149 (72)||458 (311)||<0.001||0.29||<0.001||0.049|
|AST, mg/dL||87 (76)||78 (33)||181 (122)||0.001||0.80|