Association Between Transient Elastography and Controlled Attenuated Parameter and Liver Ultrasound in Children With Cystic Fibrosis

Study Population

CFLD-NET is investigating predictors of the development of a nodular US pattern in CF in a prospective study, PUSH, whose detailed methodology has been previously reported.⁽¹⁶⁾ Briefly, between January 12, 2010, and October 17, 2013, 725 participants underwent screening US at the beginning of the PUSH study. Children 3-12 years of age were eligible for the study. Liver US studies were performed after participants had fasted for at least 4 hours and were timed to avoid coinciding with an acute respiratory exacerbation. US findings were classified by liver parenchymal patterns as normal (NL), heterogeneous (HTG), homogeneously hyperechoic (HMG), or nodular (NOD). Classification was based on consensus of three specially trained study radiologists blinded to the readings by the other radiologists, VCTE measurements, clinical information, and demographic data. Among the 725 screened participants, 245 eligible patients entered the 9-year prospective longitudinal study. A recent interim analysis that concluded the need for continuation of the study also noted the benefit of including VCTE in the study’s goals.⁽¹⁷⁾

Between March 8, 2017, and December 31, 2018, the ELASTIC study enrolled participants from the PUSH cohort at eight of the 11 CFLD-NET centers where FibroScan was available. All participants who fulfilled inclusion criteria for PUSH at the eight participating centers were eligible for enrollment. Local internal review board approval was obtained at each participating center.

FibroScan Examination

FibroScan was completed at the ELASTIC baseline visit, with two additional annual examinations to be completed at future visits. Comprehensive clinical data and biosamples were collected annually, and research gray-scale US was performed every 2 years according to the PUSH protocol. FibroScan measurements were completed using a standardized protocol at each of these visits by a designated investigator(s) who had undergone appropriate training by Echosens to ensure consistent and standardized acquisition of complete data. The thoracic perimeter of the patient was measured and recorded to determine the size of the probe to be used per standard protocol as follows: >75-cm M-probe, 45-cm to ≤75-cm S2 probe, and ≤45-cm S1 probe.

The results of LSM (in kilopascals) were reported as the median of 10 valid measurements and were considered valid if the ratio of the interquartile range (IQR) and median was ≤0.3. If no valid LSM measurements were obtained at the first visit, an additional visit was scheduled to allow an additional attempt. To account for the described patchy nature of CFLD,⁽⁸⁾ a secondary site at least one intercostal space away was identified, and the same protocol repeated. If no valid LSM measurements were obtained at the secondary site, this was recorded but a second visit was not scheduled. For participants in whom the M-probe was used, CAP (in decibel-milliwatts [dB/m]) was also recorded using the same 10 valid measurements used for LSM. CAP can only be determined with the M probe, so only those participants for whom the M probe was appropriate had determination of CAP. Four CAP measurements below the detection limit (100 dB/m) were excluded, as this was felt to be a poor study with inconclusive findings, as recommended by the manufacturer. In addition to LSM, CAP and IQR, the number of invalid measurements and success rate to obtain a valid scan were recorded.

ELASTIC Baseline Data

At enrollment into ELASTIC, demographics were collected in addition to most-recent US grade and Pseudomonas status. Conventional biomarkers of liver disease including platelet count, AST, ALT, and GGT within 1 year of the FibroScan were recorded. AST-to–platelet ratio index (APRI; AST/upper limit of normal of AST/platelet count), Fibrosis-4 index (FIB-4; age × AST)/(platelet count × √ALT), and GGT-to-platelet ratio (GPR) were calculated. Additionally, spleen size was measured on the participant’s most-recent US as the longest craniocaudal dimension and expressed as a spleen-size-for-age z-score (SSAZ) using age-specific normal ranges.⁽¹⁸⁾ Reference ranges are given in Supporting Table S1.

Statistical Analysis

To evaluate the feasibility of VCTE in children and young adults with CF, we calculated the proportion of participants with valid LSM measurements at the primary liver location on first attempt or with multiple attempts among all participants for whom FibroScan was attempted with 95% confidence intervals (CIs).

To examine the variability of LSM and CAP measurements taken at two different liver locations in the same patient, we calculated the intraclass correlation coefficient (ICC) to assess similarity between these two measurements. Because LSM is severely skewed, to avoid the influence of extreme leverage points and inflated ICC values, LSM was log-transformed before calculation. ICC values less than 0.5, between 0.5 and 0.75, between 0.75 and 0.9, and greater than 0.90 are indicative of poor, moderate, good, and excellent reliability, respectively. A recent meta-analysis of LSM by VCTE suggested a cutoff of 5.95 kPa to prompt an investigation for liver disease among patients with CF.⁹ Using this threshold, we examined the consistency of LSM from the two liver locations using a 2 × 2 table with McNemar test and kappa statistics.

To assess whether LSM and CAP are associated with US pattern, we used the analysis of variance test to compare the log-transformed LSM and CAP values at the primary liver location among all US groups. Because US was performed on a different day than the day of FibroScan, the US grade closest in time to the baseline FibroScan visit was used. Post hoc tests were conducted to compare pair-wise group difference.

We then assessed whether LSM was associated with conventional biomarkers of liver disease–severity (ALT, GGT, GPR, APRI, FIB-4, platelet count, and SSAZ) graphically penalized splines. Using the LSM values from the primary liver location, we calculated Spearman correlation coefficients and their CI between LSMs with each of these conventional biomarkers. Statistical significance was defined as P < 0.05.

Study Population and Baseline Characteristics

A total of 141 PUSH participants were enrolled into ELASTIC between March 2017 and December 2019 (Fig. 1). Baseline VCTE was attempted in 137 participants, of whom 5 were less than 8 years of age. The demographic and laboratory information for 133 participants with valid LSM measurements is given in Table 1. Within a year before VCTE, 30.8% participants had a Pseudomonas-positive sputum culture. At the US visit closest (mean: 7.8 months before VCTE scan; range: 31.6 months before to 9.6 months after) to the ELASTIC baseline visit (before or after), most of the participants (54.1%) had a NL US pattern, and 13.5%, 12.8%, and 19.5% had HTG, HMG, and NOD patterns, respectively. The median (IQR) platelet count, GPR, and FIB-4 of ELASTIC participants was 281.5 10³/mm³ (232.0, 347.0), 0.06 (0.04, 0.13), and 0.25 (0.19, 0.33), respectively, and within normal limits. The median values for GGT (17 U/L [IQR: 13.0, 40.0]), APRI (0.4 [IQR: 0.27, 0.71]), and US SSAZ (0.9 [IQR: -0.11, 2.33]) were within normal ranges, but mean values for GGT (30.8 U/L), APRI (0.6), and US SSAZ (1.2) trended toward abnormally high values, suggesting skewedness of the data.

Feasibility of FibroScan and Reproducibility at Two Liver Locations

Among the 137 participants for whom FibroScan was attempted, 120 (87.6%; 95% CI: 80.9, 92.6) had valid scans at the first probe location on first attempt. The rate of a valid scan on the first attempt was 80% (95% CI: 28.4%, 99.5%) for participants less than 8 years of age (n = 5) and 87.9% (95% CI: 81.1%, 92.9%) for 8 years old and older (n = 132). Among these 120 cases, all but one also had a valid scan at the second probe location. Among the 17 participants who failed the initial scan, 14 underwent repeat scan(s) with 12 successful studies at the second attempt; 1 participant required a third scan to obtain a valid measurement, while the final subject did not have a successful measurement even with an additional attempt. The scan success rate at the first probe location was 97.1% (95% CI: 92.7%, 99.2%) overall. The final success rate at both probe sites was 96.4%.

In total there were 19 failed scans in 17 patients (Supporting Table S2). Incorrect probe size (as determined by central review by the study’s data coordinating center) was the most commonly reported reason (n = 10), followed by calibration issues (n = 5), uncooperative (n = 2), and IQR/median ratio >0.3 (11%). Each participant with a failed scan was invited to return to their local center and have their scan repeated with the appropriate probe size. Only the scan with the correct probe size was included in the data set for analysis. Given the relatively small sample size of participants younger than 8 years of age (n = 5), we did not perform any statistical test on the difference in feasibility between younger and older participants with CF in this cohort.

The median (IQR) LSM at the first probe location was 5.9 kPa (4.6, 8.0), and ranged from 2.7 to 75 kPa. There were 83 participants for whom the M probe was appropriate and thus were eligible to have CAP determined. Four (3 HTG and 1 NL) were excluded due to CAP below detection limit (100 dB/m), as recommended by the manufacturer. The 79 available CAP measurements demonstrated a mean (SD) of 228.6 dB/m (55.5) and median (IQR) of 225 dB/m (183, 262), and ranged from 122 to 371. The only difference between those with and without CAP determination was that those with CAP determination were older compared to those without CAP (Supporting Table S3).

The ICC for LSM (log-transformed) and CAP between the first and second probe locations was excellent: 0.955 (95% CI: 0.937, 0.968) and 0.912 (95% CI: 0.862, 0.944), respectively (Fig. 1). However, further review showed that 39.4% of participants had a > 1 kPa difference between LSM measurements at the two locations. This percentage was 61.5% among the 26 subjects categorized as having a NOD pattern. Overall, 8 (6%) of subjects had a > 3 kPa difference between the two probe locations (Table 2). Among NOD subjects, 6 (23%) had a difference > 3 kPa. Using the 5.95-kPa threshold that has been suggested as a cutoff to prompt evaluation for liver disease in CF,⁽⁹⁾ 16.7% of participants would have had discordant LSM values (1 value ≥ 5.95 and 1 < 5.95 kPa) between the two probe locations. The marginal distribution of this dichotomized version was not significantly different between the two locations (P = 0.09), and the corresponding kappa statistic was 0.67 (0.54, 0.79) (Supporting Table S4). The clinical impact is that performance of LSM at a single location would have led to evaluation for liver disease in 48% versus 55% of participants if two probe locations were used, if using the proposed cutoff value of 5.95 kPa (Supporting Table S4).

Association Between LSM and US Consensus Grade

LSMs differed significantly among US groups (Fig. 2A). The median (IQR) LSM was 5.1 kPa (4.2, 7.0), 6.1 kPa (4.7, 7.0), and 9.1 kPa (6.3, 15.8) for NL, HTG, and NOD participants, respectively. NOD participants had a significantly higher median LSM level than both NL (P < 0.0001) and HTG (P = 0.0013) participants. Among subjects with HMG US patterns, the median (IQR) LSM was 5.9 kPa (5.2, 7.8), which was also significantly lower than the NOD group (P = 0.0008). The highest LSM of 75 kPa (upper limit of FibroScan LSM) was observed in a NOD participant. There was not a significant difference in the data if only the participants with an US within 1 year of the VCTE are used (versus all) (Supporting Table S4).

Association Between CAP and US Grade

Among the 79 patients with valid CAP measurements, CAP level differed significantly among US groups (Fig. 2B). The HMG US group had the highest mean (SD) CAP of 271 dB/m (61), which was significantly higher than all the other three groups; the mean (SD) CAP was 219 dB/m (47) for NL (P = 0.0027), 218 dB/m (60.7) for HTG (P = 0.044), and 223 dB/m (56) for NOD (P = 0.041). Of note, there was a weak and insignificant association between CAP and body mass index z-score (r_s = 0.18; P = 0.1545). Similar to LSM, there was not a significant difference between values obtained at the two locations (P = 0.41), and the corresponding kappa statistic was 0.66 (0.49, 0.83), as shown in Supporting Table S5.

Association Between LSM and Lab Values/US Findings

LSM was significantly associated with conventional laboratory biomarkers of liver disease (platelet count, APRI, FIB-4, GGT, and GPR) and US spleen size z-score (Fig. 3). Increased LSM was associated with lower platelet count (Spearman correlation coefficient r_s = 0.28; P = 0.0071) and a higher APRI (r_s = 0.38; P = 0.0002), FIB-4 (r_s = 0.36; P = 0.0007), GGT (r_s = 0.35; P = 0.0017), GPR (r_s = 0.35; P = 0.003), and US spleen size z-score (r_s = 0.27; P = 0.0073). Laboratory biomarkers among subjects with a NOD US pattern accounted for most of the values seen at the extremes of the respective graphs and matches the association between these markers and US grade as previously reported by our group.⁽¹⁹⁾ We have included details of LSM and those participants with clinically detected portal hypertension in Supporting Table S6.