C A P based on TRANSIENT ELASTOGRAPHY for HEPATIC STEATOSIS

Discussion

Hepatic steatosis, particularly because of NAFLD, is common and increasing in prevalence. In some patients, steatosis may progress to NASH, cirrhosis and end-stage liver disease.^{[1, 4]} In light of the growing burden of NAFLD and anticipated development of specific therapies, reliable noninvasive methods for grading steatosis are needed.^[9] We have demonstrated that the CAP is correlated with steatosis independent of inflammation and fibrosis, and can be used to noninvasively identify steatosis with good performance. Specifically, for significant steatosis (≥10% of affected hepatocytes), the AUROC of the CAP was 0.81; a CAP threshold of 283 dB/m was 76% sensitive and 79% specific for this outcome. Similar findings were reported in a study of 615 patients with HCV in whom the CAP had an AUROC of 0.80.^[27]A threshold of 222 dB/m was 76% sensitive and 71% specific in this cohort. However, our data are somewhat less optimistic than described by Sasso et al. in a study of 115 patients with various liver disorders.^[15] In this study, the AUROC for significant steatosis was 0.91; a CAP cut-off of 238 dB/m was 91% sensitive and 81% specific. These discrepancies may relate to differences in the study populations including disease aetiologies, the prevalence of obesity and extent of subcutaneous adiposity, and the severity of steatosis, which may influence CAP performance because of spectrum bias.^[10]For example, the mean BMI in our cohort was 32 kg/m² and 65% of patients had significant steatosis. In the other studies, mean BMI was 24–25 kg/m² and 31–58% had significant steatosis.^{[15, 27]} Future studies in larger cohorts, including ideally an individual patient data meta-analysis, will be useful for refining the operating characteristics of the CAP, including the optimal cut-offs in different disorders. Our data suggests that the performance of the CAP did not differ substantially between conditions.

We also examined the diagnostic performance of the CAP for quantifying steatosis according to the NAS classification.^[4] CAP performed well for identifying S1–S3 (≥5%) and S2–S3 (>33%) steatosis with AUROCs of 0.79 and 0.76 respectively. However, because the CAP was not significantly different between patients with S2 and S3 steatosis (Fig. 2), severe (>66%) steatosis was sub-optimally identified (AUROC 0.70). These findings were corroborated in our analysis evaluating the CAP's ability to discriminate individual steatosis grades, which overall, revealed reasonable performance (Obuchowski measure = 0.89). However, a problem with the CAP is that the optimal cut-offs identified for each outcome – based on the maximal sum of sensitivity and specificity – were similar (Table 2), which makes grading steatosis with this technology difficult. Similarly, although the CAP reliably differentiated steatosis at least 2 grades apart, the identification of single-grade differences was poor (Table 3). This limitation, which also applies to other surrogate markers of liver histology (e.g. serum fibrosis markers and FibroScan®) is in part because of the imprecision of the CAP, but also the limitations of liver biopsy. Indeed, in a study that evaluated sampling error of biopsy among 51 NAFLD patients who underwent dual pass biopsies, discordance in steatosis grading was observed in 22% of cases.^[7] In 18% of patients, the difference in steatosis severity exceeded 20%. It is conceivable that the CAP actually provides a more accurate assessment of steatosis within the entire liver as it samples a volume ~100-times larger than biopsy.^[15] Furthermore, the reliability and variability in the pathologic grading of steatosis, even by experts, is poor.^[28–30] Steatosis evaluation can also be influenced by tissue fixation and staining methods.^{[31, 32]} Therefore, although we included only high-quality biopsies and centralized histological grading by experts, we cannot exclude an 'imperfect gold standard bias'. Future CAP studies including more objective assessments of steatosis (e.g. computerized morphometry)[28] would be useful to investigate these issues.

Previous reports have shown higher rates of discordance in fibrosis staging using the FibroScan® in patients with highly variable LSMs.^{[22, 23]} Therefore, we examined the impact of CAP variability – as assessed by IQR/MCAP – on the diagnostic performance of this tool. Although we did not observe significant differences in the AUROCs of the CAP for significant (≥10%) steatosis between patients with high and low IQR/M, highly variable measurements (IQR/MCAP ≥15) were less accurate for ≥5% steatosis in a post hoc analysis. These findings require confirmation. The AUROC for significant steatosis was higher among patients with no to minimal fibrosis (AUROC 0.89 vs. 0.72 with moderate to severe fibrosis), suggesting a potential role for the CAP as a screening tool in the general population. The reason for this novel finding [15, 27] remains unclear because the prevalence of significant steatosis was similar between groups (60–70%). Moreover, the CAP was not associated with fibrosis after adjustment for steatosis and inflammation. Additional studies in larger cohorts are necessary for confirmation and to examine other predictors of CAP performance.

Our study suggests that the CAP may be a worthwhile adjunct in the evaluation of patients with chronic liver disease. As LSM by FibroScan® is routine in many regions, an appealing aspect of the CAP is that it is provided automatically and immediately by the FibroScan® VCTE™ software in the same region of interest as LSM. Moreover, CAP measurement is operator-independent and requires no specific training. Another potential application of CAP is the exclusion of steatosis in donors for living-related liver transplantation, in whom steatosis increases the risk of primary graft non-function.^[33] Evaluation of CAP in other clinical settings will be the focus of future investigation.

Other serum and imaging-based methods have been examined for the noninvasive assessment of steatosis. Although liver biochemistry is widely available and inexpensive, the sensitivity of these tests is sub-optimal. In our study, the AUROCs (95% CI) of ALT and GGT for significant steatosis were only 0.50 (0.40–0.60) and 0.56 (0.46–0.66) respectively (both P < 0.00005 vs. CAP). Several serum marker panels have also been proposed for quantifying steatosis. For example, the SteatoTest includes age, gender, BMI, cholesterol, triglycerides, glucose, ALT, GGT, bilirubin, haptoglobin, alpha-2-macroglobulin and apolipoprotein A1.^[34] For steatosis ≥5%, its AUROC was 0.80 in a cohort of 811 patients with various liver disorders. The proprietary nature of this algorithm and delayed results are important limitations. A recently described nonproprietary panel, the FLI, includes triglycerides, GGT, BMI and waist circumference.^[17] The FLI is associated with steatosis detected songraphically^[17] and is an independent predictor of mortality.^[18] Similarly, the HSI, which includes the ALT/AST ratio, BMI, gender and diabetes, is associated with the presence and severity of steatosis on ultrasound.^[19] In our study, the CAP outperformed these indices for the primary outcome and most secondary outcomes although the small sample size may have rendered some analyses underpowered. Ultrasound is the most common imaging method for detecting steatosis, which is recognized by a diffuse increase in hepatic echogenicity.^{[12, 13]} 

Limitations of ultrasound include its markedly reduced sensitivity for mild steatosis (under 30%),^{[35, 36]} operator and machine-dependence, the inability to reliably quantify hepatic fat content,^{[12, 13, 35]} and the potential for extensive fibrosis to increase liver echogenicity.^[13] Although promising, other abdominal imaging techniques (e.g. CT, MRI and proton magnetic resonance spectroscopy), are not widely available, expensive, lack standardization, have controversial diagnostic performance, and in the case of CT, exposure to ionizing radiation.^{[12, 13]}

Our study has several limitations. Most importantly, our study population was highly selected in that it included only patients with a BMI ≥28 kg/m²; moreover, a significant number of patients were excluded predominantly because of missing data. Therefore, the generalizability of our findings to other patient populations (e.g. a 'screening cohort' seen in primary care) requires confirmation. Second, our sample size was limited in part because of the difficulty of obtaining valid CAP measurements in obese patients using the FibroScan® M probe. Future studies are necessary to develop a CAP algorithm for the novel FibroScan® XL probe, which was designed for use in this population.^[16] Second, because of a median delay of approximately 1 month between CAP measurement and biopsy, we cannot exclude changes in steatosis that may have influenced our findings. Third, we assessed only the diagnostic accuracy of CAP although additional properties including agreement, precision, and responsiveness deserve mention. Finally, our cohort included patients with numerous liver diseases in whom inflammation and fibrosis may be staged using different scoring systems. As these classifications are not interchangeable, this issue may have influenced our multivariate analyses.

In conclusion, the CAP is a promising tool for the noninvasive detection of hepatic steatosis. Advantages of the CAP include its simplicity, operator-independence and sensitivity to lesser degrees of steatosis than are detectable using other widely available imaging modalities. Moreover, the CAP provides an immediate assessment of steatosis simultaneously with LSM used to stage hepatic fibrosis. Future studies are necessary to validate our findings in larger cohorts to define optimal CAP thresholds and to develop a CAP algorithm for the FibroScan® XL probe that will facilitate measurement in a greater proportion of obese patients.