A continuous 13C methacetin breath test for noninvasive assessment of intrahepatic inflammation and fibrosis in patients with chronic HCV infection and normal ALT

Abstract

Up to 30% of patients with hepatitis C virus (HCV) infection and normal serum alanine aminotransferase (NALT) have significant liver disease. Currently, many of these patients undergo a liver biopsy to guide therapeutic decisions. The BreathID continuous online (13)C-methacetin breath test (MBT) reflects hepatic microsomal function and correlates with hepatic fibrosis. To assess its role in identifying intrahepatic inflammation and fibrosis in NALT patients, we tested 100 patients with untreated chronic HCV infection, and 100 age- and sex-matched healthy volunteers using (13)C MBT following ingestion of 75 mg methacetin. All HCV patients had undergone a liver biopsy within 12 months of performing the MBT. Patients with a necroinflammatory grade <or=4 or >4, based on Ishak (modified HAI) score, HAIa + HAIb + HAIc + HAId, were defined as having low or high inflammation, respectively. Patients with a histological activity fibrosis stage <or=2 or >2, were defined as having nonsignificant or significant fibrosis, respectively. A proprietary algorithm to differentiate intrahepatic inflammation within chronic HCV patients with NALT achieved an area under the curve (AUC) of 0.90. Setting a threshold on the point of best agreement (at 83%) results in 82% sensitivity and 84% specificity. With application of another proprietary algorithm to differentiate patients with nonsignificant or significant fibrosis, 67% of liver biopsies performed in the patient group could have been avoided. This algorithm achieved an AUC of 0.92, with a sensitivity of 91% and a specificity of 88%. There was no correlation between body mass index (BMI) and MBT scores for patients with the same histological score. The continuous BreathID(13)C MBT is an accurate tool for measuring the degree of inflammation and fibrosis in patients with chronic HCV infection and NALT. As such, it may prove to be a powerful, noninvasive alternative to liver biopsy in the management of this patient population.

Figures

Fig. 1

A binary logistic regression model using BreathID and demographic parameters (P < 0.001 for each of the parameters) showed that the methacetin breath test (MBT) can differentiate patients and healthy volunteers with an area under the curve (AUC) of 0.7 (95% CI, 0.626–0.778), sensitivity of 56% and specificity of 86%. A binary logistic regression model using BreathID® and demographic parameters (P < 0.001 for each of the parameters) showed that the MBT can differentiate patients and healthy volunteers with an AUC of 0.67 (95% CI, 0.59–0.74), sensitivity of 56% and specificity of 86%. Using Charles E. Metz ROCKIT 1.1B2 provides the following results: binormal parameters and area under the estimated ROC: a, 0.5158; b, 0.5824; area (Az), 0.6721; area (Wilc), 0.6689. Estimated standard errors (SE) and correlation of these values: SE (a), 0.1256; SE (b), 0.0754; corr (a,b), 0.2129; SE (Az), 0.0384; SE (Wilc), 0.0380; symmetric 95% CI for a, (0.2697, 0.7619); b, (0.4347, 0.7301); asymmetric 95% CI for Az, (0.5938, 0.7435).

Fig. 2

Model to differentiate between HAIa + HAIb + HAIc + HAId ≤ 4 vs HAIA + HAIB + HAIC + HAID > 4: A proprietary algorithm that includes breath-test parameters, age and other patient data to differentiate intrahepatic inflammation (HAIa + HAIb + HAIc + HAId ≤ 4 vs > 4) within chronic hepatitis C virus (HCV) patients with normal alanine aminotransferase (NALT) achieved an area under the curve (AUC) of 0.90. Setting a threshold at the point of best agreement (at 83%), results in 82% sensitivity and 84% specificity. At the dataset’s prevalence of 68% the PPV is 92% and the NPV is 69%. Assuming a prevalence of 45.5% would lead to a PPV of 82% and an NPV of 85%. Using Charles E. Metz ROCKIT 1.1B2 provides the following results: binormal parameters and area under the estimated ROC: a, 1.9574; b, 1.0126; area (Az), 0.9155; area (Wilc), 0.9021. Estimated standard errors (SE) and correlation of these values: SE (a), 0.3453; SE (b), 0.2961; corr (a,b), 0.6238; SE (Az), 0.0305; SE (Wilc), 0.0297. Symmetric 95% confidence intervals: For a, (1.2807, 2.6342); for b, (0.4322, 1.5930); asymmetric 95% CI for Az, (0.8389, 0.9609).

Fig. 3

Receiver operating characteristic (ROC) curve describing performance of the 67% patients where significant/nonsignificant fibrosis was determined: Using Charles E. Metz ROCKIT 1.1B2 provides the following results: binormal parameters and area under the estimated ROC: a, 1.4888; b, 0.4950; area (Az), 0.9090; area (Wilc), 0.9153; estimated standard errors (SE) and correlation of these values: SE (a), 0.3047; SE (b), 0.1617; corr (a,b), 0.5744; SE (Az), 0.0384; SE (Wilc), 0.0365; symmetric 95% CI: For a, (0.8916, 2.0861); for b, (0.1782, 0.8118); asymmetric 95% CI for Az, (0.8091, 0.9636).

Fig. 4

By using the same proprietary algorithm developed to differentiate significant from nonsignificant fibrosis, 65% of the tested subjects would get an answer. Using Charles E. Metz ROCKIT 1.1B2 provides the following results: binormal parameters and area under the estimated ROC: a, 1.8231; b, 0.9943; area (Az), 0.9020; area (Wilc), 0.9005; estimated standard errors (SE) and correlation of these values: SE (a), 0.3697; SE (b), 0.2406; corr(a,b), 0.7157; SE (Az), 0.0322; SE (Wilc), 0.0378; symmetric 95% CI for a, (1.0984, 2.5477); for b, (0.5226, 1.4660); asymmetric 95% CI for Az, (0.8234, 0.9513).

Fig. 5

A proprietary algorithm that includes breath-test parameters, age and other patient data to differentiate intrahepatic inflammation (HAIa + HAIb + HAIc + HAId ≤ 4 vs > 4) applied on the 67% of the patient population assessed by the fibrosis algorithm, yields an area under the curve (AUC) of 0.89. Leaving the threshold at the point of best agreement (at 83%) found in the inflammation algorithm, results in 83% sensitivity and 81% specificity. At the dataset’s prevalence of 68%, the PPV is 91% and the NPV is 68%. Assuming a prevalence of 45.5%, this leads to a PPV of 78% and an NPV of 85%. Using Charles E. Metz ROCKIT 1.1B2 provides the following results: binormal parameters and area under the estimated ROC: a, 1.6781; b, 0.8979; area (Az), 0.8941; area (Wilc), 0.8737; estimated standard errors (SE) and correlation of these values: SE (a), 0.3529; SE (b), 0.3125; corr(a,b), 0.5320; SE (Az), 0.0419; SE (Wilc), 0.0419; symmetric 95% CI: for a, (0.9865, 2.3698); for b, (0.2855, 1.5103); asymmetric 95% CI for Az, (0.7882, 0.9552).

Fig. 6

A proprietary algorithm that includes breath-test parameters, age and other patient data to differentiate intrahepatic inflammation (HAIa + HAIb + HAIc + HAId ≤ 4 vs > 4) applied on the 33% of patient population not assessed by the fibrosis algorithm yields an area under the curve (AUC) of 0.96. Leaving the threshold on the point of best agreement (at 83%) found in the inflammation algorithm results in 82% sensitivity and 91% specificity. At the dataset’s prevalence of 68% the PPV is 95% and the NPV is 71%. Assuming a prevalence of 45.5%, this leads to a PPV of 88% and an NPV of 86%. Using Charles E. Metz ROCKIT 1.1B2 provides the following results: binormal parameters and area under the estimated ROC: a, 3.5668; b, 1.7357; area (Az), 0.9625; area (Wilc), 0.9628; estimated standard errors (SE) and correlation of these values: SE (a), 1.9626; SE (b), 1.5130; corr (a,b), 0.9003; SE (Az), 0.0419; SE (Wilc), 0.0314. Symmetric 95% CI: For a, (−0.2800, 7.4135); for b, (−1.2298, 4.7013); asymmetric 95% CI for Az, (0.7814, 0.9973).