Primary spoken language may be associated with how the Animal Naming Test, or ANT, performs when it is used to identify minimal hepatic encephalopathy, or MHE, in people with cirrhosis, a small prospective study in Taiwan suggests. The study also tested whether serum IL-6 could add diagnostic information.
Among Mandarin-speaking patients, the ANT had an AUROC of 0.760, serum IL-6 had an AUROC of 0.841, and a model combining the two reached 0.895. AUROC was the study's measure of how well each marker separated patients with and without MHE. Although the combined result was numerically higher, pairwise DeLong comparisons did not show a statistically significant advantage over the individual markers.
How the test was assessed
Researchers prospectively enrolled consecutive cirrhotic outpatients at a single tertiary referral hospital. Patients completed the ANT, PHES and MMSE, underwent eyes-closed digitalized EEG, and had serum IL-6 measured. Age-matched healthy controls were recruited from relatives.
Recruitment assessed 77 cirrhotic patients and 34 healthy controls. After 12 exclusions, 65 cirrhotic patients remained, including 16 classified as having MHE and 49 classified as non-MHE. To define MHE, the study used abnormal PHES and/or visually assessed EEG slowing in patients without overt hepatic encephalopathy. An abnormal PHES was a total score of -4 or lower, while one EEG criterion was a dominant posterior background rhythm below 8 Hz.
Diagnostic discrimination was assessed with language-stratified ROC and AUROC analyses. The researchers also used logistic regression, DeLong tests to compare AUROCs, and a composite model combining ANT with IL-6. Exploratory comparisons were adjusted with false-discovery-rate and Holm-Bonferroni corrections, with p-values below 0.05 treated as statistically significant.
The overall result
MHE was diagnosed in 16 of 65 cirrhotic patients, or 24.6%. Those in the MHE group were older on average than those without MHE, at 65.6 years versus 59.6 years, with p = 0.032. Sex distribution and education were similar; median education was 9 years versus 12 years, with p = 0.138.
Across the full group, MHE patients had lower median PHES, ANT and S-ANT scores and higher serum IL-6 than non-MHE patients. Median PHES was -2 versus 0, ANT was 15 versus 19, and S-ANT was 15.5 versus 19. The corresponding p-values were 0.001, 0.014 and 0.023. MMSE was 26.5 versus 28, a difference that was not statistically significant, with p = 0.052.
Serum IL-6 had the highest single-marker AUROC in the overall cohort, at 0.816. At a cohort-specific cutoff of 4.1 pg/mL, it detected 87.5% of MHE cases and correctly identified 69.4% of patients without MHE. Raising the fixed cutoff to 7 or 8 pg/mL lowered sensitivity to 50.0% and 43.8%, while specificity was 89.8% at both. At 11 pg/mL, sensitivity was 12.5% and specificity was 93.9%. The tradeoff shows why the cutoff cannot be treated as universal, and the study says it needs validation.
Results changed by language
In the Mandarin subgroup, median ANT scores were 12 among MHE patients and 18 among non-MHE patients, while S-ANT scores were 15 and 18. PHES was -2 versus 1, MMSE was 26 versus 29, and serum IL-6 was 8.0 versus 2.7 pg/mL. The p-values for those comparisons were 0.017, 0.020, below 0.001, 0.029 and 0.002, respectively. After false-discovery-rate correction, ANT, S-ANT, the exploratory calibrated S-ANT, PHES and IL-6 remained significant; MMSE did not.
The reported AUROCs in Taiwanese Hokkien-speaking patients were 0.679 for ANT and 0.617 for the original S-ANT. An exploratory calibrated S-ANT reached 0.776, while MMSE was 0.505, serum IL-6 was 0.735 and the ANT-plus-IL-6 composite was 0.755.
The calibrated score was a post hoc adjustment that added three points for patients older than 70 and/or those with fewer than six years of education. Its apparent AUROC was 0.776, compared with 0.617 for the original S-ANT, but the DeLong comparison gave p = 0.136. Bootstrap correction produced AUROC 0.773 rather than 0.776. Because the calibration was derived from that subgroup itself, this was internal validation, not an external test.
After adjustment, ANT remained associated with MHE in the full cohort, with an odds ratio of 0.796 and a 95% confidence interval from 0.654 to 0.969, p = 0.023. The association also remained in the Mandarin subgroup, with an odds ratio of 0.744 and a 95% confidence interval from 0.568 to 0.974, p = 0.032. It was not statistically significant in the Taiwanese Hokkien subgroup, where the odds ratio was 0.868, with a 95% confidence interval from 0.648 to 1.164, p = 0.345.
An exploratory age-stratified analysis found an ANT AUROC of 0.820 among patients younger than 65 and 0.528 among those aged 65 or older. The MHE group was older at baseline, so the age-specific findings remain exploratory despite the adjusted analyses.
A result that needs a bigger test
The picture remains provisional. This was a single-center prospective proof-of-concept cohort, and the language-stratified estimates were exploratory. The Taiwanese-calibrated S-ANT was derived within the subgroup, while the IL-6 cutoff was specific to this cohort. Neither finding had external validation.
A sensitivity analysis used EEG-defined MHE as the reference. The tested measures retained AUROCs from 0.699 to 0.766, but only 13 patients were EEG-positive, so the analysis was not language-stratified and remains exploratory.
The study does not establish a causal effect of spoken language on ANT performance, statistically significant superiority of the ANT-plus-IL-6 model in Mandarin speakers, or whether the Taiwanese-calibrated S-ANT and IL-6 cutoff work outside this cohort. Larger independent studies will need to test those findings across language groups.
The article reports acceptance on July 23, 2026, and says it is open access under a Creative Commons license. The work was supported by grants CMRPG3P0761 and CMRPG3M1932 from Chang Gung Memorial Hospital.
Paper data and sources
Original title: A Proof-of-Concept Study of Language-Stratified Assessment for Minimal Hepatic Encephalopathy: Integrating the Animal Naming Test and Serum IL-6.
Authors: Hsin-Che Lin, Cheng Jen Chen, Tsung-Han Wu et al.
Journal/Repository: The Kaohsiung journal of medical sciences
Status: Peer-reviewed
First online: 2026-08-20
DOI: 10.1002/kjm2.70271
Original paper