Diagnosis of Minimal Hepatic Encephalopathy Using Stroop EncephalApp

A Multicenter US-Based, Norm-Based Study

Sanath Allampati, MD; Andres Duarte-Rojo, MD; Leroy R. Thacker, PhD; Kavish R. Patidar, MD; Melanie B. White, RN; Jagpal S. Klair, MD; Binu John, MD; Douglas M. Heuman, MD; James B. Wade, PhD; Christopher Flud, MD; Robert O'Shea, MD; Edith A. Gavis, RN; Ariel B. Unser, BS; Jasmohan S. Bajaj, MD

Disclosures

Am J Gastroenterol. 2016;111(1):78-86. 

In This Article

Results

Cross-sectional Study

We included 437 cirrhotic patients (VA=230, Ohio (OH)=107, Arkansas (AR)=100) who were compared with 308 controls (VA=103, OH=100, AR=105). The controls had a significantly lower age, lower proportion of males, and higher education compared with cirrhotics but were similar across the three sites (Table 1). The cirrhotics were similar in regards to age and education between groups. Therefore, age-, gender-, and education-adjusted norms were created for all tests.

Cirrhotic subjects from VA had a significantly higher MELD score and proportion with prior OHE compared with the other sites (Table 2).

Unadjusted Cognitive Performance. As expected, unadjusted normative performance on individual tests of the PHES and ICT results in cirrhotics was significantly worse than that of the healthy controls in the combined group and in individual centers. A similar impairment on the EncephalApp time-related outcomes was seen between controls and patients with cirrhosis in the combined group and individual centers (Table 1). Subjects with prior OHE were significantly worse on all cognitive tests, including EncephalApp in the combined group, as well as in individual centers (Supplementary Table 1 online http://www.nature.com/ajg/journal/v111/n1/suppinfo/ajg2015377s1.html).

Norms for PHES, ICT, and EncephalApp

We created norms for each of the tests using the 308 healthy controls. As the controls were different on demographics, we used age, gender, and education adjustment. The following regression formulae were derived for adjusting values of the individual results.

PHES. An adjustment with the following regression formulae for individual tests (DST=79.21−0.41 × Age−7.15 × Male_Gender+0.99 × Education, NCT-A=16.09+0.12 × Age+2.19 × Male_Gender−0.08 × Education, NCT-B=54.78+0.35 × Age+1.77 × Male_Gender−0.86 × Education, SDT=63.99+0.02 × Age−0.01 × Male_Gender−0.72 × Education, LTT (errors+time)=110.10−0.10 × Age+2.40 × Male_Gender−1.04 × Education). A score <−4 was considered positive for MHE by PHES. Thirty-seven percent of cirrhotics were MHE positive by PHES using these adjusted norms (Table 2).

ICT. We adjusted using the following formula for ICT lures (Total_Lures=7.75+0.01 × Age+0.78 × Male_Gender+0.03 × Education). A score <0 was considered positive for MHE by ICT; 35% cirrhotic subjects had MHE by ICT alone using these norms.

EncephalApp. The following regression formula for OffTime+OnTime—OffTime+OnTime=117.43+1.08 × Age+5.54 × Male_Gender−2.17 × Education—was used. A score <0 was considered positive for MHE by EncephalApp. Fifty-four percent of cirrhotic subjects had MHE by EncephalApp based on these direct population EncephalApp norms.

EncephalApp abnormalities Based on ICT and PHES

In subjects who had MHE on ICT and PHES using traditional methods with adjusted norms ICT (>1 s.d. beyond controls) and PHES (>−4 composite s.d. score beyond controls), EncephalApp performance was significantly worse on time-related aspects in those with MHE on PHES or ICT compared with those without MHE (Table 3). On a receiver operating characteristic analysis, the maximum AUC was found to be with a cutoff point of 195.9 s for OffTime+OnTime using PHES. Using ICT lures as the gold standard, the highest AUC was found at the cutoff point of 181.9 s. Using these cutoff points, we found that 54% cirrhotics had MHE based on EncephalApp using ICT as the gold standard (>181.9 s OffTime+OnTime, AUC: 0.73) and 37% had MHE based on EncephalApp using PHES as the gold standard (>195.9 s OffTime+OnTime, AUC: 0.80).

Concordance and Agreement Between Sites

Concordance between MHE diagnosed using norms was studied between the three sites and then compared with the combined values. This showed moderate agreement between EncephalApp and ICT and better agreement between EncephalApp and PHES (Table 4).

The phi coefficients between the individual centers and the combined values calculated between sites using the three modalities showed modest to good agreement (Table 5).

Prediction of OHE Development

Entire Group Outcomes. On a median follow-up of 11 (interquartile range 8–15) months, 13% total patients developed an episode of OHE (12% VA, 14% OH, 20% AR).

Using EncephalApp MHE based on norms, on multivariable analysis, prior OHE (hazard ratio (HR) 1.9, P=0.04), MHE (HR 2.1, P=0.04), and MELD score (HR 1.4, P=0.05) were significant predictors.

When using ICT as the gold standard for EncephalApp, time to next OHE episode after baseline testing was predicted by MELD (HR1.05, P=0.03), OHE at baseline (HR 1.95, P=0.04), and EncephalApp impairment based on ICT (HR 2.1, P=0.05) but age, gender, and alcoholic etiology were not significant. The model using PHES as the gold standard for EncephalApp was similar as the model above, but none of the variables apart from OHE at baseline and MELD score were significant. Cox regression analysis for OHE episode development in all subjects was significant for EncephalApp impairment based on norms (P=0.04, Figure 1a), EncephalApp impairment based on ICT (P=0.05, Figure 1c), and borderline on EncephalApp impairment based on PHES (P=0.11, Figure 1e).

Figure 1.

Cox regression analysis for overt hepatic encephalopathy (OHE) episode development: x axis is time in months; and y axis is percentage free of OHE development. Dotted lines are minimal hepatic encephalopathy (MHE) patients, and solid lines are those without MHE based on the definitions in each figure. (a) In all subjects with EncephalApp impairment based on norms (P =0.04). (b) In subjects without OHE with EncephalApp impairment based on norms (P =0.05). (c) In all subjects with EncephalApp impairment based on inhibitory control test (ICT; P =0.05). (d) In subjects without OHE with EncephalApp impairment based on ICT (P =0.02). (e) In all subjects with EncephalApp impairment based on psychometric hepatic encephalopathy (PHES; P =0.11). (f) In subjects without OHE with EncephalApp impairment based on PHES (P =0.006).

Outcomes in Those Without Prior OHE. In the 252 total patients without prior OHE, on a median follow-up of 10 (interquartile range 8–15) months, 9% total patients developed their first OHE episode (7% VA, 9% OH, 13% AR). Using EncephalApp MHE based on norms, on multivariable analysis, MHE on EncephalApp (HR 2.5, P=0.04), and MELD score (HR 1.1, P=0.002) were significant predictors of the first OHE episode. When we defined MHE on EncephalApp based on PHES, the significant predictors were MELD score (HR 1.14, P=0.0004) and MHE (HR 7.4, P=0.0006). In the third model where MHE on EncephalApp was based on ICT, MHE continued to be a predictor (HR 4.1, P=0.02) independent of the MELD score (HR 1.14, P=0.004). Age, gender, and alcoholic etiology were not predictive in any model.

Cox regression analysis for OHE episode development in subjects without prior OHE was significant for EncephalApp impairment based on norms (P=0.05, Figure 1b), EncephalApp impairment based on ICT (P=0.02, Figure 1d), and EncephalApp impairment based on PHES (P=0.005, Figure 1f).

Therefore, EncephalApp abnormalities, whether based on direct population norms or based on existing gold standards, independently predicted the time to OHE development.

Validation Cohort

A new set of 121 cirrhotic patients were recruited for VA as a validation cohort. The mean age was 59±10 years, MELD 13.5±6.5, and years of education was 13.1±2.5 years. Etiology of cirrhosis was hepatitis C only in 41%, alcohol only in 11%, both hepatitis C and non-alcoholic steatohepatitis and alcohol in 10%, non-alcoholic steatohepatitis in 18%, and other etiologies in 20% of patients. Forty-five patients had prior OHE, all of whom were controlled on lactulose with a 15% additional rifaximin usage. The individual test results are shown in Supplementary Table 2 http://www.nature.com/ajg/journal/v111/n1/suppinfo/ajg2015377s1.html. Using the adjusted norms (age, gender, education) generated from healthy controls previously, we found that the prevalence of MHE using EncephalApp directly was 51%, 42% with EncephalApp based on PHES, and 55% using EncephalApp based on ICT. These proportions were similar to those generated by the initial cohort (54% direct EncephalApp norms, 37% EncephalApp based on PHES, and 54% EncephalApp based on ICT, P>0.05 on all comparisons between initial and validation cohorts). There was also good sensitivity (72%) and specificity (88%) in the validation cohort using EncephalApp cutoff points based on PHES. Using EncephalApp cutoff points based on ICT, there was excellent sensitivity (88%) but moderate specificity (56%) in the validation cohort compared with the initial cohort.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....