Metab Brain Dis
DOI 10.1007/s11011-014-9529-0
RESEARCH ARTICLE
Assessing inter- and intra-individual cognitive variability
in patients at risk for cognitive impairment: the case
of minimal hepatic encephalopathy
Patrizia Bisiacchi & Giorgia Cona & Vincenza Tarantino &
Sami Schiff & Sara Montagnese & Piero Amodio &
Giovanna Capizzi
Received: 23 January 2014 / Accepted: 12 March 2014
# Springer Science+Business Media New York 2014
Abstract Recent evidence reveals that inter- and intra-
individual variability significantly affects cognitive perfor-
mance in a number of neuropsychological pathologies. We
applied a flexible family of statistical models to elucidate the
contribution of inter- and intra-individual variables on cogni-
tive functioning in healthy volunteers and patients at risk for
hepatic encephalopathy (HE). Sixty-five volunteers (32 pa-
tients with cirrhosis and 33 healthy volunteers) were assessed
by means of the Inhibitory Control Task (ICT). A Generalized
Additive Model for Location, Scale and Shape (GAMLSS)
was fitted for jointly modeling the mean and the intra-
variability of Reaction Times (RTs) as a function of socio-
demographic and task related covariates. Furthermore, a Gen-
eralized Linear Mixed Model (GLMM) was fitted for model-
ing accuracy. When controlling for the covariates, patients
without minimal hepatic encephalopathy (MHE) did not differ
from patients with MHE in the low-demanding condition,
both in terms of RTs and accuracy. Moreover, they showed a
significant decline in accuracy compared to the control group.
Compared to patients with MHE, patients without MHE
showed faster RTs and higher accuracy only in the high-
demanding condition. The results revealed that the application
of GAMLSS and GLMM models are able to capture subtle
cognitive alterations, previously not detected, in patients’
subclinical pathologies.
P. Bisiacchi (*) : G. Cona : V. Tarantino
Department of General Psychology, University of Padova, Padova,
Italy
e-mail: patrizia.bisiacchi@unipd.it
S. Schiff : S. Montagnese : P. Amodio
Department of Medicine, University of Padova, Padova, Italy
G. Capizzi
Department of Statistical Sciences, University of Padova, Padova,
Italy
Keywords Inter-intra individual difference . Cognitive
control . Cirrhosis . Sub-clinical brain impairments . Hepatic
encephalopathy
Introduction
The study of inter- and intra-individual variability is, at the
moment, considered an important topic in the field of cognitive
neuroscience (MacDonald et al. 2009). The relevance of differ-
ences between individuals of the same group (inter-individual
variability or between-subject variability), and across different
moments within the same individual (intra-individual
variability or within-subject variability), has gained increasing
interest in cognitive and neuropsychological research, as well as
in a clinical setting. Importantly, high within-subject variability
in cognitive tasks has been found to correlate with clinical
diagnoses in patient populations, such as Attention Deficit/
Hyperactivity Disorders (ADHD), brain injury, dementia, de-
pression, bipolar disorder, and schizophrenia (Stuss et al. 2003;
Castellanos et al. 2006; MacDonald et al. 2006; Hultsch et al.
2008; Kaiser et al. 2008; Depp et al. 2012; Tarantino et al.
2013). Converging clinical evidence indicates that when
looking at the mean task performance of a patient, he or she
might show unimpaired results; however, his or her perfor-
mance very often shows highly inconsistent moment-by-
moment results, even in the same evaluating session and irre-
spective of the specific cognitive task administered.
This is especially true when considering sub-clinical brain
impairments or early stages of neurodegenerative disorders,
e.g., Mild Cognitive Impairment (MCI). Therefore, intra-
individual variability can be considered an early marker of brain
dysfunction (Burton et al. 2006; MacDonald et al. 2006).
Behavioral analyses of intra-individual variability of reaction
times (RTs) and accuracy in cognitive tasks have documented

that variability is not simply error variance or statistical noise but
should be interpreted as a reliable index of cognitive
malfunctioning (e.g., Castellanos et al. 2006). However, con-
ventional statistical analysis largely overshadows data on intra-
individual variability by central tendency measures. Further-
more, it does not take into account the contribution of inter-
individual variables. In this study we investigated, by means of a
relatively recent family of statistical models (GAMLSS and
GLMM), the cognitive functioning of patients with cirrhosis,
with or without Minimal Hepatic Encephalopathy (MHE).
MHE is a syndrome associated with acute or chronic liver failure
that occurs in 20 to 60 % of patients with cirrhosis and that leads
to mild cognitive deficits (Amodio et al. 2004; Dhiman and
Chawla 2009). The clinical profile of MHE is characterized by
cognitive alterations—sometimes difficult to clearly detect—
including changes in selective attention, visuomotor ability,
psychomotor speed, response inhibition and selection, and, in
particular, executive functions (e.g., Amodio et al. 1998;
Schomerus and Hamster 1998; Weissenborn et al. 2001, 2005).
Several computerized tests, such as the Sternberg para-
digm, Simon task, and Inhibitory Control Test (ICT), have
been used to investigate cognitive functioning of MHE pa-
tients (e.g., Amodio et al. 1998, 1999; Bajaj et al. 2008; Schiff
et al. 2005, 2006). The classical measures obtained from these
tests were mean values of RTs and accuracy. Taken together,
these studies showed that patients with cirrhosis having MHE
had significant slower mean RTs compared to both controls
and patients with cirrhosis without MHE (Schiff et al. 2005,
2006). On the other hand, patients with cirrhosis not having
MHE did not show significant differences in mean RTs rela-
tive to controls (Schiff et al. 2005, 2006).
Concerning the mean accuracy, some studies did not find
any difference between controls and patients with cirrhosis,
either with or without MHE (e.g., Amodio et al. 1998; Schiff
et al. 2006), whereas other studies demonstrated an impaired
performance in patients with cirrhosis and MHE compared to
controls, but not in patients with cirrhosis without MHE
(Felipo et al. 2012).
The absence of significant between-group differences and
the presence of contradictory findings might be explained by
inter- and/or intra-individual variables. Indeed, all of the afore-
mentioned studies focused on mean values of task performance,
ignoring information about the variability of RTs and accuracy
between patients and across trials. Elssas and collaborators
(1985) showed that patients with cirrhosis had a higher intra-
individual variability of RTs compared to controls and patients
with brain damage. When scrutinizing mean RT data of previ-
ously published studies, what emerges is that mean RTs in
patients with cirrhosis not having MHE were in fact intermedi-
ate between cirrhotic patients with MHE and control (Schiff
et al. 2005), or differed from controls in high cognitive demand-
ing task conditions (Schiff et al. 2006). The distributions of RTs
reported in Schiff et al. (2006) clearly revealed that patients with
Metab Brain Dis
MHE had a greater variability of RTs compared to patients
without MHE and controls. Although these findings important-
ly suggested that even cirrhotic patients without MHE might
have slight cognitive alterations, a direct comparison of standard
deviations (or other indices of variability) within and between
groups has not been performed so far.
The aim of the present study was to elucidate differences in
performance on the cognitive control task (ICT) by consider-
ing the contribution of inter- and intra-individual variability in
the performance. To this end, behavioral responses to the ICT,
namely RTs and accuracy, were obtained. These measures
were compared by adopting a statistical approach that allows
subject-varying covariates to influence the mean and the intra-
individual variance of RT, differentially affecting task perfor-
mance. If patients with cirrhosis encompass cognitive alter-
ations compared to healthy controls, these would emerge after
controlling for such predictors.
The choice of the ICT was driven by the fact that this test
has been proposed as a promising diagnostic tool for cognitive
deficits in cirrhosis (Bajaj et al. 2008; Cona et al. 2013; Schiff
et al. 2013). Furthermore, because the ICT is composed of
different task conditions, and thus entails different cognitive
processes to be performed (e.g., attention, updating informa-
tion in memory, response inhibition), analysis of variability
according to the task condition provides information about
specific processes affected in cirrhotic patients at different
stages of illness.
Material and methods
Participants
Thirty-two patients with liver cirrhosis and 33 healthy control
volunteers were enrolled. The main demographic features of the
study sample are presented in Table 1. The diagnosis of cirrho-
sis was based on case history; clinical examination; biochemi-
cal, endoscopic, and ultrasound (US) findings; or, when need-
ed, on liver biopsy. The diagnostic criteria were: the presence of
hepatic stigmata during a routine clinical examination, together
with biochemical indices of decompensated liver disease (low
serum albumin, high bilirubin, prolonged prothrombin time,
low platelet count); endoscopic or US signs of portal hyperten-
sion; or a history of previous de-compensation (ascites, jaun-
dice, bleeding from esophageal varices). The exclusion criteria
were: chronic obstructive lung disease/respiratory failure of
other origin; renal insufficiency; coronary heart disease or heart
failure of any origin; significant neurological/psychiatric dis-
ease; overt HE or bouts of overt HE within 15 days prior to the
study; bleeding or infections within 15 days prior to the stud;
psychoactive medication (benzodiazepines, antidepressant,
neuroleptic drugs); active drinking; or unwillingness/inability
to undergo the study procedure.
Metab Brain Dis
Table 1 Mean values, range of
demographic variables and as- Controls (n=33)
sessment measures divided by
group Demographic variables
Age 48 (28–74)
Education 14 (5–21)
Cognitive measures (PHES)
TMT-A (sec) 32 (13–59)
TMT-B (sec) 67 (28–177)
Serial dotting 39 (24–61)
Line tracing 66 (18–142)
MHE− patients with cirrhosis
without MHE, MHE+ patients Digit symbol 44 (20–57)
with cirrhosis having MHE Psychophysiological measures (EEG)
*Controls versus Patients MHE+ MDF – 88.8 (8–518)
(p<0.05) Delta power – 6.5 (2–12)
°Patients MHE− versus Patients Theta power – 17.1 (6–31)
MHE+ (p<0.05)
The study protocol conformed to the ethical guidelines of
the Declaration of Helsinki and was approved by the local
ethical and research committees. Signed informed consent
was obtained from all participants. The diagnosis of MHE
was based on spectral EEG features and on psychometric
hepatic encephalopathy score (PHES; Amodio et al. 2007;
Weissenborn et al. 2001). Cirrhotic patients were qualified
as having MHE if either PHES or EEG were abnormal
(Ferenci et al. 2002). On the basis of this criterion, 18 patients
were qualified as without MHE and 14 as having MHE.
All participants underwent validated paper-and-pencil neu-
ropsychological assessment, including the Mini Mental State
Examination (MMSE) and the PHES, which consists of a
battery of standardized cognitive tests aimed at the diagnosis
of MHE: the Trail Making Test part-A (TMTA), the Trail
Making Test part B (TMT-B), the Digit Symbol Test (DST),
the Serial Dotting Test (SDT), and the Line Tracing Test
(LTT). For diagnostic aims, each test score was quantified
according to age- and education-adjusted norms. All included
participants scored 27 or more on the MMSE.
The EEG was recorded according to the procedure by
Amodio et al. (1999). The outcome variables considered for
EEG evaluation were the mean dominant frequency (MDF, an
estimate of the background frequency of the EEG) and the
relative amount of slow EEG activity within the delta and
theta frequency bands. In brief, the EEG was considered to be
abnormal if the mean dominant frequency (MDF) was
≤7.3 Hz or the theta relative power ≥35 %.
Age, education, cognitive, and psychophysiological
measures were compared across groups by means of
separated univariate ANOVAs (see Table 1). Age did
not differ across groups, meaning that age was compa-
rable. Patients with MHE obtained significantly lower
scores in all tests but the TMT-B; on the other hand,
patients without MHE did not differ from the control
group in any cognitive test.
Patients MHE- (n=18) Patients MHE+ (n=14)
52 (30–74) 54 (44–64)
13(8–18) 9 (5–15)*°
31 (20–58) 49 (29–78)*°
91 (41–300) 120 (62–300)*
44 (33–78) 53 (37–78)*°
74 (12–214) 97 (51–126)*°
38 (17–60) 28 (19–37)*°
85.8 (25–154)
6.3 (2–15)
47.2 (14–72)°
Procedure
The ICT was administered immediately after the PHES evalua-
tion. The ICT was based on the original paradigm proposed by
Bajaj and colleagues (2008) and was developed by the E-prime
software (Psychological Software Tools, Pittsburgh, PA). Dur-
ing the task, black letters are presented, one after the other, in the
center of a white background computer screen for 500 ms each,
without inter-stimulus interval. During the first part of the task,
the subject is instructed to respond to the target letters ―X‖ and
―Y‖ by pressing the space bar of the computer keyboard
(Detect task condition). During the second part of the task, the
subject is instructed to respond only to ―X‖ and ―Y‖ preceded
by ―Y‖ and ―X‖ respectively (alternating target letters or Go
task condition) and to withhold response when ―X‖ and ―Y‖
are preceded by ―X‖ and ―Y‖, respectively (repeated target
letters or No-Go task condition). In other words, the subject is
instructed to respond only if an ―X‖ is preceded by a ―Y‖ or a
―Y‖ is preceded by an ―X‖. Responses must be inhibited if an
―X‖ is preceded by another ―X‖ or a ―Y‖ is preceded by
another ―Y‖. Target letters are always interspersed with other,
irrelevant, letters. The first part of the task was divided into two
blocks for a total of 500 trials (letters), 75 of which required a
response (Detect task condition). The second part comprised 6
blocks of trials, for a total of 1,753 letters, 254 of which required
a response (Go task condition) and 47 did not require a response
(No-Go task condition). The second part of the test was preced-
ed by a practice run, to allow the subject to become familiar with
the Go/No-Go task conditions.
The accuracy was expressed as percentage of correct
responses (correct responses/total number of trials*100). Re-
action times (RTs) for Detect, Go, and No-Go task conditions
were obtained. ICT is composed of many repeated trials. This
allows studying the trial-by-trial fluctuations of performance,
providing a reliable measure of intra-individual variability
both on RTs and accuracy.

Statistical analyses
Two performance measures were separately modeled: RTs, in
terms of mean and variance; and accuracy, in terms of mean
percentage of correct responses.
In the present work we used the Generalized Additive
Models for Location, Scale and Shape (GAMLSS) Rigby and
Stasinopoulos (2005), with the aim to characterize subjects’
mean values of RTs and variance. The GAMLSSs have prov-
en to be a convenient statistical tool when the response vari-
able distribution explicitly depends on covariates, as happens
in the RT distribution case. The following covariates were
entered into the model: age; block (initial, middle, final); task
condition (Detect, Go, No-Go); and group membership (pa-
tients and healthy controls). Specifically, in order to capture
possible time-on-task effects generated by practice as well as
fatigue, individual RTs were available at three different time
points, namely for three consecutive blocks of trials (initial,
middle, final). The three ICT conditions were separately con-
sidered as they require different degrees of cognitive demands.
Endpoints of 95 % two-sided confidence intervals for
model parameters were computed via bias corrected bootstrap
(Efron and Tibshirani 1993), using 1,000 replications.
Accuracy has been modeled using a binomial Generalized
Linear Mixed Model (GLMM; Fitzmaurice et al. 2004). Pa-
rameters models of the GLMM have been estimated via
maximum likelihood. Unilateral p-values for each contrast
were calculated. This choice is based on the hypothesis that
both RT and accuracy can deteriorate with task difficulty and,
conditioned to a given task, with the presence of a disease.
Results
Fitted fixed and random-effect parameters for intra-individual
mean and variance values of RTs are reported in Tables 2 and 3,
respectively, together with the unilateral p-values for specific
pairs of comparisons.
Mean values of RTs
Descriptive analyses showed that age significantly affected
RTs, with RTs being slower with advancing age. Figure 1
shows the effect of age (arbitrarily divided into under and
over 50) across groups and task conditions.
The GAMLSS model showed that, adjusting for all other
predictors, the mean values of RTs significantly increased with
age and were significantly faster in the middle blocks com-
pared to the initial ones, and in the final blocks compared to
the initial and middle ones, confirming the presence of a
practice-related improvement over time. Furthermore, mean
values of RTs in the Go and in the No-Go task conditions were
significantly slower than those recorded in the Detect
Metab Brain Dis
condition, meaning that the first two types of trials necessitate
more cognitive resources. In particular, the No-Go condition
requires an inhibition process. Hence, the speed of response to
such trials represents a sort of non-decision time up to a wrong
response.
The analysis of contrasts, reported in the bottom of
Table 2, showed that overall mean RTs significantly
differ between patients (with and without MHE) and
controls. Namely, in all task conditions, mean RTs of
patients with MHE were significantly slower than mean
RTs of the control group.
Importantly, the analysis also revealed that RTs between
patients without MHE and patients with MHE did not differ in
the No-Go task condition (p=0.089) supporting the idea that
such a task requires more cognitive resources. In addition, the
mean values of RTs of patients without MHE differed from
those of patients with MHE only in the Go task condition
(p<0.05) in which patients without MHE performed in the
normal range.
In summary, overall mean values of RTs were faster in the
control group, intermediate in the group of patients without
MHE, and slower in the group of patients with MHE. Inter-
estingly, differences between patients without MHE and con-
trols emerged only in the high cognitive demanding task
condition (No-Go).
Variance
Results show that, adjusting for the other covariates, individ-
ual RT variance significantly increases with age and decreases
in the middle and final block of trials. It means that practice
almost gradually reduces the intra-individual variability of
RTs. Furthermore, RT variance was significantly higher in
the Go and No-Go task conditions compared to the Detect
condition. The analysis of contrasts for the group × task
condition interaction revealed that in the Detect condition,
individual RT variance did not differ between controls and
patients, with and without MHE. In the Go condition, patients
with MHE showed significantly higher individual RT vari-
ance compared to controls and patients without MHE
(p<0.005 and p<0.05 respectively). Interestingly, in the No-
Go task condition, RT variance of both patients with and
without MHE were significantly larger than that of the control
(p<0.001 and p<0.05 respectively), whereas individual RT
variance of patients with and without MHE did not differ
significantly (p=0.059).
Accuracy
Estimated parameters for the GLMM model are summarized
in Table 4. The table contains the bilateral p-values for each
parameter and the unilateral p-value for each contrast. Age,
order of blocks, task conditions, group, and years of education
Metab Brain Dis

Table 2 Upper table: GAMLSS parameters of RT means. The confidence intervals were obtained by bias-corrected bootstrap. Lower table: results of
contrast analyses of GAMLSS parameters, divided by task condition. Unilateral p-values for each contrast are reported

Coefficient 95 % CI
Lower limit Upper limit
Fixed effects
Intercept 412.52a 387.25 441.67
Age 2.28a 1.67 2.95
Middle blocks −27.12a −37.35 −16.99
Final blocks −34.01a −44.62 −24.73
Go task condition 56.63a 45.73 70.57
No-Go task condition 64.10a 49.62 78.30
Patients MHE− 12.42 −8.66 39.24
Patients MHE+ 50.45a 13.46 91.17
Go: patients MHE− 8.48 −9.11 26.65
Go: patients MHE+ 13.46 −3.29 30.55
No-Go: patients MHE− 20.10 −7.80 48.27
No-Go: patients MHE+ 16.53 −3.55 39.50
Random effects
λ 69.70 48.08 94.56
Coefficient p-value
Detect Patients MHE− vs controls 12.42 0.171
Patients MHE+ vs controls 50.45* 0.019
Patients MHE+ vs patients MHE− 38.04 0.072
Go Patients MHE− vs controls 20.89 0.099
Patients MHE+ vs controls 63.91** 0.003
Patients MHE+ vs patients MHE− 43.02* 0.044
No-Go Patients MHE− vs controls 32.52* 0.034
Patients MHE+ vs controls 66.98** 0.001
Patients MHE+ vs patients MHE− 34.47 0.089

MHE− patients with cirrhosis not having MHE, MHE+ patients with cirrhosis having MHE
*p<0.05; **p<0.01
a
Significant effect (i.e., 95 % IC not containing zero value)

significantly affected accuracy. In particular, adjusting for all
other covariates, accuracy significantly decreases with age
(p<0.01). Then, accuracy significantly increases in the middle
and final blocks of trials compared to the initial one (both ps<
0.001), confirming the presence of a practice effect.
Concerning the task condition × education interaction, higher
years of education seem to be unable to support respondents in
the execution of a more demanding task. Indeed, accuracy
significantly improves with education only in performing the
Detect task condition. Contrast analyses revealed that, in
terms of accuracy, in the Detect task condition patients with
MHE showed significantly lower accuracy than the control
group (p<0.01), but they did not differ significantly from the
group without MHE (p=0.243).
Importantly, in this task condition patients without MHE
showed significantly lower accuracy than the control group
(p<0.05). In the Go condition, patients without MHE and
controls did not differ significantly, while patients with MHE
perform significantly worse than the two other groups (both ps<
0.001). In the No-Go condition, the three diagnostic groups did
not show significant differences in terms of accuracy.
Discussion
Previous studies explored the cognitive alterations associated
with cirrhosis and, particularly, with MHE (e.g., Amodio et al.
1998, 1999; Bajaj et al. 2008; Schiff et al. 2005, 2006). All of
these studies drew their conclusions on the basis of differences in
terms of group means of performance observed between cirrhot-
ic patients (with or without MHE) and healthy individuals.
Nevertheless, the mean performance on a given cognitive
task might be also the result of other variables, not strictly related
to the pathology object of the study. These include inter-
individual variables, such as age and education of the sample,
and intra-individual variables, such as the difficulty of the task or
 Metab Brain Dis

Table 3 Upper table: GAMLSS parameters of RT variance. The confidence intervals were obtained by bias-corrected bootstrap. Lower table: results of
contrast analyses of GAMLSS parameters, divided by task condition. Unilateral p-values for each contrast are reported

Coefficient 95 % CI
Lower limit Upper limit
Fixed effects
Intercept 4.471a 4.334 4.661
Age 0.006a 0.003 0.009
Middle blocks −0.279a −0.336 −0.231
Final blocks −0.366a −0.422 −0.312
Go task condition 0.227a 0.139 0.306
No-Go task condition 0.109a 0.013 0.209
Patients MHE− 0.026 −0.105 0.170
Patients MHE+ 0.143 −0.018 0.316
Go: patients MHE− −0.002 −0.110 0.119
Go: patients MHE+ 0.092 −0.011 0.214
No-Go: patients MHE− 0.198 −0.014 0.424
No-Go: patients MHE+ 0.287a 0.101 0.513
Random effect
ζ 0.226 0.189 0.274
Coefficient p-value
Detect Patients MHE− vs controls 0.027 0.335
Patients MHE+ vs controls 0.143 0.070
Patients MHE+ vs patients MHE− 0.117 0.145
Go Patients MHE− vs controls 0.025 0.379
Patients MHE+ vs controls 0.235** 0.001
Patients MHE+ vs patients MHE− 0.210* 0.015
No-Go Patients MHE− vs controls 0.225* 0.017
Patients MHE+ vs controls 0.430*** <0.001
Patients MHE+ vs patients MHE− 0.206 0.059

MHE− patients with cirrhosis not having MHE, MHE+ patients with cirrhosis having MHE
*p<0.05; **p<0.01; ***p<0.001
a
Significant effect (i.e., 95 % IC not containing zero value)

the order of task presentation. The aim of the present study was
to elucidate differences in performance on the cognitive control
task (ICT) by considering the contribution of inter- and intra-
individual variability in the performance.
We analyzed RTs and accuracy using the GAMLSS anal-
ysis. This approach has the advantage, particularly relevant in
clinical studies, of detecting subtle differences free from inter-
and intra-individual differences. In this way it was possible to
consider patients’ and healthy adults’ performance on the
cognitive task free from other influencing variables such as
education, age, etc. Obviously, there could be a large plurality
of variables that could affect performance (e.g., emotional

Fig. 1 a RT Means across groups

and task conditions of participants
<50 years; b RT means across
groups and task conditions of
participants >50 years
Metab Brain Dis

Table 4 Upper table: GMLSS parameters of accuracy. Bilateral p-values for each parameter are reported. Lower table: results of contrast analyses
divided by task conditions. Unilateral p-values for each contrast are reported

Coefficient Standard error z p-value

Fixed effects
Intercept 2.51*** 0.60 4.18 <0.001
Age −0.017** 0.006 −2.68 0.007
Education 0.12*** 0.03 3.83 <0.001
Middle blocks 0.71*** 0.08 8.31 <0.001
Final blocks 0.93*** 0.10 9.08 <0.001
Go trials −0.04 0.27 −0.16 0.874
No-Go trials −0.67* 0.33 −2.03 0.042
Patients MHE- −0.47 0.26 −1.78 0.074
Patients MHE+ −0.70* 0.30 −2.32 0.020
Education: Go −0.11*** 0.02 −5.58 <0.001
Education: No-Go −0.12*** 0.02 −5.39 <0.001
Go: Patients MHE− 0.44* 0.18 2.43 0.014
Go: Patients MHE+ −0.37* 0.18 −2.06 0.039
No-Go: Patients MHE− 0.54* 0.21 2.53 0.011
No-Go: Patients MHE+ 0.27 0.22 1.24 0.215
Coefficient Standard error z p-value
Detect trials Patients MHE− vs controls −0.47* 0.26 −1.78 0.037
Patients MHE+ vs controls −0.70* 0.30 −2.32 0.010
Patients MHE+ vs patients MHE− −0.22 0.32 −0.69 0.243
Go trials Patients MHE− vs controls −0.04 0.22 −0.16 0.434
Patients MHE+ vs controls −1.08*** 0.26 −4.04 <0.001
Patients MHE+ vs patients MHE− −1.04*** 0.28 −3.69 <0.001
No-Go trials Patients MHE− vs controls 0.07 0.25 0.27 0.606
Patients MHE+ vs controls −0.43 0.29 −1.47 0.069
Patients MHE+ vs patients MHE− −0.50 0.31 −1.60 0.055

MHE− patients with cirrhosis not having MHE, MHE+ patients with cirrhosis having MHE
*p<0.05; **p<0.01; ***p<0.001

mood state, implicit ideas); however, we took into account
those variables that were demonstrated to have a major impact
on cognitive performance.
The results on one hand confirmed some of the findings
evidenced in previous studies (cf. Bajaj et al. 2008) and on the
other hand provided new insights on the cognitive changes
associated with cirrhosis. Below, these two aspects are de-
scribed in more detail. After controlling for the effects of all of
the other variables, we found that patients with MHE had
slower RTs in all conditions as compared to healthy controls.
Moreover, they displayed an increased RT variance in the Go
and No-Go conditions compared to healthy controls. Also,
patients with MHE were less accurate than controls not only in
the Go condition, but also in the Detect condition. This means
that when controlling for age, education, and practice effect,
patients with MHE showed a worse performance compared to
controls, in both RTs and accuracy, even in a cognitive low
demanding task condition such as the Detect condition. This
pattern of results is in line with previous studies (Amodio et al.
2010; Cona et al. 2013) that described MHE-related alter-
ations in relatively low demanding conditions, that involve
basic processes, such as sustained and selective attention. On
the other hand, group differences in variability of RTs (RT
variance) emerge in more demanding task conditions (Go and
No-Go). In contrast, performance of patients with MHE in the
No-Go condition was not significantly different compared to
performance of healthy controls in terms of accuracy. How-
ever, it is important to cautiously discuss the lack of
differences.
In fact, as pointed out in Amodio et al. (2010), patients with
MHE did not correctly inhibit their response; rather they had
difficulties following the task, missing many responses, as
confirmed by the low accuracy rate in the Go condition.
Therefore, if a participant missed many responses, he or she
was more likely to perform correctly on No-Go trials (hence,
to not respond when it was required to not respond). In light of
this, the low number of errors in No-Go trials is a spurious
effect and suggests that a better evaluation of the cognitive

status of a patient with cirrhosis should be based on perfor-
mance in tasks that are not particularly complex but that
instead that involve simply basic processes, such as attentional
ones.
With respect to some of the previous investigations, the
present study has the advantage of separately considering
patients with cirrhosis having MHE and patients with cirrhosis
not having MHE. Indeed, patients with MHE showed signif-
icantly slower RTs and lower accuracy in the high demanding
condition (i.e., the Go condition) compared to patients without
MHE. The contrast analysis of RTs provides evidence of the
presence of a higher inter-individual variability in patients
with MHE compared to patients without MHE, in the Go
condition. In addition, patients with MHE showed a tendency
to have a higher error rate of inhibition in comparison to
patients without MHE and a higher intra-individual variability
in RTs for wrong responses. Taken together, these results
indicate that ICT performance can discriminate between pa-
tients with and without MHE (cf. Amodio et al. 2010; Bajaj
et al. 2008). In this regard, a new and interesting result comes
from the evaluation of the performance of patients without
MHE. Indeed, although patients without MHE showed sig-
nificant differences compared to patients with MHE in some
of the measures and conditions of ICT, nevertheless they
showed a performance in the Detect condition that was similar
to that of patients with MHE, both in terms of RTs and
accuracy. Moreover, patients without MHE showed a signif-
icant decline in accuracy compared to the control group in the
Detect condition (i.e., a cognitively low demanding condi-
tion). This is noteworthy because patients without MHE are
considered as not having any relevant cognitive dysfunction
on the basis of the classical psychometric and psychophysio-
logical tools. The present result is in line with a recent study
that showed that patients without MHE had electrophysiolog-
ical alterations similar to those of patients with MHE in only
the Detect condition (Cona et al. 2013; Schiff et al. 2013).
Therefore, these findings suggest the importance of including
a computerized task, which might mirror the Detect condition
of ICT, in order to identify even early cognitive changes.
Furthermore, patients without MHE had slower and more
variable RTs in the No-Go condition compared to healthy
individuals. Such findings suggest that differences in intra-
individual variability in RTs become evident when task con-
ditions require the involvement of higher cognitive resources.
Another finding that deserves particular consideration is
the influence of education in modulating the ICT perfor-
mance. Indeed, as the number of years of education increases,
the performance on the ICT task becomes better in the Detect
condition. In contrast, no differences in the accuracy of ICT
tasks were shown in the Go and No-Go condition as a function
of the education level. Our results indicate that a higher level
of education allowed participants to overcome their difficul-
ties in task performance, but only in low cognitive demanding
Metab Brain Dis
conditions. Such a compensation phenomenon is in line with
the cognitive reserve hypothesis (Stern 2009). According to
this hypothesis, life experiences might enhance knowledge
and allow individuals to cope for cognitive decline, which is
often exhibited by elderly individuals. Our results support and
extend such a hypothesis, showing that cognitive reserve acts
as a compensatory mechanism only when the task is relatively
easy, and becomes less effective when the task is more
demanding.
Collectively, these findings suggest that, when controlling
for variables that can vary across and within subject groups,
slight cognitive alterations are detectable even in patients with
cirrhosis who were considered as not having MHE from the
classical psychometric batteries. This leads to two main con-
siderations: first, cognitive tests with higher selectivity and
specificity are needed, in order to detect even slight alter-
ations; second, cognitive functioning of patients with cirrhosis
should be monitored from the first stages of the disease.
Acknowledgments This work was partially funded by University of
Padua CPDA108328 grant to PB and GC and by Bial Foundation Grant
No. 84/12 to P.B.
Declaration of interest The authors have declared that no competing
interests exist.
References
Amodio P, Marchetti P, Del Piccolo F, Rizzo C, Iemmolo RM, Caregaro
L, Gerunda G, Gatta A (1998) Study on the Sternberg paradigm in
cirrhotic patients without overt hepatic encephalopathy. Metab Brain
Dis 13:159–172
Amodio P, Del Piccolo F, Marchetti P, Angeli P, Iemmolo R, Caregaro L,
Merkel C, Gerunda G, Gatta A (1999) Clinical features and survivial
of cirrhotic patients with subclinical cognitive alterations detected
by the number connection test and computerized psychometric tests.
Hepatology 29:1662–1667
Amodio P, Montagnese S, Gatta A, Morgan MY (2004) Characteristics of
minimal hepatic encephalopathy. Metab Brain Dis 19:253–267
Amodio P, Biancardi A, Montagnese S, Angeli P, Iannizzi P, Cillo U,
D’Amico D, Gatta A (2007) Neurological complications after
orthotopic liver transplantation. Dig Liver Dis 39:740–747
Amodio P, Ridola L, Schiff S, Montagnese S, Pasquale C, Nardelli S,
Pentassuglio I, Trezza M, Marzano C, Flaiban C, Angeli P, Cona G,
Bisiacchi P, Gatta A, Riggio O (2010) Improving detection of
minimal hepatic encephalopathy using the inhibitory control task.
Gastroenterology 139:510–518
Bajaj JS, Hafeezullah M, Franco J, Varma RR, Hoffmann RG, Knox JF,
Hischke D, Hammeke TA, Pinkerton SD, Saeian K (2008)
Inhibitory control test for the diagnosis of minimal hepatic enceph-
alopathy. Gastroenterology 135:1591–1600
Burton CL, Strauss E, Hultsch DF, Moll A, Hunter MA (2006)
Intraindividual variability as a marker of neurological dysfunction:
a comparison of Alzheimer’s disease and Parkinson’s disease. J Clin
Exp Neuropsychol 28:67–83
Castellanos FX, Sonuga-Barke EJ, Milham MP, Tannock R (2006)
Characterizing cognition in ADHD: beyond executive dysfunction.
Trends Cogn Sci 10:117–123
Metab Brain Dis
Cona G, Arcara G, Amodio P, Schiff S, Bisiacchi PS (2013) Does
executive control really play a crucial role in explaining age related
cognitive and neural differences? Neuropsychology 27(3):378–389
Depp CA, Savla GN, de Dios LA, Mausbach BT, Palmer BW (2012)
Affective symptoms and intra-individual variability in the short-
term course of cognitive functioning in bipolar disorder. Psychol
Med 42:1409–1416
Dhiman RK, Chawla YK (2009) Minimal hepatic encephalopathy. Indian
J Gastroenterol 28:5–16
Efron B, Tibshirani R (1993) An introduction to the bootstrap. Chapman
and Hall, New York
Elsass P, Christensen SE, Mortensen EL, Vilstrup H (1985)
Discrimination between organic and hepatic encephalopathy by
means of continuous reaction times. Liver 5:29–34
Felipo V, Ordoño JF, Urios A, El Mlili N, Giménez-Garzó C, Aguado C,
González-Lopez O, Ferenci P, Lockwood A, Mullen K, Tarter R,
Weissenborn K, Blei AT (2012) Hepatic encephalopathy—defini-
tion, nomenclature, diagnosis, and quantification: final report of the
working party at the 11th World Congresses of Gastroenterology,
Vienna, 1998. Hepatology 35:716–721
Ferenci P, Zollner G, Trauner M (2002) Hepatic transport systems. J
Gastroenterol Hepatol. 2002 Feb;17 Suppl:S105–12
Fitzmaurice GM, Laird NM, Ware JH (2004) Applied longitudinal anal-
ysis. Wiley, New Jersey
Hultsch DF, Strauss E, Hunter MA, MacDonald SWS (2008)
Intraindividual variability, cognition, and aging. In: Craik FIM,
Salthouse TA (eds) The handbook of aging and cognition.
Psychology Press, New York, pp 491–556
Kaiser S, Roth A, Rentrop M, Friederich HC, Bender S, Weisbrod M
(2008) Intra-individual reaction time variability in schizophrenia,
depression and borderline personality disorder. Brain Cogn 66:73–82
MacDonald SWS, Hultsch DF, Bunce D (2006) Intraindividual variabil-
ity in vigilance performance: does degrading visual stimuli mimic
age-related neural noise? J Clin Exp Neuropsychol 28:655–675
MacDonald SW, Li SC, Bäckman L (2009) Neural underpinnings of
within-person variability in cognitive functioning. Psychol Aging
24:792–808
Rigby RA, Stasinopoulos DM (2005) Generalized additive models for
location, scale and shape. Appl Stat 54:507–554
Schiff S, Vallesi A, Mapelli D, Orsato R, Pellegrini A, Umiltà C, Gatta A,
Amodio P (2005) Impairment of response inhibition precedes
motor alteration in the early stage of liver cirrhosis: a be-
havioral and electrophysiological study. Metab Brain Dis 20:
381–392
Schiff S, Mapelli D, Vallesi A, Orsato R, Gatta A, Umiltà C, Amodio P
(2006) Top-down and bottom-up processes in the extrastriate cortex
of cirrhotic patients: an ERP study. Clin Neurophysiol 117:1728–
1736
Schiff S, D’Avanzo C, Cona G, Goljahani A, Montagnese S, Volpato C,
Gatta A, Sparacino G, Amodio P, Bisiacchi P (2013) Insight into the
relationship between brain/behavioural speed and variability in pa-
tients with minimal hepatic encephalopathy. Clin Neurophysiol. doi:
10.1016/j.clinph.2013.08.004
Schomerus H, Hamster W (1998) Neuropsychological aspects of portal-
systemic encephalopathy. Metab Brain Dis 13:361–377
Stern Y (2009) Cognitive reserve. Neuropsychologia 47:2015–2028
Stuss DT, Murphy KJ, Binns MA, Alexander MP (2003) Staying on the
job: the frontal lobes control individual performance variability.
Brain 126:2363–2380
Tarantino V, Cutini S, Mogentale C, Bisiacchi PS (2013) Time-on-task in
children with ADHD: an ex-Gaussian analysis. J Int Neuropsychol
Soc 19:1–9
Weissenborn K, Ennen JC, Schomerus H, Rückert N, Hecker H (2001)
Neuropsychological characterization of hepatic encephalopathy. J
Hepatol 34:768–773
Weissenborn K, Giewekemeyer K, Heidenreich S, Bokemeyer M,
Berding G, Ahl B (2005) Attention, memory, and cognitive function
in hepatic encephalopathy. Metab Brain Dis 20:359–367