You are on page 1of 7

Schizophrenia Research 130 (2011) 2733

Contents lists available at ScienceDirect

Schizophrenia Research
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / s c h r e s

Cholesterol and cognition in schizophrenia: A double-blind study of patients


randomized to clozapine, olanzapine and haloperidol
Menahem Krakowski a,b,, Pal Czobor a,c
a
b
c

Nathan Kline Institute for Psychiatric Research, 140 Old Orangeburg Road, Orangeburg, NY 10962, United States
New York University School of Medicine, 550 First Avenue, New York, NY 10016, United States
Department of Psychiatry and Psychotherapy, Semmelweis Medical University, Budapest, Hungary

a r t i c l e

i n f o

Article history:
Received 1 December 2010
Received in revised form 30 March 2011
Accepted 5 April 2011
Available online 6 May 2011
Keywords:
Cholesterol
Cognition
Clozapine
Olanzapine
Haloperidol

a b s t r a c t
Objective: A positive relationship between cholesterol levels and cognition has been reported in various
human and animal studies, but has never been investigated in schizophrenia. The goal of this study was to
examine this relationship in schizophrenic patients randomized to clozapine, olanzapine or haloperidol.
Method: This was a double-blind randomized prospective 12-week study. Participants received a baseline
evaluation including a cognitive battery consisting of an evaluation of psychomotor function, general
executive function, visual and verbal memory, and visuospatial ability. Their fasting serum cholesterol level
was also assessed. The participants were then randomized to clozapine, olanzapine, or haloperidol. They were
evaluated at the end of 12 weeks. A general cognitive index (GCI) derived from the cognitive battery was the
primary variable.
Results: 82 patients had both baseline and endpoint neurocognitive assessments and cholesterol levels. There
was a statistically and clinically signicant positive association between change in cholesterol levels and
change in GCI. This association was especially pronounced for verbal memory. There was no interaction
between medication grouping and cholesterol level; the positive association was observable separately in
each medication group. It was very robust and remained signicant after we controlled for glucose and
triglyceride levels, anticholinergic side effects, medication serum levels, cholesterol lowering medications,
and pre-study antipsychotic medications.
Conclusions: Cholesterol levels show a strong association with cognition in schizophrenia in all medication
groups. Further research on the role of lipid metabolism in cognition may suggest new treatments for this core
decit of schizophrenia.
2011 Elsevier B.V. All rights reserved.

1. Introduction
Cognitive impairment is a core feature of schizophrenic illness and
an important predictor of functional outcome (Green, 1996). Yet, little
progress has been made in treating it. A positive relationship between
cholesterol levels and various cognitive measures has been reported
in human and animal studies, but has not been investigated in
schizophrenia.
A cholesterol-rich diet improves performance on various learning
tasks, enhances memory acquisition and retention (Sparks and
Schreurs, 2003; Dufour et al., 2006), and facilitates classical
conditioning in diverse animal species (Schreurs et al., 2007).
Addition of cholesterol to the diet of animals decient in cholesterol
or animals that have cholesterol synthesis blocked reverses the

Corresponding author at: Nathan Kline Institute for Psychiatric Research, 140 Old
Orangeburg Road, Orangeburg, NY 10962, United States. Tel.: + 1 845 398 6542; fax: + 1
845 398 6545.
E-mail address: Krakow@nki.rfmh.org (M. Krakowski).
0920-9964/$ see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.schres.2011.04.005

learning and memory decits (Endo et al., 1996; O'Brien et al., 2002;
Voikar et al., 2002).
In humans there are both observational and randomized treatment
studies which indicate that total cholesterol level (TC) correlates
positively with cognitive abilities. There was a signicant positive
linear association between TC and a composite score measuring
multiple cognitive domains (Elias et al., 2005). Lower TC levels were
associated with poorer performance on choice reaction time task
(Benton, 1997), block design (Muldoon et al., 2001), attention/
concentration, executive function tests (Elias et al., 2005), and
visuomotor tasks (Zhang et al., 2004). Cholesterol-lowering interventions, such as treatment with statins (Muldoon et al., 2000) or diet
(Wardle et al., 2000), have been associated with poorer performance
on neurocognitive measures.
Though the blood brain barrier limits cholesterol inux from the
blood into the brain (Hibbeln et al., 2000; Bjorkhem and Meaney,
2004), changes in cholesterol in the diet or blood result in changes in
brain cholesterol content (Schoknecht et al., 1994; Sparks et al., 1995),
and are accompanied by changes in brain activity (Schoknecht et al.,
1994; Dufour et al., 2006). Some cholesterol metabolites, such as 27-

28

M. Krakowski, P. Czobor / Schizophrenia Research 130 (2011) 2733

hydroxycholesterol (Heverin et al., 2005) as well as neurosteroids,


which use cholesterol as a precursor (Compagnone and Mellon,
2000), readily cross the blood brain barrier and modulate neurotransmitter receptor activity (Tall et al., 2002).
Some authors suggest that the association between cognition and
cholesterol can be explained by the fact that serum cholesterol
concentration correlates with polyunsaturated fatty acids (PUFAs)
(Hibbeln et al., 2000), and thus serves as surrogate marker for PUFA's.
These are involved in neurotransmitter synthesis and neuronal
membrane composition. Patients with schizophrenia have decreased
levels of PUFA's (Khan et al., 2002; Yao, 2003), and it has been
hypothesized that this factor plays an important role in the illness
(Skosnik and Yao, 2003).
Plasma cholesterol can modify brain membrane fatty acid
composition and alter its biophysical properties, and thus affect the
activity of membrane neurotransmitter receptors (Burger et al.,
2000). Antipsychotic medications also modify the lipid composition
of the neuronal membrane, including membrane cholesterol and
PUFA (Horrobin et al., 1994). Thus there can be an interaction
between the effects of cholesterol and those of antipsychotic
medications on neurotransmission.
We investigate here the relationship between change in serum
cholesterol levels and change in neurocognition prospectively during
treatment with antipsychotic medication. In addition, we examine
whether this association is affected by the specic medication
assignment.
2. Material and methods
2.1. Study design
This study is a randomized, double-blind, 12-week clinical trial of
clozapine, olanzapine, and haloperidol in patients with DSM-IV
schizophrenia or schizoaffective disorder. It was part of a larger
investigation of differences among antipsychotic medications; a
detailed description of the procedures of this study and the study
owchart has already been reported (Krakowski et al., 2006).
After complete description of the study to the subjects, written
informed consent was obtained, as approved by the local Institutional
Review Boards and compliant with the Declaration of Helsinki.
Patients received baseline evaluations and were then randomized to
clozapine, olanzapine, or haloperidol using a block randomization
scheme (block size = three, no baseline stratication). The study was
divided into two periods: study weeks 16 comprised the escalation/
xed-dose period, during which the doses of study medication were
escalated to target levels (clozapine 500 mg/day, olanzapine and
haloperidol 20 mg/day), at which they remained xed until the end of
week 6.
Weeks 712 comprised a variable dose period, during which
psychiatrists, blind to treatment group assignment, could change
doses by prescribing various "levels" of medication. Permitted dose
ranges were: clozapine, 200800 mg/day; olanzapine and haloperidol, 1030 mg/day. At the end of the variable-dose period, the average
dose (mg per day) was 565.5 for clozapine (SD = 112.7 mg/day), 24.7
for olanzapine (SD = 6.1 mg/day), and 23.3 for haloperidol
(SD = 7.1 mg/day).
Throughout the study, all patients received (double-blind)
benztropine, benztropine placebo or both. Benztropine (4 mg/day)
was administered prophylactically to patients receiving haloperidol.
Patients receiving mood stabilizers or antidepressants prior to entry
into the study continued receiving these medications.
2.2. Subjects
Consenting inpatients 1860 years of age were eligible for the
study if they met criteria for DSM-IV diagnosis of schizophrenia, or

schizoaffective disorder. Patients with a history of nonresponse/


intolerance to any of the three medications were excluded, as were
patients who received depot antipsychotics within 30 days of
randomization.
2.3. Variables and measurement instruments
Screening evaluations included a diagnostic interview, the Structured Clinical Interview for DSM-IV (SCID), physical examination,
measurement of vital signs, and laboratory tests. Raters blind to
treatment group performed all clinical research assessments. Study
procedures were identical for all three groups to conceal medication
assignment.
2.3.1. Neurocognitive assessments
The neurocognitive battery was designed to examine functional
domains considered important by virtue of impairment in schizophrenia. It included Simple Motor Function (right and left hand scores
from the Finger Tapping and Purdue Pegboard Tests), Executive
Function (Wisconsin Card Sorting Test categories completed, perseverative errors and non-perseverative errors; Trail Making Test part B),
Verbal Memory (WMS-R Logical Memory, Immediate and Delayed)
and Visual Memory (WMS Figural Memory, Immediate and Delayed).
In addition, the WAIS-R Block Design and the Trail Making Test part A
were also administered.
In addition to the above tests, the Mini Mental State Exam (MMSE)
(Folstein et al., 1975) is used here as an ancillary measure of cognitive
function. It includes orientation (to time and place), immediate recall,
short-term memory, attention and calculation (serial sevens), ability
to follow verbal and written instructions, and visual constructional
ability. While it does not assess decits across cognitive domains, it
provides a functionally relevant index of cognitive impairment across
diagnoses, including schizophrenia (Chino et al., 2006).
Cognition was assessed before randomization and at the end of the
12-week trial. In case of premature termination between weeks 4 and
12, endpoint neuropsychological assessments were performed. To
minimize sedation effects, assessments were postponed until at least
24 h after the last dose of any treatment for agitation.
2.3.2. Other clinical assessments
The Positive and Negative Syndrome Scale (PANSS) (Kay et al.,
1987) was used to assess clinical symptoms by two independent
raters, at baseline, week 6, and week 12 (or endpoint). Inter-rater
reliability, estimated by ICC, exceeded 0.90. Safety measures included
vital signs, EKG's, physical exam and WBC tests. The Extrapyramidal
Symptom Rating Scale (ESRS) (Chouinard et al., 1980) was administered weekly. In addition to the regular subscales, we selected the
items assessing peripheral anticholinergic side effects (ASE's) as
providing information on anticholinergic activity, which inuences
cognition. In vitro indices of anticholinergic activity are highly
correlated with ASE's (Minzenberg et al., 2004).
2.3.3. Assessment of serum metabolites
Fasting blood samples for cholesterol were collected at baseline
and endpoint. We also assessed serum glucose and serum triglycerides. Only results of blood samples collected at morning fasting
times were included in the analyses. Plasma cholesterol, glucose, and
triglyceride levels were determined by enzymatic procedures applying the Boehringer Mannheim/Hitachi 714 automated chemistry
analyzer and using the standard analytical system packs Cholesterol/
HP, Glucose/HK, and Triglycerides GPO-TRINDER, Sigma.
Blood samples for the determination of the antipsychotic plasma
levels were drawn at baseline and at weeks 6, 8, and 1012. Samples
were drawn approximately 12 h after the last dose of medication.
Clozapine assay used a modication of a published method (Simpson
and Cooper, 1978). Olanzapine assay was a modied validated liquid

M. Krakowski, P. Czobor / Schizophrenia Research 130 (2011) 2733

chromatographic procedure with electrochemical detection (Catlow


et al., 1995). Haloperidol assay was based on a published method
(Bianchetti and Morselli, 1978).
2.4. Statistical analysis
Our objective was to investigate the relationship between change
in total cholesterol (TC) levels and change in cognition and to see
whether the same relationship exists in each medication group. The
General Cognitive Index (GCI) was used as the primary dependent
variable, as it is a global composite measure of important neurocognitive domains. We tested this relationship with the GCI in the main
analysis. In secondary analyses we examined the relationship with the
different neurocognitive domains which make up the GCI, in order to
clarify which of these domains contribute most to the general
relationship with cholesterol. We also investigated the ancillary
cognitive measure, the MMSE. TC change and medication grouping
(i.e., clozapine, olanzapine and haloperidol) were therefore used as
independent variables in the analyses, and we included also the
interaction between these two variables.
The GCI reects the equally weighted mean of all scores on the
different neurocognitive domains. The GCI and neurocognitive
domain scores are based on z-transformed values of the constituting
variables. The MMSE was also expressed in standard deviation units,
so that the slope estimate would provide a measure of the effect size.
The main independent variable was the change in TC between
baseline and endpoint.
We used a Generalized Linear Model (GLM) with last observation
carried forward analysis. Baseline total cholesterol (TC), baseline
cognitive measures, duration of stay in the study, and age were
introduced as covariates. We also looked at the effect of many other
covariates which included various demographic, treatment and
clinical variables, including gender, schizophrenic vs. schizoaffective
disorder subdiagnosis, ethnicity, anticholinergic side effects, serum
blood levels of treatment medications, cholesterol lowering medications, and antipsychotic medications used prior to entry in the study.
The latter were used as dichotomous covariates (presence/absence).
We also looked at the effect of psychiatric symptoms (PANSS Total
and subscale scores) and side effects (ESRS total and subscale scores).
As there is a complex interplay between various metabolic changes
that may impact on cognition, we studied more extensively the
inuence of weight, triglyceride and glucose levels. We looked at the
effect of baseline values and change in values over treatment for
weight, serum triglyceride and serum glucose levels on the relationship between change in cholesterol and change in GCI. We looked also
at the effect of the change in serum triglyceride and in glucose levels
on the main cognitive variable, the GCI, independently of the change
in cholesterol level, as well as interaction with cholesterol.
The relationship between the change in the GCI and the change in
TC level over the study is expressed as a regression slope. Specically,
we investigated how much of a change there is in the GCI for a change
in the range of one standard deviation (i.e., 70 mg/dl) in TC. Thus
the slope estimate provides a measure of the effect size.
We hypothesized for this specic investigation that increases in
serum cholesterol levels would be associated with improvements in
cognition and conversely, decreases in cholesterol would be associated
with deterioration in cognition.
3. Results
3.1. Sample characteristics and baseline/background information
Of the 110 patients who entered the study, 101 had baseline
cognitive assessments and cholesterol levels. Of these, 82 had both a
subsequent cognitive assessment and cholesterol level. These 82
patients, 30 each in the clozapine and olanzapine groups and 22 in the

29

haloperidol group, constitute our study sample. Their demographic


and clinical characteristics are presented in Table 1. There were no
signicant differences among the three treatment groups in demographic characteristics, illness parameters or duration of time in the
study, in baseline neurocognitive tests, psychopathology scores or
total cholesterol levels.
There were no differences in the proportion of patients receiving
other psychotropic mediations as concomitant medications. These
included antidepressants, antiepileptic medications used for mood
control and for control of impulsive behavior, as well as lithium.
With regard to the antipsychotic medications used prior to entry in
the study, 33%, 28% and 20% of the patients were on olanzapine,
risperidone and quetiapine respectively. There were no differences in
the proportion of patients who were on these medications or any
other pre-study antipsychotic agent.
Comparisons of the 82 patients and the original sample of 110 did
not reveal signicant differences on any of the demographic or clinical
variables, including baseline scores on the cognitive tests, PANSS, and
cholesterol levels. There were also no differences between these 82
patients and the 28 patients who were not included in the study.
3.2. Descriptive information about cholesterol and cognition
The baseline values for TC and for GCI for each of the three
medication groups are provided in Table 1; the changes in these
variables are provided in Table 2. The change in TC was not signicant
in the total group of subjects (mean = .45, SD = 33.9, t = .12, df = 81,
p = .9), or in each medication group (See Table 2). There was,
however, a signicant overall difference among the 3 groups

Table 1
Baseline characteristics of patients assigned to receive clozapine, olanzapine and
haloperidola.
Characteristics
Categorical variables N (%)
Demographic
Male, no. (%)
Race/ethnicity, no. (%)
White
African American
Hispanic
Other
Diagnosis, no (%)
Schizophrenia
Schizoaffective disorder
Median length of
hospitalizationb

Clozapine
(N = 30)

Olanzapine
(N = 30)

Haloperidol
(N = 22)

25 (83.3)

24 (80.0)

19 (86.4) 0.4

7 (23.3)
14 (46.7)
7 (23.3)
2 (6.7)

4 (13.3)
22 (73.3)
4 (13.3)
0 (0.0)

3 (13.6)
13 (59.1) 7.7
6 (27.3)
0 (0.0)

21 (70.0)
9 (30.0)
45

18 (60.0)
12 (40.0)
48

15 (68.2) 0.7
7 (31.8)
36
1.2

Continuous variables: mean (standard deviation)


Mean age at
randomization, y
Mean duration of illness, y
Prior psychiatric
hospitalizations
Total cholesterol
level mg/dl
Global Cognitive Indexc
Motor function
Executive function
Verbal memory
Visual memory
MMSE
PANSS total

0.83

0.46

0.69
0.54

34.6 (11.3)

34.6 (8.9)

32.1 (10.1) 0.5

0.62

14.9 (8.7)
11.8 (9.4)

16.3 (10.7)
12.0 (10.2)

14.1 (10.0) 0.3


9.4 (4.8)
0.7

0.74
0.53

186.6 (39.6) 191.9 (52.4) 174.8 (42.9) 0.91 0.41


.01 (.54)
.00 (.67)
.07 (.64)
.07 (1.08)
.11 (1.06)
25.0 (2.9)
87.3 (15.0)

.09
.06
.03
.26
.40
25.5
81.0

(.57)
(.68)
(.57)
(.78)
(.87)
(3.5)
(13.7)

.06
.11
.09
.16
.08
25.8
85.2

(.52)
(.51)
(.73)
(1.04)
(.97)
(2.8)
(12.4)

0.25
0.46
0.41
1.21
2.42
0.4
1.59

0.78
0.63
0.66
.30
.10
.70
0.21

a
For categorical variables data are presented as relative frequencies and for
continuous variables means and SD's are provided.
b
Median length of hospitalization in days upon study entry.
c
The Global Cognitive Index and the neurocognitive domain scores are based on ztransformed values of the constituting variables. The Global Cognitive Index reects the
equally weighted mean of all scores on the different neurocognitive domains.

30

M. Krakowski, P. Czobor / Schizophrenia Research 130 (2011) 2733

Table 2
Changes in total cholesterol levels and changes in cognitive function over treatment in patients assigned to the olanzapine, clozapine or haloperidol groupsa.
Clozapine N = 30

Olanzapine N = 30

Haloperidol N = 22

Change in:

Mean

SD

Mean

SD

Mean

SD

Total cholesterol level


Global Cognitive Indexb

7.43
.10

38.2
.39

1.07
1.41

.30
.17

2.13
.34

34.7
.72

.34
2.61

.74
.01

5.55
.17

25.0
.49

1.04
1.64

.31
.12

a
The values in the table are based on paired t-tests for the change between baseline and endpoint values. Positive signs for the total cholesterol levels indicate increases and
positive signs for the Global Cognitive Index indicate improvement.
b
The Global Cognitive Index reects the equally weighted mean of all scores on the different neurocognitive domains, based on z-transformed values of the constituting variables.

(ANCOVA; F = 4.32, df = 2, 81; p = .02). Specically, this was due to a


pairwise difference (p b .004) between clozapine patients, whose
cholesterol levels increased, and haloperidol patients, whose levels
decreased.
The overall change in the General Cognitive Index (GCI) over the
study for the total group was not signicant (Mean = .04; SD = .60;
t = 0.65, df = 81, p = .52). When we look at each medication group
separately, we nd a signicant change only for the olanzapine group
(Table 2). When we compared the groups, we found a signicant
overall difference (F = 6.63, df = 2, 81, p = .002), in line with results
obtained in the total group of subjects (Krakowski et al., 2008).
Improvement in GCI was greater with olanzapine than with clozapine
(p = .002) or haloperidol (p = .006), but there was no difference
between the latter two groups.

3.3. Relationship between cholesterol and cognition in the total sample


There was a signicant main effect for the change in total
cholesterol (TC) level on the GCI (F = 13.3, df = 1, 81, p b .001),
indicating a positive relationship between these two variables across
all three groups. The estimate of the slope, representing the
association between TC change over one standard deviation and
GCI change was .57 (SE = .16; t = 3.64, df = 81, p = 0.0005). Since the
GCI was expressed in standard deviation units, the above slope
estimate indicates a medium-to-high effect size (Cohen, 1988).
Several cognitive domains contributed, to some extent, to this
overall effect represented by the GCI, but only two reached statistical
signicance; the slope estimate was .65 (SE = .23; t = 2.82, df = 81,
p = 0.006) for verbal memory, and .35 (SE = .18, t = 1.98, df = 81,
p = .05) for simple motor function.
There was also a signicant association between TC change and
change in the MMSE total score (F = 8.28, df = 1,81; p = .005). The
standardized slope estimate representing this association is .82
(SE = .83; t = 2.88, df = 81, p = 0.005). This indicates a large effect
size (Cohen, 1988). In order to provide an indication of the actual
clinical effect, we looked at the MMSE raw scores. A decrease of one
standard deviation in total cholesterol resulted in a decrease of 2.44
points (SE = .78, p = .003) in total score.
We introduced multiple covariates in the above analyses in order
to control for possible confounding effects. These included measures
of psychopathology, side effects, use of other medications and other
clinical and demographic variables. When the PANSS Total score was
used as a covariate, the relationship between TC and GCI remained
highly signicant (F = 11.8, df = 1, 81; p b .001). This was also the case
when the three PANSS subscales were used as covariates. The
relationship remained also highly signicant when the ESRS Total
score (F = 14.0, df = 1, 81; p b .001) or the ESRS subscales were used as
covariates.
We examined the effect of pre-study medication on the relationship between TC and GCI, by including olanzapine, risperidone and
quetiapine, the three most commonly used pre-study medications, as
additional dichotomous (yes or no) covariates in the analyses. The
previous medication status did not change the signicance of the
results.

Our hypothesis was focused on total cholesterol level, but we


provide here additional information about low-density lipoprotein
(LDL) and high-density lipoprotein (HDL) cholesterol. We examined
the relationship between change in LDL cholesterol and GCI change, as
well as change in HDL cholesterol and GCI change in secondary
analyses. There was a signicant positive relationship between
change in LDL and change in GCI (F = 9.0, df = 1, 81; p = .004), but
the interaction between LDL change and medication grouping was not
signicant. Thus, the results for LDL cholesterol parallel the above
results for TC. This was not the case for HDL cholesterol; there was no
signicant relationship between change in serum HDL levels and
change in GCI (F = 0.8, df = 1, 81, p = .39). There was also no
signicant interaction between HDL change and medication group.
3.4. Relationship between total cholesterol and cognition in the three
medication groups
The positive relationship between change in TC and change in GCI
is not limited to the total group. It is present in each medication group,
with a signicant slope estimate in each; it was .38 (SE = .19; t = 1.96;
df = 81, p b .05) for clozapine, which represents a low to medium
effect size; it was .50 (SE = .22; t = 2.25, df = 81, p = .03) for
olanzapine, which is a medium effect size, and .85 (SE = .32;
t = 2.67; df = 81, p b .01) for haloperidol which is a large effect size
(Cohen, 1988).
Despite these differences in effect size, the interaction term
between medication grouping and change in cholesterol did not
reach statistical signicance in determining the change in GCI (p N .1).
Similarly, the interaction term between these two variables was not
signicant for the change in the MMSE (p N .1).
3.5. Triglycerides, glucose and weight
As various metabolic changes may impact on the relationship
between cholesterol levels and cognition, we looked at the possible
contribution of weight, triglyceride and glucose levels to cognition
and to the relationship between cholesterol and cognition. The
baseline and endpoint values for these variables for all the study
subjects are presented in Table 3.
We looked at the effect of the change in serum triglyceride and
change in serum glucose on the main cognitive variable, the GCI,

Table 3
Baseline and endpoint values for weight, triglycerides and glucose in the 82 subjects of
the studya.

Weight (kgs)
Triglyceride (mg/dl)
Glucose (mg/dl)

Baseline values

Endpoint values

Mean

SD

Mean

SD

91.3
160.9
92.7

17.6
92.1
24.9

93.5
177.0
99.5

19.1
101.4
42.0

2.97
1.88
1.57

.004
.06
.12

a
The table provides information about the observed mean values at baseline and at
endpoint for body weight, serum triglycerides and serum glucose in the total group of
subjects. Standard deviations and paired t-test values are also provided in the table.

M. Krakowski, P. Czobor / Schizophrenia Research 130 (2011) 2733

independently of the change in cholesterol level. The change in


triglyceride level was positively related to the change in the Global
Cognitive Index (F = 5.1, df = 1, 81; p = .03). However, when we
included the change in cholesterol level as one of the main
independent variables, the change in triglyceride level was no longer
signicant (F = .01, df = 1, 81, p = .94), while the change in
cholesterol was still highly signicant (p b .001). The interaction
between change in cholesterol and change in triglyceride was not
signicant (F = .18, df = 1, 81, p = .67). Thus the change in triglyceride did not provide any contribution to the GCI independently of the
change in cholesterol.
There were no signicant results for the change in glucose level. It
was not signicantly related to the change in the Global Cognitive
Index when it was investigated independently of the change in
cholesterol level (F = 0.4 df = 1, 81; p = .53), or when the change in
cholesterol level was included (F = 0.13, df = 1, 81; p = .72). The
interaction between change in cholesterol and change in glucose
levels was not signicant (F = 1.05, df = 1, 81; p = .31).
In addition to the above analyses we introduced baseline weight,
serum triglyceride and serum glucose levels, as well as change in
weight over the study, as covariates in our basic analyses. The
relationship between change in cholesterol and change in GCI
remained highly signicant.
3.6. Anticholinergic side effects
We wanted to see if the anticholinergic side effects (ASE's) played
a role in the relationship between cholesterol levels and cognition, as
they have been associated with cognitive impairment. We investigated rst the differences in the severity of peripheral ASE's among
the medications (as indicated by endpoint ASE items on the ESRS).
There was a signicant difference among the 3 groups (F = 4.98,
df = 2, 81 p b .01). Though the patients in the haloperidol group
received benztropine, an anticholinergic agent, it was the clozapine
group that presented with most severe ASE's (N = 29, Mean = 3.17;
SD = 1.65). They were least pronounced with olanzapine (N = 30,
Mean = 1.97, SD = 1.45), and of intermediate severity with haloperidol (N = 22, Mean = 2.68, SD = 1.25). There was a signicant
pairwise difference between clozapine and olanzapine (p = .002)
and a marginal difference between haloperidol and olanzapine
(p = .09).
We then investigated the effect of these anticholinergic side effects
on the relationship between GCI change and TC change by introducing
ASE change and baseline ASE as covariates. The relationship remained
unchanged in the total group (F = 13.7, df = 1,81; p b .001), and in
each of the 3 groups, as indicated by the regression slope (p b .05 for
clozapine and olanzapine and p b .01 for haloperidol).
3.7. Medication serum levels
In post-hoc analyses we checked whether the serum levels of the
study medications inuenced the relationship between cholesterol
and cognition in the total group, and separately in each medication
group. For the total group, since these levels are not directly
comparable across medications, we used a median split for the
medication serum levels. The median serum level at endpoint was
412.5 ng/ml, 32 ng/ml, and 10.8 ng/ml for clozapine, olanzapine and
haloperidol respectively. We introduced these levels as a dichotomous covariate (low/high) in the above analyses. The association
between change in GCI and change in TC remained highly signicant
(F = 12.51, df = 1,81; p b .001).
We also looked at the effect of the actual serum levels, as
continuous covariates separately in each medication group. The mean
serum level was 446.4 ng/ml, 38.0 ng/ml and 11.9 ng/ml, for clozapine, olanzapine and haloperidol respectively. The association, as

31

indicated by the regression slope, was signicant in each group


(p b .05 for clozapine and olanzapine and p b .01 for haloperidol).
3.8. Cholesterol lowering medications
We looked whether the use of cholesterol lowering medications
affected the relationship between cholesterol and cognition. Eleven
patients received statins during double-blind treatment. In order to
investigate their effect, we introduced the use of statins as a
dichotomous covariate (Yes/No) in the above analyses. The association between change in GCI and change in TC remained highly
signicant (F = 13.55, df = 1,81; p b .001). The association, as indicated by the regression slope, was signicant in each medication group
(p b .05 for clozapine and olanzapine and p b .01 for haloperidol). In
addition to the above analysis, we excluded the 11 patients who
received statins and repeated the analyses. Once again, the results
were unchanged (F = 13.52, df = 1,70; p b .001).
4. Discussion
Our study investigates the relationship between cholesterol levels
and cognition in schizophrenia. There is a positive longitudinal
association between change in cholesterol and change in global
cognition, as measured by the Global Cognitive Index, in the total
group of subjects, and in each of the three medication groups. Several
cognitive domains contribute to this relationship, but the most
prominent one is verbal memory. The strong positive relationship
between cholesterol levels and cognition is also evidenced on the
MMSE.
While in the current study, the overall change in cholesterol was
not clinically substantial at the group level in any of the three
medication groups, there was a clinically robust association between
cholesterol and cognition, as indicated by the magnitude of the
regression slope. In each group, across individuals, cognition
improved with an increase in cholesterol and deteriorated with a
decrease.
The different propensity of the medications to produce anticholinergic side effects is very unlikely to have confounded our results, as
the association between cholesterol and cognition was present in each
medication group separately. Furthermore, the anticholinergic side
effects do not seem to have inuenced this association, as it remained
essentially unchanged in the total group and in each medication group
separately when we used these side effects as covariates in the
analyses.
Changes in cholesterol levels retain their strong association with
changes in cognition when we introduce other metabolic factors, such
as changes in triglyceride and glucose levels, or changes in weight.
While a change in triglyceride level was related to cognition when
considered independently, the relationship disappeared totally when
change in cholesterol was introduced in the analyses. This association
between cholesterol and cognition is quite robust and is not affected
by an extensive list of other possible confounders, including
medication side effects, psychiatric symptoms, serum levels of the
study medications, the use of cholesterol lowering drugs, pre-study
antipsychotic medications, and various other clinical and demographic
variables. In addition, all the patients were on the same ward and
received a similar diet; they were all encouraged to participate in
physical exercises on a regular basis.
This positive association between change in cholesterol and change
in cognition appears to be more pronounced in subjects with a diagnosis
of schizophrenia than in the general population, as the effect size in our
study was moderate to large, whereas in the general population it is
mild to moderate (Elias et al., 2005). Our results may be interpreted in
line with the animal and human literature reviewed above which
suggests that cholesterol or some associated marker, such as PUFAs, has
an important effect on cognition.

32

M. Krakowski, P. Czobor / Schizophrenia Research 130 (2011) 2733

Cholesterol may inuence serotonergic and cholinergic neurotransmission. A reduction in serum cholesterol may result in decreased brain
serotonergic activity, as cholesterol concentration determines availability of serotonin receptors and transporters (Engelberg, 1992). As noted
above, cholesterol may be a marker for PUFAs which are also linked to
serotonergic and cholinergic function (Hibbeln et al., 2000). Other
factors should also be considered, as a complex regulatory network is
implicated in metabolic changes in patients with schizophrenia. Leptins,
for example, may also play a role. Studies have shown that leptin
enhances cognition (Li et al., 2002; Farr et al., 2006) and that atypical
antipsychotic medications, especially clozapine and olanzapine, increase leptin blood levels (Kraus et al., 1999).
The positive association between cholesterol and cognition does
not necessarily mean that cholesterol exerts a procognitive effect.
Other explanations for this association should be considered. Improvement in performance on cognitive tests may represent a practice
(learning) effect. Increases in cholesterol may allow patients to prot
more from previous practice, as has been pointed out in the literature
with regard to cognitive improvements that occur with antipsychotic
treatment (Harvey et al., 2000; Harvey and Keefe, 2001).
There are limitations in the generalizability of our results to other
schizophrenic patients in that the length of hospitalization in our
study is longer than that reported for non-state facilities; these
patients may represent a more chronic population. There was also a
high percentage of AfricanAmerican patients. In addition, our results
were based on only two time points; future studies may investigate
more extensively the time course of this association. In the future, the
use of standardized batteries, such as the MATRICS, might also
facilitate comparisons across studies.
5. Conclusions
We found a positive longitudinal association between change in
cholesterol and change in cognition in each of three medication
groups. These results were robust, as they remained highly signicant
when we controlled for many different confounders. A more extensive
investigation of lipid metabolism and its role in cognition may open
the way for new treatments of cognitive decits in schizophrenia.
Role of funding source
The project was supported by NIMH grant MH74767 and MH85322. The National
Institute of Mental Health had no role in the study design, data collection, analysis and
interpretation; the writing of the report and in the decision to submit the article for
publication.
Contributors
Dr. Krakowski was responsible for the design of the study, data collection and
analyses, and manuscript preparation. Dr. Czobor helped with the design of the study
and the interpretation of the results; he also conducted the statistical analyses. All
authors contributed to and have approved the nal manuscript.
Conict of interest
The authors have no conict of interest.

Acknowledgements
The authors thank Linda Kline, R.N., M.S., C.S., the chief coordinator of the project,
the CREF psychiatrists, Dr. Leslie Citrome, Dr. Biman Roy, Dr. Angel Cienfuegos, Dr.
William Greenberg, Dr. Fabien Tremeau and Dr. Narenda Patel, the CREF internist, Dr.
Surgit Dhami; Mike Hill and Henry Epstein, the CREF nursing staff and the Nathan Kline
Institute research staff: Melissa Benedict, Elsie Andrade, Faye Greenebaum, Gabriel
Goldfedder, Yakov Frances, Susan Grencer, Michael Radosta, Lorraine O'Donnell and the
Nathan Kline Institute research nurses: Sathamma Vaidian and Eunide Joseph. They
thank also Dr. Joseph Battaglia, Stephen Tomor and Stuart Moss.

References
Benton, D., 1997. Dietary fat and cognitive functioning. In: Hillbrand, M., Spitz, R.T.
(Eds.), Lipids, Health, and Behavior. American Psychological Association,
Washington, DC, pp. 227243.

Bianchetti, G., Morselli, P.L., 1978. Rapid and sensitive method for determination of
haloperidol in human samples using nitrogenphosphorus selective detection.
Journal of Chromatography 153, 203209.
Bjorkhem, I., Meaney, S., 2004. Brain cholesterol: long secret life behind a barrier.
Arteriosclerosis, Thrombosis, and Vascular Biology 24, 111.
Burger, K., Gimpl, G., Fahrenholz, F., 2000. Regulation of receptor function by
cholesterol. Cellular and Molecular Life Sciences 57, 15771592.
Catlow, J.T., Barton, R.D., Clemens, M., Gillespie, T.A., Goodwin, M., Swanson, S.P., 1995.
Analysis of olanzapine in human plasma utilizing reversed-phase high-performance liquid chromatography with electrochemical detection. Journal of Chromatography B, Biomedical Applications 668, 8590.
Chino, B., Mizuno, M., Nemoto, T., Yamashita, C., Kashima, H., 2006. Relation between
social functioning and neurocognitive test results using the optional thinking test
in schizophrenia. Psychiatry and Clinical Neurosciences 60, 6369.
Chouinard, G., Ross-Chouinard, A., Annable, L., Jones, B.D., 1980. Extrapyramidal
symptom rating scale. The Canadian Journal of Neurological Sciences 7, 233.
Cohen, J., 1988. Statistical Power Analysis for the Behavioral Sciences, 2nd edition.
Lawrence Erlbaum Assoc. Publishers, Hillsdale NJ.
Compagnone, N.A., Mellon, S.H., 2000. Neurosteroids: biosynthesis and function of
these novel neuromodulators. Frontiers in Neuroendocrinology 21, 156.
Dufour, F., Liu, Q., Gusev, P., Alkon, D., Atzori, M., 2006. Cholesterol-enriched diet affects spatial
learning and synaptic function in hippocampal synapses. Brain Research 1103, 8898.
Elias, P.K., Elias, M.F., D'Agostino, R.B., Sullivan, L.M., Wolf, P.A., 2005. Serum cholesterol
and cognitive performance in the Framingham Heart Study. Psychosomatic
Medicine 67, 2430.
Endo, Y., Nishimura, J.I., Kimura, F., 1996. Impairment of maze learning in rats following
long-term glucocorticoid treatments. Neuroscience Letters 203, 199202.
Engelberg, H., 1992. Low serum cholesterol and suicide. Lancet 339, 727729.
Farr, S.A., Banks, W.A., Morley, J.E., 2006. Effects of leptin on memory processing.
Peptides 27, 14201425.
Folstein, M.F., Folstein, S.E., McHugh, P.R., 1975. Mini-mental state: a practical method
for grading the cognitive state of patients for the clinician. Journal of Psychiatric
Research 12, 189198.
Green, M.F., 1996. What are the functional consequences of neurocognitive decits in
schizophrenia? The American Journal of Psychiatry 153, 321330.
Harvey, P.D., Keefe, R.S.E., 2001. Studies of cognitive change in patients with
schizophrenia following novel antipsychotic treatment. The American Journal of
Psychiatry 158, 176184.
Harvey, P.D., Moriarty, P.J., Serper, M.R., 2000. Practice-related improvements in
information processing with novel antipsychotic treatment. Schizophrenia Research 46, 139148.
Heverin, M., Meaney, S., Lutjohann, D., Diczfalusy, U., Wahren, J., Bjorkhem, I., 2005.
Crossing the barrier: net ux of 27-hydroxycholesterol into the human brain.
Journal of Lipid Research 46, 10471052.
Hibbeln, J.R., Umhau, J.C., George, D.T., Shoaf, S.E., Linnoila, M., Salem, N., 2000. Plasma
total cholesterol do not predict CSF neurotransmitter metabolites. The American
Journal of Clinical Nutrition 71 (S), 331S338S.
Horrobin, D.F., Glen, A.I., Vaddadi, K., 1994. The membrane hypothesis of schizophrenia.
Schizophrenia Research 13, 195207.
Kay, S.R., Fiszbein, A., Opler, L.A., 1987. The positive and negative syndrome scale
(PANSS) for schizophrenia. Schizophrenia Bulletin 13, 261276.
Khan, M.M., Evans, D.R., Gunna, V., Scheffer, R.E., Parikh, V.V., Mahadik, S.P., 2002.
Reduced erythrocyte membrane essential fatty acids and increased lipid peroxides
in schizophrenia at the never-medicated rst-episode of psychosis and after years
of treatment with antipsychotics. Schizophrenia Research 58, 110.
Krakowski, M.I., Czobor, P., Citrome, L., Bark, N., Cooper, T.B., 2006. Atypical
antipsychotic agents in the treatment of violent patients with schizophrenia and
schizoaffective disorder. Archives of General Psychiatry 63, 622629.
Krakowski, M.I., Czobor, P., Nolan, K., 2008. Atypical antipsychotics, neurocognitive
decits, and aggression in schizophrenic patients. Journal of Clinical Psychopharmacology 28, 485493.
Kraus, T., Haack, M., Schuld, A., Hinze-Selch, D., Khn, M., Uhr, M., Pollmcher, T., 1999.
Body weight and leptin plasma levels during treatment with atypical antipsychotic
drugs. The American Journal of Psychiatry 156, 312314.
Li, X.L., Aou, S., Oomura, Y., Hori, N., Fukunaga, K., Hori, T., 2002. Impairment of long
term potentiation and spatial memory in leptin receptor-decient rodents.
Neuroscience 113, 607615.
Minzenberg, M.J., Poole, J.H., Benton, C., Vinogradov, S., 2004. Association of
anticholinergic load with impairment of complex attention and memory in
schizophrenia. The American Journal of Psychiatry 161, 116124.
Muldoon, M.F., Barger, S.D., Ryan, C.M., Flory, J.D., Lehoczky, J.P., Matthews, K.A.,
Manuck, S.B., 2000. Effects of lovastatin on cognitive function and psychological
well-being. The American Journal of Medicine 108, 538546.
Muldoon, M.F., Flory, J.D., Ryan, C., 2001. Serum cholesterol, the brain, and cognitive
functioning. In: Waldstein, S.R., Elias, M.F. (Eds.), Neuropsychology of Cardiovascular Disease, pp. 3759. Mahwah, NJ.
O'Brien, W.T., Xu, G., Batta, A., Tint, G.S., Salen, G., Dyer, C.A., Kendler, A., Servatius, R.J.,
2002. Developmental sensitivity of associative learning to cholesterol synthesis
inhibitors. Behavioural Brain Research 129, 141152.
Schoknecht, P.A., Ebner, S., Pond, W.G., Zhang, S., McWhinney, V., Wong, W.W., Klein, P.D.,
Dudley, M., Goddard-Finegold, J., Mersmann, H.J., 1994. Dietary cholesterol supplementation improves growth and behavioral response of pigs selected for genetically
high and low serum cholesterol. The Journal of Nutrition 124, 305314.
Schreurs, B.G., Smith-Bell, C.A., Darwish, D.S., Wang, D., Burhans, L.B., Gonzales-Joekes,
J., Deci, S., Stankovic, G., Sparks, D.L., 2007. Cholesterol enhances classical conditioning
of the rabbit heart rate response. Behavioural Brain Research 181, 5263.

M. Krakowski, P. Czobor / Schizophrenia Research 130 (2011) 2733


Simpson, G.M., Cooper, T.A., 1978. Clozapine plasma levels and convulsions. The
American Journal of Psychiatry 135, 99100.
Skosnik, P.D., Yao, J.K., 2003. From membrane phospholipid defects to altered
neurotransmission: is arachidonic acid a nexus in the pathophysiology of schizophrenia? Prostaglandins, Leukotrienes, and Essential Fatty Acids 69, 367384.
Sparks, D.L., Schreurs, B.G., 2003. Trace amounts of copper in water induce betaamyloid plaques and learning decits in a rabbit model of Alzheimer's disease.
Proceedings of the National Academy of Sciences of theUnited States of America
100, 1106511069.
Sparks, D.L., Liu, H., Gross, D.R., Scheff, S.W., 1995. Increased density of cortical
apolipoprotein E immunoreactive neurons in rabbit brain after dietary administration of cholesterol. Neuroscience Letters 187, 142144.
Tall, A.R., Costet, P., Wang, N., 2002. Regulation and mechanisms of macrophage
cholesterol efux. The Journal of Clinical Investigation 110, 899904.

33

Voikar, V., Rauvala, H., Ikonen, E., 2002. Cognitive decit and development of motor
impairment in a mouse model of NiemannPick type C disease. Behavioural Brain
Research 132, 110.
Wardle, J., Rogers, P., Judd, P., Taylor, M.A., Rapoport, L., Green, M., Nicholson, P.K., 2000.
Randomized trial of the effects of cholesterol-lowering dietary treatment on
psychological function. The American Journal of Medicine 108, 547553.
Yao, J.K., 2003. Abnormalities of fatty acid metabolism in red cells, platelets and brain in
schizophrenia, In: Peet, M., Glen, I., Horrobin, D.F. (Eds.), Phospholipid Spectrum
Disorders in Psychiatry and Neurology, 2nd edition. Marius Press, Lancashire, UK,
pp. 193212.
Zhang, J., Muldoon, M.F., McKeown, R.E., 2004. Serum cholesterol concentrations are
associated with visuomotor speed in men: ndings from the third National Health
and Nutrition Examination Survey, 19881994. The American Journal of Clinical
Nutrition 80, 291298.