Content uploaded by Jonathan Rogers
Author content
All content in this area was uploaded by Jonathan Rogers on Jan 09, 2023
Content may be subject to copyright.
The role of the electroencephalogram (EEG) in determining the
aetiology of catatonia: a systematic review and meta-analysis
of diagnostic test accuracy
Paris Hosseini,
a,t
Rebecca Whincup,
b,t
Karrish Devan,
c
Dory Anthony Ghanem,
d
Jack B. Fanshawe,
e
Aman Saini,
d
Benjamin Cross,
f
Apoorva Vijay,
g
Tomas Mastellar i,
h,i
Umesh Vivekananda,
j,k
Steven White,
l
Franz Brunnhuber,
m
Michael S. Zandi,
k,n
Anthony S. David,
o
Ben Carter,
p
Dominic Oliver,
q
Glyn Lewis,
h
Charles Fry,
m
Puja R. Mehta,
n
Biba Stanton,
r,s
and Jonathan P. Rogers
c,h,
*
a
Department of Neuropsychiatry, University College London Hospitals NHS Foundation Trust, London, UK
b
Leicestershire Partnership NHS Trust, Leicester, UK
c
South London and Maudsley NHS Foundation Trust, London, UK
d
Medical School, University College London, London, UK
e
Department of Psychiatry, University of Oxford, Oxford, UK
f
Mersey Care NHS Foundation Trust, Prescot, UK
g
GKT School of Medical Education, King’s College London, London, UK
h
Division of Psychiatry, University College London, London, UK
i
Inserm U1172, CHU de Lille, Lille Neuroscience & Cognition (LilNCog), Université de Lille, Lille, France
j
Department of Clinical and Experimental Epilepsy, Institute of Neurology UCL, London, UK
k
National Hospital for Neurology and Neurosurgery, London, UK
l
Department of Clinical Neurophysiology, The National Hospital for Neurology and Neurosurgery, London, UK
m
Department of Clinical Neurophysiology, King’s College Hospital NHS Foundation Trust, London, UK
n
Queen Square Institute of Neurology, University College London, London, UK
o
Institute of Mental Health, University College London, London, UK
p
Department of Biostatistics and Health Informatics, King’s College London, London, UK
q
Department of Psychosis Studies, King’s College London, London, UK
r
Department of Neurology, King’s College Hospital NHS Foundation Trust, London, UK
s
Neuropsychiatry Service, South London and Maudsley NHS Trust, St. Thomas’Hospital, London, UK
Summary
Background Catatonia is a psychomotor syndrome that has a wide range of aetiologies. Determining whether cata-
tonia is due to a medical or psychiatric cause is important for directing treatment but is clinically challenging. We
aimed to ascertain the performance of the electroencephalogram (EEG) in determining whether catatonia has a
medical or psychiatric cause, conventionally defined.
Methods In this systematic review and meta-analysis of diagnostic test accuracy (PROSPERO CRD42021239027),
Medline, EMBASE, PsycInfo, and AMED were searched from inception to May 11, 2022 for articles published in
peer-reviewed journals that reported EEG findings in catatonia of a medical or psychiatric origin and were
reported in English, French, or Italian. Eligible study types were clinical trials, cohort studies, case–control
studies, cross-sectional studies, case series, and case reports. The reference standard was the final clinical
diagnosis. Data extraction was conducted using individual patient-level data, where available, by two authors. We
prespecified two types of studies to overcome the limitations anticipated in the data: larger studies (n≥5), which
were suitable for formal meta-analytic methods but generally lacked detailed information about participants, and
smaller studies (n< 5), which were unsuitable for formal meta-analytic methods but had detailed individual
patient level data, enabling additional sensitivity analyses. Risk of bias and applicability were assessed with the
QUADAS-2 tool for larger studies, and with a published tool designed for case reports and series for smaller
studies. The primary outcomes were sensitivity and specificity, which were derived using a bivariate mixed-effects
regression model.
Findings 355 studies were included, spanning 707 patients. Of the 12 larger studies (5 cohort studies and 7 case
series), 308 patients were included with a mean age of 48.2 (SD = 8.9) years. 85 (52.8%) were reported as male and 99
had catatonia due to a general medical condition. In the larger studies, we found that an abnormal EEG predicted a
*Corresponding author. UCL Division of Psychiatry, 6th Floor, Maple House, 149 Tottenham Court Rd, Bloomsbury, London W1T 7NF, UK.
E-mail address: jonathan.rogers@ucl.ac.uk (J.P. Rogers).
t
Joint first authors.
eClinicalMedicine
2023;56: 101808
Published Online xxx
https://doi.org/10.
1016/j.eclinm.2022.
101808
www.thelancet.com Vol 56 February, 2023 1
Articles
medical cause of catatonia with a sensitivity of 0.82 (95% CI 0.67–0.91) and a specificity of 0.66 (95% CI 0.45–0.82)
with an I
2
of 74% (95% CI 42–100%). The area under the summary ROC curve offered excellent discrimination
(AUC = 0.83). The positive likelihood ratio was 2.4 (95% CI 1.4–4.1) and the negative likelihood ratio was 0.28 (95%
CI 0.15–0.51). Only 5 studies had low concerns in terms of risk of bias and applicability, but a sensitivity analysis
limited to these studies was similar to the main analysis. Among the 343 smaller studies, 399 patients were included,
resulting in a sensitivity of 0.76 (95% CI 0.71–0.81), specificity of 0.67 (0.57–0.76) and AUC = 0.71 (95% CI
0.67–0.76). In multiple sensitivity analyses, the results were robust to the exclusion of reports of studies and in-
dividuals considered at high risk of bias. Features of limbic encephalitis, epileptiform discharges, focal abnormality,
or status epilepticus were highly specific to medical catatonia, but features of encephalopathy had only moderate
specificity and occurred in 23% of the cases of psychiatric catatonia in smaller studies.
Interpretation In cases of diagnostic uncertainty, the EEG should be used alongside other investigations to ascertain
whether the underlying cause of catatonia is medical. The main limitation of this review is the differing thresholds for
considering an EEG abnormal between studies.
Funding Wellcome Trust, NIHR Biomedical Research Centre at University College London Hospitals NHS Foun-
dation Trust.
Copyright © 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
Keywords: Catatonia; Electroencephalogram; EEG; Systematic review; Meta-analysis; Diagnostic test accuracy
Introduction
Catatonia is a psychomotor syndrome, characterised by
specific abnormalities in movement and speech with
accompanying neurovegetative and behavioural signs.
1
There are also distinct affective signs that some have asso-
ciated with catatonia.
2
Having been described originally by
Kahlbaum in 1874 as an independent entity,
3
it was
considered as part of schizophrenia for much of the 20th
century.
4
It is now recognised in both the major psychiatric
diagnostic manuals (ICD-11 and DSM-5-TR) that catatonia
mayoccurasaclinicalmanifestationofabroadspectrumof
psychiatric and general medical disorders.
5,6
Recent data
suggest it has an incidence of approximately 10 episodes per
100,000 person-years
7
and some (though not all) studies
have associated it with increased mortality, even compared
to other major psychiatric disorders.
8,9
Electroencephalography (EEG) was a technique first
developed in the 1920s by the psychiatrist Hans Berger,
with the aim of finding a physical basis for mental
function.
10
However, apart from identifying occasional
Research in context
Evidence before this study
Catatonia is a severe psychomotor syndrome that may arise
due to a psychiatric condition or a general medical condition.
Ascertaining the aetiology of catatonia is an important clinical
question, as it has substantial treatment implications. The
current evidence is mainly based on disparate case reports and
small observational studies. In this study, we aimed to
ascertain the diagnostic accuracy of the electroencephalogram
(EEG) in determining whether catatonia has a medical or
psychiatric cause. We searched Medline, EMBASE, PsycInfo,
and AMED up to May 11, 2022 for studies that included cases
of catatonia where individuals had undergone an EEG and had
received a final clinical diagnosis. Search terms combined
synonyms for catatonia with synonyms for EEG. Among the
larger studies, the sensitivity of an EEG for detecting a
medical cause of catatonia was 0.82 (95% CI 0.67–0.91) and
the specificity was 0.66 (95% CI 0.45–0.82). Less than half of
these studies had low concerns in terms of risk of bias and
applicability, but a sensitivity analysis limited to these studies
gave a similar result.
Added value of this study
To our knowledge, this is the first systematic review to
examine the diagnostic test accuracy of the EEG in catatonia.
We found that the EEG offered excellent discrimination with
an area under the ROC curve of 0.83.
Implications of all the available evidence
Our findings and other literature suggest that the EEG should
be used as part of a diagnostic work-up for catatonia in cases
where the aetiology is unclear, but it should not be the only
piece of evidence used in making a diagnosis. The substantial
minority of patients with a psychiatric cause of their catatonia
and an abnormal EEG suggests that a group of patients with
catatonia may have an identifiable electroencephalographic
process as part of their illness.
Articles
2 www.thelancet.com Vol 56 February, 2023
general medical ‘mimics’of psychiatric disorders, the
utility of the EEG in psychiatry has been limited, with
abnormalities tending to be nonspecific with poor cor-
relation to current diagnostic categories.
11
The primary
use of the EEG in contemporary clinical practice is in
the assessment of epilepsy, although it is also valuable
in evaluating levels of consciousness, in localising le-
sions, and in the diagnosis of encephalitides and sleep
disorders.
12
Attempts to characterise the EEG in catatonia date as
far back as the 1950s in various populations. Findings
have varied, including groups of spikes and abnormal
responses to photic stimulation correlating with clinical
state.
13–15
However, there has been little attempt to repli-
cate these results. More recently, Northoff and colleagues
investigated the role of movement-related cortical poten-
tials (Bereitschaftspotentials) on the EEG, finding that pa-
tients with catatonia showed significantly delayed
potentials, relative to psychiatric and healthy controls.
16
In clinical practice, one of the most challenging di-
lemmas in patients with catatonia is ascertaining
whether it is associated with a conventionally defined
primary psychiatric disorder, such as major depressive
disorder, bipolar affective disorder, schizophrenia or a
neurodevelopmental disorder, or whether it is associ-
ated with a general medical cause, such as status epi-
lepticus, autoimmune encephalitis, neurodegenerative
disease, a space-occupying lesion, or medications.
17
These varying disorders can require dramatically
different treatments, so the distinction is critical.
In current practice, a standard work-up for catatonia
may include a detailed history and physical examination
as well as a wide range of blood tests, cerebrospinal fluid
analysis, a urine drug screen, neuroimaging and EEG,
but this varies depending on the clinical scenario.
18–21
Recommendations vary, however, with some authors
suggesting that all patients with catatonia have an
EEG
18,20,22
and others advising that an EEG is just
considered in catatonia
19,23
or that it is used only in certain
circumstances.
21,24
According to recent observational data
from a large US study in acute hospitals, only 4.6% of
patients with catatonia had an EEG, compared to 6.4%
who underwent a lumbar puncture.
25
Overall, the evi-
dence base for use of the EEG remains uncertain and
practice appears to differ. There are two clinical scenarios
where there is an obvious benefit of EEG recording in
catatonia. One is in the context of possible non-convulsive
status epilepticus
26
and the other is in suspected NMDA
receptor encephalitis, where a highly specificfinding of
extreme delta brush is sometimes evident.
27
However, overall there is currently very little evi-
dence on which to base the decision as to whether an
EEG is helpful in catatonia. In particular, the sensitivity,
specificity, positive predictive value and negative pre-
dictive value of the EEG in identifying whether there is a
medical or psychiatric cause of catatonia is unclear.
Given that most studies of catatonia have small sample
sizes,
28
there is a need to synthesise data from multiple
reports to reach robust conclusions. A previous sys-
tematic review from 1998 examined EEG abnormalities
in catatonia due to a medical condition, finding that
84.7% of cases had an abnormality, most commonly
diffuse slowing, but this did not include the more recent
literature and there was no comparison group of cata-
tonia due to a psychiatric illness.
29
Moreover, the cor-
relation between specific EEG abnormalities and the
aetiology of catatonia has not been systematically stud-
ied but has the potential to be more useful than a simple
normal-abnormal EEG classification.
In terms of terminology, we note there is controversy
over the use of the traditional functional-organic
distinction, as it artificially dichotomises complex dis-
orders.
30
For the purposes of this study, we are inter-
ested in the pragmatic clinical distinction between cases
of catatonia where there is considered an identifiable
neuropathological process (which we term ‘medical’
catatonia) and those where catatonia is considered part
of a primary mental disorder (which we term ‘psychi-
atric’catatonia). While we acknowledge the imperfec-
tions of this terminology, we can benefit from a
common language within this paper.
We conducted a systematic review and meta-analysis
of the diagnostic test accuracy of the standard clinical
EEG in catatonia for ascertaining whether catatonia is
due to a medical cause (as opposed to a psychiatric
cause). As a secondary objective, we aimed to charac-
terise the specific EEG abnormalities in catatonia, both
medical and psychiatric.
Methods
Search strategy
In this systematic review and meta-analysis of diag-
nostic test accuracy, the authors used Ovid to search
Medline® All, EMBASE Classic + EMBASE, APA,
PsycInfo, and AMED (Allied and Complementary
Medicine). The overall approach to developing a search
in each database was to combine synonyms for catatonia
with synonyms for electroencephalography without
limits. The search was originally run on 23/02/2021 and
updated on 11/05/2022.
The full search strategy for all databases is available
in Supplementary Methods 1. The search strategy for
Medline is as follows:
1. catatoni*.mp. [mp = ab, hw, ti, tn, ot, dm, mf, dv,
kw, fx, dq, tc, id, tm, mh, nm, kf, ox, px, rx, ui, sy]
2. exp Catatonia/or exp Schizophrenia, Catatonic/
3. 1 or 2
4. (eeg or electroencephalogr* or electrocerebral or
telemetr*).mp. [mp = ab, hw, ti, tn, ot, dm, mf, dv,
kw, fx, dq, tc, id, tm, mh, nm, kf, ox, px, rx, ui, sy]
5. exp Electroencephalography/
6. 4 or 5
Articles
www.thelancet.com Vol 56 February, 2023 3
7. 3 and 6
8. 7 use ppezv
In addition to searching databases, the authors
examined the reference lists of included articles and
contacted significant researchers in the field to identify
further works. Duplicate articles were first identified
automatically using Ovid, then manually by comparing
similar article citations.
Selection criteria
Inclusion criteria were observational or interventional
human studies published in a peer-reviewed journal in
English, French, or Italian. Clinical trials, cohort
studies, case–control studies, cross-sectional studies,
case series, and case reports were eligible. Individuals
must have had a diagnosis of catatonia in the opinion of
the authors of the original study and an aetiology for
catatonia must have been described (at a minimum
stating whether it was medical or psychiatric). There was
no age restriction and individuals could be in any clin-
ical setting. A clinical EEG (either scalp or intracranial)
must have been performed while the individual was
experiencing catatonia and there must be a clinical
report in the article that identified –at a minimum –
whether it was considered normal or abnormal. For the
larger studies, which underwent a formal meta-analysis,
there was an additional inclusion criterion of having at
least 5 eligible patients. A cut-off of 5 was chosen as a
pragmatic compromise between reducing selection bias
and the requirements of formal meta-analytic methods
on the one hand, and the small sample sizes in most
studies of EEG diagnostic test accuracy on the other,
31,32
which we anticipated would be particularly the case for a
rare disorder.
Conference abstracts were excluded because they
generally lack detailed information about case histories,
so assumptions about missing data do not hold. Articles
in which it was not clear that individual patients had
catatonia, or an EEG was reported only during treatment
with electroconvulsive therapy or other induced seizures
were also excluded. Articles in which only quantitative
EEG (with, for example, spectral analysis) or an EEG
described only in terms of the absence of certain ab-
normalities (and thereby not commenting on whether
other abnormalities were present) were also excluded.
Two authors (P.H. and K.D.) assessed article inclu-
sion by examining titles and abstracts sequentially in
parallel, blinded to each other’s ratings. Where there
was disagreement between reviewers, the study in
question was included for the next round of screening.
Articles identified for full text screening were retrieved
by searching online catalogues and university libraries.
Where articles could not be retrieved, the authors were
contacted with a request to provide the text. Two of the
authors (P.H., K.D., R.W., D.A.G., A.S., J.P.R., T.M.,
and J.B.F.) assessed article inclusion by examining the
full texts of the identified articles in parallel, blinded to
each other’s ratings. Where there was disagreement on
the inclusion of a full text, an additional author who had
not already reviewed the full text (J.P.R. or P.R.M.)
arbitrated.
The systematic review is reported according to
PRISMA guidelines (see Supplementary Tables S1 and
S2 for checklists) and the study protocol was preregis-
tered with PROSPERO at https://www.crd.york.ac.uk/
prospero/display_record.php?RecordID=239027.
Data extraction
Where possible, data were sought at an individual pa-
tient level, but summary estimates were also included.
Definitions of each variable for which the data were
extracted are included in Supplementary Table S3. Data
were extracted by two of the authors (R.W., J.B.F.,
D.A.G., B.Cross, P.H., K.D., A.S., J.P.R. and T.M.) in
parallel, blinded to each other’s data. Where there were
discrepancies between the data extracted, a third author
from this list arbitrated. In cases of ambiguity, the
original investigators of the study were contacted for
further details.
To uniformly synthesise the EEG findings, two
neurophysiologists (C.F. and F.B.) developed a template
with the following fields: whether the EEG was normal,
the posterior background rhythm, the presence of fea-
tures of encephalopathy, the presence of features of
limbic encephalitis, whether the EEG was reactive to eye
opening, the presence of epileptiform discharges, the
presence of focal abnormalities, whether sleep was
recorded, the presence of normal sleep architecture and
the presence of status epilepticus. All EEG reports were
coded using this template by a neurophysiologist (C.F.)
and either a neurologist (P.R.M.) or a psychiatrist
(J.P.R.) in parallel with blinding. Where there were
discrepancies in the coding of EEG reports, one of the
authors (P.R.M. or J.R.) who had not already reviewed
the report arbitrated.
Given that there is a wide variety of potential causes
for catatonia, and the diagnosis of psychiatric disorders
remains clinical, we decided to use the considered final
clinical opinion of the report authors as the reference
diagnostic standard. Where catatonia was reported as
having both a medical and psychiatric cause, it was
coded as medical catatonia, as clinicians are most often
interested in ruling out medical causes.
For larger studies, the risk of bias was assessed using
the QUADAS-2 tool, which is specifically designed for
studies of diagnostic accuracy.
33
The QUADAS-2 was
independently completed by two authors (R.W. and
T.M.) and a third author (J.P.R.) arbitrated where there
were discrepancies. As recommended within the
QUADAS-2 tool, we provided some review-specific
guidance, which can be found in Supplementary
Articles
4 www.thelancet.com Vol 56 February, 2023
Methods 2. Risk of bias for the smaller studies was
assessed using a tool designed to assess the methodo-
logical quality of case series and case reports.
34
This tool
had two items that related specifically to studies of
medication effects, so these items were excluded and the
adapted tool with scoring criteria is in Supplementary
Methods 3. Two of the authors (R.W., A.V., J.B.F.,
P.H., B.Cross, J.P.R., K.D., D.A.G., or T.M.) conducted
this assessment; in cases of discrepancies, a third author
from this list arbitrated. The QUADAS-2 does not
recommend using an overall rating, but for the tool used
for smaller studies, a maximum score of 6 was possible,
so scores of 0–2, 3–4, and 5–6 were denoted as low,
moderate, and high quality, respectively.
Where duplicate publications reporting the same
individual were identified, the report with the most
detail was included.
Data analysis
When designing this meta-analysis, the authors
considered that there would be a few larger studies
(n≥5), which would be suitable for a standard meta-
analysis but would have little in the way of clinical
details about patients that would be important for
sensitivity analyses. In contrast, we anticipated that
there would be many smaller studies (n< 5) that would
likely exhibit reporting biases and be computationally
unsuitable for standard meta-analysis but would have
abundant clinical details about the patients. We there-
fore decided to conduct two separate analyses:
1. The larger studies (n≥5) would be synthesised
based on summary estimates using formal meta-
analysis methodology. Any sensitivity analyses
where data were available would be conducted on
these larger studies.
2. The smaller studies (n< 5) would be synthesised
based on individual patient data as if they were all
from one study using the binomial ‘exact’method.
The overall estimates of sensitivity and specificity
may not be as reliable as the analysis of larger
studies, but this would facilitate relevant sensitivity
analyses and more detailed description of the
patients.
Descriptive statistics for both types of studies were
calculated and tabulated.
The primary outcome was whether an EEG was re-
ported as abnormal, considered at a per-patient level. In
this paper, an abnormal EEG is considered a positive
finding, while a normal EEG is considered a negative
finding. A true positive result would be a patient with
medically caused catatonia who had an abnormal EEG.
Secondary outcome measures were specific EEG ab-
normalities. The main measures of effect were sensi-
tivity and specificity with 95% confidence intervals,
which were presented using forest plots. The analysis
was performed by using a bivariate random effects
model of sensitivity and specificity. This allowed calcu-
lation of the area under the summary receiver operating
characteristic (SROC) curve. Additional analyses were
conducted to calculate positive predictive values, nega-
tive predictive values, and diagnostic likelihood ratios.
These were used to generate a probability modifying
plot, comparing pre-test and post-test probabilities.
Calculations of positive predictive values and negative
predictive values used an estimated baseline prevalence
of medical catatonia (among all cases of catatonia) of
20% from a previous systematic review, although this
varies by clinical setting.
17
A prespecified sensitivity analysis was performed by
excluding participants who used a psychotropic drug
within 7 days prior to the EEG recording. Additional
sensitivity analyses were conducted (for smaller studies
or larger studies, as data permitted) by excluding certain
studies or participants deemed to be at high risk of bias:
studies published prior to 1980, studies published prior
to 2010, studies not deemed of high quality, studies with
concerns about the reference standard, studies where
follow-up time was potentially inadequate to be confi-
dent in the final diagnosis, studies lacking either med-
ical or psychiatric catatonia cases, individuals with a
possible prior neurological disorder, individuals not
meeting DSM-5 criteria for catatonia, individuals who
were prescribed psychotropic medications (including
benzodiazepines) in the 7 days prior to the EEG, in-
dividuals where alternative causes of catatonia had not
been adequately ruled out, and individuals where the
underlying disorder was neurodevelopmental. A pre-
specified subgroup analysis was conducted in which
individuals were divided into age groups; the groups
were children (<18 years), adults (18–64 years), and
older adults (≥65 years). Additional subgroup analyses
were conducted by sex and underlying diagnosis.
Study variability was assessed using the I
2
measure
of heterogeneity and potential sources of heterogeneity
were described and explored through subgroup ana-
lyses. Publication bias for the larger studies was
assessed within the midas package by performing a
linear regression of log odds ratios on the inverse root of
effective sample sizes.
35
The meta-analysis was performed in Stata-MP v16.1
using the midas package.
36
The forest plot was produced
using RevMan v5.4. Statistical significance was set at
0.05.
Role of the funding source
The funders of the study had no role in the study design,
data collection, data analysis, data interpretation, writing
of the report or the decision to submit it for publication.
J.P.R., R.W., and P.H. had access to the raw data. The
corresponding author had full access to all data in the
Articles
www.thelancet.com Vol 56 February, 2023 5
study and had final responsibility for the decision to
submit for publication.
Results
The search strategy yielded 1608 results, which after
deduplication left 1166 articles, which were screened
(Fig. 1). This resulted in 355 included studies with a
total of 707 patients, of which 12 were larger studies
(n≥5) and 343 were smaller studies (n< 5). All EEGs
were recorded via the scalp; no studies reporting intra-
cranial EEGs met the eligibility criteria.
Characteristics of included studies
The 12 larger studies are presented in Table 1. 6 were
from the USA, 3 from the UK, and 1 each from Italy,
Japan, and Mexico. There were 5 cohort studies and 7
case series. In total, 308 patients were included with a
mean age of 48.2 (n= 139, SD = 8.9) years. Sex was
reported in 161 patients, of whom 85 (52.8%) were male
and 76 (47.2%) female. In terms of diagnostic groups,
99 had an underlying general medical condition, 11 a
mood disorder, 137 a psychotic disorder, and 61 an
unspecified psychiatric catatonia. The results for quality
assessment of the larger studies using the QUADAS-2
tool are shown in Table 2.
Among the 343 smaller studies, there were 399 pa-
tients, of whom 302 had medical catatonia and 97 psy-
chiatric catatonia. A summary of the smaller studies and
the cases in them is presented in Table 3. Additional
data on diagnoses and treatments received are presented
in Supplementary Tables S4 and S5. A full list of the
smaller studies with their quality assessment rating is in
Supplementary Table S6.
Diagnostic test accuracy of the larger included
studies
Fig. 2 displays a forest plot for the sensitivity and
specificity of the larger studies alongside the raw data.
Of note, 6 studies included only patients with psychi-
atric catatonia,
37–42
so sensitivity cannot be derived for
these studies, while 1 study included only patients with
medical catatonia,
46
so specificity cannot be derived for
this study.
The main diagnostic test accuracy meta-analysis
found that the sensitivity (i.e., the proportion of pa-
tients with medical catatonia who had an abnormal
EEG) was 0.82 (95% CI 0.67–0.91) and the specificity
(i.e., the proportion of patients with psychiatric catatonia
who had a normal EEG) was 0.66 (95% CI 0.45–0.82).
The proportion of variance accounted for by between-
study heterogeneity was measured with an I
2
statistic
of 74% (95% CI 42–100%). The positive likelihood ratio
was 2.4 (95% CI 1.4–4.1) and the negative likelihood
ratio was 0.28 (95% CI 0.15–0.51). The diagnostic odds
ratio was 9 (95% CI 3–22). A summary receiver oper-
ating characteristics (SROC) curve displaying this result
along with the 5 studies from which both sensitivity and
specificity could be derived is shown in Fig. 3 with an
area under the SROC curve of 0.83 (95% CI 0.79–0.86),
corresponding to excellent discrimination.
48
Study 10
45
appears to be an outlier in Fig. 3, but its specificity is
based on findings in only 2 patients, so it has a wide
confidence interval, as shown in Fig. 2. For clinical
interpretation, Fig. 4 displays a probability modifying
plot, which illustrates the effect on the post-test proba-
bility of medical catatonia of an abnormal (positive) or
normal (negative) EEG for a given prior probability. If a
prevalence of medical catatonia among all cases of
catatonia of 20% is assumed,
17
the positive predictive
value is 0.37 and the negative predictive value is 0.93.
Fagan’s Bayesian nomogram assuming a baseline
probability of medical catatonia of 20% is shown in
Supplementary Fig. S1.
Model diagnostics for the larger studies are shown in
Supplementary Fig. S2, which shows a good model fit
without any outliers. No studies were considered influ-
ential based on their Cook’s distance. When publication
bias was assessed by performing a linear regression of
log odds ratios on the inverse root of effective sample
sizes, no evidence for publication bias was found with a
regression coefficient of 0.9 (95% CI −13.6 to 15.4), as
illustrated by the funnel plot in Supplementary Fig. S3.
Sensitivity analyses excluding studies that were older,
had more concerns on the QUADAS-2, had high con-
cerns about the reference standard or lacked both
medical and psychiatric catatonia cases were performed
with the results shown in Table 4.
Diagnostic test accuracy of the smaller included
studies
The results of the EEG findings for the smaller studies
are shown combined in a 2 × 2 table (Table 5). From this
table, the sensitivity was 0.76 (95% CI 0.71–0.81) and
the specificity was 0.67 (0.57–0.76). The area under the
ROC curve was 0.71 (95% CI 0.67–0.76). Sensitivity
analyses excluding the following groups were conducted
and the results are shown in Supplementary Table S7:
studies published prior to 1980, studies not deemed of
high quality, studies where follow-up time was poten-
tially inadequate to be confident in the final diagnosis,
studies where alternative causes were not adequately
ruled out, individuals with a possible prior neurological
disorder, individuals with psychotropic drug use within
7 days prior to the EEG, individuals not meeting DSM-5
criteria for catatonia and individuals where the under-
lying disorder was neurodevelopmental. There was
substantial overlap in the confidence intervals for
sensitivity and specificity with the primary analysis for
all sensitivity analyses, suggesting that the results were
robust to the exclusion of studies at high risk of bias.
Articles
6 www.thelancet.com Vol 56 February, 2023
The number of abnormal EEGs by underlying diagnosis
are presented in Supplementary Table S8. Subgroup
analyses by age, sex, diagnostic subgroup and continent
of participants are shown in Supplementary Table S9.
Subgrouping by age merits particular attention: while
the area under the ROC curve for children (0.79 [95% CI
0.68–0.87]) and adults (0.72 [95% CI 0.66–0.77]) pro-
vided acceptable discrimination, for older adults (0.53
[95% CI 0.36–0.68]) the EEG provided no discrimination
between medical and psychiatric catatonia.
48
Records identified from: (n =
1608)
Databases (n = 1437)
MEDLINE (n = 322)
EMBASE (n = 903)
PsycINFO (n = 209)
AMED (n = 3)
Reference lists (n = 171)
Records removed before
screening (n = 442):
Duplicate records removed
automatically (n = 361)
Duplicate records removed
manually (n = 81)
Titles screened
(n = 1166)
Records excluded
(n = 62)
Reports sought for retrieval
(n = 975)
Reports not retrieved
(n = 27)
Reports assessed for eligibility
(n = 948)
Reports excluded: (n = 593)
Not in an included language
(n = 99)
Not original research (n = 47)
Not peer-reviewed (n = 172)
Animal subjects (n = 3)
No catatonia identified (n =
79)
No EEG report during
catatonia (n = 161)
EEG reports not linked to
diagnoses (n = 1)
EEG report from induced
seizure (n = 31)
Studies included in review
(n = 355)
Reports of included studies
(n = 355)
Identification of studies via databases and registers
Identification
Screening
Included
Abstracts screened
(n = 1104)
Records excluded
(n = 129)
Larger studies included in formal
meta-analysis (n = 12)
Smaller studies (n = 343)
Fig. 1: PRISMA flowchart.
Articles
www.thelancet.com Vol 56 February, 2023 7
Study Setting Design Sample
size
Demographics n
medical
catatonia
Medical catatonia EEG findings n
psychiatric
catatonia
Psychiatric catatonia EEG findings
1 MacMahon
(1938)
37
UK; psychiatric
hospital
Cohort 11 –0–11 1 normal; 10 abnormal: delta rhythm
(10)
2 Walter
(1942)
38
UK; psychiatric
hospital
Cohort 6 –0–6 3 normal; 1 doubtful (considered
normal for meta-analysis); 2
abnormal
3 Stevens
(1958)
39
USA; psychiatric
hospital
Case
series
21 –0–21 20 normal; 1 abnormal: runs of high–
voltage activity (1)
4 Ishibashi
(1963)
40
Japan; psychiatric
hospital
Case
series
11 –0–11 1 normal; 8 borderline (considered
normal for meta-analysis); 2
abnormal
5 Abenson
(1970)
41
UK; psychiatric
hospital
Cohort 79 –0–79 60 normal; 19 abnormal: ‘choppy’
abnormalities (9), temporal (focal)
abnormalities (7), dysrhythmic
abnormalities (3)
6 Philbrick
(1994)
42
USA; general
hospital
Case
series
53M,2F
Age 59.6 (mean),
16.2 (SD)
0–5 4 normal; 1 abnormal: background
slowing (1)
7 Carroll
(1995)
43
USA; psychiatric
hospital or
medical
psychiatry unit
Case
series
26 15 M, 11 F
Age 48.2 (mean),
21.4 (SD)
13 2 normal; 11 abnormal: diffuse slowing (8), focal
slowing (2), bilateral spikes (1)
13 8 normal; 5 abnormal: diffuse
slowing (4), focal slowing (1)
8 Carroll
(1998)
29
USA; psychiatric
hospital
Case
series
12 Age 41.8 (mean),
17.9 (SD)
6 1 normal; 5 abnormal 6 5 normal; 1 abnormal
9 Smith
(2012)
44
USA; general
hospital
Cohort 68 28 M, 40 F
Age 51.9 (mean),
20.9 (SD)
16 1 normal; 15 abnormal: diffuse slowing (13),
focal temporal slowing (5), asymmetry (6)
a
52 13 normal; 39 abnormal: diffuse
slowing (31), focal temporal slowing
(7), asymmetry (6)
a
10 Llesuy
(2017)
45
USA; general
hospital
Case
series
20 Age 49.6 (mean),
17.7 (SD)
18 7 normal; 11 abnormal: generalised slowing (7),
generalised slowing with epileptiform activity
(3), seizures (1)
2 1 normal; 1 abnormal: generalised
slowing (1)
11 Espinola-
Nadurille
(2019)
46
Mexico;
neurosciences
hospital
Cohort 41 –41 4 normal; 37 abnormal: generalised dysfunction
(33), asymmetric activity (7), delta–brush activity
(7), epileptic activity (6), focal dysfunction (3)
a
0–
12 Ursitti
(2021)
47
Italy; children’s
hospital
Case
series
83M,5F
Age 15.1 (mean),
1.6 (SD)
5 1 normal; 4 abnormal: focal slowing (3), status
epilepticus (1), diffuse beta activity (1)
a
3 2 normal; 1 abnormal: focal slowing
(1)
a
a
Each patient may be reported to have more than one EEG abnormality in these studies.
Table 1: Characteristics of larger studies included in the meta-analysis.
Study Funding Risk of bias Applicability concerns
patient selection Index test Reference standard Flow and timing Patient selection Index test Reference standard
1 MacMahon (1938)
37
Not stated Unclear Unclear Unclear Unclear Low Unclear High
2 Walter (1942)
38
Not stated Low Low Unclear Unclear Low High Low
3 Stevens (1958)
39
Not stated High Low Low Unclear High Low High
4 Ishibashi (1963)
40
Not stated High Unclear Unclear High Low Unclear Unclear
5 Abenson (1970)
41
Not stated Low Low Low Unclear Low Low Low
6 Philbrick (1994)
42
Not stated High Unclear Low Low High Unclear High
7 Carroll (1995)
43
Not stated Low Low High Unclear Low Low Low
8 Carroll (1998)
29
Not stated Low Unclear High High Low Unclear Low
9 Smith (2012)
44
Non-commercial support
a
Low Low Low High Low Low Low
10 Llesuy (2017)
45
Not stated Low Unclear Low High Low Low Low
11 Espinola-Nadurille (2019)
46
None Low Unclear Low Low High Low Low
12 Ursitti (2021)
47
None High Unclear Unclear Low Low Low Low
a
Study was partially supported by the Center for Translational Science Activities at Mayo Clinic. The Center was funded in part by a grant from the National Center for Research Resources, a component of
the National Institutes of Health (NIH).
Table 2: Funding statements and quality assessment of larger studies using QUADAS-2.
Articles
8 www.thelancet.com Vol 56 February, 2023
Study characteristics All studies (K= 343)
Publication year, min, max 1952, 2022
Country of corresponding author, k(%)
- USA 134 (39.1)
- Japan 23 (6.7)
- India 22 (6.4)
- Germany 15 (4.4)
- Italy 15 (4.4)
-UK 14 (4.1)
- France 13 (3.8)
- Other 107 (31.2)
Study design, k(%)
- Cohort study 4 (1.2)
- Case series 66 (19.2)
- Case report 273 (79.6)
Number of patients, k(%)
-1 308 (89.8)
-2 17 (5.0)
-3 15 (4.4)
-4 3 (0.9)
Quality assessment, k(%)
- Patient(s) represent(s) the whole experience of the investigator (centre) 205 (59.8)
- Exposure adequately ascertained? 234 (68.2)
- Outcome adequately ascertained? 334 (97.4)
- Other alternative causes that may explain the observation ruled out? 128 (37.3)
- Follow-up long enough for outcomes to occur? 185 (53.9)
- Case(s) described with sufficient details? 262 (76.4)
Overall study quality rating, k(%)
- Low 41 (12.0)
- Medium 184 (53.6)
- High 118 (34.4)
Patient characteristics Medical catatonia (N= 302) Psychiatric catatonia (N= 97) Total (N= 399)
Sex, n(%)
- Male 123 (40.7) 49 (50.5) 172 (43.1)
- Female 178 (58.9) 47 (48.5) 225 (56.4)
- Not specified 1 (0.3) 1 (1.0) 2 (0.5)
Age/years, mean (SD) 35.9 (19.8) 37.8 (20.8) 36.4 (20.0)
Ethnicity, n(%)
- Asian 15 (5.0) 6 (6.2) 21 (5.3)
- Black 15 (5.0) 5 (5.2) 20 (5.0)
- White 32 (10.6) 15 (15.5) 47 (11.8)
- Other 11 (3.6) 1 (1.0) 12 (3.0)
- Not specified 229 (75.8) 70 (72.2) 299 (74.9)
Prior neurological history affecting brain, n(%)
- Present 80 (26.5) 17 (17.5) 97 (24.3)
- Absent 212 (70.2) 75 (77.3) 287 (71.9)
- Not stated 10 (3.3) 5 (5.2) 15 (3.8)
Prior psychiatric history, n(%)
- Present 113 (37.4) 64 (66.0) 177 (44.4)
- Absent 181 (60.0) 29 (29.9) 210 (52.6)
- Not stated 8 (2.7) 4 (4.1) 12 (3.0)
Medication and drug use mentioned in 7 days prior to EEG, n(%)
- Alcohol 3 (1.0) 1 (1.0) 4 (1.0)
- Recreational drugs (not alcohol) 13 (4.3) 0 (0.0) 13 (3.3)
- Antidepressants 22 (7.3) 10 (10.3) 32 (8.0)
(Table 3 continues on next page)
Articles
www.thelancet.com Vol 56 February, 2023 9
As a secondary analysis among the smaller studies,
we examined the diagnostic accuracy of each individual
EEG abnormality, as shown in Table 6. Features of
limbic encephalitis, epileptiform discharges, focal ab-
normalities, and status epilepticus were all highly spe-
cific with varying sensitivity, but the features of
encephalopathy were more sensitive and much less
specific. The EEG posterior background frequencies
were not usually specified but the available frequencies
are presented in Supplementary Table S10.
Discussion
In this systematic review and meta-analysis of diag-
nostic test accuracy, including 355 studies and 707 pa-
tients, we found that scalp EEG has excellent
discrimination in ascertaining whether catatonia was
due to a medical cause in larger studies with acceptable
discrimination in smaller studies. This result was robust
to excluding studies at high risk of bias.
EEG performance varied across age groups with ac-
ceptable performance in children and working-age adults
Patient characteristics Medical catatonia (N= 302) Psychiatric catatonia (N= 97) Total (N= 399)
(Continued from previous page)
- Antipsychotics 104 (34.4) 31 (32.0) 135 (33.8)
- Benzodiazepines 79 (26.2) 18 (18.6) 97 (24.3)
Catatonia meeting DSM-5 criteria, n(%) 227 (75.2) 70 (72.2) 297 (74.4)
Catatonia duration prior to EEG/days (n= 174)
- Mean (SD) 21.5 (47.8) 36.7 (93.1) 24.4 (59.2)
- Median (IQR) 7 (2–21) 14 (2–36) 7 (2–28)
Periodic catatonia (as identified by authors), n(%) 2 (0.7) 9 (9.3) 11 (2.8)
Underlying diagnosis, n(%)
- Catatonia due to a general medical disorder 302 (100.0) –302 (75.7)
- Catatonia due to a primary psychotic disorder –44 (45.4) 44 (11.0)
- Catatonia due to a primary mood disorder –24 (24.7) 24 (6.0)
- Catatonia NOS
a
–29 (29.9) 29 (7.3)
Duration of underlying illness prior to EEG/days (n= 266)
- Mean (SD) 515 (1985) 1616 (3377) 755 (2396)
- Median (IQR) 65 (14–1095) 65 (14–1095) 28 (10–150)
Clinical outcome of catatonia, n(%)
- Full recovery 236 (78.2) 73 (75.3) 309 (77.4)
- Partial recovery 25 (8.3) 15 (15.5) 40 (10.0)
- Continued catatonia 10 (3.3) 4 (4.1) 14 (3.5)
- Death 22 (7.3) 0 (0.0) 22 (5.5)
- Not stated 9 (3.0) 5 (5.2) 14 (3.5)
IQR = interquartile range. NOS = not otherwise specified. SD = standard deviation.
a
This category was used for psychiatric catatonia where the underlying diagnosis was unclear, the underlying diagnosis
was other than a primary psychotic or mood disorder, or catatonia was considered idiopathic.
Table 3: Characteristics of smaller studies and of patients in smaller studies.
Fig. 2: Forest plot of sensitivity and specificity of larger studies.
Articles
10 www.thelancet.com Vol 56 February, 2023
but no meaningful discrimination in older people (>65
years old). There were differences between individual
EEG abnormalities. Features of encephalopathy were
common in both psychiatric and medical catatonia, and
showed moderate sensitivity and specificity, while fea-
tures of limbic encephalitis, epileptiform discharges,
focal abnormalities and status epilepticus were much less
common with low sensitivity but very high specificity.
The strengths of this study included that the per-
formance of the EEG in catatonia was excellent and
found consistently across most studies. It is estimated
with good precision, model performance and discrimi-
nation, so it is unlikely to be due to chance. However, it
is quite possible that other findings, such as higher
sensitivity than specificity, or subgroup differences, are
due to chance given the substantial overlap in confi-
dence intervals.
There are several limitations to this review. Impor-
tantly, the included studies were observational, which
included case reports and series, typically with a high
risk of bias and small sample sizes. Specific issues are
selection bias, measurement bias and external validity,
which we consider in turn.
Selection bias is likely to have played a role in our
findings, as at least four of the 12 larger studies were
found to be at high risk of bias for patient selection in
the QUADAS-2 and only in 59.8% of the smaller studies
did the patient represent the whole experience of the
investigator. There is likely to have been reporting bias,
as a systematic review found that 20% of catatonia cases
had a medical cause,
17
while in our larger studies 32.1%
had a medical cause and in our smaller studies
77.4% had a medical cause. However, this is less of a
problem than it may initially seem because there is only
limited evidence that reporting bias causes biased re-
sults in studies of diagnostic test accuracy
49
and in most
of the included studies (particularly the smaller ones),
EEGs were only an incidental part of the paper, so it
would be unlikely for an EEG finding to substantially
influence the decision of whether to publish. There were
several studies that reported only psychiatric or medical
cases of catatonia, but a sensitivity analysis excluding
these studies did not find that the results were sub-
stantially different. Unfortunately, few of the larger
studies reported funding, although this is also unlikely
Fig. 3: Summary receiver operator characteristics (ROC) curve for
larger studies.
Fig. 4: Probability modifying plot for interpretation of EEG findings in catatonia.
Articles
www.thelancet.com Vol 56 February, 2023 11
to be a major problem in an area where the technology
is not protected by intellectual property and where there
is little pharmaceutical relevance. The proportion of
studies not retrieved was very small (2.8%), so this is
unlikely to have substantially affected the results.
In terms of measurement bias, much of the EEG
reporting was of poor quality, sometimes denoting
EEGs simply as ‘abnormal’without any indication of
which particular abnormalities were present. We were
able to partially overcome this by analysing the smaller
studies, which tended to give more detailed reports. Our
results remained robust after excluding cases where
psychotropic medications (including benzodiazepines)
had been used in the previous 7 days, but it is possible
that antiepileptic or anaesthetic drugs also played a role.
It is also possible that encephalopathic findings may
have been confused with drowsiness or sleep, as som-
nolent states may be harder to distinguish clinically in
the context of catatonia. Although it is usually possible
to distinguish sleep from encephalopathy on the basis of
the EEG,
50
this requires a sufficient length of recording,
which was generally not specified in the included re-
ports. One potential problem would be bias towards the
null hypothesis if medical causes of catatonia were not
adequately identified, resulting in misclassification of
medical cases as psychiatric ones. For the larger studies,
sensitivity analyses were conducted where studies pub-
lished before 1980 and those with low concerns about
the reference standard were excluded, each finding
similar results to the main meta-analysis (Table 4). For
the smaller studies, more data were available, so we
conducted four sensitivity analyses to try to determine
whether there was misclassification, excluding studies
prior to 1980, studies prior to 2010, studies with inad-
equate follow-up time and studies where sufficient
investigations were not performed (Supplementary
Table S7). All of these produced similar results. It
therefore does not seem likely that misclassification due
to inadequate diagnostic investigation explains our re-
sults. Among the smaller studies, it was clear in only a
minority of cases that alternative causes for catatonia
had been adequately ruled out, although a sensitivity
analysis excluding such studies was similar to the main
analysis. The other issue in terms of measurement bias
is that EEGs were often interpreted by a reporter who
already had knowledge of the reference standard,
or –conversely –the reference standard was often
established by a clinician with prior knowledge of the
EEG findings. Some larger studies avoided this, but it is
likely to be a problem in any EEG that was requested as
part of ordinary clinical care and would inflate the
supposed diagnostic test accuracy. However, a sensi-
tivity analysis of the larger studies, excluding those
where there may be concerns about the reference stan-
dard, found similar results to the main analysis.
In terms of external validity, participants came from
a range of psychiatric and medical settings. However,
the major concern is that –among those studies where
routine clinical records were used –only patients whose
clinical presentation apparently justified the use of an
EEG were included in the study. It is likely that such
patients pose additional diagnostic uncertainty, so it is
more reasonable to generalise these results to patients
where there is at least some diagnostic uncertainty. If
clinicians used the EEG more widely in catatonia, it is
likely that more cases of psychiatric catatonia would be
included, so the pre-test probability –and thus the
positive predictive value –of the EEG would be lower.
The studies also presented considerable heterogeneity
in their results. This is particularly apparent in the larger
studies. Fig. 3 suggests that there may be some negative
correlation between sensitivity and specificity, which is
the expected outcome when there is a threshold effect in
a diagnostic test.
51
In a test, such as EEG, where a report
is qualitative, there are often implicit thresholds, above
which different studies or clinicians consider the inves-
tigation to be abnormal,
52
and there is prior evidence that
neurophysiologists do exhibit such reporting thresholds.
53
This alters the metrics for sensitivity and specificity
within an individual study, but the bivariate model used
in this meta-analysis takes into account this threshold
when producing summary estimates. It does, however,
render interpretation more difficult, as it is not clear at
what threshold of considering an EEG to be abnormal the
summary estimates are taken. Individual EEG abnor-
malities are probably more straightforward to interpret in
this regard, as it is clearer what is being considered
abnormal. Another substantial source of heterogeneity
was age, which we explored with a subgroup analysis,
finding much less support for the utility of the EEG
Analysis Number of
studies (k)
Number of
subjects (n)
Sensitivity
(95% CI)
Specificity
(95% CI)
Area under ROC
curve (95% CI)
I
2
Primary analysis 12 308 0.82 (0.67–0.91) 0.66 (0.45–0.82) 0.83 (0.79–0.86) 0.74
Only studies published from 1980 onwards 7 180 0.83 (0.69–0.91) 0.58 (0.34–0.78) 0.80 (0.76–0.83) 0.64
Only studies with low concerns in at least 5 domains of QUADAS-2 5 234 0.82 (0.64–0.92) 0.55 (0.31–0.76) 0.77 (0.73–0.81) 0.59
Only studies with low concerns about reference standard 6 234 0.76 (0.45–0.93) 0.69 (0.38–0.89) 0.79 (0.75–0.82) 0.62
Only studies containing both medical and psychiatric catatonia cases 5 134 0.80 (0.63–0.91) 0.56 (0.31–0.79) 0.77 (0.73–0.81) 0.62%
Table 4: Sensitivity analyses for larger studies.
Articles
12 www.thelancet.com Vol 56 February, 2023
among older adults than in other age groups, which may
be due to the increased prevalence of nonspecificslowing
in general among older people.
54,55
Moreover, it is
possible that additional heterogeneity was introduced by
varying definitions, severities and subtypes of catatonia.
While a sensitivity analysis of smaller studies restricting
to those cases that met DSM-5 criteria for catatonia was
similar to the main analysis, it is possible that the EEG
findings differ in cases, for example, where catatonia has
been defined according to the Northoff Catatonia Scale
2,56
or catatonia is particularly severe. It might be of particular
relevance to understanding any heterogeneity to investi-
gate the EEG findings in malignant catatonia, periodic
catatonia or neuroleptic malignant syndrome in future
studies.
One particularly interesting finding in our results is
that a significant minority of patients with a supposed
psychiatric cause for their catatonia have an abnormal
EEG, most commonly with features of encephalopathy,
which were present in 22 out of 97 (23%) patients in the
smaller studies and at least 48 out of 209 (23.0%) pa-
tients in the larger studies. A previous review has found
that encephalopathic features were the most common
EEG abnormalities in catatonia due to a medical con-
dition, but the current study extends this to catatonia
due to a psychiatric condition. Since encephalopathy is
defined as a pathobiological process in the brain, which
distinguishes it from primary psychiatric disorders, this
finding is surprising and intriguing. There is a
longstanding literature on EEG abnormalities across
psychiatric disorders, but the abnormalities described
hitherto have not been specific to any diagnostic entity.
57
We suggest four possible reasons for the generalised
slowing in catatonia. Firstly, EEG slowing may reflect an
undiagnosed medical condition. There is a substantial
overlap between catatonia and delirium,
58
which has an
encephalopathic EEG correlate, and older reports would
not have recognised NMDA receptor encephalitis.
59
Moreover, ictal slowing can occur,
60,61
although the
absence of evidence for epilepsy in most of these case
reports means that this is unlikely to be a major expla-
nation. Secondly, EEG abnormalities could reflect
various medical complications which have arisen as a
result of catatonia, such as sepsis, cardiac arrhythmia,
renal failure, neuroleptic malignant syndrome and he-
patic dysfunction.
86
Thirdly, some psychotropic drugs,
particularly clozapine,
62
have been associated with EEG
slowing, although our sensitivity analysis, excluding
such cases suggests this is not a major factor. Finally, it
is theoretically possible that a mental state itself could
lead to EEG abnormalities. Catatonia can certainly
generate a marked sympathetic response with fever and
tachycardia being common in severe cases
63
and even
occasionally bilateral dilated pupils unreactive to light.
64
In conclusion, our results are similar to a previous
systematic review of EEG abnormalities in 105 patients
with catatonia, which found that the majority of medical
catatonia cases had an abnormal EEG, usually generalised
Medical catatonia (N= 302) Psychiatric catatonia (N= 97)
EEG normal, n(%) False negatives
72 (23.8)
True negatives
65 (67.0)
EEG abnormal, n(%)
a
True positives
230 (76.2)
False positives
32 (33.0)
- Features of encephalopathy - 160 (53.0) - 22 (22.7)
- Features of limbic encephalitis - 8 (2.6) - 0 (0.0)
- Epileptiform discharges - 75 (24.8) - 6 (6.2)
- Focal abnormality - 73 (24.2) - 5 (5.2)
- Status epilepticus present - 28 (9.3) - 0 (0.0)
a
Some EEGs had more than one abnormality, so figures on the types of abnormalities add up to more than the total number of abnormal EEGs.
Table 5: EEG results by specific EEG abnormality for smaller studies.
EEG abnormality
a
Sensitivity (95% CI) Specificity (95% CI) Area under the ROC curve (95% CI)
Any abnormality (primary analysis) 0.76 (0.71–0.81) 0.67 (0.57–0.76) 0.71 (0.67–0.76)
Features of encephalopathy 0.58 (0.52–0.64) 0.77 (0.68–0.85) 0.68 (0.63–0.73)
Features of limbic encephalitis 0.03 (0.01–0.05) 1.00 (0.96–1.00) 0.51 (0.46–0.56)
Epileptiform discharges 0.25 (0.20–0.30) 0.94 (0.87–0.98) 0.59 (0.54–0.64)
Focal abnormality 0.24 (0.20–0.30) 0.95 (0.88–0.98) 0.60 (0.55–0.65)
Status epilepticus 0.09 (0.06–0.13) 1.00 (0.96–1.00) 0.55 (0.50–0.60)
a
Categories of abnormalities are not mutually exclusive, as many EEGs showed more than one abnormality.
Table 6: Diagnostic test accuracy by individual EEG abnormality for smaller studies.
Articles
www.thelancet.com Vol 56 February, 2023 13
slowing.
29
However, our study takes this further by
incorporating many more studies and comparing the
EEG findings in medical versus psychiatric catatonia
cases. Selection and measurement bias are both likely to
be present, but sensitivity analyses suggest that they did
not have a major effect on our results. The fact that an
EEG was likely to be used only in cases of diagnostic
uncertainty does limit the external validity of our con-
clusions to such cases. Notwithstanding these limita-
tions, it is reasonable to conclude that the EEG is of value
in discerning whether catatonia has a psychiatric or
medical aetiology, but its interpretation relies on the pre-
test probability, the specific EEG findings and the results
of other investigations.
In terms of the implications for future research, our
first suggestion is methodological. EEGs were reported
inconsistently and often minimally, lacking important
details; we call for a minimum reporting standard for
EEGs in case reports and series, specifying at a mini-
mum what abnormalities were present, what the pa-
tient’s state of consciousness was at the time of the
recording, what medications had been taken in recent
days and who reported the recording. Future studies of
the EEG in catatonia should use mixed samples of
catatonia secondary to both psychiatric and medical
disorders with blinding of reporting staff to the sup-
posed diagnosis; such studies could be conducted
retrospectively with existing EEG recordings. Systematic
longitudinal follow-up would be important. Given our
finding that a large minority of patients have a clear EEG
abnormality, catatonia would be an obvious target dis-
order in studies of quantitative EEG analysis.
The main implication for clinical practice is that the
EEG should be considered in cases of catatonia where
there is diagnostic uncertainty to support in establishing
whether there is a medical or psychiatric underlying
disorder. Although it is a safe, non-invasive test, its
diagnostic accuracy is such that it should not be used
alone but belongs as part of a comprehensive work-up,
including history, collateral history, physical examina-
tion and other investigations. A normal EEG increases
the confidence that catatonia has a psychiatric origin. An
abnormal EEG must be interpreted depending on the
specificfinding: features of encephalopathy have only a
moderate specificity, whereas features of limbic en-
cephalitis, epileptiform discharges, focal abnormalities
and status epilepticus are highly specific for a medical
cause of catatonia. However, caution is required in those
aged over 65, where diagnostic accuracy is poor.
Contributors
J.P.R., P.R.M., and B.S. conceived the project. J.P.R., P.R.M., B.S., P.H.
and K.D. designed the project with input from F.B., M.S.Z., A.S.D.,
B.C., D.O., G.L. and C.F. F.B. and C.F. designed the EEG extraction
form. P.H., K.D., R.W., D.A.G., A.S., J.P.R., P.R.M., T.M. and J.B.F.
assessed article inclusion. R.W., J.B.F., D.A.G., B.C., P.H., K.D., A.S.,
J.P.R. and T.M. extracted data. C.F., P.R.M. and J.P.R. coded EEGs.
R.W., A.V., J.B.F., P.H., B.C., J.P.R., K.D., D.A.G. and T.M. conducted
the assessment of risk of bias and applicability. J.P.R. conducted the
analysis with advice from B.C. and D.O. F.B., C.F., U.V., and S.W.
advised on interpretation of the neurophysiological findings. J.P.R. led
the study and wrote the first draft of the manuscript. All authors had the
opportunity to provide input on the final manuscript.
Data sharing statement
Records of article screening, extracted data and statistical analysis from
the study are available from the corresponding author on reasonable
request at jonathan.rogers@ucl.ac.uk.
Declaration of interests
G.L. declares payments made to his institution by the Wellcome Trust
and the NIHR UCLH BRC. J.P.R. declare payments to his institution for
his salary by the Wellcome Trust. M.S.Z. declares salary support to
support research time from the NIHR UCLH BRC. M.S.Z. declares
honoraria for one lecture each for each of the four mentioned in the last
three years: Norwegian Neurological Society; Copenhagen Neuropsy-
chological Society, Rigshospitalet; Cygnet Healthcare; and UCB Pharma.
M.S.Z. declares travel and hotel support for a stay in Florence from the
European Association of Neurology (EAN) for an EAN meeting on
autoimmune encephalitis in April 2022. M.S.Z. represents neurology in
the UK for the Association of British Neurologists for matters related to
Covid in meetings with NHS England and Royal College of Physicians.
All other authors declare no competing interests.
Acknowledgements
J.P.R. and P.R.M. are supported by a Wellcome Trust Clinical Training
Fellowship (102186/B/13/Z). M.S.Z., G.L., and A.S.D. are supported by
the UK NIHR University College London Hospitals Biomedical
Research Centre. The funders of the study had no role in the study
design, data collection, data analysis, data interpretation, writing of the
report or the decision to submit it for publication.
The authors were not paid to write this article by a pharmaceutical
company or other agency. J.P.R., R.W. and P.H. had access to the raw
data. Authors were not precluded from accessing data in the study and
they accept responsibility to submit for publication. J.P.R. was respon-
sible for the decision to submit the manuscript.
We would like to thank Paul Lee and the staff of the South London
and Maudsley Trust Library for their kind assistance in obtaining arti-
cles for this study.
Appendix A. Supplementary data
Supplementary data related to this article can be found at https://doi.
org/10.1016/j.eclinm.2022.101808.
References
1Fink M, Taylor MA. Catatonia: a clinician’s guide to diagnosis and
treatment. Cambridge Univ Pr; 2006.
2Northoff G, Koch A, Wenke J, et al. Catatonia as a psychomotor
syndrome: a rating scale and extrapyramidal motor symptoms. Mov
Disord. 1999;14:404–416.
3 Kahlbaum KL. Die Katatonie: oder das Spannungsirresein, eine klini-
sche Form psychischer Krankheit; 1874. https://archive.org/details/
39002079238854.med.yale.edu. Accessed August 18, 2017.
4Kendler KS. The development of Kraepelin’s mature diagnostic
concept of catatonic dementia praecox: a close reading of relevant
texts. Schizophr Bull. 2020;46:471–483.
5American Psychiatric Association. Diagnostic and statistical manual
of mental disorders: DSM-5. Arlington, VA: Amer Psychiatric Pub
Incorporated; 2013.
6 World Health Organization. ICD-11 for mortality and morbidity
statistics. https://icd.who.int/browse11/l-m/en; 2018. Accessed
February 4, 2019.
7Rogers JP, Pollak TA, Begum N, et al. Catatonia: demographic,
clinical and laboratory associations. Psychol Med. 2021:1–11.
Articles
14 www.thelancet.com Vol 56 February, 2023
8Funayama M, Takata T, Koreki A, Ogino S, Mimura M. Catatonic
stupor in schizophrenic disorders and subsequent medical com-
plications and mortality. Psychosom Med. 2018;80:370–376.
9Niswander GD, Haslerud GM, Mitchell GD. Effect of catatonia on
schizophrenic mortality. Arch Gen Psychiatry. 1963;9:548–551.
10 la Vaque TJ. The history of EEG hans berger: psychophysiologist. A
historical vignette. J Neurother. 1999;3:1–9.
11 Boutros N, Galderisi S, Pogarell O, Segmiller F. EEG in psychoses,
mood disorders and catatonia. In: Standard electroencephalography in
clinical psychiatry. Wiley; 2011:113–132.
12 Teplan M. Fundamentals of EEG measurement. Meas Sci Rev.
2002;2:1–11.
13 Hill D. EEG in episodic psychotic and psychopathic behaviour. A
classification of data. Electroencephalogr Clin Neurophysiol.
1952;4:419–442.
14 Hill D. Clinical applications of EEG in psychiatry. J Ment Sci.
1956;102:264–271.
15 Gjessing L, Harding G, Jenner F, Johannessen N. The EEG in three
cases of periodic catatonia. Br J Psychiatry. 1967;113:1271–1282.
16 Northoff G, Pfennig A, Krug M, et al. Delayed onset of late
movement-related cortical potentials and abnormal response to
lorazepam in catatonia. Schizophr Res. 2000;44:193–211.
17 Oldham MA. The probability that catatonia in the hospital has a
medical cause and the relative proportions of its causes: a sys-
tematic review. Psychosomatics. 2018;59:333–340.
18 Daniels J. Catatonia: clinical aspects and neurobiological correlates.
J Neuropsychiatry Clin Neurosci. 2009;21:371–380.
19 Francis A. Catatonia: diagnosis, classification, and treatment. Curr
Psychiatry Rep. 2010;12:180–185.
20 Morrison JR. Catatonia: diagnosis and management. Hosp Commu-
nity Psychiatry. 2006;26:91–94. https://doi.org/10.1176/ps.26.2.91.
21 Bhati MT, Datto CJ, O’Reardon JP. Clinical manifestations, diag-
nosis, and empirical treatments for catatonia. Psychiatry (Edgmont).
2007;4:46–52.
22 Rasmussen SA, Mazurek MF, Rosebush PI. Catatonia: our current
understanding of its diagnosis, treatment and pathophysiology.
World J Psychiatry. 2016;6:391.
23 Oldham MA, Lee HB. Catatonia vis-à-vis delirium: the significance
of recognizing catatonia in altered mental status. Gen Hosp Psy-
chiatry. 2015;37:554–559.
24 Taylor MA, Fink M. Catatonia in psychiatric classification: a home
of its own. Am J Psychiatry. 2003;160:1233–1241.
25 Luccarelli J, Kalinich M, McCoy TH, et al. The occurrence of
catatonia diagnosis in acute care hospitals in the United States: a
national inpatient sample analysis. Gen Hosp Psychiatry.
2022;77:141–146.
26 Ogyu K, Kurose S, Uchida H, Kanemoto K, Mimura M,
Takeuchi H. Clinical features of catatonic non-convulsive status
epilepticus: a systematic review of cases. J Psychosom Res.
2021;151:110660.
27 Schmitt SE, Pargeon K, Frechette ES, Hirsch LJ, Dalmau J,
Friedman D. Extreme delta brush. Neurology. 2012;79:1094–1100.
28 Weleff J, Barnett BS, Park DY, Akiki TJ, Aftab A. The state of the
catatonia literature: employing bibliometric analysis of articles from
1965-2020 to identify current research gaps. J Acad Consult Liaison
Psychiatry. 2022. https://doi.org/10.1016/J.JACLP.2022.07.002.
29 Carroll BT, Goforth HW, Boutros NN, Anfinson TJ, Werner L.
Electroencephalography in catatonic disorders due to general med-
ical conditions and psychiatric disorders. CNS Spectr. 1998;3:57–61.
30 Bell V, Wilkinson S, Greco M, Hendrie C, Mills B, Deeley Q. What
is the functional/organic distinction actually doing in psychiatry
and neurology? Wellcome Open Res. 2020;5:138. https://doi.org/10.
12688/WELLCOMEOPENRES.16022.1.
31 Takwoingi Y, Riley RD, Deeks JJ. Meta-analysis of diagnostic ac-
curacy studies in mental health. Evid Based Ment Health.
2015;18:103–109.
32 Bachmann LM, Puhan MA, ter Riet G, Bossuyt PM. Sample sizes
of studies on diagnostic accuracy: literature survey. Br Med J.
2006;332:1127–1129.
33 Whiting PF, Rutjes AWS, Westwood ME, et al. QUADAS-2: a
revised tool for the quality assessment of diagnostic accuracy
studies. Ann Intern Med. 2011;155:529–536.
34 Murad MH, Sultan S, Haffar S, Bazerbachi F. Methodological
quality and synthesis of case series and case reports. BMJ Evid
Based Med. 2018;23:60–63.
35 Dwamena BA, Sylvester R, Carlos RC. MIDAS: Stata module for
meta-analytical integration of diagnostic test accuracy studies.
RePEc. 2007:S456880.
36 Dwamena BA. midas: a program for meta-analytical integration of
diagnostic accuracy studies in Stata. 2007.
37 MacMahon JF, Walter WG. The electro-encephalogram in schizo-
phrenia. J Ment Sci. 1938;84:781–787.
38 Walter WG. Electro-encephalography in cases of mental disorder.
J Ment Sci. 1942;88:110–121.
39 Stevens JM, Derbyshire AJ. Shifts along the alert-repose continuum
during remission of catatonic stupor with amobarbital. Psychosom
Med. 1958;20:99–107.
40 Ishibashi T, Sato T, Asano S, Hiraga H. Clinical electroencepha-
lographic study of atypical endogenous psychosis. Psychiatry Clin
Neurosci. 1963;16:330–354.
41 Abenson MH. EEGs in chronic schizophrenia. Br J Psychiatry.
1970;116:421–425.
42 Philbrick KL, Rummans TA. Malignant catatonia. J Neuropsychiatry
Clin Neurosci. 1994;6:1–13.
43 Carroll BT, Boutros NN. Clinical electroencephalograms in patients
with catatonic disorders. Clin Electroencephalogr. 1995;26:60–64.
44 Smith JH, Smith VD, Philbrick KL, Kumar N. Catatonic disorder
due to a general medical or psychiatric condition. J Neuropsychiatry
Clin Neurosci. 2012;24:198–207.
45 Llesuy JR, Coffey MJ, Jacobson KC, Cooper JJ. Suspected delirium
predicts the thoroughness of catatonia evaluation. J Neuropsychiatry
Clin Neurosci. 2017;29:148–154.
46 Espinola-Nadurille M, Flores-Rivera J, Rivas-Alonso V, et al. Cata-
tonia in patients with anti-NMDA receptor encephalitis. Psychiatry
Clin Neurosci. 2019;73:574–580.
47 Ursitti F, Roberto D, Papetti L, et al. Diagnosis of pediatric anti-
NMDAR encephalitis at the onset: a clinical challenge. Eur J Pae-
diatr Neurol. 2021;30:9–16.
48 Hosmer DW, Lemeshow S, Sturdivant RX. Applied logistic regression.
3rd ed. Wiley; 2013.
49 McInnes MDF, Moher D, Thombs BD, et al. Preferred reporting
items for a systematic review and meta-analysis of diagnostic test
accuracy studies: the PRISMA-DTA statement. JAMA.
2018;319:388–396.
50 Johnson EL, Kaplan PW. Clinical neurophysiology of altered states
of consciousness: encephalopathy and coma. Handb Clin Neurol.
2019;161:73–88.
51 Chapter 8: meta-analysis of test performance when there is a “gold
standard”| Effective Health Care (EHC) Program. https://
effectivehealthcare.ahrq.gov/products/methods-guidance-tests-met
aanalysis/methods. Accessed July 5, 2022.
52 Leeflang MMG, Deeks JJ, Takwoingi Y, Macaskill P. Cochrane
diagnostic test accuracy reviews. Syst Rev. 2013;2:82.
53 Jing J, Herlopian A, Karakis I, et al. Interrater reliability of experts
in identifying interictal epileptiform discharges in electroencepha-
lograms. JAMA Neurol. 2020;77:49–57.
54 Woodruff DS, Kramer DA. Eeg alpha slowing, refractory period,
and reaction time in aging. Exp Aging Res. 2007;5:279–292.
55 Scally B, Burke MR, Bunce D, Delvenne JF. Resting-state EEG
power and connectivity are associated with alpha peak frequency
slowing in healthy aging. Neurobiol Aging. 2018;71:149–155.
56 Hirjak D, Kubera KM, Wolf RC, Northoff G. Going back to
Kahlbaum’s psychomotor (and GABAergic) origins: is catatonia
more than just a motor and dopaminergic syndrome? Schiz-
ophr Bull. 2020;46(2):272–285. https://doi.org/10.1093/schbul/sb
z074.
57 Fenton GW. The electroencephalogram in psychiatry: clinical and
research applications. Psychiatr Dev. 1984;2:53–75.
58 Ramirez-Bermudez J, Medina-Gutierrez A, Gomez-Cianca H, et al.
Clinical significance of delirium with catatonic signs in patients
with neurological disorders. J Neuropsychiatry Clin Neurosci.
2022;34:132–140.
59 Dalmau J, Tüzün E, Wu H, et al. Paraneoplastic anti-N-methyl-D-
aspartate receptor encephalitis associated with ovarian teratoma.
Ann Neurol. 2007;61:25–36.
60 Kennedy JD, Schuele SU. Neocortical temporal lobe epilepsy. J Clin
Neurophysiol. 2012;29:366–370.
61 Blumenfeld H, Rivera M, McNally KA, Davis K, Spencer DD,
Spencer SS. Ictal neocortical slowing in temporal lobe epilepsy.
Neurology. 2004;63:1015–1021.
62 Jackson A, Seneviratne U. EEG changes in patients on antipsy-
chotic therapy: a systematic review. Epilepsy Behav. 2019;95:1–9.
63 Mann SC, Caroff SN, Bleier HR, Welz WK, Kling MA, Hayashida M.
Lethal catatonia. Am J Psychiatry. 1986;143:1374–1381.
64 Agrawal AK, Das S. The catatonic pupil: an unprivileged entity.
Asian J Psychiatr. 2018;32:75–76.
Articles
www.thelancet.com Vol 56 February, 2023 15
