Intensive Care Med (2018) 44:1284–1294

https://doi.org/10.1007/s00134-018-5305-7

SYSTEMATIC REVIEW

Optic nerve sheath diameter measured

sonographically as non‑invasive estimator
of intracranial pressure: a systematic review
and meta‑analysis
Chiara Robba1,2* , Gregorio Santori3 , Marek Czosnyka4,5, Francesco Corradi6, Nicola Bragazzi7,
Llewellyn Padayachy8, Fabio Silvio Taccone9 and Giuseppe Citerio10

© 2018 Springer-Verlag GmbH Germany, part of Springer Nature and ESICM

Summary
Purpose: Although invasive intracranial devices (IIDs) are the gold standard for intracranial pressure (ICP) measure-
ment, ultrasonography of the optic nerve sheath diameter (ONSD) has been suggested as a potential non-invasive ICP
estimator. We performed a meta-analysis to evaluate the diagnostic accuracy of sonographic ONSD measurement for
assessment of intracranial hypertension (IH) in adult patients.
Methods: We searched on electronic databases (MEDLINE/PubMed®, ­Scopus®, Web of ­Science®, ­ScienceDirect®,
Cochrane ­Library®) until 31 May 2018 for comparative studies that evaluated the efficacy of sonographic ONSD vs. ICP
measurement with IID. Data were extracted independently by two authors. We used the QUADAS-2 tool for assessing
the risk of bias (RB) of each study. A diagnostic meta-analysis following the bivariate approach and random-effects
model was performed.
Results: Seven prospective studies (320 patients) were evaluated for IH detection (assumed with ICP > 20 mmHg
or > 25 ­cmH2O). The accuracy of included studies ranged from 0.811 (95% CI 0.678‒0.847) to 0.954 (95% CI
0.853‒0.983). Three studies were at high RB. No significant heterogeneity was found for the diagnostic odds ratio
(DOR), positive likelihood ratio (PLR) and negative likelihood ratio (NLR), with I2 < 50% for each parameter. The pooled
DOR, PLR and NLR were 67.5 (95% CI 29‒135), 5.35 (95% CI 3.76‒7.53) and 0.088 (95% CI 0.046‒0.152), respectively.
The area under the hierarchical summary receiver-operating characteristic curve (AUHSROC) was 0.938. In the subset
of five studies (275 patients) with IH defined for ICP > 20 mmHg, the pooled DOR, PLR and NLR were 68.10 (95% CI
26.8‒144), 5.18 (95% CI 3.59‒7.37) and 0.087 (95% CI 0.041‒0.158), respectively, while the AUHSROC was 0.932.
Conclusions: Although the wide 95% CI in our pooled DOR suggests caution, ultrasonographic ONSD may be a
potentially useful approach for assessing IH when IIDs are not indicated or available (CRD42018089137, PROSPERO).
Keywords: Intracranial pressure, Invasive intracranial devices, Optic nerve sheath diameter, Ultrasonography, Adult
patients, Meta-analysis

*Correspondence: kiarobba@gmail.com
2
Neurosciences Critical Care Unit, Addenbrooke’s Hospital, University
of Cambridge, Box 1, Hills Road, CB2 0QQ Cambridge, UK
Full author information is available at the end of the article
Chiara Robba and Gregorio Santori contributed equally as first authors

Introduction
Increased intracranial pressure (ICP) is an important
cause of secondary brain injury, and its association with
poor outcome has been extensively demonstrated [1].
Therefore, monitoring and treatment of ICP are invalua-
ble in the management of neurosurgical and neurological
critically ill patients. The gold standard for ICP measure-
ment is invasive techniques (i.e., ventriculostomy and
intraparenchymal microtransducers), associated with
risks, such as infection and haemorrhage [2, 3]. Thus,
an accurate non-invasive method to measure elevated
ICP would be most appropriate in the management of
critically ill patients. Several non-invasive techniques for
the evaluation of ICP have been developed [4], includ-
ing brain imaging techniques based on morphological
changes associated with ICP increases [such as magnetic
resonance imaging (MRI), computed tomography (CT)]
and indirectly transmitted ICP (fundoscopy, tympanic
membrane displacement).
Although some non-invasive methods have been
shown to be promising, none of them has come to
replace the invasive methods [4]. Among non-invasive
methods, the measurement of optic nerve sheath diam-
eter (ONSD) has gained particular interest. The sheath
enveloping the optic nerve is in continuity with the dura
mater of the brain and the subarachnoidal interspace
filled with cerebrospinal fluid (CSF), accounting for a
direct connection between the two compartments. As
the ONS is distensible, CSF pressure variations influ-
ence the volume of ONSD with fluctuations in the ante-
rior retrobulbar compartment, about 3 mm behind the
globe [5]. ONSD measurement for ICP detection based
on CT, MRI and ultrasound was previously studied [6, 7].
Ultrasonography has the potential advantage of being a
repeatable, bedside safe and low-cost tool, with no risks
of radiations and side effects. Sonographic ONSD meas-
urement is performed using a linear probe, which is care-
fully placed on the closed upper eyelid without exerting
pressure on the eye, with patient in the supine position.
In the two-dimensional mode, ONSD is then measured
3 mm behind the globe using an electronic calliper.
According to some authors, there is a linear relation-
ship between perioptic CSF pressure and ICP [6, 8],
and ONSD changes almost directly with ICP showing
good sensitivity and specificity [9]. Nevertheless, some
authors reported poor correlation with low sensitiv-
ity and specificity [10], while there is a substantial het-
erogeneity regarding the cut-off (ICP > 20 mmHg, > 25
or > 20 cmH2O) for defining intracranial hypertension
(IH). To date, the clinical validity/reliability of using
ONSD as a non-invasive ICP estimator has not been fully
established, although in 2011 Dubourg et al. published
the first meta-analysis on this topic [11], and since then
1285
Take home message
Optic nerve sheath diameter (ONSD) is gaining popularity as an easy,
bedside, non-invasive ICP estimator
The available evidence, summarised in this systematic review and
meta-analysis, suggests its potential utility for estimating increased
intracranial pressure non-invasively in critically ill patients
several authors extended the use of this technique [9, 12].
However, with an updated approach, diagnostic meta-
analyses concerning ONSD as a potential ICP estimator
can be improved in many ways (study selection criteria,
evaluation of cut-off values for defining IH, effect size,
statistical methods) adding new elements at the previ-
ous one. Therefore, we performed a systematic review
and meta-analysis following more stringent criteria for
the selection of studies and management of the cut-off
values assumed for IH in adult patients, with the aim to
provide an updated and reliable synthesis of the available
evidence.
Methods
We adhered to the Preferred Reporting Items for System-
atic Reviews and Meta-Analysis of Diagnostic Test Accu-
racy (PRISMA-DTA) [Electronic Supplementary Material
(ESM) Table 1A, B] [13]. We registered our protocol with
the International Prospective Register of Systematic
Reviews (PROSPERO) on 22 February 2018 (registration
number: CRD42018089137).1
Data sources and search strategy
A systematic literature search was performed for the
period 1 January 1980–31 December 2017 and further
MEDLINE/PubMed®, ­Scopus®, Science Citation I­ndex®
updated until 31 May 2018 in the following databases:
Science®, ­ScienceDirect® and
®
Expanded from Web of ­
Cochrane ­Library . The search was primarily set by
including only original studies published in English in
peer-review journals. A control search without language
limitations was also performed. More details regarding
the search strategy, terms and queries are presented in
ESM Table 2.
Study screening and selection
The studies were independently screened by two authors
(CR, GS) looking at study titles and abstracts for potential
eligibility. Screening questions were answered and pilot
tested with a subset of records before implementation.
1
The protocol of this study was prospectively developed with the differ-
ent steps started on 1 December 2017 (as pointed out in the PROSPERO
“Anticipated or actual start date” section) and was submitted with some
delay to the PROSPERO staff, which, after revision, did not find any flaw or
methodological incoherence. Available at: https​://bit.ly/2ISbI​bp.
1286
Disagreement was assessed using κ statistics and was
resolved through discussion; a third reviewer (FC) was
involved if necessary. We included only studies where ICP
assessment with ultrasonographic ONSD was compared
with invasive (intraparenchymal, intraventricular or lum-
bar puncture) measurement. The invasive ICP measure-
ment (IICPM) was assumed as the standard reference.
Studies conducted in healthy volunteers or in animal mod-
els were excluded, as were any other publications lack-
ing abstract and/or quantitative details. When multiple
publications of the same research group/center described
potentially overlapping case series, we used the more
recent publication, if eligible. We excluded any contact
with the authors of the original studies to obtain essential
information not available in the published format. More
details about extracted data are reported in ESM Table 3.
We established further inclusion criteria to re-eval-
uate the first pool of studies preliminarily selected for
the quantitative synthesis: (1) the diagnostic accuracy of
ONSD ultrasonography should be compared with IICPM
in adult patients (> 18 years old); (2) the IH should be set
with ICP > 20 mmHg or > 25 cmH2O; (3) sonographic
ONSD and IICPM should be simultaneously performed
in the same patients or with a time window formally/
technically < 1 h; (4) the sample size should be ≥ 15
patients; (5) enough information should be reported to
construct a 2 × 2 table; (6) the correlation coefficient (r)
between mean ONSD and IICPM should be reported.
Only studies that met all the above criteria were enrolled
for quantitative synthesis.
Appraisal of study quality
The overall quality of evidence at the outcome level was
assessed according to the Grading of Recommendations,
Assessment, Development and Evaluations (GRADE)
system [14]. Risk of bias (RB) of each included study
was assessed using the Quality Assessment of Diagnostic
Accuracy Studies (QUADAS)-2 tool [15]. Two reviewers
were contents experts (CR, FC), and one reviewer (GS)
was an experienced biostatistician/epidemiologist. The
contents experts only assessed potential publications
with respect to the appropriateness of the research ques-
tions tested. The biostatistician only evaluated the appro-
priateness of methods employed. Disagreement was
resolved by consensus.
Statistical analysis
Diagnostic accuracy of each included study was calcu-
lated from a 2 × 2 table for sensitivity (Sens), specificity,
false-positive rate (FPR), diagnostic odds ratio (DOR)
and positive and negative likelihood ratio (PLR; NLR).
The DOR was assumed as the main indicator of the whole
diagnostic performance. No study presented a full 2 × 2
table, while they returned the ROC curve (plot of sensi-
tivity vs. 1-specificity) to establish the optimal cut-off to
optimise ONSD sensitivity and specificity. A χ2 test was
carried out to assess equality of sensitivities and specifi-
cities; 95% confidence intervals (CIs) for sensitivities and
specificities were calculated with the Wilson method and
placed in forest plots. The heterogeneity of DOR, PLR
and NLR was estimated by the Cochran’s Q statistic. The
I2 was calculated to assess variability due to heterogeneity
rather than chance, assuming heterogeneity with I2 > 50%.
Since sensitivities and false-positive rates are interre-
lated, bivariate approaches to the meta-analysis of diag-
nostic accuracy [16] and their hierarchical summary
receiver-operating characteristics (HSROC) have become
the gold standard for diagnostic meta-analysis (DMA)
[17]. The bivariate approach implies a binomial model
conditional on primary studies’ true sensitivity and false-
positive rates, with a bivariate normal model for the logit-
transformed pairs of sensitivities and false-positive rates.
The final model is a linear mixed model with known vari-
ances of the random effects, similar to the computational
approach by Reitsma et al. [16], with variance compo-
nents estimated by restricted maximum likelihood. The
outputs of the bivariate diagnostic random-effects model
(pooled DOR, PLR, NLR) were calculated according to
Zwinderman and Bossuyt [18]. For the HSROC curve, we
calculated the area under the curve (AUC), partial AUC
(FPR-restricted and normalised), an average operating
point central to the observed data (θ) and curve asymme-
try (β). The HSROC curve was not extrapolated beyond
the range of the original data. Meta-regression analysis
was performed to evaluate the potential impact of pre-
defined covariates on the Sens and FPR of the bivariate
diagnostic random-effects model [19].
The presence of publication bias (PB) was evaluated by
combining trim and fill with ln DOR [20]. The trim and
fill method (TFM) estimates the number of missing stud-
ies from a meta-analysis because of the suppression of the
most extreme results on one side of the funnel plot. Then,
this method augments the observed data and recomputes
the summary estimate based on the complete data. The
trim and fill outputs were obtained with 100 iterations.
The same studies enrolled for DMA were further evalu-
ated by assuming r as effect size. The DerSimonian-Laird
random-effects model [21] was applied to the Fisher’s
r-to-z transformed correlation coefficients. PB was esti-
mated by extensive funnel plots.
Statistical significance was assumed for p < 0.05. Sta-
tistical analysis was carried out using the R software/
environment (version 3.4.2. R Foundation for Statistical
Computing). The numerical and graphical outputs were
obtained by using mada [22], metafor [23] and metamisc
[24] R packages.
 1287

Fig. 1 PRISMA flow diagram of the search results for original studies published in English and selection of eligible studies

Results screening records, 65 studies written in English were

Study selection
The initial database searching returned 883 records
written in English. A control search without language
limitations found 20 other studies, none of which met
the enrolment criteria. After removing duplicates and
assessed for eligibility (Fig. 1). Of these, 57 did not fit the
eligibility criteria (ESM Table 4). In the group of eight
studies including 352 adult patients initially considered as
potential candidates for quantitative synthesis, one [25]
did not report either r or a correlation plot to indirectly
1288

estimate this coefficient. Thus, we finally selected seven

0.60
0.76
0.50
0.76
0.69
0.59
0.68
ICP intracranial pressure, IH intracranial hypertension, ONSD optic nerve sheath diameter, PO prospective observational, D diagnostic phase I-II validity study, O observational, PBO prospective blinded observational, EVD
studies including 320 adult patients for the quantitative

external ventricular drainage, LP lumbar puncture, IP intraparenchymal, M miscellaneous, ICH intracerebral haemorrhage, SAH subarachnoid haemorrhage, TBI traumatic brain injury, Sim simultaneously, r correlation
analysis.

Threshold (mm)
Study characteristics
The main characteristics of the selected studies are sum-

5.60
5.85
6.30
4.80
5.20
5.00
5.86
marised in Table 1. All studies used a prospective design
and enrolled consecutive adult patients with suspicion
of IH. The ICP control was made through external ven-

ICP-ONSD window
tricular drainage, lumbar puncture and/or intraparen-
chymal probes. The cut-off values for IH were placed
with ICP > 20 mmHg in five studies [12, 26–29] and

Not Sim

Not Sim

Not Sim
ICP > 25 cmH2O (18.3 mmHg) in the remaining two [30,

Sim

Sim
Sim
Sim
31]. The ONSD threshold values that optimised sensitivi-
ties and specificities evaluated by the ROC curves ranged

Pathology

ICH + SAH
from 4.80 to 6.30 mm. The procedures, probes and skills

TBI + ICH
for ONSD sonographic measurement are reported in

TBI
M
M
M
M
Table 2. The diagnostic accuracy parameters for each
included study are reported in ESM Table 5.

Male (%)
Quality of evidence and risk of bias

32
77
13
40
57
67
68
Following the GRADE system, the overall quality of evi-
dence for the included studies was assessed as very low
Agea

55.9

36.5
53

53
52
60
38
(ESM Table 6). The RB evaluated with the QUADAS-2
tool was low in three studies [12, 26, 27], unclear in one
Patients (n)

study [30] and high in three studies [28, 29, 31] (ESM
Table 7).
62
64
30
65
53
15
31
Publication bias and heterogeneity of the included studies
Table 1 Characteristics of the studies included in the quantitative synthesis

The TFM returned two simulated missing studies on the

25 ­cmH2O

25 ­cmH2O
IH cut-off

20 mmHg
20 mmHg

20 mmHg
20 mmHg

left side and no missing studies on the right side of the 20 mmHg
funnel plot, with all included studies within the pseudo-
confidence region (Fig. 2). For each study, the sensitiv-
coefficient between the means of sonographic ONSD and invasive ICP measurements
ICP control

ity and specificity of sonographic ONSD measurement

IP or EVD

IP or EVD
IP or EVD

compared with ICP by invasive devices are shown in

Fig. 3. The χ2 test for equality did not reach statistical
EVD
EVD

LP

IP

significance for sensitivities (p = 0.972) and specificities

(p = 0.375). No significant heterogeneity was found for
Design

DOR (Q = 3.507, p = 0.743; I2 = 0.00%), PLR (Q = 5.154,

PBO
PO
PO

PO
PO
PO
D

p = 0.524; I2 = 0.00%) and NLR (Q = 1.486, p = 0.966;

I2 = 0.00%).
United Kingdom
South Korea

Diagnostic meta‑analysis for the included studies

Country

France

The pooled DOR returned by the bivariate diagnostic

Spain

USA
USA
Italy

random-effects model was 67.5 (95% CI 29‒135), the

pooled PLR was 5.35 (95% CI 3.76‒7.53), and the pooled
Reported as mean or median
del Saz-Saucedo et al. 2016

NLR was 0.088 (95% CI 0.046‒0.152). The area under the

HSROC curve was 0.938 (θ = 2.346; β =1.476), whereas
Moretti and Pizzi 2009

Geeraerts et al. 2007

Kimberly et al. 2008

the partial AUC was 0.916 (Fig. 4). To evaluate the poten-
Rajajee et al. 2011
Robba et al. 2017

tial impact of predefined covariates on Sens and FPR of

Jeon et al. 2017

the bivariate diagnostic random-effects model, we per-

formed an extensive meta-regression analysis. No sta-
Study

tistical significance was found for the patient age (Sens

a
 1289

Table 2 Procedures, probes and skills described for sonographic measurement of the optic nerve sheath diameter in the
studies included in the quantitative synthesis
Study Description of US for ONSD Probe Skills described

Jeon et al. 2017
Robba et al. 2017
del Saz-Saucedo et al. 2016
Rajajee et al. 2011
Moretti and Pizzi 2009
Kimberly et al. 2008
Geeraerts et al. 2007
US ultrasonography, ONSD optic nerve sheath diameter
Temporal part of the closed ProSound Alpha 6, Hitachi Two investigators with more than 17
upper eyelid, with coupling gel, two Medical Corp., Tokyo, Japan examinations
measurements on each eye in trans- 13 MHz US probe
verse plane
3 mm behind the retina in both eyes 11L4, Xario 200; Toshiba; Zoetermeer, Three experienced operators
patients in supine position The Netherlands
The final ONSD value was calculated by 7.5-MHz linear probe
averaging four measured values
3 mm behind the lamina cribrosa, Toshiba AplioXG US system One single operator with 10 years of
patient in supine position, three meas- 4.8–11-MHz linear probe experience
urements on each eye
3 mm behind the retina, three meas- Sonosite™ M-Turbo One operator with 3 years of
urements on each eye (SonoSite Inc., Bothell, WA, USA) experience, and one operator with
6–13-MHz linear probe 2 months of experience
3 mm behind the globe, patients in Hitachi EUB 405, Hitachi Two experienced operators
supine position, two measurements Medical Corp., Tokyo, Japan
on each eye 7.5-MHz linear probe
3 mm posterior to the orbit, patients in Sonosite™ Micromax Two experienced operators
supine position, three measurements (Sonosite Inc., Bothell, WA, USA)
on each eye 5–10-MHz linear probe
3 mm behind the globe, supine posi- HP Sonos 5550, Hewlett Two investigators trained in ocular
tion, two measurements for each eye Packard, Les Ulis, France pathology
(sagittal and transverse planes) 7.5-MHz linear probe

p = 0.656; FPR p = 0.933), pathology (Sens p = 0.809;
FPR p = 0.091), ICP-ONSD window (Sens p = 0.690; FPR
p = 0.495), cut-off value assumed for IH (Sens p = 0.811;
FPR p = 0.475), ONSD threshold (Sens p = 0.857; FPR
p = 0.829), r (Sens p = 0.889; FPR p = 0.687) and RB (Sens
p = 0.813; FPR p = 0.470).
Publication bias and heterogeneity in a subset of included
studies
We repeated the analysis described above on the sub-
set of five studies (275 patients) that assumed IH for
ICP > 20 mmHg (Table 1). The TFM returned no simu-
lated missing study on either side of the funnel plot, with
all included studies within the pseudo-confidence region
(ESM Fig. 1). For each study of the subset, the sensitivity
and specificity of sonographic ONSD measurement com-
pared with ICP by invasive devices are shown in ESM
Fig. 2. The χ2 test for equality did not reach statistical
significance for sensitivities (p = 0.937) and specificities
(p = 0.212). No significant heterogeneity was found for
DOR (Q = 3.013, p = 0.556; I2 = 0.00%), PLR (Q = 4.121,
p = 0.390; I2 = 2.928%) and NLR (Q = 0.983, p = 0.912;
I2 = 0.00%).
Diagnostic meta‑analysis for a subset of included studies
In the subset of included studies that assumed IH for
ICP > 20 mmHg, the pooled DOR returned by the
bivariate diagnostic random-effects model was 68.10
(95% CI 26.8‒144), the pooled PLR was 5.18 (95% CI
3.59‒7.37), and the pooled NLR was 0.087 (95% CI
0.041‒0.158). The area under the HSROC curve was
0.932 (θ = 3.682; β = 2.188), whereas the partial AUC was
0.921 (ESM Fig. 3).
By evaluating the potential impact of predefined
covariates on the bivariate diagnostic random-effects
model for the subset of included studies, we did not find
any statistical significance in meta-regression analysis
for patient age (Sens p = 0.952; FPR p = 0.636), pathol-
ogy (Sens p = 0.779; FPR p = 0.113), ICP-ONSD win-
dow (Sens p = 0.876; FPR p = 0.607), ONSD threshold
(Sens p = 0.510; FPR p = 0.649), r (Sens p = 0.807; FPR
p = 0.851) and RB (Sens p = 0.876; FPR p = 0.607).
Additional analyses
Further analyses concerning the r between the means of
sonographic ONSD and IICPM were reported in the sup-
plementary material (ESM Sect. 1, Figs. 4, 5).
Discussion
Raised ICP can occur in multiple clinical settings and is
a common life-threatening condition. The invasiveness of
the ‘gold standard’ devices for detecting raised ICP and
the risk for severe complications (infection, haemorrhage,
malfunction) preclude their use in several situations [32].
1290
Fig. 2 Funnel plot for the trim and fill method, with original studies as solid circles and simulated studies as open circles
The measurement of ONSD for ICP detection was exten-
sively studied using MRI, CT or ultrasound [4]. Sono-
graphic ONSD has gained interest in the last years as a
promising bedside tool for the non-invasive detection of
raised ICP in a binary mode, with several authors dem-
onstrating that this technique provides good diagnostic
accuracy [12, 27], although these studies suffered from
several methodological flaws to provide definitive results.
A critical point of studies regarding ICP evaluation is
represented by the unit of measurement and the cut-off
value assumed for IH. Although IH is frequently defined
as ICP ≥ 20 mmHg [33], a non-negligible part of studies
assumes as reference value ICP ≥ 25 cmH2O, and some-
times even ≥ 20 cmH2O [9] or 30 cmH2O [35]. However,
the ICP cut-offs expressed as mmHg or ­cmH2O are not
equivalent [34, 35], thus introducing a further element of
heterogeneity.
In this study we found high pooled DOR, PLR and NLR
returned by the bivariate approach, suggesting a good
diagnostic accuracy for ONSD measured sonographically
in the detection of ICP. Moreover, the area under the
HSROC curve strengthened the reliability of sonographic
ONSD measurement for ICP detection (Fig. 4). The same
order of magnitude for the pooled DOR and area under
the HSROC curve was confirmed in the subset of five
studies that assumed IH for ICP > 20 mmHg. However, in
the whole set of included studies the forest plot for sen-
sitivity and specificity showed for this latter a relatively
lower performance (Fig. 3). The studies included in our
analysis were quite heterogeneous regarding patients’
pathology (Table 1), and this covariate was the closest to
approaching significance for the potential impact on FPR
in the meta-regression analysis. On the other hand, in the
subset of studies that established IH for ICP > 20 mmHg,
we found no significance by entering patients’ pathology
in the meta-regression for FPR. Notably, no significant
heterogeneity was found for DOR, PLR and NLR, with
I2 systematically lower to 5% in both all included studies

Fig. 3 Sensitivity (a) and specificity (b) of sonographic ONSD compared with invasive ICP measurement for diagnosis of intracranial hypertension in
each included study
Fig. 4 HSROC curve of sonographic ONSD compared with invasive
ICP measurement for diagnosis of intracranial hypertension. The
HSROC curve was not extrapolated beyond the range of the original
data
and in the subset with ICP > 20 mmHg. In the analysis
performed assuming r as effect size, we found a slightly
increased heterogeneity (ESM Sect. 1). Conversely, we
1291
found no heterogeneity by limiting the same model to the
five studies that assumed IH for ICP > 20 mmHg, suggest-
ing that the evaluation of r as effect size could be a prom-
ising additional output, especially when this ICP cut-off
value was assumed for IH.
Our results substantially confirmed the previous find-
ings of Dubourg et al. [11], although with several meth-
odological differences. Our enrolment criteria were
more protective towards the potential overlapping of the
same patients from different included studies. Moreover,
by applying the QUADAS-2 tool we assign a low [27],
unclear [30] and high [28] RB for the three shared stud-
ies with Dubourg et al., whereas these same studies were
originally assumed of high quality [11]. We entered the
RB as covariate in the meta-regression analysis, without
finding any significance. We also evaluated the over-
all quality of evidence of the included studies with the
GRADE system [14] by assigning a whole judgment of
“very low”. This finding should be reasonably extensible
to each sample of ONSD studies, considering their recur-
rent weaknesses (small sample sizes, no evaluation of
statistical power, different ICP cut-off values, clinical het-
erogeneity and wide 95% CI).
A critical point for any meta-analysis is the PB for
many potential reasons (delayed publication, selective
reporting of positive results, more impact on studies with
small sample size) [36]. For the whole set of included
1292
studies, we found that each study fell within the pseudo-
confidence region, although two simulated missing stud-
ies were returned on the left side (Fig. 2). In the subset of
studies with ICP > 20 mmHg, the funnel plot for the TFM
showed no simulated missing study, with all studies of the
subset within the pseudo-confidence region (ESM Fig. 1).
Also, in the funnel plot related to the TFM, we observed
that all selected studies fell within the pseudo-confidence
region. Moreover, the small studies (< 50 patients) did not
have better sensitivity and specificity than larger studies
(Fig. 3), thus reinforcing that PB can be excluded.
Since 2011, PROSPERO has become the reference
open-access facility to prospectively register non-
Cochrane reviews [37]. While a previous study found
no evidence that review/protocol registration reduced
outcome reporting bias [38], the PROSPERO registra-
tion is recommended for all non-Cochrane reviews [39].
We registered our protocol/review in PROSPERO with
a delay from start date. Although this asynchronism is
clearly not desirable, no discrepancy occurred between
this review and the declared primary outcome, an even-
tuality demonstrated in about a third of the systematic
reviews registered in PROSPERO [39].
Our results seem to support the use of ONSD as a
non-invasive tool for the assessment of IH in a binary
mode. Measuring ONSD with a duplex Doppler machine
is quick, safe and inexpensive; it does not require probe
fixation or specific dedicated hardware, and ultrasonog-
raphy devices are available in most emergency depart-
ments and intensive care units. However, this method
presents several limitations. Differently from invasive
ICP monitoring, ONSD is not able to return continuous
measurements. The inter-/intra-observer variability may
be high [11]. Although the learning curve for experienced
sonologists may just include from 17 to 25 scans [40, 41],
pitfalls and detection of artefacts are common during the
examination. Also, ONSD use was questioned in patients
with subarachnoid haemorrhage [42]. Therefore, ONSD
should not be used to substitute invasive devices in neu-
rocritical care patients when indications are met. How-
ever, current guidelines on TBI management are unclear
about ICP indications [43], and although all the authors
agree about the importance of assessing and managing
ICP [44, 45], there is a huge variability among different
centres concerning the indications [46]. Therefore, whilst
the superiority of invasive ICP monitoring is still advo-
cated, the sonographic ONSD measurement may be of
potential benefit in several scenarios where IICPM is not
immediately available or indicated (general ICU patients,
coagulopathy, liver failure, metabolic coma) as well as
in the intraoperative settings [47, 48] and emergency
departments [49].
Strengths and limitations
Our meta-analysis has several strengths and limitations.
Differently from the previous meta-analysis of Dubourg
et al. [11], we used pooled DOR instead of pooled sensi-
tivity and specificity to evaluate the diagnostic accuracy
of sonographic ONSD measurement for IH detection,
this latter approach being potentially misleading [50],
especially when different threshold values occur between
studies. Moreover, in both the meta-analysis of Dubourg
et al. [11] and the recent protocol of Koziarz et al. [51],
no specifications were provided regarding the methods
used for calculation of the pooled DOR, PLR and NLR.
We calculated these parameters as outputs of the bivari-
ate diagnostic random-effects model using the method
of Zwinderman and Bossuyt [18]. Furthermore, in our
meta-analysis we evaluated the PB by using the TFM
with ln DOR, which a previous simulation study rec-
ommended as the more reliable approach for evaluating
performance of the PB tests in DMA [52]. In addition,
neither the previous meta-analysis of Dubourg et al. [11]
nor the protocol of Koziarz et al. [51] provided a formal
distinction between the criteria used for assessing IH
(ICP > 20 mmHg or ICP > 25 ­cmH2O). We treated the
cut-off assumed for IH as a predefined covariate for the
meta-regression analysis, and we replicated the DMA
for the subset of included studies that assumed IH with
ICP > 20 mmHg. Finally, we also evaluated the r between
the means of sonographic ONSD and IICPM, both as a
predefined covariate for the meta-regression analysis and
as effect size in a random-effects model. Notably, we used
an open source environment for all statistical analyses,
making our study potentially reproducible [53].
The main limit of our study is the low power, although
we included more studies (seven) and patients (320)
than the meta-analysis of Dubourg et al. (six studies
for 231 patients [11]). Our patient number was slightly
higher (275) than in Dubourg et al. also in the subset of
five studies that established IH with ICP > 20 mmHg.
The wide 95% CI calculated for the pooled DOR and the
increased risk of false positives in the subgroup analy-
sis suggest caution in our estimates. Moreover, the low
number of studies included in our meta-analysis clearly
affects the reliability of the meta-regression models. We
excluded studies that did not report/allow obtaining all
the required data, without contacting the corresponding
authors; this strategy may have increased the study selec-
tion bias. The delayed PROSPERO registration represents
an additional limitation of this review. The PB evalua-
tion is elusive in DMA, especially when fewer than ten
studies are included. The TFM seems to suggest that PB
can be excluded, especially for the subset of studies that
assumed IH with ICP > 20 mmHg. A further limitation of

our study is represented by the heterogeneity of patients’
pathology, considering that the ONSD thresholds for
subarachnoid haemorrhage, intracerebral haemorrhage
and stroke were not as well defined as for traumatic brain
injury. Thus, the aim to obtain a resolutive cut-off value
for sonographic ONSD overcomes the possibilities of this
review because of the extreme heterogeneity of thresh-
olds in the included studies. However, our study suggests
that a more rigorous approach to establish IH should be
used and that the ICP > 20 mmHg seems to be associated
with more reliable results.
Conclusions
Sonographic measurement of ONSD may be a potentially
useful technique for assessing IH in a binary mode (pre-
sent/absent) when invasive/monitoring methods are not
desirable or available. However, because of the limited
number of patients and the low quality of the included
studies in our meta-analysis, larger high quality studies
are needed to establish the appropriate cut-off value for
IH and the applicability of this technique in specific clini-
cal conditions.
Electronic supplementary material
The online version of this article (https​://doi.org/10.1007/s0013​4-018-5305-7)
contains supplementary material, which is available to authorized users.
Author details
1
Anaesthesia and Intensive Care, San Martino Policlinico Hospital, IRCCS
for Oncology, Genoa, Italy. 2 Neurosciences Critical Care Unit, Addenbrooke’s
Hospital, University of Cambridge, Box 1, Hills Road, CB2 0QQ Cambridge, UK.
3
Department of Surgical Sciences and Integrated Diagnostics (DISC), Univer-
sity of Genoa, Genoa, Italy. 4 Department of Clinical Neurosciences, Division
of Neurosurgery, Brain Physics Laboratory, Cambridge Biomedical Campus,
Addenbrooke’s Hospital, University of Cambridge, Cambridge, UK. 5 Institute
of Electronic Systems, Warsaw University of Technology, Warsaw, Poland.
6
Anaesthesia and Intensive Care Unit, E.O. Ospedali Galliera, Genoa, Italy.
7
Department of Health Sciences (DISSAL), University of Genoa, Genoa, Italy.
8
Paediatric Neurosurgery Unit, Division of Neurosurgery, Faculty of Health
Sciences, Red Cross War Memorial Children’s Hospital, University of Cape Town,
Cape Town, South Africa. 9 Department of Intensive Care, Clinique Universitaire
de Bruxelles (CUB) Erasme, Brussels, Belgium. 10 School of Medicine and Sur-
gery, University of Milan-Bicocca, Milan, Italy.
Compliance with ethical standards
Conflicts of interest
None of the authors have any potential conflict of interest associated with this
study.
Ethics and dissemination
Formal ethical approval was not required as primary data were not collected.
Received: 25 April 2018 Accepted: 4 July 2018
Published online: 17 July 2018
References
1. Marmarou A, Anderson RL, Ward JD et al (1991) Impact of ICP instability
and hypotension on outcome in patients with severe head trauma. J
Neurosurg 75(Suppl 1):S59–S66
1293
2. Holloway KL, Barnes T, Choi S et al (1996) Ventriculostomy infections: the
effect of monitoring duration and catheter exchange in 584 patients. J
Neurosurg 85:419–424. https​://doi.org/10.3171/jns.1996.85.3.0419
3. Hoefnagel D, Dammers R, Ter Laak-Poort MP, Avezaat CJJ (2008) Risk fac-
tors for infections related to external ventricular drainage. Acta Neurochir
150:209–214. https​://doi.org/10.1007/s0070​1-007-1458-9
4. Robba C, Bacigaluppi S, Cardim D, Donnelly J, Bertuccio A, Czosnyka M
(2016) Non-invasive assessment of intracranial pressure. Acta Neurol
Scand 134:4–21. https​://doi.org/10.1111/ane.12527​
5. Moretti R, Pizzi B (2011) Ultrasonography of the optic nerve in neurocriti-
cally ill patients. Acta Anaesthesiol Scand 55:644–652. https​://doi.org/10.1
111/j.1399-6576.2011.02432​.x
6. Sekhon MS, Griesdale DE, Robba C et al (2014) Optic nerve sheath
diameter on computed tomography is correlated with simultaneously
measured intracranial pressure in patients with severe traumatic brain
injury. Intensive Care Med 40:1267–1274. https​://doi.org/10.1007/s0013​
4-014-3392-7
7. Geeraerts T, Newcombe VFJ, Coles JP et al (2008) Use of T2-weighted
magnetic resonance imaging of the optic nerve sheath to detect raised
intracranial pressure. Crit Care 12:R114. https​://doi.org/10.1186/cc700​6
8. Geeraerts T, Merceron S, Benhamou D, Vigué B, Duranteau J (2008) Non-
invasive assessment of intracranial pressure using ocular sonography in
neurocritical care patients. Intensive Care Med 34:2062–2067. https​://doi.
org/10.1007/s0013​4-008-1149-x
9. Amini A, Kariman H, Arhami Dolatabadi A et al (2013) Use of the sono-
graphic diameter of optic nerve sheath to estimate intracranial pressure.
Am J Emerg Med 31:236–239. https​://doi.org/10.1016/j.ajem.2012.06.025
10. Strumwasser A, Kwan RO, Yeung L et al (2011) Sonographic optic nerve
sheath diameter as an estimate of intracranial pressure in adult trauma. J
Surg Res 170:265–271. https​://doi.org/10.1016/j.jss.2011.03.009
11. Dubourg J, Javouhey E, Geeraerts T, Messerer M, Kassai B (2011)
Ultrasonography of optic nerve sheath diameter for detection of raised
intracranial pressure: a systematic review and meta-analysis. Intensive
Care Med 37:1059–1068. https​://doi.org/10.1007/s0013​4-011-2224-2
12. Robba C, Cardim D, Tajsic T et al (2017) Ultrasound non-invasive meas-
urement of intracranial pressure in neurointensive care: a prospective
observational study. PLoS Med 14:e1002356. https​://doi.org/10.1371/
journ​al.pmed.10023​56
13. McInnes MDF, Moher D, Thombs BD et al (2018) Preferred reporting items
for a systematic review and meta-analysis of diagnostic test accuracy
studies: the PRISMA-DTA statement. JAMA 319:388–396. https​://doi.
org/10.1001/jama.2017.19163​
14. Guyatt GH, Oxman AD, Vist GE et al (2008) GRADE: an emerging consen-
sus on rating quality of evidence and strength of recommendations. BMJ
336:924–926. https​://doi.org/10.1136/bmj.39489​.47034​7
15. Whiting PF, Rutjes AWS, Westwood ME et al (2011) QUADAS-2: a revised
tool for the quality assessment of diagnostic accuracy studies. Ann Int
Med 155:529–536. https​://doi.org/10.7326/0003-4819-155-8-20111​
0180-00009​
16. Reitsma JB, Glas AS, Rutjes AWS, Scholten RJPM, Bossuyt PM, Zwinder-
man AH (2005) Bivariate analysis of sensitivity and specificity produces
informative summary measures in diagnostic reviews. J Clin Epidemiol
58:982–990. https​://doi.org/10.1016/j.jclin​epi.2005.02.022
17. Liu Z, Yao Z, Li C, Liu X, Chen H, Gao C (2013) A step-by-step guide to
the systematic review and meta-analysis of diagnostic and prognos-
tic test accuracy evaluations. Br J Cancer 108:2299–2303. https​://doi.
org/10.1038/bjc.2013.185
18. Zwinderman AH, Bossuyt PM (2008) We should not pool diagnostic
likelihood ratios in systematic reviews. Stat Med 27:687–697. https​://doi.
org/10.1002/sim.2992
19. Doebler P, Holling H, Böhning D (2012) A mixed model approach to
meta-analysis of diagnostic studies with binary test outcome. Psychol
Methods 17:418–436. https​://doi.org/10.1037/a0028​091
20. Duval S, Tweedie R (2000) Trim and fill: a simple funnel-plot-based
method of testing and adjusting for publication bias in meta-analysis.
Biometrics 56:455–463. https​://doi.org/10.1111/j.0006-341X.2000.00455​.x
21. DerSimonian R, Laird N (1986) Meta-analysis in clinical trials. Control Clin
Trials 7:177–188. https​://doi.org/10.1016/0197-2456(86)90046​-2
22. Doebler P (2017) mada: meta-analysis of diagnostic accuracy. R package
version 0.5.8. https​://cran.r-proje​ct.org/web/packa​ges/mada/index​.html.
Accessed 19 June 2018
1294
23. Viechtbauer W (2010) Conducting meta-analyses in R with the metafor
package. J Stat Softw 36:1–48. https​://doi.org/10.18637​/jss.v036.i03
24. Debray T, de Jong V (2017) metamisc: diagnostic and prognostic meta-
analysis. R package version 0.1.7. https​://cran.r-proje​ct.org/web/packa​
ges/metam​isc/index​.html. Accessed 19 June 2018
25. Mehrpour M, Oliaee Torshizi F, Esmaeeli S, Taghipour S, Abdollahi S
(2015) Optic nerve sonography in the diagnostic evaluation of pseu-
dopapilledema and raised intracranial pressure: a cross-sectional study.
Neurol Res Int 2015:146059. https​://doi.org/10.1155/2015/14605​9
26. Rajajee V, Vanaman M, Fletcher JJ, Jacobs TL (2011) Optic nerve ultra-
sound for the detection of raised intracranial pressure. Neurocrit Care
15:506–515. https​://doi.org/10.1007/s1202​8-011-9606-8
27. Moretti R, Pizzi B (2009) Optic nerve ultrasound for detection of intrac-
ranial hypertension in intracranial hemorrhage patients confirmation of
previous findings in a different patient population. J Neurosurg Anesthe-
siol 21:16–20. https​://doi.org/10.1097/ANA.0b013​e3181​85996​a
28. Geeraerts T, Launey Y, Martin L et al (2007) Ultrasonography of the optic
nerve sheath may be useful for detecting raised intracranial pressure
after severe brain injury. Intensive Care Med 33:1704–1711. https​://doi.
org/10.1007/s0013​4-007-0797-6
29. Jeon JP, Lee SU, Kim SE et al (2017) Correlation of optic nerve sheath
diameter with directly measured intracranial pressure in Korean adults
using bedside ultrasonography. PLoS One 12:e0183170. https​://doi.
org/10.1371/journ​al.pone.01831​70
30. Kimberly HH, Shah S, Marill K, Noble V (2008) Correlation of optic nerve
sheath diameter with direct measurement of intracranial pressure. Acad
Emerg Med 15:201–204. https​://doi.org/10.1111/j.1553-2712.2007.00031​
.x
31. del Saz-Saucedo P, Redondo-González O, Mateu-Mateu Á, Huertas-Arroyo
R, García-Ruiz R, Botia-Paniagua E (2016) Sonographic assessment of the
optic nerve sheath diameter in the diagnosis of idiopathic intracranial
hypertension. J Neurol Sci 361:122–127. https​://doi.org/10.1016/j.
jns.2015.12.032
32. Anderson RC, Kan P, Klimo P, Brockmeyer DL, Walker ML, Kestle JR (2004)
Complications of intracranial pressure monitoring in children with
head trauma. J Neurosurg 101(Suppl 1):53–58. https​://doi.org/10.3171/
ped.2004.101.2.0053
33. Dubourg J, Messerer M, Karakitsos D et al (2013) Individual patient
data systematic review and meta-analysis of optic nerve sheath
diameter ultrasonography for detecting raised intracranial pres-
sure: protocol of the ONSD research group. Syst Rev 2:62. https​://doi.
org/10.1186/2046-4053-2-62
34. Nabeta HW, Bahr NC, Rhein J et al (2014) Accuracy of noninvasive
intraocular pressure or optic nerve sheath diameter measurements for
predicting elevated intracranial pressure in cryptococcal meningitis.
Open Forum Infect Dis 1:ofu093. https​://doi.org/10.1093/ofid/ofu09​3
35. Wang LJ, Chen LM, Chen Y, Bao LY, Zheng NN, Wang YZ, Xing YQ (2018)
Ultrasonography assessments of optic nerve sheath diameter as a
noninvasive and dynamic method of detecting changes in intracranial
pressure. JAMA Ophthalmol 136:250–256. https​://doi.org/10.1001/jamao​
phtha​lmol.2017.6560
36. Sterne JAC, Gavaghan D, Egger M (2000) Publication and related bias in
meta-analysis: power of statistical tests and prevalence in the litera-
ture. J Clin Epidemiol 53:1119–1129. https​://doi.org/10.1016/S0895​
-4356(00)00242​-0
37. PROSPERO. York, England: Centre for reviews and dissemination, Univer-
sity of York. http://www.crd.york.ac.uk/PROSP​ERO/. Accessed 19 June
2018
38. Tsujimoto Y, Tsujimoto H, Kataoka Y et al (2016) Majority of systematic
reviews published in high-impact journals neglected to register the
protocols: a meta-epidemiological study. J Clin Epidemiol 84:54–60. https​
://doi.org/10.1016/j.jclin​epi.2017.02.008
39. Tricco AC, Cogo E, Page MJ et al (2016) A third of systematic reviews
changed or did not specify the primary outcome: a PROSPERO reg-
ister study. J Clin Epidemiol 79:46–54. https​://doi.org/10.1016/j.jclin​
epi.2016.03.025
40. Ballantyne SA, O’Neill G, Hamilton R, Hollman AS (2002) Observer vari-
ation in the sonographic measurement of optic nerve sheath diameter
in normal adults. Eur J Ultrasound 15:145–149. https​://doi.org/10.1016/
S0929​-8266(02)00036​-8
41. Tayal VS, Neulander M, Norton HJ, Foster T, Saunders T, Blaivas M (2007)
Emergency department sonographic measurement of optic nerve
sheath diameter to detect findings of increased intracranial pressure
in adult head injury patients. Ann Emerg Med 49:508–514. https​://doi.
org/10.1016/j.annem​ergme​d.2006.06.040
42. Bäuerle J, Niesen WD, Egger K, Buttler KJ, Reinhard M (2016) Enlarged
optic nerve sheath in aneurysmal subarachnoid hemorrhage despite
normal intracranial pressure. J Neuroimaging 26:194–196. https​://doi.
org/10.1111/jon.12287​
43. Bratton SL, Chestnut RM, Ghajar J et al (2007) Guidelines for the manage-
ment of severe traumatic brain injury. VI. Indications for intracranial
pressure monitoring. J Neurotrauma 24(Suppl 1):S37–S44. https​://doi.
org/10.1089/neu.2007.9990
44. Robba C, Citerio G (2017) Focus on brain injury. Intensive Care Med
43:1418–1420. https​://doi.org/10.1007/s0013​4-017-4869-y
45. Asehnoune K, Balogh Z, Citerio G et al (2017) The research agenda
for trauma critical care. Intensive Care Med 43:1340–1351. https​://doi.
org/10.1007/s0013​4-017-4895-9
46. Cnossen MC, Huijben JA, Van Der Jagt M et al (2017) Variation in monitor-
ing and treatment policies for intracranial hypertension in traumatic brain
injury: a survey in 66 neurotrauma centers participating in the CENTER-
TBI study. Crit Care 21:233. https​://doi.org/10.1186/s1305​4-017-1816-9
47. Robba C, Cardim D, Donnelly J et al (2016) Effects of pneumoperitoneum
and Trendelenburg position on intracranial pressure assessed using
different non-invasive methods. Br J Anaesth 117:783–791. https​://doi.
org/10.1093/bja/aew35​6
48. Robba C, Bragazzi NL, Bertuccio A et al (2017) Effects of prone position
and positive end-expiratory pressure on noninvasive estimators of ICP: a
pilot study. J Neurosurg Anesthesiol 29:243–250. https​://doi.org/10.1097/
ANA.00000​00000​00029​5
49. Blaivas M, Theodoro D, Sierzenski PR (2003) Elevated intracranial
pressure detected by bedside emergency ultrasonography of
the optic nerve sheath. Acad Emerg Med 10:376–381. https​://doi.
org/10.1111/j.1553-2712.2003.tb013​52.x
50. Gatsonis C, Paliwal P (2006) Meta-analysis of diagnostic and screening
test accuracy evaluations: methodologic primer. AJR Am J Roentgenol
187:271–281. https​://doi.org/10.2214/AJR.06.0226
51. Koziarz A, Sne N, Kegel F et al (2017) Optic nerve sheath diameter sonog-
raphy for the diagnosis of increased intracranial pressure: a systematic
review and meta-analysis protocol. BMJ Open 7:e016194. https​://doi.
org/10.1136/bmjop​en-2017-01619​4
52. Bürkner PC, Doebler P (2014) Testing for publication bias in diagnostic
meta-analysis: a simulation study. Stat Med 33:3061–3077. https​://doi.
org/10.1002/sim.6177
53. Santori G (2016) Research papers: journals should drive data reproduc-
ibility. Nature 535:355. https​://doi.org/10.1038/53535​5b