This was a prospective, observational study following the Helsinki Declaration and approved by the institutional review board. Patients older than 18 years who were admitted to the ICU of the Hospital de Clínicas and Asociación Española Primera en Salud, Montevideo, Uruguay, from February 1, 2016 to March 31, 2018 were included.

The exclusion criteria were patients with previous specific traumatic or structural ophthalmological pathology.

Upon admission to the ICU, consent was obtained from patients’ family members. Disease severity was determined following the Glasgow Coma Scale. Electronic medical records were reviewed to obtain demographics (age, gender, race), vital signs (heart rate, respiratory rate, blood pressure, temperature), and hemodynamic parameters (if available).

All patients were sedated with midazolam and fentanyl administered intravenously, according to the treating physician's discretion. Following the decision to place an ICP monitoring device (by the attending intensive care physician and neurosurgeon), the patients were enrolled in this study.

ICP was measured directly through a subdural or ventricular screw or catheter inserted by a neurosurgeon. Intracranial hypertension was considered present when the patient had ICP above 20 mmHg for a period of time greater than 5 min.

The operators measuring the ONSD with ultrasonography were blinded to invasive measurements of ICP.

Ocular ultrasound was performed by physicians trained in this technique and according to pre-established protocols [9,10,11,12,13]. On each patient, measurements were performed by a single operator.

The learning curve for optic nerve ultrasound appears to be rapid. In the case of an experienced sonographer, it has been described in the literature that 10 measurements with three abnormal examinations are sufficient to develop technical proficiency. In the case of less experienced operators, up to 25 examinations may be necessary [14, 15].

Retrobulbar ONSD measurement has become possible thanks to the introduction of high-frequency transducers and ultrasound (US) units with a spatial resolution of less than 0.4 mm [16]. According to current protocols, high-frequency linear probes (> 7.5 MHz) should be used to visualize and evaluate the Optic Nerve sheath [13, 17].

All measurements used were carried out with the use of portable, cart-based US equipment, with a linear transducer, at a frequency of 7.5–15 MHz.

The ultrasound presets were established via the "small parts" mode. This setting produces a default field of 4–5 cm depth.

Following FDA regulations for the mechanical index and thermal index for ophthalmic US, we used values less than 0.23 for the mechanical index (MI) and less than 1.0 for the thermal index (TI). Although we used the small parts preset, we lowered the mechanical index below 0.23 by varying the acoustic power [13].

For all patients, the ultrasound beam was focused on the retrobulbar area and the gain was adjusted to obtain the optimal contrast between the optic nerve and the periorbital fat.

The patients were evaluated in the supine position, with the head elevated at 35º. The transducer was placed on the upper eyelid with the patient’s eyes closed, until a hypoechoic line with clearly defined margins posterior to the eyeball was observed.

The probe was always placed gently with the eyelids closed, never on the cornea or sclera, to avoid abrasions or other injury.

The ONSD is measured 3 mm behind the retina (Fig. 1), from which point a transverse line is drawn from the inner edge to the inner edge of both vertical hypoechoic lines (inner edge of the dura mater). This has come to be considered the point at which the maximum widening of the ONSD occurs due to the effect of increased ICP [9]. The optic disc is seen as a hyperechoic line at the posterior pole of the eyeball. The normal ultrasound appearance of the optic nerve is considered from the center to the peripheral: hypoechoic nerve fibers closely surrounded by the pia mater with echogenic appearance; the subarachnoid space appears anechogenic or hypoechoic and is surrounded by a hyperechoic dura mater and periorbital fat [7, 13, 18].

ONSD ultrasound. The ONSD is measured 3 mm behind the retina, from which point a transverse line is drawn from the inner edge to the inner edge of both vertical hypoechoic lines (inner edge of the dura mater)

For each optic nerve two measurements were made—one in the sagittal plane and the other in the transverse plane—by rotating the probe clockwise (Fig. 2). The mean value obtained for both eyes was retained as the final ONSD value.

On the left, transducer with transversal orientation for ONSD measurement. To the sagittal flat right with the transducer in the longitudinal direction

In accordance with the Independence assumption, a single ONSD measurement was taken for each patient. The measurements were performed in the ICU after insertion of invasive ICP monitoring.

In all cases, the ICP reading was collected at the end of the ocular ultrasound in order to avoid bias in the measurement.

For the statistical analysis of our data, tables and numerical summaries were made, such as the mean, standard deviation (SD), and range, of the different variables. Using logistic regression models, the relationship between possible predictors and the dichotomous variable ICP > 20 mmHg was studied. Using a simple linear regression model, the relationship between the continuous variables ICP and ONSD was analyzed. Likewise, diagnostic graphs of the model were made.

The ROC curve was graphed and the cut-off points for intracranial hypertension were analyzed by measuring the ONSD. Using bootstrap (non-parametric method) the confidence interval for the area under the curve (AUC) was constructed. Data analysis was performed using the programming language and statistical software R (version 3.6.1).