Setting and participants
This observational study was conducted in a 16-bed intensive care unit (ICU) affiliated with an academic hospital in eastern China. The research protocol was approved by the second Hospital affiliated Zhejiang University Institutional Review Board prior to the start of recruitment and data collection. Almost all of the enrolled neurocritical patients had no ability to express themselves at early stage. So after admitted to our unit, the doctor had a talk with the patients’ immediate family members (spouse, children, parents, et ac) and then they signed an informed consent about patient-related medical treatment. We screened all patients who underwent invasive ICP monitoring after a hemicraniectomy. Exclusion criteria were age < 18 years, ONSD measurement unavailable within 6 h postoperatively, ocular trauma or pre-existing ocular disease, and unsuitable optic nerve sheath images.
Study protocol
To avoid operator differences, one experienced doctor (WC) performed the ocular ultrasound for every eligible patient and then numbered and stored the ONSD images. Blinded to the ONSD measurements, another doctor (LQ) was in charge of collecting the relevant data from the ocular ultrasound examinations. Before collecting the data, we ensured that the ICP monitors had been zeroed. We followed the patients for 6 months after the injury and evaluated them using the Glasgow Outcome Score (GOS). We classified the patient outcomes as unfavorable (GOS 1, dead; GOS 2, vegetative state; GOS 3, severe disability) and favorable (GOS 4, moderate disability; GOS 5, return to normal life). Two experienced doctors (LS and GY) rechecked the ONSD images and recorded the data. Instead of averaging the values for the two eyes, we classified the ONSD measurements from both eyes into ipsilateral ONSD (craniectomy side, ONSDips) and contralateral ONSD (the side opposite the craniectomy, ONSDcon) according to the side of the craniectomy. Then, we analyzed the ONSD measurements according to the ICP and outcome.
ONSD measurement
A portable ultrasound machine (M9, Mindray, Shenzhen, China) with a linear array probe (13–6 MHz) was used. As described by Blehar et al. [15], we determined ONSD in the visual axis by placing the probe over the closed eyelid. The probe was moved slightly until the optic nerve appeared as a linear hypoechoic object with defined margins behind the globe. After freezing the screen image, we determined ONSD manually 3 mm behind the globe with mechanical calipers (Fig. 1).
Management of neurocritical patients
All postoperative neurocritical patients received standard intensive care according to the guidelines for management of TBI or intracranial hemorrhage. Ventilation was used to maintain normal oxygenation (PaO2 80–120 mmHg, PaCO2 35–45 mmHg). Midazolam, fentanyl, or sufentanil was infused continuously for sedation and analgesia. Hyperosmolar therapy with 20% mannitol or 3% hypertonic saline was applied when the ICP exceeded 20 mmHg. The cerebral perfusion pressure was maintained at 60~70 mmHg with fluids and norepinephrine as needed. Hypothermia treatment, neuromuscular blocking drugs, and barbiturates were used if indicated clinically.
Statistics analysis
Variables with normal distributions are expressed as means ± SD, and those with non-normal distributions as medians with interquartile range (IQR). Categorical variables are expressed as n (%). We used Student’s t-test or the Kolmogorov–Smirnov test to compare baseline characteristics between patients with favorable and unfavorable outcomes. We constructed Bland–Altman plots and used intraclass correlation coefficient (ICC) analysis to determine the agreement between ONSDips and ONSDcon. Spearman rank correlation was used to evaluate the association of both ONSD measurements with ICP. Receiver operator characteristic (ROC) curves were used to analyze predictors of an unfavorable outcome at 6 months (GOS 1–3). The cutoff values of predictors of a poor outcome (GOS 1–3) were determined using ROC curves. Then, we calculated the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for each cutoff. Statistical analyses were performed using MedCalc for Windows software (ver. 11.4; MedCalc Software). P-values < 0.05 were considered to be statistically significant.