Assessing prognostic value of optic nerve sheath diameter and Marshall computed tomography score in severe traumatic brain injury: A prospective study

INTRODUCTION

Traumatic Brain Injury (TBI) stands as a major contributor to morbidity and mortality in India, with staggering annual estimates of 1.5 to 2 million injuries and a grim toll of approximately 1 million fatalities. The leading culprits behind these grim statistics are Road Traffic Injuries (RTIs), accounting for about 60% of cases, followed by falls (20–25%), and acts of violence (10%).¹ Pathophysiology, TBI initiates a cascade of damage: the primary injury inflicts immediate structural harm, manifesting as contusions, intracranial hemorrhages, or diffuse axonal injury. This is often compounded by secondary brain injuries, a delayed yet devastating progression driven by mechanisms like cerebral ischemia, swelling, and escalating intracranial pressure—further exacerbating outcomes.

Radiological techniques such as computed tomography (CT) scans are critical in determining the severity of TBI and may be helpful in prognostication. The Marshall CT classification and the Rotterdam CT classification score [RCTS] are central classification systems. Marshall et al. established a metric categorization based on CT scans to determine injury severity.² The condition of the basal cisterns determines the classification, the degree of midline shift, the existence of a mass lesion, and surgical evacuation criteria. Authors have independently predicted mortality rates and clinical outcomes.² Elkbuli et al. studied the utility of the Marshall and Rotterdam classifications in predicting mortality in severe TBI patients.³ They observed that the odds ratio of dying from severe TBI with both scores was ≥ 4, with p < 0.05. Thus, they concluded that high scores were associated with mortality.³ Hereby, we conducted the study with the primary objective of this study was to evaluate the prognostic value of optic nerve sheath diameter (ONSD) and Marshall CT scores in predicting mortality and clinical outcomes (Glasgow Coma Scale (GCS) at discharge) in patients with severe TBI. Additionally, we aimed to correlate these findings with invasive ICP measurements.

METHODOLOGY

This prospective, single-center clinical study was meticulously conducted from January 2024 to December 2024 at the Level 1 Trauma Center of the All-India Institute of Medical Sciences (AIIMS), Rishikesh, following rigorous ethical oversight. Approval was obtained from the Institutional Ethics Committee (AIIMS/IEC/20/718), and the trial was duly registered with the Clinical Trials Registry-India (CTRI) under registration number REF/2021/04/042529. The study cohort comprised adult patients (aged >18 years) presenting with severe TBI, as defined by GCS scores of less than 8, alongside abnormal non-contrast CT scans indicative of potentially salvageable conditions—such as intracranial hematomas, contusions, cerebral swelling, herniation, or compression of basal cisterns. These inclusion criteria aligned with the Brain Trauma Foundation (BTF) guidelines for intracranial pressure (ICP) monitoring. Exclusion criteria were carefully applied to eliminate confounding factors, specifically patients with concomitant globe injuries, pre-existing optic nerve abnormalities, or polytrauma involving abdominal injuries, as the latter could unpredictably alter measured ICP values. The sample size was based on a convenience sample of eligible patients during the study period.

Upon arrival at the emergency department, all patients were systematically managed according to Advanced Trauma Life Support (ATLS) protocols, followed promptly by a non-contrast head CT scan. Once the patient was stabilized, they underwent a non-contrast head CT scan and other relevant radiological investigations, and the Marshall CT score was applied. Subsequently, for those patients who were planning non-operative management, an external ventricular drainage (EVD) device was placed for ICP monitoring. The patients requiring surgical intervention underwent decompressive craniotomy with intraparenchymal ICP sensor placement for ICP monitoring. Post-procedure, all patients were transferred to the Trauma Intensive Care Unit (ICU) for close monitoring and standardized management. ONSD measurements were performed using a Philips ultrasound machine equipped with a high-frequency (13–6 MHz) linear array transducer. With participants positioned supine and their heads elevated to 30 degrees, the transducer was gently applied over closed eyelids (protected by a transparent covering and ultrasonic gel) at a 30-degree horizontal angle. Measurements were taken in the transverse plane, capturing the widest ONSD 3 mm posterior to the retina in both eyes. To ensure accuracy, six readings per eye were averaged. Three readings were taken per eye (total of six readings per session), which were then averaged to provide a single ONSD value for that time point. A total of 96 measurements were recorded per patient over 4 days (3 readings × 2 eyes × 4 sessions/day × 4 days). To account for repeated measurements, data were averaged per session for correlation and comparative analyses. Statistical significance was defined as p < 0.05. These were rigorously compared with concurrent invasive ICP measurements. To mitigate observer bias, ICP values were recorded by dedicated nursing staff, while the principal investigator performed all ONSD evaluations. A baseline ONSD threshold of >4.5 mm was considered indicative of elevated ICP (>15 mmHg). It was hypothesized that regular ONSD monitoring would enhance the precision of correlating dynamic changes in ONSD with ICP fluctuations throughout the monitoring period.

Data was meticulously recorded in an MS Excel spreadsheet and analyzed using Statistical Product and Service Solutions (SPSS) version 25 (IBM Corp.). Descriptive statistics were employed: continuous variables were expressed as means with standard deviations (SD) or medians with interquartile ranges (IQRs), while categorical variables were summarized as frequencies and percentages. Depending on data distribution, Pearson's or Spearman's correlation coefficients were calculated to examine linear relationships between variables, complemented by a correlogram illustrating correlation strengths (with insignificant correlations marked ×). A p-value of <0.05 was uniformly adopted as the threshold for statistical significance across analyses.

RESULTS

During the study period, a total of 659 patients were admitted, of whom 120 were diagnosed with severe TBI. After excluding 42 patients with associated globe injuries, 29 with intra-abdominal injuries, and 15 whose attendants declined consent, the final cohort comprised 24 male patients, with over 50% under 30 years of age. RTI was the most common mechanism of injury, followed by falls from height. Upon arrival, the mean GCS score was 6.29, indicating severe TBI. Ten patients underwent frontotemporal decompressive craniotomy, while fourteen were managed conservatively.

A significant positive correlation was observed between ONSD and ICP (p < 0.05). On the right side, mean ONSD values were 0.40 ± 0.03 cm (ICP <15 mmHg), 0.46 ± 0.09 cm (ICP 15–22 mmHg), and 0.59 ± 0.10 cm (ICP >22 mmHg), with a moderate correlation (rho = 0.45). Similarly, on the left, mean ONSD values were 0.41 ± 0.04 cm, 0.47 ± 0.08 cm, and 0.59 ± 0.07 cm, respectively, with a moderate correlation (rho = 0.49, p <0.001) [Table 1]. The distribution of CT Marshall grades was as follows: 12.5% Grade 1, 16.7% Grade 2, 45.8% Grade 3, and 25.0% Grade 4. Correlations with the Marshall Score were also explored. A weak, non-significant negative correlation was found between initial GCS and Marshall Score (rho = −0.12, p = 0.583) [Figure 1], while a moderate, non-significant negative correlation existed with final GCS (rho = −0.38, p = 0.068) [Figure 2]. The final GCS is the GCS at the time of discharge/death. Baseline ICP showed a mild positive correlation (rho = 0.38, p = 0.070) [Figure 3], and weak correlations (rho = 0.20– 0.39) were noted at various ICP measurement intervals, though none were statistically significant. No correlation was seen at 90 hours [Table 1]. Combining ONSD >4.5 mm and Marshall grade >3 yielded a sensitivity of 61.5% and a specificity of 90.9% for predicting an ICP >15 mmHg, although the results were not statistically significant [Table 2].

Close

Figure 1: Correlation between initial glasgow coma scale (GCS) and marshall score computed tomography (CT)

Export to PPT

Close

Figure 2: Correlation between final glasgow coma scale (GCS) and marshall score computed tomography (CT)

Export to PPT

Close

Figure 3: Correlation between intracranial pressure (ICP) (Baseline) and marshall score computed tomography (CT)

Export to PPT

The mean hospital stay was 24.04 ± 16.31 days (range: 7–53 days), with a weak, non-significant positive correlation (rho = 0.16, p = 0.461) between stay duration and Marshall score. Overall, 66.7% of participants were discharged, and 33.3% died. The Marshall score did not differ significantly between the discharge (mean 2.75 ± 1.00) and death (mean 3.00 ± 0.93) groups (W = 54.000, p = 0.536) [Figure 4].

Close

Figure 4: Association between outcome and marshall score computed tomography (CT)

Export to PPT

DISCUSSION

TBI remains a leading cause of disability among trauma victims in India, imposing a substantial public health burden. Epidemiological data indicate that the age group most vulnerable to TBI is 20–29 years, followed closely by those aged 30–39 years.¹ Corroborating these trends, our study observed a similar age distribution, with 33.3% of patients (eight individuals) falling within the 21–30-year bracket, and 25% (six patients) in the 31–40-year age range, underscoring the impact on young, economically productive populations. In terms of etiology, RTIs are the predominant cause of TBI in India, accounting for approximately 60% of cases, followed by falls (20%–25%) and violence (10%).¹ Our findings align with this pattern, identifying RTIs as the leading mechanism of injury (54.2%), followed by falls from height (37.5%). Given the severity of such injuries, accurate monitoring of ICP becomes critical, particularly when clinical neurological assessment is compromised or when the risk of elevated ICP is high, as emphasized in the BTF guidelines.⁴ ICP monitoring enables timely interventions that can mitigate secondary brain insults, thereby improving outcomes. For instance, an extensive retrospective study by Farahvar et al involving 2,134 patients with severe TBI demonstrated that ICP-directed therapy, guided by monitoring, was associated with significantly reduced mortality, with 1,202 patients monitored versus 244 without monitoring.⁵ While treatments such as osmotic diuretics and hyperventilation are effective in lowering ICP, their prolonged use carries risks, including electrolyte imbalances and ischemic damage. Thus, ICP monitoring not only guides titration of these interventions but also refines management, a strategy supported by multiple studies showing decreased mortality in head-injured patients with ICP surveillance.^6–8

The management of severe TBI in resource-constrained settings has been explored in notable studies, such as the benchmark evidence from south American trials: Treatment of intracranial pressure (BEST-TRIP) trial (2014), which demonstrated that in low-resource countries, patients can be effectively monitored and treated using frequent clinical assessments and CT evaluations, even without invasive ICP monitoring.⁹ Building on predictive strategies, Alali et al developed a clinical decision rule to identify intracranial hypertension in severe TBI, incorporating Marshall CT grades III and IV as major criteria and grade II as a minor criterion.¹⁰ Their model demonstrated high sensitivity (93.9%) but low specificity (42.3%), suggesting its utility as a screening tool for selecting patients for ICP monitoring.¹⁰ In our study, Marshall Grade 3 was most prevalent (45.8%), with an overall mean score of 2.83 ± 0.96. We observed a moderate positive correlation between baseline ICP and Marshall Score (r = 0.40), although this did not reach statistical significance (p = 0.070), suggesting a potential but variable relationship.

In contrast, Miles et al. focused on pediatric TBI. They identified predictors of intracranial hypertension, noting that hypoxia, female sex, and higher Injury Severity Scores (ISS) were stronger indicators than CT classifications (Rotterdam or Marshall scores).¹¹ Extending this, our analysis of ICP at staggered intervals versus Marshall scores revealed only weak correlations (r = 0.20–0.39), with negligible associations at specific time points (36, 48, 54, 60, 78, 84, 96 hours) and no correlation at 90 hours, all of which lacked statistical significance. These findings suggest that the reliability of Marshall grading alone in tracking dynamic ICP changes is limited.

Encouragingly, Majeed et al reported that combining CT-based ONSD ≥6.0 mm with Marshall grade ≥3 yielded high sensitivity (92.5%) and specificity (92.6%) for predicting elevated ICP, proposing a refined criterion for monitoring.¹² Aligning with this, our study found that pairing ONSD >4.5 mm with a Marshall grade >3 achieved a sensitivity of 61.5% and a specificity of 90.9% for an ICP >15.

LIMITATIONS

The study focused exclusively on adult patients with severe TBI, deliberately excluding pediatric patients to maintain homogeneity in the cohort. The decision to limit inclusion to severe TBI cases, rather than incorporating moderate TBI, was made to avoid potential confounding variables that could broadly influence the study's results, ensuring a more precise evaluation of the parameters under investigation. Also, the study had a small sample size and a lack of multivariable adjustment for demographic variables. Future research on a larger scale is needed to establish definitive protocols.

CONCLUSION

We strongly recommend incorporating ONSD assessment into the routine monitoring of all TBI patients, given its noninvasive nature, straightforward learning curve, and robust correlation with ICP. The initial CT scan findings alone can serve as a valuable predictor of elevated ICP; however, combining these with ONSD measurements significantly enhances predictive accuracy. This integrated approach is particularly advantageous in resource-limited settings, such as India, where accessibility to advanced monitoring tools may be constrained, making it a practical and efficient strategy for improving patient care.