Outcome Prediction in Aneurysmal Subarachnoid Hemorrhage with World Federation of Neurological Societies grade V(OPAS-V)

Shuhei Yamada, Takeo Nishida, T Takenaka, Hiroki Yamazaki, Ryota Nakagawa, Masatoshi Takagaki, Yoshihiro Yano, Hajime Nakamura, Shingo Toyota, Toshiyuki Fujinaka, Takuyu Taki, Toshiaki Fujita, Haruhiko KishimamedRxiv 2023 · DOI

Abstract

Aneurysmal subarachnoid hemorrhage (aSAH) classified as World Federation of Neurological Societies (WFNS) grade V is associated with high mortality and poor prognosis. While aggressive treatment can lead to favorable outcomes in a subset of these patients, clinicians lack reliable tools to predict recovery at the time of admission. This study developed OPAS-V, a decision tree model designed to predict clinical outcomes for grade V aSAH patients. Using a derivation cohort of 201 patients and a validation cohort of 12, the model utilized three routine metrics: lymphocyte percentage, age, and the glucose-to-potassium ratio. The model achieved an Area Under the Curve (AUC) of 0.828 in the derivation cohort and 0.700 in the validation cohort, suggesting that simple admission findings can effectively aid in determining treatment strategies for severe aSAH.

Introduction

Aneurysmal subarachnoid hemorrhage (aSAH) remains a devastating neurological event. Patients presenting with WFNS grade V—the most severe classification—often arrive in a comatose state, facing mortality rates that frequently exceed 50%. Due to this bleak outlook, the decision to pursue "maximal treatment," such as aggressive surgical intervention, is often fraught with ethical and clinical uncertainty.

However, existing literature suggests that between 14% and 25% of grade V patients can achieve a good functional recovery if treated aggressively. The primary challenge for neurosurgeons is the "admission gap": the inability to distinguish, at the moment of hospital entry, which patients belong to this salvageable minority. This study introduces the OPAS-V model to bridge that gap using readily available clinical data.

Methods and Model Development

The researchers conducted a multicenter study across four Japanese institutions, analyzing a derivation cohort of 201 patients (admitted 2012–2021) and a prospective validation cohort of 12 patients (admitted 2021–2022).

To build the predictive model, the team employed the Classification and Regression Tree (CART) method. CART is a non-parametric algorithm specifically chosen for its ability to identify non-linear relationships and "tipping points" within complex clinical data without requiring the rigid assumptions of standard linear regression. The primary outcome was the Modified Rankin Scale (mRS) at discharge, with a "good outcome" defined as functional independence (mRS 0–2).

The OPAS-V Decision Logic

Despite evaluating a wide array of radiological and clinical variables, the CART algorithm identified just three routine metrics as the most powerful predictors of recovery:

  1. Lymphocyte Percentage: If lymphocytes are > 49.9%, the probability of a good outcome is significantly higher (50% in this high-risk population).
  2. Age: For those with lower lymphocyte counts, age becomes the primary filter. Patients over 50 years old faced an overwhelmingly poor prognosis (97% poor outcome rate).
  3. Glucose-to-Potassium Ratio: For patients under 50 with low lymphocyte counts, a glucose/potassium ratio < 3.2 served as a positive indicator, pointing toward a 70% chance of a good outcome.

This hierarchy suggests that the body’s systemic inflammatory and metabolic response to "Early Brain Injury" (EBI)—represented by lymphocyte levels and glucose/potassium shifts—is a more granular predictor of recovery than traditional neurological scales alone.

Results and Performance

The OPAS-V model demonstrated strong predictive capabilities:

  • Derivation Cohort: The model achieved an AUC of 0.828 (95% CI: 0.712–0.944) and an overall accuracy of 93.0%.
  • Validation Cohort: In the prospective test group, the model maintained an AUC of 0.700 and an accuracy of 83.3%.

Multivariate analysis further confirmed that age (p=0.001), the presence of intracerebral hemorrhage (p=0.005), and lymphocyte percentage (p=0.013) were the most robust independent predictors of poor outcomes in this population.

Discussion and Clinical Implications

The significance of OPAS-V lies in its simplicity. By relying on routine laboratory findings obtained at admission, it provides a data-driven framework for clinicians to discuss prognosis with families and make informed decisions regarding surgical intervention. The inclusion of inflammatory markers like lymphocytes highlights the growing understanding that the systemic response to the initial hemorrhage is as critical to the final outcome as the primary injury itself.

Limitations

While promising, the study has limitations. The prospective validation cohort was small (n=12), and the study was conducted entirely within Japanese medical centers, which may limit its immediate generalizability to different demographic groups or healthcare systems. Additionally, the model focuses on outcomes at hospital discharge; because neurological recovery can continue for months or years, the long-term potential of some patients may be underestimated. Finally, the model was built using patients who received surgery, meaning the outcomes for those deemed "too sick for intervention" remain a variable outside this current logic.

Conclusion

OPAS-V represents the first decision tree model specifically tailored for WFNS grade V aSAH. By distilling complex clinical presentations into three simple metrics, it offers a practical tool for the high-stakes environment of emergency neurosurgery, potentially identifying patients who, despite their critical state, have a meaningful chance at recovery.

Summary updated

Abstract

Abstract Background Aneurysmal subarachnoid hemorrhage (aSAH) with World Federation of Neurological Societies (WFNS) grade V has a high mortality rate and poor prognosis. Some patients with WFNS grade V aSAH have had good outcomes after aggressive treatment; however, outcome predictions based on routine examinations and findings obtained at admission are yet to be reported. This study aimed to develop a decision tree model for predicting outcomes of patients with WFNS grade V aSAH to aid decision-making for treatment strategy. Methods A multicenter study with retrospective and prospective data collected from 201 (derivation cohort) and 12 (validation cohort) patients with WFNS grade V aSAH, respectively, was conducted. Clinical outcomes were divided into good (Modified Rankin Scale [mRS] score at the time of discharge: 0–2) and poor (mRS score: 3–6) outcomes. A decision tree model was developed for the derivation cohort using the classification and regression tree method with clinical data including laboratory findings; it was named OPAS-V (Outcome Prediction in Aneurysmal Subarachnoid hemorrhage with WFNS grade V). The performance of the model was evaluated by area under the curve (AUC) and overall accuracy in both cohorts. Results OPAS-V comprised 3 metrics; the percentage of lymphocytes (&lt;49.9% or not), age (&gt;50 yrs or not), and glucose to potassium ratio (≥3.2 or not). The model achieved an AUC of 0.828 (95% confidence interval: 0.712–0.944) and overall accuracy of 0.930. Moreover, the model performed well in the validation cohort with an AUC of 0.700 (95% confidence interval: 0.200–1) and overall accuracy of 0.833. Conclusions This study developed the first decision tree model for predicting outcomes of patients with WFNS grade V aSAH, based on simple findings obtained at admission. This may aid clinicians in determining treatment strategies for severe conditions such as WFNS grade V aSAH.
BibTeX
@misc{yamada2023outcome,
  author = {Shuhei Yamada and Takeo Nishida and T Takenaka and Hiroki Yamazaki and Ryota Nakagawa and Masatoshi Takagaki and Yoshihiro Yano and Hajime Nakamura and Shingo Toyota and Toshiyuki Fujinaka and Takuyu Taki and Toshiaki Fujita and Haruhiko Kishima},
  title = {{Outcome Prediction in Aneurysmal Subarachnoid Hemorrhage with World Federation of Neurological Societies grade V(OPAS-V)}},
  year = {2023},
  doi = {10.1101/2023.06.07.23291115},
  howpublished = {medRxiv},
}
Figure