Predicting subdural hematoma outcomes as a function of PACE score

Article information

Korean J Cerebrovasc Surg. 2026;.jcen.2026.E2025.10.003

Publication date (electronic) : 2026 January 21

doi : https://doi.org/10.7461/jcen.2026.E2025.10.003

Ryan W. Sindewald, Brian R. Hirshman, Nicholas S. Hernandez, Kiefer J. Forseth, Arvin R. Wali, Michael G. Brandel, Jeffrey S. Pannell, Jeffrey A. Steinberg, Nikhil K. Murthy, Alexander A. Khalessi, David R. Santiago-Dieppa

Department of Neurosurgery, UC San Diego School of Medicine, CA, USA

Correspondence to Ryan W. Sindewald Department of Neurosurgery, UC San Diego School of Medicine, 9500 Gilman Dr. La Jolla, CA 92093, USA Tel +1 760 532 5453 E-mail rsindewald@health.ucsd.edu

Received 2025 September 10; Revised 2025 December 18; Accepted 2026 January 6.

Abstract

INTRODUCTION

Subdural hematoma (SDH) is one of the most common neurosurgical pathologies, with an incidence expected to increase with the aging population [5]. Risk factors for SDH include male sex, coagulopathy, and advanced age [2,4]. Additionally, patients on chronic antiplatelet and anticoagulant medications are also at increased risk for SDH formation. Management of SDH may include conservative management, middle meningeal artery (MMA) embolization, and/or surgical evacuation.

MMA embolization is considered a safe and effective treatment for chronic SDH [2,7]. MMA embolization is performed using one or more embolisates, including particles (e.g., polyvinyl alcohol [PVA]), liquid agents (e.g., Onyx, n-BCA), and coils [1,2]. While the superiority of these different embolisates remains currently debated, recent investigations have shown that coils alone, or in combination with other embolisates, are a feasible method for MMA embolization [2,3,6]. The superiority of distally penetrating embolic agents (particles, Onyx, n-BCA) compared to proximally occluding agents (coils) raises speculation on the superior mechanism of embolization. Advocates of distally penetrating agents postulate that penetration of the embolizing material to capillary beds in the dura mater reduces microhemorrhages and may improve the efficacy of the procedure [2]. Advocates of proximal embolization hypothesize that proximal occlusion of the MMA may reduce local angiogenesis from pathologic neomembranes adjacent to the hematoma, which may produce a more successful result. In an effort to explore the superiority of these agents and mechanisms for embolization, our group established the Patency After Coil Embolization score‒a clinical scoring system for quantifying residual flow through the MMA following coil embolization. A PACE score is assigned after the final angiogram run following final coil deployment in MMA coil embolization. PACE 0 is defined as no flow past the coil mass, PACE 1 is defined as trace filling of frontal or parietal branches, PACE 2 is defined as filling of the frontal and parietal branches without filling of distal capillary beds, and PACE 3 is considered a failed embolization, with evidence of filling to the distal capillary beds of the dura mater. A visual representation of PACE scores can be found in Fig. 1. The study found that a PACE score of 0 was associated with significantly fewer interval acute blood products discovered on follow-up imaging when compared to PACE 1 scores, suggesting that total occlusion of the proximal MMA may improve hematoma resolution. However, the primary outcome of the previous study was binary and not designed to differentiate between scores. For this reason, comparison of resolution proportions between PACE scores is of significant interest for understanding the value of the PACE scoring system. We hypothesize that complete proximal embolization of the MMA may eliminate residual perfusion to the neomembranes of the hematoma, reduce angiogenesis and micro-hemorrhages, which in turn may lead to a higher degree of hematoma resolution at follow-up.

Fig. 1.

PACE scores 0-3 shown on angiography with corresponding follow-up CT scans. Red arrows demonstrate acute blood products within subdural collections.. PACE, Patency after Coil Embolization; CT, computed tomography

To explore this relationship, we next sought to examine the relationship between the PACE score, size of the initial hematoma prior to embolization, time to follow up, and size of the hematoma at long-term follow-up. In this study, we present a novel analysis of patients undergoing coil embolization of subdural hematoma characterized by the PACE score and found that lower the PACE scores are associated with reduced hematoma thickness as measured by the cross-sectional area of the hematoma at follow-up.

MATERIALS AND METHODS

RESULTS

Forty-three consecutive patients who met the inclusion criteria were included in the analysis, with a total of 63 SDHs when bilateral SDHs were considered separately (20 patients with bilateral SDH). Mean age (±standard deviation) for the cohort was 67.0±15.7, and 32.5% of the cohort was female (Table 1). Forty-six of 63 SDHs achieved PACE 0 embolization, and 17 had a PACE 1 embolization. A schematic of the preoperative hematoma CSA and the percent change at follow-up, arranged by the PACE score, is found in Fig 2. Histograms for the percent change of hematoma CSA separated by the PACE score can be found in Fig. 3.

Table 1.

Patient demographic information with univariate comparison between PACE score groups

Fig. 2.

Violin plot demonstrating the distribution of CSA at the time of embolization and the corresponding CSA at the time of follow-up. CSA, cross-sectional area

Fig. 3.

Histograms demonstrating the distribution of percent solutions stratified by PACE score. Complete resolution=100% hematoma resolution, Large reduction=70%-99% hematoma resolution, minimal change ＜70% hematoma resolution. PACE, Patency after Coil Embolization

A univariate analysis found that the average PACE 0 cross-sectional areas were significantly smaller compared to PACE 1 (p=0.006). Similarly, PACE 0 patients had longer follow-up time on average compared to PACE 1 patients (p=0.021). Data on SDH CSA characterized on pre- and post-embolization imaging, time to surveillance CT, and percentage resolution are found in Table 2.

Table 2.

Preoperative mean hematoma area, follow-up time, and Hounsfield units with univariate comparison between PACE score groups

Using an ROC curve and the Youden method as described above, we calculated an area under the curve (AUC) of 0.684, and the Youden method determined an optimal threshold of 0.709. The corresponding ROC curve is demonstrated in Fig. 4. Hematoma resolution was dichotomized using the calculated threshold of 0.709, and the logistic regression model was built to adjust for the PACE score, initial hematoma CSA, days to follow up, and preoperative Hounsfield units. Again, the threshold of 0.709 determined by the Youden method was used to dichotomize the hematoma as either resolved or persistent. A total of 39 patients achieved a larger resolution than the threshold, with an event per variable of 0.975 in the multivariate model. A multivariate analysis showed no relationship between hematoma resolution and initial SDH CSA (p=0.431), duration of time to follow up (p=0.749), or preoperative Hounsfield units of the hematoma (p=0.443) (Table 3). We did find that the PACE score was independently associated with hematoma resolution at follow-up in multivariate regression analysis (p=0.013) (Table 3).

Fig. 4.

Receiver operating characteristic (ROC) curve demonstrating the optimal cutoff threshold for differentiating between PACE 0 and PACE 1 SDH resolution with 95% confidence interval. PACE, Patency after Coil Embolization; SDH, subdural hematoma

Table 3.

Logistic regression model for hematoma resolution reaching the ROC calculated threshold adjusted for initial hematoma size

DISCUSSION

In this study, we sought to establish a relationship between a reduction in the cross-sectional area of subdural hematomas which were treated with coil embolization of the MMA and characterized by the previously described the PACE score. We organized hematomas by the PACE score, and as either resolved or persistent based on a threshold determined by the Youden method, then built a multivariate regression model to examine the relationship between these variables. Our results show that a PACE 0 embolization result (total proximal occlusion of the MMA without flow distal to the coil mass) is independently associated with a greater degree of hematoma resolution on follow-up imaging. We specifically sought to examine differences in hematoma resolution between PACE 0 and PACE 1 scores, as we hypothesize that complete proximal embolization of the MMA to the foramen spinosum using coils is a superior method for embolization in the treatment of subdural hematoma.

The proposed mechanism of action of MMA embolization on SDH resolution lies in decreasing blood flow and pressure to the neomembrane, reducing microhemorrhage, and ultimately allowing fibrinolysis to resolve the hematoma over time. This hypothesized mechanism is supported by a prior finding of our group that a PACE 0 score showed less acute blood products on follow-up imaging when compared to PACE 1 embolization results. It stands to reason that any degree of filling of the MMA beyond the coil mass may allow for persistent perfusion of neomembranes, may induce angiogenesis, and therefore increase the frequency of developing acute subdural blood products, and decrease the frequency of hematoma resolution.

Initial univariate analysis demonstrated significant differences in preoperative hematoma size and follow-up time between PACE 0 and PACE 1 embolizations. However, within the multivariate analysis adjusting for the PACE score, follow-up time, and initial size, the only statistically significant variable was the PACE score. The fact that SDH size was not significantly associated with the rate of resolution, when accounting for PACE score, was an interesting and unexpected result.

Given the prevalence of decreased but residual flow through the MMA in PACE 1 embolizations, the rate of resolution quantified by SDH CSA would be expected to differ from a PACE 0 embolization. The rate of hematoma resolution likely has many factors, including but not limited to initial size, time from initial imaging to follow-up, patency of MMA, and chronicity of the SDH. This can make direct comparison of rates difficult and inaccurate due to multiple confounding factors. Therefore, a multivariate model was more suitable for investigating the differences in the PACE score outcomes.

The receiver operating curve generated a cutoff for percent hematoma resolution, generated by the Youden method, was 70.9%. The method suggests that a cutoff of 70.9% hematoma resolution at the time of surveillance CT is the optimal proportion for differentiating hematoma CSA changes between PACE 0 and PACE 1 patients. After adjusting for the SDH size, time to follow up, and preoperative Hounsfield units in multivariate logistic regression, the PACE score is the most significant predictor of hematoma resolution percentage on follow-up using this method.

The 70.9% cutoff had a sensitivity 95% confidence interval of 0.412, 0.882, and a specificity 95% confidence interval of 0.587, 0.848. This threshold is used to statistically differentiate between PACE 0 and PACE 1 embolizations in this model and suggests that patients with PACE 0 embolization should have greater than 70% of the hematoma CSA resolved on follow-up while PACE 1 will have less than 70% resolved. This is in alignment with the 39.1% incidence of acute blood products found in the PACE 1 group compared to 1.92% in the PACE 0 group of the initial PACE score study, which suggested continued neomembrane perfusion and subsequent microhemorrhage at the time of follow-up. Incorporation of other important clinical information and its effects on hematoma resorption is limited due to the large sample sizes required to create multivariate models. What can be determined from the present information is that angiographically confirming and describing residual blood flow at the time of embolization is of significant prognostic value, and PACE 0 is a useful target for optimizing long-term hematoma resolution. This threshold also gives a generalized expectation of how much hematoma should be resolved by follow-up. The PACE score at the time of embolization may help determine whether or not a patient needs follow-up imaging sooner than the typical 90-day surveillance. Additionally, patients with a PACE 0 at the time of embolization and who have hematomas that resolve more slowly than expected may have developed new collaterals or had a failed embolization altogether. A follow-up angiogram with the purpose of determining the cause of delayed resolution may be considered.

PACE scores 2 and 3 are used to describe even larger residual blood flow through the coil mass, with PACE 2 defined as blood flow to both the frontal and temporal branches of the MMA, and PACE 3 defined as flow to the distal branches of the MMA, and is essentially a failed embolization. Theoretically, patients with PACE 2 or 3 scores would demonstrate even lower percentage resolution due to the effects of MMA patency on chronic SDH formation, expansion, and recurrence, though this was not specifically examined in the present study. In order to confirm this, a significantly larger sample size would be required and is of interest for future studies.

The limitations of this study include sample size, usage of a single-center institutional data set, and lack of follow-up angiography to determine neovascularization postoperatively. While larger sample sizes would result in better generalizability, the event per variable for the logistic regression is sufficient for some interpretation, especially given that none of the variables were of borderline significance. Larger sample sizes would also narrow the confidence interval of the ROC curve and thus the corresponding threshold. Follow-up angiography months after embolization is not routinely collected in this patient population, especially for those with decreasing hematoma size and improving symptoms. Due to this limitation, it’s not possible to know whether some of the PACE 1 score patients have developed collaterals following embolization, which circumvent the coil mass and therefore would likely have a similar effect as a patent MMA. However, embolizations are scored based on immediate angiographic runs. If collaterals did form in PACE 1 patients, this would still suggest that PACE 1 is of some prognostic value, given the statistical significance of the model. Similarly, antiplatelet and anticoagulant use in patients is not expected to affect the PACE score and are not included in the current model. A significantly larger sample size would be required in order to incorporate these variables in the model.

Hematoma size changes were calculated using the cross-sectional area of the largest single slice preoperatively and the CSA of the same slice at follow-up. This limits the measurement of hematoma resolution as it does not account for irregular shapes in hematomas. However, we would anticipate that hematomas in most cases would resolve uniformly, and the uniformity of resolution would be independent of the PACE score. Thus, we expect the measurement to be sufficient for demonstrating differences between categorical variables.

One additional source of potential bias is the necessity for excluding patients who were treated with a craniotomy. The purpose of the study was to measure differences in hematoma resolution rates between patients who received only coil embolization as treatment for the subdural hematoma, and thus by definition did not include patients who required evacuation. In Hernandez et al., there was no statistically significant difference between PACE scores and the incidence of craniotomy following embolization. A separate study analyzing rates of rescue surgery between PACE scores with a larger sample size would be beneficial for describing the prognostic value of the PACE score.

CONCLUSIONS

In this study, we aimed to determine an association between the PACE score and percent resolution of subdural hematomas at follow-up after coil embolization of the middle meningeal artery. We built a multivariate regression model and found that a PACE 0 embolization result was an independent predictor of at least 70.9% hematoma resolution at follow-up. This finding supports our hypothesis that complete proximal coil embolization of the MMA may be a superior embolization technique for the treatment of subdural hematoma when compared to other embolization methods.

Article information Continued