Development of the Chronic Subdural Hematoma Grading System to Predict Postoperative Recurrence


 Objective Assessing the risk of postoperative recurrence of chronic subdural hematoma (CSDH) is a clinical focus. To screen the main factors associated with the perioperative hematoma recurrence. We also propose a new prognostic grading system and compare it with previous grading systems to deliver a quick and effective system.Methods We included 242 unilateral patients with CSDH as the training group for modeling. Factors predicting postoperative recurrence requiring reoperation (RrR) were determined using univariate and multivariate regression analyses. The cut-off value for the brain re-expansion rate was determined through receiver operating characteristic curve analysis. Based on these, we developed a new prognostic scoring system and conducted preliminary verification. A verification group including 119 patients with unilateral CSDH was used to verify the predictive performance of the new and other grading systems.Results The key factors for predicting unilateral CSDH recurrence were cerebral re-expansion rate (≤ 40%) at postoperative days 7 – 9 and the preoperative computed tomography density classification (isodense or hyperdense, or separated or laminar types). Cerebral atrophy played a key role in brain re-expansion. The CSDH prognostic grading system ranged from 0 to 3. An increased score was associated with a more accurate progressive increase in the RrR rate. Our grading system demonstrated the best predictive performance compared with other systems (area under the curve = 0.856).Conclusions Our prognostic grading system could quickly and effectively screen high-risk RrR patients with unilateral CSDH. However, increased attention should be paid to brain re-expansion rate after surgery in patients with CSDH.


Introduction
Chronic subdural hematoma (CSDH) is a common neurological disease in the elderly. The incidence among people over 65 years of age is 80.1/100,000/year 13 , and the average age of onset is 76.8 years 26 .
Currently, surgery is the chief mode of treatment along with the use of drugs as a supplement. Minimally invasive surgery can quickly and effectively remove the hematoma, relieve the pressure on the brain tissue, and improve the clinical symptoms of the patient. Improved surgical skills and perioperative management have signi cantly reduced the postoperative recurrence requiring reoperation (RrR) rate.
Assessment of the risk of postoperative recurrence in CSDH has been proven challenging in clinical research. The risk factors of recurrence reported so far mainly include the general clinical characteristics of the patient and the surgical methods, perioperative management methods, and imaging characteristics used 5,17,21,22,28 . Among them, imaging characteristics play a very important role.
The changes in perioperative imaging in CSDH include the following: the volume and maximum width of the hematoma, the volume and maximum width of the effusion, the distance of the midline shift, the volume of gas, the computed tomography (CT) density of the hematoma, and the effusion or signal manifestation of the MRI image of the hematoma and effusion. CT is routinely used for perioperative examination because of the ease of simple operation and the relatively low costs involved. This study retrospectively analyzed the general clinical characteristics and CT imaging parameters of patients with CSDH to determine the factors related to postoperative recurrence. The key factors were selected to establish a recurrence risk model grading system, which was compared with other previously published grading systems 8,29,32 . The aim of this study is to develop a convenient and effective recurrence grading model system for clinical use.

Methods
Patients A retrospective review of 242 patients with unilateral CSDH who were treated via surgical evacuation between July 2017 and October 2020 at The First Hospital of Jilin University was conducted. All patients were evaluated for appropriateness of surgical intervention using CT. The patients underwent burr-hole irrigation with saline under general anesthesia, following which a catheter attached to a closed-system drainage was constructed. The drainage tube was usually removed 24 h after the surgery.
The following clinical and demographic data were recorded: sex, age, atrophy, history of trauma, smoking, alcohol abuse, comorbidities (hypertension, diabetes mellitus, heart disease, and cerebral infarction), anticoagulant and antiplatelet use, coagulation evaluation (platelet count, INR, and APTT), and complication (epilepsy).
CT scanning was performed once preoperatively and twice postoperatively (before drainage tube removal on postoperative day 1 and patient discharge on postoperative days 7-9).
A picture archiving and communication system was used to gather preoperative and postoperative radiographic data, which included the preoperative hematoma characteristics (volume, density characteristics 23 , and maximal thickness), postoperative effusion characteristics (volume, density characteristics, and maximal thickness), postoperative residual air volume, pre-and postoperative midline shift, and cerebral re-expansion rate (which was calculated using the equation: preoperative hematoma volume − postoperative effusion volume/preoperative hematoma volume × 100%) from the CT scans of the head. Quantitative imaging characteristics were analyzed using the Philips IntelliSpace Discovery 3.0 software (ISD3.0, Philips, US). For quantitative volumetric analysis, the hematoma, effusion, and residual air margins were traced for each axial slice, and the volumes were calculated using the software. The quantitative image analysis was performed by a neuroradiologist who was blinded to the CSDH recurrence.
The recurrence measure was reoperation after the rst surgery due to hematoma reaccumulation by CT scan and reappearance of neurological de cits with 6 months (exclude postoeprative acute subdural hematoma).
All patients underwent a six-month follow-up. Ethical approval for this study was obtained from the Institutional Review Board of The First Hospital of Jilin University (IRB00008484). The requirement to obtain informed consent from the patients for the use of the materials was waived based on the retrospective nature of the study under the approval of the IRB.

Statistical Analyses
Research data were described using categorical variables (percentage of patients) and continuous variables (mean ± standard deviation). The risk factors of recurrence were rst performed by univariate analysis. The χ2 test or Fisher's exact test was used for categorical variables, and the Student's t-test or Man--Whitney U test was used for continuous variables. Subsequently, a logistical regression model was used for multivariate analysis, and variable selection was based on a p value of < 0.01. Regression coe cients were scaled and rounded off to obtain the weighting of variables. The ability of predicting the postoperative recurrence of CSDH was determined using the receiver operating characteristic (ROC) curve. The cut-off value was de ned as the highest sum of sensitivity and speci city calculated based on the Youden index. Finally, the scoring system for predicting the recurrence of CSDH was internally veri ed, and the ROC curve was used to compare different scoring systems. All data were analyzed using the SPSS version 22.0 software (IBM, Armonk, New York) and the MedCalc version 19.0 software (MedCalc Software, Ostend, Belgium). A two-tailed p < 0.05 was considered statistically signi cant.

Clinical Characteristics of the Patients
Of the 242 patients who underwent surgery for unilateral CSDH, 211 were males and 31 were females (age range, 21-91 years; mean age, 64.7 ± 14.4 years). Patient demographic and clinical data are shown in Tables 1. Reoperation was performed on 14 patients (5.8%) (Supplemental Table 4).

Risk Factors for RrR of CSDH
Associations between various individual variables and reoperation are show in Tables 1. Univariate analyses showed that age > 65 years, atrophy, preoperative CT classi cation based on the density of the hematoma, and postoperative radiographic factors (effusion volume, cerebral re-expansion rate, maximal effusion thickness [> 20 mm], and midline shift [> 5 mm]) were associated with a signi cantly higher RrR (Table 1). Furthermore, effusion volume, cerebral re-expansion rate, maximal effusion thickness, and midline shift demonstrated more sensitivity on days 7-9 than on day 1 (Supplemental Table 1 and Fig. 1). Multivariate analysis showed that isodense/hyperdense or separated/laminar types of CT images and cerebral re-expansion rate were independent risk factors for potential CSDH recurrence at postoperative days 7-9 (Tables 1).
Predicting the Formula of Cerebral Re-expansion In previous studies, the hematoma clearance rate was determined by calculating the change in the maximum thickness or volume of hematoma before and after surgery (Supplemental Table 2). We found that this formula could indirectly re ect the re-expansion rate in cerebral tissues. In addition, we proposed that the midline shift could re ect the cerebral re-expansion rate before and after surgery. ROC curve analysis was used to compare the abilities of three cerebral re-expansion formulas to predict RrR.
Pairwise comparisons of the AUCs of the three cerebral expansion formulas did not show any statistically signi cant differences (Supplemental Table 5 and Fig. 2). Moreover, the maximum thickness reexpansion rate was clinically easy to calculate and did not require any other software. The cerebral reexpansion rate was used based on the maximum thickness for the analysis of the clinical data in this study.

Risk Factors for Cerebral Re-expansion
The factors that affected the cerebral re-expansion rate were evaluated. ROC curve analysis was used to determine the cut-off point (41.18%). Re-expansion rates of > 40% and ≤ 40% were considered good reexpansion and partial re-expansion, respectively (Fig. 2). Univariate analysis showed that the cerebral reexpansion rate was signi cantly related to cerebral atrophy. Age (> 65 years) and brain trauma (> 30 days) were associated with cerebral re-expansion ( Table 2). Multivariate analysis showed that cerebral atrophy was the only factor related to the cerebral re-expansion rate ( Table 2).

RrR Grading System of CSDH
The preoperative hematoma density on CT according to the classi cation described by Nakaguchi et al. 23 and re-expansion at postoperative days 7-9 were identi ed as independent recurrent factors in unilateral CSDH patients (Table 3). These results of the regression modeling were the basis for the development of the new CSDH classi cation system. Risk factors related to RrR were used in a statistical selection test to identify and create the most effective model for a scoring system based on high-risk patient groups for RrR. The new grading system developed in this study consisted of the preoperative hematoma density on CT and the re-expansion at postoperative days 7-9, and scores were assigned based on the strength and regression coe cients associated with RrR. The CSDH prognostic grading system ranged from 0 to 3 (Table 4). An increase in the score was associated with a more accurate progressive increase in the RrR rate (Nagelkerke R 2 = 0.504; p < 0.001).

Internal Validation and Comparison of the RrR Model
Internal validation with the other cohorts (n = 119, Supplemental Table 3) was performed to examine the predictive power. Furthermore, the Changchun model was compared with different grading models to predict the RrR (Supplemental Table 6) 8,27,29 . The results showed that the Changchun model was more accurate than the different models in its ability to predict the RrR (p < 0.001). Likewise, the ROC curve analysis revealed that the Changchun model was better at predicting the recurrence (AUC = 0.856; Fig. 3 and Table 5).

Discussion
Several factors affect the recurrence of CSDH after surgery, including the general clinical characteristics of the patient, surgical skills, perioperative management, and imaging characteristics, which are closely related to the recurrence 5,17,21,22,28 . Owing to advancements in research, clinicians can reduce surgical complications by improving the surgical skills and strengthening perioperative management, yet cases of postoperative recurrence continue to be reported. This may be attributed to the structural characteristics of the brain tissue and the pathological characteristics of the hematoma. The brain tissue is similar to an elastic sponge. A high-quality sponge has good resilience and can reexpand quickly after decompression, whereas poor quality sponges rebound slowly after compression and have poor recruitment effects 9 . The cord separation of the hematoma cavity is likely to cause poor drainage. Furthermore, the presence of fresh blood in the hematoma uid indicates that the disease is in the active phase and might be associated with postoperative recurrence 6,21 . These characteristics can be observed by analyzing the imaging parameters during the perioperative period.
The clinical factors related to RrR were retrospectively analyzed in this study. As reported previously, age (> 65 years) was related to recurrence 3,12 . The imaging characteristics during the perioperative period play an important role in the assessment of the RrR. The univariate analysis shows that the preoperative CT classi cation, volume of effusion, midline shift, effusion thickness, and cerebral re-expansion rate after surgery were related to recurrence. These results are consistent with those reported previously 2,9,24,28 .
Cerebral re-expansion rate is calculated as the change in brain tissue volume before and after surgery. Under the condition of a xed cranial cavity volume, the effusion and volume change of a hematoma before and after surgery can indirectly re ect the cerebral re-expansion rate 11,15 . However, some studies de ned brain re-expansion rate as the change in the maximal thickness of the hematoma and the maximal thickness of the effusion before and after surgery 19,20,25 (Supplemental Table 2). The volumes of the hematoma and the effusion in the formula for the brain re-expansion rate need to be calculated using software, which is not convenient for clinical application. In this study, three formulas were compared based on the volume, maximal thickness, and midline shift ratios. The formulas demonstrated similar predictive effects, especially the ones based on the maximal thickness ratio and the volume ratio (AUC difference = 0.002; p = 0.983; Supplemental Table 5 and Fig. 2). The calculation of the maximal thickness of the hematoma and effusion does not require software assistance and can be conveniently applied in the clinical setting. Therefore, we use the formula based on the maximal thickness ratio as that for the cerebral re-expansion rate in the grading system.
Cerebral atrophy was found to affect the re-expansion of the brain tissue (p = 0.002). In addition, age (> 65 years) and injury time (> 30 days) had a tendency to in uence the re-expansion (p = 0.091 and p = 0.057). Multivariate analysis revealed that atrophy was the only factor that affected the re-expansion rate (p = 0.015). Previous studies have found that cerebral atrophy, long injury time (> 30 days), and old age are high-risk factors for CSDH recurrence, and these factors are related to the cerebral re-expansion rate 19,20,25 . This means that the cerebral re-expansion rate is a hub in the collection of the above risk factors.
The pathological characteristics of CSDH are closely related to the characteristics of the CT imaging 23 .
The density of the images on the CT scan is closely related to hematoma recurrence 18 . The homogeneous type includes three subtypes (hypodense, isodense, and hyperdense). The separated type is de ned as a higher density component under a lower density component, and there is a clear boundary between them. If two components are mixed without a boundary, it is called the gradation type. The laminar type is de ned as a hematoma that presents with a dense layer running along the inner membrane. The trabecular type is de ned as a hematoma with a low iso-density component and a highdensity septum that separates the inner and outer membranes. In the pathophysiology of CSDH, the hypodense and gradation subtypes are considered to have a moderate tendency to re-bleed, and the trabecular type is considered to be the regression stage of these lesions 23 . The isodense, hyperdense, laminar, and separated types have a high-risk of recurrence 28 . Conversely, the hypodense, gradation, and trabecular types have a low risk of recurrence.
The data of 242 patients were used to establish a model scoring system to assess the recurrence of CSDH following which another group (119 patients, June 2015 to July 2016) was used to verify the grading system. The factors predicting postoperative recurrence were screened out, and those that met the criteria were incorporated into the multiple regression analysis model. It was concluded that the cerebral re-expansion rate and preoperative CT imaging classi cation were important independent predictors of RrR. According to the ROC curve analysis, the critical threshold of the postoperative cerebral re-expansion rate (cut-off point, 40%) was determined. According to the intensity and regression coe cients associated with RrR to assign scores to establish a grading system. The prediction performance of the new model was compared with those of the previously published models in the validation group, and the new model had a better evaluation effect (AUC = 0.856). The main parameters of the model were easy to collect clinically and could quickly screen the RrR high-risk patients, thereby providing a reference for guiding the treatment.
Past studies have established models to evaluate high-risk patients for RrR. They have good clinical application values but are associated with some shortcomings. The Alberta grading system only includes preoperative clinical and imaging parameters without the postoperative factors and cannot fully re ect the perioperative imaging changes 8 . In the Oslo grading system, the preoperative hematoma volume cutoff point is 130 mL, and the postoperative residual cavity volume cut-off points are 80 and 120 mL 29 . in the Xining grading system, the thresholds for the volume of the hematoma before the operation and the volume of the postoperative residual cavity are 121 and 72 mL, respectively 33 . However, it is not suitable for different races to use the same xed volume parameter threshold due to limitations in clinical application. The Xining grading system adopts the Nomogram Model, which is a relatively innovative method, but the outcome of predicting the recurrence is binary. The proportion of postoperative gas accumulation in the Wuhu grading system is an important factor 27 . However, with improvements in surgical skills, the amount of postoperative gas produced is reduced and does not affect the patient's prognosis 7 . On the contrary, the cerebral re-expansion rate can more accurately re ect the changes in the patient's perioperative imaging.
Comparisons of the relationship between bilateral CSDH and cerebral re-expansion rate indicated that the postoperative re-expansion ability of bilateral CSDH was weaker (p = 0.028; Supplemental Table 3). This result was consistent with those reported in previous studies 15 . Consequently, the unilateral grading system cannot be applied to patients with bilateral CSDH. We need develop a model to predict the recurrence of bilateral CSDH 32 .
The parameters of postoperative day 1 were mostly used in some studies 29,32 . In the current study, the imaging parameters of postoperative day 1 were compared with those of days 7-9. The 7-9th day parameters demonstrated better predictive abilities of the recurrence of CSDH (Supplemental Table 1 and Fig. 1). The re-expansion is greatest during the rst week after surgery and slows down considerably after that 9 .
Cerebral re-expansion is very important to reduce recurrence. The current methods used to increase the reexpansion rate of the brain tissue include the following: intraoperative aspiration of pneumocephalus via a subdural drain following evacuation 4 , neuroendoscopic removal of the residual septa and trabecula structures to promote brain expansion 14 , postoperatively performed supervised Valsalva maneuver (SVM) 31 , administration of at least 2000 mL per 3 days 10 , and early mobilization 16 . These methods reduce potential subdural space and promote cerebral expansion, thereby decreasing RrR.
One of the limitations of this study is that it is a single-center study. Hence, further veri cations using multicenter studies are warranted.

Conclusions
In this study, cerebral re-expansion was found to play an important role in the RrR of CSDH. On the basis of the preoperative CT imaging classi cation and cerebral re-expansion rate, a grading model system for assessing recurrence was established. This model was validated and compared with previously used models, and it was found to be the most effective. These ndings suggest that the postoperative recurrence of CSDH can be reduced by increasing the cerebral re-expansion rate.

Declarations
Funding Study Funded by National Natural Science Foundation of China (No. 81201980, No. 81572476).

Con icts of interest/Competing interests The authors declare no competing interests.
Availability of data and material The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. Authors' contributions Li Bie contributed to the study conception and design. Also, all authors were involved in material preparation and data collection. Data analysis was performed by Shuai Han. The rst draft of the manuscript was written by Li Bie, and all authors commented on previous versions of the manuscript. All authors read and approved the nal manuscript.    *p < 0.05 Table 5.