SIH-EBP score for prediction of efficacy of epidural blood patching in patients with spontaneous intracranial hypotension

Abstract

Background:

Epidural blood patching (EBP) is the primary treatment for spontaneous intracranial hypotension (SIH), although multiple attempts may sometimes be necessary. The SIH-EBP score, with a cutoff of ≥3, predicts the response to the first EBP. However, its generalizability requires further confirmation. This study aims to validate the clinical utility of the SIH-EBP score and determine the optimal cutoff for predicting the response to the first EBP in an independent cohort of patients with SIH.

Methods:

This retrospective study included patients with SIH who received at least one EBP at a tertiary medical center. Clinical data were extracted from electronic medical records, and brain and spinal magnetic resonance images were reviewed.

Results:

Ninety-six patients (58 female [F]/38 male [M], mean age: 42.67 ± 10.16 years) were screened, with 49 (32 F/17 M, mean age: 41.20 ± 9.13 years) analyzed, including 30 (22 F/8 M, mean age: 41.10 ± 10.14 years) (61.2%) responders. There was a positive correlation between SIH-EBP scores and responder rates (p = 0.001). A cutoff score of ≥3 was associated with a higher response rate than a score of <3 (80.0% vs 41.7%, p = 0.006) (sensitivity = 73.7%, specificity = 66.7%, accuracy = 69.4%). The optimal cutoff in this cohort was ≥2 (Area under curve (AUC) = 0.77, p < 0.001) (sensitivity = 52.6%, specificity = 90.0%, accuracy = 75.5%).

Conclusion:

In this cohort, the SIH-EBP score correlated with response rates to the first EBP. Although a score of ≥3 remains a valid predictor of treatment response, a cutoff of ≥2 proved to be more accurate and specific. However, its practical use is limited by a sensitivity of 52.6%. Further studies are needed to verify its role in other populations.

1. INTRODUCTION

Spontaneous intracranial hypotension (SIH) is caused by spinal cerebrospinal fluid (CSF) leaks without preceding trauma or medical procedures that could compromise the integrity of the spinal dural sac.¹ The incidence of SIH is five cases per 100 000 individuals annually,² and it predominantly affects middle-aged women, with a female-to-male ratio of 2:1.³ Leakage of CSF from the spinal dural sac leads to CSF hypovolemia and frequently results in descent of the brain. Clinically, this downward movement of the brain causes traction on pain-sensitive structures, resulting in characteristic orthostatic headaches.⁴ In addition, downward displacement of brain tissue can tear bridging veins in the cerebral convexity, potentially leading to subdural hematoma and brain herniation.⁵ A decrease in intracranial CSF volume may also cause compensatory dural sinus engorgement and reduced cerebral venous flow, possibly followed by cerebral venous thrombosis.⁶ Prompt diagnosis and treatment are crucial to accelerate recovery and mitigate the risk of these serious complications.

Early epidural blood patching (EBP) is currently the treatment of choice for SIH, especially when conservative treatment fails, and the average response rate to a single EBP of approximately 64% according to a recent meta-analysis.² Therefore, repeat or multiple EBPs may be necessary. Clinical predictors of response to EBPs include female sex, older age, and the volume of blood used in the patching.^7,8 In addition, several radiologic factors have been associated with treatment outcomes, such as the presence of brain sagging, the angles between the vein of Galen and straight sinus, and between the midbrain and pons on brain magnetic resonance imaging (MRI), as well as the length of anterior epidural collections and the number of CSF leakages on spinal MRI.^7,9–12 However, the predictive value of these indicators varies across studies, and only a few have been consistently validated.

Scoring systems are widely used for diagnosis, risk stratification, and prognostication in various neurological disorders.^13–15 For example, the Bern score, based on brain MRI findings, was developed to predict the presence of spinal CSF leaks, primarily serving as a diagnostic tool.¹⁶ More recently, the SIH-EBP score, a grading system consisting of two clinical and two radiologic variables—sex, age, midbrain-pons angle, and anterior epidural CSF collections—was proposed to predict the treatment response to the first EBPs. This score ranges from 0 to 5.¹² Patients with an SIH-EBP score of ≥3 were found to have a 10-fold increase in the odds of having a treatment response to the first EBP, compared to those with a score of <3.¹² However, the generalizability of the SIH-EBP score has not yet been confirmed in other patient populations, which may limit its widespread use.

The aim of this study was to evaluate the clinical utility of the SIH-EBP score in patients with SIH and to establish an optimal cutoff score for predicting the treatment response after the first EBP in an independent cohort.

2. METHODS

2.1. Patients and clinical data

This retrospective study examined consecutive patients diagnosed with SIH who received at least one EBP at Taichung Veterans General Hospital, a tertiary medical center in central Taiwan, from January 2015 to January 2023. Clinical data were extracted from electronic medical records, and brain and spinal MRIs were reviewed. The analysis included demographics such as age, sex, headache characteristics, and relevant clinical manifestations (eg, nausea, vomiting, neck stiffness, tinnitus, or photophobia), along with EBP volumes, time from symptom onset to diagnosis, time from symptom onset to EBP, and treatment responses. Patients were included if they (1) visited the neurology outpatient service or the emergency department of Taichung Veterans General Hospital with symptoms and signs suggestive of SIH, (2) were hospitalized for a diagnostic work-up for SIH, and (3) received a diagnosis of SIH based on the criteria of the Third Edition of the International Classification of Headache Disorders (ICHD-3).¹ Exclusion criteria were (1) spontaneous recovery or recovery after conservative treatment, (2) recovery after treatment at another hospital, or (3) unavailability of medical records or MRIs.

The treatment protocol for patients with SIH at our hospital included intensive hydration for the first 3 days after diagnosis. If there was no satisfactory improvement in clinical symptoms, targeted EBPs were arranged. Patients who fully recovered after hydration were excluded from the analysis. This study was approved by the institutional review board of Taichung Veterans General Hospital (protocol number CE22242B), and the requirement for obtaining informed consent was waived due to the retrospective nature of the study.

2.2. Neuroimaging

All patients underwent conventional brain and spine MRI and whole-spine magnetic resonance myelography (MRM) for diagnostic purposes upon consultation. Key MRI findings before treatment, including the midbrain-pons angle and anterior epidural fluid collections, were recorded. Post-EBP MRIs, which included noncontrast brain MRI and whole-spine MRM, were routinely performed to assess the treatment effects of EBPs. A 1.5 T MRI scanner (MAGNETOM Aera; Siemens Healthcare, Erlangen, Germany) was used for imaging. The conventional brain MRI protocol included axial spin echo T1-weighted images (T1WI) with a repetition time (TR) of 500 milliseconds and an echo time (TE) of 10 milliseconds; axial fast spin echo T2-weighted images (T2WI) with TR 3200 milliseconds and TE 115 milliseconds; and gadolinium (Gd)-enhanced spin echo T1WI in the axial, sagittal, and coronal planes. For the whole spine, sagittal T1WI, T2WI, STIR, and whole-spine MRM using three-dimensional (3D) sampling perfection with optimized contrast using different flip-angle evolution (3D-SPACE) sequences were acquired with the following parameters: TR = 3000 milliseconds, TE = 560 milliseconds, isotropic voxel size = 0.9 mm³, matrix size = 320 × 320 pixels, and field of view (FOV) = 200 mm. Techniques such as fat suppression and generalized autocalibrating partially parallel acquisition (GRAPPA) imaging reconstruction with an acceleration factor of two were utilized. Volumetric images were obtained in the coronal plane, covering the cervical-to-thoracic and thoracic-to-lumbar regions of the spine.

2.3. SIH-EBP scores

The SIH-EBP scores comprise two clinical variables (sex and age) and two radiological variables (midbrain-pons angle and anterior epidural CSF collections).¹² Sex, age, and midbrain-pons angle were rated on a scale from 0 to 1, while anterior epidural CSF collections were rated from 0 to 2. The total of these ratings constitutes the SIH-EBP score. Patients with scores of ≥3 were shown to respond well to the first EBP. The midbrain-pons angle was measured on the midline sagittal T1WI of the brain. The length of the anterior epidural CSF collection, measured in vertebral segments, was evaluated using both sagittal T2WI spinal MRI and MRM. All imaging findings were interpreted by an experienced neuroradiologist (H.-C.C.) and a neurologist (T.-W.H.). In cases of inconsistencies, a consensus was reached through discussion between the two specialists.

2.4. Targeted EBP injection

For all patients, a semi-targeted EBP injection was performed in one or two vertebral segments below the level where the most prominent abnormal CSF signals in neural sleeves were identified on MRM images. The procedure was conducted by an experienced anesthesiologist using a 20-gauge epidural Tuohy needle and a midline approach, with the patient in a lateral recumbent position. During the injection, the loss of resistance technique was used to locate the epidural space in the lumbar spine, and the hanging-drop technique was used for the cervical or thoracic spine.¹⁷ Autologous blood injection (ABI) was carried out until the patient reported back pain, headache, or any other discomfort. The volume of blood injected was documented. After the injection, patients were advised to remain supine for at least 2 hours.

2.5. Treatment response assessment

In addition to clinical observation, post-EBP spinal MRIs and MRMs were utilized to assess treatment responses. After the EBP procedure, patients’ ability to perform basic movements such as sitting, standing, and walking without significant discomfort was evaluated. Responders were defined as patients who could maintain an upright posture for 1 to 3 hours without experiencing severe headaches requiring bed rest or those who showed significant clinical improvement and disappearance of spinal CSF leakage on post-EBP spinal MRI or MRM. Nonresponders were those whose symptoms persisted, necessitating further EBPs.

2.6. Statistical analyses

The data analysis was conducted using SPSS software (v22; SPSS Inc, Chicago, IL). The normality of the distribution of continuous variables was assessed using the Kolmogorov-Smirnov test. For continuous variables, the nonparametric Mann-Whitney test was used, as most data were not normally distributed. Comparisons of nominal variables between responders and nonresponders were carried out using the Fisher exact test and the Chi-square test. In addition, the receiver operating characteristic (ROC) curve, along with Youden J statistic¹⁸ was used to determine the optimal cutoff value of the SIH-EBP scores for predicting treatment response in our patients. The statistical significance level was set at p < 0.05.

3. RESULTS

3.1. Patient characteristics

During the study period, a total of 96 (58 female [F]/38 male [M], mean age: 42.7 ± 10.2 years) patients with SIH were screened. Forty-seven patients were excluded due to recovery after conservative treatment (n = 43), recovery after treatment at another hospital (n = 2), or unavailability of medical records (n = 1) or MRIs (n = 1) (Fig. 1). Consequently, 49 (32 F/17 M, mean age: 41.2 ± 9.1 years) patients with SIH who received at least one targeted EBP in our hospital were included in the analysis. Of these, 30 (22 F/8 M, mean age 41.1 ± 10.1 years) (61.2%) were responders. Compared with nonresponders, responders had shorter anterior epidural collections (12.6 ± 4.3 vs 16.2 ± 3.9 segments, p = 0.004) and were less likely to experience photophobia (0% vs 15.8%, p = 0.025). However, the demographics and other clinical and radiologic features at baseline were similar between the two groups (Table 1).

Fig. 1.

Flow diagram of patient recruitment. EBP = epidural blood patch; MRI = magnetic resonance imaging; SIH = spontaneous intracranial hypotension.

Table 1.

Clinical and radiological features of the responders and nonresponders

Characteristic	Responders	Nonresponders	p
	(N = 30)	(N = 19)
	N (%)	N (%)
Age (mean ± SD), y	42.1 ± 10.1	39.8 ± 7.3	0.394
Women (%)	22 (73.3%)	10 (52.6%)	0.138
Midbrain-pons angle (mean ± SD), °	41.8 ± 10.4	40.9 ± 9.9	0.758
Length of anterior epidural collections (mean ± SD)	12.6 ± 4.3	16.2 ± 3.9	0.004
Anterior epidural CSF collection			0.021
No collection	0	0
1-7 Segments	4	0
8-18 Segments	26	13
≥19 Segments	0	6
Clinical feature
Nausea	18 (60%)	10 (52.6%)	0.612
Vomiting	14 (46.7%)	8 (42.1%)	0.754
Neck stiffness	13 (43.3%)	6 (31.6%)	0.411
Tinnitus	11 (36.7%)	7 (36.8%)	0.990
Hypacusia	2 (6.7%)	1 (5.3%)	0.304
Photophobia	0	3 (15.8%)	0.025
EBP volume (mean ± SD), mL	19.6 ± 7.1	18.3 ± 8.4	0.550
Time from onset to diagnosis (mean ± SD), d	13.5 ± 13.8	9.8 ± 8.6	0.308
Time from onset to EBP (mean ± SD), d	22.4 ± 14.6	16.5 ± 8.5	0.117
SIH-EBP score	2.6 ± 0.9	1.8 ± 0.8	0.006

CSF = cerebrospinal fluid; EBP = epidural blood patch; SIH = spontaneous intracranial hypotension.

3.2. SIH-EBP scores and treatment response

In total, 13 (26.5%) patients had an SIH-EBP score of 1, whereas 11 (22.4%), 21 (42.9%), and 4 (8.2%) had scores of 2, 3, and 4, respectively. Responder rates were positively correlated with SIH-EBP scores (p = 0.001, Cochran-Armitage test for trend) (Fig. 2A). Responders had higher SIH-EBP scores compared with nonresponders (1.8 ± 0.8 vs 2.6 ± 0.9, p = 0.006) (Fig. 2B). Patients were subsequently dichotomized based on the previously reported cutoff score of 3. Those with an SIH-EBP score of ≥3 were more likely to respond to the first EBP compared with those who scored <3 (80.0% vs 41.7%, p = 0.008) (Fig. 2C). The overall accuracy of predicting EBP response using a cutoff value of ≥3 was 69.4%, with sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of 73.7%, 66.7%, 58.3%, and 80.0%, respectively. The Youden index was 0.404.

Fig. 2.

Relationship between responder rates and SIH-EBP scores. A and B, Responder rates in relation to SIH-EBP scores. C, Responder rates between SIH-EBP scores with a cutoff value of 3. EBP = epidural blood patch; SIH = spontaneous intracranial hypotension.

3.3. Derivation of cutoff score in the current cohort

Based on the ROC curve analysis (Fig. 3A), the cutoff value of the SIH-EBP score for predicting treatment response was determined to be 2 in the current cohort (Youden index = 0.426) (area under the ROC curve = 0.77, p < 0.001). The responder rates were higher in patients who scored ≥2 compared to those who scored <2 (75.0% vs 23.1%, p = 0.003) (Fig. 3B). Using a cutoff value of ≥2, the accuracy in predicting a positive response to the first EBPs was 75.5%, with sensitivity, specificity, PPV, and NPV of 52.6%, 90.0%, 76.9%, and 75.0%, respectively.

Fig. 3.

Cutoff of SIH-EBP score in the current cohort. A, ROC curve of current cohort. B, Responder rates between SIH-EBP scores with a cutoff value of 2. EBP = epidural blood patch; ROC = receiver operating characteristic; SIH = spontaneous intracranial hypotension.

4. DISCUSSION

The present study validated the role of the SIH-EBP score in predicting treatment outcomes in an independent cohort of patients with SIH. Patients with higher scores were more likely to experience better outcomes. An SIH-EBP score of ≥3 continued to be a reliable predictor for a treatment response to the first EBPs in this cohort. Although a cutoff score of ≥2 was associated with higher overall accuracy, specificity, and PPV in the current cohort, its clinical utility is limited by a sensitivity of 52.6%. Incorporating the SIH-EBP score in the treatment of SIH could assist in patient stratification. Early identification of patients at risk of treatment failure could prompt treating physicians to consider more aggressive approaches, such as repeat or multiple EBPs, potentially speeding recovery, and reducing the risk of complications.

One of the most important strengths of this study is the verification of the reproducibility of a scoring system in an independent cohort. The clinical utility of the SIH-EBP score was examined across different hospital settings and MRI platforms, demonstrating its validity and generalizability. Despite the retrospective nature of this study, the data were of relatively high quality. All patients were hospitalized for diagnostic work-up and management, and treatment outcomes were evaluated in a standardized manner. Furthermore, brain and spinal MRIs were interpreted by a neurologist and an experienced neuroradiologist, both of whom have relevant publications in this field.^11,19

In the present cohort, the only radiologic variable associated with treatment response was the length of anterior epidural collections. This finding aligns with previous reports.^7,20 Patients with longer anterior epidural collections may experience more pronounced shrinkage of the dural sac, indicating greater disease severity,²¹ which could make the patients less responsive to treatment with EBPs.⁷ Although the differences in age and sex distributions between groups were not significant, there was a trend toward older age and a higher prevalence of females among responders, consistent with findings from other studies.^7,12,22 The current study may not have been adequately powered to demonstrate statistical significance. However, these studies involved only Asian patients, and further research is needed to confirm these findings in other ethnic groups.

There are some discrepancies between the findings of the present study and those in prior studies, particularly regarding the role of the midbrain-pons angle and the cutoff SIH-EBP score in predicting treatment response. Although previous research indicated that patients with smaller midbrain-pons angles were less likely to respond to EBPs,^7,12 our study did not show a significant difference between responders and nonresponders, aligning with a report from Korea.¹⁰ This discrepancy might have influenced the differing cutoff scores identified. The original report that proposed the scoring system determined an optimal cutoff score of 3,¹² while it was 2 in our cohort. This incongruence could be attributed to variations in the study population and the definitions of “targeted” EBPs (1-2 segments below the level of the most prominent CSF leaks vs directly at the levels of identified CSF leaks) and treatment response criteria (combination of clinical and radiological findings vs purely clinical outcomes). Technical aspects could also play a significant role. Although the angles were supposed to be measured on midline sagittal images, sometimes measurements had to be taken on an image slice near the midline rather than exactly at the midline, which could have introduced bias in rating this variable. In our cohort, while a cutoff score of ≥2 was associated with slightly higher accuracy, PPV, and Youden index, a sensitivity of 52.6% might limit its clinical utility. Identifying potential responders is crucial, as such an approach could help avoid unnecessary procedures. In practice, a score of ≥3 may prove to be a more effective cutoff in predicting treatment response.

There are some limitations to this study. First, as a retrospective study, only a limited number of patients were included, and some essential information could be incomplete or unavailable. However, the primary objective was to validate the findings of prior reports, not to explore unknown variables, and only patients with reliable medical records or MRIs were included in the analysis. In fact, only two patients were excluded due to the unavailability of medical records or MRIs. Second, the cohort was recruited from a tertiary medical center, and the analysis included only patients who received at least one EBP, excluding those who recovered after conservative treatment. It is possible that the clinical characteristics could differ among patients managed in different settings and between those treated with and without EBPs. Selection bias remains a concern, and the generalizability of the findings could be an issue. However, SIH is a rare neurological disorder that can lead to potentially life-threatening complications, and many patients would eventually be referred to medical centers. Besides, the primary objective was to verify the clinical utility of the SIH-EBP score in patients with a potential need for EBPs. Third, technical difficulties in measuring the midbrain-pons angle could arise; the midline sagittal slice is not always available, and the borders of the midbrain and pons on sagittal images are often curved, which could potentially limit the implementation of this parameter in clinical practice. Finally, the treatment protocols and definitions of treatment response differed from the study that proposed the scoring system, which could have contributed to the disparity in the cutoff scores identified. However, the findings were generally consistent with the original report, suggesting that the scoring system remains a valuable tool in an independent clinical sample.

In conclusion, in this independent cohort of patients with SIH, the SIH-EBP score was positively correlated with the response rate to the first EBPs, and patients with higher scores experienced better outcomes. A cutoff of ≥3 for the SIH-EBP score remained valid in predicting treatment response. Incorporating the SIH-EBP score into the clinical management of SIH could aid in risk stratification. However, given the limitations of this study, additional research, preferably multicenter prospective trials, is necessary to further confirm the external validity of the SIH-EBP scores across different ethnic or racial groups and clinical settings. Moreover, exploring whether advanced imaging techniques, such as phase-contrast MRI, or automatic tools like 3D reconstruction or Artificial Intelligence-assisted measurements, could enhance the accuracy of outcome prediction after EBPs is warranted.