Long-Term Epidural Patching Outcomes and Predictors of Benefit in Patients With Suspected CSF Leak Nonconforming to ICHD-3 Criteria

Abstract

Background and Objectives

Spinal CSF leaks lead to spontaneous intracranial hypotension (SIH). While International Classification of Headache Disorders, Third Edition (ICHD-3) criteria necessitate imaging confirmation or low opening pressure (OP) for SIH diagnosis, their sensitivity may be limited. We offered epidural blood patches (EBPs) to patients with symptoms suggestive of SIH, with and without a documented low OP or confirmed leak on imaging. This study evaluates the efficacy of this strategy.

Methods

We conducted a prospective cohort study with a nested case-control design including all patients who presented to a tertiary headache clinic with clinical symptoms of SIH who completed study measures both before and after receiving an EBP between August 2016 and November 2018.

Results

The mean duration of symptoms was 8.7 ± 8.1 years. Of 85 patients assessed, 69 did not meet ICHD-3 criteria for SIH. At an average of 521 days after the initial EBP, this ICHD-3–negative subgroup experienced significant improvements in Patient-Reported Outcomes Measurement Information System (PROMIS) Global Physical Health score of +3.3 (95% CI 1.5–5.1), PROMIS Global Mental Health score of +1.8 (95% CI 0.0–3.5), Headache Impact Test (HIT)-6 head pain score of −3.8 (95% CI −5.7 to −1.8), Neck Disability Index of −4.8 (95% CI −9.0 to −0.6) and PROMIS Fatigue of −2.3 (95% CI −4.1 to −0.6). Fifty-four percent of ICHD-3–negative patients achieved clinically meaningful improvements in PROMIS Global Physical Health and 45% in HIT-6 scores. Pain relief following lying flat prior to treatment was strongly associated with sustained clinically meaningful improvement in global physical health at an average of 521 days (odds ratio 1.39, 95% CI 1.1–1.79; p < 0.003). ICHD-3–positive patients showed high rates of response and previously unreported, treatable levels of fatigue and cognitive deficits.

Discussion

Patients who did not conform to the ICHD-3 criteria for SIH showed moderate rates of sustained, clinically meaningful improvements in global physical health, global mental health, neck pain, fatigue, and head pain after EBP therapy. Pre-treatment improvement in head pain when flat was associated with later, sustained improvement after EBP therapy among patients who did not meet the ICHD-3 criteria.

Classification of Evidence

This study provides Class IV evidence that epidural blood patch is an effective treatment of suspected CSF leak not conforming to ICHD-3 criteria for SIH.

Introduction

Spinal CSF leaks result in spontaneous intracranial hypotension (SIH), a syndrome that, in its most recognizable form, comprises orthostatic headaches, neck pain, nausea, and vestibulocochlear symptoms.¹ The International Classification of Headache Disorders, 3rd Edition (ICHD-3) diagnostic criteria for SIH require either of the following to satisfy the diagnosis: an opening pressure less than 60 mm CSF or evidence of a CSF leakage on imaging.² However, converging evidence now suggests that these diagnostic criteria may be inadequately sensitive to identify many patients suffering from SIH, particularly in the chronic phase. With increasing duration of a spinal CSF leak, opening pressure is less likely to be low and brain MRI is less likely to demonstrate obvious features of SIH.^3-10 We now know that opening pressure does not distinguish between those with and without imaging-proven CSF leaks, with studies suggesting that between 61% and 94% of patients with imaging-proven CSF leaks have normal opening pressures.^6,8,11,12 In addition, neuroimaging may also be inadequately sensitive to exclude SIH in the clinical setting. A recent meta-analysis concluded that 1 in 5 patients with SIH had normal brain MRI, whereas previous reports have suggested 1 in 4.^13,14 Spinal imaging has similar sensitivity limitations: CSF-venous fistulas (CVFs), in which CSF is lost directly into paraspinal veins, escaped detection by the medical field until 2014 because they cannot be seen with spinal MRI or conventional CT myelography. While initially thought to be rare, CVFs may account for the underlying pathology in up to half of patients with SIH.^15-20

In a tacit recognition that the proposed ICHD-3 criteria might be specific but not sensitive, the ICHD-3 recommends “in patients with typical orthostatic headache and no apparent cause, and after exclusion of postural orthostatic tachycardia syndrome, it is reasonable in clinical practice to provide autologous lumbar epidural blood patch (EBP).”² However, little data have been available to evaluate the efficacy of this approach.

In 2016, our institution began offering EBPs to patients with orthostatic headaches and other elements of a compelling history suspicious for SIH whether or not they satisfied the ICHD-3 criteria. We prospectively measured validated metrics of both CSF leak and global health before and after EBP as part of a quality assurance program. We hypothesized that a strategy of patching patients with symptoms of CSF leak, regardless of baseline neuroimaging findings or opening pressures, would be associated with clinically meaningful improvement in validated metrics of quality of life and SIH symptoms.

Methods

Study Design

This was a prospective cohort study with a nested case-control approach aimed at identifying predictors of clinically meaningful improvement in global physical health after epidural patching. Patients who demonstrated improvement in the Patient-Reported Outcomes Measurement Information System (PROMIS) Global Physical Health measures exceeding a predefined published threshold for clinically meaningful change were considered “cases” and were assessed compared with “controls” who did not demonstrate such improvement. The Stanford University Institutional Review Board approved this study, and because of data deidentification, written consent was waived.

Study Cohort

The cohort comprised 85 consecutive patients who underwent epidural patches at our institution for suspected CSF leak from August 2016 to November 2018 and completed study measures both before and after receiving an epidural patch (Figure 1). Patches were used as part of routine care. Patients were offered or declined patches based on holistic assessment of a detailed history, review of current symptoms, all available imaging, laboratory test results, and/or opening pressure. In general, if a patient suffered from an orthostatic headache and vestibulocochlear symptoms without a better alternative diagnosis to account for these symptoms despite a negative workup for CSF leak, they were offered a patch. Patients with mild symptoms, with significant responses to pharmacologic headache management, and with intermittent rather than daily symptoms were generally excluded. Imaging was specifically not used to exclude patients from an EBP in the presence of these clinical symptoms. The clinical features of our cohort are provided in Table 1. Follow-up data were collected through November 2022. Patients completed validated measures of CSF leak–related symptoms and health-related quality of life. These were augmented by specific questions regarding the orthostatic nature of their symptoms and demographic information. Measures were administered through a Health Insurance Portability and Accountability Act–compliant online platform (Qualtrics, Provo, UT).

Figure 1. Flowchart Illustrating Patients Selected for Analysis.

Table 1.

Baseline Characteristics of Study Population

Characteristic	Total patients with suspected CSF leak (N = 85), n (%)	Met ICHD-3 criteria (n = 16), n (%)	Not meeting ICHD-3 criteria (n = 69), n (%)	p Value
ICHD-3 criteria met		16	69
Age, y	42.5 ± 13.7	41.8 ± 14.2	42.6 ± 13.7	0.84
Sex
Male	19 (22.4)	4 (25.0)	15 (21.7)	1.00
Female	66 (77.7)	12 (75.0)	54 (78.2)	1.00
Symptom duration, y	8.7 ± 8.1	8.4 ± 7.7	8.7 ± 8.3	0.91
CSF leak symptom
Head pain or pressure	84 (98.8)	16 (100)	68 (98.6)	1.00
Neck pain or stiffness	70 (82.3)	10 (62.5)	60 (90.0)	0.05
Tinnitus	54 (63.5)	12 (75.0)	42 (60.9)	0.44
Cognitive impairment	77 (90.5)	13 (81.3)	64 (92.8)	0.34
Fatigue	81 (95.2)	15 (93.8)	66 (95.7)	1.00
Orthostatic features
Orthostatic head pain	71 (83.5)	14 (87.5)	57 (82.6)	0.92
Time upright before head pain starts (min)		120 ± 225	63 ± 84	0.39
Quartile 1	0–5.0	0–15.0	0–5.0
Quartile 2	5.0–30.0	15.0–30.0	5.0–30.0
Quartile 3	30.0–108.8	30.0–121.0	30.0–90.4
Quartile 4	108.8–841.0	121.0–841.0	90.4–366.2
Head pain severity (0–10)
After 1 h upright	5.3 ± 2.4	4.9 ± 3.0	5.4 ± 2.2	0.52
After 1 h flat	2.7 ± 2.5	2.2 ± 2.8	2.8 ± 2.4	0.41
After night recumbent	2.2 ± 2.5	2.2 ± 2.2	2.3 ± 2.6	0.92
Migrainous symptom
Photophobia (sensitivity to light)	62 (72.9)	7 (43.8)	55 (79.7)	0.009
Phonophobia (sensitivity to sound)	62 (72.9)	10 (62.5)	52 (75.4)	0.46
Nausea or vomiting	56 (65.9)	7 (43.8)	49 (71.0)	0.08
At least 1 migrainous feature	76 (89.4)	12 (75.0)	64 (92.8)	0.10
Symptom measures
PROMIS Global Physical Health	32.7 ± 6.8 (85)	35.4 ± 6.9	32.0 ± 6.6	0.09
PROMIS Global Mental Health	36.8 ± 8.4 (85)	39.7 ± 8.9	36.2 ± 8.2	0.17
Headache Impact Test-6	66.9 ± 6.8 (83)	63.7 ± 10.0	67.6 ± 5.6	0.15
Neck Disability Index	45.1 ± 20.6 (85)	38.1 ± 21.1	46.8 ± 20.3	0.15
Dizziness Handicap Inventory	42.5 ± 27.2 (85)	38.6 ± 30.2	43.4 ± 26.6	0.57
PROMIS Applied Cognition	39.6 ± 7.6 (83)	37.8 ± 8.7	40.0 ± 7.4	0.38
PROMIS Fatigue	66.8 ± 7.5 (85)	66.8 ± 7.4	66.8 ± 7.6	0.95
Rhodes Index of Nausea, Vomiting, and Retching (nausea and vomiting)	1.1 ± 1.1 (85)	0.7 ± 0.8	1.1 ± 1.1	0.25
Tinnitus Handicap Inventory	19.7 ± 22.1 (85)	27.6 ± 23.5	17.8 ± 21.5	0.12

Abbreviations: ICHD-3 = International Classification of Headache Disorders, 3rd edition; PROMIS = Patient-Reported Outcomes Measurement Information System.

Data Time Points

Metrics were collected before each patch and again after being seen at follow-up visits. Owing to the varied number of patches received by patients, data were analyzed at 3 primary points: (1) before the first patch, (2) after the first patch (“early results”), and (3) after the final patch (“durable results”).

Patient and Clinical Characteristics

Where suitable, PROMIS metrics were used. PROMIS, an NIH-developed patient-reported outcome system, produces T-scores standardized to a US mean of 50, with a 10-point change corresponding to SD change in the US population.²¹

Measures

PROMIS Global Health SF v1.1

This is a 10-item quality of life instrument producing physical and mental health summary scores. Higher scores indicate better health. It is reliable across various neurologic syndromes.^22-24

PROMIS Fatigue 8a

This is an 8-item assessment of fatigue over the past week. The T score is centered at 50 for the US mean, with a 10-point change representing 1 SD. Scoring is adjusted for 1 omitted item.^25-27

PROMIS Short Form v1.0—Applied Cognition—Abilities 8a

This 8-item tool measures cognitive self-assessments over the past week. Lower scores indicate greater impairment and have been linked with disability-induced unemployment in patients with multiple sclerosis.^28,29

Headache Impact Test

This is a 6-item patient-reported tool quantifying the adverse headache impact. Higher scores indicate greater impact, categorized into 4 severity levels.^30,31

Neck Disability Index

This is a 10-item survey for neck-related disability. Scores range from 0 to 50, with higher scores indicating increased disability.³²

Dizziness Handicap Inventory

Measures perceived dizziness-related handicap are calculated using a 25-item questionnaire, scored from 0 to 100, with categories for symptom severity.^33,34

Tinnitus Handicap Inventory

This is a 25-item validated tool for tinnitus severity. It is scored from 0 to 100, with classifications based on severity.^35,36

Rhodes Index of Nausea, Vomiting, and Retching

This is an 8-item measure for gastrointestinal symptoms. Higher scores suggest more severe symptoms. Patients without symptoms were given a zero score.³⁷

EBPs were targeted in all cases to overt leaks when possible and to imaging features associated with increased risk of CSF leak when a leak was not visible, including prominent meningeal diverticula, distended nerve roots, and disk osteophytes impinging on the dura^12,38-40 (eFigure 1).

Missing Data

When data were missing, if the scoring protocol for the questionnaire allowed for accurate scoring of the instrument, then the instrument was scored and included. Missing data were not imputed or carried forward from previous measures.

Statistical Analysis

Statistical analyses were conducted using R (version 4.3.0). Patient data were summarized using frequency, percentage, mean, and SD. Continuous variables were compared using t tests or Wilcoxon rank-sum tests while paired data used t tests or Wilcoxon signed-rank tests, all depending on distribution. Shapiro-Wilk test checked score normality. Categorical variables underwent Fisher exact or Pearson χ² tests. Odds ratios (ORs) were calculated for categorical variables as the ratio of the probability of the outcome event occurring in the exposure vs non-exposure group. ORs for continuous variables were calculated using logistic regression to derive the log odds. A significance level of 0.05 was used for all tests.

Imaging Assessment

Brain and full spine MRI was performed using protocols designed to look for signs of intracranial hypotension and spinal epidural fluid. The spine sequences included a precontrast 3D T2 fat-suppressed (IDEAL [Iterative Decomposition of water and fat with Echo Asymmetry and Least-squares estimation]) and sagittal 3D steady-state gradient echo (FIESTA) sequence with multiplanar reformats. Myelograms were performed as conventional CT myelograms with 10 cc of iopamidol (Isovue-M-300). CT myelography included bone and soft-tissue windows of the spine from the mid skull through the sacrum with 0.625- and 2-mm slice thickness. Opening pressure was measured during CT myelography with the patient in the prone position, with previous systematic exploration showing high correlation between opening pressures measured in the right lateral decubitus and prone position.⁴¹ All images were interpreted by a board-certified neuroradiologist.

Data Availability

After publication, data will be accessible for up to 4 years to those researchers seeking collaborative endeavors. Acquisition requires approval from an independent institutional review board review committee.

Results

Baseline Characteristics of Patients

The cohort consisted of 85 patients (mean age 42.5 ± 13.7; 78% female) including 69 patients who did not meet the ICHD-3 criteria for SIH—hereafter referred to as “ICHD-3 criteria–negative”—and 16 patients who did meet the ICHD-3 criteria for SIH—hereafter referred to as “ICHD-3 criteria–positive” (Table 1). Imaging results and opening pressure assessments constituting ICHD-3 criteria are tabulated in eTable 1. Baseline characteristics of the cohort stratified by whether the patients did or did not meet the ICHD-3 criteria for SIH are presented in Table 1. Ninety-nine percent of patients reported head pain, and 84% reported orthostatic headache. Sixty-four percent reported tinnitus, 81% reported abnormal hearing, 91% reported cognitive impairment, and 81% reported fatigue. Eighty-nine percent reported at least 1 ICHD-3 migraine feature. The impact of these symptoms was severe. The mean Global Physical Health T score was 32.7 ± 6.8 (4th percentile for the United States) and Mental Health T score was 36.8 ± 8.4 (10th percentile for the United States). The mean Headache Impact Test (HIT)-6 score was 67 ± 7, indicating severe headache impact.³¹ The average neurocognitive abilities T score was 39.6 ± 7.6 (16th percentile) while the fatigue score was 66.8, denoting fatigue levels higher than 95% of the US populace. Symptoms had persisted for an average of 8.7 years.

Time until head pain worsened upon standing varied widely and is detailed by quartile in Table 1. The earliest quartile onset occurred immediately to 5 minutes after standing while the slowest group experienced onset between 108 and 841 minutes upright. Using a 0–10 visual analog scale, average head pain was 5.3 ± 2.4 after an hour upright, reducing to 2.7 ± 2.5 after an hour lying down and further decreasing to 2.2 ± 2.5 after a night flat.

Baseline Differences in Symptoms and Measures Between Patients Conforming and Not Conforming to ICHD-3 Criteria for SIH

Whether meeting the ICHD-3 criteria or not, patients showed nearly identical symptom presentations. CSF leak symptom frequency and severity did not significantly differ between patients who did not meet the ICHD-3 criteria and patients who did meet the ICHD-3 criteria (Table 1). Parameters such as orthostatic head pain, pain onset timing, and severity both upright and lying down were comparable. Likewise, tinnitus, nausea, cognitive issues, and fatigue were uniformly distributed. The presence of at least 1 migraine-associated symptom (photophobia, phonophobia, nausea, or vomiting) was also reported similarly (75 vs 93; p = 0.10). On all symptom-specific and global health metrics, symptom severities matched between groups and duration of illness was also nearly identical (8.4 vs 8.7 years; p = 0.91). However, minor differences emerged in neck discomfort and photophobia.

Among the 16 patients who met the ICHD-3 criteria for SIH, 14 had an opening pressure measurement, of which only 4 had a pressure <60 mm CSF (sensitivity 28%; 95% CI 0.08–0.58). All 10 of the remainder had a positive imaging study for CSF leak, but a normal opening pressure.

Epidural Patch Characteristics

Patients received a mean of 3.6 ± 2.9 EBPs, at a mean of 2.5 ± 1 spinal levels treated per patch. Average days between patches was 71 ± 80. 8.4% of patches were cervical, 39% were thoracic, 44% were lumbar, and 9% were sacral. Forty-three percent of epidural patches used fibrin sealant, and 57% used autologous blood. No patient received a combination of blood and fibrin simultaneously.

Measures and Outcomes

A mean of 225 days (median 164 days) elapsed between the first epidural patch and the first post-patch completion of measures, which formed the time frame for assessment of early outcomes after patching. The last set of health measures occurred a mean of 521 days (median 332 days) after the first patch and 377 days after the last patch, which formed the time frame for assessment of durable outcomes after patching.

Epidural Patching Outcomes by ICHD-3 Criteria

Patients Who Did Not Meet ICHD3 Criteria: Early Outcomes

At an average of 225 days after the first patch, patients who did not meet ICHD-3 criteria demonstrated a statistically significant improvement in their PROMIS Global Physical Health score +3.2 (95% CI 1.6–4.8; p < 0.001) (Table 2). Similarly, their HIT-6 scores, which measure head pain, significantly decreased (−3.6; 95% CI −5.8 to −1.4; p < 0.001). Fatigue, as measured by PROMIS Fatigue, also showed significant improvement (−2.2; 95% CI −4.5 to −1.0; p < 0.003). Although there was a reduction in neck pain and nausea, these changes were not statistically significant. No health metric deteriorated after patching.

Table 2.

Epidural Patch Outcomes in Patients Not Meeting ICHD-3 Criteria

Survey scores	N	PRE-PATCH Mean ± SD	POST-PATCH Mean ± SD	Difference (95% CI)	p Value
Early outcomes: Change from pre-patch to first post-patch assessment (mean 225 d)
PROMIS physical health	69	32.0 ± 6.6	35.2 ± 6.7	3.2 (1.6 to 4.8)	<0.001
PROMIS mental health	69	36.2 ± 8.2	37.0 ± 9.0	0.9 (−0.8 to 2.5)	0.29
Headache Impact Test-6	67	67.6 ± 5.6	64.3 ± 9.0	−3.6 (−5.8 to −1.4)	0.001
Neck Disability Index	69	46.8 ± 20.3	43.2 ± 20.5	−3.5 (−7.8 to 0.8)	0.11
Dizziness Handicap Inventory	69	43.4 ± 26.6	41.4 ± 25.6	−2.0 (−6.8 to 2.9)	0.42
PROMIS Applied Cognition	68	40.0 ± 7.4	40.0 ± 7.8	0.0 (−1.7 to 1.8)	0.97
PROMIS Fatigue	69	66.8 ± 7.6	64.6 ± 7.2	−2.2 (−4.5 to −1.0)	0.003
Tinnitus Handicap Inventory	69	17.8 ± 21.5	19.9 ± 20.1	2.0 (−4.0 to 8.0)	0.60
Rhodes Index of Nausea, Vomiting, and Retching	69	1.13 ± 1.14	0.96 ± 1.02	−0.2 (−0.5 to 0)	0.06
Durable outcomes: Change from pre-patch to last post-patch assessment (mean 521 d)
PROMIS physical health	69	32.0 ± 6.6	35.4 ± 7.2	3.3 (1.5 to 5.1)	<0.001
PROMIS mental health	69	36.2 ± 8.2	38.0 ± 8.6	1.8 (0 to 3.5)	0.046
Headache Impact Test-6	67	67.6 ± 5.6	64.1 ± 8.23	−3.8 (−5.7 to −1.8)	<0.001
Neck Disability Index	69	46.8 ± 20.3	42.0 ± 19.2	−4.8 (−9.0 to −0.6)	0.03
Dizziness Handicap Inventory	69	43.4 ± 26.6	41.0 ± 26.0	−2.4 (−7.5 to 2.7)	0.35
PROMIS Applied Cognition	68	40.0 ± 7.4	40.2 ± 8.1	0.1 (−1.6 to 1.9)	0.87
PROMIS Fatigue	69	66.8 ± 7.6	64.5 ± 7.2	−2.3 (−4.1 to −0.6)	0.009
Tinnitus Handicap Inventory	69	17.8 ± 21.5	22.2 ± 21.9	4.4 (−2.0 to 12.0)	0.20
Rhodes Index of Nausea, Vomiting, and Retching	69	1.13 ± 1.14	1.02 ± 1.15	−0.1 (−0.5 to 0.1)	0.14

Abbreviations: ICHD-3 = International Classification of Headache Disorders, 3rd edition; PROMIS = Patient-Reported Outcomes Measurement Information System.

Patients Who Did Not Meet ICHD3 Criteria: Durable Outcomes

When assessed an average of 377 days after the last patch (mean 521 days after the first patch), the patients who had not met ICHD-3 criteria because of negative imaging and normal opening pressure showed further progressive improvements. Global physical health improved further (+3.3; 95% CI 1.5–5.1; p < 0.001) (Table 2). HIT-6 head pain scores (−3.8; 95% CI −5.7 to −1.8; p < 0.001) and fatigue (−2.3; 95% CI −4.1 to −0.6; p < 0.009) further improved.

Global mental health and neck pain improvements, which were not statistically significant after the first patch, improved to a statistically significant degree with further patching and more time. Global mental health, as captured by the PROMIS Global Health Mental, improved to 1.8 (95% CI 0–3.5; p < 0.046) while neck pain (Neck Disability Index) decreased to −4.8 (95% CI −9.0 to −0.6; p < 0.03). No health metric deteriorated after patching.

Patients Who Did Not Meet ICHD-3 Criteria: Clinically Meaningful Response Rate

The degree of improvement large enough to be considered clinically meaningful has been previously described for both PROMIS Global Physical Health and HIT-6. For PROMIS Global Physical Health, the minimal clinically meaningful difference has been estimated to be 1.7.⁴² The within-patient minimal meaningful clinical difference for the HIT-6 has been reported to be 2.5.⁴³

Of the 69 patients who did not meet ICHD-3 criteria, 37 (54%) experienced a clinically meaningful improvement in PROMIS Global Physical Health after patching. In addition, 30 of these 69 patients (45%) experienced a clinically meaningful improvement in HIT-6 after patching. Sixty-four percent of patients who did not meet the ICHD-3 criteria had a clinically meaningful improvement in either Global Physical Health or HIT-6 (Table 3).

Table 3.

Clinically Meaningful Improvement After Epidural Patching by ICHD-3 Criteria for Spontaneous Intracranial Hypotension^a

	Total meaningful improvement, N (%)	ICHD-3 criteria negative (N = 69), N (%)	ICHD-3 criteria positive (N = 16), N (%)	Odds ratio (95% CI)	p Valueb
PROMIS Global Physical Health	50 (58.8)	37 (53.6)	13 (81.3)	3.74 (0.98–14.3)	0.08
Headache Impact Test-6 (HIT-6)	41 (48.2)	30 (44.8)	11 (68.8)	2.71 (0.85–8.67)	0.15
PROMIS Global Physical Health or HIT-6	58 (68.2)	44 (63.8)	14 (87.5)	3.98 (0.84–18.9)	0.12

Abbreviations: ICHD-3 = International Classification of Headache Disorders, 3rd edition; PROMIS = Patient-Reported Outcomes Measurement Information System.

^aChange from pretreatment to final assessment completed mean of 521 days after the first patch and 377 days after the last of an average of 3.6 ± 2.9 epidural patches.

^bICHD-3 criteria negative vs criteria positive.

The association of baseline symptoms of CSF leak and patch outcomes are presented in Table 4. Two related specific clinical characteristics of orthostatic headache among patients who did not meet the ICHD-3 criteria measured before patching were associated with the likelihood of a durable clinically meaningful improvement in Global Physical Health at a mean of 377 days after their last patch. These were the degree of relief of head pain that a patient attained after laying flat for 1 hour (OR 1.39; 95% CI 1.10–1.79; p < 0.003) and the degree of relief after laying flat all night (OR 1.27; 95% CI 1.04–1.60; p < 0.02). Each additional 1 point drop in the 0–10 head pain score after laying flat for an hour before epidural patching was associated with a 39% increased odds of having a clinically meaningful response in Global Physical Health 377 days after the last EBP.

Table 4.

Association of Pretreatment Characteristics With Long-Term Clinically Meaningful Improvement After Epidural Patching in ICHD-3–Negative Patients^a

Characteristic	Not improved (N = 32)	Clinically meaningful improvement (N = 37)	Odds ratio	p Value
Symptom measures	Mean ± SD; n (%)	Mean ± SD; n (%)
Baseline PROMIS global health
Physical	33.6 ± 7.4	30.7 ± 5.6	0.93	0.08
Mental	35.2 ± 9.6	37.0 ± 6.9	1.03	0.38
Orthostatic features
Orthostatic head pain	25 (78.13)	32 (86.49)	1.79	0.55
Time upright before head pain starts (min)	76.3 ± 97.6	54.0 ± 72.2	1.00	0.44
Head pain severity (0–10)
After 1 h upright	5.7 ± 2.4	5.1 ± 2.0	0.88	0.29
After 1 h flat	3.7 ± 2.4	2.1 ± 2.2	1.39b	0.003
After night recumbent	3.0 ± 2.9	1.6 ± 2.1	1.27b	0.02
Migrainous symptoms
Photophobia (sensitivity to light)	26 (83.87)	28 (77.78)	0.67	0.75
Phonophobia (sensitivity to sound)	23 (74.19)	28 (77.78)	1.22	0.96
Nausea or vomiting	22 (70.97)	25 (69.44)	0.93	1.00
Symptom duration, y	10.09 ± 9.89	7.49 ± 6.50	1.04	0.35

Abbreviations: ICHD-3 = International Classification of Headache Disorders, 3rd edition; PROMIS = Patient-Reported Outcomes Measurement Information System.

^aOdds of clinically meaningful change in PROMIS Global Health Physical from pretreatment to final assessment (mean of 377 days after the last of an average of 3.6 ± 2.9 epidural patches).

^bOdds ratio of being a long-term clinically meaningful responder per each 1 point reduction in head pain when flat before treatment.

Among patients who did not meet the ICHD-3 criteria in whom head pain was reduced to zero after an hour of lying flat, 12 of 13 (92.3%) had a durable clinically meaningful response in global physical health when assessed an average of 377 days after their last patch (Figure 2). Among patients who reported their head pain was reduced to a 1 or less after being flat an hour, 16 of 21 (76.2%) patients had a durable clinically meaningful response. By contrast, among patients who reported that pain severity remained at least 5 out of 10 after being flat for 1 hour, 4 of 14 (28.6%) had a clinically meaningful improvement at long-term follow-up.

Figure 2. Relationship of Head Pain When Flat and Global Physical Health Outcomes After Epidural Patching in ICHD-3–Negative Patients.

Decreased head pain after lying flat for 1 hour predicted greater likelihood of clinically meaningful improvement in PROMIS Global Physical Health 521 days after initiation of epidural patching (OR 1.39; 95% CI 1.10–1.79; p < 0.003) among patients not conforming to the ICHD-3 criteria for spontaneous intracranial hypotension. The ratio of patients improving a full US population SD in global physical health (10 points on the T-score) compared with those who declined a full SD was 13 to 1 (p = 0.0018) favoring a treatment effect. ICHD-3 = International Classification of Headache Disorders, 3rd edition; OR = odds ratio; PROMIS = Patient-Reported Outcomes Measurement Information System.

As illustrated in Figure 2, among the patients who did not meet the ICHD-3 criteria, 13 of 69 (19%) had a much more robust response at long-term follow-up—improving a full US population SD in Global Physical Health (10 points on the T-score). The ratio of patients improving a full US population SD in Global Physical Health (10 points on the T-score) compared with those who declined a full SD was 13 to 1 (p = 0.0018), favoring a treatment effect.

Comparing Outcomes of Patients Who Did and Did Not Meet ICHD-3 Criteria

Patients who did not meet the ICHD-3 criteria improved to the same degree as patients who met the ICHD-3 criteria with no significant difference in any health metric between the 2 groups after a first patch (Table 5). However, when assessed after repeated patches and with long-term follow-up, patients who met the ICHD-3 criteria improved more than patients who did not meet the ICHD-3 criteria in 3 domains—global physical health (PROMIS Global Health Physical: +5.0 [95% CI 0.6–9.4]), fatigue (PROMIS Fatigue: −6.3 [95% CI −10.6 to −0.3]) and neurocognitive abilities (PROMIS Neurocognitive abilities: +7.8 [95% CI 2.5–13.0]).

Table 5.

Results of Epidural Patching by Patient ICHD-3 Status

Survey scores	Pre to post-patch change for ICHD-3 negative (n = 69)	Pre to post-patch change for ICHD-3 positive (n = 16)	Difference	p Value
	Mean ± SD	Mean ± SD	(95% CI)
Early outcomes: Change from pre-patch to first post-patch assessment (mean 277 d)
PROMIS physical health	3.2 ± 6.7	5.2 ± 8.6	2.0 (−2.8 to 6.8)	0.40
PROMIS mental health	0.9 ± 6.8	2.4 ± 9.0	1.5 (−3.5 to 6.6)	0.53
Headache Impact Test-6	−3.6 ± 8.9	−3.5 ± 12.6	0.1 (−6.9 to 7.1)	0.98
Neck Disability Index	−3.5 ± 17.9	−10.0 ± 16.4	−6.5 (−16.1 to 3.1)	0.17
PROMIS Fatigue	−2.2 ± 7.6	−3.5 ± 8.3	−1.3 (−3.8 to 3.3)	0.98
Rhodes Index of Nausea, Vomiting, and Retching	−0.2 ± 1.0	−0.2 ± 0.9	0.0 (−0.3 to 0.5)	0.97
PROMIS Applied Cognition	0.03 ± 7.4	3.8 ± 9.9	3.8 (−2.1 to 9.7)	0.19
Tinnitus Handicap Inventory	2.1 ± 18.3	−3.0 ± 25.4	−5.1 (−18.0 to 4.0)	0.46
Dizziness Handicap Inventory	−2.0 ± 20.2	−7.1 ± 26.0	−5.1 (−19.7 to 9.3)	0.47
Durable outcomes: Change from pre-patch to last assessment (mean 521 d from first patch, 377 d from last patch)
PROMIS physical health	3.3 ± 7.5	8.3 ± 7.7	5.0 (0.6 to 9.4)	0.03
PROMIS mental health	1.8 ± 7.3	5.9 ± 8.7	4.1 (−0.8 to 9.0)	0.09
Headache Impact Test-6	−3.8 ± 7.9	−5.9 ± 13.33	−2.1 (−9.5 to 5.1)	0.54
Neck Disability Index	−4.8 ± 17.5	−12.5 ± 17.0	−7.7 (−17.5 to 2.1)	0.12
PROMIS Fatigue	−2.1 ± 7.7	−8.4 ± 9.9	−6.3 (−10.6 to −0.3)	0.03
Rhodes Index of Nausea, Vomiting, and Retching	−0.1 ± 1.0	−0.3 ± 0.8	−0.2 (−0.5 to 0.2)	0.55
PROMIS Applied Cognition	0.1 ± 7.4	7.9 ± 8.7	7.8 (2.5 to 13.0)	0.006
Tinnitus Handicap Inventory	4.4 ± 19.8	−10.0 ± 27.0	−14.4 (−30 to 0)	0.13
Dizziness Handicap Inventory	−2.4 ± 21.3	−9.0 ± 28.9	−6.6 (−22.7 to 9.5)	0.40

Abbreviations: ICHD-3 = International Classification of Headache Disorders, 3rd edition; PROMIS = Patient-Reported Outcomes Measurement Information System.

Clinically Meaningful Improvement

In the cohort as a whole, 50 of 85 patients (60%) experienced a clinically meaningful improvement in PROMIS Global Physical Health after patching (Table 3), including 37 of 69 (54%) of those not meeting the ICHD-3 criteria and 13 of 16 of those meeting the ICHD-3 criteria (81%). While only 19% of patients met the ICHD-3 criteria, 68% (58 of 85 patients) experienced a clinically meaningful improvement in either PROMIS Global Physical Health or HIT-6 at 377 days after a mean of 3.6 epidural patches.

No patients in this cohort developed any serious adverse events (e.g., epidural hematoma, abscess, or neurologic injury). Frequency of adverse events such as lower back pain and rebound intracranial hypertension were not formally recorded.

Classification of Evidence

This study provides Class IV evidence that epidural blood patch is an effective treatment of suspected CSF leak not conforming to the ICHD-3 criteria for SIH.

Discussion

We report on a consecutive cohort of 85 patients presenting with symptoms of CSF leak. Eighty-one percent did not meet the ICHD-3 diagnostic criteria for low-pressure headaches, yet over half of this subset experienced sustained, statistically significant, and clinically meaningful improvements in global physical health outcomes after epidural patching. Two constituents of orthostatic headache—decreased head pain after attaining a flat position for an hour and decreased head pain after staying flat overnight—were significantly associated with the subsequent likelihood of sustained and clinically meaningful improvement in global physical health after patching. Specifically, among patients who did not meet the ICHD-3 criteria who reported their head pain was reduced to a 1 out of 10 or less after being flat for an hour, 76.2% experienced a durable, clinically meaningful response. This pattern of response argues against the null hypothesis that improvements were due to a placebo effect, regression to the mean, or other nonspecific treatment effects. It also suggests that degree of relief of head pain when flat, rather than severity of head pain when upright, is the clinically relevant dimension of orthostatic headache that treaters should elicit from a patient with a suspected CSF leak.

In our cohort, relying on the ICHD-3 criteria to determine suitability for epidural patches would have disadvantaged patients. Fifty patients experienced a clinically significant improvement in global physical health, a noteworthy 74% of which were classified as not meeting the ICHD-3 criteria. For every patient who conformed to the ICHD-3 criteria and benefited significantly from epidural patching, nearly 3 others who demonstrated similar meaningful improvement would have been disqualified from this intervention if positive imaging or low opening pressure had been required for their inclusion. Our findings are the first to provide robust evidence supporting the ICHD-3 diagnostic criteria addendum that it is reasonable to pursue a course of EBP in patients with characteristic orthostatic headaches when an alternative clear cause is not apparent. Nonetheless, while a greater number of responders arose from the patients not conforming to the ICHD-3 criteria, a greater response rate was seen among patients who did conform to the ICHD-3 criteria (81% vs 54%). These findings suggest that clinicians may find that their success rate per patch is higher if they choose to only patch those patients meeting the ICHD-3 criteria; however, a more inclusive set of eligibility criteria is needed to maximize the number of patients with clinically meaningful responses.

Our findings among patients who met the ICHD-3 criteria are equally noteworthy. Among the patients who met the ICHD-3 criteria who had confirmatory imaging or opening pressure, their diagnosis was only established after a mean of 8.4 years of symptoms previously ascribed to other diagnoses. The role of misdiagnosis among patients with CSF leak has been described elsewhere, but our data suggest a much longer time frame than previously reported.⁴⁴ Perhaps more remarkable, durable clinically meaningful improvement was seen after epidural patching in 13 of 16 of these patients despite nearly a decade of symptoms. We are unaware of any comparable long-term outcome data consisting of prospectively ascertained validated outcome metrics, documenting the effectiveness of epidural patching among patients with such longstanding symptoms.

Opening pressure alone would have resulted in misdiagnosis in 72% of patients with an imaging-confirmed CSF leak. This further substantiates growing data that opening pressure lacks sensitivity for detecting a CSF leak.^6,8

Fatigue as a core symptom of SIH has not been previously described. A recent systematic review and meta-analysis did not include fatigue in a table of 15 CSF leak symptoms but noted its possibility as less common.¹³ By contrast, a recent narrative review suggested that up to 25% of consecutive patients in their series self-reported fatigue.¹ Previous reports have not used validated instruments to assess fatigue presence, severity, or treatment responsiveness in patients with confirmed or suspected CSF leak. Fatigue was reported in 15 of 16 patients (94%; 95% CI 0.7–1.0) meeting strict ICHD-3 SIH criteria. On average, fatigue intensity in these patients was severe, with a T-score of 66.8 ± 7.4 (Table 1), surpassing the 95th percentile of the US population. In patients with confirmed SIH, fatigue severity resembled that in a large ME/CFS cohort with a mean PROMIS Fatigue T score of 68.6,⁴⁵ exceeding PROMIS Fatigue T scores in 2 sizable multiple sclerosis cohorts by nearly 10 points (a full SD from the US population).²⁶

Fatigue in patients who met the ICHD-3 criteria for SIH responded to epidural patching. After patching, average fatigue T-scores reduced by 8.6 points (Table 5), decreasing nearly a full SD within the US population. The frequency, severity, and treatment responsiveness of fatigue in patients who met the ICHD-3 criteria suggest it could be a core disabling feature of SIH. Further research should examine whether specific treatments targeting fatigue improve quality of life in patients with refractory SIH.

The extent of perceived neurocognitive deficits or responsiveness to treatment are also not well described in detail in the literature on SIH outside the context of comorbid superficial siderosis or behavioral-variant frontotemporal dementia.^46,47 Previous reports have not reported the frequency, severity, and responsiveness to treatment of perceived neurocognitive impairment in patients with CSF leak using validated instruments. Of the patients who met the ICHD-3 criteria within our cohort, 81% endorsed cognitive impairment before epidural patching as defined by “difficulty thinking, concentrating, or remembering.” The mean PROMIS Applied Cognition T-score before epidural patching was 37.9 ± 8.7, placing such patients at the 11th percentile for the US population. After patching, T-scores improved markedly: 7.9 ± 8.7 (10 points reflecting a SD in perceived neurocognitive function in the US population). Further research is needed to validate these findings and objectively characterize the specific neurocognitive deficits in patients with SIH.

In our cohort, 75% of patients who met the ICHD-3 criteria reported at least 1 migraine-associated symptom, most commonly phonophobia (62.5%). Migraine symptoms at baseline did not distinguish between patients with or without a positive CSF leak diagnosis, nor did they predict treatment outcomes after epidural patching (OR 1.81, 95% CI 0.28–11.6; p < 0.87). These data suggest the presence of migrainous features should not inhibit further evaluation for a CSF leak and possible treatment when otherwise appropriate.

In our study, only 19% of patients with orthostatic headache showed positive imaging or low opening pressure, aligning with previous prospective evaluations.⁴⁸ While focusing solely on patients who meet the ICHD-3 criteria may enhance uniformity and facilitate comparison across studies, it sacrifices generalizability and overlooks a much larger patient population lacking established diagnostic markers.

This study has important limitations. In the absence of a control group given sham patches, we cannot exclude nonspecific treatment effects as responsible for some of the improvement seen. In addition, it is possible that some portion of the improved patients had a primary headache disorder and the improvement was not related to the EBP. However, patients' prepatch relief when flat was strongly associated with the subsequent likelihood of clinically meaningful benefit over a year after patching. Furthermore, the positive effects of patching seemed to increase, rather than diminish, over extended time periods and repeated patching. This pattern among responders strengthens a causal inference that patches had a specific treatment effect.

In addition, nonrandom completion of a post-treatment questionnaire (nonparticipation bias) could limit generalizability to outside groups of patients with orthostatic headache. We compared available baseline measures of the patients who did not complete follow-up with those who did and found no significant differences, suggesting the participating cohort is representative of the larger group of all patients who received patches after completing baseline measures (eTable 2).

Another limitation involves CVF. One recent report suggests that 10% of patients with an orthostatic headache and negative brain and spine MRI later have a CVF identifiable with lateral decubitus digital subtraction myelogram—a technique not available at the time of our data acquisition.¹⁸ These data would suggest that 10% of our subset of patients who did not meet the ICHD-3 criteria might have had an undiagnosed CVF and might have responded to epidural patching on that basis.⁴⁹

HIT-6 head pain scores improved to 64.1 ± 8.23 (p < 0.001) at a mean 521 days after the first patch, and these scores after patching are still in the severe range on average. In addition, differing estimates of the within-person clinically meaningful change in the HIT-6 exist and would yield different calculated rates of meaningful responders.

Owing to the exploratory analysis of this project, p values were not adjusted for multiple testing. This raises the risk that any one of the significant findings reflects a Type I error due to multiple testing; significant results must be interpreted with caution and bear repetition.

In the present cohort characterized by severe symptoms of CSF leak, a majority of patients did not fulfill the ICHD-3 diagnostic criteria. Despite this, a therapeutic strategy using permissive epidural patching led to clinically meaningful and statistically significant enhancements across a spectrum of validated measures, both symptom-specific and related to overall health-related quality of life. The relative absence of head pain after lying flat for an hour and when in the supine position overnight forecasted sustained, meaningful improvement in global physical health after EBP. Further research confirming these findings is indicated.

Glossary

CVF

CSF-venous fistula

EBP

epidural blood patch

ICHD-3

International Classification of Headache Disorders, Third Edition

PROMIS

Patient-Reported Outcomes Measurement Information System

SIH

spontaneous intracranial hypotension

Appendix. Authors

Name	Location	Contribution
Ian Carroll, MD, MS	Department of Anesthesiology, Perioperative and Pain Medicine, and Stanford Headache Clinic, Stanford University, Palo Alto, CA	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data
Lichy Han, MD, PhD	Department of Anesthesiology, Perioperative and Pain Medicine, Stanford University, Palo Alto, CA	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Niushen Zhang, MD	Department of Neurology, Stanford University, Palo Alto, CA	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; analysis or interpretation of data
Robert P. Cowan, MD	Departments of Neurology and Neurosciences and (by courtesy) Anesthesia, Stanford University, Palo Alto, CA	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; analysis or interpretation of data
Bryan Lanzman, MD	Department of Radiology, Stanford University, Palo Alto, CA	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Syed Hashmi, MD	Department of Radiology, Stanford University, Palo Alto, CA	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; analysis or interpretation of data
Meredith J. Barad, MD	Department of Anesthesiology, Perioperative and Pain Medicine, and Department of Neurology, Stanford University, Palo Alto, CA	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data
Addie Peretz, MD	Department of Neurology, Stanford University, Palo Alto, CA	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Leon Moskatel, MD	Department of Neurology, Stanford University, Palo Alto, CA	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Oyindamola Ogunlaja, MBBS, MSc	Department of Neurology, Stanford University, Palo Alto, CA	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Jennifer M. Hah, MD	Department of Anesthesiology, Perioperative and Pain Medicine, Stanford University, Palo Alto, CA	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design
Nada Hindiyeh, MD	Headache Neurology, Metrodora Institute, West Valley City, UT	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Carol Barch, MN	Department of Neurology, Stanford University, Palo Alto, CA	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Selene Bozkurt, PhD	Department of Biomedical Informatics, Emory University, Atlanta, GA	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Tina Hernandez-Boussard, PhD	Department of Medicine (Biomedical Informatics), Stanford University, Palo Alto, CA	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Andrew L. Callen, MD	Department of Radiology, University of Colorado Anschutz Medical Campus, Aurora	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data

Study Funding

This work was funded by the Considine CSF Leak Fund, an unrestricted philanthropic gift to Stanford University.

Disclosure

N. Hindiyeh reports work for Eli lilly, Impel, Amgen, and Zosano. C. Barch reports consultant work for Abbvie—Migraine. S. Bozkurt reports work as an academic collaborator and paid consultant for Flatiron Health, Inc. T. Henrnadez-Boussard reports consultant work for Grai-Matter and Board Membership for Athelo Health. A.L. Callen reports consultant work for Eli Lilly. The other authors report no disclosures. Go to Neurology.org/N for full disclosures.