When Orthostatic Headache Misleads: Positional Headache Phenotypes Across Spinal CSF Leak Types and Implications for Patients Without SIH Findings on Brain MRI

<h3>BACKGROUND AND PURPOSE:</h3> Orthostatic headache is considered the hallmark of spontaneous intracranial hypotension from spinal CSF leaks, yet whether positional headache patterns differ across leak mechanisms is unknown. We compared positional headache phenotypes across confirmed leak types and assessed whether classic orthostatic headache preferentially identifies CSF-venous fistulas (CVFs) in patients with limited brain MRI support for underlying SIH. <h3>MATERIALS AND METHODS:</h3> Patients from a single-center CSF leak registry were classified by myelographic diagnosis: confirmed CVF, ventral tear, lateral tear, or myelography-negative. Positional headache was classified from deidentified clinical notes using structured large language model–assisted abstraction into predefined categories (classic orthostatic, partial orthostatic, evolving, mixed, non-positional), with each classification linked to a direct quotation from the source note. For each patient, the initial encounter note within our program was used to capture presenting symptom phenotype prior to intervention. A focused subanalysis compared low Bern score (≤4) CVF patients with the myelography-negative cohort. <h3>RESULTS:</h3> Of 298 patients classified by myelographic diagnosis (104 CVF, 47 ventral tear, 19 lateral tear, 128 myelography-negative), clinical positional data were available for 129. Classic orthostatic headache predominated in dural tears (44%) and myelography-negative patients (50%) but was present in only 25% of CVF patients. Among low Bern score patients without spinal epidural fluid collections, classic orthostatic headache was present in only 17% of CVF positive patients versus 50% of myelography-negative patients (OR 0.23; 95% CI 0.07–0.75; p=.02). This effect was stable across sensitivity comparator definitions and alternative Bern score thresholds. Partial orthostatic headache was the dominant low Bern CVF pattern (52%), and subtle cisternal measurements did not discriminate between groups. <h3>CONCLUSIONS:</h3> Positional headache phenotype differs by spinal CSF leak mechanism. Dural tears more often conformed to the classic orthostatic paradigm, whereas classic orthostatic headache was a minority phenotype in CVF. Among patients without typical brain MRI findings of SIH, classic orthostatic headache was more frequently observed in myelography-negative patients than in those with confirmed CVF. These findings suggest that the absence of textbook orthostatic headache should not diminish clinical suspicion for CVF in appropriately selected patients.