Digital Subtraction Myelography for the Detection of Type 1 Spinal CSF Leaks: Evaluation of Temporal Characteristics and Diagnostic Value

BACKGROUND AND PURPOSE: Ventral dural tears (type 1 leaks) are reported to be the most common cause of spontaneous intracranial hypotension (SIH) and may require high-dynamic myelography for detection. The aim of this cross-sectional study was to evaluate the temporal characteristics and diagnostic value of digital subtraction myelography (DSM) in type 1 leaks.

MATERIALS AND METHODS: Between April 2022 and August 2024, one hundred four consecutive patients with type 1 leaks were retrospectively identified. The institutional diagnostic standard included DSM as a first-line examination; when deviating from this standard, patients were excluded. A previously described positioning technique of patients was used, enabling examination even in the challenging cervicothoracic junction of the spine. We evaluated the time for the contrast agent to first appear in the epidural space after reaching the level of the leak intrathecally and the overall diagnostic yield of DSM.

RESULTS: One hundred of 104 patients (49 women) were included. The mean age was 49 (SD, 11.9) years; mean body mass index, 24.8 (SD, 4.29); and the median Bern SIH score, 4 (interquartile range, 5). Type 1 leaks most commonly occurred at the T1/2 and T2/3 level (each, 20/100; range, C6/7–L1/2). The mean time for the contrast to be visible in the epidural space was, on average, 1.5 seconds (range, 0–9 seconds) with 1 frame per second (fps) acquisition being sufficient for all except 1 patient. DSM as a first-line investigation made the diagnosis in 76/100 patients, confirmed in all patients undergoing surgery (74/76). Twenty-four of 100 patients required ≥1 subsequent dynamic CT myelography (dCTM) for a definite diagnosis on another day, with 21/24 leaks confirmed at surgery. The Bern SIH score was significantly lower in dCTM compared with the DSM group (3.25 versus 5; P = .009), whereas age (P = .548) and body mass index (P = .185) were not found to have an impact.

CONCLUSIONS: DSM demonstrated a high diagnostic yield for type 1 leaks when used as a first-line investigation. We have confirmed the high-flow characteristics of these leaks, suggesting that the high temporal resolution of DSM was ideally suited for their detection, with 1 fps being overall sufficient. A lower Bern SIH score could favor patients for the primary use of dCTM; however, further research may clarify why DSM occasionally misses diagnoses.