Prevention of Postoperative Cerebrospinal Fluid Leaks After Translabyrinthine Tumor Resection With Resorbable Mesh Cranioplasty

Jacob B Hunter,Alex D Sweeney,Matthew L Carlson,George B Wanna,Alejandro Rivas,Kyle D Weaver,Lola B Chambless,Reid S Thompson,David S Haynes,Marc L Bennett
DOI: https://doi.org/10.1097/MAO.0000000000000835
Abstract:Objectives: To evaluate the effectiveness of resorbable mesh cranioplasty at reducing postoperative cerebrospinal fluid (CSF) leak and pseudomeningocele formation after translabyrinthine tumor resection. Study design: Case series with chart review. Setting: Tertiary academic referral center. Patients: Fifty-three consecutive cases using a resorbable mesh cranioplasty after translabyrinthine tumor resection were reviewed. Intervention: Temporal bone defects were repaired with a dural substitute, layered fat graft, and a resorbable mesh plate secured with screws. Main outcome measures: Primary outcome measures included the incidence of postoperative CSF wound leak or rhinorrhea, pseudomeningocele formation, and surgical site infection. Results: Fifty-three cases (average age, 54.0 yr; range, 19.3-75.1 yr) were analyzed. The average body mass index was 30.8 kg/m2 (range, 17.9-48.3 kg/m2), and the average tumor size was 18.8 mm (range, 8-38 mm). One patient (1.9%) experienced CSF rhinorrhea on postoperative Day 16, which resolved after transmastoid middle ear and eustachian tube packing. One patient (1.9%) experienced a surgical site infection requiring surgical debridement and mesh removal 4 months after surgery. Compared with 1,441 prior translabyrinthine surgeries analyzed from our institution using a traditional fat graft closure without mesh, the rate of postoperative CSF leak was significantly less using the resorbable mesh cranioplasty technique (p = 0.0483). Conclusion: Resorbable mesh cranioplasty is a safe and effective method to reduce postoperative CSF leak and pseudomeningocele formation after translabyrinthine craniotomy for tumor excision.
What problem does this paper attempt to address?