Pneumocephalus patterns following endonasal endoscopic skull base surgery as predictors of postoperative CSF leaks.

OBJECTIVES: Postoperative pneumocephalus is a common occurrence after endoscopic endonasal skull base surgery (ESBS). The risk of cerebrospinal fluid (CSF) leaks can be high and the presence of postoperative pneumocephalus associated with serosanguineous nasal drainage may raise suspicion for a CSF leak. The authors hypothesized that specific patterns of pneumocephalus on postoperative imaging could be predictive of CSF leaks. Identification of these patterns could guide the postoperative management of patients undergoing ESBS.
METHODS: The authors queried a prospectively acquired database of 526 consecutive ESBS cases at a single center between December 1, 2003, and May 31, 2012, and identified 258 patients with an intraoperative CSF leak documented using intrathecal fluorescein. Postoperative CT and MRI scans obtained within 1-10 days were examined and pneumocephalus was graded based on location and amount. A discrete 0-4 scale was used to classify pneumocephalus patterns based on size and morphology. Pneumocephalus was correlated with the surgical approach, histopathological diagnosis, and presence of a postoperative CSF leak.
RESULTS: The mean follow-up duration was 56.7 months. Of the 258 patients, 102 (39.5%) demonstrated pneumocephalus on postoperative imaging. The most frequent location of pneumocephalus was frontal (73 [71.5%] of 102), intraventricular (34 [33.3%]), and convexity (22 [21.6%]). Patients with craniopharyngioma (27 [87%] of 31) and meningioma (23 [68%] of 34) had the highest incidence of postoperative pneumocephalus compared with patients with pituitary adenomas (29 [20.6%] of 141) (p < 0.0001). The incidence of pneumocephalus was higher with transcribriform and transethmoidal approaches (8 of [73%] 11) than with a transsellar approach (9 of [7%] 131). There were 15 (5.8%) of 258 cases of postoperative CSF leak, of which 10 (66.7%) had pneumocephalus, compared with 92 (38%) of 243 patients without a postoperative CSF leak (OR 3.3, p = 0.027). Pneumocephalus located in the convexity, interhemispheric fissure, sellar region, parasellar region, and perimesencephalic region was significantly correlated with a postoperative CSF leak (OR 4.9, p = 0.006) and was therefore termed "suspicious" pneumocephalus. In contrast, frontal or intraventricular pneumocephalus was not correlated with postoperative CSF leak (not significant) and was defined as "benign" pneumocephalus. The amount of convexity pneumocephalus (p = 0.002), interhemispheric pneumocephalus (p = 0.005), and parasellar pneumocephalus (p = 0.007) (determined using a scale score of 0-4) was also significantly related to postoperative CSF leaks. Using a series of permutation-based multivariate analyses, the authors established that a model containing the learning curve, the transclival/transcavernous approach, and the presence of "suspicious" pneumocephalus provides the best overall prediction for postoperative CSF leaks.
CONCLUSIONS: Postoperative pneumocephalus is much more common following extended approaches than following transsellar surgery. Merely the presence of pneumocephalus, particularly in the frontal or intraventricular locations, is not necessarily associated with a postoperative CSF leak. A "suspicious" pattern of air, namely pneumocephalus in the convexity, interhemispheric fissure, sella, parasellar, or perimesencephalic locations, is significantly associated with a postoperative CSF leak. The presence and the score of "suspicious" pneumocephalus on postoperative imaging, in conjunction with the learning curve and the type of endoscopic approach, provide the best predictive model for postoperative CSF leaks.