Spontaneous CSF leaks into the anterior skull base nasal sinuses are often associated with meningoencephaloceles and occur in patients with idiopathic intracranial hypertension (IIH). Endonasal endoscopic repair has become the primary method of choice for repair. The authors sought to evaluate the success rate of endoscopic closure and to identify predictive factors for CSF leak recurrence. A consecutive series of endonasally repaired anterior skull base meningoencephaloceles was drawn from a prospectively acquired database. Lumbar punctures were not performed as part of a treatment algorithm. All patients had at least 5 months of follow-up. Chart review and phone calls were used to determine the timing and predictors of recurrence. Demographic information and details of operative technique were correlated with recurrence. Two independent radiologists reviewed all preoperative imaging to identify radiographic markers of IIH, as well as the location and size of the meningoencephalocele. From a total of 54 patients there were 5 with recurrences (9.3%), but of the 39 patients in whom a vascularized nasoseptal (n = 31) or turbinate (n = 8) flap was used there were no recurrences (p = 0.0009). The mean time to recurrence was 24.8 months (range 9-38 months). There was a trend to higher BMI in patients whose leak recurred (mean [± SD] 36.6 ± 8.6) compared with those whose leak did not recur (31.8 ± 7.4; p = 0.182). Although the lateral recess of the sphenoid sinus was the most common site of meningoencephalocele, the fovea ethmoidalis was the most common site in recurrent cases (80%; p = 0.013). However, a vascularized flap was used in significantly more patients with sphenoid (78.3%) defects than in patients with fovea ethmoidalis (28.6%) defects (Fisher's exact test, p = 0.005). Radiographic signs of IIH were equally present in all patients whose leak recurred (75%) compared with patients whose leak did not recur (63.3%); however, an enlarged Meckel cave was present in 100% (2/2) of patients whose leaks recurred compared with 13.3% (4/30) of patients whose leaks did not recur (p = 0.03). The average meningoencephalocele diameter tended to be larger (1.73 ± 1.3 cm) in patients with recurrence compared to those without recurrence (1.2 ± 0.66 cm; p = 0.22). A ventriculoperitoneal shunt was already in place in 3 patients, placed perioperatively in 5, and placed at recurrence in 2, none of whose leaks recurred. Recurrence after endonasal repair of spontaneous CSF leaks from meningoencephaloceles can be dramatically reduced with the use of a vascularized flap. Although failures of endonasal repair tend to occur in patients who have higher BMI, larger brain herniations, and no CSF diversion, the lack of vascularized flap was the single most important risk factor predictive of failure.