How long is the tail end of the learning curve? Results from 1000 consecutive endoscopic endonasal skull base cases following the initial 200 cases.

Endoscopic endonasal approaches (EEAs) to the skull base have evolved over the last 20 years to become an essential component of a comprehensive skull base practice. Many case series show a learning curve from the earliest cases, in which the authors were inexperienced or were not using advanced closure techniques. It is generally accepted that once this learning curve is achieved, a plateau is reached with little incremental improvement. Cases performed during the early steep learning curve were eliminated to examine whether the continued improvement exists over the "tail end" of the curve. A prospectively acquired database of all EEA cases performed by the senior authors at Weill Cornell Medicine/NewYork-Presbyterian Hospital was reviewed. The first 200 cases were eliminated and the next 1000 consecutive cases were examined to avoid the bias created by the early learning curve. Of the 1000 cases, the most common pathologies included pituitary adenoma (51%), meningoencephalocele or CSF leak repair (8.6%), meningioma (8.4%), craniopharyngioma (7.3%), basilar invagination (3.1%), Rathke's cleft cyst (2.8%), and chordoma (2.4%). Use of lumbar drains decreased from the first half to the second half of our series (p <0.05) as did the authors' use of fat alone (p <0.005) or gasket alone (p <0.005) for dural closure, while the use of a nasoseptal flap increased (p <0.005). Although mean tumor diameter was constant (on average), gross-total resection (GTR) increased from 60% in the first half to 73% in the second half (p <0.005). GTR increased for all pathologies but most significantly for chordoma (56% vs 100%, p <0.05), craniopharyngioma (47% vs 0.71%, p <0.05) and pituitary adenoma (67% vs 75%, p <0.05). Hormonal cure for secreting adenomas also increased from 83% in the first half to 89% in the second half (p <0.05). The rate of any complication was unchanged at 6.4% in the first half and 6.2% in the latter half of cases, and vascular injury occurred in only 0.6% of cases. Postoperative CSF leak occurred in 2% of cases and was unchanged between the first and second half of the series. This study demonstrates that contrary to popular belief, the surgical learning curve does not plateau but can continue for several years depending on the complexity of the endpoints considered. These findings may have implications for clinical trial design, surgical education, and patient safety measures.