Midline suprasellar tumors represent a unique surgical challenge for neurosurgeons. A variety of growths occur in this space, including craniopharyngioma, meningioma, germinoma, metastasis, pituitary adenoma, lymphoma, and glioma. Resection has remained a mainstay of treatment for many of these tumors, although significant debate exists as to the benefit of attempted gross-total resection (GTR).^{5,9,12,13,19,27,41,48,49,52}

Transcranial approaches have been traditionally used to remove suprasellar masses. Pterional, bifrontal, and subfrontal approaches have all been described. These procedures can require a significant amount of brain retraction, which may result in edema, contusion, or venous compromise.^{14,38,45,47,53} In addition, the dissection often requires manipulation of neurovascular structures such as the internal carotid artery and its branches, as well as the cranial nerves, including the optic apparatus.^{29,32,35,38,52,53} Moreover, direct visualization and/or dissection of the tumor may be difficult due to its location below the optic nerves, filling of the third ventricle, or extension into the posterior fossa. Not surprisingly, complications involving injury to these neurovascular structures via these approaches have been reported.^{5,14,38,45,47,53}

In recent years, expanded endoscopic endonasal transsphenoidal surgery (EETS) has become increasingly popular for access to the midline suprasellar space.^{2,4,7,16,18,36} Via transplanum extensions, this approach provides direct access to the infrachiasmatic, olfactory, and subfrontal regions; and via transclival extensions, to the prepontine and perimesencephalic regions. The direct visualization to the skull base afforded by this approach minimizes brain retraction and manipulation of neurovascular structures, and provides a wide operative field compared with standard microscope-based transsphenoidal approaches.^{3,15,16,24,28,31}

Although both approaches have their advantages and advocates, there is no clear consensus whether one is superior to the other. The most direct way to resolve this issue would be to perform a comparison of surgical outcomes between these 2 approaches. Several authors have documented their experience with endoscopic endonasal approaches for suprasellar lesions.^{2,7,16,17,30,31,34} Similarly, there are an abundant number of reports of outcomes following transcranial approaches but no series directly comparing these approaches in a single- or multiinstitutional cohort. One systematic review has compared EETS to cranial approaches based on reported case series.²⁸ Because the goals, such as the aggressiveness of resection, may vary among institutions, it is difficult to compare outcomes in a systematic review of this nature.

To better address this issue, we compared the surgical outcomes, including extent of resection (EOR), complication rate, and progression-free survival (PFS), between the 2 approaches for cases treated at a single institution.

Methods

All patients with midline suprasellar lesions that were resected at Cedars-Sinai Medical Center between January 2000 and March 2013 were identified by retrospective review of the Department of Pathology surgical case log. Patients with histologically verified pathologies of craniopharyngioma, pituicytoma, and xanthogranuloma were included in this review. Meningioma, pituitary adenoma, glioma, and other nonstalk-based lesions were excluded, as were stalk-based lesions requiring only biopsy, such as lymphoma, hypophysitis, or germinoma. These inclusion and exclusion criteria resulted in a subset of tumors of similar size and operative requirements, with similar operative techniques and expected goals of surgery. For each case, the patient’s medical chart was reviewed. Demographic information including patient age, sex, presenting symptoms, hormonal status, visual field examination, and previous therapies were recorded.

All patients who underwent operation before the year 2000 were excluded from this study to let the data reflect more current techniques and approaches, as well as a more unified institutional philosophy regarding the goals of surgery. Additionally, patients with primarily intrasellar tumors were not included in this study, because the aim of the study was to compare operative techniques for midline suprasellar/third ventricular lesions. Xanthogranuloma and pituicytoma cases were included because these lesions appear radiographically and clinically similar to craniopharyngioma, were presumed to represent craniopharyngioma prior to surgery, and the surgical strategy used mimicked that for craniopharyngioma. Surgeons only became aware of the alternate pathological type on postoperative histological examination. Therefore, the expected surgical morbidities and outcomes in these cases are similar enough to those of craniopharyngioma to warrant inclusion. Finally, cases with insufficient or incomplete records were also excluded.

The patients included in this study were arranged into 2 groups: those who underwent a craniotomy (including bifrontal, pterional, orbitozygomatic, or transcallosal), and those who underwent EETS. Because the primary objective of the analysis was to determine if either method was associated with more complications, more complete resection, or improved survival prognosis, the analysis focused primarily on surgical cases rather than individual patients. Since this study focused on comparing surgical procedures, patients who underwent reoperation for recurrence of their tumor, but via a different approach from that of the index surgery, were designated to both groups. For example, a patient who initially underwent craniotomy and later underwent EETS for a recurrence was evaluated for complications, EOR, and survival in both the craniotomy and EETS groups.

For each operative case, tumor volumes were measured from a pre- and postoperative MRI scan, using the line measurement tool on our picture archiving and communication system (PACS) station (GE Healthcare). Preoperative MRI scans were always obtained within 2 weeks prior to surgery, whereas postoperative scans were obtained either 1–2 days or 3 months after surgery, depending on surgeon preference and clinical indication. Tumor volumes were computed using the following equation: Tumor volume = (A × B × C)/2, where A, B, and C represent the dimensions of the tumor in 3 orthogonal planes. This equation assumes that the shape of the lesion approximates roughly a sphere. Although this measurement is only a reasonable estimate rather than an exact measure of tumor volume, it provided adequate assessment of the EOR with some degree of quantitation. All tumor volumes were recorded in units of cubic centimeters. The EOR for each case was computed by the following equation: EOR = 1 - postoperative volume/preoperative volume.

Additional characteristics of the tumor were also recorded, including the consistency of the tumor (solid vs cystic), as indicated by the surgeon’s operative note and the lesion’s radiographic appearance. Tumors were classified based on their location in relation to the infundibular stalk. Using the classification scheme proposed by Kassam et al.,²⁵ lesions located primarily anterior to the stalk were classified as Type I, lesions located intrinsic to the stalk were designated as Type II, and lesions posterior to the stalk were classified as Type III. The pathologies for all cases were recorded. Craniopharyngiomas were divided into adamantinomatous or papillary subtypes.

For each surgery, intraoperative and postoperative complications were tabulated. These included vascular, neurological, hormonal, and CSF leak complications. Follow-up data for each case were also tabulated, including whether the patient underwent reoperation or radiation for recurrent or progressive tumor, length of PFS, and time of last follow-up. The PFS was calculated as the time from initial surgery to radiation and/or additional surgery for MRI-documented evidence of tumor growth. This included both growth of existing residual disease and recurrence of disease in a patient with no radiographic evidence of disease on prior MR images. For cases with no progression or recurrence (i.e., censored), PFS was the same as the length of time from initial surgery to last follow-up.

A significant concern in a retrospective review of this nature is that selection bias influenced the surgical approach and that outcomes are reflective of this bias. For example, tumors with a more solid component might tend to be selected for craniotomy, whereas more cystic tumors would be selected for EETS. In turn this could affect results, such as EOR. To address this potential confounding factor, 3 neurosurgeons who were experienced in tumor resection (> 10 years clinical practice) and who were not involved in the cases were asked to evaluate the preoperative images and clinical data for each patient, and to select the surgical approach he or she would prefer for that case. The neurosurgeons were blinded to each other, to the actual surgeon or patient identifiers, and to any information about outcomes. These data were tabulated, compared with the actual procedure performed, and subjected to concordance analysis. In the absence of a randomized clinical trial (RCT), these data are useful to determine if certain tumors were always preferentially selected for craniotomy or EETS, or if the surgical approach was principally attributable to surgeon preference.

Statistical analysis for differences in demographic variables, EOR, complication rates, types of complications, PFS, and recurrence rate were calculated between craniotomy and EETS cases. Descriptive statistics were reported for all patients as well as for cohorts classified according to treatment group. Continuous variables were described using means, medians, and interquartile ranges (IQRs), whereas categorical variables were reported in terms of frequencies. Univariate comparisons were conducted using Student t-test, chi-square, and Fisher’s exact tests where appropriate. The Kaplan-Meier estimation method was used to obtain PFS functions for patients treated with craniotomy versus EETS surgery, and the curves were compared using the log-rank test. Cohen’s kappa test was used to evaluate concordance between surgical approaches among surgeons. A p value ≤ 0.05 was considered to be statistically significant. All statistical analyses were conducted in SAS 9.2 (SAS Institute).

Results

A total of 67 patients treated between 2000 and 2013 were identified. Cases treated prior to 2000, those with incomplete records, and those involving primarily intrasellar tumors were excluded, leaving 46 patients comprising 53 operative cases. Nineteen (35.8%) surgeries were performed via the EETS approach, whereas 34 (64.2%) were performed via a craniotomy. Among the 19 EETS procedures, 11 (57.9%) were initial operations and 8 (42.1%) treated a recurrent or progressive tumor that had previously been resected via an EETS or craniotomy. Of the 34 craniotomy cases, 28 (82.4%) were deemed initial operations and 6 (17.6%) procedures treated cases of recurrent or progressive disease. The proportion of first-time operations between the 2 groups differed, with a greater percentage of first-time resections performed in the craniotomy group (82.4% vs 57.9%, p = 0.05).

The mean age at surgery for the entire cohort (n = 53) treated with either a craniotomy or EETS procedure was 45.2 years, with an IQR of 38–59 years (Table 1). The average age among craniotomy (45.3 years) and EETS (45 years) patients was comparable (p = 0.96). Fewer female patients were treated with a craniotomy compared with EETS (44.1% vs 52.65%, p = 0.55). A description of preoperative symptoms showed that visual disturbances (54.7%) were most commonly observed, followed by endocrinopathy (41.5%), neurological symptoms (32.1%), and headaches (26.4%). The rates of preoperative symptoms were comparable between cases treated with a craniotomy and EETS, with the exception of a moderate difference in neurological symptoms (craniotomy: 41.2%, EETS: 15.8%; p = 0.07). Because preoperative symptoms encompassed a wide spectrum of conditions, Table 2 provides a detailed summary of the presenting signs and symptoms in each surgical group.

Table 3 demonstrates histological characteristics in the 2 groups. All tumors were midline suprasellar/third ventricular tumors, with the overwhelming majority (92.5%) of cases comprising craniopharyngioma. Other pathological types were pituicytoma and xanthogranuloma. Among the craniopharyngiomas, the frequency of subtypes based on histological characteristics (adamantinomatous vs papillary) and infundibular classification are also listed in Table 3. No significant difference in tumor pathology or consistency was seen between the 2 cohorts. Among craniopharyngiomas, no significant difference was seen in the distribution of infundibular class between the 2 groups. The preoperative tumor volume was similar between the 2 groups (9.5 cm³ vs 9.0 cm³, p = 0.84).

Within the EETS group, the predominant surgical approach was a transsphenoidal and transplanum approach. In one case a combined transplanum and transclival approach was used. In another case early in the series, a sublabial approach was used. In the craniotomy group, the surgical approach varied significantly depending on the anatomical location of the tumor and surgeon preference. The approaches included pterional (35.3%), bifrontal (35.3%), supraorbital keyhole craniotomy (8.8%), interhemispheric transcallosal (8.8%), transventricular (5.9%), orbitozygomatic (2.9%), and subtemporal (2.9%).

The mean EOR for all cases within each group is summarized in Table 4. Given that reoperations performed for recurrent or progressive tumors may entail more challenging resections than first-time cases due to more aggressive pathology, scarring, and difficult anatomical location, EOR was also subdivided and evaluated separately for first-time cases and reoperation cases. No significant difference was found in the EOR between the 2 cohorts, both overall (90.7% vs 85.6%, p = 0.77) and separately for first-time (p = 0.43) and redo operations (p = 0.36).

New endocrine deficits were the most frequent complications, representing 50.9% of all complications, followed by cranial nerve injuries (including visual deficits), representing 15.1% of all complications. Comparisons of specific complication rates between the craniotomy and EETS groups are depicted in Table 5. The most commonly encountered complications in both groups were panhypopituitarism and diabetes insipidus (DI), occurring with similar frequency in both groups (38.2% vs 42.1% for panhypopituitarism, and 52.9% vs 31.6% for DI).

The complications were grouped into endocrinological complications, cranial nerve injury including optic nerve injury, hydrocephalus/hygroma, stroke/hemorrhage, CSF leak, death, and other (Table 6). The incidence of cranial nerve injury including visual deficit was significantly higher in the craniotomy group (23.5% vs 0%, p = 0.04). These included injuries to the optic, oculomotor, trochlear, olfactory, and frontalis branch of the facial nerve. The incidence of CSF leaks was significantly higher in the EETS group (0% craniotomy vs 26.3% EETS, p = 0.004). Other complications, including hydrocephalus (p = 0.69), stroke (p = 1.00), and death (p = 1.00) were similar between the 2 groups.

The mean follow-up time in the craniotomy group was 1049 days. The mean follow-up time in the EETS group was 1022 days (p = 0.80). The PFS times as well as the rates of reoperation/radiation for recurrent or progressive tumor are shown in Table 7. Due to the fact that reoperation cases may have an increased likelihood of more aggressive and/or difficult-to-resect tumors, the PFS in this group may be different than in cases in which a first-time operation was performed. Therefore, the PFS for first-time and reoperation cases was calculated separately. The mean PFS times were similar between the 2 cohorts (760 vs 780 days, p = 0.78), including between the first-time operation (p = 0.91) and reoperation (p = 0.27) subcohorts. In addition, Kaplan-Meier curves were constructed for the PFS for both cohorts (Fig. 1). There was no significant difference in PFS curves for the craniotomy and EETS cohorts based on the log-rank test (p = 0.99).

Kaplan-Meier curves comparing PFS for patients undergoing EETS versus craniotomy for midline suprasellar tumors. Figure is available in color online only.

The rate of further treatment for recurrent or progressive disease, whether surgical or radiation, was similar in the 2 groups (23.5% vs 21.1%, p = 1.00) (Table 7). Surgical retreatment included both craniotomy and EETS. Additionally, the percentage of cases that received further treatment for recurrent or progressive disease following GTR during the index procedure was 8.8% and 0% (p = 0.55) for the craniotomy and EETS groups, respectively.

The overall rate of craniotomy in this cohort was 64.2% and for EETS it was 35.8%. Analysis of the surgical approach preferences of 3 independent surgeons who were not involved in the treatment of this cohort (Table 8) found that one of the surgeons had a strong preference toward craniotomy (85%), another had a strong EETS preference (87%), and the other had a similar preference toward craniotomy (53%) or EETS (47%). In evaluating intersurgeon concordance we found a kappa coefficient of −0.039 (p = 0.7624). This statistic indicates a genuine disagreement between the surgeons as to the best approach. That is, surgeons agree on a particular surgery less than would be expected by chance alone. Further analysis identified only 9 cases (17%) in which all 3 blinded reviewers agreed on the same surgical approach (5 EETS, 4 craniotomy), which is less than the 25% odds that there would be agreement based on pure chance. Similarly, there were only 6 cases (11%) in which all 3 blinded reviewers chose concordantly with the actual surgery performed, whereas there was a 25% chance that this would occur randomly. Taken as a whole, these data firmly support the notion that a surgeon’s expertise was the strongest factor driving surgical choice in this series. No other metric was identified as a factor in determining surgical approach, including tumor size, tumor location, tumor shape or consistency, patient age, history of previous surgery, or clinical symptoms.

An analysis of surgical trends over time suggests an ongoing change in philosophy toward each surgical method in our institution (Fig. 2). The EETS approach was first used in 2005, with the arrival of a surgeon trained in those techniques. Despite the fact that the overall number of cases being resected remained stable, use of EETS increased while the number of craniotomies decreased.

Bar graph showing percentage of cases treated with EETS versus craniotomy between 2000 and 2013. Figure is available in color online only.

Discussion

Tumors occupying the midline suprasellar space represent a diverse array of pathological types. Of these, craniopharyngiomas are the most common, accounting for 1%–4% of all intracranial tumors.⁴³ Resection and radiation have been the mainstays of treatment.^23,27,52 Total resection of craniopharyngiomas has been associated with improved long-term outcomes according to several studies;^{5,9,12,13,19,41,48,49,53} however, long-term outcomes in patients with GTR are similar to those receiving partial resection with adjuvant radiation,⁵² leading many surgeons to advocate for less aggressive resections to reduce surgical morbidities.

The constraints imposed by the nearby critical neurovascular structures in the suprasellar region have led to the development of several approaches. Pterional, orbitozygomatic, and subfrontal approaches rely on the surgeon developing “corridors” to access the suprasellar space, requiring retraction of brain tissue, and working between neurovascular structures.^{29,32,35,38,52,53} This may result in highly morbid perioperative complications, such as cerebral edema, venous infarct, carotid artery injury, stroke, and cranial nerve injury.

First reported by Locatelli et al., endoscopic approaches to the skull base have become an increasingly popular method of treatment of suprasellar lesions.^{2,4,7,16,18,36} EETS has several advantages over traditional open approaches, including direct access to the skull base without retraction of neural tissue or manipulation of neurovascular structures.^{3,15,16,24,27,31} Additionally, the third ventricle and hypothalamus can be well visualized using endoscopic endonasal approaches.¹⁵

Extended microscopic transsphenoidal approaches have been described for treatment of suprasellar lesions.^22,29,40 EETS has allowed significantly wider surgical exposure than the traditional microscopic approach. Transplanum and transclival extensions of the expanded endoscopic endonasal approach permit access to the suprasellar and prepontine spaces, respectively, thus providing a significantly larger working window for resection.⁴⁴ In addition, EETS can allow access to tumors located in the retroinfundibular space via transposition of the pituitary gland.^17,26

Despite these advantages, several limitations do exist to endoscopic endonasal approaches. Tumors extending lateral and/or encasing structures such as the supraclinoid carotid artery may be difficult to resect via EETS, and may be more amenable to a transcranial approach.^8,10 Moreover, a steep learning curve exists for the performance of these approaches, and often requires a neurosurgeon and otolaryngologist trained in skull base approaches.³⁹

At our institution, although the total number of surgeries for midline suprasellar tumors has remained fairly constant, there has been a steady increase in the use of EETS since 2007, with a concomitant decrease in craniotomies. A similar trend has been noted in many other centers as well. We suspect that this trend will continue and that it reflects a growing consensus that EETS has several advantages over craniotomy for these growths.

Conclusions

Both EETS and craniotomy can successfully remove craniopharyngiomas and similar midline suprasellar masses. The EOR, recurrence rate, and PFS appear to be similar when either EETS or craniotomy was used at our center over a 13-year period. Whereas the rate of CSF leakage was found to be significantly higher in the EETS cohort, the rate of cranial nerve injury and other neurovascular complications was found to be significantly higher in the craniotomy cohort. In light of the long-term consequences of neurological injury and the relative ease with which CSF leaks can be repaired, we believe that our data support the conclusion that EETS surgery is the preferred approach for removal of suprasellar craniopharyngiomas and similar lesions whenever feasible, and we advocate for increased use of the EETS approach.

Acknowledgment

We thank Serguei Bannykh, MD, PhD, Department of Pathology, Cedars-Sinai Medical Center.

Author Contributions

Conception and design: Mamelak, Wu, Bonert, Carmichael. Acquisition of data: Mamelak, Jeswani, Wu, Carmichael, Chu, King. Analysis and interpretation of data: Mamelak, Jeswani, Nuño, Bonert, Black, Chu, King. Drafting the article: Mamelak, Jeswani, Bonert. Critically revising the article: Mamelak, Nuño, Black. Reviewed submitted version of manuscript: Mamelak, Jeswani, Nuño, Wu, Carmichael. Statistical analysis: Mamelak, Jeswani, Nuño. Administrative/technical/material support: Carmichael. Study supervision: Mamelak, Black.