Randomized controlled trials (RCT), prospective cohort studies, and case series that compared EEA with TCA for resection of craniopharyngiomas in patients above 18 years of age were included. Inclusion of case series ensured that the review thoroughly consolidated available evidence. Articles where data for craniopharyngiomas in adult populations could not be extracted or calculated, either due to mixed cohorts or unavailability of individual data, were excluded. No restrictions regarding year of publication or language of the article were applied.

On July 9, 2021, the databases of PubMed, Web of Science, and Cochrane Library were systematically retrieved. The following keywords were used: ‘endonasal,’ ‘transsphenoidal,’‘transplanum,’ ‘trans sellar,’ ‘endoscopic,’ ‘endoscopic,’ ‘endoscopy,’‘endoscopy,’ ‘EEA,’ ‘transcranial,’ ‘pterional,’ ‘bitozygomatic,’‘frontolateral,’ ‘bifrontal,’ ‘frontobasal,’ ‘interhemispheric,’‘transpetrosal,’ ‘subtemporal,’ ‘transcallosal,’ ‘transcortical,’ ‘trans lamina terminalis,’ ‘TCA,’ ‘craniotomy,’ ‘microscopic,’ ‘microsurgical,’‘microsurgery,’ ‘microsurgery,’ ‘minimally invasive,’‘craniopharyngioma,’ ‘craniopharyngeal neoplasm,’ ‘craniopharyngeal neoplasm,’ ‘craniopharyngeal tumor’ (Supplementary Material in the online-only Data Supplement). The references of the included studies were also checked to find possible relevant articles. The search included registries of registered trials. The titles and abstracts of the articles were reviewed independently by two reviewers and the full texts of the articles that either reviewer found relevant were acquired with the input of a library consultant. Full texts were then assessed independently by two reviewers for relevancy and disagreements were settled by discussion and by consultation with a third reviewer.

Data were extracted independently by two reviewers. A third reviewer was consulted in cases of disagreements. Data on the following variables were extracted: study population (number, age, and gender), tumor characteristics (volume, location, extension, consistency and calcification), and patient outcomes (postoperative complications, recurrence, 30-day mortality, and length of stay). Primary outcomes were 30-day mortality following surgery and postoperative CSF leak, stroke (ischemic), hypopituitarism, recurrence on subsequent follow-up, visual impairment, and hydrocephalus. Secondary outcomes identified were postoperative diabetes insipidus, length of stay, grand total resection, and intracranial infection.

The methodological quality of the included RCTs was evaluated independently by two reviewers based on the Cochrane Handbook for Systematic Reviews of Interventions, version 6.0 [10]. The following 5 item scales were assessed: 1) random sequence generation—selection bias; 2) allocation concealment—selection bias; 3) blinding of the participants and personnel— performance bias; 4) blinding of outcome assessments—detection bias; and 5) selective reporting—reporting bias. Each of the items needed to be measured as "Yes" (low risk of bias), "No" (high risk of bias), or "Unclear" (unclear risk of bias). The risk of bias summary figure was obtained using the Robvis visualization tool [11]. The methodological quality of the cohort studies was assessed using the Newcastle Ottawa Scale [12]. The scale uses a star system to assess quality of studies based on selection, comparability of groups, and ascertainment of outcome. Bias was assessed at the study level. The methodological quality of the case series was evaluated using the tool for evaluating the methodological quality of case series and case reports. It consists of eight items, categorized into four domains: selection, ascertainment, causality, and reporting. Five items pertinent to this review were selected by the authors. Each item was given 1 point and measured as “Yes” (low risk of bias) or “No” (high risk of bias). The total score was then calculated by adding the score allotted to each item. Disagreements between the two reviewers in judgment of the quality of the study were settled by discussion and consultation with a third reviewer. The conflict of interest and source of funding reported by all the studies included was also recorded.

In reporting of categorical outcomes, risk ratios (RRs) with 95% confidence intervals (CIs) were used. Continuous data was analyzed as standard mean differences with similar 95% CIs. Pooled participant data from the identified cohort studies were included in a random-effects model (Mantel-Haenszel method) meta-analysis using RevMan Version 5.4 and displayed in Forest plots [13]. Heterogeneity of subgroups was calculated using I² statistic with an I² >40% assumed to represent significant heterogeneity, as recommended in the Cochrane Handbook for Systematic Reviews of Interventions, version 6.1 [10].

A total of 344 articles were identified from literature databases and through reference and citation searches. After exclusion of duplicates, there were 227 remaining, which were assessed for relevance by reviewing the title and abstract. As a result, 26 articles were found relevant, and, despite the absence of language restriction in the search and selection strategy, the full text was available all in English. After the exclusion criteria were applied, a total of eight articles were included in the final systematic review (Fig. 1) [214151617181920].

A total of 11,395 subjects were compared for EEA and TCA in the eight included studies [214151617181920]. Age and sex ratio (male:female) were reported in seven studies [2141617181920]. The mean age ranged from 40.3±15.5 to 52.4±17.2 years and the male:female ratio ranged from 0.31 to 3 in EEA and 0.55 to 2.66 in TCA (Table 1). Four studies used an extended endoscopic approach (Supplementary Table 1 in the online-only Data Supplement).

Six studies were reported as good quality [21417181920] and two were reported fair quality [1516] according to the Newcastle Ottawa Scale Criteria (Table 2). All studies received the maximum in the ‘selection’ and ‘ascertainment of outcome’ criteria. There were concerns present in the comparability of cohorts as data was collected using a retrospective design and cohorts were not matched for characteristics of the craniopharyngioma studies and were unable to control for tumor characteristics and age and sex of patients [14151617].

Data collected from cohort studies showed a significantly higher 30-day mortality rate in the TCA group, with low heterogeneity from our analysis. The lower rates of mortality may reflect recent improvements in surgical technique and perioperative management. Tumor recurrence was higher in the transcranial group; data were however heterogeneous and non-significant, warranting further prospective trials for evaluation.

Endoscopic approaches may result in significantly shorter hospital stays for patients in comparison to TCA. This can be affected by postoperative complication rates that have been previously identified as independent predictors of hospital length of stay [21]. Our analysis showed significantly higher rates of postoperative hydrocephalus, stroke, hypopituitarism, and visual impairment in the TCA group. Rates of postoperative diabetes insipidus and intracranial infection were non-significant in either cohort, and we did not find sufficient data in the included studies to warrant comparison regarding cardiovascular, respiratory, or wound complications.

A higher risk of CSF leak was seen after EEA surgery; this is expected as skull base approaches can increase the risk of CSF leak [22]. This has been greatly reduced with the use of vascularized nasoseptal flaps, as has been studied in recent literature [35]. Advanced endoscopic skull base reconstruction techniques show promising results for improvement of CSF leak after surgery, and should be further evaluated.

GTR was evaluated in five studies—although data were homogenous, the results were non-significant. Therefore, surgeons can be more confident in endoscopic resection providing the same extent of resection as TCAs.

A meta-analysis previously conducted in 2018 took into account all adult craniopharyngioma cohorts and compared outcomes between EEA and TCA groups [23]. A key limiting factor was that only one study was included that directly compared both approaches. Pooled analysis showed higher rates of visual and endocrine symptoms in the EEA group, and no differences in hypopituitarism or diabetes insipidus. Recurrence was also similar in both groups. However, there was significant heterogeneity reported. Our study builds on these findings by providing stronger evidence through comparable cohort studies and homogenous data. Previous systematic reviews have recommended endonasal approaches to be indicated in midline adult-onset craniopharyngiomas, showing improved GTR and outcomes, both endocrinological and visual [24]. They also showed that the rate of CSF leak was strongly diminished with the use of multilayer reconstruction techniques. TCAs were advocated for use in laterally extending or purely intraventricular tumors. However, recent literature has shown endoscopic surgery for craniopharyngiomas with intraventricular extension shows similar outcomes to transcranial surgery [2526]. Similarly, CSF leak rates were brought down with the use of nasoseptal flaps in this approach.

There is a discrepancy in study samples sizes, with one study accounting for a majority of the included patient population for certain outcomes [17]. This is to be expected as this study was done through retrospective national database assessment. However, the authors only focused on specific clinical outcomes, and therefore other studies played a larger role in our analysis of surgical outcomes (GTR, recurrence). Secondly, a majority of studies were retrospective chart reviews, with few studies blinding for outcome assessment where possible (radiological tumor outcomes for instance). Lastly, patient cohorts were not matched to account for confounding factors. The size of tumor and its anatomical location can affect the outcomes of a surgical approach and are important scientific questions.

Our study is the first meta-analysis on adult craniopharyngiomas to pool together studies where cohorts were compared. We included standardized outcomes and showed statistical significance within our data set.

This study demonstrates lower complication rates, shorter hospital stays, and lower mortality associated with endoscopic adult craniopharyngioma surgery in comparison to open approaches. EEA may however be correlated with higher rates of CSF leak, warranting further subset analysis of the effect seen by improved skull base reconstruction and vascularized nasoseptal flaps techniques. These findings do not account for difference in tumor size and extension between the EEA and TCA cohort. Further research on patients with comparable tumor characteristics is required to fully assess outcomes, particularly for rates of tumor recurrence and achieving GTR.