The endoscope-assisted supraorbital “keyhole” approach for anterior skull base meningiomas: an updated meta-analysis

This review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement [81]. A search strategy was created using the keywords “Meningioma,” “Tuberculum Sellae,” “Olfactory Groove,” and synonyms (Appendix A). Studies were included if they reported on (1) patients with olfactory groove (OGM) or tuberculum sellae (TSM) meningiomas; (2) patients undergoing surgery using the mTCA, EEA or eSKA approaches; and (3) surgical outcomes and complications. Exclusion criteria included case reports, commentaries, congress abstracts, reviews, animal studies, studies describing a combined surgical approach (for example EEA + mTCA), studies in paediatric patients (< 18 years old), re-operations, and cadaveric studies. A date filter was applied, with articles from 2004 to 2020 being included—reflecting a period of the contemporary adaptation of endoscopic and keyhole approaches and the continuous improvement of traditional microsurgical approaches for the relevant pathologies [11, 15, 32, 106].

Both PubMed and Embase databases were searched on 19 April 2020. Duplicates were removed using Endnote X9. Independent title and abstract screening of updated results was performed in duplicate by two authors (DZK, HJM). Review of full-text articles ensued according to the inclusion/exclusion criteria. Any discrepancies in selection were settled out by discussion and mutual agreement.

Data points extracted from the included articles comprised of patient characteristics (age, sex distribution), tumour characteristics (surgical approach used, sum of sample, tumour grade, tumour diameter or volume, follow-up length), outcomes (GTR, visual improvement), and complications (CSF leak, 30-day mortality, intra-operative arterial injury). World Health Organization (WHO) grading included recording the proportion of WHO Grade 1 tumours [72]. Of note, the grading system was revised in 2016 to include brain invasion as a criterion to upgrade Grade 1 tumours to Grade 2 [72]. Therefore, if any studies pre-2016 reported brain invasion amongst Grade 1 tumours, the respective tumours were upgraded accordingly [72]. Gross total resection (GTR) referred to Simpson Grades 1 and 2 as per our original methodology [84, 116]. Visual improvement was in the context of those with preoperative visual problems only. Mortality (within 30 days after resection) was recorded on an all-cause basis.

Owing to the not uncommon reporting of follow-up time as a median number of months (as opposed to mean), the estimated mean number of months was calculated as per recommendations of Hozo et al. [53]. Of note, in sample sizes greater than 25, the sample’s median follow-up is presented as the best estimate of the mean [53].

Importantly, studies that did not report specific outcomes for OGM/TSM and approach combination were excluded from the final meta-analysis. These studies were considered for qualitative analysis if the relevant the tumour (TSM or OGM) and approach (mTCA, EEA, or eSKA) combination of interest was > 50% of the study population [9, 48, 98, 105, 111]. Similarly, articles that grouped TSM cases with planum sphenoidale meningiomas [3, 92] were considered for qualitative review only (owing to the similarity of these tumour groups) but not included in the final meta-analysis.

Study quality was assessed with a modified New-Castle Ottawa Scale (mNOS), which assesses two domains: sample selection and outcome reporting. The modification made to the original NOS was the exclusion of the “comparability” domain as this is not applicable to case series [126]. The scale is scored out of 6 (3 for selection domain, 3 for outcome domain). Publication bias was assessed using Begg’s tests [8] and by generating funnel plots with and without trim-and-fill method [34].

A meta-analysis was conducted using R 3.6.1 (The R Foundation, Austria) applying the “meta” package. Pooled incidence (using the random-effect model method of DerSimonian and Laird [33]) was calculated for each approach (eSKA, mTCA, EEA), tumour (TSM, OGM), and outcome (GTR, arterial injury, visual improvement, CSF leakage, and mortality) combination. Study heterogeneity was assessed by calculating I-squared values [52] (I² > 50% considered significant) and Cochran’s Q test (p < 0.10) [36, 52]. Sensitivity analysis was performed by running the above analyses on a low risk of bias sub-group (mNOS score greater than or equal to 4).

Additionally, a univariate meta-regression was performed to explore the effect of mean age (continuous variable) and male percentage (continuous variable) on each approach, tumour, and outcome combination. Meta-regression was only performed if 8 or more studies were available for the outcome/approach/tumour combination being explored. This threshold was chosen (a deviation from the standard threshold of 10) on a pragmatic basis, to reflect the relative paucity of literature from the newer eSKA approach [51]. This threshold was also applied to the performance of Begg’s test, trim-and-fill analysis, and the generation of funnel plots.

In all, after removing duplicates, 2285 articles were identified (Fig. 1). After screening for titles and abstracts, 2044 articles were excluded and 241 full texts were reviewed to yield 96 included studies. Fifty-three TSM-only case series were included in the meta-analysis of which 21 involved the EEA [3, 12, 13, 17, 19, 22, 37, 41, 49, 60, 61, 63, 67, 74, 91, 92, 120, 125, 131], 37 the mTCA [1, 3, 6, 13, 20, 21, 26, 29, 38, 43, 45, 55, 60, 63, 64, 66, 68, 74‐76, 78, 79, 82, 85, 88, 92, 95, 97, 98, 108, 112, 114, 120, 124, 127, 130], and 5 the eSKA [16, 35, 41, 67, 78] with 10 of these papers covering multiple approaches [3, 14, 41, 60, 63, 67, 74, 78, 92, 120]. Twenty-eight OGM-only case series were included in the meta-analysis of which 5 involved EEA [4, 28, 62, 70, 92], 24 in mTCA [5, 7, 10, 23, 25, 27, 28, 40, 42, 44, 47, 56, 57, 70, 83, 87, 89, 96, 99, 101, 109, 117, 121, 123], 3 in eSKA [4, 39, 92], and with 4 of these studies detailing multiple approaches [4, 28, 70, 92]. Additionally, 15 studies explored both OGM and TSM [9, 30, 31, 50, 54, 58, 65, 93, 94, 103, 105, 113, 122, 128, 129]. Resultantly, the TSM group totalled 2191 patients and OSM group totalled 1519 patients (Tables 1 and 2).

The median number of patients per study was 20 (range: 3–95) for TSM and 19.5 (range: 2–129) for OGM. The average percentage of male patients was 24% for TSM and 31% for OGM. The median mean of age was 54.2 years for TSM and 54.75 years for OGM. The median mean of follow-up time for TSM was 32 months (reported in 55/67 studies) for and 44.5 months for OGM studies (reported in 39/43 studies). The modified NOS score varied between 2/6 and 6/6 amongst the TSM and OGM case series, with predominant factors affecting this variance being a description of follow-up and outcome reporting. Summary characteristics by approach (eSKA, EEA, or mTCA) are highlighted in Tables 1 and 2. Individual study characteristics are displayed in Tables 5 and 6 (Appendix B).

The pooled incidence of surgical outcomes of a subgroup of low-risk studies (defined as mNOS score greater than or equal to 4) is presented in Appendix E. This analysis, when compared with the total group analysis, yielded overall similar results for GTR, mortality, and intraoperative arterial injury. CSF leak incidence after EEA was apparently lower (in both OGM and TSM), and visual improvements after EEA (in the TSM group) were more marked in the lower risk of bias studies.

In the context of heterogeneous study populations and outcome reporting, the endoscope-assisted supraorbital “keyhole” approach appeared to be associated with similar effectiveness (GTR, visual improvement) and safety (CSF leak, 30-day mortality) compared with the mTCA and EEA alternatives based on our findings. Case selection and an understanding of relative indications are important in selecting the most appropriate approach for anterior skull base meningioma resection.

As previously found, the EEA was associated with the highest rates of visual improvement across OGM and TSM groups. However, this advantage of EEA may be offset when considering the safety profile of the three approaches, with the EEA having the highest incidence of post-op CSF leak (statistically significant in the TSM sub-group). In contrast, the mTCA had a slightly higher incidence of GTR than eSKA and EEA (eSKA > EEA) across TSM and OGM groups. Interestingly, the eSKA displays intermediate results in terms of efficacy (GTR and visual improvement) and complications (CSF leak). Results for intra-operative arterial injury and 30-day mortality incidences are similar and overlapping across the 3 approaches. Indeed, the eSKA, as a relatively new technique, is less well explored. When compared with the mTCA, the eSKA—as a minimally invasive technique—offers a smaller craniotomy scar, less brain exposure, and less brain/nerve retraction [105]. Thus, theoretically, it shares similar limitations to the minimally invasive EEA [92]—potentially making total resection of larger tumours or tumours with significant local invasion difficult [24, 92, 105]. However, when performed with the benefit of neuronavigation, neuroendoscopy (12/13 of eSKA studies in our meta-analysis), and appropriate surgical training, the eSKA has been used to resect relatively large tumours of the anterior skull base [4, 41, 67, 105].

All 3 approaches likely have their own role in the management of anterior skull base meningiomas, with their varying safety and efficacy profiles as evidenced above. Case selection will be paramount in establishing a role for each technique/combination of techniques [4, 92, 105]. Indeed, case selection of eSKA is currently considerably variable, owing to its novelty and ongoing refinement [70, 92]. The selection of the preferred approach for each case must be taken in the context: (a) patient-related factors (demographic, presentation, preferences), (b) tumour-related factors (size, consistency, extension, location—such as relation to optic foramen or cribriform plate), and (c) surgeon experience, surgeon preference, and surgical goals (such as GTR or STR, visual or olfactory preservation) [2, 77, 92, 105, 118]. Ottenhausen et al. presents a concise decision-making algorithm (based on tumour anatomy and resulting functional deficits), which incorporates the specific characteristics of eSKA, EEA, and mTCA approaches [92]. In this algorithm, the eSKA is suitable for TSMs with lateral extension up to the internal carotid arteries (ICA) and anterior clinoid processes (ACP), or lateral extension beyond the lamina papyracea (LP). Additionally, the eSKA is suggested for OGMs with (1) preserved olfaction and (2) disrupted olfaction without cribriform plate invasion but with significant anterior (up to the frontal sinus) or lateral extension. In contrast, the EEA is proposed for TSM without significant lateral extension (ICA/ACP/LP as above) and OGMs without significant lateral extension (where olfaction is disrupted). Finally, an mTCA or a combined EEA + eSKA approach is suggested for OGMs and TSMs with a significant anterior or lateral extension (unless there is no cribriform plate invasion, in which case, eSKA alone may be possible) [92]. Of note, other algorithms cite > 5mm sellar extension and optic canal involvement as factors favouring EEA in TSM [60]. During EEA for TSM, decompression of the optic canal from below avoids excessive vascular manipulation, can be achieved before or after tumour resection, and is well suited to tumours with extension into the inferomedial aspect of the optic canal [2, 69]. Decompression of the involved optic canal is described in mTCA approaches with early decompression (before tumour resection) favoured [20, 80, 90]. In eSKA, studies describe both early and late bilateral canal decompression with optimum timing being less clear [16, 67]. More generally, within the literature, consensus for the ideal surgical approach in various contexts is not clear [41, 60, 67, 74, 92, 99]. Indeed, in light of the COVID-19 pandemic—which elucidated to the risk endonasal surgery may pose (exposing theatre staff to high viral loads and potentially serious infection)—this case selection process is likely to be a dynamic field in the near future [71, 100, 119].

Previous meta-analyses have compared the EEA and mTCA (not eSKA) with varying results.

In terms of GTR, Muskens et al. (co-author) previously found higher GTR incidence with mTCA in OGM at 88.5% (CI 85.9–90.7%) versus EEA 70.9% (CI 60.3–79.9%) [84]—in line with our findings. This corroborates with other meta-analyses. Ruggeri et al. explored OGM and TSM, finding a higher GTR rate (p < 0.01) in mTCA (88.13%) than EEA (78,42%) [110]. Similarly, Komotor et al. highlighted a 92.8% GTR rate in mTCA, compared with 63.2% in EEA (0.001) in the context of TSM and OGM [59], whilst Shetty et al. explored GTR in OGM, finding a significantly (p < 0.01) higher rate in mTCA (90.9%) than in EEA (70.2%) [115].

Regarding visual outcomes, a recent comparative meta-analysis by Lu et al. suggests improved visual outcome in OGM resection using the EEA (vs. mTCA) (OR, 0.318; p = 0.04) but not statistically significant in TSM [73]. This is slightly different from our updated findings and previous findings of Muskens et al. [84], in which the visual outcome advantage of EEA was most prominent in the TSM group. In other analyses, an early (2013) study by Clark et al. displayed higher (p < 0.01) visual improvement incidence in TSM with EEA (50–100 % in included studies) compared with mTCA studies (25–78 %) [24]. Shetty et al. explored OGM alone and found 80.7% visual improvement in the EEA studies group versus 12.83% in the mTCA group (p < 0.01) [115]. Ruggeri et al. replicated these findings when taking OGM and TSM as a collective group, with EEA displaying an 80.1% incidence visual improvement, significantly (p < 0.01) higher than mTCA (62.2%) [110].

In terms of CSF leak rate, Muskens et al. highlighted this as a disadvantage to the EEA in both TSM (EEA: 19.3% (95% CI 14.1–25.8%), mTCA 5.8% (95% CI 4.3– 7.8%)) and OGM (EEA: 25.1% (95% CI 17.5–34.8%), mTCA 10.5% (95% CI 8.2–13.4%)) [84]. This finding is echoed throughout relevant secondary literature, with Lu et al. highlighting a higher CSF leak incidence in EEA (vs mTCA) in TSM (OR 3.854; p = 0.013) and Shetty et al. showcasing a 25.7% CSF leak occurrence in EEA versus 6.3% in mTCA (p < 0.01) [73, 115]. In taking TSM and OGM, together, Komotor et al. demonstrated a higher CSF leak incidence of 21.3% in EEA, compared with 4.3% in mTCA (p < 0.01), whilst Ruggeri et al. illustrated 18.84% CSF leak occurrence in EEA versus 5.95% in mTCA (p < 0.01) [110, 115].

Finally, when considering 30-day mortality, significant associations have been difficult to establish both in our study and the literature. Ruggeri et al. found mortality rates of 2.3% in mTCA and 1.03% in EEA in TSM and OGM, but this did not reach statistical significance (p = 0.154) [110]. Similarly, differences in mortality explored by Muskens et al. were inconclusive in TSM (EEA: 5.2% (95% CI 3.4–10.8%), mTCA 2.7% (95% CI 1.8–4%)) and OGM (EEA: 4.3% (95% CI 1.5–11.6%), mTCA 3.9% (95% CI 2.7–5.8%)) [84]. A similar situation is found with intra-op arterial injury incidence with most syntheses not highlighting significant differences, echoed by our updated analysis [24, 59, 110, 115].

The principal limitations of this study are the likely prevalent publication bias and heterogeneity of the primary literature synthesized, more specifically heterogeneity in the reporting of baseline characteristics and outcomes. This is reflected in the I² and Cochran Q tests highlight in Figs. 2 and 3, corroborating with respective funnel charts (Appendix B). Development of core data set, through a multi-stakeholder consensus process for example, would be useful for future pooled analysis in the field. Secondly, it is likely the study populations examined are considerably variable owing to the surgical decision-making process that informs the choice of approach. Larger, more extensive tumours may be more likely to undergo traditional open approaches (of which there are many variants) in order to achieve acceptable tumour resection [84]. This is reflected in Tables 1 and 2 where larger (diameter and/or volume) tumours are included in the mTCA group compared with the EEA group. Unfortunately, we were not able to perform meta-regression based on tumour size or grade owing to heterogeneous data reporting, potentially adding to confounding factors [18, 46]. Additionally, the overwhelming majority of studies included were case series, and thus, our interpretation of our results should be tempered to reflect this. Finally, owing to the novelty of the approach, there is a relative paucity in the amount of included eSKA studies. Although overall, the results of the main analysis are similar to that of the sensitivity analysis subgroup, the number of low risk of bias studies analyzed is also limited. Future studies in the field must improve on methodological design, with an emphasis on comparative studies, in order to facilitate more robust data synthesis.