The endoscopic endonasal approach is not superior to the microscopic transcranial approach for anterior skull base meningiomas—a meta-analysis

The mainstay of treatment for tuberculum sellae meningiomas (TSMs) and olfactory groove meningiomas (OGMs) is surgery. Goals of surgery include obtaining tissue for histopathological diagnosis and relieving pressure caused by the tumor on neighboring structures such as the olfactory nerves, anterior cerebral arteries, optic nerves, and pituitary gland. At the same time, these structures are very susceptible to manipulation, and damage to these structures can lead to great morbidity [51].

Traditionally, TSMs and OGMs are resected using a microscopic transcranial approach (mTCA). Various approaches have been described, including interhemispheric, pterional, bifrontal, and subfrontal mTCA [1, 2, 5‐7, 9, 47, 51, 56, 64, 70]. In the last decade, however, as a result of the evolution of endoscopic surgery for pituitary adenomas, these meningiomas have been increasingly resected using an endonasal endoscopic transsphenoidal approach (eTSA), as first described by Jho et al. in 2004 [38]. Although the endoscopic approach is generally viewed as less invasive, with some studies suggesting that eTSA caused fewer postoperative changes on magnetic resonance imaging (MRI) compared to mTCA possibly indicating less manipulation [22], it has been suggested that eTSA results in higher rates of CSF leaks and potentially different outcomes (e.g., less GTR) [18, 42]. However, a direct comparison between eTSA and mTCA is currently lacking. Therefore, the aim of this systematic review and meta-analysis was to evaluate which approach (eTSA vs. mTCA) offers the best surgical outcomes.

After removing duplicates, 1684 articles were identified. After screening for titles and abstracts, 1426 articles were excluded and 216 full texts were reviewed (Fig. 1). For TSM, 44 case series (of which 11 were in eTSA, 29 in mTCA, and 4 in both) were included in the meta-analysis for the different outcomes, including a total of 1444 patients [3, 5, 8, 11‐13, 15, 16, 20, 21, 23, 25, 29, 30, 32, 34‐36, 40, 41, 43, 45, 47‐53, 56, 58, 61‐63, 65, 66, 68, 69, 72, 73, 77, 79, 81, 82]. As for OGM, 25 case series (of which 6 were in eTSA, 18 in mTCA, and 1 in both) were included describing outcomes in 891 patients [2, 4, 6, 7, 17, 19, 22, 24, 25, 35, 37, 40, 44, 47, 55, 57, 60, 62, 64, 67, 68, 70, 75, 76, 78].

The median number of patients per study was 24 for TSM (Table 1) and 29 for OGM (Table 2). The average percentage of male patients was 27% for TSM and 32% for OGM. The median age was 51.0 for TSM and 52.0 for OGM. The median follow-up time was 6.0 years based on 35 studies for TSM [3, 5, 8, 12, 13, 15, 16, 21, 25, 29, 30, 32, 34, 36, 43, 45, 47‐53, 56, 61, 62, 65, 66, 68, 72, 73, 77, 79, 81, 82] and 7.0 years based on 20 studies for OGM [2, 4, 6, 7, 17, 19, 22, 24, 25, 37, 44, 47, 55, 57, 60, 62, 67, 68, 76, 78]. The modified NOS score varied between three and four of seven among the TSM and OGM case series [3, 5, 8, 11‐13, 15, 16, 20, 21, 23, 25, 29, 30, 32, 34‐36, 40, 41, 43, 45, 47‐53, 56, 58, 61‐63, 65, 66, 68, 69, 72, 73, 77, 79, 81, 82]. Outcomes of the meta-analysis for TSM (Table 3) and OGM (Table 4) are shown.

For TSM, GTR after eTSA was reported in 14 studies [8, 11‐13, 16, 20, 23, 29, 30, 40, 43, 61, 62, 79] and after mTCA was reported in 31 studies [3, 5, 11, 13, 15, 21, 23, 25, 29, 32, 34, 36, 45, 47‐49, 51‐53, 56, 58, 63, 65, 66, 68, 69, 72, 77, 79, 81, 82]. In a fixed effect model, the overall incidence for GTR was not significantly different comparing eTSA (incidence = 83.0%; 95% CI = 76.7–88.0%, p-heterogeneity = 0.74, I² = 0%, 221 patients) to mTCA (incidence = 85.8% (95% CI = 83.6–87.9%, p-heterogeneity = 0.07, I²: 28.4%, 1223 patients) (p-interaction value = 0.34). In meta-regression, TSM studies with lower percentage of males had a higher rate of GTR (p = 0.03). Studies conducted in Europe and Africa had significantly higher rates of GTR than those in North America (p = 0.02). Begg’s test for publication bias was non-significant (p = 0.31) (Table 3).

For OGM, GTR was specifically addressed in 7 eTSA [4, 22, 24, 35, 40, 44, 62] studies and 18 mTCA studies [2, 6, 7, 17, 19, 22, 25, 37, 47, 55, 57, 60, 64, 67, 68, 70, 75, 76]. Unlike TSM, the overall fixed incidence of GTR was significantly lower in eTSA (incidence = 70.9%; 95% CI = 60.3–79.9%, p-heterogeneity = 0.45, I² = 0%, 86 patients) compared to mTCA (88.5%; 95% CI = 85.9–90.7%, p-heterogeneity = 0.06, I²:36.5%, 786 patients) (p-interaction < 0.01; Fig. 2). In meta-regression, only higher age was associated with lower GTR in resected OGM with the eTSA approach with borderline significance (p = 0.05). Begg’s test for publication bias was non-significant (p = 0.48) (Table 4).

Visual outcomes were reported in 12 studies for eTSA [8, 12, 16, 23, 29, 30, 35, 40, 43, 61, 62, 79] and 28 studies for mTCA [3, 5, 13, 15, 21, 23, 25, 29, 32, 34, 36, 47‐51, 56, 63, 65, 66, 68, 69, 72, 73, 77, 81, 82] with a total of 1139 patients presenting with visual problems [3, 5, 8, 12, 13, 15, 16, 21, 23, 25, 29, 30, 32, 34‐36, 40, 43, 47‐51, 53, 56, 61‐63, 65, 66, 68, 69, 72, 73, 77, 79, 81, 82]. Postoperative visual improvement was significantly higher for eTSA (incidence = 77.7%; 95% CI = 70.3–83.7%, p-heterogeneity = 0.37, I² = 7.90%, 167 patients) than mTCA (incidence = 60.7%; 95% CI = 57.3–64.0, p-heterogeneity < 0.01, I² = 77.4%, 1139 patients) in fixed-effect models (p-interaction < 0.01). Because age and male percentage were not provided for this subgroup of patients who presented with visual problems, only continent could be assessed as a source of heterogeneity, which was not a significant source of heterogeneity for TSM resection using eTSA or MTCA. Begg’s test for publication bias was non-significant (p = 0.14) (Table 3). One study specifically addressed visual improvement per approach in TSM resection, finding that eTSA was associated with more visual acuity improvement (≥5%; p-value: 0.01), but not with improvement of visual field deficits (p-value = 0.61) [41].

Visual improvement in OGM patients was described four eTSA studies [4, 40, 44, 62] and nine mTCA studies [6, 7, 47, 57, 60, 68, 70, 75, 78] with 224 patients presenting with visual symptoms. The resulting fixed overall improvement rate was 64.5% (95% CI: 37.9–84.4%, p-heterogeneity = 0.03; I² = 65.5%) for eTSA compared to 50.6% (95% CI = 42.9–58.4%, p-heterogeneity <0.01, I² = 68.6%) for mTCA; however, this difference was not significant (p-interaction value: 0.33). Continent was not identified as a significant source of heterogeneity for eTSA (p = 0.34) and mTCA (p = 0.57). Begg’s test for publication bias was non-significant (p = 0.25) (Table 4).

CSF leak occurrence after TSM resection was extracted from 15 eTSA studies [8, 11, 16, 20, 23, 29, 30, 35, 40, 43, 61, 62, 79, 81] and 24 mTCA studies. The overall incidence of postoperative CSF leakage was significantly higher in patients treated with the eTSA approach (incidence = 19.3%; 95% CI = 14.1–25.8%, p-heterogeneity = 0.50, I² = 0%, 225 patients) than with mTSA (incidence = 5.81%; 95% CI = 4.33–7.75%, p-heterogeneity = 0.93, I² = 0%, 879 patients) in fixed models (p-interaction value <0.01, Fig. 3a). Age, gender, and continent were not identified as sources of heterogeneity using meta-regression (all p-value > 0.05). Begg’s test revealed no significant publication bias (p = 0.98) (Table 3).

In OGM, 7 eTSA studies [4, 22, 24, 35, 40, 44, 62] and 17 mTCA studies [2, 6, 7, 17, 19, 22, 25, 37, 55, 57, 60, 64, 67, 68, 70, 75, 76, 78] including 889 patients described whether patients developed a CSF leak postoperatively. The overall incidence in fixed models was statistically significantly higher (p-interaction < 0.01) for eTSA (incidence = 25.1%; 95% CI = 17.5–34.8%, p-heterogeneity = 0.22, I² = 25.8%) than mTCA (incidence = 10.5%; 95% CI = 8.22–13.4%, p-heterogeneity <0.01, I² = 60.2%) (Fig. 3b). In meta-regression, only older age was significantly associated with a lower CSF leakage rate for mTCA (p < 0.01). For eTSA, age, gender, and continent were not identified as potential effect modifiers (p-interaction for all > 0.05). Begg’s test indicated no significant publication bias (p = 0.30) (Table 4).

For intraoperative arterial injury, outcomes were extracted from 12 eTSA studies [8, 11, 16, 23, 29, 30, 35, 40, 43, 61, 62, 79] and 27 mTCA studies for TSM [3, 5, 11, 12, 15, 21, 23, 25, 29, 32, 36, 45, 48, 49, 51, 52, 56, 58, 63, 65, 68, 69, 72, 77, 81, 82]. The overall incidence of intraoperative arterial injury was significantly higher for eTSA (incidence = 4.89%; 95% CI = 2.33–9.94%, p-heterogeneity = 0.97, I² = 0%, 225 patients) than for MTCA (incidence = 1.86%; 95% CI = 1.13–3.05%, p-heterogeneity = 0.99, I² = 0%, 225 patients) in fixed effect models (p-interaction value = 0.03; Fig. 4). Trial-level covariates such as age, continent, and gender did not significantly contribute to any heterogeneity in the models for both eTSA and mTCA (all p-interaction values > 0.05). There was a significant publication bias, indicating that study results with higher arterial injury incidence tended not to be published (Begg’s test p-value < 0.01) (Table 3). However, the imputed overall incidence estimate for TSM was not materially different from the original incidence rate (not shown).

For OGM, the incidence of intraoperative arterial injury was extracted from 858 patients in 7 eTSA studies [4, 22, 24, 35, 44, 62] and 17 mTCA studies [2, 6, 7, 17, 19, 22, 25, 37, 55, 57, 60, 64, 67, 68, 70, 75, 76, 78]. For eTSA, the fixed overall incidence of intraoperative arterial injury was 3.88% (95% CI = 1.55–9.43%, p-heterogeneity = 0.98, I² = 0%). Although lower, the incidence for mTCA was 1.62% (95% CI = 0.87–2.98%, p-heterogeneity = 0.99, I² = 0%) but not significantly different (p-interaction = 0.12). Covariates such as age, gender, and continent were not identified as sources of heterogeneity for both eTSa and mTCA procedures (all p-interaction > 0.05). Although Begg’s test for publication bias indicated the presence of publication bias (p-value < 0.01), Egger’s test did not (p-value = 0.50) (Table 4). Moreover, the imputed overall incidence estimates for OGM were not materially different from the original incidence values (not shown).

Mortality after TSM surgery was described in a total of 10 eTSA studies [8, 11, 23, 29, 40, 43, 61, 62, 79] and 30 mTCA studies [3, 5, 11‐13, 15, 21, 23, 25, 29, 30, 35, 36, 45, 48, 49, 51‐53, 56, 58, 63, 65, 68, 69, 72, 77, 81, 82]. eTSA resulted in a 30-day mortality incidence of 5.15% (95% CI = 2.39–10.8, p-heterogeneity = 0.85, I² = 0%, 194 patients), which was not significantly different from mTCA (incidence = 2.67%; 95% CI = 1.77–4.02, p-heterogeneity = 0.99, I² = 0%, 962 patients) in fixed models (p-interaction = 0.14). Age, gender, and continent did not appear to have different incidence values based on the meta-regression results for both eTSA and mTCA (all p > 0.05). Begg’s test p-value for publication bias was significant, indicating that articles with higher mortality rates tend not to be published (p < 0.01) (Table 3); however, the trim-and-fill method suggested that the imputed overall incidence estimates for TSM were not materially different from the original incidence values (not shown).

For OGM, 7 eTSA studies [4, 22, 24, 35, 40, 44, 62] and 19 mTCA studies [2, 6, 7, 17, 19, 22, 25, 37, 47, 55, 57, 60, 64, 67, 68, 70, 75, 76, 78] including described mortality incidence. For eTSA, the overall 30-day mortality incidence was 4.27% (95% CI = 1.50–11.6%, p-heterogeneity = 0.94; I² = 0%; 82 patients), which was not significantly different from the mortality incidence in the mTCA group (incidence = 3.92%, 95% CI = 2.66–5.75, p-heterogeneity = 0.74, I² = 0%; 779 patients) in fixed models (p-interaction = 0.88). In a meta-regression for gender, it was identified that studies with a lower male percentage were significantly associated with a higher mortality incidence for mTCA (p = 0.02) but not for eTSA (p = 0.34), while age and continent were not. Begg’s test for publication bias was non-significant (p = 0.21) (Table 4).

For all the above-mentioned results, the random-effect models yielded similar results (Tables 3 and 4).

For blood loss, operating time, and length of hospital stay, a quantitative meta-analysis was not feasible because of the paucity of studies reporting them; hence, these few studies were systematically reviewed. In TSM, mean blood loss ranged from 448 to 970 ml in three studies describing mTCA compared to 200 to 617 ml for eTSA [21, 30, 41, 47]. The mean operating time ranged from 375 to 444 min for eTSA in two studies and from 116 to 426 min for mTCA in four studies [21, 23, 41, 47, 69]. Hospital length of stay ranged from 6 to 21 days in one study in patients treated by an eTSA [23].

For OGM, blood loss was only reported in one case series in patients operated with an interhemispheric approach (mean: 570.9 ml, SD: 442) [47]. The mean hospital length of stay for eTSA ranged from 11 to 13.5 days in two studies [9, 13] compared to 8.5 to 18 days for mTCA [7, 22, 24, 78]. Of these studies, one described the mean length of stay in both approaches, with a mean length of stay of 11 days for eTSA compared to 8.5 days in mTCA (p = 0.54) [22]. Operating time ranged from 6 to 10 h in one study reporting outcomes from eTSA [24]. In a study examining patients with an interhemispheric approach, the mean operating time was 209 min (standard deviation: 103) [47].

In this meta-analysis, eTSA was not shown to be superior to mTCA for resection of both OGMs and TSMs. Only in patients with preoperative visual deficits due to TSM, eTSA seems superior to mTCA, but with great hetereogeneity. In patients with TSM, eTSA resulted in higher rates of visual improvement, similar rates of GTR, and more CSF leaks and intraoperative arterial injury, while in patients with OGM, results of both techniques were similar for visual improvement and intraoperative arterial injury, but worse in patients operated with eTSA for GTR and CSF leaks. There seems to be no substantial difference in perioperative blood loss, operating time, or length of hospital stay between the two approaches. There was no substantial difference between incidence rates in the fixed- and random-effect models. This could be explained by a relative lack of difference between the study populations in the studies, which could have been implicated in the case of a difference between the models. However, mTCA was associated with considerable heterogeneity for outcomes visual improvement in TSMs and CSF leak for OGMs, which could reflect a relatively greater inter-study variability for these outcomes.

Although no significant difference was identified in GTR rate for TSM, mTCA resulted in higher GTR rates in OGM. As OGMs are located more anteriorly than TSM, an extended eTSA approach is needed for OGM, which requires more extensive drilling of the anterior skull base and a potential suboptimal view because of the angle of the scope. However, it should also be noted that GTR was not always the primary the goal of surgery (e.g., the goal could be preserving vision) [43, 72]. Furthermore, many other factors seem to influence GTR rate. One factor may be the learning curve associated with eTSA, as seen with pituitary adenoma resection [10, 14, 46]. Also, tumor factors such as large size and vascular enhancement can significantly lower the GTR rate for eTSA, as seen in one study in TSM [43]. Furthermore, presence of a “cortical cuff” (a layer of brain between the tumor capsule and cerebral vessels) on MRI was associated with more GTR in OGM [40].

For visual improvement, it remains to be determined whether eTSA is truly associated with more visual improvement than mTCA in TSM as correction for the heterogeneity among mTCA studies could not be done. Therefore, the difference witnessed may be insignificant as seen with OGM. Furthermore, as the variance in reporting of tumor size did not allow for it to be incorporated in a meta-regression, the TSMs in the eTSA group may be smaller compared to the mTCA group. However, regarding visual outcomes, one study looking at the mTCA approach suggests that visual outcomes are associated with age and duration of visual symptoms but not with actual tumor size [28].

For both OGM and TSM, eTSA was significantly associated with more CSF leakage. However, prophylactic lumbar drain placement varied greatly; in some studies, almost all patients were given a prophylactic preoperative lumbar drain, while in other studies none of the included patients were drained [8, 24, 30, 35, 40, 44]. Also, the different studies used different reconstruction techniques (e.g., introduction of a vascularized flap and use of certain glues), although this caused no considerable heterogeneity among the studies [40, 44, 62]. Another factor in the postoperative CSF leakage rate may be the neurosurgeon’s level experience. Although the difference was not significant, in a small number of patients, one group had two leaks in their first group of patients (n = 8) compared to none in the latter group (n = 12) [40]. Also, use of a vascularized flap for reconstruction of the skull base seems to bring the CSF leakage rate down considerably [40, 43, 62]. Still, this rate is considerably higher than the overall incidence calculated for mTCA. Further improvement with more sophisticated reconstruction techniques following eTSA may bring the rate of CSF leakage down to those reported for mTCA.

Previously, two reviews have described a comparison between eTSA and mTCA for both TSM and OGM. The first review identified a higher GTR rate and less CSF leakage associated with mTCA for both OGM and TSM (p < 0.01 for both, using the chi-squared test and Fisher’s exact test, respectively), which is similar to our findings except for the GTR rate for TSM [42]. A second review found significantly more visual improvement (p < 0.01) and CSF leakage (p < 0.01) for eTSA and no difference in mortality (p = 0.15) for TSM and OGM together, similar to our findings. eTSA was also found to be associated with a lower GTR rate (p < 0.01) compared to mTCA, which was only the case in OGM in this meta-analysis [71]. Finally, the authors of a meta-analysis for TSM found that eTSA was significantly associated with CSF leakage (OR: 3.9; 95% CI: 1.15–15.75, p < 0.05) and visual improvement (OR 1.5; 95% CI 1.18, 1.82, p < 0.05), which again is similar to our results [18].

Strengths of this study include an extensive review of the literature and evaluation of outcomes such as arterial injury, length of hospital stay, and blood loss. The use of both fixed- and random-effect models, evaluation of heterogeneity between the included studies, and assessment of publication bias ensures a rigorous evaluation of outcomes with appropriate valuation of the results. All outcomes were also subjected to meta-regression for various study characteristics where possible to try to identify sources of heterogeneity between the studies.

There are several limitations of this meta-analysis. First, the decision to discard studies published before 2004 produces a limitation. The decision to do so was based on the assumption that also mTCA outcomes improve over time with continual innovation and that meningiomas were not reported to be resected with an eTSA before that time [26, 38]. Regarding the included studies, only case series were identified, resulting in the inability of calculation overall odds ratios. There is probably also a great difference between the population of patients who were deemed eligible for an eTSA resection compared to those resected with mTCA because of the size, extension, and invasion of the tumors (confounding by indication). Furthermore, one could argue that only looking at perioperative outcomes may not be conclusive as especially recurrence happens during follow-up. However, as the GTR and World Health Organization (WHO) grade remain the main prognostic factors for predicting recurrence, opting for eTSA should be done with great caution as high-grade meningiomas may be harder to resect completely [59, 74]. However, it was not possible to correct for meningioma size, which is unfortunate as very small meningiomas may show very different results. Furthermore, it was not possible to correct for WHO grade, which could theoretically alter the results [31]. Also, the choice of approach varied greatly among mTCA approach studies [1, 2, 5‐7, 9, 47, 51, 56, 64, 70].

Indications for eTSA vary between groups. One group reported operating on all midline meningiomas regardless of size, extension, or configuration except for those tumors that extend from the anterior clinoid process [43]. It has also been suggested that if the tumor extends laterally over the internal carotid artery, chances of GTR are limited [61]. Others have suggested that larger tumors, tumors that extend laterally, involve vasculature, or are calcified are also lesser candidates [23, 44]. Therefore, confounding by indication cannot be ruled out, especially since the patients in these studies were not randomized to either treatment. As a result, the exact indications and contraindications for eTSA remain to be determined.

Future studies should, therefore, focus on identifying clear indications for eTSA for OGM and TSM and its safety by direct comparison in a randomized study. Such a study should ideally be conducted in a research setting by experienced surgeons, as its safety has not been prospectively compared to mTCA and as both approaches seem to come with a considerable learning curve, which results in different outcomes [43]. Given the observation that younger patients seem to benefit more from eTSA compared to older patients (p = 0.02, n = 34), it is not unlikely that specific groups might benefit more from one of the approaches [39]. Probably, patients with relatively small (<3 cm) midline TSMs would probably be the best early candidates. These patients may benefit from a potential higher incidence of visual improvement postoperatively and the relative invasiveness of the eTSA approach. Further evaluation could be focused on characteristics such as size, a cortical cuff, and WHO grading to identify the best potential candidates for either approach [40]. However, due to the low incidence of TSMs and OGMs in general and the great variety in anatomical characteristics among them, this may be challenging. Therefore, other trial designs—e.g., a registry—should be considered when answering this question. Also, future improvement of the instruments used (e.g., 3D endoscopes or glues) may improve the results obtained by eTSA over time [33].

This meta-analysis indicates that the endoscopic transsphenoidal approach (eTSA) has not been shown to be superior to the microscopic transsphenoidal approach (mTCA) for either olfactory groove meningiomas (OGMs) or tuberculum sellae menigniomas (TSMs). More specifically, eTSA was associated with lower GTR rate for OGMs and higher rate of arterial injury for TSMs compared to mTCA. Furthermore, eTSA was associated with more CSF leaks in both OGMs and TSMs compared to mTCA. On the other hand, eTSA was associated with a higher rate of visual improvement postoperatively compared to mTCA in TSMs, which was not observed for OGMs. All conclusions should, however, be interpreted with caution because of the limitations of this study.