Type and anatomical site of the approach

All of the patients were treated using a 3DHD endoscope. Surgery was performed using a “two-nostril—four-hand” technique, previously described [13, 14].

Patients were divided into two different groups according to the type and the anatomical site of the approach: standard or expanded approach. Standard approach consisted of entirely trans-sellar corridor after anterior sphenoid sinus wall removal and sellar floor opening.

Expanded approach consisted of sagittal or coronal extension of the standard one. Unilateral or bilateral posterior ethmoidectomy, different types of maxillectomy and eventually pterygoid removal provided extension in a coronal plane. The extension on the sagittal plane, instead, included a trans-tuberculum, trans-planum sphenoidale and/or trans-ethmoidal approach. The extended trans-clival approach allowed access to the posterior fossa.

CSF leak grading

CSF leaks occurring during surgical procedures were divided into 3 groups: intra-operative no CSF leaks (INL), intra-operative low flow CSF leaks (ILFL) and intra-operative high flow CSF leaks (IHFL). As commonly accepted [15, 16], ILFL included dural opening with minimal flow observed or no involvement of basal cisterns or ventricle, while basal cisterns or ventricular opening defined the IHFL group. Moreover, among patients belonging to the INL group, the presence of thin arachnoid membranes or small dural tears, the early descent of supra-sellar cisterns and the coexistence of independent risk factor for POL and meningitis (e.g. high BMI, older age) [10] identified the INL* subgroup (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Materials and methods.

Intra-operative CSF leak grading system, reconstruction methods and reconstructive strategies.

https://doi.org/10.1371/journal.pone.0245119.t001

Techniques and strategies of skull base reconstruction

Skull base reconstruction strategy was divided according to 1) size of defect; 2) anatomical site of surgical approach; 3) grade of IOL and 4) type of tumor. Available methods to plan proper reconstruction strategies were (Table 1): A) Synthetic dural substitute (Redura, Lyoplant Onlay, Duragen) used in a inlay or onlay fashion, albeit different matrixes were not individually analyzed; B) Autologous abdominal fat; C) Ileo tibial tract (ITT), harvested from the lateral side of thigh; D) sinus mucosal flap, harvested from the posterior wall of sphenoid sinus and positioned on the defect; E) Free graft of nasal mucosa from inferior or middle turbinate; F) Naso-septal flap (NSF) harvested from the septal muco-periosteum and muco-perichondrium, preserving the patency of sphenopalatine artery [17]; G) Inverted U-shaped rhino-pharyngeal mucosal flap harvested detaching mucosa and muscles from the lower clivus and cranio-vertebral junction (CVJ). No materials like bone, cartilage or rigid artificial substitute were used.

The strategies adopted for reconstruction in this series were summarized in 3 major groups (Table 1):

1. Minimal Reconstruction: in cases with a small intracranial opening without CSF leakage, a sponge was used to fill the surgical cavity fixed with fibrin glue. Mucosal flap from sphenoid sinus could have also been used in this strategy (when/if preserved).
2. Standard Reconstruction: dural synthetic substitutes were used for small intracranial openings and for ILFL. In this group autologous abdominal fat and inferior or middle turbinate mucosa was often used respectively to fill the cavity and to cover the defect. Fibrin glue was then used to adhere the dural substitute and fat together. Additionally, patients belonging to the INL* subgroup, who underwent fat and/or reconstruction, have been included in this group as “incomplete type 2 reconstruction”.
3. “Sandwich” Multilayered reconstruction (Fig 1): this strategy was used for large defect, highly risky sites in the surgical approach (anterior fossa and clivus) and IHFL (Fig 1A). Also, patients belonging to the INL* subgroup, in which a pedicled flap reconstruction was performed, have been included in this group as “incomplete type 3 reconstruction”. The first layer consisted in an inlay-positioned synthetic dural substitute or ITT (Fig 1B); then, autologous abdominal fat was placed on the previous layer to fill the empty space (Fig 1C); finally, a second membranous layer was positioned in an onlay fashion, or between the dura and inner bone surface (Fig 1D). At this point, autologous fat was used to cover the lateral edges of the reconstruction outside the bone, tailoring the closure especially in cases with irregular osteo-dural defects (Fig 1E). Every layer was secured with fibrin glue. NSF or rhino-pharyngeal flap, according to the location of the approach, was then rotated on the skull base defect and tightened with fibrin glue (Fig 1F). A paramount step was to fill the empty dead space. The NSF needed to be laid with the muco-perichondral side in direct contact with posterior wall of the sphenoid sinus, in order to avoid a subsequent mucocele formation. Attention was paid to flatten any bony asperity. Poly-vinyl sponges were used to take in site the flap and to pack nasal cavities. Foley catheter was used to take in place the rhino-pharyngeal flap.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. The “sandwich” multilayer closure.

Illustrations (A-F) depict a step-by-step multilayer closure after transtuberculum-transplanum approach for skull base tumors, such as meningioma or craniopharyngioma. A, shows a detail of the neurovascular structures that can be seen in this type of approach, when the tumor is removed; B, a first layer of derived dural is positioned “inlay”, with its edges under the dura; C, an autologous fat graft is collected at the abdomen and it is used to fill the dead spaces (it is considered intraosseous); D, a second layer of collagen derived matrix is positioned “onlay”, between the dura and bone edges, or over the bone defects; E, another layer of abdominal fat graft is positioned (extraosseous); F, finally NSF, harvested at the beginning of the surgery, is carefully rotated on the skull base defect, where the mucosa has already been removed to avoid mucocele formation.

https://doi.org/10.1371/journal.pone.0245119.g001

Nonetheless, although the aforementioned strategies were adopted in the majority of cases, several not standardizable combinations of available materials could have been used for borderline cases (incomplete type 2 and 3 reconstructions).

Lumbar drain positioning criteria

The type of surgical approach and consequent probability for high flow CSF leak represented the criteria for Lumbar Drain (LD) positioning. More specifically, it was used in every expanded approach with osteo-dural defect in the anterior cranial fossa, in the clival region and in the sellar region, and/or when supra-sellar extension was needed (trans-tuberculum and trans-planum approaches). LD was positioned before surgery, under general anesthesia and opened only after surgery.

Statistical analysis

Descriptive statistics were reported as mean and standard deviation for continuous variables. Comparison of proportions were performed with Chi-squared test for categorical variables and, when needed (>20% of values < = 5 and/or presence of values<1), with Cramer’s Phi and V coefficients to verify association between variables. Relationships between dependent and independent variables were evaluated using logistic regression. Moreover, 3 casual extractions from the database were made in order to obtain parametric variables. Statistical significance was defined with a p-value < 0.05. All statistical analyses were performed using SPSS Statistics software.

Reconstruction techniques and POL rate

The development of different materials and different strategies for reconstruction resulted in reduction of POL rate, from the higher values reported in the first papers (10%–40%) [17, 23–35], to the conspicuously lower rate described more recently (from 2,6% to 8,9%) [11, 12, 18]. In this study the overall POL rate was 5.2%, thus confirming the encouraging trend described by the evidence.

Many reconstructive methods and different materials have been reported in the literature [8, 15] but it is likely that the introduction of NSF and the principle of multilayered closure constituted the real game-changer in skull base reconstruction.

Hadad et al. described in 2006 a novel technique for reconstruction of skull base defect using a pedicled NSF, harvested from the nasal septum and vascularized by sphenopalatine branches which provided a reported 5% of postoperative CSF leak rate [36].

Wormald et al. reported the “bath-plug” technique, Leng at al. defined the “gasket seal” closure, Amit et al. described the “champagne cork” technique (CCT), while recently, Cavallo et al. have reported the “3F” strategy (fat—flap—flash) [7, 27–29].

Common principles of new strategies involve (1) filling the empty dead space, (2) obtaining a watertight closure and (3) creating meticulous multilayer dural closure to reduce the entity of intra-operative CSF flow acting on the skull base defect [3, 11, 18].

In addition, different soft materials have been described for dural reconstruction, fat and ITT were the most used [3, 29]. ITT has been described as the ideal graft for dural replacement thanks to its availability, handling and strength [8]. However, also synthetic dural substitutes have been produced with the effort of miming ITT properties [3, 7]. Different ideas have been reported about rigid materials application [16, 30].

IOL grade

According to the literature, a graded approach to skull base repair after endonasal surgery should be considered in the endoscopic era [15]. Esposito et al. classified IOL as follows: Grade 0, no leak observed; Grade 1, small leak without obvious diaphragmatic defect; Grade 2, moderate leak; or Grade 3, large diaphragmatic/dural defect [14]. Sigler at al. focused on the entity of the leak’s flow, defining the high flow leak characterized by large dural defect and basal cisterns or ventricular opening, while small dural defect and moderate CSF leak defined the low flow leak [30].

In this study, Grade 1 or 2 of the Esposito’s system were considered a low flow CSF leak (ILFL), while Grade 3 were considered a high flow leak (IHLF), as reported in other papers [31].

The literature results showed that multilayered closures could be considered essential in the reconstructive phase of EEA in both cases of ILFL and IHFL [2, 3, 12, 21]. Furthermore, NSF combined with multilayered dural reconstruction should be performed in cases of IHFL, while no differences were reported adding vascularized flap in cases of ILFL [15, 18].

In this series, IOL grade could be considered the main predicting factor for POL risk (Table 4).

POL rate, resulted higher in the IHFL group (25.5%) than in both the ILFL and INL groups (10.5% and 1.5% respectively) (Table 3). Moreover, considering single grades of IOL, the IHFL group showed higher probability to have POL (Exp(B) 16.25, p.001), while the INL group resulted in a protective factor with a higher probability to not show POL (Exp(B) 0.016, p.002).

This result could be considered as another validation of the crucial role of CSF leak flow acting on skull base osteo-dural defect and it should be carefully analyzed during reconstruction planning.

Anatomical site and type of surgical approach

Anatomical site and type of surgical approaches still remain a critical aspect of reconstructive algorithm [32]. Indeed, higher post-operative CSF rates have been described in expanded approaches than in standard ones [33]. The extent of skull base defect should be always assessed because frontal extension or caudal extension were associated with higher risk for post-operative CSF leakage [3, 10]. Moreover, meningiomas or craniopharyngiomas were more frequent in this location and higher POL rate was described for these tumors, due to their arachnoid invasion [15]. Some authors have also described the association between POL and the size of the defect (higher risk for defect larger than 3 cm) [10, 31].

In this series, the type of IOL was found to have a key role in determining significant differences among different anatomical sites of the approach (Table 4, Fig 2). IHFL was higher in expanded approaches (88.2%) and in anterior or posterior extensions. Furthermore, POL rate increased in both standard and expanded approaches when IHFL grade was encountered (83.3% and 17.8% respectively). Even considering the anatomical site of the approach, POL was higher when IHFL grade was registered (Fig 2).

These results further highlight the importance of IOL grade in predicting POL risk, considering different surgical approaches, in order to choose the best strategy for reconstruction.

Synnderman et al. recently reviewed different reconstructive strategies for anterior and posterior cranial fossa. A three-layer reconstruction with dural substitute, fascia lata and NSF was described for the anterior cranial fossa [3, 10], while vascularized multilayer reconstruction with fat was considered mandatory for posterior fossa defects to reduce the risk of pontine herniation and also because clival tumors often needed adjuvant radiotherapy [3, 15, 18]. Sellar region was considered a low risk zone due to the low flow of CSF, if no supra-sellar extension was needed (12% of POL described). In the latter case, multilayered reconstruction and NSF were suggested [21, 34].

In this series, reconstructive strategies were prevalently chosen according to expected IOL grade.

POL rate resulted higher in type 3 reconstruction (13%) than in type 2 (6.5%) and type 1 reconstruction (1.7%). This result could be linked to the fact that type 3 reconstruction was used predominantly for procedures with larger osteo-dural defects and higher IOL grades. Moreover, in addition to the volume of the defect that could increase the POL rate, as recently described by Turri-Zanoni [10], it is well known among skull base surgeons that extended approaches and complex reconstructions were frequently adopted for more invasive and more complex malignancies (e.g. craniopharyngioma, chordoma, tuberculum-sellae meningiomas, aggressive and/or giant pituitary adenoma), which could require a more aggressive intradural dissection with consequent higher risk for high grades IOL and POL rates [15, 31]. For the abovementioned reasons, different reconstructive strategies were analyzed according to different IOL grades. Considering the IHFL group, POL rate resulted lower (4.3%) in patients undergoing a type 3 reconstruction (Table 5). This association was stronger considering fat positioning, which was the only significant factor linked to lower POL rate in IHFL group. Albeit thorough analysis of different dural substitutes was beyond the aim of the authors, the use of 2 dural layers reconstruction showed to reduce the rate CSF leak (Table 5). Moreover, although several complications related to dural substitutes positioning have been described in literature [34], namely infections and material extrusion, no complications ascribable to dural substitute issues were reported in the presented series.

Again, the same results were not significant in patients with both low flow or no IOL, further strengthening the role of this aspect in analyzing different strategies for multilayer reconstruction (Fig 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. A case example of "sandwich” multilayered reconstruction.

Preoperative sagittal (A) and coronal (B) MRI scans, show a T1w hypointense suprasellar 37x25 mm lesion with ring enhancement. C-G intra-operative views of reconstruction steps after a craniopharyngioma removal. C, Transtuberculum-transplanum approach for sopra-sellar pathology; note the optic chiasm pushed forward by the tumor; D, “inlay” synthetic dural substitute; E, "onlay” synthetic dural substitute, which is positioned over a first layer of autologous fat graft; F, pieces of autologous fat graft; G, NSF is fashioned over the layers previously described; three-months post-operative sagittal (H) and coronal (I) MRI scans (T1w with gadolinium), show GTR and the multiple layers adequately placed. No postoperative CSF-leaks was described in the current case.

https://doi.org/10.1371/journal.pone.0245119.g004

To summarize, “sandwich” multilayer reconstruction with vascularized flaps should be always considered when IHFL occurs or could be planned considering the type and the anatomical site of surgical approach. The accuracy of different layers positioning and the capability to fill the empty dead space could be considered the target, in order to reduce the CSF flow acting on the skull base defect.

LD positioning

In this study LD was placed predominantly in patients who underwent expanded EEA, according to different IOL grade (11.8% in ILFL group, 66.6% in IHFL group).

The rate of post-operative CSF is reported in Table 5.

LD positioning during EEA has been debated in literature and no agreement has been found yet [21, 34, 35, 37–39].

Recently Zwagerman et al. published a prospective, randomized controlled trial reporting that perioperative LD, used in the contest of NSF and multilayered reconstruction, significantly reduced the rate of postoperative CSF leaks among those patients who might to have high risk of CSF leak (8,2% in LD group vs 21,2% in the control group) [37]. Nonetheless, multilayer reconstruction and firm closure of skull base defect should be mandatory in LD groups to avoid pneumocephalus sustained by unidirectional valve mechanism [37, 40].

In contrast, Eloy et al. retrospectively review 59 patients who underwent EEA repair of high flow CSF leak with NSF without LD, reporting 0% of POL rate, even if the majority of patients had sopra-sellar pathologies [38].

In this study, no significative statistic differences were found between LD and no LD group. Considering the IHFL group, POL rate resulted lower in LD positioning group (23.5% vs 29.4%), although this difference did not reach the statistical significance. Thus, no conclusions about the utility of LD positioning could be drawn from this study, even if results seem to support a contribution in reducing POL rate, especially in IHFL group.

No association between LD and meningitis was found in this study, despite the literature reports a positive correlation between longer duration of LD and higher rate of infections [41].

Re-do surgery and radiation

Re-do surgeries and radiotherapy were considered a significant part of skull base tumors management, so their impact on POL have been analyzed in different studies [18, 42]. Nishioka et al. reported that POLs were significantly increased in case of prior trans-sphenoidal surgery or radiotherapy [42]. This could be related to bone changes and mucosal devascularization due to radiotherapy, while sphenoid scars could contribute to the higher risk described for second surgery [18]. In this series, no significative different POL rates were found between patients who underwent previous radiotherapy and/or surgery. Nevertheless, vascularized multilayer reconstruction was strongly recommended in these patients in order to provide better healing and prevent bone damage due to radiotherapy [3, 43].

A decreasing trend through the years in preventing CSF leaks

A decreasing trend in POL rate was observed during the years (from 14.7% to 2%; p.002) (Fig 3) and these results were confirmed even stratifying patients for IOL grade and for type of surgical approach, confirming trends described from other authors [10, 44–46]. Analyzing the curve, two peaks in POL were observed, the former at the beginning and the latter in the middle of the center’s endoscopic learning curve. This feature was confirmed even in the expanded approach and in the IHFL group. The first peak could be related to the unavoidable drawbacks of acquiring a basic expertise in reaching a proper setting, an adequate affinity among the team members, other than a personal surgical experience and skills. The second peak, instead, could be explained by the following growth of the number of more complex operations (e.g. more expanded approaches). This has probably led to initial higher rates of POL, followed by a concrete reduction in the last two years. Moreover, analyzing the number of patients without POL through the years, a constant increasing trend was observed and this was confirmed also after stratification (Fig 3). These details could be related to the evolution of reconstructive strategy adopted during the years, with more tailored reconstruction in these high-risk patients, according to the expected IOL grade.

Limitation

Principal limitation of this study is its retrospective nature. The choice of reconstructive strategy and general management were made in absence of a standardized protocol. Moreover, the adopted reconstructive strategies, summarized in Table 1, could not be considered mandatory, since they represent an indicative and no standardizable algorithm resulting from different methods adopted and lessons learned during the authors’ institution experience. Nevertheless, although the analysis does not allow to make stratified recommendations about the best reconstructive strategies, the role of risk factors and reconstructive techniques could be reliably highlighted. This was, in fact the aim of this paper.

Another limitation of this study lies in the fact that a consistent subgroup of patients who underwent minimal skull base reconstruction have been included. Nonetheless, albeit it is widely accepted among skull base surgeons that small sellar defects without intra-operative CSF leak might not need reconstruction, the risk of post-operative leak could not be considered nil among these patients. Moreover, the presence of small dural tears or coexisting independent risk factors for CSF leak could explain the reported rates of POL. Besides, although the use of fibrin glue and sphenoidal sinus mucosa are not widely considered a reconstructive strategy, the use of fibrin glue for the treatment of post-operative CSF leak have been described [47]; additionally, the rationale behind sphenoid sinus mucosa repositioning lies into the evidence of re-epithelization and subsequent coverage of the defect. Hence, from this perspective, patients treated with minimal reconstruction strategy have been included.

Nonetheless, further considerations would need a prospective analysis with the aim to analyze additional risk factors and to validate reconstructive strategies.