Systematic Review: Success Rate of Endoscopic Endonasal versus Combined Endonasal and Transorbital Neuroendoscopic Approach for Nontraumatic Cerebrospinal Fluid Leak Repairs in the Lateral Recess of Sphenoid Sinus

Abstract

Background  Cerebrospinal fluid (CSF) leaks from the lateral recess of the sphenoid sinus (LRS) occur due to a skull base defect and are important to treat due to the associated morbidity, e.g., life-threatening meningitis. Nontraumatic CSF leaks have a predilection toward obesity which is a rising phenomenon. LRS is notoriously difficult to access because of its lateral location and its associated neurovascular complications. An alternative surgical corridor has been explored which is the transorbital neuroendoscopic (TONES) approach.

Objective  To compare the success rate of the endoscopic endonasal with the TONES approaches.

Rationale  This is the first systematic review on the endoscopic endonasal and combined transorbital approaches to treat CSF leaks from the LRS.

Method  A PRISMA-concordant systematic review. PubMed, MEDLINE, EMBASE, Web of Science, and SCOPUS were searched. The studies underwent abstract and full-text screening by two reviewers. The data collected included patient demographic, surgical approach, reconstruction method, layers and materials, follow-up period, ROBINS-I bias, complications, and success rate.

Results  In total, 26 of 4,385 studies were included for further synthesis. Of these studies, a total of 336 patients were identified from a cohort of 910 patients. The endoscopic endonasal approach showed a repair success rate of 95.24% and the combined TONES and endonasal approach showed a success rate of 100%.

Conclusion  Both the endoscopic endonasal and transorbital approach provide a good success rate. However, due to the small TONES sample, large, randomized control trials are needed.

Introduction

The lateral recess of the sphenoid sinus (LRS) exists as a hyperpneumatized variant of the sphenoid sinus in which the sphenoid sinus cavity extends laterally beyond an imaginary line connecting the anterior opening of the Vidian canal and the extracranial end of the foramen rotundum, known as the VR line, in the coronal plane. ¹ The LRS is found within the anterior skull base which extends from the posterior wall of the frontal sinus anteriorly to the cribriform plate of the ethmoid bone and the sphenoid planum and lesser wing of the sphenoid in the sagittal plane more posteriorly. ²

Cerebrospinal fluid (CSF) leaks from the skull base occur when a fistula forms due to an osteodural disruption, ³ allowing communication between the subarachnoid space and the sinonasal tract, ⁴ which presents as colorless unilateral rhinorrhoea. ⁵ Spontaneous CSF (sCSF) rhinorrhea of the skull base is a multifactorial manifestation influenced by a dynamic association between various factors such as obesity, obstructive sleep apnea, intracranial hypertension, increased pneumatization, and skull base thinning which risk the structural integrity of the skull base. ⁶

The prevalence of sCSF leaks from the LRS has been reported variably with a recent systematic review indicating that the lateral sphenoid sinus was one of the most common locations in anterior skull base defects. ⁷

Endoscopic endonasal approaches to access the LRS include the transpterygoid approach and the transethmoidal approach. ^{8 9} Typically, a transpterygoid endonasal endoscopic approach is employed to reach the LRS via the posterior wall of the maxillary sinus and the pterygopalatine fossa (PPF). ¹⁰ Although the transpterygoid endoscopic endonasal approach provides appropriate access to the LRS, it comes with important considerations. The transpterygoid approach involves extensive dissection of the PPF and the surrounding area where important neurovascular structures may be compromised such as the pterygopalatine ganglion which when transected can lead to palatal hypoesthesia that has functional ramifications with regard to deglutition. ¹¹ In addition to this, dry eye symptoms may be experienced by patients when the Vidian nerve or the pterygopalatine ganglion is damaged or transected which normally supply the lacrimal gland. ¹¹ Moreover, the division of the maxillary nerve which can result in persistent or transient facial hypoesthesia and or paraesthesia could impact a patient's quality of life. ¹² The sphenopalatine artery and the internal maxillary artery which are encountered may also require ligation. ¹³ Other drawbacks include the risk of bleeding due to the high vascularity of the area which can impede visualization and the potential of inadequate exposure which can lead to poor instrumentation. ¹⁴

The endoscopic endonasal approach to the skull base has also been considered as a partially sterile approach ¹⁵ because it involves the microbe-colonized sinonasal cavity and the sterile cranial fossa. ¹⁶ Bendersky et al also criticized the endoscopic transpterygoid approach for not actually being a minimally invasive technique. ¹⁵

The transorbital neuroendoscopic (TONES) approach was developed by Moe et al ¹⁷ which provides access to the anterior skull base via the orbit in a sterile, minimally invasive manner, and with full control of the orbital contents and the optic nerve. ¹⁸ Currently, there are only two systematic reviews on the use of the TONES approach one by Vural et al ¹⁹ and another by Houlihan et al ²⁰ which look at various skull base pathologies including CSF leaks. Both of these studies ^{19 20} demonstrate that the TONES approach is feasible and promising but requires further research.

To date, there has been no systematic review of endoscopic endonasal repair of CSF leaks specifically from the LRS regardless of etiology. A recent systematic review on the surgical repair of sCSF leaks found 31 studies with an anterior skull base leak which had an overall postoperative CSF leak rate of 9% with all the studies included undertaking a purely endoscopic endonasal repair apart from two studies. ⁷ An earlier systematic review on the endoscopic repair of various CSF leak etiologies including 52 studies found a success rate of 90.6% (95% confidence interval [CI]: 80–92.5%) for the primary repair, similar to the prior study. ²¹ However, this study did not conduct a sub-group analysis on patients with sCSF leaks and moderate heterogeneity was found between the series with an I ² of 29.2% (95% CI: 0–50%). ²¹

The primary objective of this study is to investigate the repair success rate of the combined endoscopic endonasal and TONES approach in comparison to the endoscopic endonasal approach for nontraumatic defects in the LRS. The success rate was defined as the rate of CSF leak repairs occurring without postoperative leak recurrence at the site of the original leak. The secondary outcomes are eliciting the reported complications that are associated with either approach.

Materials and Method

Study Design

This is a systematic review performed following the Preferred Reporting Items for Systematic Reviews (PRISMA) guidelines. ²² The systematic review protocol was prospectively registered on the PROSPERO (CRD42023393592) on the January 27, 2023.

Eligibility Criteria

The eligibility criteria for study inclusion and exclusion are demonstrated in Table 1 . Studies were excluded if they did not follow or report the population, intervention, comparator, and outcomes (PICO) of interest highlighted below. Studies were grouped for synthesis according to the endoscopic approach (endonasal vs. transorbital).

Table 1. Eligibility criteria.

PICO	Inclusion criteria	Exclusion criteria	Limits
Population	• Nontraumatic CSF leak (congenital and spontaneous etiology) • Lateral sphenoid recess leak location	• Traumatic CSF leak • Skull base defects not in the lateral sphenoid recess • Other pathologies of the skull base, e.g., tumors	• Human studies • Adult population
Intervention	• Transorbital neuroendoscopic (TONES) approach	• Open repair approach • Cadaver	• None
Comparator	• Endoscopic endonasal approach	• Open repair approach	• None
Outcomes	• Postoperative CSF leak rate • Complications: dry eye, facial numbness, dental/palatal numbness, neurological, meningitis, others reported • Revision rates • Number of procedures	• None	• None
Study design	• Primary studies • Reviews that contain primary data • Case studies for transorbital approach	• Secondary data, e.g., reviews • Case study (only one patient) for endonasal approach	• English language

Abbreviations: CSF, cerebrospinal fluid; PICO, population, intervention, comparator, and outcomes.

Information Sources

EMBASE, MEDLINE, PubMed, SCOPUS, and Web of Science were searched by the first reviewer.

Search Strategy

The search strategy was developed in collaboration with an information specialist librarian (J.M.) for the databases MEDLINE and EMBASE. The databases were searched from the date of inception to February 7 or 8, 2023. An example search strategy is demonstrated using the MEDLINE search strategy on Table 2 ; for the full search strategy for each database, see Appendix A .

Table 2. Search strategy on PubMed.

PubMed

Search: (((anterior skull base[Title/Abstract]) OR (lateral sphenoid recess[Title/Abstract])) AND (cerebrospinal fluid leak[Title/Abstract])) OR (cerebrospinal fluid rhinorrhoea[Title/Abstract]) = 288

Appendix A. Search strategies for PubMed, MEDLINE, Embase, Web of Science, and SCOPUS.

PubMed	Search: (((anterior skull base[Title/Abstract]) OR (lateral sphenoid recess[Title/Abstract])) AND (cerebrospinal fluid leak[Title/Abstract])) OR (cerebrospinal fluid rhinorrhoea[Title/Abstract]) = 288
MEDLINE	Ovid MEDLINE(R) ALL <1946 to February 07, 2023> 1 Skull Base/5806 2 “anterior skull base.”mp. 1847 3 ((“lateral recess” adj4 sphenoid).ab. or “lateral recess.”mp.) adj4 sphenoid.ti. [mp = title, book title, abstract, original title, name of substance word, subject heading word, floating sub-heading word, keyword heading word, organism supplementary concept word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 38 4 “sphenoid lateral recess.”mp. 11 5 1 or 2 or 3 or 4 7062 6 Cerebrospinal Fluid/ 19414 7 “Cerebrospinal Fluid.”mp. 131924 8 “Cerebral Spinal Fluid.”mp. 2316 9 csf.mp. [mp = title, book title, abstract, original title, name of substance word, subject heading word, floating sub-heading word, keyword heading word, organism supplementary concept word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 113210 10 6 or 7 or 8 or 9 192489 11 leak*.mp. [mp = title, book title, abstract, original title, name of substance word, subject heading word, floating sub-heading word, keyword heading word, organism supplementary concept word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 134291 12 rhinorrh?ea.mp. [mp = title, book title, abstract, original title, name of substance word, subject heading word, floating sub-heading word, keyword heading word, organism supplementary concept word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 7137 13 11 or 12 139626 14 repair.mp. [mp = title, book title, abstract, original title, name of substance word, subject heading word, floating sub-heading word, keyword heading word, organism supplementary concept word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 397003 15 surg*.mp. [mp = title, book title, abstract, original title, name of substance word, subject heading word, floating sub-heading word, keyword heading word, organism supplementary concept word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 3585882 16 endoscop*.mp. [mp = title, book title, abstract, original title, name of substance word, subject heading word, floating sub-heading word, keyword heading word, organism supplementary concept word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 284041 17 14 or 15 or 16 3944747 18 transorbital.mp. [mp = title, book title, abstract, original title, name of substance word, subject heading word, floating sub-heading word, keyword heading word, organism supplementary concept word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 1034 19 endonasal.mp. [mp = title, book title, abstract, original title, name of substance word, subject heading word, floating sub-heading word, keyword heading word, organism supplementary concept word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 7032 20 transnasal.mp. [mp = title, book title, abstract, original title, name of substance word, subject heading word, floating sub-heading word, keyword heading word, organism supplementary concept word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 4400 21 18 or 19 or 20 11875 22 5 and 10 and 13 and 17 and 21 527
Embase	Embase <1974 to 2023 February 07> 1 Skull Base/ 14837 2 “anterior skull base.”mp. 2614 3 ((“lateral recess” adj4 sphenoid).ab. or “lateral recess.”mp.) adj4 sphenoid.ti. [mp = title, abstract, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword heading word, floating subheading word, candidate term word] 47 4 “sphenoid lateral recess.”mp. 18 5 1 or 2 or 3 or 4 16238 6 Cerebrospinal Fluid/ 121794 7 “Cerebrospinal Fluid.”mp. 225871 8 “Cerebral Spinal Fluid.”mp. 3607 9 csf.mp. [mp = title, abstract, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword heading word, floating subheading word, candidate term word] 177610 10 6 or 7 or 8 or 9 299416 11 leak*.mp. [mp = title, abstract, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword heading word, floating subheading word, candidate term word] 200827 12 rhinorrh?ea.mp. [mp = title, abstract, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword heading word, floating subheading word, candidate term word] 19194 13 11 or 12 218160 14 repair.mp. [mp = title, abstract, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword heading word, floating subheading word, candidate term word] 551932 15 surg*.mp. [mp = title, abstract, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword heading word, floating subheading word, candidate term word] 4801808 16 endoscop*.mp. [mp = title, abstract, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword heading word, floating subheading word, candidate term word] 512865 17 14 or 15 or 16 5337335 18 transorbital.mp. [mp = title, abstract, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword heading word, floating subheading word, candidate term word] 1238 19 endonasal.mp. [mp = title, abstract, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword heading word, floating subheading word, candidate term word] 9847 20 transnasal.mp. [mp = title, abstract, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword heading word, floating subheading word, candidate term word] 6147 21 18 or 19 or 20 16374 22 5 and 10 and 13 and 17 and 21 1007
Web of Science	24 #15 AND #17 AND #19 AND #21 AND #23 and Preprint Citation Index (Exclude – Database) 1333 23 #11 OR #12 OR #13 22 #11 OR #12 OR #13 and Preprint Citation Index (Exclude – Database) 21 #8 OR #9 OR #10 20 #8 OR #9 OR #10 and Preprint Citation Index (Exclude – Database) 19 #7 OR #6 18 #7 OR #6 and Preprint Citation Index (Exclude – Database) 17 #4 OR #5 16 #4 OR #5 and Preprint Citation Index (Exclude – Database) 15 #1 OR #2 OR #3 14 #1 OR #2 OR #3 and Preprint Citation Index (Exclude – Database) 13 TS = (transnasal) 12 TS = (endonasal) 11 TS = (transorbital) 10 TS = (endoscop*) 9 TS = (surg*) 8 TS = (repair) 7 TS = (rhinorrh*ea) 6 TS = (leak*) 5 TS = (csf) 4 TS = (cerebrospinal fluid) 3 TS = (lateral sphenoid recess) 2 TS = (anterior skull base) 1 TS = (skull base)
SCOPUS	1,227 document results ((TITLE-ABS-KEY (skull AND base)) OR (anterior AND skull AND base) OR (TITLE-ABS-KEY (lateral AND sphenoid AND recess))) AND ((cerebrospinal AND fluid) OR (TITLE-ABS-KEY (csf))) AND ((TITLE-ABS-KEY (leak*)) OR (TITLE-ABS-KEY (rhinorrh?ea))) AND ((TITLE-ABS-KEY (repair)) OR (TITLE-ABS-KEY (surg*)) OR (TITLE-ABS-KEY (endoscop*))) AND ((TITLE-ABS-KEY (transorbital)) OR (TITLE-ABS-KEY (endonasal)) OR (TITLE-ABS-KEY (transnasal)))

Selection Process

A two-reviewer blinded process was conducted to independently screen the retrieved studies using Rayyan. ²³ Identified studies were loaded into Endnote ²⁴ then uploaded into Rayyan using the Endnote-compatible format. Discrepancies were discussed, if not resolved were arbitrated by a third independent reviewer (H.N./J.M.). ²⁵ During the full-text screening, authors were contacted, where the contact information was available, for further clarification on the previously outlined PICO.

Data Collection process

Data were independently extracted by two reviewers (U.N.A./M.S.S.) using an Excel sheet. Data collection was initially piloted using the first five studies so that adjustments could be made to the design of the data sheet to allow for more cohesive and coherent data. Where individual participant data were available for age, body mass index (BMI), and follow-up time, this was used to calculate a mean using Standard Deviation Calculator. ²⁶ Corresponding authors were contacted to clarify data or obtain further data such as regarding the success rate of the endoscopic repair for the specific population of interest which may have not been reported. Automation tools were not used to aid with the data collection.

Data Items

The Data Items Collected Are Presented in Appendix B

Appendix B. Data collection items.

Characteristics of study participants	• Mean age • Mean BMI • Gender • Diagnostic method • Previous diagnosis of IIH • Radiological signs of IIH • Mean ICP • Side of CSF leak • Endoscopic repair approach used • Postoperative antibiotic use
Characteristics of included studies	• Citation • Study location • Sample size • Participant selection methods • Mean length of follow-up • Ethical approval
Intervention	• Use of lumbar drain • Use of CSF shunt • Use medicine to treat increased ICP • Endonasal technique used • Transorbital technique used • Reconstruction method:  – Material used on each layer  – Number of layers • Nasal packing
Outcomes/results	• The primary outcome of included study • Reporting of postoperative leak • Reported complications  – Dry eye  – Dry nasal mucosa/crusting  – Facial paraesthesia/hypoesthesia  – Facial pain  – Dental/palatal hypoesthesia  – -meningitis

Abbreviations: BMI, body mass index; ICP, intracranial pressure; IIH, idiopathic intracranial hypertension.

Data regarding adjunct treatment such as the use of intracranial pressure (ICP) control medication, CSF shunts, and lumbar drains were collected as recent prospective case series and systematic review found that in a total of 679 sCSF leak patients, successful primary repair was 92.82% in patients with acetazolamide treatment or CSF shunts, whereas a significantly decreased success rate ( p  < 0.001) was found in those not managed for elevated ICP with a success rate of 81.87%. ²⁷

Data regarding the reconstruction method in the skull base repair used by the studies were also collected and tabulated, as although a recent meta-analysis by Abiri et al ²⁸ showed no significant impact of the materials and method used on the success of CSF leak repair, it is an important aspect to account for.

Study Risk-of-Bias Assessment

The Risk of Bias in Non-randomized Studies of Intervention tool (ROBINS-I) ²⁹ was used to assess the included studies. The Risk of Bias Visualization tool (ROBVIS) ³⁰ was used to generate visual plots which are a representation of the risk-of-bias results for each domain from ROBINS-I. This was conducted by U.N.A. and reviewed by M.S.S.

Effect Measures

The primary outcome in this study was the success rate of the endoscopic endonasal repair and the TONES repair. Success was defined as the lack of postoperative CSF leaks, and this was measured as a percentage or in terms of proportion of success.

Synthesis Methods

Results and data collected were synthesized and represented in tables to allow for schematic analysis.

Results

Study Selection

Database searches identified a total of 4,382 articles, ultimately yielding 36 studies for inclusion, of which a further 10 were excluded. The PRISMA flowchart for this is shown in Fig. 1 .

Fig. 1.

Flow diagram of search results and included studies following title and abstract screen and full-text screen. *No data on success rate of patients in the study with a lateral sphenoid recess leak. **Reason for exclusion was not labelled during screening.

Study Characteristics

All the studies included were between the years 2007 and 2023, as shown in Table 3 . The study designs of the included studies were retrospective, prospective, case series and case reports. Twenty-four of these studies had an endonasal endoscopic repair to the LRS nontraumatic CSF leaks and 2 studies had a transorbital approach to the nontraumatic CSF leaks in the LRS. The studies included for further synthesis had a total of 339 patients with nontraumatic CSF leaks in the LRS that were repaired via either an endonasal endoscopic approach or a transorbital endoscopic approach from studies with a total cohort size of 910 patients. A total of 16 studies used a transpterygoid approach for repair, 3 used a transsphenoidal approach, 2 studies used a transethmoidal approach, 2 studies used a prelacrimal recess approach, and 2 studies used a combined transorbital and endonasal approach. Four studies used more than one endonasal endoscopic approach for their patients.

Table 3. Characteristics of studies eligible for tabulation synthesis.

Citation	Country	Ethical approval	Study design	No. of patients		Endoscopic repair approach	Follow-up time in LRS patients (mo)
				Cohort	LRS patients		Mean + SD	Range
Abdelazim et al, 2018 33	Egypt	Yes	Prospective	17	6	Transpterygoid	–	0.75–1
Babu et al, 2019 34	India	–	Case series	3	3	Transpterygoid	–	Minimum 3 months (3+)
Bozkurt et al, 2020 35	Italy	Yes	Retrospective	39	32	Transpterygoid	–	–
Campbell et al, 2016 36	United States	Yes	Retrospective	32	12	Transpterygoid	–	–
Castelnuovo et al, 2007 37	Italy	–	Retrospective	15	10	Transpterygoid	37.6 ± 21.7	–
El Tarabishi et al, 2016 56	Egypt	Yes	Prospective	10	10	Transphenoidal	46.9 ± 8.26	–
He and Zhen, 2022 38	China	Yes	Retrospective	12	11	Transpterygoid	43.36 ± 18	9–65
Huang et al, 2022 39	China	Yes	Retrospective	49	31	Transpterygoid, prelacrimal recess	–	–
Jolly et al, 2023 40	UK	Helsinki Declaration	Retrospective	33	3	Transpterygoid	–	–
Kirtane et al, 2012 47	India	n/a	Retrospective	15	13	–	–	–
Lopatin et al, 2003 48	Russia	–	Retrospective	21	5	–	–	–
Lopatin et al, 2011 49	Russia	–	Retrospective	173	26	–	–	–
Maxfield et al, 2020 41	United States	Yes	Retrospective	19	16	Transpterygoid	16.6	1–68
Muscatello et al, 2010 42	Italy	–	Retrospective	8	2	Transpterygoid	14.5 ± 6.26 (2)	10–19
Nix et al, 2016 51	UK	–	Retrospective	76	4	-(transpterygoid?)	–	–
Rathod et al, 2021 43	India	Helsinki Declaration	Case series	11	10	Transpterygoid	29.7 ± 25.9	3–96
Sreenath et al, 2021 13	United States	Yes	Retrospective	33	33	Transpterygoid	–	–
Tabaee et al, 2010 44	Brazil	Yes	Retrospective	13	13	Transpterygoid, transphenoidal, transethmoidal	56.4 ± 39.6	8–145.2
Tomazic and Stammberger, 2009 45	Austria	–	Retrospective	5	5	Transpterygoid, transphenoidal	6.5	–
Ulu et al, 2018 46	Turkey	–	Retrospective	13	9	Transpterygoid	73.3 ± 21.5	–
Xu et al, 2022 53	China	Yes	Retrospective	15	14	Transethmoidal, prelacrimal recess	–	–
Jiang et al, 2014 57	United States	Yes	Retrospective	48	8	–	–	–
Zocchi et al, 2021 58	Italy	–	Retrospective	167	38	–	–	–
Zoli et al, 2016 50	Italy	–		23	23	-(transpterygoid?)	–	4–167
Lubbe et al, 2020 31	South Africa	Yes	Case report	1	1	Combined transorbital precaruncle and endonasal	21	21
Ramakrishna et al, 2016 32	United States	Yes	Retrospective	60	1	Combined transorbital lateral retrocanthal and endonasal	–	–

Abbreviations: LRS, lateral recess of sphenoid patients of interest; SD, standard deviation.

Note: “—“ denotes data not reported; “?”– denotes where outlined surgical approach appears to be transpterygoid but it is not confirmed.

The studies in Appendix C were from 2004 to 2021 and were not included for further synthesis as the etiology was not clear in four studies, the success rate was not clear in three studies, only one relevant case was found in one study, the lateral sphenoid recess was not explicitly mentioned in one study, and one study was a supplementary article to a study already included. Since these elements were not clarified in the article or by the author when contacted, further synthesis was not conducted as further synthesis could be misleading.

Appendix C. Studies excluded from further synthesis.

Citation	Country	Ethical approval	Study design	No. of patients		Endoscopic repair approach	Reason for excluding from synthesis
				Cohort	LRS patients
Al-Nashar et al, 2004	United States	–	Retrospective	13	–	Transpterygoid	Etiology not confirmed
Bendersky et al, 2011	Argentina	–	Case series	2	2	Transcranial and transphenoidal	Only one case of LRS undergoing endoscopic repair
El Tarabishi et al, 2016	Egypt	Yes	Prospective	7	7	Transphenoidal	Supplementary analysis to already included study [duplicate]
Gore, 2017	United States	–	Case series	7	–	Transphenoidal	Etiology not confirmed
Kim-Orden et al, 2019	United States	Yes	Retrospective	20	–	–	No explicit mention of LRS
Kreatsoulas et al, 2021	United States	Yes	Retrospective	46	10	–	Etiology not confirmed
Martinez-Capoccioni et al, 2015	Spain	–	Retrospective	25	2	–	Success rate not confirmed
McCormick et al, 2021	United States	–	Prospective	55	24	–	Success rate not confirmed
Rajasekar et al, 2019	India	Yes	Retrospective	7	3	Transpterygoid	Etiology not confirmed
Sanghvi et al, 2021	United States	yes	Retrospective	33	10	–	Success rate not confirmed

Abbreviation: LRS, lateral recess of the sphenoid sinus.

The demographic of the patients with nontraumatic CSF leaks in the LRS demonstrated in Appendix D showed 11 studies reported more than 50% incidence in females, whilst 2 studies reported more than 50% incidence in males, 1 study had an equal distribution, and 9 studies did not report this specifically in the population of interest. The average age and standard deviation of the patients of interest included ranged from 40.7 ± 7.21 to 60.5 ± 9.19. Seven studies reported a left-sided incidence of more than 50%, 4 studies reported more than 50% in the right, and 1 study with equal distribution, 12 did not report this. The mean BMI ranged from 25.17 ± 3.2 to 39.5 ± 9.4.

Appendix D. Characteristics of study participants of interest.

Citation	Gender, n (%)		Average age ± SD	Side of leak, n (%)		Mean BMI + SD
	Male	Female		Right	Left
Abdelazim et al, 2018	–	–	–	–	–	–
Babu et al, 2019	2 (66.7)	1 (33.3)	53.67 ± 7.59	2 (66.7)	1 (33.3)	–
Bozkurt et al, 2020	11 (34.375)	21 (65.625)	–	–	–	30.2
Campbell et al, 2016	–	–	–	–	–	–
Castelnuovo et al, 2007	6 (40)	9(60)	–	7 (46.7)	8 (53.3)	30 ± 5
El Tarabishi et al, 2016	3 (30)	7 (70)	40.7 ± 7.21	3 (30)	7 (70)	35.5 ± 2.84
He and Zhen, 2022	4 (36.4)	7(63.6)	56.36 ± 9.71	6 (54.5)	5 (45.5)	28.25 ± 6.2
Huang et al, 2022	–	–	–	–	–	–
Kirtane et al, 2012	6 (46.2)	7 (53.8)	46.3 ± 12.2	7 (53.8)	6 (46.2)	–
Lopatin et al, 2003	–	–	–	–	–	–
Lopatin et al, 2011	–	–	–	–	–	–
Maxfield et al, 2020	–	–	54.0625 ± 7.307	8 (50)	8(50)	–
Muscatello et al, 2010	1 (50)	1 (50)	60.5 ± 9.19	–	–	–
Nix et al, 2016	–	–	–	–	–	–
Rathod et al, 2021	4 (40)	6 (60)	48.6 ± 12	4 (40%)	6 (60%)	25.96 ± 1.5
Sreenath et al, 2021	30 (90.9)	3 (9.1)	53 ± 13.6	14 (42.4)	19 (57.6)	39.5 ± 9.4
Tabaee et al, 2010	5 (38)	8 (62)	57.1 ± 14.3	5 (38)	8(62)	–
Tomazic and Stammberger, 2009	1 (20)	4 (80)	51.2	3 (60)	2 (40)	31.9
Ulu et al, 2018	2 (22.2)	7 (77.8)	47.78 ± 10.63	2 (22.2%)	7 (77.8)	–
Xu et al, 2022	6 (42.9)	8 (57.1)	50.21 ± 8.77	–	–	25.17 ± 3.2
Jiang et al, 2014	–	–	–	–	–	35.8 ± 6.9
Zocchi et al, 2021	–	–	–	–	–	–
Zoli et al, 2016	7 (30.4)	16 (69.6)	52	11(47.8)	12(52.2)	–

Abbreviations: BMI, body mass index; SD, standard deviation.

Appendix E represents the studies in which the diagnostic modalities β-2 transferrin, computed tomography (CT), CT cisternography, magnetic resonance imaging (MRI), MRI cisternography, or intrathecal fluorescein were used on some or all of the patients included in the studies. Twelve studies mention the use of β-2 transferrin. A total of 17 studies mention the use of CT with 3 studies mentioning the use of CT cisternography. A total of 14 studies report the use of MRI in some/all their patients, with eight studies mentioning MRI cisternography and eight studies mention the use of intrathecal fluorescein.

Appendix E. CSF leak diagnosis and localization tools.

Citation	β-2 transferrin (y/n)	CT		MRI		Intrathecal Fluorescein (y/n)
		y/n	Cisternography (y/n)	(y/n)	Cisternography (y/n)
Abdelazim et al, 2018	Y	Y	–	Y	Y	Y
Babu et al, 2019	–	Y	–	–	Y	–
Bozkurt et al, 2020	Y	Y	–	Y	Y	Y
Campbell et al, 2016	Y	Y	–	Y	–	Y
Castelnuovo et al, 2007	Y	N	–	–	–	Y
El Tarabishi et al, 2016	Y	Y	–	Y	Y	Y
He and Zhen, 2022	–	Y	–	Y	–	–
Huang et al, 2022	–	Y
Jolly et al, 2023	–	–	–	–	–	–
Kirtane et al, 2012	Y	–	Y	–	–	–
Lopatin et al, 2003	–	Y	–	Y	–	–
Lopatin et al, 2011	–	Y	Y	–	Y	–
Maxfield et al, 2020	–	Y	–	–	–	–
Muscatello et al, 2010	Y	Y	–	Y	–	Y
Nix et al, 2016	–	–	–	–	–	–
Rathod et al, 2021	–	Y	–	Y	Y	–
Sreenath et al, 2021	–	Y	–	Y	–	–
Tabaee et al, 2010	Y	Y	Y	Y	–	Y
Tomazic and Stammberger, 2009	–	–	–	–	–	Y
Ulu et al, 2018	Y	–	–	–	Y	–
Xu et al, 2022
Jiang et al, 2014	–	Y	–	Y	–	–
Zocchi et al, 2021	Y
Zoli et al, 2016	Y	Y	–	Y	–	–
Lubbe et al, 2020	Y	Y	–	Y	–	–
Ramakrishna et al, 2016	–	Y	–	Y	–

Abbreviations: CT, computed tomography; MRI, magnetic resonance imaging.

Note: Y, yes; N, no; “–” = data not reported.

Appendices F , G , and H demonstrate the reconstructive techniques used in terms of the number of layers, reconstructive materials used, and supportive materials such as surgical strips, gauze, gelo-foam, tissue glue, sphenoid sinus fat plug, as well as the use of the nasal pack. Appendices G and H were designed to display studies that reported a different method of reconstruction depending on the defect size.

Appendix F. Mode of skull base reconstruction.

Citation	Reconstruction not based on defect size
	Number of layers	Materials	Surgicel strips	Gauze	Gelo-foam	Fibrin glue/tissue sealant	Nasal pack	Sphenoid fat plug
Babu et al, 2019	3	F, FL, FMF /pedicled NSF	Y	–	–	Y	Y	–
Campbell et al, 2016	3	Fa underlay, septal/septal MT flap, FMF/vascularized flap	–	–	Y	Y	Y	–
El Tarabishi et al, 2016	6	F, Fa, tragal cartilage, B, FMF	Y	–	Y	Y	–	–
He and Zhen, 2022	3	F, FMF, mucosal flap	–	Y	–	–	–	–
Jolly et al, 2023	2	NSF/F and B/DuraGen/biodesign	Y	–	–	Y	Y	Y
Kirtane et al, 2012	2	F and Fa	–	–	Y	Y	Y	Y
Lopatin et al, 2003	2	F and FL	–	–	–	Y	Y	–
Lopatin et al, 2011	2	F/nasal septum cartilage/	–	–	–	Y	Y	–
Maxfield et al, 2020	2	Cartilage/alloderm and FMF/alloderm/NSF/conchal perichondrium	–	–	–	–	–	–
Muscatello et al, 2010	2	Collagenous matrix underlay and MP flap overlay	–	–	–	Y	–	–
Nix et al, 2016	4	F, FL, F, NSF	–	–	–	Y	Y	–
Rathod et al, 2021	2	MT/ethmoid bone graft and FL overlay	Y	–	Y	Y	N	Y
Sreenath et al, 2021	1	Free mucosal MT/nasal floor graft	Y	–	–	Y	Y
Tabaee et al, 2010	2	NSF, and FL /MT/F/temporalis fascia/dural substitute	–	–	–	–	–	–
Tomazic and Stammberger, 2009	3 or4	FL, FL/AF, cartilage	–	–	–	Y	Y	Y
Ulu et al, 2018	4	AF, pedicled NSF/MT flap, AFa, balloon stent	–	–	–	Y	–	–
Xu et al, 2022	4	Artificial dura, F, FMF, B	–	Y	Y	Y	–	–
Zocchi et al, 2020	–	–	–	–	–	–	–	–
Zoli et al, 2016	3	FL, MP, B/F	–	–	–	–	–	–
Lubbe et al, 2020	4	AF, DuraGen seal, MP, AF	–	–	–	–	Y	–
Ramakrishna et al, 2016	2	Dural patch of irradiated cadaveric dermis and NSF	–	–	–	–	–	–

Abbreviations: AF, abdominal fat; AFa, abdominal fascia; B, bone; F, fat; Fa, fascia; FL, fascia lata, FMF, free mucosal flap; MT, middle turbinate; MP, mucoperiosteum; NSF, nasoseptal flap.

Note: Y = yes, N = no, “–" = not reported.

Appendix G. Mode of reconstruction.

Citation	Defect size <5 mm
	Number of layers	Materials	Surgicel strips	Gauze	Gelofoam	Fibrin glue/tissue sealant	Nasal pack	Sphenoid fat plug
Abdelazim et al, 2018	3	F, FL, NSF	Y	–	–	N	Y	Y
Bozkurt et al, 2020	2	Septum/MT and dural substitute inlay (FL/temporalis fascia/pericranium), and MP/pedicled vascular flap	Y	–	–	Y	Y	–
Castelnuovo et al, 2007	2	Dural substitute/FL, and B/cartilage	Y	–	–	–	–	Y

Abbreviations: AF, abdominal fat; AFa, abdominal fascia; B, bone; F, fat; Fa, fascia; FL, fascia lata, FMF, free mucosal flap; MT, middle turbinate; MP, mucoperiosteum; NSF, nasoseptal flap.

Appendix H. Mode of reconstruction.

Citation	Defect size >5 mm
	Number of layers	Materials	Surgicel strips	Gauze	Gelofoam	Fibrin glue/tissue sealant	Nasal pack	Sphenoid fat plug
Abdelazim et al, 2018	4	Fa, septal cartilage, FL, NSF	Y	–	–	N	N	Y
Bozkurt et al, 2020	3	Septum/MT and dural substitute inlay (FL/temporalis fascia/pericranium), and MP/pedicled vascular flap	Y	–	–	Y	Y	–
Castelnuovo et al, 2007	3	Dural substitute/FL, B/cartilage, MP	Y	–	–	Y	–	Y

Abbreviations: AF, abdominal fat; AFa, abdominal fascia; B, bone; F, fat; Fa, fascia; FL, fascia lata; FMF, free mucosal flap; MT, middle turbinate; MP, mucoperiosteum; NSF, nasoseptal flap.

Note: Y = yes, N = no, “–" = not reported.

Appendix I demonstrates the various adjunct treatments used for the patients in the included studies including CSF shunts, lumbar drains, and the use of ICP control medication such as furosemide and acetazolamide. Ten studies report the use of lumbar drains in some/all their patients, 3 studies reported the use of CSF shunts, 5 studies reported the use of acetazolamide, and 2 studies mentioned the use of furosemide.

Appendix I. Adjunct treatment.

Citation	Lumbar drain use	Use of CSF shunt in cohort	ICP control medication postoperatively in the cohort		Preoperative antibiotics
			Acetazolamide	Furosemide
Abdelazim et al, 2018	–	–	–	–	Y
Babu et al, 2019	Y	–	Y	–	–
Bozkurt et al, 2020	–	–	Y	N	Y
Campbell et al, 2016	Y	Y	Y	Y	–
Castelnuovo et al, 2007	N	–	–	–	–
El Tarabishi et al, 2016	Y	–	Y	–	Y
He and Zhen, 2022	–	–	–	–	Y
Huang et al, 2022	–	Y	–	–	–
Kirtane et al, 2012	–	–	–	–	Y
Lopatin et al, 2003	Y	Y	–	–	Y
Lopatin et al, 2011	Y	–	–	–	–
Maxfield et al, 2020	–	–	–	–	–
Muscatello et al, 2010	N	–	–	–	–
Nix et al, 2016	N	–	–	–	Y
Rathod et al, 2021	Y	–	–	–
Sreenath et al, 2021	Y	–	–	–	Y
Tabaee et al, 2010	Y	–	–	–	–
Tomazic and Stammberger, 2009	Y	–	–	–	–
Ulu et al, 2018	–	N	N	–	–
Xu et al, 2022	–	–	–	Y	–
Jiang et al, 2014	Y	Y	Y	–	Y
Zocchi et al, 2021	Y	–	–	–	–
Zoli et al, 2016	Y	Y	–	–	Y

Abbreviations: CSF, cerebrospinal fluid; ICP, intracranial pressure.

Note: Y = yes, N = No, “—” = not reported.

Risk of Bias in Studies

ROBINS-I was used to assess the risk of bias in the 24 studies included using an endonasal approach in this systematic review ( Fig. 2 ). The overall judgements from the study ROBINS-I assessment demonstrated 3 studies with moderate risk of bias, 7 studies with no information, and 13 studies with serious risk of bias.

Fig. 2.

ROBINS-I scores.

Results of Individual Studies

Table 4 demonstrates the success rate of the endoscopic endonasal and combined TONES endonasal repair success rates and complications. The two combined TONES and endonasal repair approaches had a 100% success rate. Seventeen of the endoscopic endonasal approaches had more than 90% success rate and 15 had a 100% success rate of the 24 endonasal approaches.

Table 4. Success rate and complications of CSF leak repair.

Citation	Success n (%)	Recurrence n (%)	Leak at different site	Complications
				Dry eye	Dry nasal mucosa/crusting	Facial hypoesthesia/paresthesia	Other
Abdelazim et al, 2018 33	6 (100)	0	–	–	–	–	–
Babu et al, 2019 34	3 (100)	0	–	–	3	–	–
Bozkurt et al, 2020 35	32 (100)	0	–	–	–	–
Campbell et al, 2016 36	11 (91.7)	1 (8.3)	0	–	–	–	–
Castelnuovo et al, 2007 37	10 (100)	0	0	–	–	–	Arachnoid cyst
El Tarabishi et al, 2016	10 (100)	0	0	No	No	–	No
He and Zhen, 2022 38	12 (100)	0	0	–	–	–	No
Huang et al, 2022 39	31 (100)	0	–			Yes	Dental/palatal numbness
Jolly et al, 2023 40	3(100)	0	–				–
Kirtane et al, 2012 47	13 (100)	0	0	–	–	–	–
Lopatin et al, 2003 48	4 (80)	1 (20)	–	–	–	–	–
Lopatin et al, 2011 49	17 (65.4)	9 (34.6)	–	–	–	–	–
Maxfield et al, 2020 41	15 (93.75)	1 (6.25)	–	–	–	–	–
Muscatello et al, 2010 42	1 (50)	1 (50)	–	–	–	–	–
Nix et al, 2016 51	4 (100)	0	–	–	–	–
Rathod et al, 2021 43	10 (100)	0		1 (10)	–	–
Sreenath et al, 2021 13	33 (100)	0	–	1 (3.2%)	–	3 temporary V2 anesthesia 1 prolonged (>6 months)
Tabaee et al, 2010 44	11 (84.62)	2 (15.38)	–	–	–	1 (7.69%)	Meningitis
Tomazic and Stammberger, 2009 45	3 (60)	2 (40)	–	–	–	1 numbness of right upper jaw (v2)	Meningitis, brain abscess
Ulu et al, 2018 46	8 (88.89)	1 (11.11)	–	–	–	–	Infection at abdominal graft donor area
Xu et al, 2022 53	14 (100)	0	–	0	–	–
Jiang et al, 2014	6 (87.5)	1 (12.5)	–	–	–	–	–
Zocchi et al, 2021	38 (100)	0	3%	–	–	–	–
Zoli et al, 2016 50	23 (100)	0	4.35%				Seizure
Lubbe et al, 2020 31	1 (100)	–	–	–	–	–	–
Ramakrishna et al, 2016 32	1 (100)	–	–	–	–	–	–

Abbreviation: CSF, cerebrospinal fluid.

Results of Syntheses

The success rate for the endoscopic endonasal approach was 95.238% ( n  = 336), whilst the combined TONES endonasal approach had a success rate of 100% ( n  = 2). Fig. 3 demonstrates that via a risk ratio there was no significant difference in the baseline characteristics in the LRS compared with repairs of other leak types. Similarly, Fig. 4 shows risk ratios where there was no significant difference in the baseline characteristics of the combined TONES and endonasal repair approach in patients with a CSF leak in LRS and other repair types.

Fig. 3.

Baseline characteristics of repair success risk ratio of leaks in the lateral recess of sphenoid sinus versus other leak types.

Fig. 4.

Baseline characteristics of combined transorbital and endonasal repair success risk ratio of leaks in the lateral recess of sphenoid sinus versus other leak types.

Discussion

Success Rate

The two combined TONES and endonasal approach studies both only had one relevant patient each with a 100% success rate.

The success rate for the endoscopic endonasal approach was found to be at a rate of 95.24% ( n  = 336). In total, 15 of the 24 studies included with an endoscopic endonasal approach had a 100% success rate of CSF leak repair and 17 had more than 90% success rate.

A systematic review of endoscopic endonasal repair for all CSF leak types and locations showed an overall success rate of 90.6% (95% CI: 80–92.5, p  = 0.035) which, although had a lower effect estimate than this systematic review, was statistically significant hence having a lower success rate than the current review. ²¹ More recently, a systematic review demonstrated an average of 9% failure in the anterior skull base for sCSF repairs, again a lower success rate compared to this current review. ⁷

Combined TONES and Endonasal Approach

Two studies were included which both utilized a combined TONES and endonasal approach, which were conducted by Lubbe et al ³¹ and Ramakrishna et al. ³² From the articles screened, the articles included did not have a pure TONEs approach as initially expected.

Ramakrishna and colleagues briefly described that a precaruncle combined TONES and endonasal approach was utilized for their patient with a Sternberg's canal in their retrospective review of TONEs procedures for various pathologies. ³²

On the other hand, Lubbe et al ³¹ provide a detailed description of the approach used which was a contralateral precaruncle and binostril endonasal mode of access that involved lamina papyracea dissection, posterior septectomy, breaching of the middle turbinate (MT), complete ethmoidectomy, anterior sinus wall sphenoidotomy extending to the Vidian canal laterally, and rostrum drilling between the sphenoid ostia. ³¹

The advantages of this approach were the avoidance of PPF dissection, direct surgical trajectory, increased working angles, and optimized visualization easier for a four-handed approach, and good cosmesis. ^{31 32}

In one systematic review, the TONES approach had a sample of 30 CSF leak patients and a success rate of 93%. ²⁰ Another systematic review concluded that the TONES approach is promising and feasible. ¹⁹

Transpterygoid Approach

The following studies had a transpterygoid approach of endoscopic endonasal repair for CSF leaks for 6 patients in Abdelazim et al, ³³ 3 in Babu et al, ³⁴ 32 in Bozkurt et al, ³⁵ 12 in Campbell et al, ³⁶ 5 in Castelnuovo et al, ³⁷ 9 in He and Zhen, ³⁸ 20 in Huang et al, ³⁹ 3 in Jolly et al, ⁴⁰ 13 in Maxfield et al, ⁴¹ 2 in Muscatello et al, ⁴² 10 in Rathod et al, ⁴³ 33 in Sreenath et al, ¹³ 3 in Tabaee et al, ⁴⁴ 3 in Tomazic and Stammberger, ⁴⁵ and 9 in Ulu et al. ⁴⁶ This was the most common approach and has been trusted for its ability to access the LRS.

Two studies reported the use of a binostril approach including the studies by Bozkurt et al, ³⁵ who preferred a two-surgeon and four-instrument approach, and Ulu et al. ⁴⁶ However, one study by Rathod et al ⁴³ reported the use of a uninostril approach. The rest of the studies did not report on the use of either uninostril or binostril approach.

Three studies report the use of a posterior septectomy with Babu et al ³⁴ doing so to improve instrumentation, Bozkurt et al ³⁵ mention this is done to allow for a “cross-court” angle where this needed to be used, and Ulu et al ⁴⁶ also rationalized that this step was to allow for free motion of endoscopic tools and form a single cavity. Tabaee et al ⁴⁴ report that the cartilaginous and vomer parts of the nasal septum were resected to improve access to the sphenoid cavity. On the other hand, Rathod et al ⁴³ explain that the use of the plasma ablation tool allowed them to avoid the need for a septectomy and hence a binostril approach as the instrument had multiple functions such as irrigation, suction, ablation, and coagulation hence evading the need to switch between multiple instruments.

The MT resection was reported in eight studies. Partial MT resection was reported in the following studies: Abdelazim et al, ³³ Castelnuovo et al, ³⁷ Kirtane et al, ⁴⁷ and Rathod et al, ^{35 43} and also reported the partial resection of the MT and medialization. MT resection without explicit detail on the completeness of the resection was reported by Bozkurt et al, ^{13 35} Sreenath et al, ¹³ and Ulu et al ⁴⁶ who report the resection of the MT or preparation of the MT as pedicled flap for skull base reconstruction in some cases. Tomazic and Stammberger ⁴⁵ reported that all turbinates were preserved. The MT was reported to be lateralized rather than resected in all cases in the study by Lopatin et al ⁴⁸ and Lopatin et al, ⁴⁹ as this was deemed to not be beneficial and the MT was returned to its medial position for better access and visualization when needed.

The superior turbinate resection was mentioned by Ulu et al, ⁴⁶ who state that this is conducted when required without further explanation. Rathod et al ⁴³ mention the lateralization of the superior turbinate rather the resection. As mentioned previously, Tomazic and Stammberger ⁴⁵ preserved all the turbinates.

The inferior turbinate was only reported to be resected in one study, which was by Abdelazim et al, ³³ which noted that the posterior half of the inferior turbinate was resected.

An uncinectomy was performed as the first step in the studies by He and Zhen, ³⁸ Castelnuovo et al, ³⁷ who reported the use of a subtotal uncinectomy, Kirtane et al, ⁴⁷ and Rathod et al. ⁴³

A middle meatal approach to perform a maxillary antrostomy was described in the studies by Abdelazim et al ³³ which was wide, Castelnuovo et al ³⁷ and extends posteriorly to the pterygoid process, Huang et al, ³⁹ Rathod et al. ⁴³ He and Zhen ³⁸ and Sreenath et al ¹³ report the use of wide antrostomy although the meatus through which this was conducted was not clearly stated as well as Tabaee et al. ⁴⁴ Zoli et al ⁵⁰ who also report the technique of a wide middle antrostomy although the name of the endoscopic technique used was not explicitly mentioned. Similarly, Nix et al ⁵¹ when contacted reported the use of an antrostomy on their LRS patients but the endoscopic endonasal approach was not explicitly confirmed.

Complete ethmoidectomy of the anterior and posterior ethmoid cells was conducted in the studies by Castelnuovo et al, ³⁷ He and Zhen, ³⁸ Huang et al, ³⁹ Kirtane et al, ⁴⁷ Sreenath et al, ¹³ Tabaee et al, ⁴⁴ and Tomazic and Stammberger. ⁴⁵

Sphenoidotomy of the anterior sphenoid sinus wall was reported in the studies by Abdelazim et al, ³³ Nix et al, ⁵¹ Rathod et al, ⁴³ Sreenath et al, ¹³ and Ulu et al, ⁴⁶ who reported the use of a drill. Sphenoidotomy was reported to have been performed in the studies by Castelnuovo et al, ³⁷ in the standard way, Tomazic and Stammberger, ⁴⁵ and Huang et al, ³⁹ a wide sphenoidotomy was reported by Tabaee et al ⁴⁴ and He and Zhen, ³⁸ which was lateral and beyond the V-R line. A sphenoidotomy was also reported to have been conducted by Zoli et al, ⁵⁰ who did not confirm the nature of their endoscopic endonasal approach.

The sphenoid rostrum was reported to be resected in the studies by Bozkurt et al. ³⁵ The mucosa overlying the sphenoid rostrum was reflected in the study by Sreenath et al. ¹³

The posterior wall of the maxillary sinus was removed to expose the PPF in the studies by Abdelazim et al ³³ who drilled the medial half of the posterior maxillary sinus, partially by Castelnuovo et al, ³⁷ He and Zhen, ³⁸ Huang et al, ³⁹ the posteromedial maxillary sinus wall by Sreenath et al, ¹³ Tabaee et al ⁴⁴ who used a drill, (Tomazic and Stammberger, 2009). ⁴⁵

The sphenopalatine artery encountered following the exposure of the PPF in the studies by Abdelazim et al ³³ was cauterized with bipolar diathermy, retracted inferiorly by Sreenath et al, ¹³ mobilized and either cauterized or ligated by Tabaee et al, ⁴⁴ dissected and coagulated by Huang et al, ³⁹ and Tomazic and Stammberger ⁴⁵ controlled the bleeding from the sphenopalatine artery by electrocautery.

The base/wedge of the pterygoid process of the sphenoid bone was drilled to expose the LRS in the studies by Abdelazim et al, ³³ Bozkurt et al ³⁵ who adapt the extent of bone removal based on the size of the defect and avoid removal of the lateral and inferior aspects of the pterygoid bone, Castelnuovo et al, ³⁷ He and Zhen ³⁸ also drilled the pterygoid base and extended this laterally, Huang et al ³⁹ drilled, Rathod et al ⁴³ also drilled the base and reported sacrificing the Vidian nerve, Sreenath et al ¹³ fully expose the superomedial aspect using a drill and the Vidian V2 nerve is identified and spared, Tabaee et al ⁴⁴ who drilled the pterygoid process, the base was either drilled or resected with a Kerrison punch by Tomazic and Stammberger. ⁴⁵

Reconstruction of the Skull Base

The method of reconstruction of the skull base varied greatly amongst the studies ( as outlined in Appendices G H I ) as well as between patients in aspects such as the technique, materials, number of layers used, the order in which different materials were placed, and adhesives to secure the reconstruction. Three studies have highlighted that their reconstruction method depends on the size of the skull base defect such as by Abdelazim et al, who used fat, fascia lata, and a nasoseptal flap in defects less than 5 mm, whereas they used fascia, septal cartilage, fascia lata, and a nasoseptal in defects >5 mm. ³³ Bozkurt et al describe a triple-layer technique (intradural–epidural–extracranial) for large defects and a double-layer technique for smaller defects (epidural–extracranial) using various materials. ³⁵ Similarly, Castelnuovo et al described using a triple-layer technique for large defects where layers are placed intradurally, epidurally, and an on-lay and a double-layer technique for small defects. ³⁷

The remaining studies did not specify whether the size of the skull base defect affected the technique opted for reconstruction. The materials used by these studies varied and included the use of fat as mentioned in 14 studies, fascia in 13 studies, free mucosal graft in 4 studies, the nasoseptal flap in 7 studies, MT flap in 5 studies, cartilage in 4 studies, bone in 5 studies, and artificial/substitute dura in 4 studies as per Appendix G .

The decision to obliterate the sphenoid sinus cavity with a fat plug was also diverse with six studies ^{33 37 43 45 47 52} reporting the use of this technique. Abdelazim et al described it using a fat plug in order to stabilize the “sandwich graft.” ³³ Castelnuovo et al also filled the sphenoid sinus with abdominal fat although they mentioned the challenge of having a small enough sphenoidotomy to allow the anterior sphenoidal wall to be supportive but also big enough to allow for accurate submucoperiosteal dissection. Castelnuovo et al reported a rare complication of an arachnoid cyst development in one of the patients which they attribute to the sphenoid sinus fat obliteration hence subsequently only using a multilayered skull base plasty. ³⁷ Jolly et al also described it using a fat graft plug in some of their patients although they explain that multilayered approach is more robust. ⁵² Kirtane et al describe it using a bath plug technique to obliterate the sphenoid sinus cavity where the fat is packed intracranially with protrusion of the fat extracranially. ⁴⁷ Rathod et al described plugging of the sphenoid sinus with fat wrapped in surgical after which they placed customized bone graft. ⁴³ Tomazic and Stammberger used a modified “bath-plug” technique using abdominal fat in three of their patients to act as a primary seal and to reduce the pressure exerted on the overlying fascia lata. ⁴⁵

In addition, supportive materials and/or adhesives were used to seal the reconstructed skull base defect which included the use of surgical, gauze, gelo-foam, fibrin-glue, or other tissue sealants which was heterogenous amongst the studies.

Complications

Complications reported for patients who underwent the endonasal endoscopic route of repair included dry eye, dry nasal mucosa, facial paresthesia, or hypoesthesia, palatal numbness, meningitis, arachnoid cysts, brain abscess, seizure, and infection at abdominal graft site highlighted by 11 studies. ^{13 33 34 35 37 39 43 44 45 50 53}

There were no reported complications for the combined TONEs and endonasal approach. ^{31 32}

Limitations

This study was not without its limitations. The screening process which used the Rayyan software ²³ was not always accurate in detecting duplicates which in some cases had to be manually removed. This was further complicated because the duplicates could only be deleted once by the first reviewer and hence there was no double-checking.

This systematic review excluded studies with a foreign language hence introducing the probability of language bias which is a type of publication bias. ^{54 55}

The quality of systematic reviews is dependent on the quality of the included studies. The results of the ROBINS-I assessment showed that most of the studies had a serious risk of confounding due to uncontrolled co-interventions such as lumbar puncture, CSF shunt use, and acetazolamide due to variable use within the sample. Most of the studies included were small retrospective studies and case series making them at high risk of bias.

There are various nuances to the repair of the LRS defect such as the method of reconstruction of the skull base and adjunct treatments which could not be explored further since not all the studies used these consistently across their patients and hence any confounding effect may vary on a patient-by-patient basis.

One of the main problems with the studies included was the lack of reporting of negative findings. This is especially relevant for potential confounding factors such as the use of lumbar drains, CSF shunts, and acetazolamide usage.

Reporting of the LRS CSF leaks in the literature is often under the umbrella of anterior skull base CSF leaks or anterior skull base leaks so data can be hard to extract. Moreover, the extent of LRS incidence could potentially be underestimated.

This study could have been improved by potentially obtaining the individual participant data from all the studies included to improve the consistency, reporting, and heterogeneity. The transorbital approach showed promising results and, in the future, larger trials need to be conducted to determine a more representable success rate of the TONES approach or rather the combined endoscopic endonasal and TONES approach on LRS CSF leak repairs. Moreover, a focus on all CSF leaks from the LRS regardless of etiology could have provided a larger sample size and potentially statistically significant results as a systematic review on endoscopic LRS leak repairs has not been conducted regardless of etiology.

Conclusion

In conclusion, the combined TONES and endonasal approach showed a success rate of 100%. However, this only included two studies both with only one patient. The endoscopic endonasal approach demonstrated a success rate of 95.238% ( n  = 336) whilst the combined TONES endonasal approach had a success rate of 100% ( n  = 2). This is the first systematic review to summarize data regarding endoscopic repairs of LRS CSF leaks. Although both studies demonstrated good success rates, large, randomized control trials comparing the two surgical approaches are needed.