A systematic review and meta-analysis of the hip capsule innervation and its clinical implications

Detailed understanding of the innervation of the hip capsule (HC) helps inform surgeons’ and anaesthetists’ clinical practice. Post-interventional pain following radiofrequency nerve ablation (RFA) and dislocation following total hip arthroplasty (THA) remain poorly understood, highlighting the need for more knowledge on the topic. This systematic review and meta-analysis focuses on gross anatomical studies investigating HC innervation. The main outcomes were defined as the prevalence, course, density and distribution of the nerves innervating the HC and changes according to demographic variables. HC innervation is highly variable; its primary nerve supply seems to be from the nerve to quadratus femoris and obturator nerve. Many articular branches originated from muscular branches of the lumbosacral plexus. It remains unclear whether demographic or anthropometric variables may help predict potential differences in HC innervation. Consequently, primary targets for RFA should be the anterior inferomedial aspect of the HC. For THA performed on non-risk patients, the posterior approach with capsular repair appears to be most appropriate with the lowest risk of articular nerve damage. Care should also be taken to avoid damaging vessels and muscles of the hip joint. Further investigation is required to form a coherent map of HC innervation, utilizing combined gross and histological investigation.

Aims and objectives. This systematic review and meta-analysis aims to provide an encompassing synthesis of contemporary literature on the HC's innervation from gross anatomical studies. This includes additional analyses that have not been performed to date 4,46 : a meta-analysis of the prevalence of nerves innervating the HC, discussion of the course, density, and distribution of articular nerves, and changes according to demographic, arthropometric variables and underlying pathology.

Methods
Selection of studies. A systematic review of the literature on the HC's innervation was performed to identify peer-reviewed articles published until October 2020 according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines 47 . Keywords relating to the HC's innervation in gross anatomical studies (macroscopic studies) were searched using the following online databases: Amed (1985 <), Embase (1947 <), and Medline (1949 <) via Ovid, also PubMed, ScienceDirect, Scopus and Web of Science (all approximately 1900 <). Search terms are outlined in Table 1. Wiley and SpringerLink electronic databases were also screened ensure articles published in relevant anatomical journals from a wider date range were searched. The reference lists of previous literature reviews 4,46 and the articles identified in this search were screened. To locate relevant articles that cite the included articles Web of Science Core Collection and ResearchGate.net were screened.
When the databases included search options, these were configured to remove non-primary research articles and duplicates. Articles were inputted into an Excel spreadsheet to remove duplicates, then were screened by one author (J.T.) by their title and abstract, then by their full text. All screened articles that included macroscopic study of the HC or surrounding structures (i.e., musculature, bones and vessels) from either surgical or cadaveric studies were reviewed in their full text, to ensure all incidental findings on the HC's innervation were included. The following exclusion criteria were applied during the literature search: -Animal studies -Non-peer reviewed articles -Non-primary research articles -Non-macroscopic studies Table 1. Search terms used in search strategy for this systematic review. Using the Boolean search method, the search lines were combined with the word ' AND' . *Truncation of a word, this searches for words that begin with the word preceding the asterisk.
Critique of the literature. Papers were critiqued by one researcher (J.T.) according to the several criteria adapted from the AQUA 48 and QUACS 49 scales to assess the limitations to the conclusions made in cadaveric studies. This included assessing aspects of the methodology and potential variables affecting the reliability of the results presented, including determining if data was prospectively or retrospectively collected and if magnification was used. Additionally, the replicability of the study, data presentation and consideration of limitations were assessed.
Analyses. The HC's innervation was evaluated by collating information on the course and density of innervation and information on prevalence and distribution underwent analytical processing.
Prevalence. Meta-analysis was performed to determine the pooled prevalence estimate (PPE) of each nerve innervating the HC. Analysis of the effect size was determined using a random effects model computed in Microsoft Excel 2016 with the MetaXL add-on version 5.3 (EpiGear International Pty Ltd, Queensland, Australia). The random effect model was selected as information from cadaveric samples generally have high heterogeneity 50 . By applying a random effects model this assumes a normal distribution within the sample, that the effects are estimated and different studies are not identical 51 . The PPE analysis included a double arcsine prevalence transformation, with a continuity correction of 0.5 and 95% confidence intervals (CI), in order to address the issues of variance instability 52 . The prevalence was obtained by multiplying the computed PPE by 100 to form a percentage. Several inclusion criteria were required for the PPE to be assessed: -Studies with over 10 hip joints were investigated.
-The prevalence of a certain nerve innervating the HC was calculated when there were more than two studies investigating that nerve. More than 10 studies on the topic were regarded as a substantial number to form valid conclusions from the literature. -Only prospective studies were included, retrospective studies were excluded as there is a risk of selection bias.
The variance between studies included in the meta-analysis was studied using the I 2 statistic (which assesses the amount of heterogeneity between studies). The following standard percentages were used to assess variance: < 40% indicates low heterogeneity, 30-60% suggests moderate, 50-90% implies substantial and 75-100% may be considerable heterogeneity 51 . Publication bias was analysed by visually assessing the funnel plots, doi plots and Luis-Furuva-Kanamori (LFK) index 53 . No publication bias was assumed from symmetry in the LFK index and in the absence of symmetry publication bias was assumed. An LFK index within ± 1 from 0 indicates no asymmetry, out of ± 1 but within ± 2 as minor asymmetry, and > ± 2 is to mean major asymmetry 53 . Distribution. The nerve distribution was assessed by collating information from the literature regarding which region of the HC was supplied by each nerve. A heat map was created for each nerve, to represent the areas of agreement of the distribution of innervation between studies. The opacity from black to white displayed in the heat map directly related to the percentage of agreement between studies, where black represents 100% agreement and white represents no agreement.

Results
Few papers on the hip capsule innervation were identified from the extensive search. In total, 36 papers were identified as eligible for inclusion from 49,555 records overall (see Fig. 1). These were published between 1857 and 2020. Twenty-six were available in English, five were published in French [54][55][56][57][58] , three in German 59-61 , one in Russian 62 and one in the Czech language 63 . Translations were performed by native speakers. Four studies potentially include duplicated results 60,[63][64][65] . However, as they discuss different aspects of the HC innervation, they were included in the descriptive analysis but duplicates were excluded from the meta-analysis.
Varying methodologies were employed to study hip capsule innervation. Despite all the studies including dissection of cadaveric material, the methodologies employed differed substantially and several papers provided limited information making it difficult to assess the quality of the studies included (see Table 2). Numerous authors acknowledge that factors may have affected the results presented: post-mortem changes 66,67 , their dissection [66][67][68][69] , embalming 70 , the limited sample size 65,67,71 and the morphology of the articular nerves, which are noted to be thin and may have been damaged 56,57,62,72 . It is unclear if spatial distortion or excision of nerves during dissection affected the results [66][67][68][69] . Several authors partially accounted for potential limitations by excluding cadavers from their study based on their observations or records of the following: pathology 67 www.nature.com/scientificreports/ which some authors thought this was essential when investigating this region 77 . Some studies used histological stains of capsular tissue alongside their dissection 63,79 . However, they did not clearly present their findings 63,79 . Several factors prevent a full assessment of the quality of studies. Many studies omitted to explain their methodology [54][55][56][57][58][59]61,63,[84][85][86]88,89 or failed to provide sufficient information that would enable it to be replicated 60,65,75,77,78,87 . While other studies were retrospective observational analyses [54][55][56][57][58]61,74,81,86,88 . Furthermore, some papers did not include images of their dissection [54][55][56][57][58][59][60][61][62]69,72,73,75,[77][78][79][86][87][88][89] . No paper discussed the degree of degradation of the tissue or time from death to embalming as a measure of quality control. The studies included in this review have limitations as expected from this study type, but many authors are aware of these when drawing their conclusions.

Summary of descriptive data.
Limited information on the sample characteristics were presented in the literature. The study characteristics are presented in Table 3. All studies included in this review investigated cadaveric material. Studies included samples sizes ranging between 1-1000 (mean 92) hip joints, though seven studies did not specify the number of hip joints studied 54,55,58,59,61,62,88 . Available data in the literature was limited, as only eight studies indicated the age of the cadavers in their experiment 64,66,67,70,74,[82][83][84] , five studies detailed the ethnicity 64,65,71,75,79 and one study noted the average weight 82 . No other demographic or anthropometric variables were described in the studies included.  22 . The dotted line represents the articles located from the search of the databases undergoing forwards and backwards chain sampling to located further articles.    48,49 . N no, NS not stated, P prospective, R retrospective, U unclear, Y yes.  72 . One study explicitly noted an absence of articular branches originating from the pudendal nerve 60 , however, it is important to note that their study does not confirm its absence. Other studies suggested that sympathetic fibres supply the HC, but did not prove this histologically 58,87 .
The roots of the nerves providing articular branches to the HC have only been reported for the obturator nerve, AON and AFN and are highly variable. In one study of Ethiopian cadavers, the obturator nerve originated primarily from L2,3,4 and in some cases from L3,4 71 . When the AON was present, it originated from either L2-3 80 80,85 or from the femoral nerve 80 . It is unclear which nerve root the AON most commonly originates from 80,85 . When the AFN was present its nerve roots were from L2,3 63 or L2,3,4 75 and therefore was more predictable.
Information on the course of hip capsule's articular branches is sparse. The course of the HC's articular branches was described by few authors and lack a uniform description of their course in relation to the muscles, osseous landmarks, or vessels. Articular branches from the femoral 58,76,77 , obturator 58 and superior gluteal 83 nerves were described as coursing with vessels. Specifically, these articular branches coursed deep to the femoral artery 58 , also along the obturator 58 , the lateral circumflex 58 and posterior circumflex arteries 58,63 . However, this has not been assessed systematically. Based on the literature, articular branches appear to cross all regions of the HC 57,58,64,78,83,87 . Some branches have a long course and do not terminate proximal to where the main trunk overlies the capsule, for example some articular branches from the obturator and femoral nerve coursed posteriorly 55,58,63,76 . To date, no study has described the course of the AON, AFN, sciatic or inferior gluteal nerves in detail. Articular branches appear to course indirectly across the anterior and posterior HC to their point of termination, but the current literature remains unclear with regards to the specific location of these nerves.
Hip capsule innervation from the obturator nerve and nerve to quadratus femoris is most prevalent. The NQF appears to be the most frequent nerve to innervate the HC, followed by the obturator then the femoral nerve ( Table 4). The AON appears to be present in 15% [95% CI (0.09-0.22), I 2 = 73%] of the specimens studied in this review 60,67,68,70,77,79,80,83,85 . When the AON is present the PPE that it innervates the HC is 85% [95% CI (0.58-1.00), I 2 = 86%] 60,67,68,70,77,79,80,83,85 . The prevalence of the AON innervating the HC in the specimens included in the review is presented in Table 4. Innervation of the HC from the AON and obturator nerve appears to be unrelated and highly variable 55,60,67,79,80,83 . The reported PPE of the obturator and femoral nerves are limited, as some studies which reported absence 55 and presence of articular branches 58,59,61 did not indicate the number of samples studied, so could not be included in the PPE (Figs. 2 and 3).
Considerable heterogeneity was present between studies, except for the AFN and NQF, which had low heterogeneity and the inferior gluteal nerve, which has low-moderate heterogeneity, as indicated by the high I 2 statistic. The AON also had substantial heterogeneity. High publication bias was present within the meta-analysis for the Table 4. Prevalence of nerves innervating the hip capsule. The LFK index of the accessory femoral nerve and nerve to quadratus femoris could not be calculated as only two studies were included in their meta-analysis of prevalence. CI confidence interval, LFK index Luis-Furuva-Kanamori index, NS not stated. www.nature.com/scientificreports/ obturator, sciatic and inferior gluteal nerves as represented by the major asymmetry in the LFK index. As a result of considerable heterogeneity and LFK index, the confidence intervals of the prevalence should be considered. The confidence intervals show that there is minimal variation in the PPE for the AON, NQF and inferior gluteal nerves, but these are based on few studies with small sample sizes. It is unclear if these reported PPE's are more reliable. Furthermore, PPE specific innervation patterns across the whole HC could not be determined as limited data was available. The NQF and obturator nerves appear to be most prevalent, but the reported figures may not be representative of the wider population.
Most articular branches to hip capsule are located anteriorly. A variable number of articular branches innervate the HC from each nerve ( Table 5). These articular branches were described as thin, small and delicate structures 54,57,58,62,72,[77][78][79]82,87 , which measure up to 17 μm in diameter 78 . Further detailed information was provided for the obturator nerve but no other nerve, its articular branches span the HC in a 7-38 mm band, which appears to be wider medially 69 . Based on the density of innervation, the HC appears to receive greater innervation from nerves located anteriorly.

All regions of the hip capsule receive innervation from the lumbosacral plexus. More informa-
tion is available on the distribution of the obturator and femoral nerve than other nerves that innervate the HC. Both of these nerves supply the anterior hip capsule (AHC). Distribution of the branches from the femoral nerve appears to be nearly homogenous across the AHC in comparison to other nerves (Fig. 4). Strong general consensus indicates that the obturator nerve and AON innervate the anterior medial HC, with the articular branches of the obturator nerve commonly present over the radiographic teardrop of the inferomedial acetabulum 67 . General consensus indicates that the AFN supplies the AHC 56,76,77 , but no study to date provides more detailed  www.nature.com/scientificreports/ information on the distribution of the AFN. Furthermore, one study which studied the HC's innervation in a clockface orientation noted an absence of innervation in the anterior superior aspect of the HC at around the one o' clock to half two position 83 . However, other studies did not study the HC in this detail, nor did they report an absence of innervation in this region.
The posterior capsule appears to be supplied by the superior gluteal, inferior gluteal, sciatic nerves, NQF and also the femoral nerve 58,87 in a few cases. More specifically, general agreement indicates that the sciatic nerve supplies the posterosuperior capsule 55,61,76,83 and the NQF innervates the posterior inferolateral portion [76][77][78]83,87 . The posteroinferior portion of the HC seems to be supplied by the inferior gluteal nerve 83 , however, this is based on a couple of samples in one relatively small-sample anatomical study (n = 20). Current literature remains unclear as to whether there are different innervation patterns and if any variables may predict these patterns.  www.nature.com/scientificreports/ Information on the hip capsule's innervation according to demographic, anthropometric variables and pathology is lacking. To date, no studies have made an attempt to quantify differences in the HC's innervation according to demographic or anthropometric variables. Nor have these studies stated if individuals with pathology were included in their sample. Furthermore, the details of demographic and anthropometric variables of included cadavers were not routinely collected or at least not reported in detail. Based on the limited information, no differences appear to be present as a result of age 58,60,63 , however, no study has made attempt to compare this directly. A few studies investigated both sides of the HC 68,70,71,73,80,[82][83][84][85]89 , one of which noted individual differences in the innervation pattern between sides, but omitted to compare this statistically 57 . It remains unclear, if there are differences in prevalence, density, distribution of innervation according to ethnicity, age, sex or other variables.

Discussion
The primary aim of this systematic review and meta-analysis was to provide an encompassing synthesis on the macroscopic studies investigating the HC's innervation. Several key findings have been identified from this review, most notably that specific information on the innervation of the HC is lacking. There was high heterogeneity between studies, therefore the presented PPE and distribution patterns may not be applicable to all individuals.
Articular branches to the hip capsule originate from many nerves of the lumbosacral plexus. Strong evidence indicates that the HC is supplied by a variety of branches of the lumbosacral plexus, primarily by articular branches from the NQF and obturator nerve. There is however considerable heterogeneity between studies. The PPE's have many limitations as they are formed from studies with different methodologies and some are based on only a few studies with small sample sizes particularly for the AON, NQF and inferior gluteal nerve. Consequently, it seems likely that studies employing larger sample sizes or by different examiners may produce different results in the future. On the other hand, numerous articular branches were noted making it unlikely that dissections have been consistently misinterpreted.
Other nerves including the femoral, superior gluteal and sciatic may contribute to HC's innervation in some individuals. However, this is uncertain as the root of these articular branches were not traced proximally and histological evaluation has not been performed to date. Rarely, the HC is supplied by other nerves of the lumbosacral plexus. Primarily, this highlights that injuries to the NQF and obturator nerve should be crucially avoided during THA and these nerves should be targeted during RFA in order to reduce pain. Clinicians should consider that articular branches from other nerves provide substantial nerve supply to the HC, therefore patients may continue to have some pain following RFA.
The inferior and medial regions of the posterior hip capsule seem the most suitable areas for capsular incision during THA, in order to reduce the risk of injuring articular nerves. The anterior inferomedial region appears to have the greatest abundance of nerve supply, but all regions of the HC appear to be innervated. These findings do not completely corroborate with findings from a recent review of microscopic studies 44 , which reported that the superolateral region had the majority of mechanoreceptors. This indicates that the nerve may not terminate as type I-III mechanoreceptors in the same region as where the nerve enters the HC at the gross level, i.e., macroscopically visible nerves may have an intra-ligamentous course within the HC. When considering the macroscopic and microscopic innervation of the HC, the inferior and medial regions of the posterior HC appear to receive the least innervation. Risk of damaging articular nerves may therefore be lowest when incising these regions during THA. THA via the posterior approach with capsular repair may be most appropriate for the non-risk patient in order to reduce the risk of disrupting neural feedback loops, which could contribute to lower post-operative hip joint dislocation rates. Alternative approaches should continue to be considered for at-risk patients, i.e. those of older age, with a BMI of > 30 and with mobility or neuromuscular pathology 38 . The recent literature is contradictory with regards to which surgical approach produces the highest rate of instability post-operatively [36][37][38][39][40][41] . Therefore, implementing a combination of best practices which reduce the risk of dislocation may be more important when planning surgery than the surgical approach alone, including repairing the HC 8-10 to restore biomechanical stability, preserving musculature and other surrounding soft tissue 8 , as well as correct joint alignment 9,35 .
The anterior approach during THA appears most appropriate to reduce pain post-operatively, as the current literature reports lower patient pain scores following this approach 90 . The literature on microscopic innervation of the HC provides limited evidence with regards to the distribution of FNEs 44 . However, the capsule clearly appears to have a role in nociception and more robust immunohistochemical experiments mapping the capsule will allow more solid conclusions and recommendations to be formed.
Furthermore, further research is required to determine the course of nerves towards and across the surface of the HC, as well as the differences in the number of nerves and terminal endings present in the surrounding tissues. Recent research highlights that the hip capsule may have the least innervation in comparison to the different tissue layers of the hip joint, including the skin, subcutaneous fat, muscle and fascia 91 . It may therefore be more important to consider the innervation of all of the soft tissue layers, variation between individuals and feasibility to avoid large nerve bundles when planning surgery. A comprehensive map of the nerves and end-terminals within all tissue layers surrounding the hip joint is lacking to date. Additionally, it is important to avoid vessel damage during THA as in doing so it is likely clinicians will also avoid damaging nerves, this supports previous microscopic findings 44  www.nature.com/scientificreports/ inferomedial region appears to be a suitable location. However, further study is required that considers the cumulative density and distribution of articular branches in each region, as well as the modality of these nerves.
The origin of the capsular nerve branches resembles the innervation pattern of the hip muscles. Many of the articular branches are derived from branches to musculature, rather than the common trunk of the nerve. These findings align with Hilton's law 45 and suggest that these muscles may contribute to proprioceptive control of the joint in response to changes in tension of the HC. To date, no study has recorded the origin of articular nerves and their relationship to musculature that acts on the joint, as a result the risk of nerve damage during THA and RFA is unknown. The literature also highlights that during pelvic surgery near the obturator canal there is considerable risk that both the sensory supply to the HC and the motor innervation to the muscles surrounding the joint could be damaged. Nerve damage is unlikely during RFA, as sensory testing is commonly performed prior to lesioning and some practitioners check for absence of motor stimulation 4 . Muscular weakness following RFA has not been commonly reported, however, this may be as muscular strength before and after RFA are not standardly recorded 4 . Albeit rare, these complications have been noted following THA due to insufficient muscular fixation 34 , muscular or nerve damage 92,93 . This can result in dislocation 34 , falls, numbness, muscle palsy or pain 92 . The rates of muscular weakness appear to differ according to the approach used 30,32 or the experience of the surgeon 94,95 . It remains unclear which approach has the lowest risk of functional impairment 31,33 . Nonetheless, most of these nerve palsies appear to recover within 1-3 years 92,93 and muscular weakness seems to improve with physiotherapy 30 or further surgery 96 . Considering this, clinicians should take care when performing RFA and surgical procedures to avoid damaging nerves, periarticular soft tissue and muscles in order to preserve potential proprioceptive and nociceptive functioning. Additionally, where possible, surgeons should repair the HC and overlying muscles. Furthermore, increased accuracy during these procedures may be possible with intra-procedural imaging, such as polymetric imaging during THA 97 or visualisation of RFA lesions using near field ultrasound 98 . Additionally, it may be beneficial to measure motor strength of the muscles related to the targeted nerves both pre and post intervention in order to identify individuals at increased risk of dislocation as a result of potential damage to neural feedback loops. It is important to gain greater understanding of the innervation of the HC and associated muscles to aid clinicians in preserving function of the hip joint. Further research should employ techniques that minimise the risk of spatial distortion or damage to the articular branches to increase the accuracy of the description of these branches.

Differences in innervation of the hip capsule appear to be present between individuals.
No information is available with regard to the relationship between demographic and anthropometric variables and their effect on HC innervation, but differences have been observed between individuals. This is partially due to incomplete descriptions of the sample characteristics limiting possible comparison between studies. The high heterogeneity reported between studies highlights that subgroup analysis should be performed in future research. It is important to understand the differences in innervation according to demographic variables as this may highlight that an individualized approach to surgery may be required. Male patients are at greater risk of revision following THA due to dislocation, particularly when the posterolateral approach is employed 42 . Whereas, age of the patient did not appear to alter the risk of revision due to dislocation 42 . Current systems available to adapt THA to address the sex-specific differences in boney morphology, such as the femoral head size and neck offset are thought to be sufficient to produce satisfactory outcomes and reduce this risk 99 . However, procedural changes according other demographic variables, such as ethnicity may be required. There are differences in the arrangement of the lumbar plexus according to ethnicity 100 . These alterations to innervation patterns may also relate to HC innervation, but this has not been studied to date. Differences have also been noted in microscopic innervation between individuals according to age 101 and pathology 1 . Specifically, previous microscopic investigation showed a decreased density of mechanoreceptors in the HC of individuals with osteoarthritis compared to those with no known pathology 1 . However, it is remains unknown if these differences may be observed macroscopically, as it has not been studied to date. Further studies should include larger sample sizes in order to substantiate the differences in HC innervation according to demographic, anthropometric variables or pathology (e.g. sex, side, age, pelvic dimensions). This information may increase understanding why there is conflicting evidence on which approach produces lower dislocation rates as different approaches may be required during THA or RFA in specific patient populations. Articular nerve density, distribution and prevalence of capsular branches seem highly dependent on the quality of the studies in current literature. The reported data on HC innervation are at risk of invalid measurement, as these are highly dependent on method employed to visualise these nerves. Numerous biases are present within the available literature that will affect the given summary of results in this review and meta-analysis. Firstly, a substantial number of studies were retrospective studies and case reports, which therefore introduces the risk of selection bias, detection and publication bias. In order to reduce the effect of this bias, these studies were excluded from the meta-analysis. Secondly, no study stated that more than one author reviewed the structures identified. Thirdly, it is possible that the articular branches may have been misidentified, as limited information is available on the methods employed, which prevents the quality of the work from being assessed. Fourthly, the PPE reported is based on a small number of studies. It is possible that the reported prevalence may differ in the wider population.
Under-reporting of HC innervation may have occurred as the small articular branches may be damaged easily during dissection 69 . To decrease the risk of under-reporting, several authors used magnification to aid in identifying these structures, one of which deemed this necessary 77 . Certain embalming methods reduce the risk of articular branches being damaged as they make structures more easily identifiable and forms the surrounding www.nature.com/scientificreports/ tissue in a suitable consistency for blunt dissection. Ethanol-based 102 , phenoxyethanol or Thiel 103 embalmed cadavers may be more suitable for performing future macroscopic studies of nerves, ligaments and muscles in terms of haptics, compared to the stiffer tissues resulting from formaldehyde and ethanol embalming 104 . These specific embalming methods have not been employed to date when assessing HC innervation, this is likely because of the known regional differences in anatomical departments which favor one special embalming technique. Authors should consider the fixation of cadavers used in future work thoroughly as Thiel embalming does not permit for histologic assessment of small nerves 104 . Inaccurate reporting of HC innervation may be present in the current literature, as to date no study has used dyes, nor confirmed their findings with histological analysis. It may be possible that structures were misidentified. There are inconsistencies in the current literature with regards to the methodologies employed, for example in the use of magnification and embalming technique used. This highlights many areas for potential errors in the reporting of results. Additionally, identifying these inconsistencies allows future work to be more robust. However, the quality of anatomical dissection studies is often criticised and many of the limitations are unavoidable and a product of dissection studies. Supplementary investigations, such as histology, immunohistochemistry, or Sihler's staining which aid in the identification of nerves would be extremely beneficial and would add credibility to the findings on HC innervation.
Limitations. Difficulties were presented by performing meta-analysis of gross anatomical studies, as methodologies and sample sizes were variable, few studies could be compared, and several factors increased the risk of invalid reporting. This is clearly represented by the high heterogeneity reported between the studies included.
It is possible that suitable articles may have been missed through the application of exclusion criteria, search filters, limited keywords searches and articles whose full text could not be located. However, to minimise the risk of bias and underreporting of the literature, articles in all languages were included if accessible to the authors. Some bias may persist as unpublished work was excluded, however, this was recommended previously 50 , as these may have not undergone a rigorous peer review. Finally, location bias may affect the results as articles in less accessible journals may have been missed by our search criteria. To address this issue, a variety of databases was searched.

Conclusion
The gross anatomical innervation of the HC seems highly variable between individuals. Based on the current literature, the authors recommend that RFA procedures should continue to target the anterior inferomedial aspect of the HC but remain aware that other nerves are spared during this procedure and this may not completely resolve pain. In order to spare articular nerves and potential proprioceptive functioning during THA, the posterior approach appears to be most appropriate for non-risk patients. The HC should be repaired following THA as this is proven to lower dislocation risk and may aid in restoring HC innervation. Surgeons should continue to consider other factors which reduce dislocation risk when planning surgery. A coherent map of HC innervation is required to be able to improve the patient's prognosis following RFA and THA. This research should use additional investigations to support dissection findings, such as microscopy.

Data availability
The articles included in this systematic review and meta-analysis may be found using the search strategy outlined in the methods section.