The Reconstructive Pelvic Anatomy classification system

Abstract

Aims

Pelvic reconstruction in the setting of significant bone loss has become increasingly complex, with advancements such as custom 3D-printed prostheses now used in both large oncological resections and revision total hip arthroplasty (rTHA) with massive bone loss. Surgeons must consider how to best reconstruct the pelvis to optimize patient outcomes. The Reconstructive Pelvic Anatomy (RPA) classification system is proposed to aid in communication, reconstructive planning, and developing research cohorts suitable for modern pelvic reconstructions. This study aimed to 1) describe the development of the RPA classification as a novel classification system for pelvic reconstruction; 2) demonstrate the content and construct validity of the system; and 3) investigate its interobserver and intraobserver agreement.

Methods

A group of orthopaedic surgeons and researchers, based in Sydney, Australia, developed the system’s four domains, which together form the overall classification. A total of 23 pelvic reconstruction surgeons provided an assessment of each domain’s clarity and relevance, producing an Item-Content Validity Index (I-CVI). They then applied the classification to the surgical plan and postoperative radiograph of 20 pelvic reconstructions twice, at a minimum of two weeks apart. Construct validity was assessed by agreement between observers and authors. Interobserver and intraobserver agreement was assessed using the kappa statistic.

Results

A total of 23 surgeons completed the assessment of content validity, and 18 surgeons completed both surveys. Each domain demonstrated content validity for both relevance and clarity (I-CVI > 0.78). Authors and observers had moderate agreement on the classification supporting construct validity (k = 0.6) with a median agreement of 90% for individual domains. Interobserver agreement was moderate overall (k = 0.54), with individual domains having substantial or better agreement (k = 0.73 to 0.82) except for anterior fixation (k = 0.57). Intraobserver agreement of individual domains was substantial or almost perfect (k = 0.76 to 0.84) with moderate agreement for the overall classification (k = 0.6).

Conclusion

This study establishes the content and construct validity of the RPA classification system as a novel classification system for pelvic reconstruction, and its reliability to support its use as a communicative tool, and in research and surgical planning.

Cite this article: Bone Jt Open 2026;7(7):871–881.

Introduction

Reconstruction of the pelvis may be required to manage massive bone loss in the context of bone tumours, trauma, deformity, and failed hip arthroplasty.^1-3 Early pelvic reconstructions (like the ‘saddle prosthesis’) have been associated with high rates of failure and morbidity.^3,4 Custom-made prostheses (including 3D-printed or additive manufactured) have become increasingly used due to their design versatility, superior functional outcomes, and lower incidence of structural failure and aseptic loosening.^5-8 The significant bone defects encountered and diverse pelvic implant designs in pelvic reconstruction require a multidisciplinary approach when planning and performing pelvic reconstructions.^9,10 Reconstructions of all kinds present challenges with bone stock, implant fixation, and restoring pelvic ring stability, irrespective of whether the pathology is oncological or a revision total hip arthroplasty on a patient with a large acetabular defect.^1,11-13 A classification system to aid in communication is beneficial in planning such procedures, and will assist in comparing different reconstructions.

Enneking and Dunham’s classification¹⁴ is widely used to classify bone resections in the pelvis. This divides the hemipelvis into four zones: iliac wing, periacetabular region, pubic rami, and sacrum, but does not describe the technique for subsequent pelvic reconstruction. With the introduction of custom-made prostheses, this classification is less suited to describing more complex pelvic defects associated with many reconstructive procedures, inviting the proposal of alternatives.^7,8 Houdek et al¹⁵ and Zhang et al¹⁶ proposed classification systems for sacropelvic resections, describing the sagittal extent of sacral involvement in sacropelvic resections.¹⁶ Recently, the West China Classification System was proposed to supplement Enneking’s classification, focusing on hip joint loss and resection of the sacroiliac joint (SIJ) or pubic symphysis.¹⁷ Each group sought to produce a classification which served as a communicative tool and provided standardized procedures based on their classification.^14-16 These classifications are based on resections and do not address the bone/implant interface or fixation technique, both of which are relevant in pelvic reconstruction.^18,19 Despite being proposed as a communicative tool, none of these classification systems have undergone reliability or validity assessment.

The Reconstructive Pelvic Anatomy (RPA) classification system has thus been developed as a classification system for pelvic reconstruction that is suitable for modern reconstructive techniques. As a valid and reliable classification system for pelvic reconstruction, the intention is for it to provide a common language for communication and surgical planning. The ability to categorize implants by their characteristics may also help with comparing surgical and functional outcomes, as well as studying their mode of failure.

This study aimed to: 1) describe the development of a classification system for pelvic reconstruction; 2) demonstrate the content and construct validity of the classification among a group of pelvic reconstruction and revision hip arthroplasty surgeons; and 3) evaluate the interobserver and intraobserver agreement among this group.

Methods

Development of the classification system

This classification was developed by a panel of five orthopaedic surgeons (three of whom perform pelvic reconstructions) (RB, DF, IAH, VSS, MG) and four research members (WJJC, KA, PH, JZ). The goal was to develop a clear, reliable, and clinically relevant classification system for pelvic reconstruction. Initial conceptualization was presented by two surgeons (DF, RB), then further clarified through a consensus exercise. The group discussed the relevant anatomical boundaries that differentiated any pelvic resection and implant biomechanically, and the appropriate description for each domain of the classification. The classification was then piloted internally by surveying two of the surgeons, with ten cases selected by two of the authors (RB, WJJC). The group reviewed the results and adjusted the descriptions and anatomical boundaries of resection based on the feedback to improve relevance and clarity. The panel determined that the subsequent assessment of the pelvic reconstructions would be made using a postoperative anteroposterior (AP) pelvis radiograph with a 3D image of the preoperative surgical plan serving as an adjunct.

Patients

This study was approved by the Ethics Review Committee of the Sydney Local Health District (Protocol No. X23-0490 and 2023/ETH02889). A retrospective review of patients who underwent a pelvic reconstruction procedure by any of four orthopaedic pelvic reconstruction surgeons (RB, DF, MG, PDS) at a tertiary centre in Sydney, Australia, between August 2013 and January 2023 was conducted through an institutional database. Inclusion criteria were any patient who had a pelvic reconstruction procedure, including for both tumour-associated and non-tumour cases requiring reconstruction (such as revision total hip arthroplasty (rTHA)), within the study period. Exclusion criteria were any patient who did not have both a postoperative AP pelvis radiograph and preoperative 3D-printed surgical plan available, which excluded a total of 17 patients. Indications for all procedures are in Table I.

Table I.

Indications for pelvic reconstruction.

Indication, n (%)	N = 93
Non-tumour	31 (33)
Aseptic loosening	17 (18)
Osteolysis	5 (5)
Pelvic dissociation	3 (3)
History of prosthetic infection	2 (2)
Fracture	2 (2)
Dislocation	1 (1)
Developmental dysplasia of the hip	1 (1)
Tumour associated reconstruction	62 (67)

All reconstructions were classified according to the classification system using the postoperative radiograph and the preoperative surgical plan by the principal and senior author (WJJC and RB) separately, with disagreements resolved by a third author (DF). All cases were categorized based on their anterior and posterior osteectomy, with 13 different categories of osteectomy found (Table II).

Table II.

Distribution of the extent of osteectomies (described in results) for all patients who underwent pelvic reconstructions (n = 93).

		Anterior osteectomy
		0	1	2	3
Posterior osteectomy	0	N/A	N/A	N/A	N/A
	1	N/A	40 (44%)	1 (1%)	3 (3%)
	2	N/A	11 (12%)	8 (9%)	6 (7%)
	3	N/A	2 (2%)	5 (5%)	2 (2%)
	4	6 (7%)	5 (5%)	3 (3%)	1 (1%)

N/A, not applicable.

Survey cases

One case was selected at random from each of the 13 categories of resection using computer-generated randomization. Categories that reflected over 5% of the distribution had a second case randomly selected. A survey of 20 cases was created, reflecting both the variety and distribution of osteectomies, while maintaining a randomized process.

Reliability and validity assessment

Surgeons (observers), with experience in revision hip and pelvic reconstructive surgery, were invited by email to assess the content validity of the classification system in survey format. The survey included questions regarding participant demographics and experience. Observers were asked to apply the classification system to the 20 survey cases described earlier using the postoperative pelvic radiograph and the preoperative surgical plan. Finally, participants were asked to reclassify cases at a minimum of two weeks later to assess intraobserver agreement.

Each observer was initially provided with the information sheet including a description of the classification system and the criteria described in Table III, a diagram of each resection boundary in the classification (Figure 1), as well as six example cases of the classification being applied. Observers were not provided with the justification for each criterion described in Table III.

Table III.

The reconstructive pelvic anatomy classification system criteria for posterior and anterior osteectomy and fixation of a pelvic reconstruction, with reasoning for each.

Classification	Criteria	Reasoning
Posterior osteectomy
0	No acetabular involvement	Potentially no need for reconstruction/implant
1	Acetabular involvement without extension into the greater sciatic notch (notch intact)	Necessitates arthroplasty (acetabular reconstruction) Simpler dissection with smaller moment at implant interface Large amount of residual iliac bone for fixation
2	Extension into sciatic notch without involvement of the SIJ (notch not intact)	Requires a more complex dissection with likely sacrospinous ligament sacrifice Creates larger moment arm necessitating design with resistance to bending Modest risk to superior gluteal neurovascular bundle Presence of sciatic buttress for fixation
3	Involvement of SIJ with ≥ 50% of the SIJ intact	Complex dissection with osteotomy around superior gluteal neurovascular bundle Larger reconstruction likely requiring fixation across SIJ Likely to have preservation of at least three cortices for improved SIJ screw fixation
4	Involvement of SIJ with < 50% of the SIJ intact	Complex dissection with potential for sagittal osteotomy or resection through SIJ Dissection superior gluteal neurovascular bundle and likely sacral nerve roots Larger reconstruction requiring fixation into sacrum Likely to be mostly fixed into cancellous bone only
Posterior fixation
A	Not fixed	Included for classification completeness, this occurrence would be rare or unlikely
B	Fixed: fixation not crossing SIJ	Minimal impact on SIJ joint motion, relies on fixation within the residual ilium
C	Fixed with fixation screws crossing SIJ	Significant impact on SIJ motion Risk to sacral nerve roots Likely improved bending moment resistance Better quality and volume of bone for fixation
Anterior osteectomy
0	No acetabular involvement	Avoids the need for arthroplasty Potentially limited supra-acetabular bone for fixation
1	Acetabular involvement without extension into pubic rami	Necessitates arthroplasty Simple dissection Achievable in most approaches Substantial and accessible bone for anterior fixation
2	Extension into pubic rami without extension into the pubic symphysis (symphysis spared)	More complex approach and dissection, potentially necessitating second incision Potential risk to femoral and obturator neurovascular bundles If fixation used, larger implant with minimal bone stock and potentially difficult access for placement
3	Extension into or beyond the pubic symphysis (symphysis sacrificed)	Large implant Necessitates second incision Risk to femoral and obturator neurovascular bundles
Anterior fixation
A	Not fixed	Leaving the pelvic ring ‘open’ Avoiding loosening of anterior fixation
B	Fixed: fixation not crossing pubic symphysis joint	Most closely replicates the native pubic symphysis joint motion Smaller volume of bone for fixation options
C	Fixed with fixation screws crossing pubic symphysis joint	Significant impact on pubic symphysis joint stiffening Implication for transmission of force to implant bone interface Improved fixation options

SIJ, sacroiliac joint.

Fig. 1.

This illustration depicts the margins of classification for both a) posterior and b) anterior osteectomy of the Reconstructive Pelvic Anatomy Classification System. This diagram was used when classifying the case bank and was given to observers. SI, sacroiliac.

Content validity assesses how well the classification reflects the aspects of the phenomena under study and is indicative of clinical relevance.^20,21 To evaluate content validity, a described technique was used where observers were asked to provide an assessment of relevance and clarity of each of the domains within the classification (anterior osteectomy, posterior osteectomy, anterior fixation, posterior fixation) when describing a pelvic reconstruction. The relevance and clarity of each domain was rated on a four-point scale (not, somewhat, quite or highly relevant/clear).²²

No ‘gold standard’ classification system exists as a comparator for the RPA classification. Construct validity is, therefore, best measured by providing a series of tests/questions where the answers are known, and assessing if the classification acts as expected.²³ Thus, construct validity was assessed by evaluating agreement between the observers and the authors. Classifications by observers at both survey timepoints were used to assess interobserver and intraobserver reliability. Data were collected electronically via REDCap (Vanderbilt University, USA).

Statistical analysis

Descriptive statistics were used to summarize participant characteristics. Categorical data were presented as frequencies (percentage), and continuous data were presented as mean (SD). Answers to the Content Validity Survey were recoded, using the technique described by Polit et al:²² 0 = not or somewhat clear/relevant; 1 = quite or highly clear/relevant). An Item-Content Validity Index (I-CVI) was produced as an average score. Items scored I-CVI > 0.78 were considered evidence of good content validity.²²

Agreement between authors and observers was calculated using Cohen’s Kappa statistic to reflect construct validity.²⁴ Fleiss’ Kappa coefficients were calculated to evaluate agreement between individual observers (interobserver agreement). Cohen’s Kappa coefficient was calculated to evaluate agreement of assessments made at different time points by one observer (intraobserver agreement). The interobserver and intraobserver agreement was calculated for individual domains and for the ‘overall classification’, this being that observers either agree on the overall four-character code, or they do not.

The strength of agreement was interpreted as poor (< 0.20), slight (0.21 to 0.40), moderate (0.41 to 0.60), substantial (0.61 to 0.80), or almost perfect (0.81 to 1.00).²⁴

Statistical analyses were performed using R Statistical Software v. 4.4.2 (R Core Team 2024; R Foundation for Statistical Computing, Austria). Fleiss’ Kappa was assessed using the irr and DescTools packages.^25-27 Cohen’s Kappa was assessed using the irr and psych packages.

Results

Classification system

The panel determined that a pelvic reconstruction classification system was best achieved by focusing on both anatomy and fixation. The RPA classification uses four domains which are reflected as a four-character code. This is the posterior and anterior osteectomy margins or bone/implant interface, as well as fixation onto the remaining bone for each osteectomy. The four domains are categorized alphanumerically as nXnX (n = number, X = letter). The numerical variables reflect the extent of the posterior and anterior osteectomy relative to the acetabulum, while the letters reflect the fixation onto the posterior and anterior bone interface (or lack thereof).

The first two characters represent the posterior osteectomy and fixation at the bone-implant interface, and the second two characters represent anterior osteectomy and fixation (nXnX = posterior osteectomy/posterior fixation/anterior osteectomy/anterior fixation). To define the posterior osteectomy, one must start at the pubic symphysis and work superiorly/posteriorly with respect to the pelvic ring until the level of the posterior osteectomy is reached. The first character defines the bony limit of the posterior osteectomy (with five options: 0 to 4) and the second character describes fixation used posteriorly (with three options: A to C). The third character defines the bony limit of the anterior osteectomy (with four options: 0 to 3), beginning at the sacroiliac joint (SIJ) and moving anteriorly/inferiorly, with respect to the pelvic ring, until the level of the anterior ostectomy is reached. The fourth character defines the fixation used anteriorly (with three options: A to C). All classifications must have four variables. The specific criteria that define osteectomy and fixation are shown in Figure 1 and Table III with the panel’s reasoning.

Determination of the criteria was firstly along bony biomechanical determinants, and secondarily complexity of the surgery required to achieve the resection. These included neurovascular structures at risk, necessity of second incision, and extent of soft-tissue dissection, muscle and tendon release (such as gluteal muscle attachment and the sacrospinous ligament), and biomechanical factors like preservation of native joint motion, sufficient bone stock for fixation, residual bone sufficient for cortical fixation, and magnitude of moment arm on the bone-prosthesis interface were also considered.

Survey

In total, 23 surgeons completed the initial survey (completing the validity survey) and 18 completed the second survey. All surgeons were actively performing pelvic reconstruction as part of their practice with over 55% of surgeons having performed eight or more in the preceding three years. This is indicative of substantial engagement given the subspecialty nature of pelvic reconstruction and the inherently low volume of these procedures. Information regarding the background of surgeons is presented in Table IV.

Table IV.

Observer characteristics.

Variable	Total (n = 23)
Sex, n (%)
Female	2 (9)
Male	21 (91)
Years of practice
Mean (SD)	13.0 (9.34)
Median (IQR)	10.0 (6.0 to 19.0)
Country of practice, n (%)
Australia	16 (70)
New Zealand	5 (22)
Other	2 (9)
Perform pelvic reconstruction procedures, n (%)
Yes	23 (100)
No	0 (0)
Number of pelvic reconstructions in the past 3 years, n (%) *
1 to 3	4 (18)
4 to 7	6 (27)
More than 8	12 (55)
Perform revision hip procedures, n (%)
Yes	23 (100)
No	0 (0)
Number of revision hip arthroplasties in the past 3 years, n (%) *
1 to 3	1 (5)
4 to 7	4 (18)
More than 8	17 (77)

^*Excludes one missing value.

Content validity

Each domain exhibited an I-CVI > 0.78 for both clarity and relevance. All domains received an I-CVI > 0.95 for relevance when describing a pelvic reconstruction (Table V).

Table V.

Item-content validity index of the Reconstructive Pelvic Anatomy classification system.

Variable	Posterior oseteectomy (n = 23)	Posterior fixation (n = 23)	Anterior osteectomy (n = 23)	Anterior fixation (n = 23)
Clarity	0.96	0.78	0.83	0.91
Relevance	1.0	1.0	0.96	0.96

Construct validity

Observers and authors had moderate agreement (k = 0.60) on the overall classification with each domain showing either substantial or almost perfect agreement in the initial survey (Table VI). Out of a possible 80 domains for 20 cases, the median percentage agreement between observers and surgeons was 90% (IQR 84% to 94%) (n = 23).

Table VI.

Assessment of agreement for all domains and the overall classification.

Variable	Initial agreement between observers and authors (n = 23) – Cohen’s Kappa, mean (SD)	Initial interobserver agreement (n = 23) – Fleiss’ Kappa (95% CI)	Follow-up interobserver agreement (n = 18) – Fleiss’ Kappa (95% CI)	Intraobserver agreement (n = 18) - Cohen’s Kappa, mean (SD)
Posterior oseteectomy	0.81 (0.14)	0.746 (0.731 to 0.761)	0.726 (0.706 to 0.746)	0.76 (0.16)
Posterior fixation	0.73 (0.15)	0.665 (0.639 to 0.691)	0.762 (0.727 to 0.798)	0.79 (0.16)
Anterior osteectomy	0.82 (0.15)	0.782 (0.765 to 0.798)	0.821 (0.799 to 0.843)	0.82 (0.16)
Anterior fixation	0.79 (0.32)	0.558 (0.537 to 0.579)	0.570 (0.542 to 0.598)	0.84 (0.25)
Overall classification	0.60 (0.21)	0.498 (0.492 to 0.504)	0.538 (0.529 to 0.547)	0.60 (0.21)

Interobserver agreement

The overall classification exhibited moderate interobserver agreement (k = 0.54 (follow-up survey)). Each domain showed substantial agreement except for anterior fixation which had moderate agreement (k = 0.57 (follow-up survey)) (Table VI).

Intraobserver agreement

Intraobserver agreement of the overall classification was moderate (k = 0.6). Intraobserver agreement for all individual domains was substantial or almost perfect (Table VI).

Discussion

All domains in the RPA classification system exhibited content validity with > 78% of experts who perform the procedure considering the domains clear and relevant to pelvic reconstructions. Authors and observers had moderate agreement for the overall classification (k = 0.6), with a median of 90% levels of agreement for individual domains, supporting construct validity of the classification. The RPA classification’s four domains allow for versatility that is suitable to modern 3D reconstructions. For example, Figure 2 shows three reconstructions with acetabular and partial iliac wing resection, however, these reconstructions require different prostheses and surgical plans due to variables such as sciatic notch involvement, proximity of the bone/implant interface to the SIJ, and fixation (or lack thereof) anteriorly. The RPA classification can delineate the important differences between such reconstructions, and the validity of this classification is supported by this study.

Fig. 2.

This diagram depicts postoperative radiographs and preoperative surgical plans. a) A periacetabular reconstruction without screws crossing a joint – Reconstructive Pelvic Anatomy (RPA): 1B1B. b) A pelvic reconstruction that extends posteriorly to involve the sacroiliac joint (SIJ), with > 50% SIJ retained, and screws crossing the SIJ – RPA: 3C1B. c) A reconstruction resecting the sciatic notch posteriorly with screws crossing the SIJ and a complete resection of pubic rami without fixation anteriorly – RPA: 2C3A.

The RPA classification’s use of four different domains functions as four classifications in one. While this addresses the complexities of considerations in pelvic reconstructions, utilizing four domains increases the opportunity for disagreement between observers. The most compelling evidence to support the reliability of the classification system is the median of 90% agreement for individual domains between observers and authors, with substantial individual domain agreement (k = 0.73 to 0.82). The overall interobserver (k = 0.6) and intraobserver (k = 0.54) agreement reflect a statistical reality in a multidomain classification system in which agreement must occur across four different domains. If individual domain agreement was 90%, when aggregated, one would only see agreement on the overall classification 66% of the time. These results reflect high individual domain performance aggregated across four domains.

The RPA classification fares favourably in comparison to other multidomain classification systems. The commonly used Tile classification for pelvic fractures has shown an interobserver agreement of k = 0.44 when it uses only two domains with only three options per domain.²⁸ Other multidomain classification systems, such as the AO/Orthopaedic Trauma Association (OTA) classification for pelvic ring fractures and the Paprosky classification,²⁹ have all exhibited lower interobserver and intraobserver reliability despite having fewer domains.^30,31

The overall reliability is not a failure but an expected characteristic of a multidimensional classification system. The four-domain structure is necessary to capture the complexity of pelvic reconstruction, which involves addressing bone loss patterns/osteectomy as well as reconstructive solutions. This level of granularity cannot be achieved with a single-domain classification system, and so the RPA classification cannot be compared fairly with typical single-domain classification systems.

While this is the case, several strategies could be used to improve the RPA classifications reliability in clinical and research applications, including: 1) the use of 3D imaging (CT or 3D reconstruction), which could aid in observers identifying precise anatomical landmarks; 2) standardized training modules for observers with borderline classifications; and 3) establishing a library of reference cases to aid in decision making.

In the setting of complex pelvic defects, reconstructive surgery aims to implant a prosthesis that allows for osseointegration and withstands biomechanical stress.^18,32,33 Enneking and Dunham’s classification is intuitive and facilitates communication in pelvic resections. However, advancements in pelvic reconstructive surgery have invited proposals of classification systems that consider anatomical landmarks beyond those originally described by Enneking.^5-8,32,34-37 Several descriptive classification systems exist focusing on pelvic resection without describing the reconstruction.^15-17 This study provides alternative anatomical boundaries which have undergone validity assessment among a group of pelvic reconstructive surgeons to support their relevance and clarity in practice (Table III). Several key anatomical landmarks formed part of this classification system.

The sciatic notch was considered a relevant landmark as described osteotomies require a more complex dissection and isolation of major vessels and nerves, as well as protection of pelvic viscera, when compared with cases requiring only acetabular reconstruction.^38,39 Resection of the iliac wing into the SIJ significantly increases the load-bearing stress that spans the sacroiliac screw fixation, elicited by standing in the setting of complex reconstructions using 3D-printed reconstructions.¹¹ Furthermore, endoprosthetic hemipelvic reconstruction involving the SIJ has shown significantly reduced patient functional outcomes, supporting the consideration of SIJ sacrificing reconstructions as a distinct class. Resection is equivalent to fusion of the SIJ resulting in loss of this pelvic motion segment. A delineation of 50% of SIJ retention was made, as it is known that the residual ilium facilitates improved anchorage for pelvic endoprostheses.^40,41 More significant SIJ resection impedes one’s ability to span three cortices with fixation, requires cancellous fixation in the sacrum, and makes the prostheses vulnerable to shear loading.^41-44

Anteriorly, resection of the pubic bone necessitates a more complex surgical procedure than standard arthroplasty, and has implications for soft-tissue reconstruction of the abdominal wall and for pelvic floor integrity.⁴⁵ Resection of the pubic symphysis has biomechanical implications due to the interdependence of SIJ and pubic symphysis joints, meaning that if the ring is left open, the posterior ring will sustain greater stress.^43,46 However, if either the sacroiliac or pubic symphysis joint is involved in the construct, one still sacrifices its native biomechanics, reducing its physiological micromotion and thus concentrating forces at the bone-implant interface.⁴³ It should be noted that anterior resection is predominantly used in the setting of tumour resections, as opposed to rTHA, which typically considers acetabulum and posterior ilium bone loss. Nonetheless, the RPA classification system offers the versatility to be applied to both.

Tumour resection and rTHR have distinct differences in pathology and operative planning. Principles of tumour resection require clear margins utilizing predictable, well-defined osteotomies,^1,17,47 whereas rTHA seeks to be bone-preserving in the setting of sclerosis, variable bone loss, and chronic pelvic discontinuity.^2,47,48 The RPA classification focuses on the bone-implant interface, which makes it versatile enough to be applicable to both tumour and rTHA. A key difference, however, is that in the setting of rTHA, the classification will correspond to the extent of pre-existing bone loss rather than a planned osteotomy/resection. Despite this, the biomechanical principles are similar between rTHA and pelvic reconstruction, such as secure fixation, which withstands physiological load in the setting of bone defects, restoration of pelvic ring mechanics, intraoperative precision, and osseointegration, justifying a unified classification.^{17,18,34,48-50} However, the differences between the two procedures could warrant supplementary additions in future iterations.

The RPA classification focuses specifically on the boundaries of fixation and whether screws span key joints (SIJ and pubic symphysis) as opposed to specific fixation hardware or technique. A wide variety of fixation methods are used in pelvic reconstruction: custom triflanges, traditional cortical screws, and highly porous trabecular materials, among other devices.^1,3,29,45,51 Implant designs continue to evolve rapidly in this space, with variability among manufacturers. The classification’s emphasis on joint spanning fixation (or lack thereof) provides a durable classification that can persist despite hardware evolution. Implant-specific features could later be considered as a supplement to the classification system.

A novel aspect of the RPA classification is its inclusion of ‘fixation’ as a further domain when describing pelvic reconstruction. Use of cortical screw fixation has been shown to have a correlative relationship with screw retention, stability, and mobilization in the setting of fracture and sacropelvic fusion.^19,52 It is believed that implant fixation with locking cortical screws has a similar relevance in pelvic reconstruction.

The RPA classification system offers anterior fixation category ‘A’, indicative of an open pelvic ring. This accommodates evolving practices and evidence in the management of both chronic pelvic discontinuity and Enneking II/III tumour resections. In the setting of chronic pelvic discontinuity, it has been suggested that anterior/inferior fixation is not required, with adequate fixation being achieved with monoflange/triflanges posteriorly alone.⁵³ In tumour reconstructions, some suggest that posterior fixation to restore the weightbearing axis is sufficient and closure of the pelvic ring anteriorly is not necessary.^11,32 The RPA classification system offers the versatility to support comparative research between ring closure and non-closure in both rTHA and tumour resection.

The strength of this study lies in its comprehensive assessment of reliability and validity of a novel classification system by 24 surgeons who perform pelvic reconstruction with an average of 13 years of experience. A limitation of the study was the use of plain radiographs and a single image mapping plan to delineate anatomical boundaries in the survey. CT or MRI may have afforded more precise anatomical assessment of the pelvic reconstructions; these were not routinely performed postoperatively, and distributing 3D imaging or the full planning templates in an extensive survey was logistically challenging. Additionally, while the classification system is intended to be used for all pelvic reconstruction, the study includes only custom-made 3D-printed designs which, although indicative of modern advancements, is not reflective of all pelvic reconstructive techniques. Further study is required to determine the RPA classification’s use beyond custom-made implant cases. Another limitation is that only 18 of the 23 surgeons completed both surveys; although this was not ideal, recent examples of reliability studies in novel classification systems have seen far fewer observers used.^54,55 Our study uses a high number of observers and further verifies their expertise by collecting information on their years of experience and the frequency with which they undertake pelvic reconstruction as part of their practice.

The RPA classification can be used reliably by different clinicians, and is consistent when applied by the same clinician, which supports its utility in preoperative planning and in making reproducible surgical decisions.⁵⁶ Furthermore, the RPA classification has been shown to be perceived as relevant and clear to modern pelvic reconstructions when considered by experienced surgeons who perform the procedure. This study supports its use as a valid and reliable communicative tool, with the intention that the classification’s reliability may enable consistent data collection to allow for analysis and comparison of outcomes, as well as surgical planning.⁵⁷ Further research is required to investigate this aim.

Take home message

- This study provides a novel classification system for pelvic reconstruction that is suitable for modern reconstructive techniques: the reconstructive pelvic anatomy classification system.

- It demonstrates that this classification system is both valid and reliable among surgeons who perform this procedure. This classification system can be used in surgical planning, communication and research.

Data Availability

The data that support the findings for this study are available to other researchers from the corresponding author upon reasonable request.