Causes and Frequencies of Reoperations After Endoprosthetic Reconstructions for Extremity Tumor Surgery: A Systematic Review

Abstract

Background

Excision of bone tumors and endoprosthetic reconstruction allow patients early weightbearing and a potential functional advantage compared with amputation. These reconstructions do not restore the limb to normal status, however, and patients are subject to complications that may result in revision or loss of the limb. Because better understanding of these complications based on current information might help the patient and surgeon in decision-making, we undertook a systematic review of studies published on this topic.

Questions/purposes

(1) What are the primary modes and proportion of failure of tumor endoprostheses in patients undergoing reconstruction after excision of primary extremity bone sarcomas?

Methods

We systematically searched MEDLINE, Embase, and the Cochrane Library for all studies published from April 15, 1998, to April 15, 2018. Three reviewers independently reviewed studies reporting endoprosthetic reconstruction survival and events requiring revision for primary extremity bone tumors treated with endoprosthetic reconstruction for inclusion and performed independent data extraction. We excluded all studies with fewer than five patients, any systematic review/meta-analyses, and any study not reporting on primary extremity bone tumors. All discrepancies were resolved by the study’s senior author. Data extracted from included studies were any reoperation event for wound dehiscence, any operative fixation for a pathologic fracture, and any revision of the primary endoprosthesis for implant wear or breakage, deep infection not amenable to prosthesis retention, or for local recurrence. We assessed the overall quality of the evidence with the Methodological Index for Non-Randomized Studies (MINORS) approach with a higher MINORS score representative of a more methodologically rigorous study with a total possible score of 16 points for noncomparative and 24 points for comparative studies. Forty-nine studies met criteria for inclusion from an initial search return of 904 studies, of which no studies were randomized controlled trials. From a total patient population of 2721, there was a mean followup of 93 months (range, 1-516 months) with loss to followup or death occurring in 447 of 2118 (21%) patients with six studies not providing loss to followup data. The mean MINORS score was 14 for prospective studies and 11 for retrospective studies.

Results

Overall, there were 1283 reoperations among the 2721 (47%) patients. Reoperation for mechanical endoprosthetic events (soft tissue dehiscence or periarticular soft tissue instability, aseptic loosening, or implant wear/fracture) occurred in 907 of 2721 (33%) patients. Aseptic loosening occurred at a mean of 75 months (range, 1-376 months) in 212 of 315 patients (67%). Deep infection requiring removal of the initial prosthesis occurred in 247 of 2721 (9%) patients with deep infection occurring at a mean of 24 months (range, 1-372 months) in the 190 infections (77%) with time to infection data available. Local recurrence rates requiring revision or amputation occurred in 129 (5%) of all patients. There was an overall primary endoprosthesis survival rate without any surgical reintervention of 63% among reporting studies at a mean of 79 months followup.

Conclusions

Failures of endoprosthetic reconstructions after extremity tumor surgery are common, most often resulting from implant wear or fracture, aseptic loosening, and infection. Importantly, the aggregated data are the first to attempt to quantify the time to specific complication types within this patient population. Deep infection not amenable to endoprosthesis retention appears to occur approximately 2 years postoperatively in most patients, with aseptic loosening occurring most commonly at 75 months. Although endoprosthetic reconstruction is one of the most common forms of reconstruction after bone tumor resection, the quality of published evidence regarding this procedure is of low quality with high loss to followup and data quality limiting interstudy analysis. The quality of the evidence is low with high loss to followup and inconsistent reporting of times to reintervention events. Although the most common modes of endoprosthetic failure in this population are well known, creation of quality prospective, collaborative databases would assist in clarifying and informing important elements of the followup process for these patients.

Level of Evidence

Level IV, therapeutic study.

Introduction

Advanced endoprosthetic implant designs and fixation methods have given orthopaedic oncologists increased options to reconstruct osseous defects after tumor excision [45]. Endoprosthetic reconstruction has several advantages compared with other reconstruction techniques, including wide availability, versatility, improvement of ambulatory status, and early rehabilitation, all while enabling a surgeon to remove all gross disease to minimize recurrence rates [3, 49, 62]. Complications of endoprosthetic reconstruction include infection, aseptic loosening, structural/hardware failure, and local recurrence. Any of these complications can diminish prosthesis survival [53, 62].

Infection of tumor endoprostheses is a major complication that complicates somewhere between 2% and 20% of these reconstructions [18, 40, 47, 48]. Endoprosthesis survival rates are widely variable and affected by both malignancy type and study sample size with some reports quoting 5- and 10-year survival rates > 80% with overall malignancy survival rates ranging from 61% to 92% [23, 62]. Despite improvements in implant design, implant-related complications and infections continue to complicate endoprosthetic reconstructions [23]. Reported reintervention for oncologic endoprosthesis reconstructions is higher than that for primary joint arthroplasty for all anatomic sites [23]. Racano et al. [53] previously conducted a systematic review on infection events in the endoprosthesis reconstruction population after long-bone tumor surgery. This review reports infection rates of 10% in the tumor endoprosthesis reconstruction population while not clearly identifying important intervals of when infections present [51]. Furthermore, there is currently no aggregated data on accepted complication values for all possible modes of postoperative complications. Most concerning, there is a wide discrepancy among orthopaedic oncologists with regard to followup and surveillance practices with 95% of recently surveyed surgeons admitting to following nonevidence-based surveillance protocols, calling for improved surveillance guidelines [21]. Thus, we sought to conduct a systematic review of the available evidence on modes of complications for primary extremity tumor endoprosthetic reconstructions, because this information may also aid surgeons in their discussions with patients about potential complications and longevity of these implants when deciding on limb salvage options.

Materials and Methods

We followed the protocol outlined in the Cochrane Handbook for Systematic Reviews of Interventions [34]. We report our findings in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [46].

Search Strategy and Criteria

We searched MEDLINE (1946 to the present), Embase (1974 to the present), and the Cochrane Library (no date limit) for articles published from April 15, 1998, up to and including April 15, 2018. Various combinations of subject headings and subheadings were utilized and supplementation with free text increased our sensitivity (see Appendix, Supplemental Digital Content 1, http://links.lww.com/CORR/A128). Additionally, we performed a manual review of the reference lists of all included full-text studies to identify any potential additional studies for inclusion. The “related articles” function in PubMed was also used to identify any potential further applicable studies.

Inclusion and Exclusion Criteria

Three unblinded reviewers (AM, MV, PT) performed triplicate independent study title and abstract screening for eligibility from an initial search return of 904 studies (Fig. 1). Any disagreements identified were resolved by the study’s senior author (RV). We included all English language studies that reported postoperative complications with endoprosthetic reconstruction in the setting of a primary extremity sarcoma. Given that functional outcomes in endoprosthetic reconstruction are strongly correlated to and directly relate to a number of implant and soft tissue-related complications requiring surgical reintervention, we included those studies that provided prognostic and/or functional patient-reported outcomes for a primary extremity bone tumor population that had undergone endoprosthetic reconstruction [32]. We excluded all case reports with less than five patients to minimize the risk of falsely under- or overrepresenting complications of endoprosthesis complications as a percentage of included patient numbers. In addition, we excluded all review articles and systematic reviews/meta-analyses. We also excluded studies with < 12 months of followup or if they reported endoprosthetic reconstructions for bony metastatic disease without reporting endoprosthetic survival or reoperation events for primary bony extremity sarcomas separate of metastatic tumor reconstructions. Additionally, given the expected high rates of loss to followup from death resulting from the patient’s primary diagnoses, any study not reporting on reintervention events among patients who died during a study followup period were excluded. Ultimately, after reviewing 74 studies in full, 49 studies with a total of 2721 patients met criteria for inclusion (Fig. 1).

Fig. 1.

Depiction of the flow of selection of studies for inclusion and exclusion is shown.

Assessment of Study Quality

The bulk of included trials centered on average age at index surgery of approximately 20 to 30 years ranging from patients as young as 3 years to 86 years (see Appendices, Supplemental Digital Content 2, http://links.lww.com/CORR/A129, and Supplemental Digital Content 3, http://links.lww.com/CORR/A130). Among 47 studies reporting gender demographics, there were 1401 of 2526 (55%) male patients with a mean followup of 94 months (range, 1-516 months). Most studies were retrospective in nature with population sizes highly variable (5-278 patients). Nearly all included studies were single-center reporting univariate statistics without comparators. Across all included studies, the overall quality of reporting was low and highly heterogeneous.

A quality assessment analysis of all included studies was performed using the Methodological Index for Non-Randomized Studies (MINORS) criteria [61]. Each of the 12 items of the MINORS criteria is scored 0 to 2 for an ideal score of 16 for noncomparative studies and 24 for comparative studies with higher scores representing better study quality [61]. The MINORS criteria were applied to each included study in triplicate by the reviewing authors (AM, MV, PT) with discrepancies again resolved by the senior author (RV).

We found no published, completed randomized controlled trials reporting on prognostic factors for reintervention among endoprosthetic reconstructions for extremity tumor surgery. Additionally, the types of studies reporting endoprosthetic survival and complications are primarily medium-sized (55 [range, 5-278]) patient retrospective (45 studies) studies with few prospective studies (four studies) (Table 1). The most frequently reported site of malignancy and endoprosthetic reconstruction was the distal femur (1567 patients) and proximal tibia (476 patients), with osteosarcoma (1558 patients), and Ewing’s sarcoma (258 patients) as the most commonly reported primary extremity malignancies. Across the included studies, the average MINORS score was 14 for prospective studies and 11 for retrospective studies (ideal score of 16 for noncomparative studies and 24 for comparative studies). The bulk of the included studies was rated down for inconsistent followup reporting and loss to followup as well as retrospective sample size calculations.

Table 1.

Study characteristics for included studies

Data Collection and Extraction

For consistency as well as completeness of data extraction, triplicate data extraction (AM, MV, PT) was used with comparison of all results. Any discrepancies with respect to extracted data were resolved by the senior author (RV). From each study, the age at the time of surgery, site of primary malignancy, tumor type, and length of followup were extracted. To determine the modes of complications among endoprosthetic reconstructions for the included studies, complication modalities were classified according to a modification of the Henderson classification with a focus on only those complications requiring surgical intervention or prosthesis removal in the case of infection and local recurrence [32]. Henderson et al. [32] previously identified six primary modes of endoprosthetic “failure” after tumor excision and reconstruction including mechanical failures (soft tissue failure [Type 1]; aseptic loosening [Type 2]; structural failure [Type 3]) and nonmechanical failures (infection [Type 4]; tumor progression [Type 5]; pediatric failures [Type 6]). When studies reported complications without utilizing the Henderson classification, the data extraction authors determined the most applicable complication type by the Henderson classification definition with any discrepancies clarified by the senior author (RV).

Data Synthesis

We calculated interobserver agreement for the assessments of study eligibility using the Cohen κ coefficient and interpreted the κ values according to Landis and Koch as poor (κ value 0), slight (0.01-0.20), fair (0.21-0.40), moderate (0.41-0.60), substantial (0.61-0.80), or near perfect (0.81-1.00) [44]. From the initial screening, the three review authors achieved a Fleiss κ of 0.78, corresponding to high substantial agreement [16].

Results

Overall there is a high proportion of surgical reintervention among endoprosthetic reconstructions for primary extremity tumor reconstructions with an overall reintervention proportion of 1283 of 2721 (47%) for all modes of complication (Table 2). Among included prospective and retrospective studies, the most common mode of surgical reintervention was mechanical complications (periprosthetic fracture, implant fracture, bushing/implant wear) (436 of 2721 [16%]) followed closely by revision for aseptic loosening (315 of 2721 [12%]). The time to mechanical complication requiring reoperation was available for only 260 of 436 patients (60%), highlighting a time to failure mean of 48 months (range, 0-329 months), whereas aseptic loosening failures were reported in 212 of 315 patients (67%) at a mean of 75 months (range, 6-376 months). Deep surgical site infection not amenable to primary implant retention occurred in 247 of 2721 (9%) patients at a mean followup of 24 months (range, 1-372 months) in the 190 of 247 (77%) patients with available time to infection data. Interestingly, the overall combined incidence of tumor progression/recurrence at the reconstruction site was reported in only 129 of 2721 (5%) of all patients. Time to local recurrence data were only available for 49 of 129 patients (38%) with reported time to local recurrence of 36 months (range, 5-299 months). Across all included trials, the proportion of pediatric complications according to the Henderson classification (pediatric “failures” of limb length discrepancy or angular deformity about a joint) were poorly reported. Although Type 6 “failures” can incorporate all pediatric complications according to the Henderson classification, many authors classified pediatric complications as one of Types 1 to 5 instead, whereas other studies reported these failures as pediatric (Type 6) regardless of complication type when a patient had a complication when younger than age 18 years. It was difficult to determine postoperative complication types universally among any studies including a broader age range incorporating both adults and pediatric patients. Additionally, functional outcome scores among pediatric patients with a surviving primary endoprosthesis were inconsistently reported across the included studies. However, 29 included studies (59%) in which reported functional outcome scores were available reported outcome scores using the Musculoskeletal Tumor Rating Scale (MSTS) with a mean of 25 out of a possible score of 30 (83%) (range, 19-30).

Table 2.

Henderson classification of endoprosthetic complications among all included studies

Discussion

Endoprosthetic reconstruction continues to be widely used for primary extremity tumor reconstruction. We have provided the first comprehensive aggregate data on overall endoprosthesis reconstruction outcomes and when these complications can be most commonly expected to occur. There is a high reported proportion of reoperations for endoprosthetic reconstructions, primarily implant-related complications requiring surgical intervention (436 patients [16%]) and aseptic loosening (315 patients [12%]) with very low reported proportions of local tumor recurrence (129 patients [5%]). The most accurately reported complication for revision of endoprostheses was in the setting of deep surgical site infection not amenable to implant retention with an event proportion of 247 (9%) patients at an average of 24 months postoperatively.

This study had a number of limitations. There was wide variability among included studies with respect to functional patient-reported outcomes and reasons for complications. The associated heterogeneity limited a formal comparative statistical analysis. Given the retrospective nature of the majority of included studies and relatively high mortality among this population, followup was inconsistent among the included studies. This raises the concern for transfer bias. Although our review demonstrated a mean followup of 94 months, it was not possible to determine losses to followup nor the time at which patients were lost to followup in many studies. Importantly, six studies did not report a loss to followup at all, whereas only 18 studies provided a mean time to loss of followup. Of studies reporting a loss to followup, a mean of 21% of patients was lost to followup within these studies. This finding may artificially decrease our complication rate because it is possible that many of these patients, who were lost to followup (death from disease in most instances), may have had one or more complication during their postoperative course had they survived longer. Furthermore, few studies provided SDs for any mean data points reported. With the inconsistency within the provided data among the included studies, we were limited in our ability to provide any statistical comparisons across studies. This prevented our analysis comparing complication rates for given endoprostheses, sites of reconstruction, and tumor type. There was minimal selection bias within the included studies because operative indication, reconstructive technique, and endoprosthesis type were typically clearly detailed. However, every included study is at risk of assessment bias because all studies describe the operative surgeon determining when a complication had occurred. It was often difficult to determine within soft tissue complications what reoperation was performed, whereas conversely, revision surgery for deep prosthetic infections was easily determined. Furthermore, with regard to mechanical complications requiring revision surgery, no studies quantified the degree of bushing wear leading to bushing revision. This is important considering bushing revision represents significantly less morbid revision surgery than mechanical complications from periprosthetic fracture or implant fracture. Additionally, we did not perform a detailed search of nonpeer-reviewed studies, which may limit our ability to identify any forthcoming unpublished studies, including any potential randomized trials [65]. The discrepancy among the lengths of followup (mean, 94 months [range, 1-516 months]) among included studies potentially underreports failures, most probably aseptic loosening, within studies with shorter followups. Finally, the inconsistently reported pediatric complications (limb length discrepancy most importantly) limit our ability to reliably conclude what factors most predict a pediatric complication because often these results were not reported separately of adult complications in the same study.

Our study demonstrates the majority of patients (63%) have a primary endoprosthesis surviving at a mean of 79 months. This represents the largest compilation of all endoprosthesis data for primary bone tumor reconstructions. However, the proportion of complications after endoprosthetic repair is still high despite increased use of endoprostheses over the past several decades [20, 25, 45]. The majority of included studies report on patients with endoprosthetic reconstruction about the knee. Sharil et al. [59] demonstrated that functional patient-reported outcomes after endoprostheses around the knee are good to excellent with good ROM, allowing for daily activities [7, 10, 17, 43, 59] with patients who do predictably well. However, important distinctions in endoprosthesis survival are still unknown for specific reconstruction locations and remain so after this analysis given the data-reporting limitations of the available studies. Jeys et al. [41] report a 10-year survival rate of upper limb endoprostheses of 85%, differing from their 53% lower limb reconstruction survival rates. Overall, they reported limb salvage maintenance after 20 years in 84% of patients [41]. Song et al. [62] reported a 10-year prosthetic survival of their 117 implanted endoprostheses for tumor reconstruction. In their analysis, bone resection > 40% was associated with mechanical prosthesis failure with an optimal cutoff of 43% for resection. Our study showing a proportion of aseptic loosening of approximately 12% is comparable to previously reported event rates of aseptic loosening on a large scale [22, 62].

Concerning for all orthopaedic oncologists, the quality of the available evidence on endoprosthetic reconstructions limits any meaningful interstudy comparison. This draws particular attention to the fragility of all followup protocols within this population. Uncertainty still exists with regard to specific site endoprosthetic reconstructions because the largest aggregate body of data on endoprosthetic reconstructions, which is presented in this study, highlights an overall high proportion of all-cause complications (47%) with a mean 21% loss to followup. We are unable to provide followup recommendations specific to surgical reconstruction site, bone resection amount, implant type, nor primary malignancy given the heterogeneity within the data. The evidence base that informs the reasons for all-cause failures postendoprosthetic reconstruction has not previously been synthesized. Of important clinical relevance is the lack of evidence-based guidelines to inform orthopaedic oncologists of ideal surveillance protocols for their patients after endoprosthetic reconstruction. All studies were rated down by the MINORS score for high loss to followup and biases within their data reporting. Although it is commonly accepted that the quality of evidence surrounding endoprosthesis reconstruction for primary malignancy is low, we wished to demonstrate the paucity of quality data that currently exists to guide practice management for a difficult patient population.

Although it is known that the current level of evidence on survival rates of endoprostheses in this population is low, our combined results highlight a much more concerning phenomenon within orthopaedic oncology, the reliance of clinical management on poorly conducted observational studies. Our systematic review has demonstrated that the myriad of endoprosthetic failures that can occur is stratified with infection requiring revisions typically at the 24-month postoperative period and local recurrence reported at an average of 36 months postoperatively with revision for implant wear/fracture or aseptic loosening at 48 and 75 months postoperatively, respectively. However, these rates are subject to the frailty of the reported data quality. The inconsistency within the endoprosthesis reconstruction population with respect to loss to followup and time to complications highlights the importance of collaborative efforts that ideally should be established to create a prospective database of all endoprosthetic reconstructions. It is only through the creation of such large-scale databases that we will be able to delineate the true complication event rates and better understand how to manage this patient population postoperatively. In the meantime, we hope that the results of this study can help guide surgeons when they discuss potential complications of endoprosthetic reconstructions to their patients.