Graft survival and prognostic factors of recycled autografts after limb salvage surgery in patients with sarcoma of the long bone: a systematic review and meta-analysis of individual participant data.

Introduction
Bone sarcomas are rare malignancies that predominantly affect children and adolescents, with more than half arising in the long bones of the extremities [1–4]. With advances in multidisciplinary care, limb-salvage surgery has become the standard treatment for massive bone defects over amputation, improving quality of life and achieving a success rate of 60–80% [5–7]. After resection of a long-bone tumor, reconstruction of the resultant bone defect is critical to restore skeletal stability and function. Common reconstruction options include endoprostheses, allografts, composite allograft-prosthetic constructs, or recycled autografts using the patient’s own excised bone [8–11]. Each approach has distinct advantages and limitations in the context of limb preservation.
Reimplantation of the resected tumor-bearing bone after sterilization–i.e. a recycled autograft–is an appealing biological option. This technique uses the patient’s native bone (perfectly matching the defect) and avoids donor site morbidity, immunologic rejection, and disease transmission from allografts [12–15]. It has been especially valuable in settings with limited allograft availability. A critical step in recycled autograft surgery is extracorporeal tumor inactivation to eradicate any residual cancer cells before reimplantation. Several sterilization techniques have been developed for this purpose, including high-temperature methods (autoclaving or pasteurization), high-dose extracorporeal irradiation, and cryogenic freezing with liquid nitrogen, while aiming to preserve the bone’s structural integrity. Studies have shown that, when properly performed, these sterilization methods can effectively eliminate viable tumor cells, with a generally low risk of local recurrence attributable to the reimplanted graft [14, 15]. Recycled autograft limb salvage thus offers the promise of anatomic restoration using the patient’s own bone without substantially compromising oncologic safety.
Despite these advantages, the long-term graft survival of recycled autografts remains a concern. Devitalizing and reimplanting the bone can render it more brittle and biologically inert, which may lead to delayed union, non-union, or graft fracture during the healing period [13, 15]. Infection is another known risk, as with any large bone graft procedure. Tumor recurrence within or around the graft, though uncommon, can also threaten the construct [16]. Any of these complications can result in graft failure. Reported outcomes in the literature vary widely: some series have found that over 80% of recycled autograft reconstructions remain survived at 5 years [17–19] whereas other reports describe 5-year graft survival rates of only 30–50% [20, 21]. This variability likely reflects differences in patient populations, tumor location, and definitions of “failure” used across studies. Such heterogeneity highlights the uncertainty in prognostication and the need for more evidence to guide surgeons on the expected durability of recycled autografts.
To date, evidence on the long-term outcomes of massive recycled autologous bone grafts used for limb salvage in long-bone sarcomas remains limited. Most available data originate from single-institution retrospective cohorts or small case series focusing on specific sterilization techniques, making it difficult to compare graft survival outcomes across different methods. As a result, a comprehensive synthesis of graft survival, failure patterns, and associated prognostic factors across sterilization techniques is still lacking. Our study aims to systematically aggregate data from all available evidence to provide insights into the viability of the recycled autograft and factors that influence their survival over time.

Methods
We conducted a systematic review and meta-analysis to synthesize graft survival probability and mode of failure of the recycled autograft. In addition, we performed individual participant data (IPD) meta-analysis to synthesize individual evidence on factors associated with outcomes following recycled autograft reconstruction in long-bone sarcomas.

Study registration and systematic searching
The protocol of this systematic review and meta-analysis of individual patient data was registered in the International Prospective Register of Systematic Reviews (CRD42021245290). This study was reported in compliance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis for Individual Patient Data systematic reviews (PRISMA-IPD) [22] and the checklist is provided in Supplemental Table 1.
We performed a systematic search of observational studies that reported survival outcomes and follow-up time in PubMed, Embase, and Scopus from their inception to June 2, 2025, without language restriction. The search term used a combination of keywords related to autograft, recycle, bone sarcomas, limb salvage, bone reconstruction and relevant controlled vocabulary. The full search strategy for each database is reported in Supplemental Table 2.

Record screening and selection criteria
All records from each database were combined, and duplicates were removed. Titles and abstracts were independently screened by two authors (JK and DP). Subsequently, the full texts of relevant records were retrieved and independently reviewed. In cases of disagreement, the full texts of discordant articles were kept for a subsequent discussion with clinical specialists (DP, SP, RM). Exclusions of full-text articles were recorded and described in the study flow diagram.
We included observational studies, including both comparative studies and single-arm case series, conducted in any setting, provided that each study reported at least five patients with long-bone sarcoma treated using recycled autografts. Case reports and studies with fewer than 5 cases were excluded. Cases involving benign bone tumors, metastatic bone tumors, sarcomas of non-long bones (such as the pelvis, spine, hand, and foot), and cases involving autograft prosthesis composites were excluded. Studies that reported graft survival along with the corresponding survival time were included. The key elements of study setting, populations, interventions, and outcomes were demonstrated in the PICOTS format (Supplemental Table 3).

Outcomes and definitions
The primary outcome of this study was graft survival, which represents functional outcome of recycled autografts in long-bone sarcoma. Graft failure was defined as any case in which the recycled autograft was removed, regardless of whether revision surgery or conversion to endoprosthesis, arthrodesis, rotationplasty, or amputation was subsequently performed. As definitions of graft failure varied across studies, a summary of study-specific definitions is provided in Supplemental Table 4. The number of failures and graft survival rates were extracted at different time intervals, including 1-year, 2-year, 5-year, and 10-year. The follow-up time was recorded, and the mortality status occurring before graft failure was considered as competing risk. The time to failure was defined as the recorded time when graft removal occurred, when amputation or conversion to an endoprosthesis was performed, or when graft survival time was reported. Time-to-event data were extracted only when individual patient–level information could be clearly identified from the original articles, including tables, text descriptions, figures, or supplementary materials, allowing unambiguous linkage of survival time to a single patient. Studies reporting only aggregated survival outcomes without patient-level attribution were not included in the individual participant data analysis. Secondary outcomes included the mode of graft failure and associated prognostic factors. Complications leading to failure of biological reconstruction were reviewed and categorized according to the Henderson classification [23]. This classification includes type 1 soft–tissue failure, type 2 graft–host nonunion, and type 3 structural failure, which are grouped as mechanical type; and type 4 infection and type 5 tumor progression, which are combined as non-mechanical types. Prognostic factors were pre-defined based on literatures and expert opinion [24, 25], including age less than 13 years old, gender, chemotherapy, histological subtypes, devitalization (sterilization) methods, and type of bone and resection.

Data extraction
Data from eligible studies were extracted using a standardized data collection form. At the study level, we extracted data as follows: study characteristics, including, the first author, year and country of publication, recruitment period, study design and setting, total sample size, and devitalization methods. At the individual participants’ level, the graft failure status and follow-up time for each patient were extracted from tables presenting patient characteristics, results, outcomes, or from the text describing functional and oncological outcomes. We collected patient characteristics, including the histological type of bone sarcoma, sex, age at diagnosis, bone site (i.e., femur, tibia, humerus, ulnar, radius), tumor location (i.e., proximal, distal, diaphysis), and reconstruction types (i.e., intercalary, osteoarticular, hemicortical); clinical outcomes, including graft survival with time and mode of graft failure.
Graft failure information was extracted from details related to graft removal, graft status/outcome, reconstruction outcome, complications, reoperation or revision details, and additional procedures. Only studies that reported follow-up time data for all included cases were included in the pooled IPD analysis.
The data was individually extracted and compiled using an electronic extraction form by two authors, JK and DP. The database was checked for any errors and was predefined for analysis by LL, SN and PP.

Risk of bias assessment
JK, LL, SN and DP assessed the methodological quality of the included case series in the systematic review and meta-analysis using the convergent criteria developed by Murad et al. [26]. These criteria were adapted from modifications of the Pierson, Bradford Hills, and Newcastle Ottawa scale. Each article was evaluated based on four domains: selection, ascertainment, causality, and reporting, which were further divided into five criteria. We used binary questions with yes/no options to determine whether the report met the criteria. The specific five questions were described in Supplemental Table 5.

Evidence synthesis and statistical analysis
All statistical analyses were performed using STATA version 16 (StataCorp, College Station, TX, USA) and R version 4.4.1. Qualitative synthesis was performed for demographic data, geographical regions, surgical methods, and modes of failure.
For quantitative analysis, graft survival probabilities for each individual study at four-time intervals were pooled using total number of participants and number of graft failure events. The pooled graft survival probability along with its 95% confidence interval (95%CI) was estimated using the metaprop command with a random effect model [27]. Heterogeneity was assessed using I² statistic. Subgroup analysis was performed for devitalization methods. Sensitivity analysis when studies with small sample size (less than 10 patients) were conducted. Publication bias among the included studies was evaluated using the Doi plot and the Luis Furuya-Kanamori (LFK) index [28].
We further conducted a meta-analysis for IPD to identify prognostic factors associated with graft survival. In addition, graft survival probability at 5-years and mean graft survival time were estimated. Major prognostic factors analyzed including age, sex, sterile techniques of the autograft, chemotherapy received, type of bone resection, and tumor length more than 15 centimeter. Missing values of those factors were handled using multiple imputation by chained equations (MICE), assuming missing at random (MAR) as missingness was primarily related to incomplete reporting in the original publications. A total of 30 imputed datasets were generated, the imputation process was performed in Stata, and imputed datasets were analyzed using Rubin’s rules [29, 30].
Univariable analysis was performed using a flexible parametric survival model to determine hazard ratio (HR) and 95%CI, with cluster variance correction and study-based stratification. Participant mortality was considered as competing risk and restricted mean survival time (RMST) was calculated at 60 months, representing the mean graft survival time for each prognostic factor over the time horizon [31, 32]. Multivariable analysis for HR and 95%CI was also performed adjusting for all prognostic factors. To assess the robustness of our imputation approach, we also performed sensitivity analyses using the complete case analysis approach.

Grade certainty of evidence
The certainty of evidence was evaluated according to the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework [33], which classifies certainty as high, moderate, low, or very low, considering factors such as study design, risk of bias, inconsistency, indirectness, imprecision, and publication bias. Meta-analysis results were then interpreted with GRADE rating guidance [34], taking into account the quality and certainty of evidence.

Results

Study selection and characteristics
A total of 2833 studies were identified through a systematic search. After eliminating 878 duplicates and conducting a manual search of references and reviews, 1,955 reports remained. Following the screening process, 208 full texts remained for full-text review that were eventually evaluated for eligibility. We excluded 160 studies that did not meet the inclusion criteria described in PICOTS (Supplemental Table 6) and 20 studies that did not provide individual patient data (Supplemental Table 7). Following full-text reviewing and IPD availability, 28 articles were eligible (Fig. 1), consisting of 21 (75.0%) case series and 7 retrospective cohorts. Table 1 shows the characteristics of eligible studies. All studies were conducted in a single-center setting. For devitalization methods, 10 studies (35.7%) used ECIA, 9 (32.1%) used pasteurization, 8 (28.6%) used liquid nitrogen and only one study (3.6%) used alcohol.

While most of studies were conducted in Asia, only a study was reported in Africa (Egypt) and 2 studies in Europe (United Kingdom and Germany). The majority of studies were conducted in Japan (13 studies, 46.4%), followed by China (7 studies, 25.0%), Thailand (2 studies, 7.1%) and South Korea (2 studies, 7.1%) (Fig. 2).

Individual participant characteristics
From 28 eligible studies, a total of 527 participants were identified. Of these, 132 were excluded for not meeting the population-of-interest criteria (71 did not use recycled autografts; 29 did not have long-bone sarcoma; 22 did not have malignant bone tumor ; 7 had neither long-bone sarcoma nor recycled autograft; 3 had neither long bone nor malignant bone tumor), resulting in 395 patients included in the analysis. More than 90% of cases were from Asia, while the others were from United Kingdom, Egypt and Germany (Fig. 2). The median age of all patients was 20 years (interquartile range (IQR) 15–34 years) and 56.0% of them are male. The osteosarcoma subtype constituted the largest proportion in this study (62.5%), followed by Ewing’s sarcoma (8.9%) and chondrosarcoma (6.3%) (Supplemental Fig. 1). There were 159 cases (40.2%) that used ECIA [25, 35–43], 132 cases (33.4%) used pasteurization [10, 44–51], 98 cases (24.8%) used frozen technique [19, 52–58], and 6 cases (1.52%) used alcohol-inactivation [59]. Most of the patients had bone lesions located in the distal and diaphyseal femur (34.1%) and proximal and diaphyseal tibia (27.6%). The proximal humerus accounted for 10.1%, and the distal tibia accounted for 7.8%.

Risk of bias assessment
The Supplemental Table 5 demonstrates a list of the reports and their results of quality assessment. More than half of the published reports (16 studies, 57.1%) were classified as high quality, meeting all the quality criteria for our study design. Twelve studies (42.9%) were classified as moderate quality. All studies adequately reported recycled autograft outcomes, and the follow-up periods were long enough to determine these outcomes. A high risk of bias in the patient selection domain was observed in moderate-quality reports. These studies recruited cases based on specific objectives, including bone location, reconstruction type [37, 43, 48–51, 58], histological type [25, 52, 59], age group [19], and patients with long-term follow-up of over 10 years [40].

Graft survival probability estimated at study level
The pooled estimates for graft survival probabilities at 1 year (Supplemental Fig. 2), 2 years (Supplemental Fig. 3), 5 years (Fig. 3), and 10 years (Supplemental Fig. 4) were 99.1% (95%CI 96.8–100.0; I2 22.0%), 96.8% (95%CI 92.6–99.5; I2 49.9%), 91.0% (95%CI 85.4–95.6; I2 49.6%) and 87.0% (95%CI 78.8–93.8; I2 61.6%), respectively (Supplemental Table 8). Subgroup analysis reveals that the pasteurization method for graft devitalization had the lowest failure probability while ECIA and liquid nitrogen demonstrated comparable failure probabilities (Supplemental Table 8 and Supplemental Figs. 5, 6, 7, 8). Sensitivity analysis when small studies with less than 10 patients were excluded, showed consistent results Supplemental Table 9.

For publication bias analysis, the LFK index at 1-year, 2-year, 5-year, and 10-year periods were 0.20, − 1.00, − 1.15, and − 0.50, respectively. The plots showed no asymmetry in outcomes estimated at 1-year and 10-year periods, while minor asymmetry was observed in 2-year and 5-year periods (Supplemental Fig. 9). Certainty of evidence was evaluated for survival probabilities at 1, 2, 5, and 10 years. The rate was initially set at low quality due to the observational study design, and some outcomes were downgraded due to the following factors. For 2- and 5-year survival, the certainty was downgraded to very low since these outcomes showed minor asymmetry in publication bias. For 10-year survival, the heterogeneity warrants some concern due to a high I2 value, leading to very low certainty of evidence.

Graft survival rate and associated prognostic factors
Using individual participant data from 395 patients, the median follow-up time was 67 months (IQR 40–120 months). During this period, 70 patients (17.7%) died, resulting in the patient survival rate of 82.4% (77.9–86.0) at 5-year. In addition, based on time-to-event analysis, 58 graft failure events were observed, accounting for 84.3% of graft survival probability at 5 years (95%CI 80.0–87.8). Supplemental Fig. 10 demonstrates Kaplan-Meier plot for graft failure at 5 years.
From pre-defined prognostic factors, all factors were found to be significantly associated with graft failure in univariable analysis (Table 2). Male sex was significantly associated with poor survival, with males having a more than twofold increased risk of graft failure compared to females. This corresponded to a mean graft survival time that was over 7 months shorter for males (45.6 months) than for females (53.0 months). Similarly, younger age (0–13 years old), histology subtypes, sterile techniques, chemotherapy, and types of bone resection were significantly associated with graft survival with varied mean graft survival time. The graft survival probabilities over 60 months, based on different prognostic factors, are illustrated in Fig. 4.

To evaluate whether these associations persisted after adjustment for other measured variables, variables were included in a multivariable analysis (Table 2). After adjusting, only male gender and type of bone and resection remained significant prognostic factors of graft failure. Male gender demonstrated strong association for poor graft survival with hazard ratio of 2.46 (95%CI 1.62–3.72; p < 0.001). Types of bone and resection, including distal femur–osteoarticular (HR = 3.83; 95%CI 2.08–7.11; p < 0.001) and proximal tibia–osteoarticular (HR = 3.75; 95%CI 2.15–6.56; p < 0.001) resections, also remained strongly and independently associated with poor graft survival (Table 2; Fig. 4). In addition, sensitivity analysis using complete case approach revealed consistent results (Supplemental Table 10).

Mode of graft failure
From 395 cases, there were 58 graft failure events during the follow-up period. Of these, infection was the most common mode (44.8%), followed by graft breakage or fracture (27.6%), and tumor progression at resection site (18.9%). The other mechanical modes of failure were observed in 5 cases, including 3 cases (5.2%) of graft-host nonunion and 2 cases (3.4%) of dysfunctional soft tissue (Fig. 5). Over the follow-up period, infection accounted for the highest proportion of failures during the first year (25.9%) and decreased steadily to 6.9% and 1.7% at 5 and 10 years, respectively. Structural failures were less common in the early period (5.2% at 1 year) but increased to 13.8% at 5 years before declining thereafter. During the sixth to tenth year period, graft failures were less frequent compared to the early period, with only 2 (3.4%) failures reported. In addition, soft tissue failures and graft–host nonunion were relatively rare, accounting for less than 5% at any time point (Fig. 5).

In relation to type of bone and resection, the failure rates varied according to bone location and reconstruction type (Supplemental Table 11). The highest rate of graft failure was observed in the bones surrounding the knee with osteoarticular resection (40.5%), followed by proximal tibia–osteoarticular reconstructions (37.2%), and proximal humerus–osteoarticular reconstructions (15.8%). Infection predominated in proximal tibia (68.8%) and distal femur (53.3%) failures, while tumor progression was most common in hemicortical/intercalary resections (41.2%). Structural failure was frequent in proximal humerus (50%) and proximal tibia (31.2%) reconstructions, whereas soft tissue failure and graft–host nonunion were rare.
Failure patterns differed according to the method of tumor-bearing bone sterilization (Supplemental Table 12). Recycled autografts treated with ECIA demonstrated the highest proportion of failures (22.0%), followed by liquid nitrogen (12.2%) and pasteurization (7.6%). Alcohol-treated autografts demonstrated a failure rate of 16.7%, although interpretation is limited by the small sample size. Across all sterilization methods, infection emerged as the leading cause of graft failure, particularly among ECIA-treated grafts (54.3%) and pasteurized grafts (40.0%). In contrast, structural failure was the predominant failure mode in liquid nitrogen–treated autografts (58.3%) and accounted for all failures in the alcohol-treated group. Tumor progression at the resection site occurred most frequently following pasteurization (40.0%) and ECIA (14.3%). Failures related to soft tissue complications and graft–host nonunion were infrequently observed across all groups.

Discussion
This systematic review and individual participant data meta-analysis represents, to our knowledge, the most comprehensive evaluation of recycled autograft survival in limb salvage surgery for long-bone sarcomas to date. By pooling data from 28 studies involving 395 patients, we estimate graft survival outcomes and associated prognostic factors. Literatures reveal that this reconstruction technique is widely used, especially in Asian countries such as Japan, China, and India. Our study suggested that recycled autografts may provide a favorable long-term biological reconstruction option. We identified infection as the principal mode of failure, particularly in the early postoperative period. Furthermore, our multivariable analysis identified male sex and osteoarticular reconstructions involving the distal femur or proximal tibia as factors independently associated with graft failure.

Graft survival rate of recycled autografts
In our study, graft survival using meta-analysis yielded 91.0% survival at 5 years, whereas an IPD-based flexible parametric survival model estimated 84.3%. While the IPD time-to-event analysis, which accounts for censored follow-up times and competing risks, can provide more conservative estimates, both approaches suggest a generally favorable pattern of graft survival. Our analysis aligns with the outcomes reported in several studies demonstrating survival rates at 73–87% [17, 18, 60, 61] with small discrepancies that likely resulting from patient populations, case mix, surgical techniques and definitions of failure. A similar trend was also observed in other limb salvage surgery methods, suggesting that recycled autografts may provide graft survival outcomes that are broadly comparable to both allografts (73.2–94% graft survival) [17, 62–64]. and endoprostheses (52–75%) [65, 66].
From our eligible studies, three dominant techniques were observed including extracorporeal irradiation (ECIR), pasteurization, and liquid nitrogen freezing. In our subgroup analysis, pasteurization showed the highest 5-year graft survival, though differences among techniques were not statistically significant after adjusting with relevant prognostic factors. On one hand, irradiation and pasteurization can maintain bone biomechanical strength and allow for bone union, with pasteurization perhaps better preserving osteocyte viability and marrow elements [10, 12, 67]. On the other hand, most studies report clinically comparable results across methods [12, 18, 61]. Alcohol devitalization was reported in only a handful of cases, reflecting that this method is seldom used clinically and hindering firm conclusions. Overall, all techniques appeared to provide acceptable long-term outcomes, and the choice may depend more on institutional expertise and resources than intrinsic differences in graft survival. Given the very low GRADE certainty ratings, however, these comparative impressions should be regarded as exploratory.

Failure of recycled autografts
Our study found that infection was the predominant mode of failure and most of them occurred within the first 2 years postoperatively. This aligns with the well-recognized challenge of infection in large bone reconstructions due to devitalized, avascular nature of the recycled autograft that makes it susceptible for bacterial colonization, particularly in the context of extensive surgery and soft tissue dissection [60]. In the early postoperative period, infections are likely driven predominantly by perioperative factors such as operative duration, the extent of soft-tissue dissection, wound size, and the adequacy of soft-tissue coverage over the graft. This evidence suggests that acute post-surgical factors such as healing, and infection play a dominant role in graft failure for the recycled bones. In contrast, structural failures generally arose later and at lower frequency. Atraumatic fractures were often observed as a consequence of devitalization techniques that weaken the mechanical properties or due to a compromised state during the bone remodeling process. Additionally, traumatic fractures occurred even when patients disregarded the advice of physicians to avoid weight-bearing activities [35, 68]. Later infections may be more closely related to systemic factors, including chemotherapy-related immunosuppression, transient bacteremia, and patient-level hygiene or skin integrity, although these mechanisms could not be fully dissected from the available data. These patterns mirror previously reported modes of failure, in which infections dominate early [60, 61], but contrast with reports on allograft or endoprosthetic reconstructions, where tumor progression (53% in allografts [63]; 40% in endoprostheses [66]) and mechanical failure (73.6% [69]) are more frequently observed as the dominant failure modes. In addition, tumor progression at resection site was a relatively infrequent failure mode in our study (19%) and the rate was comparable to other studies using allograft and endoprostheses. A study showed that among 91 patients with allograft bone resection, 49 complications occurred and 26% of them had tumor progression [63]. Like another study in patients who received a large endoprosthesis for tumor resection, 17.4% of all failure were classified as tumor progression [65]. This implies that the devitalizing process used to prepare recycled autografts effectively eliminates cancer cells, supporting the oncologic safety of these techniques, although our data do not allow definitive conclusions about residual local recurrence risk [16, 63, 65, 66].

Prognostic factors for graft survival
In our analysis male sex was significantly and independently associated with poorer graft survival. This finding consistent with prior evidence that male gender has been associated with impaired bone healing and higher complication rates in biological reconstructions [61]. Several biological and behavioral explanations may account for this disparity. Boys and young men typically exhibit greater and more prolonged skeletal growth than girls [70]. The higher growth potential increases the risk of limb length discrepancy (LLD) after complications, which can in turn destabilize the graft and predispose it to failure. Besides, a higher level of activity and intensity of exercise in males or younger patients could contribute to graft failure [71]. This was reflected in the higher rate of structural failures among males which 11 cases (26.8%) of the 41 total male failures involved structural complications, while females accounted for 5 (23.8%) among 17 female failures.
We found that the type of reconstruction are strongly associated with graft survival which are consistent with that in other studies using recycled autografts and endoprostheses [19, 25, 40, 47, 65]. We also observed that failure mode distributions also varied by anatomical sites. Osteoarticular reconstructions – where a devitalized bone graft replaces a joint segment such as distal femur or proximal tibia with joint surface – had substantially higher failure rates than intercalary or hemicortical grafts. Osteoarticular grafts are more complex for which devitalized subchondral surface cannot achieve biological union and rely mainly on a prosthetic component for joint function. In our study, while the proximal humerus had fewer failures due to non-weight bearing site [14], those failures that did occur were often mechanical, potentially owing to the difficulty of fixation in small metaphyseal bone and the high mobility at the shoulder. In addition, the devascularized joint surface may predispose to subchondral bone resorption and collapse over time, and the need for extensive soft tissue dissection can increase infection risk [72]. Our proximal tibia reconstructions showed strong evidence for early infection with nearly 69% of proximal tibia graft failures were infection related. This reflects the challenges of tibial reconstructions, where limited soft-tissue coverage elevates infection risk [61]. Consistent with this, prior studies have reported that patients treated with osteoarticular autografts experience higher complication and reoperation rates compared with those undergoing hemicortical or intercalary reconstruction [40]. Further, intercalary grafts and hemicortical resections preserve native joints and tend to heal more reliably which results in the lowest mechanical failure incidence. However, patients underwent endoprothesis limb salvage surgery showed higher rate of mechanical failures in intercalary and hemicortical area [73].
Failure patterns appeared to vary according to the sterilization method, suggesting potential differences in the biological and mechanical characteristics of recycled autografts. The higher failure proportion observed in ECIA may partly reflect the larger case volume and longer follow-up in this group. Infection accounted for a substantial proportion of failures and may represent a general challenge of massive bone reconstruction, with perioperative and biological factors potentially playing a greater role than the sterilization technique itself [60]. Structural failure was more frequently observed in liquid nitrogen–treated grafts, which may be related to microstructural changes associated with freeze–thaw procedures [74]. Pasteurization demonstrated a relatively lower overall failure rate; however, a higher proportion of tumor progression at the resection site was observed. Although pasteurization has been reported to preserve osteocytes and bone marrow cellularity and to maintain bone collagen structure, thereby providing favorable overall outcomes [12, 75, 76], thermal exposure at approximately 60℃ may be associated with variability in tumor cell eradication, suggesting a potential trade-off between biological preservation and oncologic safety. Failures in alcohol-treated autografts were mainly structural in nature, although interpretation is limited by the small sample size. Overall, graft failure appears to result from a combination of surgical, biological, and sterilization-related factors.

Clinical implications
Our results suggest that recycled autografts may have comparable graft survival outcomes with other limb salvage reconstructions, particularly in settings where allograft procurement is difficult and cultural preferences favor biological reconstruction. This evidence provides insights for treatment planning and may help stratify patients and tailor surgical strategies. A male patient requiring a distal femur osteoarticular reconstruction represents the highest-risk profile identified in this study, while intercalary or hemicortical resection possess better graft survival outcome. This suggests that patients undergoing limb salvage surgery at intercalary region may be considered for autograft as alternative. Our results may warrant consideration of augmented fixation techniques, and a more conservative postoperative weight-bearing protocol. Furthermore, our findings that infection was the most common mode of failure highlight the importance of meticulous infection-prevention strategies at multiple stages of care. Intraoperatively, careful handling of devitalized bone, minimization of dead space, stable fixation, and multilayered soft-tissue coverage or flap reconstruction when needed may reduce early infection risk, particularly around the proximal tibia and distal femur where soft-tissue envelopes are thin. Perioperative antibiotic protocols should be tailored to wound class and anticipated contamination, and extended prophylaxis may be considered in selected high-risk patients. Antibiotic use should be tailored according to the wound classification. When determining the duration of antibiotic therapy, the timing of wound drainage removal should be taken into consideration [14]. During chemotherapy and rehabilitation, close surveillance for wound problems, optimization of nutrition, and graded weight-bearing may help mitigate both late infection and structural failure.
Ultimately, our results support the view that surgeons may select a devitalization method based on institutional resources and personal experience without major concern that this choice alone will compromise the graft’s survival. Nevertheless, these implications should be applied cautiously with regards to observational data from case series and retrospective cohorts.

Strengths and limitations
This study represents the first systematic review and IPD meta-analysis to comprehensively evaluate recycled autograft survival in limb salvage surgery for long-bone sarcomas, using a large sample size and extended follow-up period. However, some limitations must be acknowledged. First, the meta-analysis is based on observational data, predominantly retrospective case series, which carry substantial risk of patient selection and reporting bias. Surgeons may have preferentially selected recycled autografts for good prognosis patients, and postoperative complications or long-term follow-up outcomes might not have been consistently recorded or published. These limitations may hinder the ability to draw firm conclusion and may overestimate or underestimate the true effectiveness of recycled autografts. Second, substantial heterogeneity across the included studies were observed. Patients differed in baseline demographics such as tumor histology, anatomical sites, and adjuvant treatments such as chemotherapy varied by institution. This makes it challenging to directly pool and compare results. Our subgroup analysis was only applicable for devitalization methods. Third, despite the advantages of IPD meta-analysis, our dataset lacked information for key prognostic variables, such as cancer stage, soft-tissue adequacy, immunosuppression and resection margins. These factors are known to strongly influence graft survival and complication rates. Their absence limits our ability address confounding and to claim independent causal effects for the associations observed in multivariable models. Fourth, definitions of “graft failure” and the timing and classification of complications were not fully standardized across primary studies. Although we applied a broad, inclusive definition of graft failure and a comparison of criteria across studies, some reports did not explicitly specify identical criteria, and misclassification between mechanical and non-mechanical modes of failure is possible. In addition, microbiological data, early versus late infection, and chemotherapy exposure were often incompletely reported, further contributing to heterogeneity. Together, these sources of misclassification and residual confounding mean that our findings only provided a suggestive factors associated with poor graft survival. These limitations highlight the necessity of large, well-structured, prospective cohort studies to validate these findings. Finally, the preponderance of studies from Asia with over 90% of cases originating from Asian countries, especially Japan and China. While this reflects the long-standing clinical adoption of recycled autografts in Asia, it raises questions about generalizability, including potential differences in the availability of cadaveric allografts, insurance coverage and resource constraints, cultural preferences for biological versus prosthetic reconstruction, and regional surgical techniques and rehabilitation pathways. Outcomes and decision-making in settings where allografts and megaprostheses are more readily accessible may therefore differ from those observed in this largely Asian cohort.

Conclusion
Recycled autografts are frequently used in limb salvage surgery for bone sarcoma, particularly in Eastern countries, but evidence on long-term functional outcomes remain limited. In this study, we observed a favorable 5-year autograft survival rate, comparable to other reconstruction techniques, although these estimates are based on observational data with very low certainty of evidence and should be interpreted cautiously. Our IPD meta-analysis showed that male sex and osteoarticular reconstruction were significantly associated with poorer graft survival comparing to intercalary resection which conferred more favorable outcomes. Infection was the most common cause of graft failure, while soft-tissue failure, non-union, and local tumor recurrence were rarely observed. These findings provide clinical insights that may guide resection type selection and optimize surgical planning in limb salvage using recycled autografts.

Supplementary Information
Below is the link to the electronic supplementary material.