Comparing simultaneous and staged bilateral total knee arthroplasty: a systematic review and meta-analysis

Abstract

Objective

For patients with bilateral knee osteoarthritis, staged bilateral total knee arthroplasty (StaBTKA) is generally considered lower in perioperative risk, while simultaneous bilateral total knee arthroplasty (SimBTKA) enables one-stage deformity correction and potential earlier rehabilitation, with its safety debated. This study aimed to systematically compare their clinical outcomes using available observational evidence.

Methods

PubMed, Embase, the Cochrane Library, Ovid, and Web of Science were systematically searched for studies published between January 1, 2000 and August 31, 2025. Cohort studies comparing SimBTKA and StaBTKA were included, and all eligible studies were observational in design.

Results

A total of 53 cohort studies involving 572,881 patients were included, comprising 244,207 patients in the SimBTKA group and 328,674 in the StaBTKA group. In studies reporting comparable short-term mortality windows, SimBTKA was associated with higher reported short-term mortality than StaBTKA (OR = 2.35, 95% CI 1.69–3.27, P < 0.001). SimBTKA was also associated with a higher incidence of deep vein thrombosis (OR = 1.45, 95% CI 1.37–1.53; P < 0.001) and greater transfusion requirements (OR = 4.42, 95% CI 3.11–6.28; P < 0.001). Length of stay tended to be shorter with SimBTKA, particularly in Asian studies; however, substantial heterogeneity limited the interpretability of a precise pooled estimate. In contrast, StaBTKA was associated with higher rates of superficial and deep infection (OR = 0.69, 95% CI 0.54–0.89; P = 0.004; OR = 0.65, 95% CI 0.61–0.70; P < 0.001). No significant differences were observed between the two strategies in revision rates, pulmonary embolism, neurological complications, or cardiac events.

Conclusion

SimBTKA and StaBTKA have distinct risk profiles rather than one being universally superior. SimBTKA was linked to higher reported short-term mortality, DVT, and transfusion, whereas StaBTKA was associated with higher infection rates. The mortality finding should be interpreted cautiously because survivor selection bias may underestimate true risk in the StaBTKA group. Surgical strategy should be individualized, balancing patient comorbidities, perioperative safety, treatment efficiency, and healthcare resource utilization.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13018-026-07039-0.

Background

Advanced knee osteoarthritis is frequently accompanied by bilateral varus or valgus malalignment. In patients with bilateral disease, the choice between simultaneous bilateral total knee arthroplasty (SimBTKA) and staged bilateral total knee arthroplasty (StaBTKA) remains a subject of ongoing clinical debate. StaBTKA is often favored because it is perceived to carry a lower perioperative risk. However, this approach may not fully account for the potential biomechanical changes that occur during the interstage period. After unilateral correction in staged surgery, a previously bilaterally symmetric lower-limb alignment becomes temporarily asymmetric, which may alter mechanical axis loading and limb length balance. Prior biomechanical and gait studies have suggested that such asymmetry could theoretically influence postoperative stability, load distribution, and contralateral joint mechanics [1, 2]. Importantly, these biomechanical considerations are largely derived from observational and experimental studies and have not been consistently evaluated as clinical outcomes in comparative cohort studies of SimBTKA and StaBTKA. The optimal surgical strategy for patients with bilateral knee osteoarthritis remains unclear. Most existing comparative studies focus on perioperative safety, complications, and resource utilization, while biomechanical factors are seldom incorporated into outcome assessment. Therefore, this meta-analysis compares the clinical outcomes of SimBTKA and StaBTKA based on existing cohort studies and discusses the results within the broader context of biomechanical theory, providing a reference for clinical practice.

Methods

Protocol registration

This systematic review was conducted and reported in strict adherence to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [3]. The study protocol was registered with the International Prospective Register of Systematic Reviews (PROSPERO) under the registration number CRD420251020982.

Literature search

A comprehensive literature search was performed independently by two reviewers (Hongzhao Wang AND Yabing Jiang) in PubMed, Embase, Cochrane Library, Ovid, and Web of Science databases. Studies published between January 1, 2000, and August 31, 2025, comparing the prognosis of patients undergoing simultaneous versus staged bilateral total knee arthroplasty (SimBTKA vs. StaBTKA) were included. The search strategy used the following Boolean string: (((simultaneous) OR (staged)) AND (bilateral)) AND ((Arthroplasty, Replacement, Knee) OR (Total Knee Arthroplasty) OR (Knee Arthroplasty) OR (Unicompartmental/Partial Knee Arthroplasty)). No language restrictions were imposed. Initial screening was conducted independently by the two reviewers to identify eligible studies. Discrepancies during screening were resolved through consultation with a third senior investigator (Tianyun Zhao) to ensure the rigor and objectivity of the selection process.

Study population

The study population consisted of patients undergoing primary bilateral total knee arthroplasty (TKA) performed via either a simultaneous (SimBTKA) or staged (StaBTKA) protocol.

Study selection

This review was restricted exclusively to clinical cohort studies comparing SimBTKA and StaBTKA published between January 1, 2000 and August 31, 2025. Meta-analyses were explicitly excluded from the analysis.

Data extraction and management

Citations were deduplicated using EndNote X9. Two independent reviewers (Hongzhao Wang AND Xiao Wang) screened records by title/abstract, followed by full-text review. Data extraction included study characteristics (demographics, sample size), clinical outcomes (mortality, revision, deep vein thrombosis, pulmonary embolism, infections, cardiac/neurological complications, blood loss/transfusion), and economic metrics (Length of Stay, costs).

Study definitions

Procedures performed on the same day under a single anesthetic episode were assigned to the SimBTKA group. Procedures performed on different days with at least a one-day interval, or under separate anesthetic episodes and clearly described as staged, were assigned to the StaBTKA group. Short-term mortality was defined as death reported within 30 or 90 days postoperatively. Revision was defined as any reoperation involving removal, exchange, or revision of prosthetic components, extracted according to the time frame reported by each study. Pulmonary embolism and deep vein thrombosis were extracted according to the diagnostic definitions used in the original studies. Superficial infection referred to infection limited to skin or subcutaneous tissue, whereas deep infection referred to infection involving deep tissues, joint space, or prosthetic components, broadly consistent with CDC surgical-site infection criteria. Neurological complications included stroke, cerebral infarction, transient ischemic attack, delirium, altered consciousness, or other reported neurological events. Cardiac complications included myocardial infarction, heart failure, arrhythmia, cardiac arrest, or other major cardiac events. Blood transfusion was defined as receipt of perioperative or postoperative blood transfusion. Length of stay was extracted as reported; for StaBTKA, cumulative LOS across both admissions was used when available.

Assessment of risk of bias and quality of evidence

Risk of bias was assessed independently by two reviewers (Hongzhao Wang and Yabing Jiang) using the Cochrane ROBINS-I tool for non-randomized studies of interventions. Disagreements were resolved through discussion or consultation with a third reviewer (Tianyun Zhao). The ROBINS-I tool evaluates seven domains, which were judged as low, moderate, serious, or critical risk of bias. A study was rated as low risk if all domains were low; moderate risk if at least one domain was moderate but none was serious or critical; serious risk if at least one domain was serious but none was critical; and critical risk if at least one domain was critical. A ROBINS-I traffic-light plot was generated to summarize domain-level judgments across studies. The Newcastle–Ottawa Scale (NOS) was additionally used as a supplementary methodological quality assessment for cohort studies. NOS scores range from 0 to 9 and cover selection, comparability, and outcome assessment. The detailed NOS assessment is presented in Supplementary Table. The certainty of evidence for major outcomes was assessed using the GRADE approach. The detailed GRADE certainty assessment is provided in Supplementary Table.

Statistical analysis methods

Binary and continuous outcomes were expressed as odds ratios (OR) and mean differences (MD), with 95% confidence intervals (CI) provided. Statistical significance was set at P < 0.05. Heterogeneity was assessed using the I² statistic and the Cochrane Q-test; if the criteria were met (P ≥ 0.1 and I² ≤ 50%), a fixed-effects model was used; otherwise, a random-effects model was used. For study results showing significant heterogeneity, we conducted additional subgroup analyses and meta-regression to explore potential sources of between-study differences. Study period, data source, and adjustment methods were selected as study-level covariates. The study period was categorized into three groups: ≤ 2010, 2011–2019, and ≥ 2020. Data sources were classified as administrative databases/registries or single-center/clinical cohorts. Adjustment methods were defined based on whether multivariate adjustment or propensity score matching was used. Publication bias was assessed using funnel plots and the Egger’s regression test. The certainty of evidence for each outcome was assessed using the GRADE framework. Detailed GRADE ratings are provided in Supplementary Table. Analyses were performed using RevMan 5.3 and Stata 15.0 software.

Results

Literature search results

A total of 3810 potentially relevant records were identified across the five databases. After screening, 53 studies met the inclusion criteria (Table 1) and were included in the final qualitative and quantitative analyses (Fig. 1). The 53 included studies comprised 572,881 patients, of whom 244,207 underwent simultaneous bilateral total knee arthroplasty and 328,674 underwent staged bilateral total knee arthroplasty.

Table 1.

Details of 53 studies included

Author	Year	Country	Study design	SimBTKA			StaBTKA			Time interval between stages (months)
				N (n)	Female (%)	Mean Age	N (n)	Female (%)	Mean Age
Mangaleshkar et al. [4]	2001	UK	Retrospective	54	61.1	73	34	61.7	71.7	< 12
Ritter et al. [5]	2003	USA	Retrospective	2050	56	69.9	152	77	69.2	16.8
Sliva et al. [6]	2005	USA	Retrospective	26	46	59.3	306	65.4	64.7	3.6
Stubbs et al. [7]	2005	Australia	Retrospective	61	N/A	64	38	N/A	67	< 12
Barrett et al. [8]	2006	USA	Retrospective	8324	57.8	N/A	13,039	62.9	N/A	< 12
Hutchinson et al. [9]	2006	Australia	Prospective	438	44	67	125	63	65	34
Forster et al. [10]	2006	Australia	Retrospective	28	46.4	66	36	50	68	0.23
Walmsley et al. [11]	2006	UK	Retrospective	826	N/A	N/A	1796	N/A	N/A	< 12
Stefánsdóttir et al. [12]	2008	Sweden	Retrospective	1139	59.2	70.4	3432	62.5	71.2	< 12
Hooper et al. [13]	2009	New Zealand	Retrospective	1012	N/A	65	1360	N/A	69	< 12
Yoon et al. [14]	2010	S Korea	Retrospective	119	94.1	70	119	94.1	70	12
Meehan et al. [15]	2011	USA	Retrospective	11,445	53.9	67.2	23,715	61.3	67.7	< 12
Bolognesi et al. [16]	2013	USA	Retrospective	4519	59	73.3	3788	61.3	74.1	< 12
Poultsides et al. [17]	2013	USA	Retrospective	2825	62.4	65.2	1151	67.8	69.5	6.99
Bini et al. [18]	2014	USA	Retrospective	1230	58.4	66	2123	65.8	67	< 12
Courtney et al. [19]	2014	USA	Retrospective	103	63	59.4	131	77	64.2	0.23
Niki et al. [20]	2014	Japan	Retrospective	60	83.8	73	60	83.3	72.3	8.2
Lindberg-Larsen et al. [21]	2015	Denmark	Retrospective	157	52.9	64	628	57.1	66.7	6.3
Zhao et al. [22]	2015	China	Retrospective	54	88.9	66.9	39	87.2	67.2	9
Bohm et al. [23]	2016	Canada	Retrospective	6349	59	64	25,253	61	66	< 12
Seol et al. [24]	2016	S Korea	Retrospective	759	94.3	68.3	315	92.1	66	1.22
Sheth et al. [25]	2016	USA	Retrospective	2814	57.2	64.8	5177	63.1	67.4	< 12
Hadley et al. [26]	2017	USA	Retrospective	371	69.8	N/A	67	64.2	N/A	< 6
Arslan et al. [27]	2018	Turkey	Retrospective	72	73.6	65.4	61	72.1	69.1	< 12
Chua et al. [28]	2018	Australia	Retrospective	23,136	46.2	N/A	12,951	49.4	N/A	< 6
Koh et al. [29]	2018	S Korea	Retrospective	820	95.9	68.6	633	95.9	69.7	< 12
Sobh et al. [30]	2018	USA	Retrospective	225	52	61	337	62	68	< 12
Sun et al. [31]	2018	China	Retrospective	67	71.6	69.4	68	63.2	71.6	0.5
Richardson et al. [32]	2019	USA	Retrospective	1637	55.8	N/A	6110	N/A	N/A	< 12
M. Lindberg‐Larsen et al. [33]	2019	Denmark	Retrospective	232	53.4	64.6	232	53	65	< 6
Tsay et al. [34]	2019	USA	Retrospective	27,301	56.8	65.8	45,419	62.6	66.6	6.03
Wyatt et al. [35]	2019	New Zealand	Retrospective	6440	38.7	N/A	5116	50.7	N/A	< 12
Wyles et al. [36]	2019	USA	Retrospective	188	58	61	242	64	72	< 12
Gill et al. [37]	2020	Australia	Retrospective	122	62.3	70.6	46	50	70.7	7.93
Raymond et al. [38]	2021	China	Retrospective	95	72.6	65.6	80	72.5	68.7	< 12
Eke et al. [39]	2022	Turkey	Retrospective	225	84.9	66.9	51	76	69.5	0.23
Liu et al. [40]	2022	USA	Retrospective	133	60	67.8	239	75	65.4	38.4
Pumo et al. [41]	2022	USA	Retrospective	198	42.9	63	396	47.2	63	< 12
Wilkie et al. [42]	2022	USA	Retrospective	19,334	57.7	65	19,334	57.8	64	4.23
Çelen et al. [43]	2023	Turkey	Retrospective	168	84.5	67.3	63	81	67.1	< 12
Chou et al. [44]	2023	China	Retrospective	1565	77.6	72.2	451	78	71.9	< 12
Erossy et al. [45]	2023	USA	Retrospective	19,382	58.8	65	19,382	59	65	< 12
Kirschbaum et al. [46]	2023	Germany	Retrospective	53	41.5	69.6	74	55.4	68.2	< 13
Ayekoloye et al. [47]	2023	Nigeria	Retrospective	19	94.7	65.8	12	100	69	N/A
Mingxi et al. [48]	2023	China	Retrospective	82	82.9	65.9	75	84	66.1	< 12
Accatino et al. [49]	2024	Italy	Retrospective	43	67	70.2	66	55	64.2	< 12
Ashkenazi et al. [50]	2024	USA	Retrospective	205	63.9	63.3	205	67.3	63.2	< 12
Matsumura et al. [51]	2024	Japan	Retrospective	94	78	76	94	81	76	5
Franceschetti et al. [52]	2024	Italy	Retrospective	65	65	68	108	70	70	< 12
Yalin et al. [53]	2024	Turkey	Retrospective	48	82.6	67.2	114	82.7	67.2	N/A
Tsui et al. [54]	2024	China	Retrospective	772	73.4	66.4	1544	N/A	N/A	19.2
Singh et al. [55]	2025	USA	Retrospective	89,568	53	63.51	121,115	60	65.51	< 6
Kim et al. [56]	2025	S Korea	Retrospective	7155	59	67	11,172	59.8	N/A	< 12
Total				244,207			328,674

Fig. 1.

Flow chart of study selection: UTKA—unilateral total knee arthroplasty; BTKA—bilateral total knee total knee arthroplasty

Methodological quality assessment results

Risk-of-bias assessment was performed for all 53 included studies using the ROBINS-I tool. The main methodological concerns were bias due to confounding and selection of participants, reflecting the observational design and baseline differences between SimBTKA and StaBTKA groups. The ROBINS-I traffic-light plot is presented in Fig. 2. As a supplementary quality assessment, NOS scores for all included studies are provided in Supplementary Table.

Fig. 2.

ROBINS-I traffic-light plot for risk-of-bias assessment of included studies. D1: bias due to confounding; D2: bias due to selection of participants; D3: bias in classification of interventions; D4: bias due to deviations from intended interventions; D5: bias due to missing data; D6: bias in measurement of outcomes; D7: bias in selection of the reported result

Mortality

We summarized the definitions of mortality and event counts across the included studies, as shown in Table 2. Moderate heterogeneity was observed among the included studies (I² = 40.7%, P = 0.042); therefore, a random-effects model was applied. The pooled analysis showed that SimBTKA was associated with a significantly higher risk of short-term mortality compared with StaBTKA (OR = 2.35, 95% CI 1.69–3.27, P < 0.001), as shown in Fig. 3. Visual inspection of the funnel plot revealed a symmetrical distribution, and Egger’s regression test did not indicate significant publication bias (P = 0.325) (Fig. 11A). To assess the potential influence of survivor selection bias, a tipping-point sensitivity analysis was performed by hypothetically adding unobserved inter-stage deaths to the StaBTKA group. The mortality association remained statistically significant until 133 additional deaths were assigned to the StaBTKA group, at which point the lower bound of the 95% CI crossed 1.0. This suggests that the observed mortality difference was relatively robust to a moderate degree of potential under-ascertainment in the StaBTKA group.

Table 2.

Mortality definitions and event counts across included studies

Author Year	SimBTKA		StaBTKA		Mortality definition
	Events	Total	Events	Total
Chua et al. [28]	40	23,136	8	12,951	30d
Eke [39]	1	225	0	51	30d
Franceschetti et al. [52]	0	65	0	108	30d
Lindberg‐Larsen et al. [33]	0	232	0	232	30d
Mangaleshkar [4]	4	54	0	34	30d
Meehan et al. [15]	43	11,445	84	26,350	30d
Singh [55]	33	89,568	15	121,115	30d
Stefánsdóttir et al. [12]	11	1139	5	3432	30d
Wilkie et al. [42]	20	19,334	0	19,334	30d
Wyles et al. [36]	0	188	0	242	30d
Sliva et al. [6]	0	26	1	306	60d
Bini et al. [18]	2	1230	0	2123	90d
Bolognesi et al. [16]	33	4519	13	3788	90d
Çelen et al. [43]	1	168	0	63	90d
Courtney et al. [19]	0	103	1	131	90d
Lindberg-Lars et al. [21]	0	157	6	628	90d
Ritter et al. [5]	14	2050	1	152	90d
Sheth et al. [25]	8	2814	5	5177	90d
Tsay et al. [34]	83	27,301	63	45,419	90d
Walmsley et al. [11]	8	826	5	1796	90d
Yoon et al. [14]	0	119	0	119	In-hospital
Bohm et al. [23]	10	6349	15	25,253	In-hospital
Ayekoloye et al. [47]	0	19	1	12	In-hospital
Hutchinson et al. [9]	1	438	1	125	In-hospital
Stubbs et al. [7]	0	28	0	36	In-hospital
Arslan et al. [27]	1	72	1	61	1 year
Gill et al. [37]	1	122	0	46	1 year
Kim et al. [56]	14	7155	37	11,172	1 year
Koh et al. [29]	2	820	6	633	1 year
Yalin et al. [53]	0	48	0	114	1 year
Niki et al. [20]	0	60	0	60	> 1year

Fig. 3.

Forest plot of short-term mortality after SimBTKA and StaBTKA. Studies reporting comparable 30-day, 60-day, or 90-day mortality outcomes were included. Random-effects model

Fig. 11.

Funnel plots for publication bias assessment across outcomes. Egger’s regression test showed no significant evidence of funnel plot asymmetry for mortality (P = 0.325), revision rate (P = 0.952), pulmonary embolism (P = 0.085), deep vein thrombosis (P = 0.264), superficial infection (P = 0.992), deep infection (P = 0.194), neurological complications (P = 0.723), or cardiac complications (P = 0.234). The x-axis represents effect size, and the y-axis represents the standard error of effect size

Revision rate

For revision rate, the initial meta-analysis using all studies yielded I² = 57%, indicating moderate heterogeneity. A random-effects model was applied, giving an OR of 0.94 (95% CI 0.80–1.12). Sensitivity analysis excluding Meehan et al. [15] reduced heterogeneity to I² = 16%; therefore, a fixed-effects model was applied for the subsequent meta-analysis. The pooled results indicated no statistically significant difference in revision rates between the SimBTKA and StaBTKA groups (OR = 1.02, 95% CI 0.95–1.10; P = 0.52), as illustrated in the forest plot (Fig. 4). In addition, visual inspection of the funnel plot revealed a symmetrical distribution, and Egger’s regression test did not suggest the presence of publication bias (P = 0.952) (Fig. 11B).

Fig. 4.

Forest plot of revision rate after SimBTKA and StaBTKA. Fixed effects model

Pulmonary embolism

Heterogeneity testing revealed substantial between-study heterogeneity (I² = 85.6%, P < 0.01); therefore, a random-effects model was applied for the meta-analysis. The pooled analysis demonstrated no statistically significant difference in the incidence of PE between the SimBTKA and StaBTKA groups (OR = 1.22, 95% CI 0.89–1.67; P = 0.22), as shown in the forest plot (Fig. 5). Visual inspection of the funnel plot suggested an approximately symmetrical distribution, and Egger’s regression test did not indicate significant publication bias (P = 0.085) (Fig. 11C). Given the substantial heterogeneity observed in PE (I² = 86%), additional univariable and exploratory multivariable meta-regression analyses were performed. Study era, data source, and adjustment method did not significantly explain the heterogeneity in univariable analyses. In the multivariable model, no independent moderator was identified, and substantial residual heterogeneity remained. Detailed results are shown in Table 3.

Fig. 5.

Forest plot of pulmonary embolism after SimBTKA and StaBTKA. random effects model

Table 3.

Univariable and multivariable meta-regression analyses for heterogeneous outcomes

Outcome	Model	Moderator	P value	Residual I2	Interpretation
Pulmonary embolism	Univariable	Study era	0.859	78.04%	Not significant
Pulmonary embolism	Univariable	Data source	0.528	82.74%	Not significant
Pulmonary embolism	Univariable	Adjustment method	0.524	82.56%	Not significant
Pulmonary embolism	Multivariable	All covariates	0.923	79.24%	No independent moderator
Cardiac complications	Univariable	Study era	0.130	90.29%	Trend, not significant
Cardiac complications	Univariable	Data source	0.608	94.80%	Not significant
Cardiac complications	Univariable	Adjustment method	0.977	94.74%	Not significant
Cardiac complications	Multivariable	All covariates	0.309	90.82%	No independent moderator
Blood transfusion	Univariable	Study era	0.760	93.24%	Not significant
Blood transfusion	Univariable	Data source	0.361	92.58%	Not significant
Blood transfusion	Univariable	Adjustment method	0.956	92.46%	Not significant
Length of stay	Univariable	Region	0.010	99.80%	significant

All covariates: Study era, Data source, Adjustment method. P value indicates the significance of the moderator; for study era and multivariable models, it refers to the joint test. Residual I² indicates unexplained heterogeneity after adjustment for moderator(s).

Deep vein thrombosis

Heterogeneity assessment did not reveal significant between-study heterogeneity (I² = 25.2%, P = 0.109); therefore, a fixed-effects model was employed for the meta-analysis. The pooled results demonstrated that the risk of DVT was significantly higher in the SimBTKA group than in the StaBTKA group (OR = 1.45, 95% CI 1.37–1.53; P < 0.001), as illustrated in the forest plot (Fig. 6). Visual inspection of the funnel plot indicated an approximately symmetrical distribution, and Egger’s regression test did not suggest significant publication bias (P = 0.264) (Fig. 11D).

Fig. 6.

Forest plot of deep vein thrombosis after SimBTKA and StaBTKA. Fixed effects model

3.7 Superficial Infection: Initial heterogeneity assessment indicated moderate between-study heterogeneity (I² = 57.2%, P < 0.001); therefore, a random-effects model was applied for the meta-analysis. The pooled analysis demonstrated that the risk of superficial infection was significantly higher in the StaBTKA group than in the SimBTKA group (OR = 0.69, 95% CI 0.54–0.89; P = 0.004), as shown in the forest plot (Fig. 7). Visual inspection of the funnel plot revealed an approximately symmetrical distribution, and Egger’s regression test did not indicate significant publication bias (P = 0.992) (Fig. 11E).

Fig. 7.

Forest plot of superficial infection after SimBTKA and StaBTKA. Random effects model

Deep infection

Assessment of heterogeneity did not demonstrate significant between-study heterogeneity (I² = 7%, P = 0.36); therefore, a fixed-effects model was applied for the meta-analysis. The pooled results indicated that the risk of deep infection was significantly higher in the StaBTKA group than in the SimBTKA group (OR = 0.65, 95% CI 0.61–0.70; P < 0.001), as shown in the forest plot (Fig. 8). The funnel plot appeared symmetrical, and Egger’s regression test did not suggest significant publication bias (P = 0.194) (Fig. 11F).

Fig. 8.

Forest plot of deep infection after SimBTKA and StaBTKA. Fixed effects model

Neurological complications

Heterogeneity testing demonstrated significant between-study heterogeneity (I² = 43.8%, P = 0.015); therefore, a random-effects model was used for the meta-analysis. The pooled analysis showed no statistically significant difference in the incidence of neurological complications between the SimBTKA and StaBTKA groups (OR = 1.20, 95% CI 0.99–1.45; P = 0.06), as illustrated in the forest plot (Fig. 9). Visual inspection of the funnel plot suggested an approximately symmetrical distribution, and Egger’s regression test did not indicate significant publication bias (P = 0.723) (Fig. 11G).

Fig. 9.

Forest plot of neurological complications after SimBTKA and StaBTKA. Random effects model

Cardiac complications

Heterogeneity assessment revealed substantial between-study heterogeneity (I² = 95%, P < 0.001); therefore, a random-effects model was applied for the meta-analysis. The pooled analysis demonstrated no statistically significant difference in the risk of postoperative cardiac complications between the SimBTKA and StaBTKA groups (OR = 0.87, 95% CI 0.66–1.16; P = 0.35), as shown in the forest plot (Fig. 10). Visual inspection of the funnel plot suggested an approximately symmetrical distribution, and Egger’s regression test did not indicate significant publication bias (P = 0.234) (Fig. 11H). Because substantial heterogeneity was observed for cardiac complications (I² = 95%), meta-regression analyses were conducted. Study era showed a trend toward moderating the effect estimate, but the overall effect did not reach statistical significance. Data source and adjustment method did not significantly explain heterogeneity. In the multivariable model, no independent moderator was identified, and substantial residual heterogeneity remained. Detailed results are shown in Table 3.

Fig. 10.

Forest plot of cardiac complications after SimBTKA and StaBTKA. Random effects model

Blood transfusion

Substantial heterogeneity was observed among the included studies (I² = 92.2%, P < 0.001); therefore, a random-effects model was used for the meta-analysis. The pooled results demonstrated that the transfusion rate was significantly higher in the SimBTKA group than in the StaBTKA group (OR = 4.42, 95% CI 3.11–6.28; P < 0.001), as illustrated in the forest plot (Fig. 12). Given the substantial heterogeneity in transfusion outcomes (I² = 92%), univariable meta-regression analyses were performed. Study era, data source, and adjustment method did not significantly explain the observed heterogeneity. Multivariable meta-regression was not performed because only 15 studies were available, and including multiple covariates could increase the risk of overfitting. Detailed results are shown in Table 3.

Fig. 12.

Forest plot of blood transfusion rate after SimBTKA and StaBTKA. Random effects model

Length of stay

Given the extreme heterogeneity in LOS (I² = 99.9%), subgroup analysis and univariable meta-regression were performed according to geographic region. Asian studies generally showed larger reductions in LOS for SimBTKA, whereas Non-Asian studies showed smaller or inconsistent differences. Figure 13 presents the individual study estimates and their confidence intervals for each subgroup. Meta-regression showed that geographic region significantly moderated the LOS effect estimate (P = 0.010). However, substantial residual heterogeneity remained (I² residual = 99.80%). Therefore, LOS was reported descriptively by subgroup rather than as a precise pooled mean difference. Detailed results are shown in Table 3.

Fig. 13.

Subgroup forest plot of LOS by geographic region

Discussion

Total knee arthroplasty (TKA) serves as the definitive treatment for end-stage knee osteoarthritis (OA), particularly indicated for patients exhibiting persistent, severe pain and functional limitation refractory to systematic conservative management [57]. In cases where OA has progressed to advanced stages, TKA has been substantiated to significantly alleviate pain and restore joint function. With the accelerating aging of the population and shifting lifestyle patterns, the disease burden of knee OA continues to escalate. Previous simulation-based studies predict that the sustained rise in obesity rates among the elderly in the United States and the United Kingdom over the next two decades will yield significant health and economic repercussions [58]. Obesity is recognized as a critical risk factor for various chronic conditions; similarly, advancing age and increased weight-bearing load play pivotal roles in the onset and progression of knee OA [59–62]. Given these trends, the demand for TKA is projected to surge further.

Mortality

The interpretation of mortality requires caution because mortality definitions differed across the included studies. To reduce clinical and methodological heterogeneity, the primary mortality analysis was restricted to studies reporting comparable short-term mortality windows, mainly 30-day or 90-day mortality. Studies reporting only in-hospital mortality, 1-year mortality, or mortality beyond 1 year were not pooled in the primary analysis because these time frames may reflect different risk periods and may be influenced by length of hospitalization or long-term comorbidity burden. Even after restricting the analysis to comparable short-term mortality windows, SimBTKA was associated with higher reported mortality. This finding is consistent with previous reports by Fu et al. [63], Liu et al. [64], and Makaram et al. [65], suggesting that SimBTKA may be associated with an increased risk of short-term perioperative mortality. The observed excess risk may be attributable to the greater cumulative surgical burden under a single anesthetic exposure, longer operative duration, and the resultant increase in systemic physiological stress. However, this finding should be interpreted with caution because of potential survivor selection bias in the StaBTKA group. In most included studies, the StaBTKA group consisted of patients who successfully underwent both staged procedures. Patients who died or experienced serious complications after the first surgery and therefore did not proceed to the second procedure were typically not counted in the StaBTKA group. This design may underestimate adverse events in the StaBTKA group and exaggerate the apparent mortality difference between SimBTKA and StaBTKA. The tipping-point sensitivity analysis showed that the mortality finding would lose statistical significance only after assigning 133 additional unobserved inter-stage deaths to the StaBTKA group. This suggests that the observed association was relatively robust to moderate under-ascertainment of mortality in the staged group. Nevertheless, the true number of patients who died or experienced major complications after the first-stage procedure but did not proceed to the second stage was not reported in most included studies. Therefore, the mortality result should still be interpreted as a safety signal rather than a definitive causal estimate.

Pulmonary embolism and deep vein thrombosis

For pulmonary embolism, neither study era, data source, nor adjustment method significantly explained the heterogeneity, either in univariable or multivariable models. This suggests that PE estimates may have been influenced by other unmeasured factors, such as thromboprophylaxis protocols, diagnostic strategies, baseline thromboembolic risk, perioperative mobilization, and differences in outcome definitions. Therefore, although no statistically significant difference in PE incidence was identified between SimBTKA and StaBTKA, the pooled estimate should be interpreted with caution. In contrast, the risk of DVT was significantly higher in the SimBTKA group than in the StaBTKA group. From a pathophysiological perspective, SimBTKA is associated with a greater surgical burden, prolonged anesthesia duration, and more pronounced hemodynamic stress. The convergence of these multiple stressors within a short time frame may increase the risk of lower-extremity venous thrombosis [66]. Conversely, staged procedures provide an interval for physiological recovery and optimization of modifiable risk factors between surgeries, which may facilitate venous return and reduce the incidence of thrombotic complications.

Infection and revision rate

Based on CDC criteria [67], StaBTKA exhibited significantly higher incidences of both superficial and deep infections compared to SimBTKA. This difference should be interpreted with caution, as many studies have compared outcomes on a per-patient basis rather than on a per-procedure basis. Since StaBTKA involves two separate surgical incisions, two perioperative exposure phases, and two opportunities for postoperative infection, the cumulative risk of infection may appear higher even if the risk of infection per procedure is similar. Future studies should report infection rates for both the patient and the knee side to allow for more clinically meaningful comparisons. Heterogeneity regarding superficial infections primarily stemmed from methodological discrepancies; notably, Çelen et al. [43] utilized a cumulative follow-up framework that extended the risk exposure window for StaBTKA. Regarding 1-year revision rates, no significant intergroup difference was observed, consistent with prior reviews [63–65]. However, Fu et al. [63] reported higher revision rates in StaBTKA, likely correlated with the increased incidence of deep infection, which remains a leading indication for revision [68]. Meehan et al. [15] used life-table methods with scenario-based assumptions to estimate risks for patients who did not complete the second-stage procedure; although methodologically innovative, such model-based estimates differ from observed cohort data and introduce assumption-related variability, limiting direct comparability with conventional studies.

Neurological complications

Neurological complication did not differ significantly between the SimBTKA and StaBTKA groups, a finding that contrasts with some previous meta-analyses [63–65]. Although SimBTKA theoretically concentrates bilateral surgical stress within a single perioperative period and may increase exposure to factors such as blood loss, hypoxemia, and metabolic disturbances that could predispose to neurological injury [69], this was not reflected in the pooled results. The discrepancy is likely related to the lack of standardized definitions and substantial heterogeneity in the reporting of neurological outcomes across the included studies.

Cardiac complications

In the present meta-analysis, no significant difference in the incidence of postoperative cardiac complications was observed between SimBTKA and StaBTKA groups. This may be attributable to standardized perioperative cardiovascular risk management, including preoperative assessment and optimization of hypertension, coronary artery disease, and cardiac dysfunction, meticulous intraoperative control of anesthetic depth and hemodynamic parameters, and continuous postoperative monitoring with timely interventions. Under these multilayered strategies, the additional perioperative stress associated with SimBTKA may not translate into a detectable increase in cardiac complication risk. Meta-regression analyses indicated a trend for study era to moderate the effect estimate, which may reflect temporal improvements in perioperative cardiovascular assessment, anesthetic management, postoperative monitoring, and enhanced recovery protocols. However, no independent moderator was identified in the multivariable model, and substantial residual heterogeneity persisted, suggesting that unmeasured factors, such as patient selection, baseline cardiovascular risk, and institutional practices, may contribute to variability in reported cardiac outcomes.

Blood loss and transfusion

Data on intraoperative blood loss and transfusion rates were systematically summarized across the included studies (Table 4). A total of seven studies reported postoperative total blood loss; however, although Stubbs et al.^[7]provided blood loss data, no statistical comparison was performed. Among the remaining six studies, findings regarding blood loss were inconsistent: three studies reported significantly greater blood loss with SimBTKA, whereas the other three found no significant difference between SimBTKA and StaBTKA. Accordingly, the current evidence is insufficient to establish a consistent and statistically robust difference in blood loss between the two surgical strategies. For blood transfusion, substantial heterogeneity was observed across studies (I² = 92%). Univariable meta-regression indicated that study era, data source, and adjustment method did not significantly explain the heterogeneity. Because only 15 studies reported transfusion outcomes, multivariable meta-regression was not performed to avoid overfitting. The remaining heterogeneity may reflect differences in transfusion thresholds, perioperative blood management protocols, use of antifibrinolytic agents, baseline anemia, surgical technique, and institutional practice patterns. Therefore, pooled estimates for transfusion should be interpreted cautiously and primarily as directional rather than precise quantitative effects.

Table 4.

Comparative outcomes of blood loss and transfusion in SimBTKA versus StaBTKA

Author Year	BL(ml)		BT				P
	SimBTKA-ml	StaBTKA-ml	SimBTKA-units	StaBTKA-units	SimBTKA-N	StaBTKA-N	BL	BT-units	BT-N
Sliva et al. [6]	N/A	N/A	2	1	14	11	N/A	N/A	P = 0.001
Stubbs et al. [7]	1701.8	896	3.59	2.14	58	42	N/A	N/A	< 0.01
Forster et al. [10]	3312	2578	6	5.8	N/A	N/A	0.004	N/A	N/A
Yoon et al. [14]	1299	1302	N/A	N/A	N/A	N/A	> 0.05	N/A	N/A
Poultsides et al. [17]	N/A	N/A	N/A	N/A	2718	846	N/A	N/A	< 0.001
Courtney et al. [19]	N/A	N/A	1.74 ± 1.5	1.45 ± 1.4	73	85	N/A	< 0.001	< 0.001
Zhao et al. [22]	2448.9 ± 636.3	1798.4 ± 567.2	9.6 ± 2.3	3.2 ± 1.6	54	19	< 0.001	< 0.001	< 0.001
Bohm et al. [23]	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	< 0.001
Hadley et al. [26]	N/A	N/A	1.39	0.66	N/A	N/A	N/A	0.042	N/A
Arslan et al. [27]	N/A	N/A	1.6 ± 0.78	0.72 ± 0.71	N/A	N/A	N/A	< 0.001	N/A
Sobh et al. [30]	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	< 0.001
Sun et al. [31]	2377.56 ± 479.85	1754.43 ± 369.63	1335.57 ± 451.48	579.62 ± 254.36	N/A	N/A	< 0.001	N/A	< 0.001
Gill et al. [37]	N/A	N/A	2	0	64	20	N/A	0.039	0.001
Wan et al. [38]	N/A	N/A	0.18 ± 0.45	0.09 ± 0.76	16	6	N/A	0.116	0.078
Eke et al. [39]	N/A	N/A	3	6	N/A	N/A	N/A	< 0.001	N/A
Liu et al. [40]	93.34	95.52	N/A	N/A	N/A	N/A	0.95	N/A	N/A
Wilkie et al. [42]	N/A	N/A	N/A	N/A	1,634	409	N/A	N/A	< 0.001
Çelen et al. [43]	N/A	N/A	0.7 ± 1.0	0.3 ± 0.5	71	15	N/A	< 0.001	0.01
Chou et al. [44]	N/A	N/A	N/A	N/A	1,393	151	N/A	N/A	< 0.01
Erossy et al. [45]	N/A	N/A	N/A	N/A	1,401	433	N/A	N/A	< 0.0001
Mingxi et al. [48]	N/A	N/A	4.9 ± 0.8	3.9 ± 1.1	N/A	N/A	N/A	0.04	N/A
Accatino et al. [49]	N/A	N/A	N/A	N/A	9	12	N/A	N/A	0.27
Matsumura et al. [51]	595	640	N/A	N/A	6	4	0.897	N/A	0.514

BL= blood loss; BT= blood transfusion.

Length of stay and hospitalization costs

Geographic region significantly moderated the LOS effect estimate, indicating that the apparent reduction in hospitalization with SimBTKA was closely related to healthcare-system factors. In general, Asian studies reported greater reductions in LOS, whereas Non-Asian studies showed smaller or less consistent differences. These discrepancies may be explained by differences in discharge standards, perioperative management, rehabilitation pathways, reimbursement policies, and the way LOS was calculated for StaBTKA, particularly whether the two admissions were counted cumulatively. Because heterogeneity was extremely high (I² = 99.9%), the overall pooled LOS estimate should not be interpreted as a precise measure of effect. Instead, the LOS results are better understood as a context-dependent trend, with subgroup findings by region offering more useful clinical information than a single summary estimate. Hospitalization cost was also difficult to synthesize quantitatively. The included studies reported costs in different currencies, time periods, and healthcare systems, and few provided sufficient information for standardized conversion or purchasing power adjustment. For this reason, we summarized the cost evidence qualitatively rather than performing a formal meta-analysis (Table 5). Eleven studies reported lower direct hospitalization costs for SimBTKA compared with StaBTKA, supporting the view that simultaneous surgery may reduce expenses by completing both procedures within one treatment episode. This is consistent with the findings of Susan et al. [70]. Nevertheless, the economic advantage of SimBTKA is likely to vary across countries and institutions, depending on reimbursement structure, length-of-stay policy, rehabilitation arrangements, and perioperative resource use. Therefore, while SimBTKA may have a favorable cost-efficiency profile in selected patients and settings, this conclusion should not be interpreted as universally applicable across all healthcare systems.

Table 5.

Comparison of the cost of SimBTKA and StaBTKA

Author Year	Country (currency)	SimBTKA(cost)Mean ± SD	StaBTKA(cost)Mean ± SD	P
Sliva et al. [6]	USA(Dollar)	$14,291.85	$18,958.39	< 0.0001
Meehan et al. [15]	USA(Dollar)	$49,218 ± 23,872	$62,606 ± 34,857	< 0.001
Zhao et al. [22]	China(Dollar)	$12,223.4 ± 938.5	$14,733.2 ± 1,268.7	< 0.001
Sobh et al. [30]	USA(Dollar)	$24,596 ± 5652	$24,915 ± 5,756	0.428
Sun et al. [31]	China(RMB)	¥104,836.84 ± 5 371.45	¥126,836.57 ± 1,794.16	< 0.001
Wyles et al. [36]	USA(Dollar)	$22,057	$31,145	< 0.001
Gill et al. [37]	Australia(AUD)	A$36,892 ± 6515	A$43,281 ± 12,509	N/A
Pumo et al. [41]	USA(Dollar)	$6,209	$6,852	0.028
Wilkie et al. [42]	USA(Dollar)	$29,053 ± 19,047	$33,522 ± 13,909	< 0.001
Chou et al. [44]	China(TWD)	NT$188,888.0 ± 18,960.6	NT$206,550.2 ± 24,753.1	< 0.01
Erossy et al. [45]	USA(Dollar)	$28,296 ± 18,488	$33,202 ± 15,240	< 0.0001
Ayekoloye et al. [47]	Nigeria(Dollar)	$6,400	$7,100	N/A
Mingxi et al. [48]	China(RMB)	¥126,972.7 ± 10,618.2	¥154,025.4 ± 9,545.7	< 0.001
Tsui et al. [54]	China(Dollar)	$37,798.52	$46,220.74	N/A
Singh et al. [55]	USA(Dollar)	$98,623.15 ± 65,635.53	$117,107.82 ± 66,623.23	< 0.001

A biomechanical perspective

Biomechanical outcomes were not directly evaluated in this meta-analysis and therefore cannot be used to support definitive conclusions regarding the superiority of either surgical strategy. Previous biomechanical and gait studies have suggested that staged procedures may temporarily create inter-limb asymmetry in alignment, loading, and functional recovery during the interstage period, whereas simultaneous procedures may allow more synchronous bilateral rehabilitation [71]. However, because gait parameters, muscle strength, limb-loading patterns, and long-term functional outcomes were not consistently reported in the included studies, these biomechanical considerations should be regarded only as exploratory background and should not outweigh individualized perioperative risk assessment.

Challenges

The optimal interval between staged procedures remains unclear, ranging from a few days to several months, which limits comparability between the two groups. There are differences in patient selection criteria: SimBTKA patients are typically younger, in better health, or treated at medical institutions with higher surgical volumes. Even after adjusting for these factors, residual confounders remain. Inconsistent definitions of outcomes across studies have led to increased heterogeneity regarding complications such as PE, DVT, cardiac events, neurological issues, and infections. Most studies indicate that SimBTKA is associated with lower direct hospitalization costs; however, cross-national comparisons remain insufficient. Due to variations in the interval between stages, there is significant clinical heterogeneity within the StaBTKA group. Staged surgery with shorter intervals may result in patients undergoing a second major operation before they have fully recovered, whereas staged surgery with longer intervals may be clinically equivalent to two separate unilateral procedures. Given the available data and inconsistencies across study reports, a formal subgroup analysis stratified by interval is not feasible. We acknowledge that this heterogeneity may be one of the factors contributing to the observed differences in mortality, thromboembolic events, and blood transfusion requirements, and this should be taken into account when interpreting the pooled estimates.

Limitations

This study has several limitations that should be considered when interpreting the results. First, all included studies were observational cohort studies, and most employed a retrospective design. Therefore, the pooled results are inherently susceptible to selection bias, residual confounders, and selection bias resulting from differences in patient baseline characteristics, surgical indications, surgeon experience, and perioperative management. Although some studies employed multivariate regression or propensity score matching, unmeasured confounding factors such as cardiopulmonary reserve, surgeon experience, antithrombotic prophylaxis regimens, and rehabilitation intensity could not be fully controlled. Second, there were significant differences in the definition of StaBTKA across the included studies. Some studies defined staged surgery as procedures completed during the same hospital stay or within a few days, while others used intervals of several months or more than a year. This variation may influence complication rates, blood transfusion requirements, rehabilitation progress, and cumulative length of hospital stay. Therefore, StaBTKA should not be regarded as a single, homogeneous intervention across all included studies. Third, there were inconsistencies in the definition and reporting of several outcome measures. Mortality is typically reported in the immediate postoperative period, but the definitions of complications such as pulmonary embolism, deep vein thrombosis, cardiac events, neurological complications, superficial infections, and deep infections vary across studies. Furthermore, some large-scale database studies rely on diagnostic or surgical codes, which may introduce classification bias; in contrast, smaller clinical studies employ medical record reviews or institution-specific criteria. Fourth, there are differences in how outcomes are reported across studies. Some outcomes are reported on a “per patient” basis, while others may be reported on a “per knee” or “per hospital stay” basis. This issue is particularly pronounced for outcomes such as infection, revision surgery, blood transfusion, and length of hospital stay, as StaBTKA involves two surgeries and two periods of exposure. These differences may affect the interpretation of cumulative risks and may partially explain the observed heterogeneity. Fifth, significant heterogeneity was observed in several pooled outcome measures, particularly pulmonary embolism, cardiac complications, transfusion rates, and length of hospital stay. This heterogeneity may reflect differences in patient selection, healthcare systems, surgical protocols, antithrombotic prophylaxis, transfusion thresholds, discharge criteria, and rehabilitation pathways. Despite the use of a random-effects model and sensitivity analyses, the high heterogeneity reduces the certainty and generalizability of the pooled estimates. Sixth, due to differences in currency, healthcare systems, reimbursement policies, study periods, and cost structures, a quantitative synthesis of economic outcomes was not possible. Most cost analyses focused solely on direct inpatient costs and did not systematically include rehabilitation, readmissions, outpatient care, indirect costs, or long-term economic consequences. Finally, there is insufficient reporting on biomechanical and long-term functional outcomes. Although staged surgery may theoretically cause temporary bilateral lower limb asymmetry and changes in weight-bearing distribution between stages, most included studies focused on perioperative complications rather than gait, alignment, patient-reported outcomes, functional recovery, or long-term prosthesis survival rates. Future prospective studies should employ standardized definitions, predefined staging intervals, risk-stratified patient selection, and long-term follow-up to better identify which patients are most likely to benefit from each surgical strategy.

Conclusion

SimBTKA and StaBTKA appear to have different risk profiles rather than one approach being clearly superior to the other. In this meta-analysis, SimBTKA was associated with higher reported short-term mortality, a higher risk of DVT, and greater transfusion requirements, whereas StaBTKA was associated with higher rates of superficial and deep infection. SimBTKA also tended to shorten hospital stay and improve treatment efficiency, particularly in Asian studies, but this finding should be interpreted cautiously because LOS varied substantially across healthcare systems and showed marked heterogeneity. No significant differences were found in revision rates, pulmonary embolism, neurological complications, or cardiac complications. These findings should be interpreted in the context of the observational nature of the included studies. Differences between groups may partly reflect patient selection, baseline imbalance, and unmeasured confounding. In particular, the mortality result requires caution because patients who died or developed serious complications after the first-stage procedure may not have proceeded to the second operation and therefore may not have been captured in the StaBTKA group. This survivor selection bias could underestimate the true mortality risk of staged surgery. Therefore, the observed mortality difference should be viewed as an important safety signal, not as definitive evidence that one strategy is universally safer. The choice between SimBTKA and StaBTKA should be individualized according to patient comorbidities, thromboembolic risk, infection risk, perioperative tolerance, expected efficiency, and healthcare resource considerations.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1 ^{(294.3KB, docx)}

Supplementary Material 2 ^{(154.6KB, pdf)}

Supplementary Material 3 ^{(100.9KB, docx)}

Abbreviations

TKA

Total knee arthroplasty

OA

Osteoarthritis

SimBTKA

Simultaneous bilateral total knee arthroplasty

StaBTKA

Staged bilateral total knee arthroplasty

DVT

Deep vein thrombosis

PE

Pulmonary embolism

OR

Odds ratio

CI

Confidence interval

MD

Mean difference

CDC

Centers for disease control and prevention

Author contributions

Conceptualization: Tianyun Zhao. Methodology: Hongzhao Wang. Formal analysis: Hongzhao Wang. Investigation: Yabing Jiang. Data curation: Xiao Wang. Writing—original draft preparation: Hongzhao Wang, Yabing Jiang, Xiao Wang. Writing—review and editing: Tianyun Zhao. Supervision: Tianyun Zhao.

Funding

Not applicable.

Data availability

All data generated or analyzed during this study are included in this published article. Raw analysis data are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.