Literature management software (EndNote X9, Thomson Scientific) was used to collate the search results and remove duplicates. Two investigators independently screened the titles and abstracts of the deduplicated records, excluding those clearly irrelevant. The full texts of the remaining articles were then retrieved and assessed for eligibility based on the predefined inclusion and exclusion criteria. The quality of the finally included studies was assessed, and relevant data were independently extracted by the same two researchers. To ensure the consistency and reliability of the screening and data extraction process, inter-rater agreement was calculated using Cohen’s kappa statistic at both the title/abstract screening and full-text assessment stages, yielding kappa values of 0.85 and 0.88, respectively, indicating excellent agreement. The extracted data were summarized and cross-checked. Any discrepancies in the data extraction process were resolved through consultation with a third independent researcher. If necessary, corresponding authors were contacted to obtain complete original data.

The quality of all included studies was independently assessed by two reviewers. The Newcastle–Ottawa Scale (NOS) was used to evaluate the quality of cohort studies and case–control studies (13). In the NOS assessment, studies are scored out of a maximum of 9 points. Those with a score ≥7 are considered high quality, scores of 5–6 indicate moderate quality, and scores <5 are deemed low quality. For RCTs, the Cochrane risk-of-bias tool in Review Manager (RevMan) version 5.3 was utilized (14). The assessment focused on seven key domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other potential sources of bias. A risk-of-bias graph was generated accordingly. Any disagreements in the quality assessments were resolved through discussion with a third independent reviewer. The inter-rater reliability for the risk-of-bias assessments, as measured by Cohen’s kappa, was 0.82, indicating strong agreement between the two reviewers.

The extracted data encompassed the following two primary categories: (1) Baseline characteristics, which included the first author, year of publication, country/region, study design, sample size, patient age, type of surgery, specific intervention regimens, and follow-up duration; and (2) Outcomes of interest, specifically the incidence rates of PE, DVT, major bleeding (defined as bleeding leading to impaired organ/limb/muscle function requiring re-operation), minor bleeding (e.g., ecchymosis, hematoma, and wound bleeding), wound complications (e.g., infection and drainage), 90-day all-cause mortality, and readmission rates. It is acknowledged that the diagnostic criteria for VTE events (DVT and PE) varied across the included studies. The outcome data for VTE included both symptomatic events confirmed by objective imaging (such as compression ultrasonography, venography, or computed tomography pulmonary angiography) and, in some studies, events detected through systematic screening imaging protocols. Data regarding the specific diagnostic methods employed in each included study were extracted and considered during the analysis and interpretation of the results.

Meta-analysis was performed using RevMan version 5.3 software.¹ For dichotomous data, the odds ratio (OR) was used as the effect measure. All effect measures are reported with their 95% confidence intervals (CIs). Heterogeneity was assessed using the I² statistic and p-value: if p > 0.1 and I² < 50%, heterogeneity among studies was considered low, and a fixed-effect model was applied; if p ≤ 0.1 and I² ≥ 50%, a random-effects model was used, and subgroup or sensitivity analyses were conducted to explore sources of heterogeneity. Sensitivity analysis was performed by sequentially excluding individual studies to assess the impact of each study on the overall effect size. Publication bias was evaluated using funnel plots. A significance level of p ≤ 0.05 was set for all statistical analyses. We recognized that variations in the dosage (e.g., aspirin 81 mg vs. 325 mg; enoxaparin 40 mg once daily vs. 30 mg twice daily) and duration of prophylaxis among the included studies represent important sources of clinical heterogeneity. A priori, we planned to explore the potential impact of these factors through subgroup analyses if sufficient data were available.

The quality of evidence for each primary outcome (incidence of pulmonary embolism, deep vein thrombosis, major bleeding, and minor bleeding) was assessed using the GRADE (15) (Grading of Recommendations, Assessment, Development, and Evaluation) framework. Evidence from randomized controlled trials (RCTs) initially rated as high quality may be downgraded for five factors: risk of bias, inconsistency, indirectness, imprecision, and publication bias. Evidence from observational (cohort) studies began as low quality and could be further downgraded for the same reasons, or upgraded for factors such as a large effect size or a dose–response gradient. The evidence quality for each outcome was categorized into one of four levels: high, moderate, low, or very low. This assessment was performed independently by two reviewers, with any disagreements resolved by consensus.

A total of 1,881 records were initially identified through database searches. After removing duplicates and a rigorous screening process based on titles, abstracts, and full texts, 8 studies meeting the predefined criteria were included for the final analysis [comprising 4 RCTs (6, 11, 16, 17) and 4 cohort studies (7, 18–20)]. The study selection process is detailed in Figure 1. These studies comprised a total of 126,367 patients, with 43,441 patients in the aspirin group and 82,926 patients in the enoxaparin group. See Table 2 for the basic characteristics of the included studies.

The GRADE assessments for the key efficacy and safety outcomes are summarized in Table 4. The quality of evidence was rated as moderate for the outcomes of pulmonary embolism and major bleeding, indicating that we are moderately confident in the effect estimates, and the true effect is likely to be close to the estimate, but there is a possibility that it is substantially different. The evidence for deep vein thrombosis was rated as low quality, and for minor bleeding as moderate quality. Downgrading was primarily due to inconsistency (substantial unexplained heterogeneity) and imprecision (wide confidence intervals crossing the line of no effect or minimal important difference).

This meta-analysis found no significant difference in the efficacy of aspirin vs. enoxaparin for the prevention of PE (OR = 1.14, 95% CI: 0.66–1.95), a finding consistent with several recent studies. A large randomized controlled trial (the METRC study) by O’Toole et al. (21) reported that aspirin was non-inferior to low-molecular-weight heparin (enoxaparin) in preventing 90-day all-cause mortality in patients undergoing surgery for extremity fractures, with mortality rates of 0.78% vs. 0.73%, respectively (absolute difference 0.05%, p < 0.001). Notably, the incidence of PE was identical between the two groups in that study (21), supporting our conclusion that there was no difference in PE prevention.

Regarding DVT prevention, our primary analysis showed no significant difference between the two agents (OR = 1.13, 95% CI: 0.65–1.98, p = 0.67). However, after excluding a study with high heterogeneity (6) in a sensitivity analysis, the incidence of DVT was significantly lower in the enoxaparin group compared to the aspirin group (OR = 0.78, 95% CI: 0.64–0.96, p = 0.02). This result aligns closely with the findings of the CRISTAL study group cluster-randomized trial by Sidhu et al. (11), which demonstrated a significantly higher rate of symptomatic VTE in the aspirin group than in the enoxaparin group (absolute difference 1.97%, p = 0.007) (6, 10). A potential explanation for this discrepancy lies in the distribution of DVT types; the CRISTAL study group trial observed that aspirin was less effective in preventing distal DVT (10), which accounted for 68% of the VTE events. In our meta-analysis, inconsistent anatomical classification of DVT across the included studies likely contributed to the high heterogeneity (I² = 82%) observed in the primary analysis. It is important to note that patient population characteristics may significantly influence the outcomes. A 2022 study on patients with splenic rupture combined with lower extremity fractures found no significant difference in VTE rates between aspirin monotherapy and LMWH monotherapy (22). Conversely, a 2023 meta-analysis indicated that body mass index (BMI) did not significantly modify the comparative effectiveness of aspirin and enoxaparin (23), suggesting that the efficacy comparison might vary across different risk-stratified patient subgroups.

This study identified a significant advantage of aspirin regarding minor bleeding events (OR = 0.57, 95% CI: 0.49–0.66, p < 0.00001), while no significant differences were observed compared to enoxaparin in terms of major bleeding (OR = 0.60, 95% CI: 0.40–0.89, p = 0.01), wound complications (OR = 1.03, 95% CI: 0.60–1.78, p = 0.90), or mortality (OR = 0.81, 95% CI: 0.37–1.78, p = 0.60). This safety profile is consistent with prior research. The METRC study reported no significant difference in major bleeding events between aspirin and enoxaparin (21), and the CRISTAL study group trial similarly found comparable rates of major bleeding between the two groups.

Notably, the 43% reduction in the risk of minor bleeding (OR = 0.57) associated with aspirin observed in our study holds significant clinical implications. Although minor bleeding events are not directly life-threatening, they can substantially impact patient quality of life, medication adherence, and healthcare resource utilization. The mechanistic basis for this safety difference likely stems from the distinct anticoagulant mechanisms of the two drugs (22). Enoxaparin exerts a potent anticoagulant effect by inhibiting activated factor X (FXa), whereas aspirin produces a relatively milder antithrombotic effect by inhibiting platelet cyclooxygenase-1 (COX-1), thereby reducing thromboxane A2 generation (24–28). This difference in the mechanism of action also explains why enoxaparin demonstrates stronger efficacy in thromboprophylaxis but is associated with a higher bleeding risk.

Combining therapies may be a strategy to balance these effects. A 2022 study indicated (22) that the VTE incidence in the LMWH plus aspirin combination group (5.26%) was lower than in either the aspirin monotherapy (13.33%) or LMWH monotherapy (12.50%) groups, without increasing the bleeding risk. This suggests that a combination regimen might achieve an optimal benefit-to-risk balance for high-risk patients.

In summary, based on moderate-quality evidence, there is no significant difference between aspirin and enoxaparin in preventing pulmonary embolism or major bleeding. However, the superiority of enoxaparin in reducing DVT incidence is supported by only low-quality evidence, and the significant reduction in minor bleeding with aspirin is supported by moderate-quality evidence.

Our findings provide empirical support for the inclusion of aspirin as a viable prophylactic option in clinical guidelines. The comparable efficacy of aspirin and enoxaparin in preventing pulmonary embolism and major bleeding in our analysis, coupled with aspirin’s superior safety profile regarding minor bleeding, aligns with the recommendations of guidelines such as those from the American Academy of Orthopedic Surgeons (AAOS) and the American Society of Hematology (ASH), which endorse or suggest aspirin as an option for patients undergoing major orthopedic surgery, particularly those at standard or low risk for VTE (2, 11). Conversely, our sensitivity analysis suggesting a potential superiority of enoxaparin in preventing deep vein thrombosis supports the rationale for preferring LMWH in guidelines for patients identified as high-risk for VTE (e.g., those with a prior VTE history, active cancer, or thrombophilia).

Building upon this personalized paradigm, a hybrid or step-down strategy has emerged to further optimize the risk–benefit balance, particularly for patients at moderate to high risk. This approach involves initiating prophylaxis with a potent anticoagulant (e.g., enoxaparin or a direct oral anticoagulant) during the immediate high-risk postoperative period, followed by a transition to aspirin for the remainder of the prophylactic course. This strategy leverages the temporal decline in VTE risk, to provide strong early protection while limiting exposure duration to agents with higher bleeding potential. Supporting evidence for this model includes a study demonstrating the efficacy of risk-stratified prophylaxis using low-dose aspirin (29), and a recent large-scale evaluation confirming that a hybrid anticoagulant-to-aspirin regimen was associated with lower composite complication rates than aspirin monotherapy for patients at elevated VTE risk within an institutional risk-stratified framework (30).

Therefore, our meta-analysis contributes to the evolving evidence base for personalized prophylaxis: aspirin represents an effective and safer oral alternative for standard-risk patients, while a risk-stratified approach that may include initial anticoagulation (with or without subsequent transition to aspirin) remains crucial for those at elevated thrombotic risk.

Compared with previous important meta-analyses in this field, the present study demonstrates significant advances in the timeliness of included literature, sample size, and precision of the conclusions. The review by Nadi et al. (31) found no statistically significant differences between aspirin and enoxaparin across all thrombotic and bleeding outcomes but emphasized that the quality of evidence was “very low,” resulting in substantial uncertainty. Similarly, Farey et al. (32) found no significant differences; however, their analysis was limited by including only 4 RCTs (n = 1,507), which constrained statistical power and precluded definitive conclusions for key outcomes such as pulmonary embolism and mortality, particularly expressing caution regarding the safety profile in total hip arthroplasty patients.

Notably, two recent 2024 studies published after our literature search cut-off further inform this discussion. The network meta-analysis by Cheok et al. (33), which specifically evaluated thromboprophylaxis following hip and knee arthroplasty, concluded that aspirin (<325 mg daily), enoxaparin, and dabigatran offer an overall satisfactory efficacy and safety profile, strongly supporting the non-inferiority of aspirin in this context. Additionally, the systematic review by Williamson et al. (34), while focused on hip fracture surgery, found aspirin to be non-inferior to other agents for VTE prevention and potentially associated with reduced mortality, adding to the broader evidence for aspirin’s utility in orthopedic thromboprophylaxis.

Building on this evolving evidence base, our study incorporates 8 studies published through 2025, encompassing a total of 117,367 patients, thereby substantially enhancing statistical power. It not only confirms the non-inferiority of aspirin in preventing pulmonary embolism but, more importantly, through sensitivity analysis that eliminated heterogeneity interference, it is the first to clearly demonstrate the superiority of enoxaparin over aspirin in reducing the incidence of deep vein thrombosis (OR = 0.78, 95% CI: 0.64–0.96, p = 0.02). Concurrently, it establishes with highly significant evidence the definite advantage of aspirin in reducing the risk of minor bleeding (OR = 0.57, p < 0.00001).

Therefore, while building on the general framework of previous research and aligning with recent high-quality analyses, this study utilizes more recent and extensive data to provide more precise quantitative evidence the risk–benefit profiles of the two agents, offering stronger guidance for clinical decision-making.

This study has several limitations. First, significant heterogeneity was observed among the included studies, attributable to clinical variations in patient populations, prophylactic regimens, and methodology. Notably, the dosage and duration of prophylaxis varied substantially (e.g., aspirin 81–650 mg/day; enoxaparin 30 mg twice daily to 40 mg once daily), and the definitions for VTE diagnosis were inconsistent (systematic screening vs. clinical diagnosis). These sources of clinical and methodological heterogeneity, particularly the inconsistent diagnostic criteria, influenced the incidence of DVT and contributed to the statistical heterogeneity observed in our analyses. A planned subgroup analysis by surgical type (THA vs. TKA) was also not feasible due to the lack of procedure-specific data in most studies.

Second, the wide variability in prophylactic regimens is a key constraint on interpreting our pooled results. The anti-thrombotic effect and bleeding risk of these agents are dose-dependent. Consequently, the summary effect estimates should be interpreted as an average across a spectrum of clinically used doses rather than reflecting the effect of a specific, optimized regimen. Our inability to perform a meaningful subgroup analysis by dose, due to the limited number of studies within each dosing category, precludes conclusions about which specific dosing strategy optimizes the benefit–risk profile for either agent.

Third, the follow-up duration was primarily 90 days, which may not capture later complications. Furthermore, subgroup data for specific high-risk populations (e.g., those with extreme obesity or prior VTE) were insufficient, limiting the analysis of treatment effects in these cohorts.

Fourth, the generalizability of our findings may be limited by the geographic and economic context of the included evidence. The majority of studies were conducted in high-income countries. Differences in patient factors, healthcare systems, and socioeconomic realities may affect the applicability of these results to other settings, such as in Asia or in low-resource hospitals, where drug adherence patterns and healthcare access differ.

Fifth, the exclusion criteria for important modifiers of drug metabolism—such as renal impairment and obesity—varied across studies. This variability may influence the applicability of our findings to patient subgroups in whom drug dosing or pharmacokinetics are altered.

The favorable clinical profile of aspirin also suggests potential economic advantages. The significant reduction in minor bleeding events likely translates into lower costs associated with managing these complications (e.g., fewer dressings, clinic visits, or patient evaluations). Furthermore, oral administration of aspirin eliminates the need for syringes, nursing time for education and administration, and potential complications associated with subcutaneous injections, which are required for enoxaparin. This convenience may also improve adherence in the outpatient setting. When combined with aspirin’s inherently lower drug acquisition cost, these factors strongly indicate its superior cost-effectiveness for VTE prophylaxis in standard-risk patients undergoing major orthopedic surgery. This economic benefit further solidifies aspirin’s role as a valuable prophylactic option, particularly in resource-conscious healthcare systems.