Pertrochanteric fractures are common in patients with osteoporosis due to aging or medical diseases, or in patients encountering high-energy hip trauma. Their increasing prevalence—either in senile women (1, 2) or youths with certain medical or social conditions—underscores the growing economic burden on both the patients and society (3). Over the years, surgical methods and instruments have been continuously invented and modified, with emphasis on patient quality of life and prognosis. Extramedullary internal fixation, after decades of comparison, is now deemed less beneficial than intramedullary internal fixation (4). Among intramedullary options and instrument choices (e.g., gamma nail, proximal femoral nail (PFN), proximal femoral nail anti-rotation (PFNA), InterTAN), PFN and PFNA are the most widely applied. The debate between PFN and PFNA is trivial (5–7) compared to that between extra- and intramedullary fixations, as both instruments yield satisfactory results for patients with pertrochanteric fractures. Nevertheless, in unstable pertrochanteric fractures, surgeons face uncertainty and prepare themselves for revision if internal fixation fails. Hemiarthroplasty (HA) is the conventional salvage surgery following proximal hip instrument failure. However, the fixation–failure–revision procedure compromises prognosis, quality of life, and survival. Surgeons are increasingly presenting the idea of primary rather than revision arthroplasty in unstable pertrochanteric fractures, considering it no longer an extreme, extensive, or aggressive procedure (8–10). Comparisons between HA and PFN/PFNA have persisted for years, with varying conclusions: some favor HA (11, 12) and some favor PFN/PFNA (13, 14), while others report inconclusive (15, 16) or statistically insignificant results (17, 18). Considering the potential selection bias, inappropriate data merging, and heterogeneity explanations, we think it is necessary and important to conduct a comprehensive meta-analysis with all previously eligible studies to evince a convincing relationship between HA and PFN/PFNA.

Following the PICOS framework, this study aimed to determine whether hemiarthroplasty (Intervention) leads to better outcomes compared to internal fixation (Comparison) for patients with unstable pertrochanteric fracture (Population). Outcomes were assessed across 18 different domains covering function, complications, and perioperative matters. We included evidence from randomized controlled trials (RCTs), retrospective studies, and prospective cohort studies (study design). We surveyed five central online databases—PubMed, Europe PMC, Web of Science, Embase, and CENTRAL (Cochrane Central Register of Controlled Trials)—for eligible studies. The analysis was conducted according to the PRISMA principles (19). The certainty of evidence was evaluated as per the GRADE system (20). Study quality was assessed with the Cochrane Risk of Bias (RoB) tool (21) and the Newcastle–Ottawa Scale (NOS) (22) for several research designs. To ensure methodological rigor and minimize potential bias, the study selection and data extraction processes were conducted in strict accordance with PRISMA guidelines. Two reviewers independently screened all titles and abstracts, followed by a double-blind full-text review. Data extraction was performed by two independent investigators using a pre-piloted, standardized form. Discrepancies were resolved through consensus or adjudication by a third senior investigator. The five databases were selected to ensure the most comprehensive coverage of mainstream biomedical literature. To uphold the global generalizability of our findings, no language restrictions were applied during the search or screening phases. Furthermore, to mitigate the risk of incorporating low-quality evidence, we strictly restricted our inclusion to peer-reviewed, published literature. While gray literature was acknowledged, its exclusion was a deliberate methodological choice to safeguard the internal validity of the meta-analysis, ensuring that all synthesized data had been subjected to rigorous, independent scholarly scrutiny. This approach minimizes the potential for ascertainment bias and strengthens the overall quality of the evidence base.

For unstable pertrochanteric fractures, many surgeons claim that internal fixation has limited performance and therefore prefer arthroplasty in the event of anticipated device failure. For those who have hip issues or femoral head problems, arthroplasty is a logical choice. However, the removal of a healthy femoral head only due to concerns about the high risk of internal fixation failure is debatable. We conducted this up-to-date, comprehensive systematic review to provide clarity and elaborate results through data and appropriate analyses. Surgeons must distinguish statistical significance from clinical relevance when interpreting these findings. The high statistical heterogeneity observed across the included studies limits the direct translation of these pooled estimates into standardized clinical decision-making. Although high statistical heterogeneity (I² > 90%) was noted, it was not pervasive across all analyses as previously suggested. A clear divergence was identified: High I² values were confined to the nine domains involving subjective, patient-reported outcomes, whereas the nine objective domains demonstrated minimal heterogeneity (I² < 10% or 0%). Given that the directionality of effect remained consistent across all 18 domains—as confirmed by extensive subgroup and meta-regression analyses—the high I² in subjective metrics reflects real-world clinical diversity rather than methodological instability. By synthesizing I², PI, consistency in effect direction, and Egger's test results, our conclusions remain conservative and data-driven. We intentionally reported these results to remind readers that while objective outcomes are highly stable, the interpretation of subjective, high-heterogeneity results requires appropriate caution.

It is important to note that the PI reflects the range of true variance (τ2). While our PIs were wider than the CIs and crossed the null line, this pattern represents the inherent variability of the treatment effects across different clinical settings. Clinical significance should be interpreted by focusing on the PI's coverage area relative to the horizontal axis (clinical effect size). The fact that the PI spans the line of no effect suggests that while the treatment is effective on average, its real-world application may yield varying results, which serves as a crucial reference for personalized clinical decision-making.

For function comparison, we chose the Harris hip score as an index, and we reported results at consistent stages. In the first 3 months after surgery, the HA group performed better than the PFN/PFNA group, with a difference score of 12.92, which we consider substantial given a total of hundred points on the HHS scale. The HHS scale is mainly composed of five domains: pain, disability, function, deformity, and range of motion. We assume that the HA group will achieve greater early pain relief and experience early withdrawal of the cane or other support devices. Deformity and range of motion were only covered at a maximum of 4 and 5 points, respectively, which would not generate a difference of more than 12.92 points. At the 6-month stage, the HA group remained superior to PFN/PFNA, with a trivial difference of 5.61 points. On a 100-point scale, this is still meaningful. In fact, we will have more solid analysis results if the authors report the HHS in its four domains, i.e., Pain score, Function score, Deformity score, and Range of motion score, which will help us figure out which domain weighs more and matters, but few studies reported this. At the 12-month stage, the data indicated a seemingly insignificant result with publication bias, which means these results should be interpreted with caution with limited weight. Notably, the 12.92 points only revealed statistical significance, and any further clinical interpretation should be made carefully. Over the years, different studies have reported consistent, similar, or even contrasting results regarding functional outcomes measured using the HHS scale. The HHS remains a conventional and effective tool. However, after conducting a clearly specified and comprehensive synthesis of all studies that compare HA and PFN/PFNA to date, publication bias remained, suggesting that authors should use a more appropriate evaluation tool for HA and PFN/PFNA. The ambulation time had publication bias, and the theoretical result after the trim and fill process did not alter the conclusion. In clinical practice, ambulation time is largely determined at the surgeon's discretion. Concern about fixation failure, especially in unstable pertrochanteric fractures, may deter surgeons from aggressive prognosis plans, thus delaying the ambulation time of patients undergoing PFN/PFNA surgery. That may explain the bias, and we suggest treating ambulation time with extreme caution.

The 22.28-day advance proves the superiority of HA, but we have two concerns regarding this. One is that there are no objective guidelines or criteria that define a precise time for surgeons to instruct patients to begin full weight-bearing, because no patient's fracture type and fixation stability are identical. The second concern is that it appears unfair to compare arthroplasty and internal fixation surgery with respect to full weight-bearing time. These aspects contributed to the high I². In each study, consistently, HA demonstrated advantages that contributed to faster and better recovery, though the findings were not robust.

Longer operation time is associated with a higher risk to patients. However, whether a discrepancy of approximately 15 min defines that HA is better than PFN/PFNA is debatable. However, we must carefully consider whether a 15 min duration is clinically significant, especially for patients who can tolerate long surgeries.

Hospital stay appeared identical between the two groups. Hospital stay is affected by factors such as patients’ recovery, hospital administrative requests, and patients’ financial considerations. While these issues are important, they might not be consequential.

Blood loss measurement and transfusion requirements were at the discretion of the surgeons. This may explain the presence of high heterogeneity. In general, this difference is not decisive but could influence surgeons’ decision-making.

Mortality is meaningful, particularly as patients are senile or suffering from an extremely high-energy trauma. We categorized mortality into early and final with a 3-month border and detected no discrepancies. Surgeons should consider that both techniques are equivalent in terms of mortality when managing unstable pertrochanteric fractures.

A clear comprehensive scrutiny and explanation of complications can reduce surgeons’ uncertainty when choosing between arthroplasty and internal fixation, which is at the heart of the present debate. In general, the complication rates showed no difference between groups. The heterogeneity rate was as low as 6%, indicated in the results. However, we considered this result insufficient and proceeded with another analysis, separating the complications into the implant-related complications and implant-unrelated complications. Effectively, the separation procedure served as a subgroup analysis, even though the heterogeneity was low. The findings were interesting and meaningful. First, the implant-unrelated complications revealed no difference, which was consistent with the general complications. In an empirical sense, arthroplasty can be expected to give patients earlier and greater range of motion, and this will decrease the incidence of implant-unrelated complications such as bedsores, DVT, and superficial infection. But data did not support this assumption. This 12-study result, with a low I² of 9% and an insignificant difference, indicates that surgeons should not take such complications into consideration when making a surgical plan vis-à-vis unstable pertrochanteric fractures. By contrast, implant-related complications—including cut-out, failure, peri-implant fracture or peri-prosthesis fracture, and peri-implant infection—were deemed more closely associated with implant choice. HA only demonstrated a 0.57 times higher risk compared to PFN/PFNA. Although publication bias was detected, it was due to the small-study effect, as discussed earlier. This result is vital. However, we still insist on equivalence between PFN/PFNA and HA. The re-operation rate was not significantly different. Significant implant-related complications that did not require reoperation suggest that the instruments had limited effects on the consequential complications, i.e., those that require reoperation. Incidentally, four of six studies in the re-operation group were also included in the implant-related complication groups, which limited the bias caused by different baselines. Even when excluding the remaining two studies, the conclusion is consistent. We believe that surgeons should not overestimate the effect of complications, since the evidence-based results prove that there is no difference between the two methods for unstable pertrochanteric fractures. According to these results, HA may experience more complications, but they are not severe enough to alter re-operation rates.

We attempted to specify complications for deeper analysis, but only bedsores, DVT, and superficial infection were qualified for merging and analysis. Superficial infection demonstrated a trivial difference, which can be explained by the fact that the HA is more invasive and has a larger approach/incision.

While the certainty of evidence was categorized as “Low”—due to the inherent constraints of surgical research, NRSs, and our stringent inclusion criteria—the consistency of findings across 18 domains and the reported PIs provide a stable basis for these observations. Clinical application of these results should be approached with caution, considering individual patient contexts and the specialized nature of these surgical interventions.

Overall, HA demonstrated advantages, with better early HHS scores, around 100 mL less blood loss and transfusion, a shorter operative duration by nearly 15 min, a lower instrument complication rate that does not lead to more re-operations, and a slightly lower superficial infection rate. General and implant-unrelated complications were comparable between the groups. No mortality difference was detected, either in the early or final stage. None of these represent decisive factors that would persuade an orthopedic surgeon to favor hemiarthroplasty over proximal femoral nail (anti-rotation). This convinces us that surgeons should not interfere with a healthy femoral head in settings of unstable pertrochanteric fracture, unless arthroplasty must be conducted.

This study has several limitations. First, orthopedic surgery trials have inherent constraints, wherein complete blinding of patients and surgeons is extremely difficult to enforce. Patients can easily infer the operation they have undergone based on the incisions and post-operative instructions, and surgeons are likewise aware of the surgery performed. This leads to potential performance bias and detection bias. Second, although we merged a total of 27 studies, the data in several domains may not be sufficient for a solid result. Future studies are needed for a more precise result. Third, conventional evaluation methods or standards, such as the HHS, may not correctly represent the true characteristics of an intervention. A more detailed HHS report or adoption of a more pertinent assessment tool may address this limitation. Furthermore, patients’ baseline characteristics likely influenced treatment allocation, introducing potential indication bias between the hemiarthroplasty and PFN groups.

We systematically evaluated potential sources of bias—including patient frailty, surgeon experience, and implant design—through granular subgroup analyses and meta-regressions across 18 distinct domains. Despite these exhaustive efforts, no single clinical variable emerged as a decisive moderator or confounding factor. This absence of a “statistical fix” for the high I² in subjective metrics suggests that this heterogeneity is inherent and diffuse, arising from the cumulative interplay of various clinical protocols rather than a single identifiable source. By reporting PI alongside these results, we provide a transparent representation of this variability, offering a more realistic expectation of treatment effects across diverse clinical settings. More studies and higher-quality data are required in the future for better results. Finally, although most non-RCT studies achieved only a low methodological quality, this is the best level they can attain short of an outstanding performance. We believe that more RCTs will make our findings more convincing.

Both HA and PFN/PFNA are effective strategies for unstable pertrochanteric fractures. HA demonstrated modest advantages in early functional recovery, a lower incidence of implant-related complications, and better in partial perioperative conditions. But no significant differences were detected in long-term function, general complications, implant-unrelated complications, certain specified complications—such as bedsores and DVT—and re-operation rates due to complications. Any clinical interpretation should be performed carefully because of the inherent limitations in NRS and the high heterogeneity in the subjective domains.

Taken together, we conclude that HA achieves outcomes broadly comparable to PFN/PFNA, with equal effects and similar clinical features. Therefore, surgeons should not favor HA when they encounter an unstable pertrochanteric fracture.