Is Surgery Better Than Conservative Treatment for Proximal Humeral Fracture: A Systematic Review and Meta-Analysis

Background: Previous studies had compared the effectiveness of surgical and conservative treatment in proximal humeral fracture (PHF), suggesting that both treatments were effective for displaced fractures. However, which treatment is more effective, especially in the elderly patient population, remains controversial to date. This study aims to evaluate the ecacy of the two treatment methods, including functional outcome scores, complications, etc. Methods: We searched four databases (MEDLINE, Web of Science, EMBASE, Cochrane Library) from the inception of the databases to March 2020. Within the scope of the search, the publication time and language of the relevant literatures were not limited. The meta-analysis directly compared the results and complications of the two groups and subgroups. Results: We included 11 comparative studies into our meta-analysis. The contents of the study included 7 randomized controlled trials (RCTs) and 4 non-randomized controlled trials (nRCTs). The pooled data demonstrated that there was no signicant difference in postoperative Constant score between surgical and conservative treatment in the RCT subgroup, but there were differences among the nRCT and prosthesis subgroup. Although there was no signicant difference between the surgery and conversation group in the major postoperative complications, the subgroup analysis showed fewer complications in the joint prosthesis replacement group. Conclusion: Both treatments could provide better clinical results for patients. The joint prosthesis replacement may offer additional benets in terms of reduced complications and postoperative function. Furthermore, due to the increasing number of prosthesis replacements and the advent of the aging age, we should consider the individual differences between patients and enrich the clinical ecacy through the personal experience of surgeons before choosing treatment options.


Inclusion and Exclusion Criteria
The search results of the systematic review were independently evaluated by the two authors by means of viewing the full text or abstract. Any controversy between the two authors in the process was nally resolved by the intervention of the third author (Chi Zhou). The included articles were RCTs and nRCTs, which directly compared the e cacy of surgery and conservative treatment of adult PHF followed up for at least 1 year. Surgical intervention included plate, prosthesis replacement and others, while conservative treatment included non-invasive operation, such as bandages, slings and so on. The main outcome measures (Constant score) should be included in the study results, and the postoperative complications should be expressed. Cadaver and repeat studies, case reports, abstracts, letters to the editor, commentary reports, reviews, animal experiments, etc. were excluded. No comparison between the two treatments was included in the literature and incomplete outcome data were also excluded.

Data Extraction
Data containing information such as basic information and main results were independently and carefully summarized from the included studies and lled into a table designed before the study by two evaluators (Xiaobo Wang, Jian Li). All differences between them nally reached a consensus through discussion. This standardized extraction table included the following: (1) study information (ie, author, year of publication, country, journal, and design of study); (2) study population information (ie, age, gender, design number of patients, and fracture classi cation); (3) surgical implant, conservative treatment, incision method and follow-up; (4) primary outcomes (ie, function score and complications).

Study Quality Assessment
In terms of research methodology, the COCHRANE tool was used to assess the risk of bias to determine whether biases affected the nal outcome in the RCTs. The nRCTs were evaluated using the NOS scale. These studies were evaluated independently by the two authors (Leilei Chen, Jingli Xu). Disagreement between them were resolved through discussion to reach a consensus or by inviting the third author to participate in the discussion.

Statistical Analysis
For dichotomous outcomes, such as complications, the risk ratio (RR) and 95% con dence intervals (CIs) were used to evaluate the relevant studies. Mean difference (MD) was used for continuous variables, including constant score. For some special continuous variables in the articles, such as means and range, we used statistical algorithms to make standard deviations estimates. In the process of studies, only means and standard deviations could be included in the analysis. In the results, p and I 2 values were used to analyze whether there was heterogeneity. When p>0.1 or I 2 < 50%, a xed-effect model was adopted to the meta-analysis, when on the contrary, a random-effect model was applied. Sensitivity analyses were performed to evaluate the stability of the results (if necessary), and subgroup analyses were performed based on RCT, nRCT, implants. Forest plots were used to show the results of individual studies and respective pooled estimates of effect size.
The publication bias of any outcomes was evaluated by funnel plots. If the sample size was small, this may not be carried out. All statistical analysis processes were ran by Review Manager (version 5.3.5 for Windows). Since STATA software could display the p-value of funnel plot, which could better quantify the result, all funnel plots were ran by STATA software (version 15.0 for Windows) during this process.

Study Selection
We identi ed 3,841 potentially relevant articles searched from four electronic databases based on our previous search strategy. A total of potentially references (n=1836) were extracted by endnote software and manual inspection. After reading the titles and abstracts, 1949 papers were excluded. We read the full text of the remaining 56 articles, and 45 were excluded for the following reasons: other surgical methods, lack of raw or complete data, randomized trial designs, letters, case reports, commentaries, and reviews. According to the inclusion criteria, 11 nal publications published from 1997 to 2019 were selected through full-text reading and included in the meta-analysis ( Figure 1). The 11 articles included complete data comparing the results of surgical and non-surgical treatment of PHF. There were 590 patients, including 298 in the surgical group and 292 in the conservative group. According to the different research methods and surgical treatment methods, we classi ed the paper into the groups of random and non-random sets, plate internal xation and prosthesis replacement. A screening of the reviews through the full text of the 11 articles, no additional research was found to be available.

Study Characteristics and Quality
Among the 590 patients in these 11 studies, most of them were elderly, mostly females. The preoperative diagnostic classi cation was mainly focused on Neer classi cation of 3-part and 4-part. All of the included articles reported on the authors, year of publication, country, treatment, follow-up, fracture classi cation, functional score, etc. (  [17][18][19][20], of which 5 were plate implants and 4 were prosthesis replacements for PHF. Most surgical incisions were through the deltopectoral incision. Conservative treatment techniques mainly included sling and bandage. The quality assessment of these studies demonstrated a moderate risk of bias (Figure 2, Figure 3 and Table 2). Whether there was a representation of the exposed cohort; 2. The non-exposed cohort study came from the same community as the exposed cohort; 3. Ascertainment of exposure studies had been conducted based on records or investigations; 4. No outcome events were studied before the study began; 5A. comparability of cohorts (based on the design or analysis) about age; 5B.
comparability of cohorts regarding disease extent and its duration; 6. Assess outcomes based on relevant evidence; 7. Follow-up was adequate for outcome; 8. adequacy of follow-up of cohorts.
Outcome Functional outcome Constant score is one of the most commonly used functional indicators in patients with PHF. Due to the signi cant in uence of different experimental design methods and different implant surgery methods on the postoperative, we conducted subgroup analysis on them respectively, hoping to make more effective observation at the initial stage.
In this study, there were 7 RCT literatures, which were tested for heterogeneity, I 2 = 0% < 50%, and the p value in the Q test was 0.77 > 0.1, indicating that there was no heterogeneity among the selected articles in this study, and xed effect was selected for meta-analysis.
Sensitivity analysis was conducted on 7 research literatures, and none of them had a signi cant impact on the results of this meta-analysis, indicating that this study had good stability ( Figure 4). It could be clearly seen from Figure 5 that the funnel plot of the RCT group was basically symmetrical, and the bias test at the same time showed that p value = 0.23 > 0.05, and there was no publication deviation in the article of this group. Similarly, there was no heterogeneity among the articles in the nRCT group. Subgroups in forest plot shown that RCT group Z = 1.4, p = 0.16 > 0.05, there was no signi cant difference between the conservation group and surgery group, but nRCT trails showed signi cant difference in Constant score (p = 0.004 < 0.05).
Among the 11 literature studies, most treatments were treated with sling or bandage. The surgery group could be divided into two methods (except Karol Zyto [10] and Massimo Innocenti [19] using k-wires): internal xation plate and prosthesis replacement. Therefore, we divided the study into two subgroups based on built-in differences. In the comparison between the plate group (conservation and surgery group), Chi 2 = 0.18, P = 0.67 > 0.05, there was no signi cant difference in the functional score of the PHFs ( Figure 6) and the prosthesis group Z = 2.35, p = 0.02 < 0.05, with signi cant difference, which may be caused by the small sample size of this subgroup, and the use of Reverse Shoulder System implant by Matthieu Chivot [20] and Yaiza Lopiz [15], while the use of hemiarthroplasty by Per Olerud [11] and Harm w. Boons [14]. Although both were arthroplasty, there were still some differences in postoperative functional scores between the two surgical methods.

Complications
We extracted most common complications of PHFs from 11 articles: nonunion and osteonecrosis. In 7 studies with complications, RR of the 7 studies was 0.67, the 95% con dence interval was 0.32 to 1.41, Z = 1.04, P = 0.30 > 0.05, and there was no signi cant difference in fracture nonunion and osteonecrosis between the two treatment from Figure 7. The funnel plot was basically symmetrical, and the bias test p = 0.76 > 0.05. There was no deviation, as we could see from Figure 8. Meanwhile, there was no signi cant heterogeneity in subgroups, but Z = 2.14, P = 0.03 < 0.05 between the subgroups, Figure 9 demonstrated that there was a signi cant difference in the prosthesis group in terms of complications.

Discussion
This meta-analysis included 11 studies (7 RCTs and 13 nRCTs) that evaluated 590 patients and compared postoperative functional scores and complications between surgery and conservative treatment. The pooled data shown no difference in functional scores between the two treatment modalities in RCT group and plate group. In general, there was no signi cant difference in the incidence of common complications between the surgery and conservation group, however, there were a signi cant difference in the subgroup of prosthesis group.
The proximal humeral anatomy is complex, fractures are extremely prone to occur during trauma [21]. It is believed that conservative treatment will not cause secondary trauma, and the medical cost is low, so it is suitable for the elderly or patients with intolerance to surgery. Although conservative treatment could lead to complications such as fracture displacement and joint stiffness, some scholars reported that nonsurgical treatment had a good clinical e cacy. Den et al. [22] treated comminuted PHFs in the elderly with humeral head replacement and nonsurgical methods respectively, and found that the long-term e cacy of non-surgical treatment was better than the former. At present, with the improvement of technology, LPHP and PHILOS plates were gradually applied and popularized. In particular, the design of PHILOS plate could break through the shortcomings of large incision exposure and severe local blood ow destruction in the traditional plate exposure process [23]. At the same time, it could also achieve reliable xation effect through the cross-locking technology of multi-angle screws. The design of rotator cuff repair hole more achieves the combination of bone and soft tissue, plays the role of internal xation bracket, and promotes the healing of fractures [24,25]. In contrast to intramedullary nailing and prosthesis replacement techniques, surgeons are more likely to master plate internal xation, which is more adaptable to PHF and easier to popularize. Nevertheless, Doshi et al. [26] proposed that the application of plate technology was based on individual differences of patients, and patients should be strictly selected and familiar with relevant anatomy. But Stanbury SJ [27] believed that RSA was superior to internal xation.
In clinical practice, we often decide whether to use prosthesis based on age and degree of fracture comminution. When using different types of prosthesis, we found that the clinical effect of hemiarthroplasty often reached the concept of "all or nothing", with good e cacy in some patients and poor in others. But RSA could often achieve better results. Emmanuel Maugendre [28] believed that RSA had unique advantages in the treatment of PHF in elderly, with a 1-year survival rate that was 7-13% higher than that of femoral neck fractures. Jason Ferrel [29] agreed that RSA signi cantly improved forward shoulder exion (range of motion greater than 10°) and had a relatively low revision rate (0.93%) compared to hemiarthroplasty. In this meta-analysis, according to the complications described in the literature [11,14,15,20], we found RSA compared to hemiarthroplasty had a lower incidence of complications after surgery, Clark, NJ. [30] clinical trials also proved the point, at the same time, they thought primary RSA was effective in the treatment of PHF, lower in medical and surgical complications, RSA was also effective in patients over 80 years of age after a comprehensive evaluation. However, RSA is not a panacea, and it has its own limitations. For example, Cho CH [31] believed that we should pay great attention to the complications of acromion fractures that commonly occur after surgery. Since RSA is limited by glenoid loosening and instability, prosthesis implantation failure may occur, Song IS et al. [32] believed that hemiarthroplasty could better solve such problems, and postoperative follow-up score and joint range of motion were satisfactory.
Surgical treatment of PHF could be performed by different incision methods. In the meta-analysis we performed above, the extracted data shown that 9/11 studies used deltopectoral approach, and the other two were minimally invasive incisions [11,16]. The choice of approach is determined by the experience of different surgeons that a particular incision is more effective, safer, or that they are better at operating a particular approach. Although the traditional deltopectoral approach had the advantage of showing clear intermuscular space, its exposure to the posterior displaced greater tuberosity was limited [33]. Guilherme Grisi Mouraria [34] demonstrated that the anterolateral approach could better expose the lateral surface of the humerus, which was conducive to the operation, but the axillary nerve injury and subacromial impingement were still complications that couldn't be ignored. Other scholars believed that the subacromion anterolateral approach through the deltoid muscle had the advantages of minimally invasive, clear exposure, and no signi cant difference with the traditional deltopectoral approach in terms of reduction quality [35]. Harmer LS et al. examined two commonly used surgical approaches from cadaver tests, and they also demonstrated that the deltopectoral approach provided a better exposure to the front operating markers, while the anterolateral acromial approach provided a better rear exposure [36].
Due to the anatomical complexity of PHF and the nature of the fractures within the joint, many clinical complications occur, such as arthritis, joint stiffness, fracture nonunion, osteonecrosis, pain, joint dysfunction, fracture displacement, infection, bone resorption, and so on. However, nonunion of fracture and osteonecrosis are the most common and most likely causes of secondary surgery. Therefore, we summarized and analyzed the data of these two complications. Boesmueller  complications in locking plate group was signi cantly higher than that in RSA group (15.8%). Therefore, it was recommended to perform RSA in the rst phase in elderly patients with complex PHF [38]. This conclusion was consistent with our understanding.
Although we took various measures to reduce the in uence of additional factors on the analysis results, there were still many potential limitations in this meta-analysis. First of all, no matter what kind of design is included in the standard, there will be methodological defects, so it is easy to cause some studies to be ignored. We consulted with professional librarians to get a better retrieval strategy, so as to minimize the impact on the results. Second, we conducted a comprehensive evaluation of RCTs and nRCTs at the initial stage, especially in terms of complications, which may lead to the deviation of the results due to problems in the trial design. Therefore, the analysis of subgroup analysis could ensure the comparability and scienti city of the results. In this process, we also found that the results of RCTs were more complete, which proved the accuracy of the design scheme and method, and the trial literature without omissions and exclusions were more accurate for the overall assessment. Thirdly, based on the grouping of the built-in objects, the overall classi cation into the plate group and the prosthesis group would bring about some differences in the results. As time goes on, more and more effective methods gradually replace the traditional methods, and meta-analysis is a systematic review process, which inevitably integrates a variety of xed surgical methods. The subgroup analysis of these two methods was adopted to reduce the difference of results caused by the diversity of treatment methods. Fourth, according to the statistics from the baseline data, many surgical methods would produce different complications, and the differences here were caused by various reasons such as fracture types, and differences in the patients under surgery, which brought di culties to the statistics of data collection. We selected the most common cases of fracture nonunion and osteonecrosis as a summary of complications, reducing individual differences in content, although this method may lead to some variation. Fifth, different classi cation, people of all ages would have different results, we conducted a comprehensive meta-analysis which would inevitably face the problem above, however, the above analysis, the cases of baseline data embodied mostly elderly patients and Neer3, 4 type fracture, common clinical characteristic of PHF determined the particularity. The relative limitation of the scope also helped us to reduce the impact on the results to some extent.
This meta-analysis is the rst of its kind to use RCT and nRCT, as well as the combination of plate internal xation and prosthesis replacement groups. In the initial strategy formulation, in order to avoid omissions, we adopted a more complete retrieval than before. In this process, unlike most articles, we searched, extracted and evaluated by two authors from different regions, and discussed in controversial cases through webinar, further reducing the in uence of additional factors. Moreover, in the past two years, there had been no analysis of such content by any author. Nowadays, with the rapid update of implants and surgical techniques, meta-analysis that cannot be updated in time will obviously miss more important information comparisons. Therefore, our assessment results were more accurate and timely. All studies included a direct comparison of conservative versus surgical treatment for PHFs, and we rigorously evaluated their quality. We conducted this study based on the PRISMA statement.

Conclusion
Our meta-analysis shown no signi cant differences in major complications between conservative and surgical approaches for PHFs, but there were some differences in the subgroup analysis, such as postoperative scores for the nRCT group and the prosthesis implant group, and postoperative complications for the prosthesis replacement group. Each patient should be considered individually before choosing the surgical procedure. Surgeons using their personal clinical experience to make preoperative planning could reduce the incidence of surgical failure. However, this study remains relatively limited, and more large-scale clinical trials, systematic reviews will be necessary in the future to con rm these ndings.

Declarations
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.

Availability of data and materials
All data and materials used in this research are freely available. All data generated or analysed during this study are included in this published article and its supplementary information les.   Forest plot representing functional scores (two implant methods) for surgery versus conservative treatment Figure 7 Forest plot representing the risk ratio of complications between the surgery group and conservation group Forest plot representing the risk ratio of complications between the surgery group and conservation group

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.