Meta-analysis in periprosthetic joint infection: a global bibliometric analysis

Background Periprosthetic joint infection (PJI) is the most serious complication of joint replacement surgery. Further comorbidities include bedsore, deep vein thrombosis, reinfection, or even death. An increasing number of researchers are focusing on this challenging complication. The aim of the present study was to estimate global PJI research based on bibliometrics from meta-analysis studies. Methods A database search was performed in PubMed, Scopus, and Web of Science. Relevant studies were assessed using the bibliometric analysis. Results A total of 117 articles were included. The most relevant literature on PJI was found on Scopus. China made the highest contributions to global research, followed by the USA and the UK. The institution with the most contributions was the University of Bristol. The journal with the highest number of publications was The Journal of Arthroplasty, whereas the Journal of Clinical Medicine had the shortest acceptance time. Furthermore, the top three frequently used databases were Embase, MEDLINE, and Cochrane. The most frequent number of authors in meta-analysis studies was four. Most studies focused on the periprosthetic hip and knee. The alpha-defensin diagnostic test, preventive measures on antibiotics use, and risk factors of intra-articular steroid injections were the most popular topic in recent years. Conclusion Based on the results of the present study, we found that there was no single database that covered all relevant articles; the optimal method for bibliometric analysis is a combination of databases. The most popular research topics on PJI focused on alpha-defensin, antibiotic use, risk factors of intra-articular steroid injections, and the location of prosthetic hip and knee infection.


Introduction
Periprosthetic joint infection (PJI) is a serious and challenging complication after joint replacement. Due to the lack of consensus on the management of PJI, physicians often face uncertainty. However, errors in diagnosis and treatment result in increased healthcare costs, reinfection, or mortality [1]. Publications play an essential role in guiding and improving disciplinary development.
Bibliometric analysis is a widely used tool that uses mathematical and statistical methods to assess research trends and growth. Another commonly used tool is meta-analysis, a statistical method of collecting and analyzing results from multiple studies to find or prove the viewpoint or relationship between variables. These two methods have been applied extensively in orthopedic research [2][3][4][5][6]; however, there were few publications on the use of meta-analysis in bibliometric studies [7,8]. To date, no such studies have been performed on orthopedic research.
The choice of database and the search strategy used are a crucial step in bibliometric studies and metaanalysis. Due to differences in exporting information between different databases, most bibliometric studies use a single database for statistics and data analysis [9,10]. Such differences regarding PJI research remained unknown. Accordingly, the present study performed a bibliometric analysis to determine the following: (1) the most suitable database (PubMed, Scopus, Web of Science) for bibliometric analysis [11]; (2) global research characteristics of PJI through the analysis of meta-analysis publications; (3) countries with the most research on the meta-analysis of PJI; (4) the diagnostic method with the highest sensitivity preoperatively, intraoperatively, and before reimplantation based on meta-analysis results; (5) the effective prevention measurement or risk factor on the meta-analysis of PJI; and (6) conclusions supported by the current meta-analysis.

Data sources and searches
We systematically searched PubMed, Scopus, and Web of Science from inception to December 2019. The search algorithm used was the following medical subject headings (MeSH) or keywords: "arthroplasty", "joint prosthesis", "joint replacement", "periprosthetic joint", "prosthetic joint", "infection", "infectious", "infected", "meta analysis", and "meta-analysis". As this study was performed using global research, there were no language restrictions.

Data collection
Data were extracted independently by two reviewers (LC and COT). Discrepancies were adjudicated by the third author (XC). Information on all eligible publications including the title, author, year of publication, country, institution, journal, keywords, citations, state of the manuscript, language, number of studies, impact factor, software, database, search algorithm, and subject information were collected. The number of citations was based on the final result, in the case that no single database covered all citation information. Subsequently, citations were collected from Google scholar. Finally, two authors (LC and COT) manually screened and analyzed the publication information in Microsoft Excel (Microsoft, Redmond, Washington, USA, 2010) and EndNote X7 (Thomson Reuters, New York, NY, USA, 2013).

Database results
Results from the search strategy demonstrated that the database with the most publications was Scopus (570), followed by Web of Science (341), and PubMed (243). The greatest number of identical articles was through the combined database of Web of Science and Scopus (Fig. 1). Finally, a total of 117 related articles were included. Of these, the database with most publications on the meta-analysis of PJI was Scopus, followed by Web of Science and PubMed. Web of Science and PubMed had most missed articles compared with other databases (Figs. 2 and 3).

Characteristics of meta-analysis of PJI research General data
Among the 117 meta-analysis articles, the earliest publications were from 2007. The greatest number of articles were published in 2018 (24), followed by 2017 and 2019 (21 each). The trend line indicates an annual increase in the number of articles (Fig. 4). One hundred and fourteen articles were in English, and three other articles were each published in Chinese, German, and Persian. In all meta-analyses, the number of studies included ranged from 4 to 203, with the highest number 12 (n = 11 publications), followed by eight (9) as well as six and eleven studies (8 each).

Countries
Nineteen countries published meta-analyses on PJI. Of these, China was the most productive country, with all publications stemming from 15 cities/provinces. The highest number of articles originated from Shanghai, followed by Beijing (Fig. 5). The country with the second highest number of publications on PJI was the US, followed by the UK (Table 1).

Institutions
A total of 76 institutions made contributions to this field. The institution with the greatest number of publications was the University of Bristol with 11 papers, followed by Shanghai Sixth People's Hospital (8). The Rothman Institute and General Hospital of the Chinese People's Liberation Army were third, with each publishing five research articles. Fourteen institutions published more than one paper, with 50% originating from China (Table 2).

Authors
The number of authors of a single article ranged from 2 to 37. The largest number of collaborating authors was four (27), followed by six (25) and 5 authors (20; Table 3). The author with most first authorships was Setor K. Kunutsor (10), followed by Xinhua Qu (3). Ten first authors wrote more than one meta-analysis, with 50% published by research institutes in China (Table 4).

Journals
Meta-analysis studies were published in 54 different journals. The journal with most publications was the Journal of Arthroplasty, with 15 publications. The Journal of Bone and Joint Surgery ranked second with eight publications, whereas PLoS ONE was third with seven. Nineteen journals had more than one publication (Table 5). In 2019, an impact factor was available for 42 journals. The list of top 10 journals with the highest impact factors is shown in Table 6.
From all publications, the date of receipt was available for 89 papers, whereas the date of acceptance for   (66 papers). From receipt to online publication, the average number of days was 157.48 (69). There were six articles accepted in less than 30 days after submission. The journal with the shortest acceptance time was the Journal of Clinical Medicine (16 days), followed by the Journal of

Search algorithm and keywords
One hundred and two meta-analyses were retrieved from the search strategy, which were exported to Microsoft Excel. All keywords or MeSH were combined. PJI-related keywords were 196, followed by diagnosis (179), prevention (82), risk factor (74), and outcome (60). All keywords are presented in Supplementary 1. From 71 publications, 389 keywords were exported. Periprosthetic joint infection (41) was the most commonly used keyword, followed by meta-analysis (29) and total knee arthroplasty (20; Table 8).

Database and software
After combining all databases from 116 articles, there were a total of 52 databases. Embase was the most described database (101), followed by MEDLINE (80), and Cochrane (74; Table 9). Three databases were most frequently searched (40), followed by four (22), and five (16). The most combined database group was Cochrane Library + Embase + MEDLINE/PubMed (10), followed by Embase + MEDLINE (6), and Cochrane Library + Embase + MEDLINE + Web of Science (5).

Subject
Location Information on the site of prosthetic joint infection from the included meta-analysis were found in 112 papers. The location with the highest number was the knee (93), closely pursued by the hip (90), shoulder (23), elbow (16), and ankle (3).
Diagnosis of PJI From 40 diagnosis-related metaanalyses, 72 tests were related to preoperative examination, followed by intraoperative methods (12), and test prior to reimplantation (14). Synovial fluid alphadefensin had highest pooled sensitivities in the list of preoperative examinations, pursued by serum IL-6 and bone scintigraphy. From all intraoperative examinations, tissue polymerase chain reaction (PCR) was the most sensitive method, followed by sonicate fluid into blood culture bottles (BCB) and PCR. Tissue culture was the most sensitive method before reimplantation, followed by the percentage of polymorphonucleocytes in synovial fluid (PMN%), and synovial fluid culture ( Table 10). The most frequent diagnostic method used was synovial fluid (16), followed by imaging (10), and periprosthetic tissue (7; Fig. 6).
Risk factor and prevention Twenty-three articles described 64 possible risk factors. The location of the risk factor was outlined in 20 studies, with the majority in the hip and knee (Table 11). Nine preventive measures were described in 17 articles, with all focusing on the hip and knee (Table 12).
Comparative analysis There were 26 comparative analytic studies from all meta-analyses, with most related to the hip and knee (11), followed by the hip as well as the hip and knee (7 each). There was no statistical difference found in 13 comparison studies (Table 13).

Discussion
This bibliometric study presents 117 meta-analysis results from three databases (PubMed, Scopus, and Web of Science), with the greatest number of relevant papers in Scopus. Furthermore, we compared all databases with or without a search strategy, with PubMed demonstrating the greatest difference among the three databases. When combined with other databases, the missing information from the search strategy could be supplemented. All results could not be found with any of the databases, with or without a search strategy, whereas the combination of PubMed and Scopus enclosed all results without a search strategy. In addition, all available information from the database and search algorithm were collected and combined. Three to five database groups were found to comprise most options for meta-analysis. Embase, MEDLINE, and Cochrane were the top three most commonly used databases and were also mostly used for meta- Table 7 The 50 most cited meta-analysis studies on PJI ranked by citation. Allogeneic blood transfusion is a significant risk factor for surgical-site infection following total hip and knee arthroplasty: a metaanalysis 74 16 Inflammatory blood laboratory levels as markers of prosthetic joint infection: a systematic review and meta-analysis 74 17 Meta-analysis of sonication fluid samples from prosthetic components for diagnosis of infection after total joint arthroplasty 68 18 Re-infection outcomes following one-and two-stage surgical revision of infected hip prosthesis: a systematic review and metaanalysis 65 19 Use of static or articulating spacers for infection following total knee arthroplasty 61 20 PCR-based diagnosis of prosthetic joint infection 57 21 Preoperative aspiration culture for preoperative diagnosis of infection in total hip or knee arthroplasty 56 22 Synovial fluid biomarkers for the diagnosis of periprosthetic joint infection: a systematic review and meta-analysis 47 23 Evaluation of white cell count and differential in synovial fluid for diagnosing infections after total hip or knee arthroplasty 35 24 Prosthesis infection: diagnosis after total joint arthroplasty with three-phase bone scintigraphy 35 25 Diagnostic performance of FDG PET or PET/CT in prosthetic infection after arthroplasty: a meta-analysis 34 26 Procalcitonin and a-defensin for diagnosis of periprosthetic joint infections 34 27 The accuracy of imaging techniques in the assessment of periprosthetic hip infection: a systematic review and meta-analysis 32 28 Control strategies to prevent total hip replacement-related infections: a systematic review and mixed treatment comparison 30 29 Outcomes following debridement, antibiotics and implant retention in the management of periprosthetic infections of the hip: a review of cohort studies 30 30 Total joint arthroplasty following intra-articular steroid injection: a literature review 30 31 Do intra-articular steroid injections increase infection rates in subsequent arthroplasty? A systematic review and meta-analysis of comparative studies 29 32 Postoperative antibiotic prophylaxis in total hip and knee arthroplasty: a systematic review and meta-analysis of randomized controlled trials 28 33 What is the accuracy of nuclear imaging in the assessment of periprosthetic knee infection? A meta-analysis 28 34 Does previous intra-articular steroid injection increase the risk of joint infection following total hip arthroplasty or total knee arthroplasty? A meta-analysis 27 35 Systematic review and meta-analysis of randomized controlled trials of antibiotics and antiseptics for preventing infection in people receiving primary total hip and knee prostheses 25 36 Diagnostic accuracy of C-reactive protein for periprosthetic joint infection: a meta-analysis 24 analysis. The available search algorithm exported from 102 publications provided a reference for scholars for a further literature search and study design. Meta-analysis could offer a useful effective reference to support or refute controversial conclusions from multiple studies. The bibliometric analysis showed that the first meta-analysis appeared in 2007, with an increasing trend in the ensuing years. The growth number likely reflects the development of the subject with an academic dispute, and the International Consensus Meeting on PJI also indicated the presence of disparate opinions on the management of PJI [54]. The current study also presented China as having the greatest number of publications in meta-analyses. This may be attributed to the fact that Chinese physicians are placed under immense pressure to publish under the healthsystem reforms [55]. Furthermore, the Chinese Association of Orthopaedic Surgeons (CAOS) play close attention to infection after joint arthroplasty. CAOS, which comprises the Chinese prosthetic joint infection society, was established in 2018 and perform PJI research by multiple centers. In China, Beijing and Shanghai had the greatest number of publication of PJI meta-analysis than other cities and is most likely related to a larger number of research institution concentrated in both regions. Institutions from the UK had the largest number of publications, with the majority from the University of Bristol. Analysis of author information showed that at least two authors were required for meta-analysis, with the most frequent number of collaborators was four. In meta-analysis studies, Setor K. Kunutsor from the University of Bristol had the most publications as the first author.
In all meta-analysis papers, the Journal of Arthroplasty had the most number of relevant papers. With more than 20 citations, PLoS ONE had the greatest number of publications from the most cited publication list. The Journal of Clinical Medicine had the minimum time from receipt to acceptance. In addition, the bibliometric method report showed most articles to be received and accepted on Wednesday.
In the top 10 most popular keywords on PJI metaanalysis, two keywords were related to treatment and diagnosis, with two-stage exchange and alpha-defensin in the top 10. Three keywords were associated with the location of PJI, with the majority on the hip and   Use of anti-granulocyte scintigraphy with 99mTc-labeled monoclonal antibodies for the diagnosis of periprosthetic infection in patients after total joint arthroplasty: a diagnostic meta-analysis 24 38 Serum and synovial fluid interleukin-6 for the diagnosis of periprosthetic joint infection 23 39 The application of sonication in diagnosis of periprosthetic joint infection 22 40 The impact of neuraxial versus general anesthesia on the incidence of postoperative surgical site infections following knee or hip arthroplasty a meta-analysis 22 41 Do 'Surgical Helmet Systems' or 'Body Exhaust Suits' affect contamination and deep infection rates in arthroplasty? A systematic review  knee. Identical results were also found in regard to the location, with the top three keywords knee, hip, and shoulder. The most frequently used software in the meta-analysis were STATA, REVMAN, and Meta-Disc. Among the diagnosis list in meta-analysis studies, the synovial fluid test was the most frequently used preoperative examination (64%). The most popular diagnostic test applied in recent years was synovial fluid alpha-defensin and has been incorporated in the 2018 Musculoskeletal Infection Society (MSIS) definition as one of the minor criteria [56]. When compared with conventional diagnostic methods, such as ESR, CRP, synovial fluid culture, and synovial fluid PMN%, alpha-defensin showed better sensitivity, especially in cases receiving antibiotics before joint puncture [57,58]. In recent years, synovial fluid alpha-defensin could be detected using two different methods. One assay is the enzyme-linked immunosorbent assay (ELISA), which is performed in a laboratory with results obtained within 24 h. The second assay is the lateral flow device, which rapidly detects infection within 20 min without the need for a laboratory. Accordingly, pooled results supported the higher sensitivity of the synovial fluid alpha-defensin ELISA compared to the lateral flow test [16,20,23,27]. The current meta-analysis demonstrated synovial fluid alpha-defensin to have the highest sensitivity in the diagnosis of PJI. As it represents a nonmicrobiological test, it could be used as a reliable reference for intraoperative microbiological diagnosis. Preoperative tests with the lowest sensitivities were synovial fluid gram staining (GS), synovial fluid procalcitonin (PCT), serum white blood cells (WBCs), and serum PCT, which were all found to have a sensitivity of less than 60%.
Sonicate fluid and periprosthetic tissue were performed most intraoperatively, whereas tissue PCR and sonicate fluid BCB were the most sensitive tests in tissue and sonicate fluid, respectively. In 2013, Qu et al. [37] performed the first meta-analysis of PCR in the diagnosis of PJI. The authors found that the tissue PCR had a higher sensitivity than synovial fluid PCR and sonicate fluid PCR (95% vs. 84% vs. 81%, respectively). However, tissue PCR showed the lowest specificity compared to synovial and sonicate fluid PCR (81% vs. 89% vs. 96%, respectively). However, this is in contrast to the study by Huang and colleagues [59], in which tissue PCR had lower sensitivity of 34% and the highest specificity of 100% among the three types. Due to limited data and that the included studies on tissue PCR were performed between 1999 and 2012 [37], the diagnostic value of tissue PCR remains unclear. The meta-analysis of sonicate fluid BCB presented a sensitivity of 85% and a specificity of 86% [46]. Compared to the conventional culture of sonicate fluid, BCB culture was more sensitive in AGS antigranulocyte scintigraphy, BCB blood culture bottles, CI confidence interval, CRP C-reactive protein, CT computed tomography, ELISA enzyme-linked immunosorbent assays, ESR erythrocyte sedimentation, GS Gram staining, LE leukocyte esterase, IL interleukin, PCR polymerase chain reaction, PCT procalcitonin, PET positron emission tomography, PMN% polymorphonucleocyte percentage, Sen sensitivity, Spe specificity, WBC white blood cell, WCC white cell count patients with or without antibiotics and also detected infection within a shorter time than normal medium sonicate fluid culture [60][61][62][63]. Yet, the drawback of sonicate fluid BCB was the rate of false-positives, which was caused by contamination during the inoculation procedure of BCB with sonicate fluid. Therefore, careful handling is required to minimize contamination [64,65]. Tissue and tissue swab GS were the two least frequently applied intraoperative tests, with a sensitivity of less than 20%. Diagnosis prior to reimplantation always posed difficulty. In the meta-analysis study by Lee and colleagues [52], tissue culture demonstrated the highest sensitivity before reimplantation, which was based on two included studies (82%). Another meta-analysis study by Bian and co-workers [53] estimated the various tests during the first stage and/or predicted failed reimplantation beyond the second stage, with tissue culture showing a sensitivity of 30%, which was based on the results of nine studies. Synovial fluid PMN% demonstrated the highest sensitivity of 70% in the study by Bian et al., while the specificity was low at 71%. Interestingly, the author found that the spacer sonication fluid culture was the most accurate method with an area under the receiver operating characteristic curve of 0.8089. There was no single test that achieved an ideal result, with combined multiple tests to evaluate infection still required [53].
There were 40 meta-analyses related to risk factor and prevention, with the majority of articles on preventive measures focusing on systemic or location antibiotics use. In regard to the risk factor, most concerns focused on intra-articular steroid injections, followed by age, diabetes mellitus, and rheumatoid arthritis.
The top three comparison studies focused on cemented vs. cementless total joint arthroplasty, the outcome of using different types of spacers, and the outcome of one-stage vs. two-stage exchange.  Cemented fixations were revealed to increase the overall PJI risk in comparison to uncemented fixations [66][67][68]. Interestingly, there was no significant difference in the eradication rate between articulating and static spacers in the infected knee replacement [69,70]. The current meta-analysis supports that the infection control or reinfection rate of one-stage or two-stage exchange did not significantly differ in the hip, knee, elbow, and shoulder [71][72][73][74][75].
There are several limitations to the present study. First, the database of present bibliometric analyses was collected from three databases. Compared with results from without the search strategy, several articles were missed when using the search strategy, especially in PubMed. However, working with multiple databases could reduce this problem. In addition, we also collected database information from all metaanalyses. Embase, MEDLINE, and Cochrane were the most widely used databases. However, whether these databases were appropriate for bibliometric analysis remains unclear and requires further investigation. Second, due to the export of all meta-analysis information between different databases with disparate formats, a visualized analysis could not be performed. Third, although meta-analysis results on diagnosis, risk factors, prevention, and comparative studies were shown, the heterogeneity and quality of included meta-analysis studies were not considered. In the subgroup diagnosis, since there is no gold standard for the diagnosis of PJI, different culture results are obtained from the various diagnostics tests. The pooled sensitivity and specificity of meta-analysis are then further affected by potential false positive or negative results. Fourth, the current study only presented metaanalysis results and did not reflect the complete perspective of PJI research. The overall trends in this field are required to further confirm.

Conclusion
The bibliometric analysis that presented global PJI research of meta-analysis studies showed an increasing trend between 2007 and 2019. The Embase database and STATA software were most frequently used for metaanalysis. Most studies focused on the periprosthetic hip and knee. The diagnostic alpha-defensin test, preventive measures on antibiotics use, and risk factors associated with intra-articular steroid injections were the most popular topics in recent years.   The hidden cost of commercial antibiotic-loaded bone cement: a systematic review of clinical results and cost implications following total knee arthroplasty Knee N Antibiotic-impregnated bone cement N the present study did not support the topic to be an effective prevention measure for PJI, Y the present study supported the topic to be an effective prevention measure for PJI  : The meta-analysis results showed that there was no significant relationship between the two topics. : The meta-analysis results demonstrated that the topic had a higher infection/reinfection rate, further compounded results, represented a risk factor, or was not an effective method of preventing infection; : The meta-analysis results showed that the topic had a lower infection rate/reinfection rate, more optimal result, represented an effective prevention measure against infection, or was not a risk factor