Unicompartmental Knee Arthroplasty Over the Past 40 Years: a Global Analysis and Review of the Most Highly Cited Articles

Background: The aim of this study was to use citation analysis to identify the top 100 papers in unicompartmental knee arthroplasty (UKA) and report the qualities and characteristics of the most inuential articles in UKA research. Methods: UKA–related articles published from 1950 to June 2020 were identied using multiple databases. The 100 most cited articles were selected for further analysis of citation count, citation density, current citation rate, authorship, theme, geographic distribution, level of evidence, and network analysis. Results: The top 100 articles' publication dates ranged from 1980 to 2015, with the largest number in the 2000s. Total citations per article ranged from 65 to 473. The USA was the most productive country, but England led the race in terms of citations. The Bone & Joint Journal was the most prolic journal with 25 articles. Price AJ was the most productive rst author and corresponding author. The most popular study theme was therapeutic research, followed by prognostic research. The most common level of evidence was level IV. The network analysis connoted that follow-up study had the highest degree of centrality. Conclusions: The study of UKA is steadily evolving. It could be predicted that there will be an increasing number of publications on UKA research in the following years according to the current global trends, the United States and England will likely maintain leadership in this area. Meanwhile, further prospective studies are required to better recognize the prognosis of UKA procedures.


Introduction
After its resurgence in popularity, unicompartmental knee arthroplasty (UKA) is now considered one of the most effective methods to treat end-stage osteoarthritis of the knee which is limited to a single compartment. [1] The concept of UKA was rst clearly proposed by McKeever and MacIntosh who introduced the utilization of a metallic prosthesis for arthroplasty of the knee in the 1950s. [2][3][4] The rst UKA prosthesis in the modern sense was designed by L. Marmor in 1972, known as the "Marmor Knee", which signi cantly enhanced the range of motion of the joint and subsequently became a model for many xed-bearing unicompartmental knee implants to imitate. [5] In its early days, however, UKA was often avoided by surgeons due to the relatively high revision risk, [6] which to some degree is still a major stumbling block today. Besides, some questions such as indications and contraindications for UKA, choice of prosthesis and long term survival were not yet understood. [7] But UKA had still unceasingly fascinated more and more scholars and surgeons since the improved component designs and advanced surgical techniques had promoted excellent clinical outcomes in knee osteoarthritis patients from the late twentieth century onward. [8,9] Having the characteristics of ligament-sparing and smaller incision, UKA is now regarded as an alternative to total knee arthroplasty in appropriately selected patients. [10,11] Bibliometric analysis has been a popular statistical and quantitative tool used by more and more researchers since Gar eld's JAMA publication in 1987. [12] It has become a popular method for analyzing the qualities and scholarly impact of publications within a research eld through identifying and evaluating some quantitative factors like total citations, citation rate and geographical distributions, aiming to allow orthopaedic surgeons to gain insight into the history, current status, and future directions of a speci c research eld. [13] To better present a visual result of the bibliometric study, the visualization of a bibliometric network has been commonly used to map the interrelationships among various nodes and reveal research hotpots. [14] The nodes in bibliometric networks can represent keywords, authors and institutions, while lines among nodes re ect the density of their relationships. [15] CiteSpace, VOS Viewer and Ucinet are widely-used software packages for bibliometric analysis and visualization. Bibliometric methodology has stimulated the trend of analyzing the top cited one hundred classics retrieved from literature database since the number of citations for an article is a valuable measure of the article's impact on the topic it covers. There has also been a trend of bibliometric analysis in orthopaedics, from investigating the general orthopaedic literature [16,17] to analyzing varied orthopaedic elds.
This study aimed to allocate the 100 most cited publications in the eld of UKA, highlighting intellectual milestones and listing analytical focuses in the eld, analyzing the qualities and characteristics of the most frequently cited original papers to allow future surgeons and researchers to comprehend better the current status and global trends in UKA research.

Materials And Methods
Search strategy and criteria Articles were obtained through searching the Web of Science Core Collection, BIOSIS Citation Index, KCI-Korean Journal Database, MEDLINE, Russian Science Citation Index, and SciELO Citation Index for all articles related to UKA. Two independent researchers performed the search at the same time in order to enhance the search sensitivity. The search terms were as follows: unicompartmental knee arthroplast* OR unicondylar knee arthroplast* OR partial knee arthroplast* OR unicompartmental knee replacement* OR unicompartmental knee replacement* OR partial knee replacement* OR unicondylar knee replacement* OR unicondylar knee prosthesis implantation* OR unicompartmental knee prosthesis implantation*.
The search was carried out on June 1, 2020 for publications published since 1950 and a total of 4,040 articles were retrieved from the database. The search results were then ltered by "journal articles" and the remaining results were 3,758. It was prede ned that only original articles were included. Therefore, reviews, meta-analyses, systematic reviews, and guidelines were excluded while registry data were included. Articles were then sorted based on the total number of citations from the highest to the lowest. The articles cited less than 60 times were excluded to ensure the quality of the research. As a result, 304 articles were included for further selection. Two independent investigators (the rst two authors in the author list) read the title and the abstract of the included articles. Articles that met the following criteria were accepted: (i) basic study, animal study, and clinical trials related to any aspect of UKA; (ii) the clinical therapeutic, prognostic, diagnostic, epidemiological studies of UKA; (iii) articles investigating materials or properties related to UKA; (iv) the registry data in relevant institutions. Disagreement on which publications to include between two reviewers were resolved by discussion. There were 113 articles remaining after reviewing the title and abstract. These articles were ranked in descending order of citations and the rst 100 most cited articles were included and pooled for data extraction (Figure 1).

Data extraction
The 100 most cited publications were pooled, and two independent well-trained investigators had reviewed all articles. The following information was extracted from all included articles: the journal name, publication date, rst and corresponding authors, year of publication, geographic origin (based on the address of the rst author), total number of citations of an article, citation density (total citations/article age), current citation rate (citations in the year 2019), research theme, and level of evidence (I through V, methodology has been described elsewhere [18]).

Statistical analysis
The Shapiro-Wilk test was utilized to con rm normality of the distribution of individual variables. Normally distributed data were presented as mean ± standard deviation. One-way analysis of variance (ANOVA) was used to compare means, and post-hoc testing was undertaken when necessary. Timedependent trends were tested using the Mann-Kendall trend test. Correlation between variables was performed using the Spearman rank or Pearson tests. A p<0.05 was considered to be statistically signi cant. The analysis was performed using IBM SPSS Statistics, version 20.0. The Ucinet for windows, version 6.212 was used to perform the analysis for the degree of centrality.

Time and citations
The top 100 cited articles on UKA were published between 1980 and 2015 and listed in (Table 1). All these articles had been published in the English language despite no language restrictions being imposed. About one half (55%) of the articles were published between 2000 and 2009, signi cantly outnumbering the articles published in the 1990s (23%), the 2010s (14%), and the 1980s (8%) (Figure 2). The year that yielded the greatest number of articles was 2005 (n=10). The number of total citations per article ranged from 65 to 473 with a median of 99 for each paper, and an interquartile range of 82-146. Citation density ranged from 2.22 to 29.17 per article. The current citation rate was between zero and 47 for each paper with a mean number of 8.33. The mean number of citations per article was greatest for articles published in the 1980s (c=142.75), followed by the 1990s (c=137.82), 2000s (c=112.95) and 2010s (c=103.57). The Kruskal Wallis H test showed no signi cant association between the decade the article was published and the total number of cites statistically (p=0.171). Mann-Kendall test showed that there was an increasing trend both between the current citation rate and the publication time (z=5.7842, p<0.01) ( Figure 3) and between the citation density and the publication time (z=6.7037, p<0.01) ( Figure   4).

Geographical distribution
The 100 articles came from 12 countries. The United States contributed to the most articles (n=38), followed by England (n=35), France (n=7), Sweden (n=7), and Australia (n=2). England enjoyed the highest overall citations (c=4,484), followed by the USA (c=4,390) and France (c=961). England was responsible for the highest number of the top 20 most cited articles (n=10), followed by the United States (n=5), France (n=3) and Sweden (n=2). A vast majority (97%) of the 100 articles came from Europe (n=58) and North America (n=39). The other three articles scattered in other areas: Australia and New Zealand. A detailed overview is presented as a world map ( Figure 5). The articles from England were mainly published in The Journal of Bone and Joint Surgery-British Volume; while the articles from the USA were mainly published in Clinical Orthopaedics and Related Research. None of the journals were observed to be the favorite for other countries.

Journals
All the articles were published in thirteen journals, among which The Journal of Bone and Joint Surgery-British Volume published the greatest number of articles (n=25). This was followed by Clinical Orthopaedics and Related Research (n=22) and The Journal of Arthroplasty (n=16) ( Table 2)

Authorships
There were sixteen authors who were rst authors of multiple articles, and fteen corresponding authors who possessed more than one article in the list ( Table 3). Majority of these authors were based in the USA (8 authors) and England (7 authors). The most productive rst author was Andrew J. Price from England with ve rst authorships, followed by H. Pandit from England with four rst authorships. The most proli c corresponding author was also Andrew J. Price with six articles, followed by David W. Murray from England with ve articles.

Institutions
Eleven institutions published more than one article in the list, contributed to 50 articles altogether ( Table  4). The United States possessed six of the 11 institutions, followed by England with two institutions. The highest number of articles came from the Nu eld Orthopaedic Centre in England which produced 21% of the 100 articles, followed by Brigham and Women's Hospital in Massachusetts in the United States (5%) and Anderson Orthopaedic Research Institute in the United States (4%).

Types of research
When categorized according to type of research, majority of the articles reported on clinical research (n=92), with the remaining being basic research (n=8). The median number of citations per article showed no signi cant discrepancy between clinical research (median 98.5, range  and basic research (median 96.5, range 75-151) by using Mann-Whitney test; p=0.414. The studies in only six of the 92 clinical articles were designed as randomized controlled trials (RCT).

Levels of evidence
The most common level of evidence for the 100 articles was level IV (n=55), followed by II (n=19), III (n=14), V (n=9) and I (n=3). The eight basic research articles were all level V. The one-way ANOVA revealed no signi cant difference in citations per article among each group of levels of evidence (p=0.351), nor did citation density and current citation rate (p=0.133 and 0.237, respectively) ( Figure 6).

Themes
The major themes of the 100 articles were as follows: therapeutic research (n=31), prognostic research (n=26), surgical indications (n=10), surgical methods and materials (n=10), and epidemiological investigation (n=7) (Figure 7). For the 92 clinical articles, majority of the studies conducted outcome follow-up (58/92). One-way ANOVA was conducted and no signi cant difference in citations per article among the various themes was detected (p=0.296). About two thirds of the clinical articles exclusively focused on UKA (60/92), outnumbering the papers that comparing the clinical outcomes between UKA and other surgical methods like total knee arthroplasty (TKA) (24/92), or high tibial osteotomy (HTO) (4/92); some articles compared UKA with TKA and HTO simultaneously (2/92).

Visualized network
The visualized network was utilized to analyze the article's key words or subject terms based on two periods of publishing time: in the 1980s and 1990s (31 articles) ( Figure 8), in the 2000s and 2010s (69 articles) ( Figure 9). The result indicated that "follow-up study" was the hottest topic in both periods. Additionally, "prosthesis failure, osteoarthritis, survival analysis and radiography" had a relatively high degree of centrality in the 1980s and 1990s; while "osteoarthritis, prosthesis failure, survival analysis, rehabilitation, comparative study and contraindications" had a relatively high degree of centrality in the 2000s and 2010s.

Discussion
To the best of our knowledge, this is the rst bibliometric analysis identifying the 100 most cited articles pertaining to unicompartmental knee arthroplasty. The maximum number of citations merely reached 473 which was a paper on a survival study for UKA patients demonstrating a ten-year survival rate as high as 98% (Table 1). It is noticeable that the range of citations per article in the list is markedly lower than similar papers published on total knee arthroplasty (median 365, range 287-2640) [14] or total hip arthroplasty (median 259, range 191-994).
[19] This can be partially explained by the following reasons: 1) the prevalence of unicompartmental disease is lower with a lower number of overall researchers focused on this topic, 2) the number of UKAs performed are less than other popular arthroplasties; for example, only 10.3% of primary knee replacements done in the United Kingdom in 2019 were UKAs, the remainder were TKAs (89.7%), [20] 3) the lower absolute citation number may be due to "lazy citations", which refers to the fact that authors tend to cite the latest reviews on the matter instead of citing the original paper which rst described the ndings. [21] In the present study, most of the articles were published after 2001 with only ve articles published after 2012 on the list. This might be due to the delay for newly published papers to be thoroughly cited. We also found that the total number of citations was greatest for articles published in the 1980s. However, the citation density and the current citation rate both demonstrated an increasing trend as the articles become more recent (Fig. 3, 4). This means authors tend to cite more current papers than more historical papers, which is appropriate given the rapid evolution of UKA knowledge during the past few decades, leading people to agree more with the recently published articles and less with the older papers.
Our ndings also revealed that 38% of the 100 articles originated in the United States which was more than England. But England leads the race in terms of quality, e.g. the number of citations and highly-cited articles. The most likely reason for the United States to be the most productive country can be its larger population size and higher gross domestic product (GDP) compared to European countries. Moreover, McKeever and L. Marmor, two pioneers in this eld, were based in the USA. In this study, the USA has six of the eleven institutions from which more than one classic article originated, re ecting the USA's great academic impact in this eld. But the higher citation rating of England could be due to the dedicated National Joint Registry and the renowned researchers in certain institutions that concentrated on UKA and produced more papers, of whom Price AJ and Pandit H from the Nu eld Orthopaedic Centre and Murray DW from the University of Oxford were the most outstanding. The Nu eld Orthopaedic Centre in England, a famous hospital with strong a liations to the University of Oxford, was the most prominent institution in UKA quality research output surpassing all other institutions in both the absolute number of classic articles as well as the number of highest-cited articles (Table 4).
In our study, the highest percentage of these articles were published in The Journal of Bone & Joint Surgery British Volume, which was renamed The Bone & Joint Journal in 2013. This journal has been widely recognized as a agship orthopaedic journal and was established in 1948. The Bone & Joint Journal has been the most popular for UKA research from England, but in the United States, we found it much less popular than Clinical Orthopaedics and Related Research which released its rst issue in 1953 ( Table 2). It was revealed that the most-cited UKA articles from the United States were published almost exclusively in American journals and articles from England were almost exclusively published in British journals, which was similar to previous studies. [16,22] Including the two journals mentioned above and along with The Journal of Arthroplasty and The Journal of bone and joint surgery American volume, these four journals together published 76% of the 100 most cited articles in the list and may keep on serving as the "main channel" for future ndings in this eld.
The level of evidence analysis revealed that majority of the 100 articles represented Level-IV and consisted of fewer articles with a high level of evidence, which contradicted the expectation that studies with a high level of evidence would be cited more frequently. Our ndings were similar to the situation reported in previous bibliometric studies. [23,24] Even the number one most cited paper only included a small number of patients and had level IV evidence. Knowing that level of evidence is used to evaluate study designs according to their protection against bias, this study echoes the concept that smaller case series or cohort study with more bias could still attract the attention of academics. For example, the article ranking 29th in the list reported an uncontrolled forty-one-patient case series and we regarded it as level IV considering its selection bias and information bias.
This study also examined the theme distribution of the 100 most cited articles. Therapeutic research and prognostic research were the top two themes (Fig. 7). The former had been frequently performed by doctors to compare UKA with TKA since the controversies between these two surgical procedures were hotly debated. Most of the therapeutic studies focused on the revision rate and the recovery of UKA procedures. Prognostic researches were also commonly conducted to determine the probability of the speci ed outcomes in UKA patients.
[25] These researchers often utilized follow-up methodology lasting at least several years. The degree of centrality analysis of the author's key words also indicated that "follow-up" has a high degree of centrality in the past four decades. But we found that retrospective studies dominated these prognostic research studies and many of them were uncontrolled case series.
The retrospective study design does truly have limitations in reducing bias when measuring predictors and outcomes while a prospective study or cohort study is reportedly a better design to resolve prognostic issues, although requires a longer follow-up duration and therefore greater expense.
[26] To better guide doctors in making decisions and estimating the future outcomes of patients who have received UKA, further well-controlled prospective studies are expected to be conducted.
The degree of centrality analysis of the author's key words was performed to predict the hotspots and future trends in the research elds of interest. As shown in the bibliometric networks, such keywords as follow-up, osteoarthritis, prosthesis failure and survival analysis were highlighted with bigger icons (Fig. 8, 9). Among varied types of studies, follow-up outcome studies are predicted to continue their high popularity in UKA research, given they are more reliable with less inherent bias. Some keywords like "minimally invasive", "robot assisted" and "mobile bearing" were burgeoning with a much higher degree of centrality after 2000 than before. Considering that UKA is now a minimally invasive procedure with potential e cacy and safety bene ts and given there is an aging population in most developed countries, it is expected that the number of UKA surgeries will continue to increasing, as will the related research. Thus, it still requires high-quality studies within the context of these aforementioned orientations.
The article with the highest impact factor among the 100 classics reported a study by Liddle AD et al. in 2014 and its total citation count ranked 15th in the list. This paper retrospectively compared the rates of adverse outcomes after TKR and UKR in more than 100,000 knees. The result showed a signi cantly higher risk of revision but lower risk of medical complications and shorter inpatient stays in patients undergoing UKR than for TKR. This research was of great importance because it was the rst to con rm the difference in mortality between UKA and TKA in comprehensively matched patients. When ranked by the citation counts in 2019, this paper ranked 1st of these 100 articles, indicating its currently highly valued. According to its citation trend, we expected that this paper would continuously be highly cited and probably become an outstanding landmark in the future.
Of all the articles selected, the oldest paper was published in 1980 and its citation number was ranked 7th. It reported the results of a follow-up study lasting an average of six years on UKA patients, showing an unfortunate result with the failure rate as high as 30%. Thus, the conclusion was that UKAs were more frequently revised than TKAs. This result re ected the historical background that UKA had not fully experienced the development of surgical techniques or component design in the 1970s. Thus the early designs of UKA failed early and often.
[27] The survival study design of this paper was repeated by many following articles in this eld.
This bibliometric analysis has provided some information of value, although there were limitations inevitably. In this study, only published articles were included but meta-analysis, meeting records, reviews and textbooks were excluded, which may cause omission bias. Moreover, as mentioned earlier, our study included only ve articles published since 2012, suggesting that our study might not be very sensitive to recent changes in UKA research trends. Like other citation analysis articles, it should be admitted that we can't have included all classics without allowing for the shortcomings of citation analysis. We tried to avoid missing classics by using extended search terms without quotation marks, utilizing a mix of databases and abandoning the search restrictions on publication time and language. Despite limitations, citation counts are still widely considered as the most currently-available standardized way to measure the impact of papers objectively in the scienti c eld. [28] Conclusion This article identi ed the top 100 most cited articles in UKA research and delineated their time and geographical distribution, research topic, authorship as well as their levels of evidence. The study of UKA has a trend of balanced development and is steadily evolving. It could be predicted that there will be an increasing number of publications on UKA research in the following years, with the United States and England maintaining leadership in this area. Meanwhile, well-controlled prospective studies should be further conducted to recognize the prognosis of UKA better. Future researchers could utilize this study to identify milestone articles and acknowledge the outstanding members who have altered the course of UKA over the decades.
Abbreviations ANOVA analysis of variance Declarations Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.

Availability of data and materials
All data generated or analysed during this study are included in this published article.

Competing Interests
The authors have no con icts of interest to declare that are relevant to the content of this article.

Authors' Funding
Not applicable.

Authors' Contributions
All authors contributed to the study conception and design. Tianlong Huang conceived and designed the work. Material preparation, data collection and analysis were performed by Buchan Jiang, Cong Wei, Wenchao Zhang, Daniel M George, Ning Tang and Tianlong Huang. The rst draft of the manuscript was written by Buchan Jiang and all authors commented on previous versions of the manuscript. All authors read and approved the nal manuscript.