A historical review and bibliometric analysis of research on fracture nonunion in the last three decades

The relationship among document types and their citations per publication, CPP_year, and the APP has recently been proposed [11]. Among the 14 document types indexed by the Web of Science, a total of 10,064 nonunion-related publications were found (Table 1). The document type of ‘articles’ was by far the most prevalent, with a total of 8976 articles (89% of 10,064 documents) and an average of 4.7 APP with a maximum of 200. The document type of the ‘proceedings’ papers had the highest CPP₂₀₁₉ of 41, which can be attributed to the 33 highly cited proceedings papers (11% of 306 proceedings papers) with TC₂₀₁₉ of 100 or more [6]. Of note, in the Web of Science document classification, one document may be classified in several categories. For example, 306 documents are classified as document types for ‘proceedings papers’ and ‘articles’, so the sum of the percentages is higher than 100% [6]. Only the 8976 articles were further analysed. In total, 12 languages were the primary languages in the articles, with English, as the most prevalent (93% of the total of 8976 articles), followed by German (324 articles; 3.6% of 8976 articles) and French (178; 2.0%). Minority languages were Czech (36 articles), Turkish (25), Portuguese (6), Spanish (5), Serbian (4), Italian (3), Russian (3), Polish (2) and Hungarian (1). English language articles had higher CPP₂₀₁₉ of 23 than non-English with CPP₂₀₁₉ of 7.5.

To understand publications in a specific research topic and their citation trends, a relationship between citations per publication (CPP_year) and article life was proposed [12]. The article life with CPP for all the 8976 nonunion articles is displayed in Fig. 1. It takes CPPs about three full years to reach a plateau and up to a peak in ten full years. A comparison of the medical-related topics was also shown in Fig. 1. Articles related to nonunion had longer citation life than some medical topics; a peak appeared in four and five full year for breast reconstruction [6] and child sexual abuse [13] research, respectively.

Both publications and citation trends in a research topic were proposed [6]. A relationship between the annual number of articles (TP) and their citations per publication (CPP₂₀₁₉ = TC₂₀₁₉/TP) throughout the years is shown in Fig. 2. The ‘jump’ between 1990 and 1991 is an artefact that arises from a policy change by the Web of Science with respect to abstract information [14]. The majority of articles have no abstract information in SCI-EXPANDED before 1991. There is a slightly fluctuating steady increase of articles from 116 in 1991 to 201 in 2003 and thereafter, the number of articles sharply increased to reach a plateau in 2015, fluctuating in the mid-550s. Based on Fig. 2, it takes CPPs about 14 years to reach a plateau. Articles related to nonunion had longer life than some medical-related topics, for example, breast reconstruction with 10 years[6] and child sexual abuse with 8 years [13]. Articles published in nonunion could be cited longer with average year of 14 than that of breast reconstruction and child sexual abuse with ten and eight years, respectively.

Journal Citation Reports (JCR) indexed 9381 journals across 178 Web of Science categories in SCI-EXPANDED in 2019. A total of 8976 nonunion-related articles in SCI-EXPANDED were published in 790 journals which are classified among the 115 Web of Science categories in SCI-EXPANDED. Table 2 shows the top ten productive Web of Science categories with number of articles (TP), APP, CPP₂₀₁₉ and number of journals in a category. A total of 7407 articles (83% of 8975 articles with category information in SCI-EXPANDED) were published in the top two categories: orthopaedics (6223 articles; 69% of 8975 articles) and surgery (4334 articles; 48%). Comparing the top ten productive categories in healthcare sciences, articles published in category of clinical neurology had the highest CPP₂₀₁₉ of 34, while articles in general and internal medicine had lower CPP₂₀₁₉ of 6.1. Articles published in category of biomedical engineering had the highest APP of 6.5, while articles in paediatrics had lower APP of 4.2. The evolution of nonunion article trends in the top five productive Web of Science categories is shown in Fig. 3. Articles in categories of orthopaedics and surgery had similar trends from 1990 to 2008; however, orthopaedics shows a sharper increasing trend since 2009. The number of articles published, as well as trends in the remaining three categories, is similar.

The top ten most productive journals are listed in Table 3, including IF₂₀₁₉, APP, CPP₂₀₁₉ and Web of Science category. All the top ten journals were classified in the category of orthopaedics and six in surgery. The Injury-International Journal of the Care of the Injured published the most articles (657 articles; 7.3% of 8976 articles). Comparison of the articles published in the top ten journals showed that articles published in the Spine had the highest APP of 5.2, while the Journal of Hand Surgery-American Volume and the Clinical Orthopaedics and Related Research had an APP of 3.8, respectively. Articles published in the Journal of Bone and Joint Surgery-American Volume had the highest CPP₂₀₁₉ of 71, while articles published in International Orthopaedics had a CPP₂₀₁₉ of 14. According to IF₂₀₁₉, the top three articles have an IF₂₀₁₉ of more than 30 (Lancet, Nature Biotechnology, and BMJ-British Medical Journal).

The 8976 articles were published by 26,079 authors among 101 different countries. Altogether, 7976 (89% of 8976 articles) were authored from a single country, with the articles being from 79 different countries. Internationally, authored collaborative articles were less common (993 articles, 11%) with authors from 85 different countries. The six publication indicators [10] were applied to compare the top ten productive countries (Table 4). Five European countries, three Asia countries and two American countries were ranked in the top ten of publications. The USA dominated the top in the six publication indicators with a TP of 3262 articles (36% of 8969 articles), an IP of 2726 articles (34% of 7976 single-country authored articles), a CP of 536 articles (54% of 993 internationally collaborative articles), an FP of 2996 articles (33% of 8969 first-author articles), an RP of 2939 articles (33% of 8832 corresponding-author articles) and an SP of 148 articles (39% of 378 single-author articles).

In total, 4192 articles (47% of 8969 articles) were articles authored within single institution and 4777 (53%) were inter-institutionally collaborative articles. The six publication indicators [10] were applied to compare the top 11 productive institutions with TP of 70 or more (Table 5). Nine of the top 11 most productive institutions are in the USA and one in Canada and China, respectively. The Mayo Clinic in the USA took the leading position for the four publication indicators with a TP of 134 articles (1.5% of 8969 articles), an IP of 67 articles (1.6% of 4192 single institute articles), an FP of 92 articles (1.0% of 8969 first-author articles) and a RP of 93 articles (1.1% of 8,831 corresponding-author articles), while the University of Toronto in Canada ranked top with a CP of 82 articles (1.7% of 4777 inter-institutionally collaborative articles). The Chang Gung Memorial Hospital in Taiwan published the most single-author article with an SP of nine articles (2.4% of 378 single-author articles). The Hospital for Special Surgery in the USA, the New York University (NYU) in the USA, the Washington University in the USA and the Shanghai Jiao Tong University in China had no single-author articles.

In experimental science, the accepted convention is that the most important positions are first and last authors and one of these usually includes the corresponding author [6]. The first author generally is the person who contributed most to the work, including conducting research and writing of the manuscript [6]. The corresponding author is perceived as the author contributing significantly to the article independently of the author position [15]. The corresponding author supervised the planning and execution of the study and the writing of the paper [16]. Table 6 listed the top ten most productive authors. P.V. Giannoudis at the University of Leeds in the UK published the most nonunion articles followed by J.B. Jupiter at the Massachusetts General Hospital in the USA and C.C. Wu at the Chang Gung University in Taiwan. Publication performance of authors was further analysed with the Y-index which is related to the number of first-author publications (FP) and corresponding-author publications (RP). The Y-index combines two parameters (j, h), to evaluate the publication potential and contribution characteristics as a single index. This indicator has been used to compare authors in medical topics, such as occupational therapy [17], child sexual abuse [13] and breast reconstruction [6]. The Y-index is defined as [7]: where j is the publication potential, a constant related to publication quantity, and h is publication characteristics which can describe the proportion of RP to FP. The greater the value of j, the greater the number of first- and corresponding-author articles by the author. Different values of h represent different proportions of corresponding-author articles from first-author articles.

Y-index was applied to evaluate the authors in 8551 (95%) of the 8976 nonunion articles with both of first-author and corresponding-author information in SCI-EXPANDED.

A total of 8551 articles were published by 25,219 authors in which 16,991 authors (67% of 25,219 authors) had no first- or corresponding-author articles with Y-index = (0, 0); 1596 (6.3%) authors published only corresponding-author articles with h = π/2; 446 (1.8%) authors published more corresponding-author articles with π/2 > h > 0.7854; 3613 (14%) authors published the same number of first- and corresponding-author articles with h = 0.7854; 242 (1.0%) authors published more first-author articles with 0.7854 > h > 0; and 2331 (9.2%) authors published only first-author articles with h = 0. Figure 4 shows distribution of the Y-index (j, h) of the top 15 authors with j ≥ 19. Each dot represents one value of the Y-index that could be one author or many authors, for example, P. Hernigou and S.D. Boden who share the same Y-index = (21, 0.9273). C.C. Wu from the Chang Gung University in Taiwan had the highest publication potential with a j of 70 and Y-index = (70, 0.7854). Wu published 42 nonunion articles including 36 first-author articles and 35 corresponding-author articles (Table 6). Wu published not only the most first-author articles and the corresponding-author articles but also the most single-single articles. However, Wu’s nonunion articles had lower CPP₂₀₁₉, followed by D. Ring from the Massachusetts General Hospital of Harvard University in the USA as the corresponding affiliation with a j of 50 and Y-index = (50, 0.9828) and P.V. Giannoudis from the University of Leeds in the UK with a j of 46 and Y-index = (46, 1.266). Comparing the top three authors had higher publication potential in nonunion articles: P.V. Giannoudis had higher CPP₂₀₁₉ of his first-author articles and corresponding-author articles than Wu and Ring, while Ring had higher CPP₂₀₁₉ for his single-author articles (Table 6). M.D. McKee with Y-index = (25, 1.131) and S. Rammelt with Y-index = (25, 0.9048) both had a j of 25. McKee and Rammelt are located on the same curve (j = 25) in Fig. 4, indicating that they had the same publication potential with the same value of j but different publication characteristics [18]. Similarly, T. Niikura (19, 1.138), M. Bhandari (19, 1.043), A. Moghaddam (19, 0.9420) and C.M. Robinson (19, 0.8380) are also located on the same curve (j = 19). Niikura published higher ratio of the number of corresponding-author articles to the number of first-author articles with an h of 1.138, followed by Bhandari with an h of 1.043, Moghaddam with an h of 0.9420 and Robinson with an h of 0.8380. These authors had the same publication potential with a j of 19 but the publication characteristics were also different. D. Ring with Y-index = (50, 0.9828) and M.J. Parker with Y-index = (20, 0.9828) are located on the line (h = 0.9828), indicating that they had the same publication characteristics but different publication potentials [19]. Ring had the greater publication potential with a j of 50 than Parker with a j of 20. It has been pointed out that a bias in analysis of authorship might occur when different authors had the same name, or one author used different names (e.g. maiden names) in their articles [20].

The total number of citations of a document in the Web of Science Core Collection is updated from time to time. Ho’s group proposed citation indicators TC_year [21] and C_year [7]. The advantage of using TC_year and C_year is that they are immutable and ensure repeatability compared with the citation index of the Web of Science Core Collection [22]. Citation frequency is considered to reflect the impact of scientific publications, although this does not necessarily correlate with the paper quality [23]. The best articles can be classified as articles that most researchers can read and cite in peer-reviewed journals [24]. The 8976 nonunion articles ranked differently if sorted by TC₂₀₁₉ than sorted by C₂₀₁₉. A total of 3521 articles (39% of 8976 articles) had no citation in the most recent year (C₂₀₁₉ = 0) and 987 (11%) articles had no citation from their publication year to the end of 2019 (TC₂₀₁₉ = 0). Furthermore, among the top 100 C₂₀₁₉ articles, 34% of the articles were among the top 100 TC₂₀₁₉ articles.

The 8976 nonunion articles were searched out with search keywords on their ‘front page’: title, abstract and author keywords from SCI-EXPANDED in the last three decades. Table 7 lists the ten most frequently cited articles in nonunion research with two citation indicators [7]. Five of the ten articles were published in the Journal of Bone and Joint Surgery-American Volume (IF₂₀₁₉ = 4.578). Nine of the ten articles were single-country articles from the USA with five articles, France with two and one by Canada and Germany, respectively. The only internationally collaborative article was published by 59 authors from South Africa, the USA, Israel, Finland, the UK, Germany, France, Australia, Canada, Belgium, Netherlands and Norway. The citation history of the top ten most frequently cited articles is presented in Fig. 5. The article by Chen et al. [25] had a sharply increased citation after its publication for ten years in nonunion studies and then decreased. It ranked top on annual citation from 2007 to 2019. An article’s impact might not be always high [22]. A highly cited article published by Friedlaender et al. (2001) had a TC₂₀₁₉ of 723 ranked fourth but had a low impact in 2019 with a C₂₀₁₉ of ten (ranked 217th) indicating that with time, the significance of the article diminished. Although some recently published articles in the past few years have great potential, their TC₂₀₁₉ is not high. The top three cited articles in 2019 by Campana et al. (2014), Xavier et al. (2015) and Goodman et al. (2013) had a C₂₀₁₉ of 117 (ranked 1st), 77 (ranked 2nd) and 73 (ranked 3rd) but had a low TC₂₀₁₉ of 262 (ranked 24th), 220 (ranked 41st) and 387 (ranked 14th).

Five articles were ranked on both the top 10 TC₂₀₁₉ as the most frequently cited articles and C₂₀₁₉ as the most impact in the most recent year articles. These five highly cited and the most impact articles in 2019 articles were summarised as follows:

C₂₀₁₉ of 69 (ranked 4th) and TC₂₀₁₉ of 1389 (ranked 1st)

In this article, Chen et al. [25] was one of the first to provide a state-of-the-art review of the advancements on basic science and clinical applications of bone morphogenetic protein (BMP) research. BMPs are extremely important in embryonic development as well as in postnatal cell functions and are involved in the heart, nervous system and cartilage development, as well as postnatal bone formation with strong bone induction properties. Therefore, at that early stage, numerous potential therapeutic applications of BMPs were outlined including bone defects and fracture nonunion. In the years following this publication, BMPs were widely studied and successfully applied in the treatment of nonunions with BMP-2 and BMP-7 currently approved for use by the US Food and Drug Administration [35] and are considered critical components in the therapeutic armamentarium for nonunions [36].

C₂₀₁₉ of 56 (ranked 5th) and TC₂₀₁₉ of 1056 (ranked 2nd)

In this pioneer basic science paper, the authors manufactured a combined construct of coral scaffold and marrow stromal cells (MSCs) to fill bony defects in sheep. Contrary to the failure of scaffolds used in the past, the idea was to use the naturally occurring coral exoskeleton which has adequate mechanical properties and increased porosity similar to cancellous bone, in order to better deliver the MSCs. The authors established ‘critical’-size bone defects of increasing width in sheep metatarsals and treated the defects with either coral exoskeleton alone, or in combination with fresh bone marrow or MSCs. They succeeded in bridging the defects after 16 weeks in the third group of combined coral exoskeleton and MSCs and this was the first study to effectively demonstrate replacement of a critical defect with mature bone that would have evolved into a nonunion otherwise.

C₂₀₁₉ of 49 (ranked 6th) and TC₂₀₁₉ of 534 (ranked 6th)

This study was the first one to provide guidelines and revive interest on the harvesting of bone marrow aspirate concentrate (BMAC) from the iliac crest and its subsequent injection in atrophic tibia nonunion sites [30]. Following bone marrow aspiration from bilateral iliac crests and its concentration via a cell separator, 20 cc of BMAC were injected into the nonunion site of 60 patients. Fifty-three patients went into union, whereas seven developed nonunion. Although the study confirmed the efficacy of the procedure from previous reports [37], it was the first to show that the number of progenitor cells as well as the need for increased concentration is critical in order to obtain union. From that point onward, BMAC in the treatment of nonunions has been widely used and the area is still under constant evolution [38].

C₂₀₁₉ of 41 (ranked 9th) and TC₂₀₁₉ of 442 (ranked 8th)

Prior to this study, even the most displaced midshaft clavicle fractures, were traditionally receiving non-operative treatment based on earlier reports [40] that had inherent biases. However, some subsequent studies showed increased level of complications such as nonunion and malunion as well as lower rates of patient satisfaction [41]. This prospective randomised clinical trial coming from eight Canadian centres and published by the Canadian Orthopaedic Association looked at 132 patients divided into two groups of similar characteristics and who received either plate fixation or non-operative treatment with a sling [39]. The authors looked at time to radiographic union, constant shoulder scores, DASH scores, complications and patient satisfaction at one year. Plate fixation showed better outcomes as far as function (scores) and malunion/nonunion complications, as well as patient satisfaction. In the plate fixation group, three out of 67 patients developed a wound infection but were successfully addressed with local wound care and antibiotics, and later plate removal with formal irrigation/debridement after union. The authors concluded that primary plate fixation of those injuries is justified. Although this study significantly helped revive the debate of operative versus non-operative management of displaced midshaft clavicle fractures, it is still unclear what the best option is, and patients must be adequately informed of the pros and cons of each treatment, as outlined in a recent systematic review also from Canada [42].

C₂₀₁₉ of 39 (ranked 10th) and TC₂₀₁₉ of 937 (ranked 3rd)

This multi-author randomised control trial by the BMP-2 Evaluation in Surgery for Tibial Trauma (BESTT) study group evaluated the safety and efficacy of the use of recombinant BMP-2 (rhBMP-2) in the treatment of open tibial shaft fractures compared to the standard of care [27]. Τhe authors looked at 121 patients with a 12-month follow-up receiving either intramedullary nail alone with routine wound care or intramedullary nailing with addition of an implant consisting of an absorbable collagen sponge containing rhBMP-2 of 6 or 12 mg over the fracture site. The study found that the patients who received the 12-mg dose rhBMP-2 implant had lower rates of hardware failure and infections and demonstrated faster wound healing and required fewer number of secondary interventions than the other two groups. Since then, there have been several clinical studies showing that rhBMP-2 (as well as rhBMP-7) enhances bone defect regeneration in adults and as stated earlier they have been approved by the US Food and Drug Administration for specific clinical indications [35].

Ho’s group proposed distribution of words in article titles, abstracts, author keywords and KeyWords Plus in different periods as information to evaluate main research focuses and find their development trends in research topics [43]. The results of keyword analyses provide information about the main and possible research foci as each word cluster comprised several supporting words. Thus, the development of the seven possible main research foci in nonunion-related research is presented. Figure 6 shows the development of the seven topics.

These seven topics identified support the efforts of clinicians over the years to understand better the characteristics, aetiology, diagnosis, prediction and treatment options that can be applied successfully to treat this difficult complication. Moreover, the burden of this condition, on the physiological and mental state of patients, has received a lot of attention. Finally, the cost-related implications of treatment have also been explored. All of the above themes represent the spectrum of research activities that have taken place up to now. It is envisaged that in the future, the same themes of research on nonunion will continue to dominate the activities of the scientific and clinical community.