Journal of Obesity & Weight Loss Therapy
Open Access

Laparoscopic adjustable gastric banding (LAGB) is one of the most popular operations performed worldwide for morbid obesity [1]. There is an ever expanding body of evidence about LAGB which includes indications, effect on weight loss and improvement of associated comorbidities, common complications and long-term outcomes. One of the methods of identifying the published research work that has had the greatest impact on our understanding of the various aspects of LAGB is by generating and analysing a citation rank list. A citation is generated when a published article formally references another article and it is now widely accepted by the academic medical community that research impact is strongly correlated with the number of citations [2]. Bibliometrics is a branch of information research which deals with the study and analysis of meta-data surrounding published material [3]. Citation analysis is a branch of bibliometrics that evaluates the impact of an article or journal based on the total number of citations received by that article. Detailed review and statistical analysis of the publications in the citation rank list provides valuable information and insight into the types of research studies that influence and generate interest in the academic world. To date, there has been no study undertaken to identify and analyse the most influential publications in the context of LABG.

Citation count is now a well-established method within the field of Bibliometrics to measure the impact of a published research study but it does have some recognised weaknesses. For example, papers with very high original impact can become a victim of their own success as far as citations are concerned due to the phenomenon of obliteration by incorporation [4]. These papers tend to be cited with increasingly low frequency with time as the original findings become universally accepted within the academic community and therefore, no longer routinely referenced in published articles. Furthermore, the process of accrual of citations is a slow process and takes several years. In an effort to circumvent this issue, other methods of assessing research impact more rapidly have been developed recently and are known as alternative metrics or “Altmetrics”. These are based on utilisation of alternative information sources that are also considered as independent or surrogate markers of impact such as frequency of online article downloads, discussion in social media platforms, use in consensus or guideline development and mention in patent applications amongst others. One of the most popular platforms that is currently favoured by many journals to assess these alternative metrics is the Altmetric Attention Score (AAS) developed by Altmetric LLP. The AAS is a relatively new concept initiated in 2011 and is derived from algorithms which assess the distribution and influence of a scientific paper through social media and until recently was known simply as the ‘Altmetric Score’ [5]. There is an urgent need to conduct research that helps to assess the utility and value of Altmetric indices over more traditional indices of academic impact such as citation number (bibliometric index), journal impact factor and level of evidence of the publication.

The primary aim of this study is to identify and perform a citation rank analysis of the 100 most cited publications that have influenced the understanding of LAGB. The secondary aim is to analyse the correlation between Total Citation Count (TCC), Level Of Evidence (LOE) and AAS of the listed publications to gain a further understanding of the dynamics between these metrics and assess the reliability of AAS as an index of academic impact.

A search of the Thomson Reuters Web of Science citation indexing database and research platform was performed using the search terms ((laparo* and adjust* and gastr* and band*) OR (LAGB or L.A.G.B.)) AND ALL FIELDS: ((endoscop* or remov* or complicat*)). The returned dataset was filtered to include only English language and full manuscripts and sorted using the TCC in descending order as per the method originally developed by Paladugu and colleagues [6]. Each manuscript was analysed to ensure that content was relevant to the study topic and the four articles were excluded as they were deemed not to have substantial content related to LAGB [7-10]. The 100 most cited manuscripts were subsequently analysed in detail and used to populate a database which included several data indices such as the publishing journal (name of journal, country of publication, 2018 impact factor and year of publication), authors (institutional affiliation and country) and article details (TCC, type of study, LOE and AAS). The quality of evidence contained within the articles was assessed and recorded according to the Oxford Evidence Based Medicine scoring system [11]. The ‘‘Altmetric it’’ application downloaded from the Altmetric.com website was used to generate Altmetric scores by utilising the journal article page containing the doi reference number [5].

The Web of Science search returned 1,473 full-length, English language papers. Table 1 lists the 100 most cited of these papers: [12-109].

Table 1: The top 100 most cited papers in LAGB1.

The total number of citations ranged from 78 for Skull et al. (Laparoscopic adjustable banding in pregnancy: Safety, patient tolerance and effect on obesity-related pregnancy outcomes) to 962 for Maggard et al. (Meta-analysis: Surgical Treatment of Obesity) (Table 1). The median citation count was 122 (interquartile range (IQR): 96.75 – 215.25) and this was not normally distributed (Skewness=2.86, Kurtosis=8.29). The oldest article within this top 100 list was a report on LAGB published in 1994 by Belachew et al. titled ‘Laparoscopic adjustable silicone gastric banding in the treatment of morbid-obesity – A preliminary report’. The most recent paper titled ‘Bariatric surgery for obesity and metabolic disorders: state of the art’ was published in 2017 by Nguyen et al.

Table 2: Journals with the top 100 most cited LAGB1 articles.

The 100 most significant papers spanning between 1994 and 2017 were published across a spectrum of 28 journals with the number of published articles per journal ranging from 1 to 36 (Table 2). The journal “Obesity Surgery” published the highest number of papers within the top 100 and also, had the highest cumulative TCC amongst all journals (n=36, 5633 citations). Amongst this list of the top 100 articles, 11 publications were in 2002 making it the most frequent year of publication. Most articles were published in journals which had an impact factor (IF) greater than 3 with only 11 articles published in journals with a lesser IF.

The country with the most publications when first author affiliation was considered was the United States of America (USA) with 41 publications and, incidentally, USA also had the highest cumulative TCC at 8,174 (43% of total citations) (Figure 1). This was followed by Italy with 15 publications. Analysis of senior author affiliation revealed that the country with the most publications and the highest cumulative TCC was also USA (n=42, 5,954 citations, 31% of total citations) followed by Italy (n=17; 5404 citations). The country with the highest AAS by first author affiliation (1,350, 77% of total AAS) as well as second author affiliation (1,350; 76% of total AAS scores) was also USA. Monash University and its affiliated institutions had the highest number of articles in the top 100 with 6 publications.

Figure 1a: Proportions of Citations by Country of First Author

Figure 1b: Proportion of AAS by Country of First Author.

There were two first authors with the highest number of publications in the top 100 with 4 articles each - Paul E. O’Brien who is affiliated with Monash University, and L. Angrisani who is affiliated with S. Giovanni Bosco Hospital and the Fondazione Institute for Spreading and Valorisation of Scientific Culture, they both had 4 publications each. L. Angrisani also had the most citations overall as first author with 1,281 citations. The senior author with the highest number of publications in the top 100 with 4 articles was G. Enzi who is affiliated with the University of Padova and the University of Padua. The senior author with the highest number of citations was Paul G. Shekelle who is affiliated with the RAND Health Division, California with 1,242 citations.

The citation rate index (CRI) was also calculated in order to control for the fact that older papers would have had more time to accrue citations. The CRI for the top 10 articles ranged from 187.5 for Angrisani et al. (Bariatric Surgery Worldwide, 2013) to 35.908 for Colquitt et al. (Surgery for Obesity, 2009) (Table 3). The highest 5 CRIs came from articles published between 2009 and 2017. The countries that published the highest number of articles in the top 10 CRI were UK and USA with 3 articles each.

Table 3: Top 10 articles with the highest CRI1.

The articles with the five highest AAS were published between 2009 and 2017 with the highest score being 413 (Perioperative Safety in the Longitudinal Assessment of Bariatric Surgery, Flum et al) (Table 4). Altmetric scores ranged between 0 and 413 (median = 0, IQR: 0 – 10.25) with 54 articles scoring 0 for AAS. The article with the highest AAS was (Perioperative Safety in the Longitudinal Assessment of Bariatric Surgery). The USA had the most articles in the top 10 AAS with 5 publications. 93 papers dealt with complications of LAGB making this the most extensively studied topic in this review followed by weight loss which was investigated in 93 papers (Table 5). Long term prognosis or outcomes following LAGB was the least discussed topic with only 4 articles dealing with this topic.

Table 4: Top 10 articles with the highest AAS1.

Table 5: Most frequently referenced topics.

Evidence levels for the top 100 articles were scored using the Oxford Evidence Based scoring system [11]. 14 papers were classed as providing level 1 evidence, 4 papers were level 2 evidences and level 3 evidences each, 62 papers were level 4 evidence and 16 papers were level 5 evidence. There was a statistically significant difference between the LOE and TCC (p=0.0029) (Figure 2a). The median number of citations received at each evidence level was level 1 165 (IQR 115.25-606.25), level 2 460 (IQR 292.25-434), level 3 190.5 (IQR 100.75-287.5), level 4 112 (IQR 89.25-177.25) and level 5 126 (IQR 103.25-227.5). There was a statistically significant difference between LOE and CRI (p=0.00002, Figure 2b). The median CRI at each evidence level was level 1 18.75 (IQR 13.75-49.75), level 2 3.05 (IQR 26.51-38.54), level 3 16.30 (IQR 6.92-41.02), level 4 7.18 (IQR 5.88-12.37) and level 5 13.33 (IQR 8.90- 21.51). There was also a statistically significant difference between LOE and the AAS (p=0.00173, Figure 2c). The median AAS received at each evidence level was level 1 26 (IQR 6.5-43), level 2 8.5 (IQR 0.75-20), level 3 1.5 (IQR 0-14.5), level 4 0 (IQR 0-3.75) and level 5 0 (IQR 0-1).

Figure 2: Relationship between Level of Evidence (LOE), Total Citation Count (TCC), Citation Rate Index (CRI) and Altmetric Attention Score (AAS). LOE has a statistically significant association with TCC (p=0.0029, Figure 2a), CRI (p=0.00002, Figure 2b) and the AAS (p=0.00173, Figure 2c) Kruskal–Wallis Test.

Figure 3: Distribution of Altmetric Attention Scores (AAS) in articles published pre-2006 and post-2006 (p=0.00001) - Kruskal-Wallis test.

Articles published from the year 2006 onwards had a significantly higher AAS compared to articles published prior to 2006 with median of 5.5 (IQR 0-23.25) and 0 (IQR 0-0) respectively (p = 0.00001) (Figure 3). The AAS of articles which were published both prior to 2006 did not have a significant correlation with the TCC (r = 0.2332, p = 0.1108) (Figure 4a) but was positively correlated with the CRI (r =0.3194, p =0.0268) (Figure 5a). The AAS of articles which were published following 2006 did not have a significant correlation with either the TCC (r = 0.5274, p = 5.8596, Figure 4b) or the CRI (r = 0.5274, p = 5.8596) (Figure 5b).

Figure 4a: Relationship between Altmetric Attenuatin Score (AAS) and Total Citation Count (TCC) Pre-2006 Spearman Rank Correlation test r=0.2332 p=0.1108.

Figure 4b:Relationship between Altmetric Attenuation Score (AAS) and Total Citation Count (TCC) Post-2006, Spearman Rank Correlation test r=0.5274 p=5.8596.

Figure 5a:Relationship between Altmetric Attenuation Scores (AAS) and Citation Rate Index (CRI) Pre-2006 Spearman Rank Correlation test r=0.3194 p=0.0268.

Figure 5b:Relationship between Altmetric Attenuation Score (AAS) and Citation Rate Index (CRI) Post-2006 Spearman Rank Correlation test r=0.5274 p=5.8596.

Bariatric Surgery is now well established as an effective treatment for morbid obesity and associated co-morbidities [110]. However, there is still ongoing debate regarding the exact indications for different bariatric operations and their long term outcomes. LAGB is one of the most widely used procedures worldwide for the treatment of morbid obesity [1]. There has been widespread interest in recent years regarding LAGB and, in particular, the indications, weight loss results, effect on co-morbidities, complications and long-term results following LAGB. It is therefore, not surprising, that the four most widely studied topics amongst the top 100 articles featured in this review include complications, weight loss, improvement of co-morbidities and indications of LAGB. This highlights that a bibliometric analysis is a very useful way to ascertain the most relevant and important topics in an area of interest as well as providing the interested reader with a ready reckoner of the most influential manuscripts in that setting.

The traditional gold standard for assessing the impact of a published manuscript has been the TCC whilst an important measure of the quality of the research is the LOE it provides. Hence, bibliometrics which utilises these metrics for analysis and comparison is considered to be a very useful tool. Our study has shown a significant difference between the LOE and TCC as well as CRI. Studies with level 1 & 2 evidences have overall higher TCC and CRI compared to studies with level 4 & 5 evidences. Whist this is not unexpected, it does demonstrate that TCC and CRI are both useful metrics to assess academic impact and studies with superior evidence quality attract more citations. In contrast, Powell et al. did not note any correlation between LOE and TCC in their study and noted that this was a surprising result in their opinion. They attributed this to the challenges inherent in linking impact with citation and research quality but an alternative explanation could be the lead time bias inherent in measuring citation count [111].

An interesting finding of this study is that 89 of the 100 articles in this study have been published since 2000. This is perhaps not surprising because surgery for morbid obesity is still a field in its relative infancy and indeed most of the major developments have taken place during this century. However, the relatively short time period that spans most of the articles in this study makes this study uniquely robust in assessing the utility of AAS as an alternative index to other bibliometric indices. All the studies comparing bibliometric indices and AAS so far have been limited by the fact that a significant proportion of their articles were published before 2000 and, therefore, inherently biased against AAS as a measure of impact. This bias arises because most social media and other platforms included in the AAS scoring algorithm have evolved after 2000 such as Twitter in 2006 and Mendeley in 2007. In fact, our study demonstrates this bias clearly and shows a significantly lower AAS for articles published pre-2006 compared with articles post- 2006. Our study also shows that AAS correlates with LOE in a trend that is similar to TCC and CRI. This shows that AAS can be used as an alternative index to citation count and CRI as a measure of article impact. Previous studies have also demonstrated correlation between AAS and TCC [112]. Another finding that supports the credibility of AAS as an alternative metric to TCC and CRI is the fact that 60% of the publications all three lists of top 10 articles ranked on basis of the highest number of citations, CRIs and ASS were the same [Tables 1, 3 & 4].

Another interesting finding of note is that there is no significant difference in the association between AAS and TCC or CRI in post- 2006 articles (Figure 4a and 5a) suggesting that perhaps CRI is a redundant index and that TCC alone should suffice. However, in pre-2006 papers AAS is significantly correlated with CRI but not TCC (Figure 4b and 5b). The explanation of this perhaps lies in the phenomenon of lead-time bias or, more accurately, the lack of it in the post-2006 articles in this study. It is likely that articles published early in the 2006-2020 period have not had sufficient time to accrue enough citations compared to articles published more recently to generate a lead-time bias. It is likely that if this study is repeated after a decade or two, then divergent results will be noted between TCC and CRI in the post-2006 cohort.

This study has some limitations due to deficiencies intrinsic to the chosen study design as well as weaknesses inherent in the bibliometric indices. Firstly, citation counts and CRI are influenced by several types of bias such as language bias, institutional bias, and publication bias amongst others. In this study the language bias resulted in high numbers of English speaking countries being represented, particularly publications from USA which has been seen previously [111,113]. Secondly, the search strategy used to identify the top 100 articles invariably includes articles that are not relevant and can also exclude high impact articles. We noted that our search strategy included 4 articles that were eventually considered as being not relevant. Thirdly, as discussed earlier, the use of AAS as a metric is biased by the fact that articles published prior to early 2000s are significantly underrepresented in social media and other online platforms used to calculate the AAS. Finally, a top 100 analysis invariably limits the total data points available for analysis across different sub-groups to exactly 100 and the statistical results should be interpreted whilst bearing this in mind.

By analyzing the most influential articles that have shaped our understanding of LAGB, this study serves as a reference of the highest impact articles that have shaped the role of LAGB as an operation for morbid obesity whilst also serving as a guide for future research. Our study shows that AAS is a valid metric for assessing the impact of a study along with traditional metrics such as TCC and CRI. However, AAS is not reliable for articles published before 2006. Furthermore, the LOE of a study is significantly associated with all three indices of study impact including TCC, CRI and AAS.