Emerging trends and knowledge structure of epilepsy during pregnancy research for 2000–2018: a bibliometric analysis

Data sources and search strategies

Literatures were retrieved online through the Social Science Citation Index and the Science Citation Index-Expanded of the Web of Science Core Collection (WoSCC) on March 7, 2019. The search terms were used for the following terms (Dataset S1) with a time span ranging from January 01, 1986 to December 31, 2018: ((“convulsion” OR “epilepsy” OR “seizure”) OR (“antiepileptic” OR “anticonvulsant” OR “AED” OR “phenytoin” OR “phenobarbital” OR “divalproex” OR “valproic acid” OR “carbamazepine” OR “oxcarbazepine” OR “levetiracetam” OR “gabapentin” OR “lamotrigine” OR “topiramate”)) AND (“pregnancy” OR “pregnancies” OR “gestation” OR “pregnant” OR “maternal” OR “fetus” OR “foetus” OR “newborn”) AND Language = English, only original articles and reviews were included. All literature retrieval and record downloads were completed within the same day so as to avoid changes in citation counts caused by daily database updates.

Data collection

Initially, the records retrieved from WoSCC were downloaded, screened, sorted and extracted by two researchers (ML Wang and WT Li) independently, any disagreements were discussed and a consensus reached. Then, these data were converted to txt format and imported into CiteSpace V 5.1.R8 SE, 64bit (Drexel University, Philadelphia, PA, USA) and the Online Analysis Platform of Literature Metrology (http://bibliometric.com/) for data analysis.

Analysis methods

We reviewed the characteristics of publications by establishing “The WoSCC Literature Analysis Report” online, including distribution of countries/regions, institutions, journals and authors, number of annual publications, citation counts and H-index. The impact factor (IF) and quartile in category of journals were obtained by querying the Journal Citation Reports (JCR) 2017 standards. Among these, the number of citations received by a paper is an indicator measuring scientific value (Eyre-Walker & Stoletzki, 2013). The IF is widely regarded as a leading proxy for importance and impact of medical journals in their respective disciplines (Roldan-Valadez et al., 2018). The H-index is a comprehensive indicator to quantify the productivity and impact of scientific research outputs of a journal, institute, scholar, and so forth. (Kulasegarah & Fenton, 2010). According to the classification of Clarivate Analytics, the journal ranking in the JCR is based on the IF of the current year, and then divided into four equal parts (25% each) of quartiles (Q) in category as Q1–Q4 (Lahuerta et al., 2017). By combining these kinds of scientific metrics, different aspects of publications (including influence, production and reputation) can be better measured.

In this study, the online analysis platform of Literature Metrology was used to analyze the trend of annual number of publications and growth trends of countries/regions. CiteSpace was used to identify research frontiers and emerging trends in this domain by exploring collaboration networks between authors/institutes/countries, analyzing the association between journals, identifying co-cited references as well as by capturing keywords with strong citation bursts over time. We set the “time slicing” in CiteSpace based on the overall time distribution of publications, and the “years per slice” was set to 1, the “top N per slice” was set to 50, which means the single network was extracted from the top 50 cited papers in a 1-year slice (Fig. S1). Different node types were selected according to the purpose of analysis, the size of node represents the number of publications or citation counts, each node was depicted with a series of citation tree-rings across a succession of time slice. The purple ring displays the structural properties of the node, and its thickness insinuates centrality degree, which is a measure associated with the transformative potential of scientific contributions. The red ring indicated a time slice in which citation bursts or an abrupt increase in references could be detected (Chen, Ibekwe-Sanjuan & Hou, 2010; Chen, 2006).

Distribution by time

A total of 2,225 records were downloaded from the WoSCC database, including 1,966 articles and 259 reviews (Fig. 1; Fig. S2). Although we attempted to retrieve all of the literature in the field of epilepsy during pregnancy from 1986 to 2018, the results of distribution by time indicate that the literature in this field was first published in 2000 (Dataset S2). The trend of annual number of publications related to epilepsy during pregnancy from 2000 to 2018 appears in Fig. 2A, overall trends present fluctuating growth from 57 articles in 2000 to 198 in 2018.

Distribution of countries/regions and institutions

The 2,225 articles concerning epilepsy during pregnancy were contributed by active authors from 50 countries/regions. The top 10 countries/regions were occupied in the relevant research included the US, England, Italy, Australia, and Canada (Fig. 2B). The US contributed the most publications (717), followed by England (220), Italy (153), Australia (137), and Canada (135). Concerning the centrality index, although Denmark had a low scientific research output, it had a greater impact on other countries’ research (centrality = 0.54), followed by Canada (0.40) and People’s Republic of China (0.27) (Table 1). Broad and extensive cooperation had been carried out between countries/regions, which can be observed in the network map of countries/regions related to epilepsy during pregnancy research (Fig. 3A).

The top 10 institutions engaged in relevant research sorted by number of publications included the University of Melbourne (68), Harvard University (60), the University of Queensland (59), Karolinska Institute (47), and Emory University (47) (Table 1). The network map of institutions involved in epilepsy during pregnancy research showed a low map density (Density = 0.0081) (Fig. 3B), suggesting that research groups were relatively dispersed, with scattered institutions and that mutual cooperation needed strengthening. The centrality index of most nodes was less than 0.15, indicating that most of the institutions’ influence was still at a low level and the degree of cooperation between institutions was inadequate.

Distribution by journals

The 2,225 articles were published in 592 journals (Dataset S3), the top 15 related journals included Epilepsia (7.326%), Epilepsy Behavior (6.067%), Seizure: European Journal of Epilepsy (4.090%), Neurology (3.461%), Epilepsy Research (2.652%), and Pediatric Neurology (2.382%). Among these, Neurology had the largest IF of 8.055, followed by Pediatrics (5.515), Epilepsia (5.067), Acta Neurologica Scandinavica (3.126), and Brain Research (3.125). According to the JCR 2017 standards, Neurology, Pediatrics, Epilepsia, and PlOS ONE were classified as Q1 sorted by the IF of the JCR category to which it belongs. Acta Neurologica Scandinavica, Brain Research, Seizure: European Journal of Epilepsy, Epilepsy Behavior, and Reproductive Toxicology were classified as Q2. Epilepsy Research, Pediatric Neurology, and European Journal of Paediatric Neurology were classified as Q3 (Dataset S4). The literature published in the journal of Neurology that involved in epilepsy during pregnancy research had the highest H-index (77), followed by Epilepsia (47), Epilepsy & Behavior (24), Seizure: European Journal of Epilepsy (20), and Epilepsy Research (20) (Table 2; Dataset S5).

Furthermore, we constructed a dual-map overlay of journals to clarify the cited relationship between journals (Fig. 4). It became clear that there were three main color citation paths in the double-map overlay of journals, the yellow path located at the top of pictures represents the papers published in molecular/biology/genetics journals that were partly self-cited and partly cited from journals in the psychology/education/social area, the intermediate green path represents papers published in medicine/medical/clinical journals, cited from journals in three fields: molecular/biology/genetics, health/nursing/medicine, and economics/political areas. The gray path at the bottom represents papers published in neurology/sports/ophthalmology journals, mainly cited from journals in molecular/biology/genetics, health/nursing/medicine, and economics/political areas.

Distribution by author

The collaboration network between authors showed many sub-networks, indicating that small and medium-sized research groups were widely distributed in this area, and there was little communication and cooperation between authors (showed in Fig. 5A). Among all retrieved documents, the top 10 productive authors sorted by number of publications included Pennell PB, Tomson T, Vajda FJE, Meador KJ, Thomas SV, and O’Brien TJ (Table 3). Among them, Prof. Pennell, from the Brigham & Women’s Hospital of Harvard University in the US, ranked first with 48 articles, followed by Prof. Tomson from Sweden Karolinska University Hospital with 44 articles. These two scholars made great achievements in the research of epilepsy during pregnancy and had high authority. In addition, the centrality of these collaborations was less than 0.1 suggesting that collaboration among authors is insufficient.

CiteSpace analyzed the information cited by the author and visualizes it in a co-cited authors network (Fig. 5B). Tomson T with 445 co-citations ranked first among the top 10 co-cited authors (Table 3), followed by Meador KJ (305), Holmes LB (284), and Pennell PB (232). These scholars conducted earlier research in the field of epilepsy during pregnancy, which laid a foundation for development of this domain. The centrality of most authors was less than 0.1, indicating that they did not form an influential core author group in the field of epilepsy during pregnancy research.

Analysis of the intellectual landscape

The network map of co-citation reference has revealed the intellectual landscape of this domain (Fig. 6A). It contained 771 references from the top 50 most-cited references per time slice for 2000–2018 (Dataset S6). Among the top 10 co-cited references sorted by the number of citations, the top ranked item was Morrow J (2006) with 162 citations in Cluster #1 (Table 4). In addition, we conducted the network map of co-citation clusters (Fig. 6B), which was one of important techniques in data mining and exploratory data analysis. The network had a high modularity of 0.8239 (showed in Fig. S3), indicating that the specialties in network map were clearly defined in terms of co-citation clusters. However, the mean silhouette value of 0.2169 was relatively low, for which the numerous small clusters possibly accounted, for the large major clusters we pay attention to in this paper is sufficiently high. Furthermore, the mean silhouette value of major clusters exceeded 0.75, indirectly indicating the reasonable and acceptable distributivity and homogeneity of clusters. Table 5 and Figure S4 demonstrates the specifics of the internal co-citation reference clusters. According to the log-likelihood ratio algorithm in CiteSpace software, Cluster #0 was labeled as “neonatal seizures” and ranked first, it contained 111 references with a mean silhouette value of 0.792, followed by the second largest cluster #1 was labeled as “fetal malformations.” Although the third largest cluster #2 was also labeled as “neonatal seizures,” cluster #2 had a different internal structure to cluster #0. A timeline visualization showed that the network was divided into distinct co-citation clusters (Fig. 6C). Tables 6 and 7 list key players of two major clusters #0 and #1 to indicate their critical research focuses.

Analysis of burst keywords

As a prominent representative of the research frontiers in a field of knowledge, burst keywords reflect the hotspots and frontiers in specific areas. As shown in Fig. 7, the time interval was plotted on the blue line and the period of burst keywords were plotted on the red line, which indicated the beginning and end of the time interval of each burst. “Management issue” (2013–2018), “brain injury” (2014–2018), “meta-analysis” (2014–2018), “in utero exposure” (2015–2018), and “autism spectrum disorder (ASD)” (2016–2018) were the keywords that had citation bursts after 2013.

General information

There were a total of 2,225 publications with a fluctuating growth trend over time. However, considering the cyclical nature of scientific research outputs, research in this area was generally increasing. The top 10 contributive countries/regions (two North American countries, five European countries, one Oceanian country, and two Asian countries) which were engaged in epilepsy during pregnancy research, contributed to 1,840 publications, accounting for 82.70% of the total number of publications. The US contributed 717 publications (about one-third of the total number of publications), reflecting its dominant position in epilepsy during pregnancy research. China and Turkey were the two-developing country in the list, which indicated they had made significant progress in epilepsy during pregnancy research over the past two decades. Furthermore, the most active collaborations were observed in Denmark, Canada, China, and the US.

Institutions with tremendous scientific research strength were observed to be mainly concentrated in higher education research institutions by examining both the publication counts and centrality, these institutions are important bases for medical scientific research and education. The top 10 institutions contributed to 447 articles, accounting for 20.09% of all publications. In this list, four of them (Harvard University, Emory University, Columbia University, and the University of California Los Angeles) were located in the US, indicating that institutions located in the US occupied the top rankings in terms of absolute contribution and relative influence, which is consistent with the analysis of contributions of countries in the field of epilepsy during pregnancy. Due to the close relationship between the level of health care and the speed of economic development, the institutional distribution table provides valuable information that helps researchers identify and choose appropriate collaborative institutions.

In terms of the top 15 academic journals, there were three journals: Neurology (8.055), Pediatrics (5.515), Epilepsia (5.067) had an IF score higher than 5, and these journals accounted for 11.87% (264/2,225) of total publications. The literature published in the journal of Neurology that involved in epilepsy during pregnancy research had the highest H-index (77), followed by Epilepsia (47), Epilepsy & Behavior (24), Seizure: European Journal of Epilepsy (20), and Epilepsy Research (20), indicating these journals were one of the most important information resources and the core journals in this area. According to the JCR 2017 standards, four journals were classified as Q1 contributed to 299 articles, accounting for 13.43% of all publications. And four journals were classified as Q2 contributed to 157 articles, accounting for 7.06% of all publications. Therefore, articles published in high-impact journals (classified as Q1, Q2) account for only one-fifth of the total numbers of publications, and the quality of related research in the field of epilepsy during pregnancy needs to be further improved.

The number of publications and count of citations could be adopted to evaluate the author’s absolute contribution and academic influence. Each of the top 10 authors identified in this analysis contributed to at least 19 publications. Therefore, they were referred to as “prolific authors.” Fortunately, five of these prolific authors (Tomson T, Meador KJ, Pennell PB, Vajda FJE, and Sabers A) were included in list of the top 10 co-cited authors, with regard to the annual co-citation counts, suggesting that these prolific authors consider both the number of articles and the quality of articles. Notably, Prof. Tomson had conducted a lot of earlier research, which laid a foundation for the development of this domain.

Intellectual base

In bibliometrics, the contemporary developmental state of one discipline could be reflected by research hotspots, and the intellectual base of this field could be constituted by the references in these articles (Egghe & Rousseau, 2002). References with higher frequencies of citation overwhelmingly centered around the topic of the major congenital malformation (MCM) and cognitive function after fetal exposure to AEDs. A prospective study of Morrow from the UK Epilepsy and Pregnancy Register (Morrow et al., 2006) assessed the relative risk of MCM from in utero exposure to AEDs and showed that the MCM rate for polytherapy exposure was greater than for monotherapy. Further research by Tomson showed that the risk of MCMs could be affected by type of AEDs, dose and other variables, which ought to be considered under the situation of managing epilepsy in women of childbearing potential (Tomson et al., 2011). Holmes explored whether MCM was caused by maternal seizures itself or by AED exposure (Holmes et al., 2001). Particularly, a longer-term follow-up by Adab in children exposed to AEDs in utero (Shorvon, 2004) found that a lower verbal IQ and developmental delay may closely relate with recurrent tonic-clonic seizures during pregnancy. A large number of studies had shown that in utero exposure to valproic acid (VPA) was associated with increased risk of impaired cognitive function compared with other widely-used AEDs (Meador et al., 2009, 2013; Shorvon, 2004).

In the network map of co-citation references, Cluster #1 contained many nodes with purple rings of betweenness centrality, and Cluster #0 contained numerous nodes with red rings of citation bursts, indicating that this was the most recently formed cluster. Therefore, we particularly focused on cluster #0 to identify emerging trends in epilepsy during pregnancy. Eight most-cited references and eight citing articles in this cluster were selected, all among citations published in the past 10 years (showed in Table 6). An article by Eadie had the highest citation coverage of 18%, their primary concern was why maternal seizure control may deteriorate during pregnancy and how this may be avoided (Eadie, 2014). Other citing articles evaluated pregnancy and delivery complications in WWE (Borthen & Gilhus, 2012) or called for more high-quality studies of malformation risks of newer AEDs in pregnancy (Bobo et al., 2012; Holmes & Hernandez-Diaz, 2012; Vajda et al., 2014). Thus, future research in the field of epilepsy during pregnancy should focus on comprehensive management of pregnant and lactating women, and the evaluation of the efficacy and safety of newer AEDs through high-quality clinical research or systematic reviews, so as to provide evidence for rational clinical application.

Research frontiers

Strengths and limitations

To the best of our knowledge, this study is the first bibliometric analysis focus on epilepsy during pregnancy trends over the past two decades. There was no time limit for our literature retrieval, the data downloaded from WoSCC database covered the vast majority of articles in the field of epilepsy during pregnancy research, data analysis was relatively objective and comprehensive, which clearly showed the past and current status of epilepsy during pregnancy and predicted the future research frontier. However, this study consisted exclusively of original articles and reviews published in 2000–2018 and indexed by the web of science (WOS) database, since books, conference abstracts, and other journal publications were not included in the range of document screening, our data may not represent all of the literature. Furthermore, we only included articles written in English in our analysis, making the analysis incomplete to some extent. In terms of retrieval of databases, we only retrieved publications from the WoSCC database. Although other databases such as PubMed, Scopus, and Embase could provide a broader range of coverage of scientific literature, the references are not available. The WOS database is currently the main citation index database based on the mutual citation relationship between documents, which allows a scientific assessment of the following contents from the perspective of literature citation, including academic value of the literature and relevant researchers, institutions, journals, and national overall research output level. Thus, the WoSCC database may be the only appropriate choice. Finally, the results of this study were stable and highly reproducible, since this study covers the vast majority of papers from 2000 to 2018, and the latest published articles may not affect the final results.