In this study, WoSCC was selected as the data source. As a high-quality digital literature resource database, WoSCC has been accepted by many researchers, and considered as the most suitable database for literature analysis (23). All publications were retrieved from the Science Citation Index Expanded (SCI-E) of the WoSCC database on November 12, 2022. We completed the search within the same day to avoid any bias caused by database updates. The following methods were conducted for search publications: topic words = (“low carbohydrate” OR “low-carbohydrate” OR “low carb” OR “low-carb” OR “ketogenic” OR “carbohydrate-restricted” OR “carbohydrate restricted” OR “restricted carbohydrate” OR “restricting carbohydrate” OR “carbohydrate restriction” OR “South Beach diet” OR “Atkins diet”). In order to more accurately analyze the current status, hotspots and frontiers trends of LChD, the publications from 2002 to 2021 were selected. Time span = January 1, 2002–December 31, 2021. To ensure the representativeness of the included studies, the types of publications were limited to “articles” and “reviews” (24). No languages limitation to avoid bias in the geographical distribution of publications. The content of literature records were “full records and cited references,” downloaded and saved in plain text document format.

We used the CiteSpace (6.1.R3) and VOSviewer (1.6.18) for a bibliometric analysis of 6,938 publications on LChD from 2002 to 2021. The java-based program CiteSpace does bibliometric analysis of publications using distribution network maps, co-citation network maps, dual maps of journal overlay, and keyword burst citation maps (25). Nodes and links are included in the visual network diagram produced by CiteSpace. Every node is a factor, such as an author, an institution, or a country (26). Links between different nodes show a network of relationships involving co-operation, co-citation, or co-occurrence (27) A wider line indicates a more effective collaboration. The higher the centrality, the larger the circle is in terms of centrality. When a node has a purple circle around it, it has a high centrality score and is therefore an important node in the field (22). VOSviewer was used to form keyword co-occurrence of overlay visualization. The colors represent the years (28). The size of the node is proportional to the frequency of keyword occurrences (29). Data was managed, charts were made, and all data tables were created using Microsoft Excel 2021 software.

A total of 6,938 publications including 5,350 articles and 1,588 reviews, related to LChD from 2002 to 2021 were retrieved by searching the WoSCC database. The flowchart was shown in Figure 1. The annual publications reflected the activities in the field and the attention given to certain areas of research (30). As seen in Figure 2, the number of annual publications on LChD showed an overall upward trend in spite of fluctuation slightly in some years over the past 20 years. It indicates that LChD research is becoming a research of great interest to scholars and has attracted great interest from scholars in recent years.

LChD publications in the past 20 years can be divided into 2 stages. The initial stage (2002–2010) was a steady growth period. The average number of publications was 188 publications every year, with the lowest number of publications being 72 publications in 2002 and the highest number being 273 publications in 2009. In 1927, a low carbohydrate ketogenic diet had been reported for epilepsy (31). As an early study in 1948, LChD was used to control of dental caries (32). Since 2002, LChD was contributed to a variety of areas, including obesity (33), diabetes (34), and cardiovascular disease (35). Although the number of papers varied at this stage, the overall trend was one of consistent growth. The second stage (2011–2021) was a sustained growth period. The average number of publications annually was 476 publications. The number of publications reached 872 in 2021. Nutrition has a significant role in daily life, and it is crucial for the advancement of social development to support research on diet and health. LChD research has gained popularity as a nutritional approach and is rapidly developing into a research hotspot.

The categories refer to the disciplines covered by the dissertation research. At top 10 categories (Table 1), Nutrition Dietetics had 1,621 publications and ranked first, followed by Clinical Neurology (1,269 publications), Endocrinology Metabolism (960 publications), Neurosciences (734 publications) and Pediatrics (472 publications). LChD research mainly covered the fields of nutrition, endocrinology, and neurosciences, reflecting the multidisciplinary nature and comprehensive knowledge.

In total, 112 countries/regions participated in 6,938 publications on LChD from 2002 to 2021. CiteSpace generated the countries/regions distribution map, and 112 nodes and 880 links were shown in the map (Figure 3). Table 2 presented the top 10 countries/regions published in LChD research field. USA had the highest number of publications, 2,862 papers, accounting for 41.25%. The Yuasa phenomenon states that the nation whose research output accounts for more than 25% of all scientific output at any given moment can be referred to as the world center of science during that time (36). As the leader in LChD research, USA published far more than a quarter of the total publications and was the world science center in the field of LChD. England (543 publications), Italy (472 publications), China (449 publications), and Germany (441 publications) followed closely behind. In terms of centrality, Canada (0.16) ranked first, followed by, Spain (0.14), Australia (0.13), England (0.13), France (0.11) and, which maintain close cooperation relationships. Countries/regions with centrality played an important role in LChD research. Germany, Canada, Australia and France each had < 450 publications, but their research roles were important. In terms of publications, China had 449 papers, but the centrality was only 0.01. It demonstrated that despite having a high publications number, China had few connections and little influence over the network map. The level of LChD research in China therefore was raised effectively by deepening the field's research, advancing cross-disciplinary and cross-field collaboration, and enhancing researchers' capacity for creative thinking and global communication.

A total of 604 institutions provided research in the field of LChD. CiteSpace generated the institutions distribution map with 604 nodes and 2,103 links (Figure 4). The institutions with large numbers of publications have been identified as influential institutions (37). Table 3 listed the top 10 institutions in publications, and they were the most influential institutions in LChD research. Universities were major institutions for LChD research. Harvard University ranking first, had 451 papers, followed by University of California System (258 publications), Johns Hopkins University (216 publications), Udice French research universities (191 publications), and University of London (174 publications). Five of the top 10 institutions were from USA, which further confirmed US predominance in the field of LChD research. Duke University, Harvard University, Johns Hopkins University and University of Toronto had close collaboration relationships.

In total of 890 authors participated in 6,938 publications on LChD from 2002 to 2021. CiteSpace generated the institutions distribution map with 890 nodes and 1965 links (Figure 5). The top 10 authors participating in the LChD research are shown in Table 4. The most productive authors were Eric Heath Kossoff (101 publications), Jeff Scott Volek (69 publications), Jong M. Rho (62 publications), William S. Yancy (45 publications), and Maria Luz Fernandez (43 publications). Eric Heath Kossoff ranked first in the number of publications devoted to the study of the effects of a high-fat, low-carb ketogenic diet on neurological disorders. He demonstrated that a high-fat, low-carb ketogenic diet reduced the number of seizures in refractory epilepsy and reported no cardiovascular or cerebrovascular events (38, 39). In addition ketogenic diets are being applied to a range of neurological disorders from autism to Alzheimer's disease (40). Jeff Scott Volek was the second position of papers. He reported that in individuals with atherosclerotic dyslipidemia, a 12-week carbohydrate restriction diet improved postprandial vascular function more than a low-fat diet (41). An study revealed that LChD (10%) not only decreased lipid deposition but avoided the buildup of plasma and aortic oxidation, decreased inflammatory cytokines within the artery wall, and prevented atherosclerosis (42). Jong M. Rho was in the third place in terms of number of publications. In addition to a high-fat, low-carbon-water ketogenic diet that improves epilepsy (43), he emphasized that a ketogenic diet enhances mitochondrial function and reduces autistic behavior in humans and rodent models of autism spectrum disorder (44, 45). The authors' collaboration displayed a geographical concentration and general decentralization.

Researchers can accurately understand the core journals in a topic by analyzing its source journals, which also serves as a reliable resource for further field research (46). A total of 1,545 academic journals published 6,938 publications in the field of research on LChD from 2002 to 2021. As shown in Table 5, the top 10 journals accounted for 17.93% of the total publications. The most productive journals were Nutrients (292 publications), Epilepsia (203 publications), Epilepsy Research (134 publications), PLoS One (116 publications), and American Journal of Clinical Nutrition (105 publications). Of the top 10 journals, eight journals' IF more than 3.0. With a maximum of 8.472, the top 2 journals had an IF >6.0. This shows that high IF journals are open to publishing LChD research.

Figure 6 illustrated the dual- map overlay of journals that produced literature linked to the topic of LChD. On the map, the right labels represented the disciplines of the journals that published the cited papers, while the left labels represented the fields of the citing journals. Citation links can show the in and out of the citation dataset. Figure 6 showed 5 reference pathways. Three yellow pathways indicate articles published in molecular/biological/immunology journals mainly citing journals in the molecular/biological/genetics field. Two green pathways suggest that articles published in medicine/clinical journals mainly cite journals in the molecular/biology/genetics/health/nursing/medicine fields. One red pathway shows that the publications from neurology/sports/ophthalmology mainly cite journals in the in molecular/biology/genetics field.

The co-cited reference analysis is one of the important indicators in bibliometric research and is usually used to explore research priorities in specific academic fields (47). CiteSpace generated the co-cited reference map, and 1,693 nodes and 9,455 links were shown in the map (Figure 7). The top 5 co-cited references in terms of frequency were in Table 6. Analysis of co-cited references provided basic data for LChD research. Noteworthy were three publications from the New England Journal of Medicine and two from the Annals of Internal Medicine, both of which have significant academic influence. The five references were all clinical trials. In most co-cited reference, obese people were given Atkins diet, and lost more weight in the first 6 months (48). Additionally, high density lipoprotein cholesterol levels increased and triglyceride levels decreased more in Atkins diet participants than in control group, indicating that Atkins diet had a higher impact on the risk factors for coronary heart disease. The second-most co-cited reference reported that patients who received a carbohydrate-restricted diet with 30 g per day or less, lost more weight than control group did and had relative improvements in their insulin sensitivity and triglyceride levels (49). The third most co-cited reference of 132 obese people on who were restricted carbohydrate intake to < 30 g per day showed more beneficial effects than those on conventional diets at 1 year; the effects of restricted carbohydrate on atherogenic dyslipidemia and glycemic control remained more favorable (34). Diet therapies were given to moderately obese subjects, and a low-carbohydrate diet and a Mediterranean diet were found to have beneficial effects on lipids and blood glucose, respectively (50). Individualized dietary regimens tailored to individual preferences and metabolism are recommended. A low-carbohydrate, ketogenic diet exhibited higher participant retention and more weight loss compared to low-fat diets in the literature with the sixth greatest co-citation frequency (51).

High cited references lay the foundation and accelerate the development of research in the field (23). The top 10 cited references were listed in Table 7. Of the top 10 references, 7 references were articles and 3 were reviews. Three references were published in the New England Journal of Medicine and two were published in Lancet. “Obesity” published by Haslam et al. in 2005, was cited 3,136 times, and ranked first. Shai et al. published in 2008 in New England Journal of Medicine of “Weight loss with a low-carbohydrate, Mediterranean, or low-fat diet” was cited 1,250 times, and ranked second. Foster et al. published in 2003 in “A randomized trial of a low-carbohydrate diet for obesity” in New England Journal of Medicinewas cited 1,124 times, and ranked third.

The map of keywords can present the main research objects and the hot topics and frontiers trends. In this study, VOSviewer software performed the keyword co-occurrence of overlay visualization (Figure 8). A total of 9,750 keywords, 172 keywords met the thresholds when the minimum number of occurrences of a keywords was 20. From Figure 8, we found that the keywords research hotspots were categorized into “ketogenic diet,” “metabolism disease,” “cardiovascular disease” and “cancer.” Bursts keywords were frequently used at a period time, reflecting the frontiers trends. We used CiteSpace software to map the top 32 keywords with the strongest citation bursts from 2002 to 2021 (Figure 9). We summarized that “oxidative stress,” “gut microbiota,” and “inflammation factors” are becoming frontiers trends of LChD research in the future.