References provide the knowledge basis and theoretical framework for subsequent research, and citation of references is crucial for scientific investigation. The fundamental theory of co-cited reference analysis lies that, supposing that two references are simultaneously cited by one literature, a “co-citation” behavior will be recognized, indicating the relevance between the references. In this case, the higher the co-citation frequency, the greater the reference value of the publication. Therefore, the highly co-cited articles are instructive for later studies in the field. Table 5 counts the top 8 co-cited references, among which a review article that offered a comprehensive description of bacterial and fungal endophthalmitis in 2017 is far ahead in the citation count (Durand, 2017). Then, most of the remaining publications concern post-cataract endophthalmitis, led by a prospective, randomized, and multicenter clinical trial initiated by the ESCRS endophthalmitis study group (Barry and Grp, 2007). The designed trial compared intracameral cefuroxime at the end of surgery and perioperative levofloxacin drops for the prevention of postoperative endophthalmitis, and logistic regression analysis was used to identify several related risk factors, which deepened the understanding of endophthalmitis after cataract extraction (phacoemulsification with intraocular lens implantation).

Based on the log-likelihood ratio (LLR) algorithm which can effectively process large-scale data and high-dimensional features and is generally applied to classification and model selection issues in various fields, CiteSpace can carve all references into several separate clusters, and papers of the same cluster represent a subtopic whose definition derives from the title terms of the citing papers in this cluster. Figure 5A retains the top 10 clusters, which are #0 cataract surgery, #1 endophthalmitis vitrectomy study, #2 endogenous endophthalmitis, #3 intravitreal treatment, #4 intravitreal injection, #5 intracameral antibiotics, #6 intravitreal voriconazole, #7 intravitreal anti-vascular endothelial growth factor injection, #8 open globe injury, and #9 25-gauge vitrectomy, respectively.

In the co-citation network, the arrows that connect clusters are called “cluster dependencies”. Like the dual-map overlay of journals, cluster dependencies represent the referential relationship between clusters, with arrows pointing to the cited cluster. Briefly, the left clusters have been tested over time and become a classic theoretical foreshadowing in the field, while the clusters on the right generally represent the latest advances and quote the published articles from the left clusters. From Figure 5A , not only can we discover close interaction among all the clusters, but the affiliation and knowledge extension between different clusters can also be intuitively visualized.

The cluster map could be converted into a timeline view to record each cluster’s emergence, evolution, and development on the timeline ( Figure 5B ). Like cluster dependencies mentioned above, a timeline diagram can also vividly depict the links in the time dimension, reflected in the lines connecting the nodes. The EVS (cluster #1) is the first to emerge and afford support for subsequent intravitreal therapy (cluster #3). Also worth noting is that cataract surgery (cluster #0) became a hot academic topic in 2000-2010 and was linked to multiple clusters. It contains several papers with frequent co-citation (large nodes) and high betweenness centrality (marked purple ring). However, with advances in aseptic operational techniques and modified phacoemulsification, post-cataract endophthalmitis has declined dramatically. Currently, endogenous endophthalmitis (cluster #2), endophthalmitis after intravitreal anti-VEGF (clusters #4 and #7), and intracameral prophylactic antibiotics (cluster #5) are the subjects more discussed.

In the timeline view ( Figure 5B ), some nodes surrounded by purple rings can be observed, representing studies with a high betweenness centrality, which often act as a bridge between different sub-directions. Among these ten clusters, nine references with the highest betweenness centrality are included in Table 6 . On the one hand, the incidence of postoperative endophthalmitis, especially following cataract surgery, was repeatedly introduced. Two studies were from the US (Aaberg et al., 1998; West et al., 2005), while the other presented data from the German population (Schmitz et al., 1999). On the other hand, the effectiveness of prophylactic antibiotic use after cataract surgery and intravitreal injection was evaluated (Ciulla et al., 2002; Garcia-Saenz et al., 2010; Bhatt et al., 2011). For example, a Spanish study compared the endophthalmitis rates in cataract surgery before and after using prophylactic intracameral cefuroxime. It concluded that cefuroxime effectively reduced the risk for acute-onset postoperative endophthalmitis (Garcia-Saenz et al., 2010).

Clusters #1 (endophthalmitis vitrectomy study) and #9 (25-gauge vitrectomy) are both related to vitrectomy ( Table 7 ), with the difference that #1 explores the therapeutic effects of vitrectomy while #9 is concerned with endophthalmitis following PPV. In 1995, a randomized clinical trial named Endophthalmitis Vitrectomy Study (EVS) was conducted through 2×2 factorial analysis (immediate pars plana vitrectomy/vitreous tap or biopsy, with or without systemic antibiotics). This study found that vitrectomy benefited bacterial endophthalmitis patients with only light perception vision. On the contrary, routine immediate PPV may not be necessary when the vision is better than light perception (Vine et al., 1995). Although this trial has been carried out for a long time, its results are recognized as a guide for follow-up research (Han et al., 1996a, Han et al., 1996b; Johnson et al., 1997; Ng et al., 2005; Grzybowski et al., 2018). Regarding postoperative endophthalmitis caused by vitrectomy, which usually occurs within 15 days after the transconjunctival sutureless operation (Scott et al., 2008), the patient’s visual outcome is often poor, albeit with a low incidence (Eifrig et al., 2004; Kunimoto and Kaiser, 2007). On this basis, retrospective statistics indicated that 25-gauge vitrectomy was more likely to contribute to endophthalmitis than 20-gauge vitrectomy (Taban et al., 2006; Chen, 2007; Kunimoto and Kaiser, 2007; Scott et al., 2008).

Providing that the outside microorganisms have access to entering the sterile intraocular tissue through openings on the ocular surface caused by various reasons, relatively common exogenous endophthalmitis may be triggered. According to the wound type, it can be subdivided into postoperative (cluster #0 cataract surgery), post-injection (cluster #4 intravitreal injection and #7 intravitreal anti-vascular endothelial growth factor injection), and post-traumatic (cluster #8 open globe injury) endophthalmitis.

Observably, most of the top co-cited papers ( Table 5 ) are from cluster #0 ( Table 8 ), showing that a lot of effort and attention has been spent on post-cataract endophthalmitis, including aspects of the incidence, risk factors, possible mechanisms, precautions, and dosing regimens (Ciulla et al., 2002; Taban et al., 2005; West et al., 2005; Barry and Grp, 2007). In addition to the previously mentioned ESCRS multicenter study, a systematic review also summarized the significant effect of incision type on the inflammatory risk, proposing that the incidence in the clear corneal incision group was 2.55 and 3.06 times (relative risk, RR) higher than the scleral incision and limbal incision group, respectively (Taban et al., 2005). The same concern was reviewed in a leading citing publication from this cluster ( Table 8 ) (Lundstrom, 2006). Another study based on the US Medicare population concluded that individuals of older age and black race were at increased risk of post-cataract endophthalmitis between 1994 and 2001 (West et al., 2005). On the other hand, several citing articles evaluated the effectiveness of different antibiotic combinations for preventing and treating postoperative endophthalmitis (Busbee, 2006; O'brien et al., 2007; Fintelmann and Naseri, 2010).

Clusters #4 and #7 are the same in terms of content, and they are both about possible intraocular inflammation after intravitreal anti-VEGF ( Table 8 ). With the gradual widespread use of anti-VEGF, post-injection endophthalmitis has begun to attract attention. Contributions in these two clusters concentrated on incidence, visual outcomes, and pathogenic microbial profiles. Although the incidence was low (varying from 0.015% to 0.05% reported in different studies) (Mccannel, 2011; Moshfeghi et al., 2011; Storey et al., 2014), visual acuity outcome was generally unsatisfactory and associated with more common and severe streptococcal infections (Mccannel, 2011; Moshfeghi, 2011; Moshfeghi et al., 2011; Xu et al., 2018). Therefore, avoiding treatment on eyes with active ocular surface or eyelid diseases and reducing droplet transmission during the injection, such as avoiding talking, coughing, sneezing, and wearing a medical mask, may be practical measures (Mccannel, 2011; Patel et al., 2021a; Singh et al., 2022).

Open globe injury (cluster #8) is a broad type of ocular trauma characterized by full-thickness lacerations of the cornea and sclera, either penetrating or perforating (Kuhn et al., 2004). According to the US Eye Injury Registry report, 3.4% of open globe injuries had endophthalmitis (Danis, 2002). Moreover, because the pathogens causing post-traumatic endophthalmitis are distinct from those in other types (Bhagat et al., 2011), this topic forms a separate cluster ( Table 8 ). Regular pathogens are coagulase-negative Staphylococci, B. cereus (Hong et al., 2016), and sometimes fungi (most commonly C. albicans and Aspergillus) (Chakrabarti et al., 2008; Chhablani, 2011). Microorganisms with different virulence could directly influence visual prognosis. Intravenous and intravitreal antibiotic treatment should be started urgently (Bhagat et al., 2011).

Accounting for only 5% to 15% of all endophthalmitis cases, endogenous endophthalmitis is blood-borne and in most cases associated with severe systemic infection. Approximately 0.05% to 0.4% fungemia and 0.04% bacteremia may be complicated with endophthalmitis (Spelta et al., 2021). Of note, in cluster #2 (endogenous endophthalmitis), two reviews written by Durand ML are recognized as the popular co-cited papers ( Table 9 ), which both detailed a variety of endogenous and exogenous endophthalmitis, either bacterial or fungal infections, and updated the research progress on this rare but severe disease (Durand, 2013, Durand, 2017). In particular, the review article published in 2017 is the most cited reference in all clusters ( Table 5 ) (Durand, 2017). Additionally, in the list of the cited articles with the most coverage, Danielescu C et al. (Danielescu et al., 2020, Danielescu et al., 2022). recently reviewed an endogenous endophthalmitis case series and the diagnosis and treatment of fungal endophthalmitis. Blood cultures revealed that hematogenous disseminated infection with fungi such as Candida was a major cause of endogenous endophthalmitis. Thus, as one of the significant quotes of cluster #2 ( Figure 5A ), cluster #6 (intravitreal voriconazole) involves the connections among the epidemiology, cultured microorganisms and visual acuity outcomes of endogenous endophthalmitis (Jackson et al., 2003; Schiedler et al., 2004; Ness et al., 2007), along with the role of intravitreal anti-fungal voriconazole in the management of culture-proven endophthalmitis ( Table 9 ) (Breit et al., 2005; Yang et al., 2011).

Finally, clusters #3 (intravitreal treatment) and #5 (intracameral antibiotics) refer to the application of antibiotics to treat or prevent acute-onset postoperative bacterial endophthalmitis ( Table 10 ). Administration of intravitreal antibiotics is the mainstay of treatment for acute-onset endophthalmitis and achieves higher intraocular antibiotic concentrations than any other modality of administration (Kresloff et al., 1998). The EVS found that all gram-positive isolates were susceptible to vancomycin, while most isolated gram-negative organisms were equally sensitive to amikacin and ceftazidime (Han et al., 1996b). On the side, following the clear benefits of intracameral cefuroxime reported by prospective ESCRS study (Barry and Grp, 2007), multiple data highlighted the inhibitory effects of intracameral antibiotic prophylaxis on postoperative endophthalmitis, where cefuroxime, vancomycin, and moxifloxacin were preferred (Barry, 2014; Chang et al., 2015; Haripriya, 2017; Haripriya et al., 2017). In a Swedish national study, the non-use of intracameral cefuroxime was even identified by logistic regression as a significant risk factor of endophthalmitis (Friling et al., 2013).

To alleviate the social burden of the alarming postoperative complication, the EVS group was established in the United States in the late 1980s to conduct the Endophthalmitis Vitrectomy Study (EVS). The group recruited 420 patients between February 1990 and January 1994, and aimed to evaluate the role of PPV and systemic antibiotics in acute-onset endophthalmitis following cataract surgery or secondary intraocular lens implantation (Vine et al., 1995). As the only multicenter, prospective, randomized clinical trial of vitrectomy to date, the EVS compared the efficacy of PPV and tap/biopsy in the management of post-cataract endophthalmitis and was dedicated to reasonable treatment guidelines (Vine et al., 1995; Grzybowski et al., 2018), thus laying the groundwork for the leading position of the United States in this field ( Table 1 ; Figure 2C ). It is worth noting that the two core co-cited authors, Han DP and Vine AK ( Table 2 ; Figure 2D ), are both from the EVS group. Because the primary observed outcome was the recoverable visual acuity after 9-12 months, the vital conclusion was that the benefit from vitrectomy over tap/biopsy was admitted in patients with only light perception vision, while no advantages were presented in patients with better visual acuity than delicate perception (Vine et al., 1995). Meanwhile, the utilization of systemic ceftazidime and amikacin exerted no direct alterations on final visual acuity or media clarity, so omitting intravenous antibiotics could reduce side effects and help control the cost and duration of hospital stay (Vine et al., 1995).

Additionally, during the following years (notably 1996-2001), the study team handled this prospective data to supplement a series of additional analyses (Han et al., 1996a, Han et al., 1996b; Bannerman et al., 1997; Barza et al., 1997; Johnson et al., 1997; Wisniewski et al., 1997; Doft et al., 1998; Han et al., 1999; Doft et al., 2000; Wisniewski et al., 2000; Doft et al., 2001), which were mainly published in several authoritative ophthalmology journals, including Archives of Ophthalmology (the predecessor of JAMA Ophthalmology), Ophthalmology, and AJO ( Table 3 ; Figure 3B ). Some recommendations or perspectives were given.

First, reduced visual acuity, conjunctival hyperemia, pain, hypopyon, and eyelid swelling were found to be typical clinical presentations of post-cataract endophthalmitis, and these features helped predict a possible cultured microbiologic spectrum (Johnson et al., 1997; Wisniewski et al., 2000). Of all intraocular sample isolates obtained, 70% were gram-positive coagulase-negative Staphylococci with S. epidermidis predominantly, which were identified to potentially originate from periocular skin flora by pulsed-field gel electrophoresis, reinforcing the essentiality for rigorous preoperative disinfection of the surgical sites (Han et al., 1996b; Bannerman et al., 1997).

Second, concerning the diagnostic techniques, vitrectomy cassette fluid did not have a higher culture-positive rate than undiluted vitreous obtained by tap/biopsy (Barza et al., 1997). Similarly, there was no significant difference between mechanized vitreous biopsy and needle aspiration regarding microbiologic yields and operative complication rates in EVS (Han et al., 1999).

Third, although visual prognosis is closely related to the type of organisms and their gram stain results, baseline visual acuity at initial diagnosis was put in a better position than microbiological factors in predicting visual outcome and determining immediate vitrectomy value (Han et al., 1996a). Moreover, patients who required additional procedures due to surgical complications and worsening intraocular infections or who developed retinal detachment after the initial treatment generally owned much worse follow-up visual acuity, and these factors were seen as a sign of more severe disease (Doft et al., 1998, Doft et al., 2000).

Fourth, diabetes mellitus, a common systemic disease, was studied separately. It was reported that diabetes seemed to be relevant to coagulase-negative micrococcal infection (Johnson et al., 1997). For diabetic patients with initial visual acuity better than light perception, the proportion achieving visual acuity of 20/40 after vitrectomy (57%) was slightly higher than that after tap/biopsy (40%). However, no statistical difference was figured out, which means the optimal treatment regimen for diabetic patients needs to be further clarified (Doft et al., 2001).

Finally, the economic implications of EVS were also taken into account, which was one of the original intentions of the study (Flynn and Scott, 2008). The charge-effectiveness analysis in 1996 estimated that assuming the EVS recommendations were followed for endophthalmitis after cataract surgery, the annual cost of hospitalization in the United States would fall by between $7.6 million and $40.0 million (Wisniewski et al., 1997).

Voices of skepticism remained. The most prominent criticism was the choice of ceftazidime and amikacin as intravenous antibiotics for post-cataract endophthalmitis, as they have poor activity against the most commonly isolated staphylococci in EVS and the inability of amikacin to cross the blood-eye barrier results in minimal intraocular concentrations (Durand, 2013). A western Australian report suggested that despite significant changes in managing postoperative endophthalmitis since the EVS, patients’ visual outcomes did not improve, which was connected to a lack of oral antibiotic therapy (Ng et al., 2005).

On the other hand, the indications for implementing PPV were also controversial. The 2002 Canadian survey showed that most Canadian vitreoretinal surgeons did not strictly follow the recommendations of EVS (Siddiqui et al., 2002). The benefits of PPV may be underestimated in the EVS due to the exclusion of severe cases with anterior chamber opacification or without light perception (Flynn and Scott, 2008), along with the absence of subjects with other types of endophthalmitis caused mainly by more virulent organisms (e.g., streptococcus). What’s more, considering that nearly 30 years have passed since the release of the EVS, during which significant progress has been made on vitrectomy, and the minimally invasive surgery allows doctors to minimize operative risks and achieve better outcomes, some researchers insist on complete and early vitrectomy for endophthalmitis (Grzybowski et al., 2018). In short, treating acute postoperative endophthalmitis should not solely focus on initial visual acuity but also consider individualized clinical manifestations and disease duration.

Unlike EVS, the ESCRS study and ASCRS survey centered on acute post-cataract endophthalmitis prophylactic approaches. All their results were published in JCRS, an influential journal of cataract and refractive surgery ( Table 3 ). Barry P (the ranking co-cited author in Table 2 ) led the implementation of the ESCRS multicenter study, which was finished together by 24 ophthalmology units in Austria, Belgium, Germany, Italy, Poland, Portugal, Spain, Turkey, and the United Kingdom, so it is not surprising that the United Kingdom becomes the other country with a high centrality outside of the United States ( Table 1 ; Figure 2C ). This European study fully affirmed the significant preventive effect of intracameral cefuroxime on post-cataract endophthalmitis at the conclusion of surgery. A lack of intracameral cefuroxime prophylaxis would result in a near five-fold risk of endophthalmitis (Barry et al., 2006; Seal et al., 2006; Barry and Grp, 2007). Concerning laboratory diagnostics, the accessional studies noted that the introduction of molecular biology techniques like polymerase chain reaction (PCR) improved the identification rate of pathogens, with a statistical association between the laboratory-proven endophthalmitis cases and clinical symptoms and signs, including eyelid swelling, pain, and vitreous opacity (Seal et al., 2008; Barry et al., 2009). Subsequent surveys towards European ophthalmic surgeons demonstrated that adopting intracameral prophylactic antibiotics became mainstream (Gore et al., 2009; Barry, 2014). Likewise, retrospective evidence from Portugal and Italy supported the protective effect of cefuroxime against endophthalmitis (Beselga et al., 2014; Grosso et al., 2016).

The prevalence of intracameral antibiotics is also reflected in the changeover in the results of the ASCRS member surveys. ASCRS launched three online surveys to its members in 2007 (Chang et al., 2007), 2014 (Chang et al., 2015), and 2021 (Chang and Rhee, 2022), aiming to understand the intentions of refractive surgeons regarding prophylactic antibiotic regimens. Right after the publication of the ESCRS results, the 2007 ASCRS questionnaire indicated a solid tendency to topically use the latest generation of fluoroquinolones, with no use of the intracameral form in a whopping 77% of the respondents (Chang et al., 2007). From 50% in 2014 to 66% in 2021, the anterior chamber injection of prophylactic antibiotics gradually replaced topical agents and was transformed into a consensus under the influence of the ESCRS research series, in which vancomycin utilization gradually declined in the United States, followed by more frequent injections of moxifloxacin (Chang et al., 2015; Chang and Rhee, 2022). Besides, commercially approved antibiotic formulations for intracameral prophylaxis are highly anticipated to avoid potential dilution and contamination risks.

Fungal infections are more common in endogenous endophthalmitis, which originates from infections of other sites (e.g., liver abscess, endocarditis, and urinary tract infections) through blood spread, because of systemic immunodeficiency (Durand, 2017). The visual outcomes are incredibly pessimistic and unsatisfactory, with blindness in most cases, and extraocular foci are often associated with high mortality (Jackson et al., 2003; Schiedler et al., 2004). As a severe complication of systemic candidemia, Candida endophthalmitis has been identified as the most frequent form of endogenous fungal endophthalmitis, followed by mold endophthalmitis, including Aspergillus and Fusarium infections (Kramer et al., 2006; Yoshida et al., 2018; Zhuang et al., 2020). For therapeutic strategies, immediate intravitreal administration combined with systemic antifungal medications and subsequent vitrectomy is beneficial. Amphotericin B and voriconazole are commonly used antifungal agents and have exhibited favorable effects in controlling inflammation (Yang et al., 2011; Bae and Lee, 2015; Zhao et al., 2015; Bienvenu et al., 2020). What’s more, several case reports published in recent years have shown that in the cases of resistant fungal endophthalmitis, intravitreal caspofungin, the first approved antifungal echinocandin, can be an ideal and safe alternative to the former two (Danielescu et al., 2017; Yadav et al., 2017; Von Jagow et al., 2020; Nakhwa, 2021).