Gram-Negative Endogenous Endophthalmitis: A Systematic Review

Background: Gram-negative bacteria are causative agents of endogenous endophthalmitis (EBE). We aim to systematically review the current literature to assess the aetiologies, risk factors, and early ocular lesions in cases of Gram-negative EBE. Methods: All peer-reviewed articles between January 2002 and August 2022 regarding Gram-negative EBE were included. We conducted a literature search on PubMed and Cochrane Controlled Trials. Results: A total of 115 studies and 591 patients were included, prevalently Asian (98; 81.7%) and male (302; 62.9%). The most common comorbidity was diabetes (231; 55%). The main aetiologies were Klebsiella pneumoniae (510; 66.1%), Pseudomonas aeruginosa (111; 14.4%), and Escherichia coli (60; 7.8%). Liver abscesses (266; 54.5%) were the predominant source of infection. The most frequent ocular lesions were vitreal opacity (134; 49.6%) and hypopyon (95; 35.2%). Ceftriaxone (76; 30.9%), fluoroquinolones (14; 14.4%), and ceftazidime (213; 78.0%) were the most widely used as systemic, topical, and intravitreal anti-Gram-negative agents, respectively. The most reported surgical approaches were vitrectomy (130; 24.1%) and evisceration/exenteration (60; 11.1%). Frequently, visual acuity at discharge was no light perception (301; 55.2%). Conclusions: Gram-negative EBEs are associated with poor outcomes. Our systematic review is mainly based on case reports and case series with significant heterogeneity. The main strength is the large sample spanning over 20 years. Our findings underscore the importance of considering ocular involvement in Gram-negative infections.


Introduction
Endogenous endophthalmitis (EE) is a rare but devastating complication of bloodstream infections found in less than 0.5% of patients with fungemia and 0.04% of patients with bacteraemia [1]. When left untreated, endogenous bacterial endophthalmitis (EBE) can damage the eye's structures, leading to visual impairment and even blindness [2]. From any possible source of infection, the aetiological agent may spread through the bloodstream across the blood-retinal barrier (BRB), eventually reaching the eye structures [2]. Nearly 10% of EBE worldwide are caused by Staphylococcus aureus (S. aureus), which is also the leading cause of EBE in the USA and Europe, comprising approximately 25% of these cases [3].
S. aureus itself can alter BRB tight junctions by disrupting the expression and/or organisation of the ZO-1 protein [4]. In other words, S. aureus possesses the ability to cause EBE regardless of pre-existing vascular leakage [5]. However, other bacteria are involved in the aetiology of EBE: Streptococcus spp. (including viridans group, S. pneumoniae, Streptococcus milleri group and group A and B streptococci), and Gram-negative pathogens

Data Items
For each selected article, we collected information regarding the number of patients with an EBE due to a Gram-negative infection, their demographic data (age, sex, and ethnicity), comorbidities, and the number of eyes involved, specifying (when available) if a right/left eye was affected. Aetiological data (Gram-negative bacteria involved, culture type and source of infection), as well as initial ocular lesions and visual acuity were reported. Furthermore, we assessed ocular and systemic complications, medical therapy (anti-Gramnegative topical, intravitreal, or systemic antibiotics along with the addition of steroids) and the eventual surgical approach employed. We reported the general and ocular outcome and of a follow-up visit within the 12 months after discharge had been performed. Missing or unclear data were reported as "non-available". Similarly, we considered "non-available" data regarding EBE due to a Gram-negative infection but impossible to extrapolate because included in more comprehensive studies concerning endophthalmitis in general.

Synthesis Methods
All the collected data were reported in a single table that was revised by an independent group. Every column was specifically associated with a different item. In the case of columns with less than 5 records, a grouping of the result was performed: i.e., in the case of poorly represented bacterial species, we preferred grouping them under an "other gram-negative aetiology" column. We limited our study to a descriptive analysis of our findings due to the wide heterogenicity of the articles selected. The percentage calculation was performed in consideration of the number of data available for each specific item. No models to identify the presence and extent of statistical heterogeneity or sensitivity analyses to assess the robustness of the synthesised results were performed.

Bias and Certainty Assessment
This is a systematic review for which a descriptive analysis has been performed due to the wide heterogenicity of the selected articles. Risk of bias or certainty (or confidence) in the body of evidence was not assessed.

Study Selection and Search Results
A total of 154 papers were identified through our search. We excluded 19 duplicate articles. A further eight analyses were removed as five were systematic reviews or metaanalyses [3,[13][14][15][16] and three were pre-clinical sciences papers [17][18][19]. Moreover, two studies were excluded because the full text was unavailable [20,21] and one paper was removed due to a lack of data regarding the peer-revision process of the journal [22]. The remaining 124 articles were assessed for eligibility. Seven were excluded as data regarding EBE due to a Gram-negative infection were available, but impossible to select [23][24][25][26][27][28][29] and two further analyses were removed as they did not fully meet the inclusion criteria [30,31]. Eventually, 115 studies were included, as shown in the following flow diagram (Figure 1).
Most studies were case reports (57, 49.6%) and retrospectively non-randomized (46, 40%). Regarding the geographic distribution of the studies, 58.1% of the articles were from Asia, 13.2% were from Europe, and 9.6% were from America. Study characteristics, patient comorbidities, and aetiologic data included are summarized in Table 1  . Most studies were case reports (57,49.6%) and retrospectively non-randomized (46, 40%). Regarding the geographic distribution of the studies, 58.1% of the articles were from Asia, 13.2% were from Europe, and 9.6% were from America. Study characteristics, patient comorbidities, and aetiologic data included are summarized in Table 1  .
Regarding the clinical outcomes at the end of the hospitalization, most patients were discharged (238, 85%), and mortality was recorded in only 15 cases. A higher percentage of NLP or no light perception (301, 55.2%) was reported as final ocular outcome compared to the initial visual acuity assessed. Only 25 studies reported follow-up information, and only five relapses occurred within 12 months after discharge.

Discussion
This systematic review estimates the clinical and epidemiological impact of Gramnegative EE, by analysing over a hundred papers spanning 20 years. EBEs, defined as the infection of intraocular tissues resulting from the hematogenous spread of bacteria to the eye, are both a diagnostic and therapeutic challenge for ophthalmologists and infectious diseases specialists [67]. Gram-negative EBEs are an undoubtedly consistent clinical reality associated with poorer outcomes due to the production of endotoxins and the phagocytosisresistant capsules conferring greater virulence [67,147].
Our search shows that, in East Asian nations, many EBEs are caused by Gram-negative bacilli, including Klebsiella pneumoniae and Escherichia coli [1]. Studies from Singapore and Taiwan showed that up to 70% of the organisms isolated from patients with EBE were Gram-negative [67]. Similarly, other analyses reported frequencies ranging from 22.2% to 77.1%, considering Gram-negative as causative agents of EBE [147]. Interestingly, Klebsiella was found to be the most common causative organism (31.7%-87.6%) followed by Pseudomonas aeruginosa [147].
In our systematic review, Klebsiella pneumoniae was the most common pathogen isolated, while liver abscesses represented the primary source of infection. Indeed, the association of liver abscesses with Klebsiella as the causative organism is observed worldwide, especially when considering hypervirulent strains (hvKp) [6]. Although the mortality rate of hvKp liver abscess is relatively low compared to that associated with pyogenic liver abscesses caused by bacteria other than K. pneumoniae, hvKp infection can lead to metastatic complications that cause significant morbidity such as, for instance, EBE [15]. Most hvKp infections are community-acquired, often afflicting individuals without any predisposing medical condition [15]. The incidence of hvKp infections seems to be rising both in Asia and Europe, and this can be explained by the rates of hvKp-carriers that range from 19% to the alarming percentage of 88% of healthy Chinese adults [15,148]. In a recent systematic review, 1 out of 22 patients with Klebsiella pneumoniae pyogenic liver abscess was found to develop EBE. This is explained by the K antigen, a capsular polysaccharide and a well-established virulence factor that makes K1 serotype infection an independent risk factor for the development of EE [15].
Although cases of EBE have been reported in otherwise healthy and immunocompetent people, EBEs are frequently associated with many systemic risk factors, including chronic immune-compromising illnesses, immunosuppressive diseases or therapies, recent invasive surgery or gastrointestinal procedures, hepatobiliary tract infections, and intravenous drug use [67]. Diabetes is the primary underlying condition associated with EBE (46-63.86%) in Asia. Considering the current scenario of the COVID-19 pandemic, the heavy use of systemic corticosteroids can predispose patients to the subsequent development of EBE via steroid-induced diabetes [147,149]. Although the pathogenic mechanism is poorly understood, it is known that poor glycaemic control might impair neutrophilic hepatic Kupffer cells' phagocytosis against the bacterial infiltrators arriving with portal blood [15]. A recent review of case series published between 2011 and 2020 stated that while diabetes mellitus remains one of the medical conditions most frequently associated with EBE, malignancies and intravenous drug use represent significant risk factors too [11]. Malignancies were thought to be prevalently associated with endogenous mould endophthalmitis, where Aspergillus spp. and Fusarium spp. were the major pathogens involved [1]. However, malignancies have also been found to be a risk factor in the case of Streptococcus spp., Pseudomonas spp., and Candida spp. endogenous endophthalmitis [11]. In our systematic review, while malignancies are well represented, a small percentage of intravenous drug use is reported. This is consistent with the current literature, since the majority of EBE in people who inject drugs are caused by Gram-positive rather than Gram-negative agents [150].
In our study, vitreous opacity and hypopyon were EBE's most described initial lesions. However, eye redness (91, 33.8%) alone or together with other ocular signs was commonly reported. This finding, together with the not always severely compromised visual acuity, enlightens the need for ophthalmologists to maintain high suspicion for EBE in patients with intraocular inflammation and significant medical comorbidities [67]. Patients with EBE usually present acutely, complaining only about decreased vision and eye pain [1]. Systemic complications or more alarming local signs such as hypopyon or vitritis might be absent during the initial evaluation [1].
The treatment of EBE should include both ocular and systemic therapy. This is a pharmacokinetic consequence, since most antimicrobial agents have a poor penetration capacity into the avascular vitreous cavity when parenterally administered [67]. Therefore, intravitreal injections are the treatment of choice for EBE. In line with our findings, the most commonly used antimicrobials for empiric treatment are third generation cephalosporines for Gram-negative microorganisms, followed by amikacin and gentamicin, which were mostly used in combination regimens [151]. The notorious Endophthalmitis Vitrectomy Study, a randomized clinical trial conducted between 1991 and 1994, stated that 89.5% of Gram-negative organisms causing endophthalmitis were susceptible to both amikacin and ceftazidime [152]. Although the emergence of multidrug-resistant bacteria is a global issue, the antibiotic susceptibility patterns of Gram-negative bacteria from vitreous isolates have not significantly changed in the United States [152].
The role of additional steroids in EBE management is controversial. A recent study in the Cochrane Library states that the currently available evidence on the effectiveness of adjunctive steroid therapy versus antibiotics alone in managing acute endophthalmitis after intraocular surgery is inadequate [153]. A combined analysis of a very limited number of studies suggests that adjunctive steroids might provide a higher chance of having a better visual outcome at three months [153]. Moreover, another study shows a higher rate of enucleation/evisceration in patients who did not receive steroid therapy [147]. These controversial findings match the differences in clinical approaches to EBE management and treatment revealed by our systematic review, where just a few patients were treated with steroids in addition to the ongoing antimicrobial regimen.
Adequate source control is often warranted in the case of EBE. Surgical intervention is generally recommended for patients infected with virulent organisms, with bilateral involvement, severe vitreous involvement, and progressive worsening [67]. Our systematic review shows that vitrectomy is the most often used surgical procedure as it helps in removing infectious organisms, toxins, and inflammatory cells from the vitreous cavity, thus leading to a better diffusion of antibiotics and a faster recovery [147]. Vitrectomy has several clinical and diagnostic implications: it might save eyes with EBE and restore vision while also providing a higher diagnostic yield compared to a vitreous biopsy, thus helping identify the causative organism [147].
Prognosis is poor in cases of Gram-negative EBE. Despite aggressive therapy, often necessitating surgical intervention, the poor clinical outcome in the case of EBE might be related to a delay in the diagnosis and treatment or the absence of worldwide shared guidelines [15]. Although several factors are associated with the visual outcome, a central role seems to be played by the pathogen involved. A recent study reported that very poor visual acuity (20/400 or worse) is associated with several Gram-negative pathogens such as H. influenzae (69%), Serratia spp. (70%), and Pseudomonas spp. (92%) [1]. Although not uniformly observed across all studies, it has been hypothesized that Gram-negative EE's poorer outcome could be linked to both Gram-negative endotoxin production and the presence of a phagocytosis-resistant capsule [147].
The findings of this systematic review should be seen in the light of some limitations. First, our research strategy includes a selection bias that cannot be eliminated. Indeed, by including the three main EBE aetiologies, "Klebsiella", "Escherichia coli", and "Pseudomonas" in the initial search, it is subordinate that their prevalence will be found to be higher. However, the selection of these species allowed the authors to consider more papers that would have otherwise been wrongfully excluded. Secondly, our systematic review is based on many case reports and case series, with only one prospective study and no RCTs. Consequently, the inclusion of retrospective studies describing aggregate data makes it hard to select data for each patient individually. Lastly, the heterogeneity of the studies included in the absence of methods to assess the risk of bias or certainty in the body of evidence restricted our review to descriptive analysis. Therefore, we limited our comprehensive analysis to a descriptive evaluation of the past 20 years' literature on Gram-negative EBE. On the other hand, the main strength of this systematic review is the large sample size. In addition, as it was noted during the search phase that most of the scientific output on the subject is produced by ophthalmologists, this systematic review presented Gram-negative EBEs from an infectious diseases specialist's point of view.

Conclusions
Although the literature on EBE mainly comprises case series or single case reports, Gram-negative EBE is an undoubtedly consistent clinical reality associated with poorer outcomes due to virulence and pathogenetic aspects of the Gram-negative's structure. Klebsiella pneumoniae is the most common causative pathogen in Gram-negative EBE, especially in the Asian population or diabetic people. Although in our study vitreous opacity and hypopyon were the most often described initial lesions of EBE, eye redness alone or together with other ocular signs was commonly reported. This enlightens the need for ophthalmologists to maintain high suspicion for EBE in patients with intraocular inflammation and significant medical comorbidities. Our findings underscore the importance of considering ocular involvement in the case of Gram-negative infections. In light of an ageing population and considering the concerning phenomenon of Gram-negative antimicrobial resistance, EBEs' appropriate management remains an open challenge for both ophthalmology and infectious disease specialists.

Conflicts of Interest:
The authors declare no conflict of interest.