Endogenous Endophthalmitis at a Tertiary Referral Center in China: A Retrospective Study Over Three Decades

Purpose To investigate the etiology, pathogens, treatment, and prognosis of endogenous endophthalmitis (EE). Methods Patients diagnosed with EE over three decades at Peking Union Medical College Hospital were retrospectively reviewed and analyzed. Clinical characteristics with different causes were evaluated, and a subgroup analysis of different initial treatment methods was conducted. Results A total of 97 eyes from 81 patients were included. Diabetes mellitus (DM) was the most common predisposing medical condition (34.6%). Klebsiella pneumoniae (31.3%) and Candida albicans (32.8%) were the most common pathogens. Liver abscess (20.6%) was the predominant cause of EE. More than 70% of patients with liver abscesses had DM, which was signi�cantly more than other etiologies (P < 0.05). EE due to liver abscess had a worse initial visual acuity (P < 0.05). Surgical interventions for EE due to liver abscess were signi�cantly shorter than those for other causes (P < 0.05), but the prognosis was poorer. Forty percent of eyes underwent evisceration/enucleation (P < 0.05). Patients who initially underwent pars plana vitrectomy (PPV) + silicone oil tamponade underwent fewer total treatments, and no eyes required additional intravitreal injections (P < 0.05).


Introduction
Endogenous endophthalmitis (EE) is a vision-threatening eye infection caused by hematogenous bacteremia or fungemia spread to the eye.Bacteria and fungi break through the blood-eye barrier and cause severe eye in ammation, leading to irreversible vision loss and eye discomfort. 1,2EE is a relatively rare entity, with an incidence of 0.04-0.4% and accounting for 2-15% of all endophthalmitis cases.It is often associated with diseases such as diabetes mellitus (DM), liver disease, malignancy, indwelling catheters, and intravenous drug abuse. 3,4ven that EE is relatively rare, most reports have presented clinical features of a single case or small case series. 2,4Single-center studies with large samples and long duration are even more limited, particularly in China.[7][8][9][10] The vision prognosis is generally considered to be poor, and no standardized treatment plan is currently available. 1,3,11 Wummarized the patients of EE in Peking Union Medical College Hospital in the past 30 years.This study aimed to investigate the etiology, pathogens, treatment, and prognosis of EE.The clinical characteristics were evaluated, and the prognosis under different circumstances was discussed to provide a reference for ophthalmologists.

Methods
We retrospectively reviewed all patients with EE admitted to Peking Union Medical College Hospital from January 1990 to October 2020 and collected the demographic characteristics, predisposing conditions, extraocular infection resources, microbiological results, treatment outcomes, initial and nal best-corrected visual acuity (BCVA), and serious complications after treatment.This retrospective study adhered to the principles of the Declaration of Helsinki of the World Medical Association.It was approved by the Institutional Review Board/Ethics Committee of Peking Union Medical College Hospital.Written informed consent was obtained from each patient prior to surgery.The diagnosis of EE was determined by clinical signs and symptoms (including eye pain, loss of vision, eyelid edema, conjunctival congestion, chemosis, anterior chamber in ammation, vitritis, decreased red re ex, and vitreous opacity on B-scan ultrasonography), sources of extraocular infection, and/or positive microbial culture for any specimen, including aqueous, vitreous, blood, or histopathology.Patients with positive intraoperative smear results were also included for those with no clear extraocular source.
Patients who had undergone ocular surgery within 1 year, had penetrating eye trauma, or had corneal ulcers were excluded.Patients with incomplete medical charts were also excluded.
Endophthalmitis treatment was determined by individual physicians.The Snellen BCVA was converted to the logarithm of the minimum angle of resolution (logMAR) equivalents.Visual acuity (logMAR) gain of greater than or equal to 0.2 (equivalent to 10 ETDRS letter gain) was considered as an improvement.No light perception (NLP), light perception (LP), hand motion, and counting ngers (CF) were assigned a logMAR value of 2.90, 2.60, 2.30, and 2.0, respectively. 12o retina specialists (W.F. Z. and X. Y. Z.) collected and evaluated the data.The mean ± SD, counts, and percentages were used to evaluate continuous and categorical variables.Continuous variables were analyzed using an independent t-test, paired t-test, or nonparametric test (if not normally distributed).The chi-square test or Fisher's exact test was used to analyze categorical variables.Binary logistic regression analysis was conducted to analyze the prognostic factors associated with the nal visual acuity (VA).SPSS 26.0 (SPSS Inc., Chicago, Illinois, USA) was used for statistical analysis.P < 0.05 was considered statistically signi cant.
Invasive Aspergillus/Candida infection was found in four immunosuppressed patients.The proportions of sources of pneumonia (seven eyes from six patients), skin and soft tissue (eight eyes from seven patients), and post-surgery (four eyes from four patients) were similar, accounting for 7.2%, 8.2%, and 4.1% of cases, respectively.Other sources of infection included brain abscesses (2.1%) and nasopharyngeal carcinoma (1.0%).The mean follow-up time was 261.01 ± 481.39 days (range 7-2550, median 50) (Table 1).Immunosuppression included long-term use of hormonal, immunosuppressants, or ongoing chemoradiotherapy.

EE with different causes
Based on infection sources and pathogen distribution, patients were divided into three groups.Patients in group 1 had liver abscesses.Group 2 comprised patients with genitourinary system infection and immunosuppression/post-surgery infection.Group 3 included patients with other and unclear causes.The clinical characteristics of EE caused by different etiologies were summarized in Table 3.Of the patients, 32.1% were identi ed from hospital consultations and 7.4% had shock or multiple organ dysfunction syndrome.More than 70% of patients with liver abscesses had DM, the proportion of which was signi cantly higher than that of other etiologies (P < 0.05).
Seven (46.7%) patients with liver abscesses also had lung infection, two (13.3%) had brain abscesses, and one (6.7%)had skin softtissue infection.This group also had worse initial logMAR BCVA (P < 0.05) than group 2 and the general data, with 90% of eyes having a VA below CF (P < 0.05).The number of eyes with retinal detachment (RD) in group 1 at initial presentation was more than that in group 3. EE due to liver abscesses was overwhelmingly caused by bacteria, particularly K. pneumoniae.However, genitourinary system infection, immunosuppression, and post-surgery infection were mostly caused by fungi, particularly C. albicans (P < 0.05).Although surgical interventions for EE due to liver abscess were signi cantly shorter than those of general data and for other causes (P < 0.05), the prognosis was poor.Fifty-ve percent of patients experienced serious complications, and the proportion of evisceration/enucleation was 40%, which was signi cantly higher than that in the other groups (P < 0.05).Six (30%) eyes underwent evisceration/enucleation at presentation.The initial logMAR BCVA of all patients was 2.15 ± 0.71, and the nal logMAR BCVA was 1.99 ± 0.97 (P = 0.263).After treatment, VA improved in 30.9% of patients, and no signi cant difference was noted in the nal VA among these groups.Final VA signi cantly improved from 2.19 ± 0.68 to 1.84 ± 0.99 in group 3 (P < 0.05).However, no signi cant improvement in the nal VA was found in groups 1 and 2 compared with the initial VA (P = 0.203 and 0.190, respectively).

Subgroup analysis according to initial treatment
Except for patients who initially received evisceration/enucleation and those who refused any surgical treatment, 83 eyes were divided into three groups according to their initial treatment.The three groups were pars plana vitrectomy (PPV) + silicone oil tamponade, PPV, and intravitreal injection of antibiotics (IVI) only.IVI was performed every time PPV/PPV + silicone oil tamponade was done.The IVI dose was reduced to 1/2 or 1/3 if tamponade was used.All the patients were systematically treated with antibiotics.Signi cant difference was not found in the initial logMAR BCVA.In the PPV group, two eyes had exudative RD at initial presentation and did not receive silicone tamponade.More eyes (14.3%) underwent evisceration/enucleation in the IVI group than in the PPV + silicone oil group.Patients who initially received PPV + silicone oil tamponade underwent signi cantly fewer total treatments, and none of the eyes required additional intravitreal injections (P < 0.05) (Fig. 2).More than 80% of eyes required two or more surgical interventions in the PPV and IVI-only groups, which was signi cantly higher (P < 0.05).In the subsequent treatment of the PPV group, six eyes (35.3%)only additional 2.67 ± 1.37 injections, three eyes (17.6%) received only additional 2.33 ± 0.47 PPV, and ve eyes (29.4%) received not only additional 3.60 ± 1.97 injections but also additional 2.20 ± 0.40 PPV.Of the eyes treated with additional PPV, three eyes received concurrent silicone oil tamponade.In the IVI group, after 1.64 ± 1.04 injections, 22 eyes (52.4%) received 1.14 ± 0.34 PPV + silicone oil tamponade and six eyes (14.3%) subsequently received PPV (two of six eyes underwent additional IVI after PPV).Eleven eyes (26.2%) received only 2.21 ± 1.52 IVI, among which six patients (eight eyes) were too ill to undergo PPV intervention.No signi cant differences were found in the nal VA after treatment among the three groups.No signi cant improvement was observed in the nal VA compared with the initial VA in the three groups (P = 0.716, 0.182, and 0.424, respectively).

Discussion
Monocentric studies on EE with large sample sizes and long duration are limited, particularly in China.In this retrospective study, we summarized the clinical features, treatment, and prognosis of EE in Peking Medical College Hospital, a large tertiary referral center in China.The most common predisposing medical condition was DM, followed by autoimmune diseases and hepatitis/liver cirrhosis.
The positive rate of vitreous/aqueous cultures was approximately 65%, and the most common bacteria and fungi were K. pneumoniae and C. albicans, respectively.Approximately 30% of patients were identi ed from hospital consultations.After treatment, VA did not improve signi cantly.EE as a result of liver abscess was overwhelmingly caused by bacteria, particularly K. pneumoniae.Both the initial condition and prognosis were the worst despite more aggressive interventions being administered.The majority of EE cases as a result of genitourinary system infection and immunosuppression was caused by a fungal infection, and VA improvement was not signi cant after the intervention.For different initial treatment measures, the number of total treatments and IVI was signi cantly reduced for eyes that initially received PPV + silicone oil tamponade.Fewer eyes required secondary or more treatments although no signi cant bene t was observed for visual prognosis.
15][16] In our study, liver abscesses were the leading cause of de nite extraocular infection sources, and more than 70% of patients with liver abscesses had DM.The initial VA was poorer; although the intervention was clearly more aggressive, the prognosis was signi cantly worse than that in the other patients.Forty percent of eyes eventually underwent enucleation/enucleation. Chemotaxis of polymorphonuclear leukocytes was speculated to be impaired in patients with DM.This abnormal chemotactic and phagocytic function of neutrophils could lead to an increased susceptibility to infection and may cause sepsis and bacteremia, especially hepatobiliary infections. 17Suspecting EE when patients with DM develop ocular in ammation is important.Liver abscess has been proven to be the major source of extraocular infection in east Asia, accounting for 19-57.5% of cases.Among them, K. pneumoniae-related bacteremia and endophthalmitis occupy the dominant position. 13,18 hen et al. showed the same trend that K. pneumoniae-related EE has poor visual effects and often requires enucleation. 2,5,19 Te main reason is that K. pneumoniae tends to develop extensive subretinal abscesses early on, destroys retinal tissue, and subsequently causes a loss of central vision.Moreover, because this damage is hidden behind an opaque vitreous, the extent of retinal necrosis is easily overlooked. 20,21 gal infections, particularly C. albicans, accounted for a large proportion of cases in our study.The majority of EE cases following urogenital infections and immunosuppression were caused by fungal infections (52.6% vs. 15.8%).Compared with bacterial endophthalmitis in liver abscesses, the treatment was delayed, and the incidence of evisceration/extraction was 7.9%.No signi cant improvement was found in VA after treatment.Candida is a part of the human ora and exists as symbiotes on the mucosal surfaces of the respiratory, gastrointestinal, and female reproductive tracts.The risk factors for candidiasis and endogenous fungal endophthalmitis are mainly associated with the suppression of immune mechanisms or procedures of the genitourinary system, which may increase the risk of blood-borne infection. 22,23 . albicans-related EE may also occur after delivery or abortion, presumably as a complication of transient candidiasis. 24,25 his nding emphasizes the suspicion of fungal endophthalmitis, especially in patients with a corresponding history.
Vitrectomy for EE treatment remains controversial because of the lack of large randomized controlled studies, but many studies have suggested that vitrectomy may be an effective approach to diagnose and treat EE. 11 In our study, the positive rate of vitreous/aqueous humor cultures was approximately twice that of blood cultures (65.7% vs. 31.3%).Compared with anterior chamber tap or vitreous tap, Chen and Weishaar PD et al showed that vitrectomy is more likely to produce positive culture results, particularly for endogenous fungal endophthalmitis. 26,27 ince endogenous fungal endophthalmitis usually begins in the choroid, it is believed that a vitreous tap may not adequately sample the vitreous cavity, particularly in mold infections.Meanwhile, vitreous specimens after vitrectomy are also subjected to microbiological testing, especially Gram staining, to quickly identify bacterial or fungal causes to aid in the diagnosis. 28In our study, eyes that initially received oil tamponade underwent signi cantly fewer treatments and did not require additional intravitreal injections.Approximately 50% of eyes in the initial IVI group also received PPV + silicone oil tamponade due to poor disease control, and more eyes underwent evisceration/enucleation, which demonstrated importance of vitrectomy and silicone oil packing in EE treatment.In the initial PPV group, more intravitreal injections and additional vitrectomy were required to control endophthalmitis progression despite initial vitrectomy.Vitrectomy is generally speculated to be bene cial for reducing microbial load, toxins, and in ammatory factors; promoting antibiotic diffusion; increasing vitreous transparency; and controlling intraocular pressure.At the same time, with the progress in surgical technology, the probability of iatrogenic injury is greatly reduced. 1,2 e hypothesized that the reason why the initial PPV group required more treatments than the PPV + silicone oil group was that the choroid and ciliary body were initially involved in EE due to higher blood ow, followed by the retina and vitreous. 1With PPV alone, in ammatory factors, cell debris, and other toxins may still damage the retina and be released into the vitreous cavity.In addition, silicone oil has been shown to inhibit bacterial proliferation.It is highly hydrophobic and has high interfacial tension, which could limit the movement of pathogens and reduce their activity.Meanwhile, its tamponade effect could reduce the probability of RD. 29 Previous reports on the effects of vitrectomy on visual outcomes have mixed results.Some studies suggested that PPV can help preserve useful vision. 30However, other studies showed that PPV is not associated with visual prognosis. 5,21 ifferences among various studies were di cult to compare because of time, etiology, and pathogen composition.In our study, no additional visual gain was observed in the initial PPV/PPV + silicone oil tamponade group.Overall, the visual prognosis of EE was poor.
This study has several limitations.First, some biases, such as treatment selection, cannot be ignored because of the retrospective nature of the study.Further prospective and randomized studies are warranted to address treatment selection in more detail.Second, different surgeons might have affected the surgical outcomes.Finally, various follow-up durations might have an effect on nal vision.
In conclusion, EE is a devastating intraocular disease with poor visual prognosis.Although more aggressive interventions were administered, the initial condition and prognosis of EE caused by liver abscess were the worst.PPV + silicone oil tamponade as an initial treatment has a higher single-procedure success rate than other treatments and may reduce the need for additional therapy.The Distribution of Causative Organisms with Different Causes.

Abbreviation List
Figure 2 Figures

Table 1
Predisposing Conditions and Infection Sources of Patients with Endogenous Endophthalmitis.

Table 2
Causative Organisms from Intraocular or Extraocular Samples.

Table 3
Clinical Characteristics of Endogenous Endophthalmitis With Different Causes.

Table 4 Subgroup
Analysis according to Initial Treatments.