Endogenous Endophthalmitis: yield of the diagnostic evaluation

Over the approximately 6.5 years included in this study, 149 patients were diagnosed with endophthalmitis at our institution. Of these, 35 patients (23.5%), including 5 bilateral, were secondary to an endogenous source. The mean age of these 35 patients was 55 ± 18 years old.

The most common identifiable source of infection was intravenous drug abuse (IVDA), observed in 8 patients (22.9%; Table 1). The number of cases with this associated risk factor was noted to increase over the course of the study, with 0 patients in 2011, 2012, and 2013, 1 in 2014 and 2015, 2 in 2016, and 4 in 2017. The most common comorbidity observed in this population was diabetes, reported in 16 patients (45.7%; Table 2).

Our study found that treating physicians performed diagnostic evaluations at their discretion without a common pathway in place. The four patients managed as outpatients did not have any diagnostic imaging obtained. Seven patients did not have imaging obtained as part of their inpatient stay, the remaining 24 patients had x-rays (100%), computed tomography (CT; 87.5%), magnetic resonance imaging (MRI; 25%), or ultrasound (54.2%) during their inpatient stay (Table 3). Percent yield of relevant findings that suggest the source of infection were calculated for each imaging type. Of the 24 patients who had x-rays during their admission, 4.4 ± 5.5 x-rays were obtained per patient, with relevant findings in 2.7%. Of the 21 patients who had CT testing, 1.6 ± 1 were obtained per patient, with a yield of 58.3%. Of the six patients with MRI testing, 1.3 ± 0.5 MRIs were obtained per patient, with a positivity of 58.3%. Of the 12 patients with ultrasound testing, 1.4 ± 1 studies were obtained per patient, with a yield of 8.3%. Ocular ultrasound was not specifically evaluated in our study as this practice was not consistent among the many providers over the course of the years reviewed.

Transthoracic echocardiogram (TTE) was obtained in 26 patients, with 92.3% negative findings. Positive findings were aortic regurgitation in one patient and pericardial effusion in another. Transesophageal echocardiogram (TEE) was then obtained in 18 patients, 4 of which revealed vegetations (22%). Two lumbar punctures were obtained, both with negative cultures.

In 35 eyes, intraocular cultures were obtained, with 10 positive samples (28.6%). If anti-infectious agents were initiated prior to intraocular culture, no culture returned as positive. With the patients who received prior antibiotics excluded, 41.7% had positive intraocular fluid cultures. When the diagnosis of endophthalmitis was made on the day of admission, the culture positivity rate was 46.7%, higher than the positivity rate when diagnosed after admission (15%; P = 0.040). Intraocular culture positivity was not significantly associated with hospital stay (12.4 ± 6.5 days for positive culture, 18.8 ± 13.8 days for negative culture; P = 0.11). The most common bacteria identified was coagulase-negative Staphylococcus (CoNS), while the most common yeast was Candida species; these pathogens were equally common (n = 3; Table 4).

Blood cultures were obtained in 33 patients, with 16 positive samples (48.5%). Methicillin-resistant Staphylococcus aureus (MRSA) was most common organism identified (Table 4). No difference in blood culture positivity was seen when diagnosis of endophthalmitis was made before or after admission (P = 0.58), and blood culture positivity was not significantly associated with length of inpatient hospital stay (P = 0.3) or duration of antibiotics (41.3 ± 19.3 days positive vs. 30.5 ± 16.2 days negative cultures; P = 0.1).

Urine culture was obtained in 15 patients, with 2 positive (13.3%). Further cultures were obtained in 11 patients; skin and abscess cultures had the highest yields (Table 4).

Most eyes (85%) were treated with a sample of intraocular fluid by needle aspirate, intravitreal injection of antibiotics, or both. Eight of these were treated with vitrectomy surgery later. Five patients were treated primarily with vitrectomy, two of these requiring a second delayed vitrectomy. The decision for vitrectomy was made by the treating vitreoretinal surgeon. There was no documented visual acuity indication for vitrectomy.

Four patients died since last follow up. One died of septic shock during the admission when endophthalmitis was diagnosed. One was discharged to hospice. One died of an arrhythmia from hyperkalemia during hemodialysis, temporally unrelated to the endophthalmitis, though the dialysis access port was considered the source of the infection. One died of unknown causes outside of our hospital system.

In the literature, rates of endogenous endophthalmitis among all causes are reported as 7.7–13.2% [9‐11]. At our center, endogenous endophthalmitis accounted for 23.5% of all endophthalmitis cases. We cannot definitively explain the higher rate seen at our center, but we attribute the large tertiary care hospital receiving many referrals throughout the region to this finding, which may reflect selection bias.

As seen in our study, IVDA is a common source of entry for pathogens into the bloodstream, accounting for 25 to 53% of endogenous endophthalmitis at some centers [12]. IVDA-associated endophthalmitis was noted with increasing frequency over the course of this study, likely correlating to the rise of IVDA in the community [13]. Common to our study, diabetes and cancer have been identified as common comorbid factors in other studies [2, 14]. In Asian countries, liver abscess is more commonly reported as the source of infection [6, 10].

We, like other studies [1], consider length of inpatient admission as a marker of the severity of infection or systemic health, as patients with a well-controlled systemic infection may be discharged to complete an antibiotic course as an outpatient. In this study, the length of systemic antibiotics (34.1 ± 18.7 days) was double the length of the average inpatient admission (16.1 ± 11.9 days). In our series, diagnosis of endophthalmitis on admission was associated with a shorter hospital stay, perhaps because appropriate treatment was able to be initiated more rapidly or because of less complex medical management required.

Intraocular culture positivity was overall 28.6%, but 41.7% when systemic anti-infective agents had not yet been administered. Likewise, when endophthalmitis was diagnosed on admission, the culture positivity rate was higher than when diagnosis was made later during hospital admission. Given the low rate of culture positivity (0%) when anti-microbial drugs had already been administered, ophthalmologists may wish to defer obtaining a vitreous aspirate, as the risk of may outweigh the benefits of identification of the infectious pathogen when the yield is poor.

Intraocular culture positivity is often low in endogenous endophthalmitis cases, reported at 14–43% in the literature of other centers [2, 4, 6, 15‐17]. Given this low diagnostic yield, endophthalmitis must often be treated empirically, such as the initiation of antifungals based on a typical fundus appearance [15]. Primary vitrectomy or polymerase chain reaction (PCR) of the vitreous aspirate may be used to increase diagnostic yield [18].

Blood culture positivity was 48.5%, similar to what has been reported elsewhere [4, 16]. As in other studies from Western centers [4], Gram-positive organisms were the most commonly identified pathogen. In some centers, fungus is frequently identified in endogenous endophthalmitis [14, 19], accounting for approximately half of positive cultures, Candida being the most common pathogen [18, 20]. Fungal infections may have better prognosis than bacterial [18, 21]; however, this outcome is not consistent across the literature [22]. Candida may be associated with IVDA [20]. Gram-negative organisms were less commonly identified in this series but are more prevalent in East Asian studies and may confer a worse prognosis [6].

In this series, CT and MRI had the highest diagnostic yield for the source or other sites of end-organ infection. Positive findings analysis of CT frequently showed pneumonia (36.8%), septic inflammation of a space/joint (31.6%), and eye or orbit findings (21%). MRI positive findings frequently found tissue inflammation. These included CNS vasculitis or chorioretinitis,(60%), and septic joint inflammation (40%).

Using these studies in a targeted fashion may thus be the preferred modality for imaging evaluation. X-ray and ultrasound had a poor yield. In some cases however, multiple images of the same modality were obtained for correct implant placement (e.g. peripherally inserted central catheter (PICC) line) or for monitoring of treatment effectiveness (e.g. pleural effusions). The acquisition of multiple images particularly in the case of X-rays may reasonably change the interpretation of diagnostic yield as the radiology read may have been influenced by the intent of the imaging study.

While the utility of intravitreal antibiotics in the management of endogenous endophthalmitis is controversial [2], most (95%) of the patients in this study received them, with multiple injections given in 52%. Vancomycin and ceftazidime were the most common intraocular antibiotics given in this study, consistent with the literature conclusions based on intraocular safety and efficacy of the most common agents [2].

No clear consensus exists on indication and timing for vitrectomy in the management of endogenous endophthalmitis. Vitrectomy may be protective to vision and preservation of the eye [2, 20]; however, no significant difference in vision outcomes was seen in this study, and our rate of enucleation was 5% (n = 2). Vitrectomy may increase diagnostic culture yield, especially if performed prior to systemic antibiotics; however, in our population, as in others in the literature [12], the sample size is too small to draw strong conclusions. Systemic disease and risks of anesthesia may prevent surgeons from pursuing vitrectomy.

Two patients in this series were treated with enucleation, accounting for 5% of the total eyes included, compared with 25% in a review [2] from 1976 to 2003. More recent studies [4, 14, 20] report enucleation rate at 8–12%. In this study, however, the poor follow up may limit conclusion regarding globe preservation, as this study included five eyes with no light perception at last exam but never followed up with our outpatient ophthalmology service and may have been enucleated elsewhere. Given the low rate of follow up seen in our series, we caution against premature discharge when the ocular infection is still poorly controlled.

As with all retrospective case series, this study limited by the inability to determine causation and prevent bias. This study is also limited by sample size; however, the number of eyes (40) included in this study is typical of similar single-centered reports in the literature [18, 20]. Our analysis of patient outcomes, especially regarding vision, is limited by the poor follow up to the ophthalmology clinic. Although social issues, re-location, or improvement in condition are all reasons for loss to follow up, we could not identify the specific reasons in our study.