Combey de Lambert A, Campolmi N, Cornut P, Aptel F, Creuzot-Garcher C, Chiquet C, for the French Institutional Endophthalmitis Study Group. Baseline Factors Predictive of Visual Prognosis in Acute Postoperative Bacterial Endophthalmitis in Patients Undergoing Cataract Surgery. JAMA Ophthalmol. 2013;131(9):1159-1166. doi:10.1001/jamaophthalmol.2013.4242
Although rare, postoperative endophthalmitis in patients undergoing cataract surgery can lead to anatomical or functional loss of the eye. Therapeutic strategies such as antibiotic prophylaxis and microbiological diagnosis are more effective with a target patient population. New prospective data are needed to identify prognostic factors.
To identify baseline factors of visual prognosis in patients with acute bacterial endophthalmitis after cataract surgery.
Prospective study of consecutive patients undergoing cataract surgery, enrolled from March 1, 2004, through December 31, 2005. We analyzed outcomes to determine the effect on the final visual outcome, defined as poor (visual acuity [VA] worse than 20/100) or good (VA 20/40 or better) using univariate and multivariate analysis.
Four academic hospitals.
Ninety-nine consecutive patients with cataract.
Main Outcomes and Measures
Factors related to the cataract surgery (complications), initial clinical presentation, and microbiological diagnosis and the final VA.
The significant baseline factors (at presentation) for good visual outcome (45% of the series) were the winter season, absence of complications during cataract surgery, initial VA, microbiological investigations revealing no microorganism or a coagulase-negative Staphylococcus species (CNSP), and fundus visibility. Quantitative factors associated with a good clinical prognosis were shorter duration of cataract surgery, younger age, and a hypopyon no greater than 1.5 mm. Significant factors associated with poor visual outcome were infection of the right eye, initial VA, corneal edema, a hypopyon larger than 1.5 mm, detection of bacterial species other than a CNSP, and the absence of fundus visibility. Multiple logistic regression analysis showed that high bacterial virulence was the only independent factor (odds ratio, 14.0 [95% CI, 2.7-71.0]; P = .001) for poor visual outcome. On the other hand, low bacterial virulence (odds ratio, 0.2 [95% CI, 0.03-0.6]; P = .01) and the absence of complications during cataract surgery (0.1 [0.01-0.4]; P = .003) were independent factors for good VA.
Conclusions and Relevance
Most clinical outcome factors in acute postoperative endophthalmitis can be identified at presentation. The bacterial virulence level is the main factor predictive of the final visual prognosis.
Postoperative endophthalmitis in patients undergoing cataract surgery is a rare (approximately 0.08% of cataract procedures)1,2 but serious complication that can lead to anatomical and/or functional loss of the eye. Significant advances have been achieved in the field of antibiotic prophylaxis (intracameral injection of cefuroxime sodium, 10 mg/mL)3- 5 and microbiological diagnosis.6- 9 The identification of baseline prognostic factors at the time of modern clear corneal cataract surgery remains an important step to identify patients who need aggressive therapy as early as possible. Polymerase chain reaction (PCR)–based techniques now provide accurate and rapid identification of the microorganisms directly involved in intraocular samples. These methods will make rapid adaptation of the therapeutic strategy to clinical and microbiological factors possible and give more objective information to the patient.
The anatomical and functional prognoses have been reported to be associated with baseline factors, such as initial vision,10- 12 absence of a red reflex, history of diabetes mellitus or glaucoma, and the bacterial species involved.10,12- 17 These data were mainly obtained 15 years ago on a selected population undergoing randomization for pars plana vitrectomy (PPV) in the Endophthalmitis Vitrectomy Study (EVS), with inclusion of patients with a clear enough cornea to allow vitrectomy and exclusion of patients who did not consent to therapeutic randomization (with visual acuity [VA] of 20/100 or better).10 Given the changes in cataract surgery techniques and different epidemiological characteristics of the population studied, new prospective data are needed for the identification of prognostic factors.
The French Institutional Endophthalmitis Study (FRIENDS) group included 100 consecutive patients undergoing cataract surgery who developed acute postoperative endophthalmitis, which allowed the prospective study of the largest series after the EVS. Previous microbiological results, using conventional cultures and panbacterial PCR analysis, have been previously published.6,18- 21 The aim of this report was to identify baseline clinical and microbiological factors of good (VA of 20/40 or better) and poor (VA of worse than 20/100) visual outcomes at 6 months in a large series of patients with acute postoperative endophthalmitis.
This prospective study included 100 patients (100 eyes) with acute (<6 weeks) postoperative endophthalmitis, enrolled from March 1, 2004, through December 31, 2005, in 4 French academic hospitals (in Dijon, Grenoble, Lyon, and Saint-Etienne). The study followed the Declaration of Helsinki guidelines for research involving human subjects and was approved by the local institutional review board (Comité de Protection des Personnes Lyon B). The inclusion criteria, data sheet, sampling techniques, microbiological procedures, antibiotics used for intravitreal injections, and follow-up visits were standardized in the 4 centers.
Patients were eligible if they had clinical signs of endophthalmitis within 6 weeks after cataract surgery. Acute postoperative endophthalmitis was diagnosed on the basis of clinical features, including pain, decreased VA, diffuse bulbar conjunctival hyperemia, chemosis, inflammation of the anterior segment (anterior chamber reaction including 1 of the following criteria: flare, cells, hypopyon height, fibrin, or cyclitic membrane), and posterior segment inflammation (all patients had vitreous infiltration diagnosed with biomicroscopy or ophthalmic ultrasonography). Exclusion criteria were endophthalmitis secondary to other surgical procedures, toxic anterior segment syndrome (n = 6), and loss to or lack of availability for follow-up (<6 months [n = 0]). One patient was excluded from the analysis given his age (2 years) and the inability to obtain a reliable postoperative vision measurement.
A baseline evaluation form was completed during the initial examination of each patient and included demographic features, medical history, ocular history, details of cataract surgery (retrospectively collected from the cataract surgeon, including capsular rupture and vitreous loss), and data of the initial ocular examination. A complete list of these features is given in Table 1.
Performance of PPV was left to the discretion of the treating physician. During the 6-month follow-up, patients were defined as having a good (final VA, ≤0.3 logMAR, or 20/40 or better) or poor (final VA, >0.7 logMAR, or worse than 20/100) visual prognosis. Incidence of phthisis was also recorded at 6 months.
On admission, an immediate tap of aqueous humor and/or vitreous was performed followed by intravitreal injection of vancomycin hydrochloride (1 mg per 0.1 mL) and ceftazidime sodium (2 mg per 0.1 mL) without corticosteroid. Samples in all eyes were collected after instillation of aqueous povidone iodine solution, 5%, in the conjunctival sac. After installing the lid speculum, a new instillation of povidone iodine solution was given. Then the conjunctival sac was washed with 20 mL of sterile balanced salt solution before sampling. Aqueous humor samples (200 μL) or vitreous samples (300 μL) were collected in a sterile syringe just before the first intravitreal injection. Undiluted vitreous samples (500 μL) were also collected during the PPV. Conventional cultures were performed by direct inoculation of intraocular samples in brain heart infusion broth (BioMérieux). These culture media were incubated for 2 weeks at 37°C and subcultured on chocolate agar supplemented with factors X (hemin) and V (NAD) (PolyViteX; BioMérieux) and Columbia agar supplemented with 5% sheep blood (Biomérieux) when bacterial growth was detected or systematically after 2 weeks of incubation if no growth was detected. For cultures with positive yields, Gram staining was performed, and bacterial identifications and antibiograms were obtained using phenotypic identification systems (Vitek II [BioMérieux] or BD Phoenix [Becton Dickinson]). Species identification was confirmed by 16S ribosomal DNA (rDNA) amplification and sequencing.
Bacterial DNA extraction, 16S rDNA PCR amplification, and sequencing were performed as previously described.6 Panbacterial PCR was able to detect 500 colony-forming units of Staphylococcus epidermidis. Amplification of the human betaglobulin gene served as an internal positive extraction and amplification control. The 16S rDNA sequences obtained were compared with those available in the GenBank, European Molecular Biology Laboratory, and DNA DataBank of Japan databases with the Bio Informatic Bacterial Identification database (http://umr5558-sud-str1.univ-lyon1.fr/lebibi/lebibi.cgi) program. Identification to the species level was defined as a 99% or greater 16S rDNA sequence similarity to that of the GenBank prototype strain sequence. Identification to the genus level was defined as a 97% or greater 16S rDNA sequence similarity to that of the GenBank prototype strain sequence. A failure to identify was defined as a less than 97% 16S rDNA sequence similarity to sequences deposited in GenBank at the time of the analysis.
Patients were then further classified into the following groups according to microbiological results: detection of no microorganism, coagulase-negative Staphylococci species (CNSP), or a more virulent species, including Streptococcus species, Staphylococcusaureus, Staphylococcus lugdunensis, and gram-negative bacteria.18
Quantitative data were expressed as mean (SD). Associations between prognosis and clinical data were primarily studied using analysis of variance for quantitative data and χ2 analysis (incorporating the Yates correction when appropriate) for qualitative data. To determine which of these variables were combined with clinical factors to predict visual prognosis, logistic regression models were fitted using a backward stepwise procedure. We calculated the odds ratio (OR), which provides an estimate of the relative risk of good or poor prognosis. Multivariate logistic regression analysis was performed using clinical factors that were found to be statistically significant in univariate analysis (P < .10). The statistical analysis was performed using commercially available software (SPSS, version 17.0 for Windows; SPSS, Inc). The tests were 2-tailed, and statistical significance was set at P < .05.
We included 99 patients who underwent cataract surgery, with a mean (SD) age of 73 (9) (range, 41-96) years and a diagnosis of acute (<6 weeks) postoperative endophthalmitis. Patients were hospitalized with a median delay of 6 (range, 1-38) days after cataract surgery (clear corneal phacoemulsification in 95 patients and manual extracapsular extraction in 4) associated with posterior (n = 89) or anterior (n = 6) lens implantation or with no intraocular lens implantation (n = 4). The median time between surgery and the onset of symptoms was 5 (range, 1-35) days. No intracameral antibiotics were used at the time of cataract surgery. Intraoperative complications (capsular rupture with or without vitreous loss) were reported in 20 cases (20%), including 12 eyes that had undergone an anterior vitrectomy during cataract surgery. Other systemic and ocular risk factors of endophthalmitis were immunosuppression (n = 7), diabetes mellitus (n = 13), use of oral corticosteroids (n = 7), atopic dermatitis (n = 1), blepharitis (n = 1), dry-eye syndrome (n = 2), and contact lens wear (n = 1).
Thirty-one patients (31%) had a final VA of worse than 20/100, including 10 patients with light perception only, and 45 (45%) had a final VA of 20/40 or better. Finally, 24 patients (24%) had a final VA ranging from 20/100 to 20/50. Main complications noted during the follow-up were retinal detachment (7 [7%]), choroidal detachment (3 [3%]), epiretinal membrane (8 [8%]), and macular edema (6 [6%]). Phthisis developed in 8 cases (8%).
The significant initial factors of good visual outcome were the winter season, absence of complications during cataract surgery, initial VA, detection of CNSP, absence of microorganism detection, and fundus visibility (Table 1). Quantitative factors associated with good clinical prognosis were a shorter duration of cataract surgery, younger age, and a smaller hypopyon. Significant factors associated with poor visual outcome were infection of the right eye, initial poor VA, corneal edema, hypopyon larger than 1.5 mm, detection of a bacterial species other than CNSP, and the absence of fundus visibility (Table 2).
Considering treatment, the number of intravitreal injections was not significantly associated with the visual outcome. Pars plana vitrectomy was more frequently used in cases of poor visual outcome and was significantly associated with earlier presentation (P = .03), detection of a virulent species (P = .008), initial VA (P < .001), and loss of red reflex (P = .01). Pars plana vitrectomy was performed in 29 of the patients with initial VA limited to light perception (91%), 26 of the patients with VA of hand motions (57%), 34 of the patients with VA of counting fingers (38%), and 4 of the patients with VA of 20/400 or worse (28%).
Multiple logistic regression analysis (Table 3) showed that high bacterial virulence was the only independent factor (OR, 14.0 [95% CI, 2.7-71.0]) for poor visual outcome. On the other hand, low bacterial virulence (OR, 0.2 [95% CI, 0.03-0.6]) and absence of complications during cataract surgery (0.1 [0.01-0.4]) were independent factors of good visual acuity (Table 4). Other factors, significantly identified using univariate analysis, were not independent factors and included diabetes, VA, corneal edema, hypopyon, fundus visibility, and PPV.
This prospective study analyzing baseline prognostic clinical and microbiological factors shows that bacterial virulence is the main independent factor of visual outcome after acute postoperative endophthalmitis. One additional independent factor of good visual outcome is the absence of complications during cataract surgery.
Most studies considering prognostic factors after acute postoperative endophthalmitis10- 16 were retrospective. The prospective design is necessary to standardize collection of baseline clinical data and microbiological methods. Our study was based on previously validated microbiological techniques.6,18,19
Our prospective study differed from the EVS by less restrictive inclusion criteria because it was not designed to randomize patients for treatment (in the EVS, inclusion of patients with cornea clear enough to allow PPV and exclusion of patients who did not consent to therapeutic randomization and with VA of 20/100 or better).10 Other differences should also be taken into account, such as the number of patients studied (99 patients in FRIENDS vs 420 in the EVS), the number of centers (4 vs 24), the duration of the study (1.5 vs 4 years), the study period (2004-2005 vs 1990-1994), and the type of surgery (phacoemulsification rate, scleral or corneal incisions, or secondary implantation). Some strengths of our study include analysis of cases with positive culture yields,15 microbiologically unproven cases (n = 30), and cases diagnosed using PCR only (25 in this series). Complications at the time of cataract surgery were also included.
The final visual outcome data reported in this recent series showed that 31% of patients had a poor visual outcome (VA, worse than 20/100), whereas 45% had a good visual outcome (VA, 20/40 or better). These results are similar to those described in the EVS (53% with a VA of 20/40 or better) and other reports (Table 5). Defining prognostic factors of poor visual outcome at admission may be useful for clinicians because these patients would probably benefit from more aggressive treatment. The initial VA level has previously shown a good correlation with the visual prognosis10,11,13,15,16,33 and was one of the most important prognostic factors in the EVS (OR, 2.0), in addition to diabetes mellitus (1.6), corneal infiltrate and/or ring ulcer (1.7), posterior capsular rupture (1.9), and rubeosis iridis (1.8). Other previously reported significant factors, such as the absence of fundus visibility10 and the absence of red reflex,10 were also identified in the present series. Diabetes mellitus is a systemic condition that potentially has an effect on the severity of endophthalmitis, especially with a higher gram-positive and CNSP infection identification rate.10 The FRIENDS group18 previously showed that eyes infected by Streptococcus species were more frequently noted in patients with diabetes mellitus. The frequency of diabetes mellitus in the present study was 13%, nearly identical to that in the EVS patients (14%).10 A gram-positive microorganism was always identified in this situation, and in more than half the cases (7 of 13), the microorganism belonged to the Streptococcus species. We found a trend toward a higher frequency of diabetes mellitus in the group with poor VA. One limitation of all reported studies is the absence of information concerning the severity and treatment of diabetes mellitus.
The season could also play a role in the severity of infection and therefore in the prognosis. Variations in the bacteria spectrum and virulence may explain the seasonal variation in the severity of postoperative endophthalmitis. A previous study18 showed that Streptococcus species, uniformly frequent (30%-36%) from winter to summer, were less frequent in autumn. These data should be taken into account when noting the increased rate of S pneumoniae found in conjunctival cultures in the same periods (March, November, and December).34 A retrospective Australian study14 showed a trend toward more isolates in the cooler, wetter seasons of autumn and winter.
The multivariate analysis identified the bacterial virulence as the main independent factor of visual outcome. We classified patients into the following 3 infection groups: virulent, CNSP, and sterile. This classification is based on the reported virulence of bacteria in cohort or case series, such as Pseudomonas species,11,35,36Enterococcus species,37Streptococcus species,11,22,38S aureus,22,39 and S lugdunensis40 in acute postoperative endophthalmitis. For instance, only 29% of patients infected with virulent gram-positive cocci had a final VA of 20/40 or better.12 Another example concerns infections with S aureus, streptococci, or enterococci, with 50%, 30%, and 14%, respectively, having a final VA of 20/100 or better in the EVS.12 Several reports have emphasized the prognostic role of bacterial virulence (Table 5),10- 16 but our study is, to our knowledge, only the second prospective study after the EVS that shows the influence of bacterial virulence on final visual prognosis, taking into account baseline clinical factors. Intraocular lesions (inflammation and retinal damage) during infection have been extensively studied in animal models of endophthalmitis, and experimental studies have strongly suggested that ocular lesions during infection are related to bacterial virulence and secondary intraocular inflammation to the host response to bacterial toxins and cell wall constituents (cellular infiltration; production of cytokines, chemokines, and adhesion molecules; and Fas ligand).41 Virulence is strongly associated with bacterial characteristics, such as production of toxins (cytolysin of Enterococcus faecalis, the quorum-sensing fsr system regulating gelatinase and serine protease, pneumolysin and autolysin of S pneumoniae, and S aureus alpha-toxin and beta-toxin), production of biofilms (Enterococcus and Staphylococcus, including CNSP), and motility in the eye (Bacillus cereus).42
Culture-negative cases are usually associated with a better final visual outcome (Table 5).31 These patients probably have a low level of inoculum undetected by conventional cultures and PCR techniques despite infection. This situation is close to that of CNSP infections, which are traditionally associated with a less severe disease course.12,22,31 However, a subpopulation of patients with CNSP infection has a severe prognosis (3 [9%] in our series). The more severe CNSP infections may be related to specific virulence traits of the involved strains, such as higher adhesion or biofilm formation.43 Further development in the field of CNSP virulence may help to better define specific phenotypic and genotypic traits associated with the occurrence of more severe endophthalmitis cases. Also, previous research43 shows that acquired antibiotic resistance is frequent in S epidermidis isolated from eyes with endophthalmitis (eg, methicillin resistance). The development of diagnostic tests allowing rapid evaluation of the virulence and antibiotic resistance traits of CNSP may help better define the population of patients with CNSP endophthalmitis with poorer outcomes. Previous studies have shown that the turnaround time of panbacterial PCR is 2 to 3 days,6 and this time could be reduced to only a few hours with real-time PCR, as recently described in the literature.7- 9
The need for PPV was associated with a poor visual outcome but also with other clinical factors correlated with the severity of the disease (poor initial VA, loss of red reflex, corneal edema, infection with a virulent species, and early onset of symptoms). For this reason, after multivariate analysis, the need for PPV was no longer identified as an independent prognostic factor. This relationship is easier to understand with the legacy of the EVS, showing the benefit of PPV in patients with more severe disease, that is, those with a VA of light perception at admission.10 Another nonrandomized study15 showed that PPV was associated with a worse prognosis owing to the indication bias of vitrectomy.
In conclusion, this prospective study shows that bacterial virulence is a major prognostic factor of final visual function in patients with acute endophthalmitis after cataract surgery. Although most patients currently undergo binary treatment, the use of PPV or intravitreal injections and rapid bacterial identification and characterization (especially for virulence and antibiotic resistance) could guide clinicians in adapting the therapeutic regimen to each patient.
Corresponding Author: Christophe Chiquet, MD, PhD, Department of Ophthalmology, University Hospital, BP217, F-38043 Grenoble, France (firstname.lastname@example.org).
Submitted for Publication: November 14, 2012; final revision received February 23, 2013; accepted February 26, 2013.
Published Online: July 25, 2013. doi:10.1001/jamaophthalmol.2013.4242.
Author Contributions:Study concept and design: Combey de Lambert, Aptel, Chiquet.
Acquisition of data: Combey de Lambert, Campolmi, Cornut, Creuzot-Garcher, Chiquet.
Analysis and interpretation of data: Combey de Lambert, Aptel, Creuzot-Garcher, Chiquet.
Drafting of the manuscript: Combey de Lambert, Campolmi, Aptel, Chiquet.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Campolmi, Aptel, Chiquet.
Administrative, technical, and material support: Campolmi, Cornut, Aptel, Creuzot-Garcher.
Study supervision: Chiquet.
Conflict of Interest Disclosures: None reported.
Group Information: The French Institutional Endophthalmitis Study group included the following participants: Christophe Chiquet, MD, PhD (study coordinator); François Vandenesch, MD, PhD, and Max Maurin, MD, PhD (statistics, methodology, and microbiological techniques); Pierre-Loïc Cornut, MD, PhD (database management); Pierre-Olivier Lafontaine, MD, Marie Passemard, MD, Catherine Creuzot-Garcher, MD, PhD, and Alain Bron, MD, PhD (Department of Ophthalmology, University Hospital of Dijon); Christophe Chiquet, MD, PhD, Aurélie Combey de Lambert, MD, Thierry Zhou, MD, Ralitsa Hubanova, MD, Karine Palombi, MD, and Jean-Paul Romanet, MD, PhD (Department of Ophthalmology, University Hospital of Grenoble); Pierre-Loïc Cornut, MD, PhD, Frédéric Rouberol, MD, and Philippe Denis, MD, PhD (Department of Ophthalmology, University Hospital of Lyon [University Hospital Edouard Herriot]); Nelly Campolmi, MD, PhD, Gilles Thuret, MD, PhD, and Philippe Gain, MD, PhD (Department of Ophthalmology, University Hospital of Saint-Etienne); André Péchinot, PhD, and Catherine Neuwirth, MD, PhD (Department of Microbiology, University Hospital of Dijon); Jacques Croizé, MD and Max Maurin, MD, PhD (Department of Microbiology, University Hospital of Grenoble); Jérôme Etienne, MD, Yvonne Benito, Eng, Sandrine Boisset, Pharm PhD, Anne Tristan, MD, and François Vandenesch, MD, PhD (Department of Microbiology, University Hospital of Lyon); Anne Carricajo, PhD, and Gérard Aubert, PhD (Department of Microbiology, University Hospital of Saint-Etienne); Frédéric Dalle, MD, PhD, and Alain Bonin, MD, PhD (Department of Mycology, University Hospital of Dijon); Bernadette Lebeau, MD, and Hervé Pelloux, MD, PhD (Department of Mycology, University Hospital of Grenoble); Frédérique de Montbrison, MD, and Stéphane Picot, MD, PhD (Department of Mycology, University Hospital of Lyon); and Hélène Raberin, MD, and Roger Tran Manh Sung, MD, PhD (Department of Mycology, University Hospital of Saint-Etienne).