Is routine ophthalmoscopy really necessary in candidemic patients?

The purpose of this study was to determine among patients with candidemia the real rate of ophthalmoscopy and the impact of performing ocular assessment on the outcome of the disease. We performed a post hoc analysis of a prospective, multicenter, population-based candidemia surveillance program implemented in Spain during 2010–2011 (CANDIPOP). Ophthalmoscopy was performed in only 168 of the 365 patients with candidemia (46%). Ocular lesions related to candidemia were found in only 13/168 patients (7.7%), of whom 1 reported ocular symptoms (incidence of symptomatic disease in the whole population, 0.27% [1/365]). Ophthalmological findings led to a change in antifungal therapy in only 5.9% of cases (10/168), and performance of the test was not related to a better outcome. Ocular candidiasis was not associated with a worse outcome and progressed favorably in all but 1 evaluable patient, who did not experience vision loss. The low frequency of ophthalmoscopy and ocular involvement and the asymptomatic nature of ocular candidiasis, with a favorable outcome in almost all cases, lead us to reconsider the need for systematic ophthalmoscopy in all candidemic patients.

Nevertheless, recent series showed a low frequency of symptomatic ocular disease and a favorable clinical outcome in almost all patients with Candida eye involvement who received systemic antifungal treatment [10]. Moreover, the recommendation to perform systematic ophthalmoscopy is based not on the findings of a randomized controlled trial, but rather on the results of old, small-scale studies that do not clearly state clinical benefits [1][2][3][4][5].
Based on data from a recent population-based study of candidemia in Spain (CANDIPOP study), we determined the frequency of ophthalmoscopy and analyzed the impact of the examination on outcome.

Study setting
The findings reported here are from a post hoc sub-analysis of the Population Study on Candidemia (CANDIPOP), a prospective, multicenter, population-based candidemia surveillance program implemented in 29 hospitals in Spain. The inclusion criteria, study population, main definitions, microbiological studies, and outcomes have been extensively described elsewhere [11]. Briefly, during the study period local laboratories daily identified patients and reported them to study coordinators, who collected data using a standardized case report form. Data included demographic and clinical characteristics, risk factors for candidemia, antifungal management and source control. Thirty-day follow-up outcome was recorded for each patient. Given the observational nature of the study, patients were managed according to routine clinical care.
Severity of the infection and Pitt bacteremia score were recorded on the day of blood culture sampling [12]. Proven catheter-related candidemia was defined according to current guidelines [13], whereas secondary candidemias required the microbiological documentation of the same Candida species at the origin of infection [11]. When there was no apparent infection at another site, candidemia was classified as primary. An episode of candidemia was defined as persistent when patients had positive follow-up blood cultures performed according to IDSA guidelines [9].
The local institutional review boards of each participating center approved this study, and written informed consent was obtained from each patient before enrollment (IRB of the coordinating center for this study: Comité Ético de Investigación Clínica, Hospital General Universitario Gregorio Marañón).

Populations
Systematic dilated ophthalmoscopy was recommended at baseline for all patients included in the study, although only pathological findings suggestive of Candida ocular involvement were recorded in the general CANDIPOP database. To overcome this limitation, we asked the 29 participating hospitals to re-check whether ophthalmoscopy had been performed in patients aged >16 years who had experienced an episode of candidemia. Eleven hospitals complied with our request. Therefore, we included the 365 episodes of candidemia from the CANDIPOP study for which information was available on whether or not ophthalmoscopy had been performed.

Collection of ophthalmologic data
The study coordinators were invited to retrospectively review the results of all ophthalmological examinations in order to categorize ocular involvement as proven, probable, or possible Candida chorioretinitis or endophthalmitis [10]. To be considered proven, ocular candidiasis had to be diagnosed based on the presence of an ocular lesion and on isolation of the microorganism from the vitreous humor by culture or histopathology-based identification. Probable Candida endophthalmitis consisted of vitritis or fluffy lesions with extension into the vitreous humor. Probable Candida chorioretinitis included deep focal white infiltrates in the retina, hemorrhages, Roth spots, or cotton wool spots. In patients with diabetes, hypertension, or concomitant bacteremia, ocular involvement was classified as possible based on previous criteria [10]. Injection of intravitreal antifungal agents or intravitreal corticosteroids was recorded, as was the need for vitrectomy.
The outcome of ocular candidiasis was considered successful when follow-up ophthalmoscopy revealed resolution of the retinal lesion or of active inflammation.

Statistical analysis
Categorical variables are presented as absolute numbers and their relative frequencies. Quantitative variables are presented as means and standard deviation (SD) if normally distributed and as median and interquartile range (IQR) if non-normally distributed. We compared categorical variables between groups using the Pearson chi-square and Fisher exact tests; we compared continuous variables using the Mann-Whitney test or a two-tailed t test.
Risk factors for 30-day mortality were analyzed using the Cox regression model assuming proportional hazards.
One of our main objectives was to analyze the impact of systematic ophthalmoscopy on the outcome of patients with candidemia. Since very early mortality is generally associated with the impossibility to perform an ophthalmoscopy, patients who died within 3 days after withdrawal of blood cultures (12 patients) were excluded from mortality analysis to rule out potential bias.
Associations are given as odds ratio (OR) with the 95% confidence interval (95% CI). Data were analyzed using SPSS Statistics for Windows, Version 17.0 (SPSS Inc., Chicago, Illinois, USA). Statistical significance was established at p < 0.05.

Frequency of ophthalmoscopy and incidence of ocular candidiasis
Ophthalmoscopy was performed in only 168 of the 365 patients with candidemia (46%). Abnormalities suggestive of Candida eye involvement were present in only 13/168 cases (7.7%).
Probable Candida endophthalmitis occurred in 2 patients, whereas probable and possible chorioretinitis were present in 8 and 3 cases, respectively. Interestingly, all patients but 1 (a patient with probable bilateral chorioretinitis who had low visual acuity) were asymptomatic when the diagnosis of ocular candidiasis was established. Accordingly, symptomatic ocular disease due to Candida occurred in 0.27% of the whole candidemic population (1 out of 365 patients) and in 7.6% of the 13 patients with ocular involvement (1 out of 13 patients).

Assessment of performing or not ophthalmoscopy on 30-day mortality
Overall, 102/365 patients (27.9%) died within 30 days of the episode; 12 of these patients died within 3 days of blood sample collection and were, therefore, excluded from the mortality analysis. Univariate analysis of risk factors for 30-day mortality in the 90 remaining patients showed that the factors associated with poor outcome were admission to the ICU, previous renal disease, HIV infection, previous corticosteroid treatment, primary source of infection, septic shock, higher Pitt score, and need for hemodialysis as a complication of candidemia. The factors associated with a better outcome were admission to a surgical ward, receipt of azoles, and ophthalmoscopy.
However, when a multivariate analysis was performed, the independent risk factors for mortality (Table 1) were septic shock at presentation of candidemia, primary candidemia, and a high Pitt score. Performance of ophthalmoscopy did not remain an independent protective factor for 30-day mortality (OR, 0.59; 95% CI, 0.34-1.05; p = 0.08).

Clinical impact and outcome of ocular candidiasis
Antifungal therapy was changed after dilated ophthalmoscopy in 10/168 patients (5.9%). Six patients required a change in the class of antifungal administered, whereas 4 patients were prescribed extended treatment.
The outcome of ocular candidiasis is summarized in Table 2. Overall, information on the evolution of ocular candidiasis was available in 7/13 patients (53.8%), since 5 died and fundoscopy follow-up was not available in the remaining patient. Antifungal treatment was considered successful in 6/7 patients. None of the patients with Candida eye involvement needed intravitreal injection of antifungals or surgery.
As for outcome, although patients with ocular involvement had a higher mortality rate compared to patients without ocular candidiasis, the difference was not statistically significant (30-day mortality rate38.5% [5/13]

Comparison between patients with and without Candida eye involvement
In an attempt to determine whether ophthalmoscopy could be avoided in patients with a low risk of ocular candidiasis, we compared patients with and without ocular candidiasis who had undergone ophthalmoscopy.
As shown in Table 3

Discussion
Our study shows that, despite the recommendation by several guidelines that systematic ophthalmoscopy should be performed in all patients with candidemia [6][7][8][9], as many as 50% of patients never actually underwent ophthalmoscopy. Furthermore, the yield of this examination is low, and ocular infection is uncommon and mostly asymptomatic. Our data also show no independent impact of ophthalmoscopy on the outcome of candidemic patients. The incidence of ocular candidiasis in patients with candidemia varies from 50% in older studies [14][15][16][17] to less than 5% in more recent ones [18,19]. In the present report, the overall incidence of ocular candidiasis was 7.7%, which is consistent with that observed by Shah et al [20], who found a 7.9% frequency of chorioretinitis, with no cases of endophthalmitis. The widespread use of early antifungal therapy and improvements in diagnosis may explain the lower incidence of ocular candidiasis observed in recent years [18,19,21]. Nevertheless, recent guidelines recommend performing ophthalmoscopy in all candidemic patients [6][7][8][9], stating that "missing and not appropriately treating Candida endophthalmitis could have great consequences for the patients" [9]. However, the quality of the evidence supporting this recommendation is low, as it is based on the clinical judgment of the Expert Panel members. This weakness in the recommendations provided by guidelines is reflected in the wide range of rates of ophthalmoscopy in candidemic patients (53% to 75%) reported in the medical literature [22,23]. In the same sense, although all patients included in the CANDIPOP study were prospectively followed by infectious disease specialists, we found that in "real life", ophthalmoscopy was performed in less than 50% of the study population. In our fully publicly funded health care system, this low percentage cannot be attributed to cost-or reimbursement-related factors.
As for clinical manifestations, we found only 1 patient with ocular symptoms, ie, an incidence of symptomatic disease of 0.27% in the whole candidemic population. Moreover, ocular candidiasis progressed favorably in all except 1 of our evaluable patients, who was asymptomatic and did not experience vision loss. These findings are consistent with those recently reported by Oude Lahof, who observed visual symptoms in only 3.3% of patients with ocular candidiasis and a favorable outcome in almost all evaluable cases [10]. Older studies reporting ocular candidiasis as a "malignant complication" were performed in an era when medical management of candidemia differed substantially from contemporary care [1][2][3][4][5][24][25][26]. Furthermore, a high proportion of patients included in such studies had a history of intravenous drug use [5] that was not observed in our series.
Oude Lashof [10] found a similar mortality rate between patients with and without ocular candidiasis (43.3% vs. 36.5%, p = 0.31). We also observed nonsignificant differences (p = 0.13), thus indicating that the presence of ocular abnormalities does not predict a poor outcome.
Similarly, although the populations undergoing or not undergoing ophthalmoscopy were likely not identical, we did not observe an impact of routine ophthalmoscopy on clinical outcome. Performance of the examination resulted in a change in antifungal therapy in only 5.6% of the 168 patients. On the contrary, ophthalmoscopy generates an increase in hospital costs ($400 per consultation), is uncomfortable for patients, and carries a small risk of acute angleclosure glaucoma [27].
Since ocular candidiasis progressed favorably in almost all patients and given the lack of data on the clinical impact of longer treatment schedules, we believe that the recommendation of systematic dilated ophthalmoscopy for all candidemic patients should be reassessed. An alternative could be a risk-based approach, in which the examination is limited to symptomatic patients, those who do not respond to treatment, or those more likely to acquire ocular candidiasis.
Regarding this aspect, our findings are consistent with those reported by other investigators [10,15,20,28,29], who found increased severity of the underlying conditions (cancer, need for hemodialysis after candidemia, and corticosteroids) and exposure to a more virulent infection (C. albicans, persistent fungemia, septic metastasis in other organs) as risk factors for ocular candidiasis. Our study is subject to a series of limitations. First, the CANDIPOP study was not designed to analyze ocular candidiasis; however, we report the broadest experience to date on ophthalmoscopy in a large population of patients who were prospectively followed by an infectious disease specialist. Second, only 46% of the candidemic patients underwent ophthalmoscopy, with the result that we may have underestimated the involvement of ocular Candida infection; however, no clinical manifestations of ocular candidiasis were observed in the group of patients who did not undergo ophthalmoscopy. Finally, no patients with a history of drug addiction were included in the study.
In conclusion, we provide data from a large series of patients with candidemia showing that ophthalmological assessment is frequently omitted and that the rate of ocular candidiasis is relatively low (7.7%), anecdotally symptomatic, and usually associated with a good outcome. A prospective clinical trial evaluating the real benefits of routinely performed ocular assessment in all candidemic patients to limit the use of such a cumbersome, low-yield examination.