Validation of the C-DU(KE) Calculator as a Predictor of Outcomes in Patients Enrolled
in SCUT and MUTT Trials
Alejandro Arboleda1, Hazem M Mousa2, Namperumalsamy V Prajna3, Prajna Lalitha3, James Feghali4, Nisha R Acharya5,
Thomas M Lietman5, Victor L Perez2, Jennifer Rose-Nussbaumer1,5
1Byers Eye Institute, Department of Ophthalmology, Stanford University, Palo Alto, California, USA; 2Foster Center for Ocular Immunology, Department of Ophthalmology, Duke Eye Center, Duke University, Durham, North Carolina, USA; 3Aravind Eye Care System, Madurai, India; 4Department of Neurosurgery, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA; 5Department of Ophthalmology and Francis I. Proctor Foundation, University of California, San Francisco, San Francisco, California, USA
Purpose: To validate the C-DU(KE) calculator as a predictive tool on a dataset from patients with culture-positive ulcers compiled from NIH-funded studies including SCUT, MUTT, and MUTT II.
Methods: Criteria from the C-DU(KE) calculator for stratifying risk in infectious keratitis were recorded from 1064 (500 bacterial and 564 fungal) cases of infectious keratitis. These variables include (1) Corticosteroid use after symptoms, (2) Decreased visual acuity, (3) Ulcer area, (4) Keratitis caused by a fungal organism, and (5) Elapsed time from diagnosis to appropriate therapy. Univariable analysis was performed followed by a stepwise multivariable logistic regression on a culture-exclusive (without culture-dependent variables) model and a culture-inclusive model to assess for association between the variables and need for surgical intervention. Then, the predictive probability of treatment failure was calculated for each patient using the derived C-DU(KE) model. Treatment failure was defined as corneal perforation, need for therapeutic penetrating keratoplasty, or corneal thinning greater than 50% of initial measurement. Discrimination was assessed using the area under the receiver operating characteristic curve (AUC) and results were compared to those from the original study.
Results: Overall, 18.3% of the current study population required intervention compared to 38.9% of the initial C-DU(KE) population. Univariable analysis showed that all five criteria were higher in patients who failed treatment. Of these, decreased visual acuity (p<0.001), larger ulcer area (p<0.001), and presence of a fungal organism (p<0.001) had a significant association with treatment failure. In the culture-exclusive multivariable model, decreased vision (OR=3.13, p<0.001) and increased ulcer area (OR=1.03, p<0.001) had significant impact on outcomes while corticosteroid use did not (OR=1.93, p=0.226). In the culture-inclusive multivariate model, decreased vision (OR=4.9, p<0.001), ulcer area (OR=1.02, p<0.001), and fungal etiology (OR=9.8, p<0.001) significantly impacted results while corticosteroid use and elapsed time to appropriate therapy did not. This analysis showed that 2 of 3 variables in the culture-exclusive model and 3 of 5 variables in the culture-inclusive model were similar in the two datasets. The AUCs were 0.784 for the culture-exclusive model and 0.846 for the culture-inclusive model, which were comparable to the original validation set.
Conclusions: The C-DU(KE) calculator was useful as a risk stratification tool for infectious keratitis in a patient population from large international studies primarily taking place in India. The results of this study support its use as a clinical aid that assists physicians in patient management.
Disclosure: N: (Arboleda, Mousa, Prajna, Lalitha, Feghali, Acharya, Lietman, Rose-Nussbaumer);
C: (Perez: Alcon, Dompe, EyeGate, Kala, Trefoil, Quidel, OBTears)
Support: National Institutes of Health/National Eye Institute: R01EY030283 (VLP), R01EY024485 (Perez), UG1 EY030417 (Rose-Nussbaumer), UG1 EY028518 (Rose-Nussbaumer), Core Grant and Prevent Blindness Unrestricted Grant (Stanford, Duke)