Abstract
Purpose :
The Collaborative Ocular Tuberculosis Study (COTS) Calculator is an evidence and experience-based online scoring tool (www.oculartb.net/cots-con) for initiating antitubercular therapy (ATT) in patients with clinically suspicious ocular tuberculosis (OTB). The objective of the study was to validate the predictive ability of COTS calculator to correctly recommend ATT.
Methods :
The minimum datapoints required for COTS calculator input (TB endemicity region for the patient, clinical phenotype, Mantoux test, IGRA test, chest radiology findings) were extracted from COTS-1 dataset, which included retrospective, observational case records of OTB patients with 6-month, 12-month or 24-month follow-up after completing treatment. The clinical phenotype of the affected eye and the right eye were included for patients with unilateral and bilateral uveitis respectively. The predictive performance of COTS calculator median score 4 and 5 was then compared to ATT initiation based on the clinician’s judgement. The accuracy of COTS calculator performance was then further validated based on treatment outcome at the three timepoints using machine learning technology.
Results :
Among 633 patients included, 544 (85.9%) patients were treated with ATT based on the treating physician discretion. Retrospective COTS calculator application stratified 283 (44.7%) patients with high or very high probability among international experts to initiate ATT (median score = 4 or 5) and 138 (21.8%) patients for very high probability alone (median score = 5). COTS calculator recommendation shared similar positive predictive value as physician’s discretion over all three timepoints. COTS calculator predictions using median score 5 show better specificity (0.43-0.61) over all timepoints than clinician’s judgement (0.15-0.19). The median scores of 4 and 5 show better sensitivity (0.82-0.87) than a score of 5, and better specificity (0.37-0.33) than clinician’s judgement. Bivariate analysis shows Mantoux and IGRA status as the most influential factor for treatment outcome at 24 months.
Conclusions :
COTS calculator provides greater predictive accuracy for less experienced clinicians to withhold ATT treatment in patients who are unlikely to have OTB, therefore reducing drug resistance and unnecessary side effects from inappropriate ATT use.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.