Validity of visual evoked response in measuring ethambutol induced optic neuritis in tuberculosis patients: Evaluation using receiver operating characteristic [ROC] curves

Significant vision loss can occur during treatment with Ethambutol [EB] in tuberculosis [TB] patients. Visual Evoked Response [VER] is often used to detect this subclinical visual impairment even before the appearance of symptoms. We assessed the usefulness of three VER parameters- P[100] latency, amplitude and interocular difference for the early diagnosis of ethambutol-induced optic neuritis [ON]. This study was carried out on 60 newly diagnosed adult cases of tuberculosis aged between 20-50 years who were randomly assigned into two groups of 30 each. Nonparametric Receiver Operating Characteristic [ROC] analysis was used to evaluate the validity of VER indexes. At a cut-off pint of 116 ms sensitivity for the diagnosis of ON was 77.8% and specificity was 81.1%. Results of the application of Bayes's theorem showed that 87% of the patients scoring 116 ms or higher would actually have ethambutol-induced ON and 99% of those scoring less than 116 ms would not have ON. The best area under curve [AUC] for ROCs, an index of diagnosing accuracy, was 0.91 for P[100] latency, suggesting very good accuracy. The results suggest that P[100] latency gives the best results for ON screening in ethambutol treated patients. Amplitude and interocular difference were reasonable alternatives. Measurement of P[100] latency of VER is a valuable tool which can be used more easily than clinical examination in detecting subclinical ethambutol-induced ON