Validation of oxidative stress assay for schizophrenia

Abstract

Accumulating evidence implicates oxidative stress in a range of diseases, yet no objective measurement has emerged that characterizes the global nature of oxidative stress. Previously, we reported a measurement that employs the moderately strong oxidant iridium (Ir) to probe the oxidative damage in a serum sample and reported that in a small study (N = 15) the Ir-reducing capacity assay could distinguish schizophrenia from healthy control groups based on their levels of oxidative stress. Here, we used a larger sample size to evaluate the Ir-reducing capacity assay to assess its ability to discriminate the schizophrenia (N = 73) and healthy control groups (N = 45). Each serum sample was measured (in triplicate) at three different times that were separated by several weeks. The Intraclass Correlation Coefficient (ICC = 0.69) for these repeated measurements indicates the assay detects stable components in the sample (i.e., it is not detecting transient reactive species or air-oxidizable serum components). Correlations between the Ir-reducing capacity assay and independently-measured total serum protein levels (r = +0.74, p < 2.2 × 10^-16) suggest the assay is detecting information in the protein pool. For cross-validation of the discrimination ability, we used machine learning and receiver operating characteristic (ROC) analysis. After adjusting for potential confounders (age and smoking status), an area under the curve (AUC) of ROC curve was calculated to be 0.89 (p = 9.3 × 10^-5). In conclusion, this validation indicates the Ir-reducing capacity assay provides a simple global measure of oxidative stress, and further supports the hypothesis that oxidative stress is linked with schizophrenia.

Keywords: Machine learning; Oxidative stress; Receiver operating characteristic (ROC); Schizophrenia; Total antioxidant status.