Evaluation of inter-rater and test-retest reliability for near-infrared spectroscopy reactive hyperemia measures.

Abstract

Near-infrared spectroscopy (NIRS) is a non-invasive tool used to measure blood flow in peripheral tissues. More information on inter-rater agreement and test-retest reliability of NIRS-based reperfusion assessments is needed. To assess inter-rater agreement for NIRS based data analysis, and evaluate the measurement's reliability across days. On three separate days (average days between visits 1 and 3: 19.4±6.9days), participants' (N=15 males, 22±2yr.) post-occlusion reactive hyperemia (PORH) was measured in the left gastrocnemius muscle using Continuous-Wave NIRS (CW-NIRS). A blood pressure cuff was placed proximal to the knee and inflated to occlude lower leg blood flow for 5min. The following CW-NIRS parameters were selected: (1) percent saturation in HbO2 (StO2%) at baseline; (2) the O2Hb range used to normalize the NIRS signal; (3) the time for the O2Hb signal to reach 50% peak post-occlusion hyperemia (T1/2), and (4) the post peak hyperemic O2Hb recovery slope (O2REC-SLP). Absolute agreement between the two analysts was calculated using two-way random effects Intraclass Correlation Coefficients (ICC2,1). Consistency between analysts and across days was calculated using two-way mixed models (ICC3,1). Mean and 95% confidence intervals (CI) of ICCs are reported. Coefficient of variation (CV) and standard error of the measurement (SEM) are reported. The ICC2,1 data indicated "adequate" to "excellent" absolute agreement between the two NIRS analysts. ICC2,3 data indicated "adequate" to "good" reliability across visits. The CV and SEM for rater 1 and rater 2 across visit were StO2 (CV: 3.79%±2.71% and 4.50%±2.37%; SEM: 3.42 and 3.82), O2Hb range (CV: 10.50±5.93 and 12.79±12.41; SEM: 3.26 and 4.71), T1/2 (CV: 11.15%±5.52% and 10.96%±4.50; SEM: 1.22 and 1.11), and O2REC-SLP (CV: 19.49%±9.99% and 18.45%±9.48%; SEM: 0.04 and 0.04). It is concluded that NIRS parameters assessed show adequate reliability between analysts and across three visits. It is recommended, when feasible and because of the absence of 100% reliability, that investigators employ more than one rater for scoring at least a portion of the data across each trial in a study's control condition in order to have the ability to estimate the magnitude of error attributable to imperfect reliability.