B58-02 Assessing Proportional Bias Variability Across Oscillometry Devices Using Two Reference Equations

Rationale In 2013, Oostveen and colleagues published normative equations for respiratory oscillometry derived from five separate devices acquired from more than 350 healthy Caucasian adults which have been widely adopted in clinical and research laboratories. Recently, Valach et al. published new reference equations derived from the Austrian LEAD Study, a sample of 887 healthy adults, using a single device (Resmon Pro Full™). The objective of this study is to evaluate the compatibility and bias of the two reference equations with oscillometry measurements from three devices (Quark i2M, Resmon Pro Full™, and Tremoflo® C-100). Methods The present study included previously deployed Veterans whose data were collected using two devices (Quark i2M and Resmon Pro Full™) and civilians using Tremoflo® C-100. Resistance and reactance measurements at 5 Hz (R5, X5) were selected for analysis. Bland-Altman plots for each reference equation were used to visually assess mean bias, or the average of the difference between the predicted value and the measured value. Linear regression was conducted to quantitatively assess Bland-Altman results for proportional bias. The models were adjusted for sex and BMI. Results In a sample of 566 Veterans and civilians, the Valach plots were shown to have slightly steeper negative regression slopes than the Oostveen plots, thus showing a greater proportional bias when comparing predicted to measured values. In particular, levels of proportional bias differed among the two reference equations on two devices capturing R5, (Resmon: β = -0.55, p < 0.001, i2M: β = -0.21, p = 0.04), and X5 values across all three included devices (i2M: β = -0.33 , p < 0.001; Resmon: β = -0.47, p < 0.001; Tremoflo: β = -0.46, p < 0.001). Demographics and relevant statistics are included in the Table. Conclusion Despite small effect sizes, the Valach equation demonstrated greater differences in regression slopes, especially at the extremes of X5 and R5 on the Resmon device, compared to the Oostveen equations. This may reflect differences in the underlying populations, as Valach’s equation was derived from a more geographically restricted cohort than Oostveen. Future studies should consider comparing device measurements across the same sample, since demographic characteristics, particularly age, differed by machine.