This study investigated whether a running-adapted version of the cycling-based "step-ramp-step" (SRS) protocol would improve prediction of O2 in treadmill exercise compared to the traditional prescriptive approach. Fourteen healthy individuals (6 females; 25 ± 6years; 66.1 ± 12.7kg) performed a treadmill-based SRS protocol including a ramp-incremental test to task failure followed by two constant-speed bouts within the moderate-(MODstep-below estimated lactate threshold; θLT), and heavy-intensity domains (HVYstep-between θLT and respiratory compensation point; RCP). Using the uncorrected O2-to-speed relationship from the ramp exercise, three constant-speed bouts were performed at 40-50% between: baseline and θLT (CSEMOD); θLT and RCP (CSEHVY); and RCP and peak (CSESEV). For CSEMOD, CSEHVY, and CSESEV measured end-exercise O2 was compared to predicted O2 based on the: (i) "SRS-corrected" O2-to-speed relationship (where MODstep and HVYstep were used to adjust the O2 relative to speed); and (ii) linear "uncorrected" data. Average treadmill speeds for CSEMOD and CSEHVY were 7.8 ± 0.8 and 11.0 ± 1.4km·h-1, respectively, eliciting end-exercise O2 of 1979 ± 390 and 2574 ± 540mL·min-1. End-exercise O2 values were not different compared to SRS-predicted O2 at CSEMOD (mean difference: 5 ± 166mL·min-1; p = 0.912) and CSEHVY (20 ± 128mL·min-1; p = 0.568). The linear "uncorrected" estimates were not different for CSEMOD (-91 ± 172mL·min-1; p = 0.068) but lower for CSEHVY (-195 ± 146mL·min-1; p < 0.001). For CSESEV (running speed: 13.8 ± 1.7km·h-1), the end-exercise O2 was not different from peak O2 achieved during the ramp (3027 ± 682 vs. 2979 ± 655mL·min-1; p = 0.231). In healthy individuals, the SRS protocol more accurately predicts speeds for a target O2 compared to traditional approaches.