Purpose: To critically appraise the utility of heart rate (HR) and power output (PO) to predict metabolic rate (MR) and oxygen consumption (V˙O2) during variable-intensity roller skiing and cycling. Methods: National-level cyclists (n = 8) and cross-country skiers (n = 9) completed a preliminary session to determine V˙O2max, and a variable-intensity protocol with 3 high-intensity stages at 90% V˙O2max for 3 minutes interspersed with 3 moderate-intensity stages at 70% V˙O2max for 6 minutes. Cardiorespiratory measures were recorded throughout. Linear HR–MR, HRV˙O2, PO–MR, and POV˙O2 regressions were computed from the preliminary session, individually, for all athletes and used to predict MR and V˙O2 from both HR and PO, separately, during the variable-intensity protocol. Mean differences with 95% limits of agreement (LOA) between measured and predicted MR and V˙O2 were calculated. Results: MR and V˙O2 estimated from HR displayed a mean bias close to zero but wide LOA. HR overestimated MR and V˙O2 during moderate intensity but underestimated MR and V˙O2 during high intensity, for both roller skiing and cycling. MR and V˙O2 estimated from PO were more consistent across the experimental trial, displaying a mean bias farther from zero but with tighter LOA. Conclusions: This study has demonstrated that HR has limited utility to predict metabolic intensity during variable-intensity roller skiing and cycling because of wide LOA. On the other hand, metabolic intensity predicted from PO had tighter LOA, suggesting better consistency. PO might provide a better prediction of metabolic intensity compared with HR, particularly when longer-duration steps are performed during preliminary testing.