Time-dependent integrated predictive modelling is carried out using the PTRANSP code to predict fusion power and parameters such as alpha particle density and pressure in ITER H-mode plasmas. Auxiliary heating by negative ion neutral beam injection and ion-cyclotron heating of He3 minority ions are modelled, and the GLF23 transport model is used in the prediction of the evolution of plasma temperature profiles. Effects of beam steering, beam torque, plasma rotation, beam current drive, pedestal temperatures, sawtooth oscillations, magnetic diffusion and accumulation of He ash are treated self-consistently. Variations in assumptions associated with physics uncertainties for standard base-line DT H-mode plasmas (with Ip = 15 MA, BTF = 5.3 T and Greenwald fraction = 0.86) lead to a range of predictions for DT fusion power PDT and quasi-steady state fusion QDT (≡PDT/Paux). Typical predictions assuming Paux = 50–53 MW yield PDT = 250–720 MW and QDT = 5–14. In some cases where Paux is ramped down or shut off after initial flat-top conditions, quasi-steady QDT can be considerably higher, even infinite. Adverse physics assumptions such as the existence of an inward pinch of the helium ash and an ash recycling coefficient approaching unity lead to very low values for PDT. Alternative scenarios with different heating and reduced performance regimes are also considered including plasmas with only H or D isotopes, DT plasmas with toroidal field reduced 10% or 20% and discharges with reduced beam voltage. In full-performance D-only discharges, tritium burn up is predicted to generate central tritium densities up to 1016 m−3 and DT neutron rates up to 5 × 1016 s−1, compared with the DD neutron rates of 6 × 1017 s−1. Predictions with the toroidal field reduced 10% or 20% below the planned 5.3 T and keeping the same q98, Greenwald fraction and βn indicate that the fusion yield PDT and QDT will be lower by about a factor of two (scaling as B3.5).