Theoretically, the normalised plasma pressure (β) at which an NTM is triggered is expected to depend on normalised Larmor radius (ρ*) and normalised collisionality (ν), and this has formed the basis for the way in which NTM onset scalings are quoted on many devices. However, new analyses of JET data shows that such ρ* - ν based scalings are non-predictive, with discharges largely following such scalings over the majority of their duration while in H-mode. Neural network techniques indicate a key additional parameter to include is sawtooth period, providing a better degree of predictability. Indeed, this parameter appears more important than ρ* in predicting NTM onset. Analysis using cases where sawteeth are modified by localised heating and current drive, indicates that it is the sawtooth which governs where it is along the NTM onset scaling trajectory that the NTM is triggered, rather than leading to departures from the scaling. Finally, exploring data from cross-device similarity experiments shows that the different devices, rather than lining up with one scaling, appear to be displaced in β -ρ* space, with similar absolute values for the range in NTM onset β in each device. This suggests that a simple r* based extrapolation for ITER may be inappropriate, and that NTM threshold levels may be more directly related to absolute value of β, suggesting a higher β threshold for ITER, at least if large sawteeth are avoided.