This paper discusses two approaches that have been proposed to account for the data scatter observed in delamination growth tests under cyclic-fatigue loading and thereby enable an estimate of a worst-case delamination growth curve for use in the damage tolerance and durability assessment of composite and adhesively-bonded airframes. The two approaches discussed are: (a) the normalisation approach, whereby the energy release rate is divided by the resistance to delamination growth, GR(a), and (b) the Hartman-Schijve approach to delamination growth. It is shown that for the cases considered this normalisation approach can be used to yield curves that are similar to the 'mean-3σ', "worst-case", i.e. upper-bound, curve obtained using the Hartman-Schijve equation. However, despite the reduction in the scatter that arises if this particular normalisation approach is adopted, there is still considerable scatter in the important "near-threshold" region. In this region the normalised curves are bounded above by the 'mean-3σ' curve obtained using the Hartman-Schijve equation. To address this issue, an alternative normalisation approach is then proposed. This alternative normalisation approach has the advantage of having reduced scatter in the near-threshold region but elsewhere is significantly more conservative than the Hartman-Schijve approach.