To the Editor: We would like to thank Crofts et al [1] for their positive and constructive comments regarding our article [2]. The main reason why we have not explored insulin response in the Danish Inter99 cohort using the latent class trajectory approach is fairly technical. Serum insulin values are usually log-transformed before analysing them as continuous outcomes because of their skewed distributions. In our study, this would mean we would model a piecewise-linear trajectory on the log-scale (Fig. 1a), which results in a rather unrealistic shape when transformed back to the original scale (Fig. 1b). Imposing the peak at 30 min is already a big restriction; therefore, we did not want to put this further constraint on the shape of the insulin response curve.

Fig. 1
figure 1

 Serum insulin trajectory on the (a) logarithmic and (b) original scale. The curves represent a fictional participant for the purpose of demonstration

Full size image

Furthermore, insulin levels are highly variable, and their analysis is laboratory-dependent and expensive, limiting their utility for clinical purposes. Their potential use in prediabetic substratification is the subject of another investigation that we are planning to pursue in the EGIR-RISC (European Group for the study of Insulin Resistance: Relationship between Insulin Sensitivity and Cardiovascular disease risk) cohort [3], using glucose and insulin measurements at more than three time-points during a 2 h oral glucose tolerance test.