For example the Mahalanobis distance or on related extensions by propensity-like scores. Future studies may explore such methods. Matching can be performed both with and, as done here, without replacement. An FM19G11 advantage of matching with replacement is that the match will not depend on initial sorting order of trial participants and that the distances will be globally minimized. However, some individuals, showing extreme values, may potentially end up with an unduly large influence on the results as a consequence of being selected multiple times. From inspection of scree plots, four iterations were chosen, which may be considered as arbitrary, but it is unlikely that notable differences in results would be obtained by choosing, for instance, three or five iterations. Several aspects of the trial participants could not be mimicked in the observational data. Despite 773 trial participants being randomized to the intervention, only 555 had information on dietary intake and, of these, 460 had information about FMI. However, the re-analysis of trial participants with data on diet and weight change between randomization and post-intervention showed results similar to the Hydroxyflutamide analyses of the initial trial. Inadequate matching on some variables was also a problem; the highest values of protein intake among trial participants could not get a good match in the observational data. This is probably because the highest intake in the trial generally goes beyond habitual intake reflected in observational data. However, the reanalysis of the DiOGenes trial data restricted to participants with protein intake below 30 E% showed a result essentially similar to the result of the original DiOGenes trial. Thus, these differences seemed not to influence the present study. Various differences were present across the trial and the observational data, which can potentially be important for the results. These are discussed in the appendix note in file S1, and include differences in measurement methods, exposure, follow-up time as well as the differences between weight change and weight loss maintenance. However, the hypothesized beneficial effect of a high-protein diet on weight control may be assumed to be unaffected by these differences, which is supported by the results of the present study. In conclusion, differences between the RCT and observational data were minimized wherever it seemed possible including dietary intake, participant characteristics and statistical analysis.