The combination of these variables captures both total adiposity and body fat distribution. Body fat distribution is a better indicator of the adverse state of obesity than adiposity itself, as demonstrated in relation to mortality in the DCH cohort. The possibility that a beneficial effect of a high protein intake on weight control is more pronounced among individuals in an adverse state of obesity, with a combination of higher BMI and WC, needs to be investigated. Vergnaud et al. found a significant interaction between BMI and protein intake in relation to weight change. An 6-Hydroxymelatonin association between greater protein intake and weight gain was seen in all three groups, but the strongest among individuals with BMI 25�C30. However, interactions with other aspects of adiposity and body fat distribution were not investigated. The literature on modifying observational data is growing. Although not related to dietary protein and weight, previous studies have mimicked a trial in observational data. Here other aspects were important to mimic, e.g. a wash-out period before initiation of a drug. Several studies have mimicked hypothetical interventions. In relation to nutritional research Lajous et al. investigated the association between change in fish intake and subsequent long-term risk of coronary heart disease by mimicking a hypothetical intervention of fish intake. The method applied in the present study has the presumed advantage of mimicking the variation in exposure level followed by the intention-to-treat type of analysis of the trial. Further, it was possible to analyze the selected cohort participants according to the randomly assigned exposure of the trial participant. Otherwise, it may be problematic to compare results from an ����as-treated���� analysis in observational data with an intention-to-treat analysis in trial data. The intention-to-treat analysis does not necessarily reflect the actual exposure. As seen in Figure 3 and Table 1, great variation of protein intake existed within the high and low protein groups of the trial. This shows that the trial did not achieve a clear distinction of exposure level in accordance with the randomization status. When matching participants from an observational cohort study with trial participants, as done in the present study, it was possible to select a subgroup similar to the trial participants including the variation in exposure level. Matching was based on the Euclidean distance metric, but other HT1171 methods could also have been used.