Statistical analysis

The sample size of ten volunteers was predicted to detect a treatment effect 1.4 times the magnitude of the within volunteer spread, with 80% power at the 5% significance level. For glucose and insulin, the within volunteer spread was approximately 10% (based on previous studies), which gave the study power to detect a treatment shift of ca. 14%. Data were analysed by analysis of variance (ANOVA) of area under the curve (AUC) and iAUC of data collected over multiple time points following consumption of the test meal, as well as the data at each time point individually. The ANOVA included terms for volunteer, baseline value and diet. A version of the ANOVA in which the diet term was split into a contrast between the meat diet and the plant diet, and the variation among the plant diets was also examined. Differences among the diets were assessed by post hoc tests with Tukey's adjustment for multiple comparisons. Data were checked for the assumptions of normality and constant variance, and were log transformed before analysis where appropriate. All analyses were carried out using Genstat 17 (Lawes Agricultural Trust, VSN international Ltd, Hemel Hempstead, UK). The plant metabolites (non-nutrients) from test meals, plasma phenylpropanoid pathway and products of protein and carbohydrate metabolism data was analysed by principal component analysis (PCA), unit variance (UV)-scaled using SIMCA 14.1 (Umetrics, Cambridge). The correlation between plasma metabolites and gastrointestinal hormones was analysed by partial least squares-discriminant analysis (PLS-DA) SIMCA 14.1 (Umetrics, Cambridge). The effect of test meals over time and between test meals (plasma) as well as the differences between the amounts of amino acids eaten for each interventions were assessed by two‐sided post hoc t tests. A version of the analysis which accounted for the varying protein intake was also conducted, by including a term for amount consumed and its interaction with the protein type.