Table S1. Papers included in the meta-analysis

Table S2. Estimated (Est.) coefficients (Coef.), and 95% confidence intervals (lower and upper confidence interval limit; LCI, UCI) from multi-level meta-regression models fitted from the obesogenic diet dataset. All models included study ID as a random effect.

Table S3. Estimated (Est.) coefficients (Coef.), and 95% confidence intervals (lower and upper confidence interval limit; LCI, UCI) from multi-level meta-regression models fitted from the obesogenic diet locomotory behaviour dataset. All models included study ID as a random effect.

Table S4. Estimated (Est.) coefficients (Coef.), and 95% confidence intervals (lower and upper confidence interval limit; LCI, UCI) from multi-level meta-regression models fitted from the caloricrestriction dataset. All models included study ID as a random effect.

Table S5. Estimated (Est.) coefficients (Coef.), and 95% confidence intervals (lower and upper confidence interval limit; LCI, UCI) from multi-level meta-regression models fitted from the proteinrestriction dataset. All models included study ID as a random effect.

Figure S1. PRISMA chart of the literature review screening process and selection process. The main data set examined the influence of obesogenic diets on time spent in the open in Elevated Plus Maze and Open Field Tests, a measure of anxiety-like behaviour used in rodent models. Auxiliary data sets examined the effect of obesogenic diets on measures of locomotion (LOC), and effects of caloric restriction (CR) and protein restriction (PR) on anxiety-like behaviour.

Figure S2. Funnel plots showing the individual effect sizes (lnRR) as a function of precision in each of the 4 datasets; risk-taking behaviour under caloric restriction (CR), protein restriction (PR), obesogenic diets (OB), as well as locomotory behaviour under OB. Solid vertical lines indicate the overall mean effect as estimated by meta-analysis. In the absence of both publication bias and heterogeneity 95% of effect sizes would be expected to fall within the white regions.

Figure S3. Bubble plot showing effect sizes (lnRR) of obesogenic diets on locomotory behaviour as a function of the duration of the dietary exposure. Individual effect sizes are scaled by their respective precision. The solid black line indicates the fit a of a multi-level meta-regression, the slope of which is statistically significant (MLMR lnRRSlope = −0.005, CI = −0.008 to −0.001; see Table A3 for all coefficients from all meta-regressions on this dataset). The shaded area indicates the region encompassed by the 95% confidence interval for the fitted values.

Figure S4. Orchard plot (as in Figure 1) showing the individual effect sizes in the caloric restriction dataset as a function of dietary exposure type. The difference between overall estimates in the two groups is statistically significant based on multi-level meta-regression (MLMR lnRRIndirect – Direct = − 0.487, CI = −0.895 to −0.078; see Table A4 for all coefficients from all meta-regressions on this dataset).

Figure S5. Orchard plot (as in Figure 1) showing the individual effect sizes in the protein restriction dataset as a function of test type; open field test (OFT) or elevated plus maze (EPM). The difference between overall estimates in the two groups is statistically significant based on multi-level metaregression (MLMR lnRREPM – OFT = −0.332, CI = −0.646 to −0.018; see Table A5 for all coefficients from all meta-regressions on this dataset).

Figure S6. Orchard plot (as in Figure 2) showing the individual effect sizes in the protein restriction dataset as a function of the sex of the animals. The difference between overall estimates in the males and females is statistically significant based on multi-level meta-regression (MLMR lnRRMix –Females = −1.471, CI = −2.843 to −0.100; see Table A5 for all coefficients from all meta-regressions on this dataset).