Regression model results for seven general indicators of illness among 1,369 cats fed meat-based or vegan diets.

Note: Cat numbers in some groups were lower than 1,369, as described under Results. The effect of each variable is reported as ‘exponentiated effect estimate [exponentiated 95% confidence interval], p-value’. Boldface indicates p < 0.05. The reference categories for variables ‘sex and neuter’ and ‘habitat’ are ‘female, spayed’ and ‘mostly indoor habitat’, respectively. The exponentiated intercept parameters encode the odds (models 1–6) of ‘unhealthy’ outcomes occurring, or the expected number of disorders (model 7), for cats with these reference characteristics. The effects of other categories state average differences compared to these reference categories. For the logistic regression models (models 1–6) these exponentiated effect estimates are odds ratios. For Quasi-Poisson model 7 the effects correspond to multiplicative changes in the expected number of health disorders per unwell cat. Some age effects were estimated nonlinearly. For these effects, the estimated degrees of freedom (edf) values are given (as a measure of how much each effect deviates from linearity), instead of a linear effect estimate. Visualizations of all control variable effect estimates and the nonlinear age effects can be found in S1–S6 Figs.