Abstract / Summary
Abstract Purpose : To estimate whether Brazil’s 2003–2004 restrictive and 2019 permissive firearm-policy shifts were followed by opposite changes in firearm-specific suicide mortality relative to non-firearm suicide, and to assess method substitution and the prespecified subgroup aged 60 years or older. Methods : We conducted a preregistered controlled interrupted time-series analysis of monthly Brazilian mortality data from January 2000 through December 2023. Firearm and non-firearm suicides were jointly modelled using negative-binomial regression with 12-month heteroskedasticity- and autocorrelation-consistent inference. Uncertainty from probabilistic redistribution of deaths of undetermined intent was propagated across 200 completed datasets. The primary estimand was the 36-month cumulative differential rate ratio (cDRR). Results : After the restrictive shift, firearm suicide declined relative to non-firearm suicide (cDRR, 0.899; 95% CI, 0.834–0.970), corresponding to 401.1 fewer fitted than counterfactual deaths (95% CI, 107.9–694.3 fewer). After the permissive shift, the cDRR was 1.054 (95% CI, 0.991–1.122), corresponding to 164.6 excess deaths (95% CI, 25.1 fewer to 354.3 excess). The between-shift cDRR ratio was 1.172 (95% CI, 1.071–1.283). Both directional national tests passed preregistered gatekeeping, although the permissive-shift estimate was imprecise and power-limited. Non-firearm suicide did not move oppositely, and the pattern was not reproduced among adults aged 60 years or older. Conclusion : Firearm-policy direction was associated with firearm-specific suicide in the hypothesised directions, more clearly after restriction than permissiveness. However, residual temporal structure, limited falsification specificity, and pandemic overlap constrain causal interpretation; directional symmetry was partially supported, not confirmed.