Statistical analysis is described in detail in Supplement 1. Briefly, General linear stepwise regression analysis was used to evaluate statistical group in relation to the potential influence of various variables: age, brain pH, sex, blood ethanol and brain freezer storage time. Variables with a significant association with group were included in the final statistical model as covariates. To compare the correlation structure between heroin and control brains a structural equation multi-group analysis was performed. Two models were estimated, one in which the correlations were constrained to be the same across the two groups and one were the correlation parameters were unconstrained. Akaike's information criteria (AIC) and chi-square test (likelihood ratio test) was used to identify the best model fit. If the correlations were found to be invariant across groups, a Z-statistic (critical ratio test) was used for pairwise comparisons between single correlation estimates. Significance was set at P < 0.05 and trends considered for P < 0.10. Statistical evaluations were carried out using JMP, Statistica and SPSS software packages.
