The participants in this study may not be representative of dependent heroin users who have never sought treatment, as most of our participants are at the severe end of the heroin dependence spectrum. However, non-treatment samples in Australia have been found to have similar characteristics to treatment samples (Ross, et al., 2005). A further limitation is that the participants were predominantly heroin dependent, not opioid analgesic dependent. Differences between these two groups may mean that different associations exist and so the results of this study may not generalize to dependent opioid analgesic users (Subramaniam & Stitzer, 2009). The study was also cross-sectional and retrospective. Although studies have shown that adult recall of child abuse can be inconsistent, the bias is towards false negatives rather than false positives (Fergusson, Horwood, & Woodward, 2000; Hardt & Rutter, 2004). There is also evidence that self-reported drug use amongst heroin users is reliable in research settings (Darke, 1998; Ross, et al., 2003). However, it is difficult to draw firm conclusions about pathways and causal relationships from a cross-sectional study. Finally, lifetime diagnoses were used so it is not possible to distinguish between those with concurrent and non-concurrent substance dependence diagnoses.
