Pathways to drug prevalence estimation: synthesizing three comments on triangulation

McKetin's story of discordant trends in Australian methamphetamine series is a great example of the type of paradox that motivated us to write the original paper [1, 2]. Hospitalization, treatment admissions, seizures and arrests all rose even as prevalence estimates from Australia's National Drug Strategy Household Survey (NHS) fell. Efforts to harmonize all of that dissonant evidence led to the plausible but potentially spurious assumption that a smaller number of people were consuming a more dangerous form of the drug. There was minimal consideration of the possibility that the consumer base grew or changed in ways that the NHS missed. Triangulation needs to be brutally unsentimental about the limitations of the various indicators, in this case a general population survey's (GPS's) weakness at describing stigmatized, low-prevalence activities such as methamphetamine use.