Why Submitted Samples Do Not Describe an Entire Market

Overview

No. A database made up of submitted samples cannot be taken as a representative snapshot of all material circulating in a market. Health Canada’s Drug Analysis Service (DAS) publishes results from samples it receives, but those data reflect what was submitted, how it was tested, and what was reported — not the total universe of products, batches, or contaminants in circulation. Conclusions about overall prevalence, average purity, or “market-wide” safety go beyond what these data can support and must remain sample-, method-, matrix-, and date-specific .

What the DAS data are — and are not

Key reasons submitted-sample databases can mislead when generalized

1) Submission bias — who and what gets sent

2) Selective quantification and method-dependent detection

3) Incomplete contaminant and adulterant reporting

4) Temporal, matrix, and batch specificity

Measurement uncertainty and the role of sampling theory

A practical evidence-reading approach for DAS data

Conclusion: keep conclusions tightly scoped Health Canada’s DAS provides valuable, verifiable analytical results for submitted samples, but those results are sample-, method-, matrix-, and date-specific and cannot alone support claims about overall market prevalence, average purity, or population-level safety. Any inference beyond the tested submissions requires explicitly designed sampling, transparent accounting for measurement uncertainty, and acknowledgement of the DAS dataset’s submission-driven limitations .

Required caveat: All conclusions must be treated as specific to the samples analysed, the methods used, the matrices tested, and the dates of analysis. No single result proves sterility, safety, efficacy, Canadian authorization, chain of custody, or batch-wide uniformity.

• The DAS data are a record of analyses performed on items that were voluntarily or mandatorily submitted to the service from various sources (law enforcement, provincial agencies, coroners, partners, etc.). The database documents the analytical outcomes for those submitted items, along with methodological limits and reporting practices .
• These entries provide useful, concrete observations about particular submissions: what was detected in that sample, by which methods, and within what measurement uncertainty. They are valuable for case-level investigations, trend signals, and methodological transparency .
• They are not a probability sample of a market. The DAS does not (and is not designed to) collect random, stratified, or otherwise statistically representative samples across user populations, production lots, distribution networks, or geography in a way that would support unbiased prevalence estimates .
• Submissions are not random. Items come from specific reporting pathways (e.g., targeted law enforcement seizures, clinical incidents, or partner requests). That creates systematic over- or under-representation of certain types of material: seized shipments may over-represent large-scale distribution, while clinical submissions may over-represent adulterated or harmful items tied to adverse events .
• Because submission is often driven by suspicion, harm, or enforcement priorities, a database will tend to concentrate on unusual, illicit, or problematic samples rather than the typical, day-to-day products that never come to DAS attention. Therefore, frequency in the database is not frequency in the market.
• What is reported depends on the analytical scope applied to each submission. Some analyses report qualitative presence/absence; others quantify specific analytes. Different methods have different limits of detection, accuracy, and uncertainty, and not every submission is analysed with every method .
• A compound not reported in a given submission can mean: it was absent above the method’s detection limit, it wasn’t targeted by that analysis, or it was present but below quantified thresholds. Without uniform, pre-specified testing across all samples, absence in the database cannot be equated to absence in the market.
• DAS reports focus on the compounds chosen for analysis and the matrices submitted. They may not include every possible contaminant (e.g., low-level impurities, unknowns not screened for, or degradation products) because exhaustive screening is rarely feasible for every sample .
• Some contaminants require specialized methods or confirmatory testing. If those aren’t applied, the database will under-report those classes of contaminants. Thus database silence on a contaminant class should be read cautiously.
• Samples are dated and associated with particular matrices (tablets, powders, liquids) and collection contexts. Any inference must be anchored to those attributes: a finding in one pill, one batch, one location, or one time period does not generalize to all pills, batches, locations, or times .
• Markets change. A sample collected months ago may no longer reflect current supply chains, formulations, or contaminant profiles. The DAS documentation itself stresses the need to treat results as specific to the sample and time tested .
• Analytical measurements carry uncertainty. Technical guidance on sampling and uncertainty emphasizes that both measurement variability and sampling design affect how (and whether) results can be generalized beyond the tested item . Without probabilistic sampling and an accounting for measurement uncertainty across a population of interest, extrapolation is unsupported.
• EURACHEM/CITAC guidance underlines that uncertainty arises from sampling strategy as much as from laboratory measurement. When the sampling frame is the set of submitted items rather than a defined market sample, the uncertainty about any market-level estimate is effectively unquantifiable .
• Read each entry as a confirmed observation about that submission: what was found, by which method, and with what limits and qualifiers. Treat these as case-level data, not prevalence statistics .
• Check the metadata: source of submission, collection context, matrix, analytical methods, detection limits, and date. Those details determine the population to which the finding might plausibly apply (often a very narrow one) .
• Look for consistent patterns across multiple, independently collected datasets before inferring trends. Repeated detection of the same compound in diverse, independently sourced submissions over time may warrant further, targeted, statistically designed studies — but repeated DAS entries alone do not constitute market-wide prevalence proof .
• Be explicit about unresolved questions. For example: Which parts of the market were never sampled? Were certain analytes excluded from testing? How might submission pathways bias the dataset? State these gaps when discussing implications.