Analytical Flexibility and the Crisis of Research Integrity: An Institutional Audit of the British Columbia Centre on Substance Use (2008–2025)

Background:The reproducibility crisis has emerged as a defining challenge in contemporary science, with mounting evidence of questionable research practices across disciplines. Substance use research, which directly informs life-and-death policy decisions for marginalized populations, remains particularly vulnerable to analytical flexibility that can compromise evidence quality. This audit evaluated the consistency, transparency and methodological rigour of a sub-set of quantitative, regression-based studies conducted by the British Columbia Centre on Substance Use (BCCSU) between 2008 and 2025.Methods:We systematically reviewed 85 peer-reviewed publications affiliated with BCCSU researchers employing cross-sectional data from major cohorts (VIDUS, ACCESS, ARYS). Studies were assessed for pre-registration status, variable selection methods, transformation justifications, sampling timeframe rationale, missing data handling and data/code transparency. Structured metadata extraction enabled quantification of researcher degrees of freedom across the institutional portfolio.Results:Zero studies (0%) were pre-registered on public platforms. Fifty-one publications (58%) relied on automated stepwise selection methods (AIC/QIC minimization) rather than theory-driven covariate specification (3.4%). Critical methodological decisions lacked transparency: 36% of studies employed unjustified variable transformations of drug use measures; 34% of studies did not report missing data handling approaches; 66% provided no justification for sampling start dates; 69% provided no justification for sampling end dates. These researcher degrees of freedom produced pervasive analytic heterogeneity that rendered the evidence base largely unsynthesizable.Conclusions:The findings reveal systemic patterns that reflect institutional incentive structures prioritizing publication quantity over methodological rigour. Widespread reliance on data-driven model selection creates a high-risk environment for p-hacking and other questionable research practices. Radical reform via preregistration mandates, standardized reporting templates and annual integrity metrics is required to restore public trust and produce policy-actionable science capable of serving vulnerable populations.