Biostatisticians

A new generation of AI-powered statistical tools — including Julius AI, ChatGPT's Advanced Data Analysis, Posit's Shiny AI modules, and embedded analytics in REDCap and Medidata Rave — is enabling clinical researchers, epidemiologists, and data managers to perform power calculations, generate Kaplan-Meier curves, run logistic regressions, and produce APA-formatted results tables without biostatistician involvement. This is not a future risk: NIH-funded research teams are actively using these tools for preliminary analyses and grant power calculations, and academic medical centers are deploying self-service statistical platforms that route only complex analyses to human biostatisticians. The demand contraction operates independently of whether individual biostatisticians adopt AI tools — even a highly productive AI-augmented biostatistician cannot generate enough new demand to offset the consultations being rerouted to self-service.

GitHub Copilot, Claude Code, and Cursor are now generating production-quality SAS and R code for the standard biostatistical analysis suite — PROC MIXED, PROC LIFETEST, PROC GENMOD, PROC LOGISTIC, lme4, survival, mice, geepack — from plain-language prompts with sufficient reliability that pharma companies are evaluating them for use in validated computing environments. More significantly, pharmaverse (an open-source consortium including Roche, Novartis, GSK, Pfizer, Johnson & Johnson) has released the admiral package for CDISC ADaM dataset construction and rtables/teal for TLF generation, effectively standardizing and templating the code that biostatistical programmers write. When LLMs are combined with these pharmaverse templates, the generation of standard analysis code for a typical Phase III trial can be largely automated. Hiring data from CROs including ICON and Parexel shows statistical programmer job postings declining 20–30% over 2023–2025.