Causal consulting for pharma
Better study interpretation for pharmacological evidence that comes with real-world constraints.
Causality Graphs works with pharma, clinical research, and biostatistics teams to map assumptions, review confounding structure, and support evidence decisions using directed acyclic graphs and dynamic causal models.
Assumption mapping
Make confounding, mediation, and selection logic explicit before interpretation hardens.
Dynamic modeling
Bring time, feedback, and evolving treatment response into the causal picture.
Decision support
Support evidence strategy when controls are limited, partial, or operationally imperfect.
Built for technical stakeholders
Built for clinical teams, biostatistics groups, translational medicine, and evidence strategy leaders.
The approach is designed for research environments where methodological rigor, internal alignment, and practical decision-making all have to coexist.
What we do
Study interpretation, assumption mapping, and evidence strategy for non-ideal designs.
Causal study design support
Frame the causal question, clarify the estimand, and make design tradeoffs visible before analysis choices become default assumptions.
DAG review and modeling
Build and review graph-based representations of exposure, outcome, mediation, and confounding pathways to support defensible interpretation.
Dynamic causal analysis
Model systems that evolve through time when treatment response, adaptation, feedback, and pathway timing cannot be ignored.
Evidence strategy guidance
Translate causal reasoning into practical recommendations for study interpretation, internal alignment, and next-step evidence decisions.
Why it matters
When trial conditions are imperfect, decision risk sits in the structure as much as the numbers.
Many pharmacological studies operate with limited controls, partial blinding, evolving treatment pathways, or observational contamination.
Standard summaries can look convincing while still hiding confounding, collider bias, selection effects, or pathway ambiguity.
Explicit causal structure makes tradeoffs discussable, assumptions inspectable, and interpretation more defensible across teams.
Techniques preview
Methods chosen to answer causal questions, not to decorate an analysis plan.
Directed Acyclic Graphs
Map assumptions, identify adjustment sets, and make hidden structure legible.
Useful when teams need a shared causal language before modeling begins.Dynamic causal models
Represent how biological and treatment systems evolve rather than treating time as a nuisance.
Useful for longitudinal response, feedback, adaptation, and mechanistic interpretation.Counterfactual reasoning
Anchor interpretation in explicit what-if contrasts instead of loose correlational claims.
Useful when decision-makers need clarity on treatment effect questions under constraints.Sensitivity analysis
Stress-test conclusions against unmeasured bias, structural uncertainty, and model dependence.
Useful when evidence quality is limited but choices still need to be made.Selected work
Representative consulting situations where explicit causal reasoning changed the recommendation.
Single-arm oncology signal review
Problem -> method -> impactChallenge: A promising response pattern was difficult to interpret without a concurrent control.
Method: DAG refinement plus explicit counterfactual framing around likely confounding and selection processes.
Outcome: The study team gained a clearer interpretation boundary and a more credible next-evidence strategy.
Longitudinal treatment response mapping
Problem -> method -> impactChallenge: Dose changes, dropouts, and symptom dynamics blurred the treatment story over time.
Method: Dynamic causal modeling to separate temporal structure, pathway timing, and evolving response states.
Outcome: The resulting model supported better reasoning about progression, timing, and endpoint relevance.
Partial-blinding evidence interpretation
Problem -> method -> impactChallenge: Operational realities introduced expectation effects and outcome interpretation risk.
Method: Structured causal assumptions, mediation review, and sensitivity framing for interpretation robustness.
Outcome: Leadership received a cleaner account of what could be claimed and what required caution.
Philosophy
The work is collaborative, assumption-aware, and designed to hold up across teams.
Make assumptions explicit before they become invisible defaults.
Separate structural signal from statistical convenience.
Use causal thinking to improve decisions, not just analysis complexity.
Team snapshot
Dr. Elena Maris
Founder and Causal Strategy Lead
Jonas Vale
Dynamic Systems Specialist
Mira Sol
Evidence Interpretation Consultant
Start a conversation