For CROs, HEOR & evidence-synthesis consultancies
Comparator analyses that hold up, client after client.
Locked recipes and two-person sign-off make the comparator analysis defensible client after client — with an inspectable statistical engine underneath every deliverable.
Template protocols · two-person sign-off · inspectable engine
Random-effects pool · primary outcome6 studies
Smith 2024
0.62 [0.41, 0.94]
Chen 2023
0.71 [0.55, 0.92]
Patel 2024
0.83 [0.62, 1.11]
Garcia 2023
0.74 [0.58, 0.95]
Okafor 2024
0.88 [0.71, 1.09]
Müller 2022
1.05 [0.78, 1.41]
Pooled (RE)
0.78 [0.65, 0.93]
← favours interventionfavours control →
study estimate (size ∝ weight)pooled (RE)
RR, log scaleThe shift
From bespoke every time to portable and defensible.
Same workflow, two states. Flip the switch to see what changes.
Delivered once · frozen on handover Live · re-runs for the client
Each engagementRebuilt from scratch, artisanalA locked recipe, reusedbespokeportable
Comparator analysisA one-off deliverable, then frozenKept current for the clientone-offliving
Audit answerReconstructed hours after the askCaptured at screen-time, queryableafterinstant
Static — a snapshot at deliveryKept current across the engagement
How it works
One engine, every client deliverable.
The same build-once, keep-live, inspect-everything mechanism — wrapped in the roles, sign-off, and audit trail a client deliverable needs.
Build
Build the comparison once
- A systematic review and a network meta-analysis
- Anchored against every relevant comparator
- Defensible from the first submission
Keep live
Re-runs on a schedule
- The whole pipeline re-runs on a cadence
- Flags when a new trial moves the result
- A signed record of what changed, each cycle
Inspect
No black box
- Deterministic statistics — same inputs, same result
- Every estimate traces to its source study
- Change an assumption and re-run
Why trust it
Deliverables that survive a sponsor audit.
Validation
We reproduced a published forest plot — exactly.
In our reproduction case study, Axelium rebuilt a published meta-analysis end to end and matched its pooled estimate. Same inputs, same numbers — because the statistical engine is deterministic.
Comparisons a committee will accept
Pairwise and network meta-analysis, plus population-adjusted indirect comparisons, with certainty grading — every step inspectable.
Network meta-analysisMAICSTCML-NMRGRADE certainty
Questions