Evaluating an Investigational MRI Contrast Agent for CNS & Body Imaging

Neurological pain disorders—such as neuropathic pain, migraine, and fibromyalgia—remain among the most difficult indications in central nervous system (CNS) drug development. Pain trials are failing not because the science is weak, but because the model of clinical development for neurological pain has not evolved in over two decades.

How Aixial’s imaging and biostatistics teams delivered regulator-ready analyses across adult and paediatric cohorts

Compliance note: This case study intentionally does not name the investigational product or comparator and does not report sponsor results. Content focuses on study design, methods, and CRO value delivery.

Study at a Glance

Populations:

• Adults: 200 participants (including 80 with CNS indications)
• Paediatrics: 40 participants (including 19 with CNS indications)

Design

• Adults: Randomised, double-blind, cross-over
• Paediatrics: Open-label, single-arm

Primary Objectives:

• Adults: Demonstrate non-inferiority of the investigational agent vs the standard agent for lesion visualisation across CNS and other body regions
• Paediatrics: Evaluate added diagnostic value of the investigational agent vs unenhanced MRI

Blinded Evaluation: 3 independent radiologists per body region, pooled as Reader A/B/C, to mitigate single-reader bias

Decision Framework: 10% non-inferiority margin (0.35 on an expected mean of 3.5); non-inferiority declared if all 3 co-primary criteria are met simultaneously for ≥ 2 of 3 readers

Sensitivity & Subgroups: Worst-score imputation for non-matching lesions; alternate reader poolings; CNS vs non-CNS subgroup analyses

Reliability: Intra- and inter-reader ICC to quantify repeatability and agreement

Study Design Overview — Investigational MRI Contrast Agent (CNS & Body)

Section	Adults N=200 (80 CNS)	Paediatrics N=40 (19 CNS)
Design	Randomised, double-blind, cross-over Open-label, single-arm Primary objective Non-inferiority vs. standard agent for lesion visualisation (CNS & other regions)	Added diagnostic value vs. unenhanced MRI for lesion visualisation
Lesion visualisation assessment	3 co-primary criteria assessed on a 4-point scale; patient-level mean across up to 10 largest & most enhancing lesions	3 criteria assessed on a 4-point scale; same lesion-level rules where applicable
Blinded evaluation	3 independent radiologists per body region; pooled as Readers A/B/C	3 independent radiologists per body region; pooled as Readers A/B/C
Reader pooling robustness	Secondary analyses with alternate random combinations of readers
Non-inferiority framework	Margin 10% (0.35 on expected mean 3.5); NI if all 3 co-primary criteria met for ≥2/3 readers	Not applicable (no NI hypothesis; focuses on added value vs. unenhanced)
Sensitivity analyses	Worst-score imputation for lesions not seen with investigational agent; include matching & non-matching lesions	Worst-score imputation approach mirrored where relevant to paediatric objective
Subgroups	CNS (brain & spine) vs. other body regions	CNS (brain & spine) vs. other body regions
Reliability	Intra- and inter-reader variability assessed via ICC (target: good-to-excellent)	Inter-reader variability assessed via ICC (target: good-to-excellent)

Why This Matters to Biotech Sponsors

Lesion visualisation is often the fulcrum for value in CNS imaging trials. Sponsors need confidence that endpoints are clinically justified (margin, scales, lesion multiplicity), reader variability is engineered and measured, analyses are regulator-ready (CDISC, traceable derivations, reproducible TFLs), and the narrative scales into pooled safety/efficacy summaries for streamlined review. Aixial addressed these needs with a design-forward and submission-back approach – planning endpoints and reader strategy up front, while building data flows that map cleanly into global submissions later.

What Aixial Delivered

1) Analysis Strategy & Reader Science

• Lesion-level rules (up to the 10 largest and most enhancing lesions) pre-specified to reflect clinical reality and control multiplicity.
• Blinded, independent readers per body region with pooling to Reader A/B/C; three secondary analyses with alternate reader combinations to demonstrate invariance of conclusions.
• Reliability quantified through intra/inter-reader ICC, with reader training and calibration documented.

2) Non-Inferiority Framework that Stands Up to Scrutiny

• Clear clinical rationale for a 10% margin (expressed on a 4-point scale as 0.35) tied to the practical meaning of ‘no clinically important difference’ in lesion visualisation.
• Decision rule requiring concordance across three co-primary criteria for ≥ 2 of 3 readers, increasing robustness to reader-specific effects.

3) Submission-Grade Standards & Traceability

• PMDA compliant CDISC SDTM & ADaM datasets verified with Pinnacle 21; eCRT packages; fully documented Table Figures and Listings (TFLs).
• Pooled analyses: Integrated summaries for safety and efficacy assembled from multiple studies to support a cohesive submission narrative.
• Audit-ready lineage: Every endpoint and table traceable back to specifications, derivations, and source data.

4) Pediatric-Specific Considerations

• Ethical guardrails (consent/assent), dose selection rationale for contrast administration, and a design suited to clinical feasibility (single-arm vs unenhanced baseline).

Best Practices We Recommend (and Implemented Here)

• Pre-specify sensitivity pathways (e.g., worst-score imputation for lesions not detected by one agent) before any unblinding.
• Engineer reliability: reader calibration sessions, qualification criteria, and ICC thresholds (e.g., good ≥ 0.75).
• Power cross-over designs properly: include within-subject correlation, sequence balance, and carryover mitigation.
• Design for submissions from Day 1: CDISC mapping plans, reusable TFL shells, and pooled-summary strategy to cut rework and shorten review cycles.
• Make the endpoint interpretable: tie the margin and scales back to clinical decision-making, not just statistics.