Synthetic Data for Workflow Utility in Clinical Trials
Rapid Research Brief

Executive Summary

Key Takeaways

• The JCTS conversation is primed. — Taekman (2026) and respondents Dadiz/Guillet have established an active editorial discourse on simulation for trial preparation in JCTS — the exact target venue.
• Synthetic data methods are mature but misaligned. — TrialSynth (Gao et al., 2024) and others achieve high-fidelity synthetic clinical trial data, but all optimize for clinical realism, not workflow development utility.
• Pipeline and workflow papers exist at the margins. — DosiTest (Kayal et al., 2022) and Petalcorin (2026) validate clinical workflows with simulated data, but target analytic outputs, not pre-data-lock code development.
• The workflow utility concept is novel. — No paper articulates the specific value proposition: synthetic data need not mimic patients perfectly — it needs to exercise the same code paths, data structures, and analytical logic.
• JCTS is the natural home. — Four of the nine most relevant papers appeared in JCTS; no other journal shows comparable topical alignment.

Key Findings

1 The JCTS Conversation — Simulation for Trial Preparation

Taekman (2026) published "On the prior use of high-fidelity simulation to improve clinical trial development and implementation" in JCTS, arguing for simulation exercises conducted before trial launch to surface operational and design flaws. Guillet and Dadiz (2026) responded in the same journal, engaging substantively with the proposal. Earlier, Dadiz, Jones, and Guillet (2024) described the practical use of simulation during preparation for a neonatal antiseizure medication RCT, documenting workflow challenges discovered through scenario-based exercises [Sources 1–3].

This conversation establishes that JCTS editors and reviewers are actively receptive to methodological papers about pre-trial preparation. However, all three papers focus on operational simulation (mock patient scenarios, role-playing) rather than data-centric preparation (synthetic datasets for pipeline development).

2 Synthetic Data Generation — Mature but Misdirected

The technical capability to generate synthetic clinical trial data is well-established. TrialSynth (Gao et al., 2024) introduced a VAE + Hawkes Process architecture achieving state-of-the-art fidelity for time-sequence clinical trial data [Source 5]. Azizi et al. (2021) validated in BMJ Open that synthetic data can serve as a statistical proxy for real clinical trial data (94 citations) [Source 7]. Mangino et al. (2026) provided a JCTS tutorial on GANs for synthetic clinical data [Source 4]. Zazzetti et al. (2025) demonstrated longitudinal synthetic data for breast cancer research [Source 8].

Critically, every paper in this cluster evaluates synthetic data by how closely it mimics real patient data — the "clinical mimicry" paradigm. None frames synthetic data as a tool for exercising analysis infrastructure where perfect fidelity is unnecessary.

3 Pipeline & Workflow Validation — The Closest Neighbors

Two papers approach the workflow concept from adjacent angles. DosiTest (Kayal et al., 2022) used Monte Carlo modeling to simulate a virtual multicentric clinical dosimetry trial, evaluating the clinical workflow pipeline itself rather than patient outcomes — the closest methodological analog [Source 9]. Petalcorin (2026) described a reproducible health informatics pipeline for simulating oncology clinical trial data [Source 6].

Both papers validate infrastructures for data processing and analysis. However, DosiTest evaluates an existing clinical workflow for uncertainty quantification, and Petalcorin targets exploratory decision-support analytics — neither addresses the pre-data-lock code development use case.

4 The Gap — Workflow Utility vs. Clinical Mimicry

Synthesizing across all three clusters reveals a clear and specific gap. The literature recognizes: (a) simulation can improve trial preparation (Taekman, Dadiz/Guillet); (b) synthetic data can be statistically valid (Azizi, TrialSynth, Mangino); (c) pipelines can be validated with simulated data (DosiTest, Petalcorin).

But no paper connects these insights to propose synthetic data as a workflow-utility substrate. The specific concept — generating synthetic RCT datasets that exercise the same analytical code paths as the real data, enabling full pipeline development and validation before data lock without touching PHI — does not appear.

Unlike the clinical mimicry paradigm (which demands high distributional fidelity), the workflow utility paradigm only requires structural fidelity: matching variable names, data types, visit structures, missingness patterns, and analysis metadata. This is a fundamentally lower bar that may enable simpler generation methods.

5 JCTS as the Venue

The Journal of Clinical and Translational Science is the natural home for this work:

• 4 of 9 relevant papers appear in JCTS [Sources 1–4]
• The simulation-for-trial-preparation conversation is actively unfolding there
• JCTS has an upcoming special issue on AI in clinical research
• The journal's scope — translational methodology — aligns precisely

Risks, Gaps & Uncertainty

• Novelty depends on framing. — The concept must be carefully distinguished from DosiTest (pipeline validation) and the Taekman/Dadiz conversation (operational simulation). The specific claim — synthetic data as a PHI-avoiding substrate for pre-data-lock code development — must be articulated precisely.
• Preprint risk for Petalcorin and TrialSynth. — Both are preprints; peer-reviewed versions could appear during manuscript preparation.
• Citation graph incomplete. — Semantic Scholar rate limits prevented exhaustive forward/backward citation tracking on Taekman and DosiTest.
• JCTS special issue deadline unknown. — Confirm submission timeline and AI theme alignment.
• Medium-High confidence. — The search was broad but not exhaustive; niche proceedings and gray literature may contain related work.

Recommended Next Actions

Annotated References