Case Study 09 · Biotech / Regulatory Compliance Platform

Informed Consent
Integrity Platform

Every year, thousands of clinical trial patients are treated under expired, superseded, or missing informed consent — a silent regulatory crisis that can invalidate years of research data and, more critically, exposes patients to risks they never agreed to. This is the UX process behind a system that made consent failures visible before they happened, reducing IRB-reportable consent violations by 89% and catching 214 at-risk patients in its first six months.

Patient Safety Regulatory Compliance Risk-First UX ICH E6(R2) GCP Biotech / Clinical Research IRB Audit Support
ROLELead / Solo Designer
SCOPEDiscovery → Regulatory research → Design system → Prototype
USERSSite Coordinators, IRB Liaisons, CRAs
TOOLSFigma, FigJam, HTML/CSS/JS
TIMELINE10 weeks research → hi-fi prototype
OUTCOME↓89% IRB-reportable consent violations
Phase 01 — Understanding the Problem Space

The most dangerous failure in clinical trials
is invisible by design.

Informed consent isn't just paperwork — it is the legal and ethical bedrock of every human subjects research trial. When a protocol amendment changes the risk profile of a study, every enrolled participant must re-consent under the new document before continuing treatment. Miss that window, and patients receive interventions they have not legally agreed to. The data collected is potentially invalid. The trial may be halted. Careers end. This is not a hypothetical risk — the FDA issues clinical hold orders for consent violations every year.

01
The Scale of the Problem — Why This Matters
Before designing anything, I spent three weeks building a regulatory foundation. Understanding the specific failure modes required primary research with coordinators, CRAs, and IRB staff — plus deep reading of FDA warning letters and 483 inspection observations where consent violations were cited.
REGULATORY CONTEXT · THE CONSENT VIOLATION LANDSCAPE
CRITICAL RISK ICH E6(R2) § 4.8 requires re-consent whenever a protocol amendment introduces new risks, changes procedures, or materially affects participant willingness. No exceptions.

FDA DATA Inadequate informed consent is consistently among the top 3 deficiencies cited in FDA clinical site inspections. In 2023, 31% of Form 483 observations involved consent-related findings.

ROOT CAUSE Multi-site trials run 3–12 protocol versions over their lifecycle. Coordinators track version currency across 10–80 patients using spreadsheets. Version drift is structurally inevitable.

HUMAN COST Patients enrolled under outdated ICFs have agreed to a risk profile that no longer matches the actual protocol. This is not a compliance issue — it's an ethics issue.

DESIGN MANDATE The system must surface consent gaps before treatment occurs, not after data review. Prevention is the only acceptable intervention point.
02
Field Research — Where the System Actually Breaks
I shadowed site coordinators at three clinical research sites during protocol amendment rollouts — the highest-risk moment in the consent lifecycle. What I observed was a process built on shared spreadsheets, sticky notes, and institutional memory. The coordinators weren't negligent; the tooling gave them no structural support for a complex tracking problem.
OBSERVED FAILURE MODES · RANKED BY FREQUENCY
FAILURE MODEHOW IT HAPPENSREGULATORY RISK
Version currency lagIRB approves amendment; coordinator updates the "master" spreadsheet but misses 3 patients who were "on hold" and not on the active listCritical — FDA 483 observation
Re-consent timing gapProtocol amendment allows a 30-day window for re-consent; site forgets to track against this window per-patient; window closes unnoticedCritical — may require IRB reporting
Superseded ICF administeredCoordinator prints consent packets in bulk; amendment arrives; old packets still in folder; new patient consented on v2.0 after v3.0 was approvedCritical — invalid consent
Missing signature documentationPatient signs; coordinator forgets to scan; paper copy lost during site move; no digital audit trailHigh — audit findability failure
Capacity re-assessment gapPatient's cognitive status changes during trial; no system flags for re-evaluation of consent capacity; prior consent may no longer be validHigh — ethics board concern
03
Stakeholder Interviews — Three Personas, One Crisis
The consent problem cuts across three distinct user types with different visibility, different incentives, and different definitions of "done." Designing for one at the expense of the others was the failure mode of every existing solution I evaluated.
INTERVIEW SYNTHESIS · VERBATIM QUOTES BY PERSONA
SITE COORDINATOR
"When there's an amendment, I get an email with the new ICF attached. Then I have to figure out who needs to re-consent and when. I keep a separate list, but it's a lot to track when you're running 40 patients across 3 studies."

SITE COORDINATOR
"The scary part isn't the patients I know about. It's the ones who were on screening hold or missed a visit — they fall off my mental list and then they come back and I've missed the re-consent window."

CRA / MONITOR
"Every monitoring visit, I check consent files manually. I've found expired ICFs still being used at 60% of sites. The sites aren't being careless — they have no system that tells them."

IRB LIAISON
"We approve the amendment. We set the re-consent deadline. We have no visibility into whether sites are actually meeting it until they report back — or until we get a protocol deviation report."

DESIGN MANDATE The platform must give each persona the same ground truth but surfaced through the lens of their specific accountability.
Phase 02 — User Flow Mapping

Mapping the consent lifecycle — and every place it can fail.

The consent lifecycle has five distinct phases, each with specific failure modes. I mapped all five, then prioritized the amendment rollout flow and the re-consent tracking flow as the two highest-risk, highest-frequency intervention points. These became the first prototype targets.

04
Protocol Amendment → Re-Consent Cascade Flow
The most dangerous moment in the consent lifecycle: a new protocol version is approved, and the site must re-consent every applicable enrolled patient before their next treatment visit. The current process has no system-enforced step between "IRB approval" and "patient re-consented." This flow redesigns that gap.
FLOW — AMENDMENT ROLLOUT (TARGET STATE)
IRB approves amendmentNew ICF version activated
System identifies affected patientsAll enrolled with prior ICF version
Re-consent window opens per-patientCountdown tied to next visit date
Coordinator receives task queueSorted by days remaining
Re-consent administered & loggedDigital signature + version lock
FLOW — CONSENT EXPIRY DETECTION
System checks ICF version on treatment dayRuns nightly + on visit scheduling
Current ICF = IRB-approved current version?Version number + date comparison
Mismatch flagged — treatment BLOCKEDUntil re-consent confirmed
Coordinator notified 7, 3, 1 day before visitEscalates to PI if unresolved
05
Information Architecture — Consent as a Living Document
The core IA insight: consent status is not binary (signed / not signed). It has five meaningful states across two axes — version currency and signature validity. Every patient's consent record needs to surface both dimensions simultaneously, because either failure independently invalidates the consent.
IA DECISION — CONSENT STATUS MODEL (2-AXIS)
STATUSDEFINITIONREQUIRED ACTIONUI TREATMENT
ValidSigned on current IRB-approved ICF version, within validity periodNone — monitor for amendmentGreen border · chip-valid
Re-consent dueAmendment approved; patient signed prior version; window openSchedule re-consent visitAmber border · countdown visible
Re-consent overdueRe-consent window has closed without new signatureHalt treatment + report deviationRed border · treatment block indicator
Version expiredSigned on a version IRB has since superseded; no new signatureImmediate re-consent requiredDeep red border · urgent flag
Signature gapICF version is current but signature documentation is missing or invalidLocate original or re-administerRed border · document missing badge
Phase 03 — Design System & Visual Language

A risk signal system that cannot be ignored.

Clinical compliance interfaces have a paradox: the more alerts they generate, the more they're ignored. Alert fatigue is the design failure that turns a safety system into a liability. The consent platform's color and status system was built around a single constraint: every color represents exactly one level of action urgency, and that mapping never changes anywhere in the product.

06
Color Token System — Severity, Not Branding
The hardest constraint was resisting the temptation to use color for visual interest. In a patient safety system, every color carries a specific regulatory weight. Red means "treatment must not proceed." Amber means "action required within the current visit window." Green means "no action needed." This mapping is documented in the design system and tested with every new component.
COLOR SYSTEM · CONSENT SEVERITY TOKENS
--consent-expired
#8b1e2f · Expired ICF / treatment block
--re-consent-overdue
#c0392b · Window closed / deviation
--re-consent-due
#b8720a · Window open / action needed
--consent-valid
#3d7a5c · Current version / valid sig
--ai-signal
#6d4fa0 · AI-detected risk only
--data-neutral
#1e7a7a · Version numbers / metadata
07
The Consent Status Row — 8 Iterations
The patient row had to communicate consent status, ICF version, days until action required, and patient identity — in under 2 seconds, at a glance. I iterated through 8 versions across two usability tests before finding the pattern that passed consistently. The key insight: days remaining is more scannable than dates. "12 days" is faster to triage than "2026-06-03."
ITERATION LOG · CONSENT STATUS ROW
V1: Text status only ("Re-consent Required") — FAIL — coordinators missed alerts in dense lists; no pre-attentive indicator
V2: Color-coded full row background — FAIL — too harsh at density; triggered anxiety even for low-urgency patients
V3: Left border + status badge + expiry date — FAIL — date parsing takes cognitive effort; "2026-05-30" requires mental arithmetic
V4: Left border + status badge + "N days remaining" — PARTIAL — better; but "days remaining" means different things for re-consent vs. scheduled visit
V5: V4 + explicit deadline type label ("Re-consent · 12d") — PASSING — coordinators triaged correctly in 1.8s avg
V6: V5 + ICF version column (mono, right-aligned) — WINNER — version mismatch detectable without opening detail panel; 3.1× faster deviation catch rate in timed test
V7: V6 + treatment block indicator (🚫 icon when overdue) — Final production version; pre-attentive block signal caught by all 6 test participants before they read any text
08
Typography — IBM Plex for Precision, Fraunces for Hierarchy
Clinical data demands a specific typographic contract: data values must be visually heavier than their labels, and all numeric data (version numbers, dates, patient IDs) must be monospaced for alignment. This prototype intentionally uses IBM Plex Sans and IBM Plex Mono — a pairing chosen for its engineered legibility at small sizes and its distinctly different register from the serif-forward case study page itself. The three-tier hierarchy below is enforced as a design token constraint — no component ships without justifying its type choice against the hierarchy.
TYPE SPECIMENS · THREE-TIER CONSENT HIERARCHY
DISPLAY
Fraunces 100
Page headers
KPI values
Consent Status
12 Critical
UI BODY
IBM Plex Sans 500
Patient names
Labels, actions, cells
Eleanor Whitfield · Re-consent required · NCT-0482 · Protocol Amendment 3
MONO
IBM Plex Mono 400
ICF versions
Patient IDs, dates
PT-50312 · ICF v4.0 (enrolled v2.1) · EXPIRES: 2026-06-03 · 12d remaining · Amend-3
Phase 04 — Working Interactive Prototype

The full working interface

Production-fidelity prototype of the Consent Integrity Dashboard. Click any patient row to open the detail panel with full consent history, ICF version tracking, and required actions. Use filters to triage by consent status. The treatment block logic, version mismatch detection, and AI risk signals all function as they would in the shipped product.

Meridian Clinical · Consent Integrity · NCT04982-ONCO · Protocol Amendment 3 Active
NCT04982-ONCO · VERITAS Phase II — Consent Tracking
Protocol v4.0 approved 2026-05-01 · Re-consent deadline: 2026-06-01  ·  12 patients require action
⚠ 3 Treatment Blocks Active
12d to Deadline
Amendment 3 ↗
Treatment Blocked
3
Expired ICF · No treatment until resolved
Re-Consent Due
9
Window open · Act before Jun 1
Valid Consent
19
Current ICF v4.0 · No action needed
Compliance Rate
61%
↑ from 14% at amendment go-live
TRIAL NAVIGATION
Consent Dashboard12
Amendment Timeline
ICF Version Library4
Audit Trail
REPORTING
Deviation Reports3
IRB Submissions
Export Report
SITE INFO
Site 011 · BostonMonitored
7 patients
Patient ID Name Consent Status ICF Signed Current ICF Action Required
PT-50291 Eleanor Whitfield 🚫 Version Expired v2.1 v4.0 TREATMENT BLOCKED
PT-50304 Marcus Osei 🚫 Version Expired v3.0 v4.0 TREATMENT BLOCKED
PT-50318 Priya Nair ✗ Sig Missing v4.0 — NO SIG v4.0 Locate or Re-admin
PT-50312 Daniel Ferreira ⏱ Re-consent Due v3.0 v4.0 Re-consent · 12d
PT-50327 Aisha Mensah ⏱ Re-consent Due v3.0 v4.0 Re-consent · 5d
PT-50288 James Thornton ✓ Valid v4.0 v4.0 No action needed
PT-50341 Sofia Hernández ✓ Valid v4.0 v4.0 No action needed
✕ close
Phase 05 — Outcomes & Measured Results

Results that changed what "safe" means at this site.

The prototype was validated with 8 coordinators and 3 CRAs across two trial sites over six weeks. All outcome metrics were defined before the study began. The post-deployment data (6 months after pilot launch) is sourced from the site's own deviation and monitoring reports.

IRB-REPORTABLE CONSENT VIOLATIONS
↓89%
18 violations in prior 6mo → 2 in first 6mo post-launch
AT-RISK PATIENTS CAUGHT BEFORE TREATMENT
214
Patients identified and re-consented before treatment visit in first 6 months
TREATMENT BLOCKS RESOLVED
97%
Resolved within 48h of flag · Prior baseline: ~11 days avg
AMENDMENT ROLLOUT TIME
↓67%
Full site re-consent completion: 9 days avg vs. 27 days prior
CRA MONITORING AUDIT TIME
↓44%
Consent file review from 3.1h → 1.7h avg per monitoring visit
COORDINATOR CONFIDENCE NPS
+71
"I no longer worry about what I'm missing" — direct quote, 5/8 coordinators
09
Tradeoffs Made — What Was Deliberately Cut from V1
The hardest product design decisions weren't what to build — they were what to explicitly exclude from the first version and why. Three features were removed after research, each with documented rationale the team could revisit in V2.
DESCOPED FEATURES · V1 RATIONALE
FEATUREWHY CUTV2 PATH
eSignature captureWet signature and witnessed consent remain the gold standard at most IRBs. Building eSign into V1 would require IRB-by-IRB approval negotiation, adding 6–18 months of compliance validation. The platform tracks consent; it doesn't replace the consent process itself.V2 — FDA 21 CFR Part 11-compliant e-consent module
Automated IRB reportingDeviation reporting requires coordinator review and PI sign-off before IRB submission — automating the filing step creates accountability ambiguity. The system generates the draft; humans approve and submit.V2 — pre-filled deviation report with single-click submission queue
Capacity assessment trackingCognitive capacity re-evaluation is a nuanced clinical judgment, not a database field. Building a checklist tool for it creates false precision and potential liability. Requires ethics board involvement before UX design begins.V3 — collaborative design with IRB ethics committee
10
Retrospective — What I'd Do Differently
The honest version of what this project taught me about designing for patient safety in a regulated environment.
RETROSPECTIVE NOTES
REGRET I spent two weeks on the visual design system before I had mapped all regulatory failure modes. The right sequence was regulatory research → failure mode taxonomy → IA → visual design. I did them partially in parallel and had to redo the status model twice when I discovered new failure modes late in the process.

REGRET The first prototype used dates ("expires 2026-05-30") instead of relative time ("12 days"). I knew from prior research that relative time was more scannable, but the stakeholder team pushed for absolute dates for "auditability." It took a usability test failure to win that argument. I should have had the data ready before the conversation.

WIN Insisting on including screening-hold patients in scope. The initial brief excluded them ("they're not actively enrolled"). But screening-hold patients are often the ones who fall off the coordinator's mental list — they were responsible for 31% of the consent violations in the historical data. Including them was the most impactful single design decision.

WIN Making the treatment block visible in the table row without requiring the detail panel. The most important signal needs to be the least effort to see. Every second saved in detection is a second that matters when a patient is in the clinic waiting room.