See selected operational reports of our studies, the frameworks we use to design them, and our general opinions about clinical research.
Alzheimer's plasma biomarkers: an LAR consent pathway, built up front
Top-tier global diagnostics manufacturer · clinically-confirmed Alzheimer's plasma biomarker collection · 200 subjects (+10 screen failures), up to 5 sites
The challenge the client gave us
A top-tier diagnostics manufacturer is developing a plasma-based Alzheimer's biomarker assay. It needs specimens from clinically confirmed Alzheimer's disease patients — confirmed by neurological exam, cognitive/functional testing, and imaging (MRI/CT/PET) or CSF/blood, with MMSE ≤ 24 and age ≥ 60 — not cognitively normal volunteers standing in for patients.
The catch is two-sided. First, cognitive impairment makes informed consent harder: many of these patients cannot legally consent for themselves, so the study only works if a legally authorized representative (LAR) consent pathway is built into the protocol from the start. Second, confirmed Alzheimer's patients are concentrated in specific practices — neurology, psychiatry, geriatric medicine, and some family medicine — not spread evenly across a research-site network.
At a glance
Subjects
~400, split between cognitively normal controls and clinically confirmed Alzheimer's
Neuro exam, cognitive/functional testing, MRI/CT/PET or CSF/blood
Sites
Up to 5, across neurology, psychiatry, geriatric medicine, family medicine
Specimen
Plasma — 3 × 10 mL K2 EDTA per subject, 1 mL aliquots
Cold chain
2–8 °C within 6 hrs; frozen ≤ −70 °C within 48 hrs
Consent
Operationalized LAR consent for cognitively impaired subjects
Timeline
Site targetingClaims-data ranking; highest-density practices first
StartupProtocol + LAR consent + Part 11 stack
ActivationUp to 5 specialty sites opened
EnrollmentMMSE ≤ 24, age ≥ 60; LAR consent; plasma to spec
Delivery~400 subjects; submission-grade set
Site targeting Claims-data ranking; highest-density practices first
Startup Protocol + LAR consent + Part 11 stack
Activation Up to 5 specialty sites opened
Enrollment MMSE ≤ 24, age ≥ 60; LAR consent; plasma to spec
Delivery ~400 subjects; submission-grade set
Fig. 1 Cumulative enrollment (line) and monthly enrollment (bars) across the collection window. The program closed at roughly 400 subjects, split between cognitively normal controls and clinically confirmed Alzheimer’s patients, with monthly cadence peaking mid-collection before the cohort caps approached saturation.
Source: Diagnostica™ at program close. Sponsor identity withheld per confidentiality terms.
What we did as a CRO
CRO responsibility
What RDI did on this study
Site selection & feasibility
Ranked physician practices by AD patient volume using insurance claims data and built the network highest-density first — going where confirmed Alzheimer's patients actually are (neurology, psychiatry, geriatric, family medicine) rather than to the easiest sites to open
Regulatory & study start-up
Built the LAR consent pathway up front and ran a 21 CFR Part 11 CTMS/EDC/eTMF stack so the documentation was submission-grade from day one
Site activation & training
Activated up to 5 sites and partnered with the practices that could actually deliver confirmed, consentable AD patients
Recruitment & enrollment
Confirmed AD by neuro exam / cognitive-functional testing / imaging or CSF, screened to MMSE ≤ 24 and age ≥ 60; for cognitively impaired subjects, coordinated the LAR to be present with the patient so consent and the blood draw happened in one visit
Kits, samples & lab
Collected plasma to spec — 3 × 10 mL K2 EDTA, 1 mL aliquots, 2–8 °C within 6 hrs and frozen ≤ −70 °C within 48 hrs
Monitoring & data
Central monitoring on the Part 11 stack; captured confirmation basis and consent pathway per subject
Project management & close-out
Delivered ~400 subjects split between cognitively normal controls and confirmed AD (plus reimbursable screen failures) against the specimen spec
What made it hard
Many subjects can't consent for themselves. Without a working LAR pathway, the protocol simply can't enroll the confirmed-dementia patients the assay is meant for.
The patients are concentrated, not distributed. Confirmed Alzheimer's diagnosis lives in specific neurology, psychiatry, geriatric, and family-medicine practices — a general research-site list won't reach them efficiently.
Confirmation is demanding. MMSE ≤ 24 and age ≥ 60 is only the screen; clinical confirmation rests on neuro exam, cognitive/functional testing, and imaging or CSF.
The specimen spec is strict — K2 EDTA plasma on a tight cold-chain clock to ≤ −70 °C, where any handling slip costs a hard-to-recruit subject's sample.
How we solved it
We built the LAR consent pathway up front, so the protocol could enroll the patients the sponsor actually needed instead of stalling on capacity-to-consent.
We coordinated the legally authorized representative to be there with the patient. For dementia subjects who can't consent for themselves, we scheduled around the LAR so consent, cognitive confirmation, and the blood draw happened in one trip.
We found the sites through claims-data targeting. Instead of a generic site list, we ranked physician practices by Alzheimer's patient volume from insurance claims and went highest-density first — then partnered with the practices that could actually help deliver confirmed, consentable patients.
We ran everything on a 21 CFR Part 11 stack, so the specimens arrived with submission-grade documentation attached.
The result
A turnkey model for diseases that affect a patient's capacity to consent: claims-data site targeting to find the concentrated patient population, plus operationalized LAR consent to actually enroll them. The collection delivered ~400 subjects split between cognitively normal controls and clinically confirmed Alzheimer's (plus reimbursable screen failures) across up to 5 specialty sites, on a submission-grade plasma specimen set.
Source: RDI program scope and specimen specification; Diagnostica™ at program close. Sponsor identity withheld per confidentiality terms; site labels anonymized. Operations report, not a clinical-results publication.
Operations report
An academic pediatric ER site, activated in 85 days
Site live 85 days after the intro call, on a path that normally runs 3–6 months. RDI's scope was deliberately narrow — find the site, make the introduction, own the contract and payments — powered by its peer-to-peer MD network in hospital critical care. The engagement later grew to ~5× its original value.
The challenge
A global diagnostics leader's point-of-care blood glucose monitor needed analytical-performance evidence from critically ill pediatric patients in both the ER and ICU, per FDA BGMS guidance — one of the hardest site populations to access. The sponsor didn't have the 3–6 months such a site normally takes to open.
Operations report
Chronic Hepatitis B: reaching patients through their physicians
149 subjects across two community clinics, with a 98% qualified rate at first-collection close. These patients are reached through the physicians who already treat them — often across a language barrier.
The challenge
A top-tier diagnostics manufacturer needed chronic Hepatitis B specimens for a serology marker from a population concentrated in specific communities. High sample quality brought them back for a second and third collection.
Operations report
1,240 healthy subjects for a reference range, locked on time
1,240 healthy volunteers from the general population across ~20 sub-cohort quotas — adults by age and sex, pediatric by Tanner stage, pregnant by trimester — collected across 12 sites, cold-chained to the lab, and locked on the sponsor's fixed date.
The challenge
A global diagnostics leader needed reference ranges for a new panel of assays: not patients, but a thousand-plus healthy people, stratified into the cohorts a normal range is built from, delivered as one clean, monitored, locked dataset by a fixed date.
Operations report
Transplant drug monitoring: a tiered site network
244 subjects across 7 transplant centers. Five drugs across heart, liver, and kidney — a matrix of sub-cohorts, each behaving like its own sub-study.
The challenge
The client needed immunosuppressant specimens across five drugs and three organ groups, with draw-time precision, the right tube per drug, and cold chain held from collection.
Operations report
Thyroid reference ranges: the right people, not more people
564 healthy adults across 9 sites. CLSI EP28 defines the demographic mix the reference group must match, so the challenge was reaching the right people at each site, not enrolling more of them.
The challenge
The client needed a CLSI EP28 reference-interval study for a thyroid panel — where the binding constraint is the demographic makeup of the healthy reference population.
Operations report
Ten blood draws per subject in five weeks, with the retention to finish
60 subjects × 10 visits, ~575 specimens through a single site, with a +376-specimen peak month. The study isn't enrolling 60 women — it's keeping all 60 through ten visits so the serial dataset doesn't collapse into dropouts.
The challenge
A top-tier manufacturer needed serial-draw evidence for a progesterone assay across the menstrual cycle: 10 visits per subject, ≥3 days apart over 5 weeks, both serum and LiHep plasma every visit, from healthy women with a narrow eligibility screen.
An assay validation scientist runs our trials.
Structure the CRO so the person who lives in specificity and traceability owns trial execution — not a handoff to generalists.
Right-size the validation study.
The conservative protocol that protects the filing can suffocate enrollment — scoping pre-pivotal work for evidence without theatrical burden.
Qualified site is not the same as a high-enrolling site.
Most CROs select on regulatory qualification; the signal that predicts enrollment is performance in the field, not paperwork.
Monitoring is evidence-gathering.
IVD monitoring as investigation in service of a regulatory story — four principles for what monitoring should produce.
Map the values, then chase them.
Value-bucket IVD recruitment: map the analytical-range distribution to populations, recruit by segment, and recalibrate weekly with the physicians who see the patients.
Patient centricity is a buzzword, not a blueprint.
Patient experience is rarely the bottleneck in IVD trials; standard-of-care-aligned design beats patient-theater checklists.
The enrollment forecast is the only thing that matters.
Daily resolution, per-site positive-rate calibration, milestones on the calendar — updated weekly so enrollment reality stays ahead of the spreadsheet.
We make our own kits.
Kit-making controls enrollment timing, supports sites operationally, and locks in sample quality — three principles for IVD logistics.
The best protocol is one page.
Compress ruthlessly, leave room for biological variance, and write for who actually runs the study — not the filing cabinet.
Kickoff documents are how projects get managed.
The kickoff document is not a meeting agenda — it is the artifact the team operates from for the life of the study.
Good site qualification unravels the physician-patient relationship.
Five principles: the physician–patient relationship is the unit of analysis — and the friction that erodes it is the qualification work most CROs skip.
The sample spec sheet, and why simplicity is a feature.
One document, six tables, and roughly 25 fields — scope alignment that auto-builds the SOW because simplicity is a feature.
Give clients what they need.
Sponsor partnership as guidance — the CRO is neither order-taker nor know-it-all, but the operator who shows the path.
The case report form is a work of art.
Four principles for IVD CRF design — the case report form is the most-used UX surface on a clinical trial. Treat it like one.
The 20-minute SIV.
The SIV is a milestone, not the work. Good site support is a system that runs continuously — before activation, during enrollment, after database lock.
Trial design framework: put physicians first.
Across 300+ IVD trials since 2008, the protocols that recruit on schedule share one property: study activities fit inside the patient visit that was going to happen anyway.
The over-engineered protocol problem.
Pre-pivotal validation protocols often recruit themselves into failure — complexity buys paperwork, not subjects.
The intended use statement is the most expensive line in your study.
Write the intended use before the protocol — getting it backwards has a real cost in what sites can execute and what the filing can claim.
What if I told you the secret to happiness was hard work you're good at?
Eighty thousand working hours — the culture treats them as debt against your real life; the alternative is craft you would pay to keep doing.
AI is the car. The driver is everything.
AI in clinical operations does not produce outcomes — drivers do; people and judgment remain the bottleneck.
The best sites don't have time for you. Cherish the ones that make time, and use it wisely.
A PI who has all afternoon for your diligence call is not a PI with a busy practice — the friction-rich signal is the one worth designing around.
Gantt charts are not project managers.
Real PMs find dry ice for sites at 6 a.m. Gantt charts are tools, not work — what IVD operations leadership actually looks like on the ground.
“Refer to the IFU” is not a clear instruction.
A license to defer specificity: the investigator does not have the IFU at consent, the CRF does not pull from it — the site needs the answer in the protocol.
Use remnant samples sparingly.
Remnants are cheap and fast, but they lack the structured data IVD validation needs — and ICD mining does not put it back.
Build a portfolio of sites the way a VC builds one of companies.
Sites are small businesses; the fix is not only sharper qualification but portfolio construction so one slow site does not sink the program.
A startup's guide to LDT development.
There's a lot of uncertainty in laboratory developed test (LDT) development right now — which rules apply, which pathway to follow. Here's how to build for competitive advantage from day one.