Choosing a clinical trial management system (CTMS): sandbox-test it against your own standard operating procedures (SOPs) before you switch

Why a demo cannot predict daily fit

A demo shows a system at its best, in the vendor's order, with clean data and no interruptions. Your site does not work that way: a coordinator reschedules a visit while answering a sponsor query, a principal investigator (PI) signs between patients, and finance reconciles a visit completed, partially completed, then repeated.

Those collisions decide fit, and they rarely appear in a demo. A cheaper system that forces workarounds in every window costs more in time and errors than a costlier system that follows your workflow. Which platform runs your work with the fewest detours?

Build a test study from your own SOPs

Create one fictional study in a sandbox environment — a safe, separate test version with no real patient data — constructed entirely from your own SOPs. Do not let each vendor build a different example. Give every candidate the identical study, tasks, and expected results, based on how your site actually operates:

• Visit schedules: Your real structure, windows, and exceptions — screening, a repeat procedure visit, an unscheduled visit, and a visit outside its window, documented that way.
• Invoicing rules: How you actually bill — pass-through costs, procedures paid per completion, holds until source is verified, and when a visit becomes invoiceable.
• Regulatory tracking: Approvals, expirations, renewals, training records, delegation logs, and the alerts you expect before something lapses.
• Roles and permissions: Who can create, edit, approve, and sign, and who must not see or change certain records.

Use fictional patients and sponsors, but keep the complexity real. If your SOPs require a second check before a visit closes, the test must require it too.

At Amavita Research Services, our Miami cardiovascular site, we work from more than 18,000 active cardiology patients and zero Food and Drug Administration (FDA) Form 483 findings — observations an inspector records when conditions may violate regulations. That record comes from procedures followed the same way every time, not software. Any system must support those procedures, not replace them.

Use one sandbox test script for every candidate

Give each vendor the same written script, score results the same way, and have daily users — coordinators, regulatory staff, finance, and the PI or a delegate — run it, not just watch it.

1. Startup: Create the study, enter the visit schedule, assign roles, load the regulatory checklist with due dates.
2. Recruitment handoff: Add fictional referrals, record screening outcomes, document one screen failure with its reason.
3. Scheduling: Book visits inside and outside the window, reschedule one, add one unscheduled visit.
4. Visit execution: Complete one visit, partially complete another, upload source, route items for PI signature.
5. Deviations and corrections: Record a missed procedure, correct an entry, confirm the audit trail shows who changed what, when, and why.
6. Regulatory maintenance: Let a document near expiration, confirm the alert fires, renew it, produce a tracking report.
7. Finance: Mark visits invoiceable, generate an invoice, record a hold, reconcile completed versus billed work.
8. Reporting: Produce visits due, overdue regulatory items, and revenue by study. Time each report.
9. Closeout: Lock and archive the study, then retrieve everything a monitor or auditor would ask for.

For each step record completion, time, workarounds, and whether the result matched your SOPs. A system needing nine screens for a three-screen task has not passed, even if it is technically possible. Set a pass standard in advance — for example, no critical step may require data kept in a spreadsheet outside the system.

Test source-data priority, not just storage

Ask early: where does source data live, and what counts as the original record?

Source documents are the first records of what happened — signed forms, procedure reports, device printouts, notes made at the visit. Electronic source is source data captured directly in electronic form, rather than transcribed later from paper. A strong system makes the link between a visit, its source, and the reported data easy to trace; a weak one stores files but disconnects them from the work.

Test it. Upload source and confirm it links to the right participant, visit, and procedure. Have the PI sign electronically and check that signature, date, and version are preserved. Correct a value after signature: the system should protect the signed record with a clear, attributable correction path, not a silent edit.

Also test retrieval: a coordinator who did not upload the document should find it in under two minutes, as a monitor would. If finding source depends on one person's habits, the system has not solved the problem.

Check integrations with the tools you already use

No site runs on one system. A new CTMS should connect to the recruitment tools your team already knows, not demand you abandon them on day one.

List the integrations that matter first: recruitment tracking tools, electronic health record (EHR) systems used to identify potential participants, electronic data capture (EDC) systems sponsors use to collect study data, calendars, and accounting software — only the tools moving data your staff currently retypes.

Verify each claimed integration in the sandbox, with test data. "Integrates with" can mean a live two-way connection, a one-way export, or a manual file upload — very different things. Ask what data moves, in which direction, and what happens when a transfer fails. Then count the retyping that remains: multiplied across every visit, it is a real cost no demo volunteers.

Get the AI roadmap in writing

Every vendor now promises AI features. Treat them as contract terms: if it matters, get it in writing.

Ask which features exist today, in production, for sites like yours, and which are in limited release or only planned. Ask what data each uses and what human review its output requires. Useful early applications are narrow — reminders, overdue items, missing documents — and each still needs a person accountable for the result.

Do not pay today for a future-tense feature, or let a promise excuse a sandbox weakness. If a roadmap item influences your choice, have its timing, added cost, and the consequence of delay or cancellation written into the agreement.

Price the switching costs nobody quotes

The subscription is the smallest part of a switch. The surprising costs arrive after signing.

• Data migration: What study, participant, regulatory, and financial data moves, in what format, and who verifies it arrived complete and correct?
• Retraining: How many hours for coordinators, regulatory staff, finance, and investigators, including the slower weeks while the workflow becomes habit?
• Parallel running: How long will old and new systems run together to protect active studies, and what does double entry cost?
• Validation and SOP updates: Which SOPs, forms, and training records must be revised to describe the new system, and who does it?
• Exit costs: If you leave later, how do you get your data out, in what format, and at what charge?

Have migration and export terms in writing, then add your internal hours at real staff cost. A system that wins on subscription often loses once these figures are on the page.

Make the decision from the scoresheet

Run the same test, score it the same way, and let evidence rank the candidates. Keep the scoresheets: they document your choice and become your training baseline.

Switch only for a system that has proved, on your own procedures, that it can carry your site. A vendor who welcomes your test is telling you something; a vendor who insists the demo is enough is telling you something too. Believe the sandbox.

Julio G. Martinez-Clark is chief executive officer (CEO) of Amavita Research Services, a Miami cardiovascular trial site, and CEO of bioaccess, a first-in-human contract research organization (CRO) running early-stage trials anywhere from Canada down to Argentina.