Clinical trial operations: site selection to database lock

Interviewer: OK, I'm recording. Walk me through running a trial end to end, like you'd explain it to someone new on the team. Wherever you want to start.

Cra: Sure. So by the time it lands on my desk the protocol's already finalized — that's the sponsor's medical and clinical people, that document is basically the bible, everything we do traces back to it. My world really starts at the next bit, which is finding sites. Feasibility.

Interviewer: Feasibility, meaning what?

Cra: Meaning, who's actually going to be able to run this thing and enroll patients. We send out feasibility questionnaires to candidate sites — how many patients fitting these criteria do you see, what's your PI's experience, do you have the equipment, are you running three competing trials in the same indication already. You score all that and you build a shortlist. And honestly that scoring is part art — sites are wildly optimistic about how many patients they'll enroll. Everybody says fifty, everybody enrolls eight.

Interviewer: So from the shortlist, how do you decide who's in?

Cra: We do a site selection visit. Go on-site — or remote these days for some of it — and actually confirm the place is real. Right staff, right facilities, an actual patient population walking through the door. And it's a real fork: some sites just don't pass, you drop them, and if you've dropped too many and you're short of your target number, you go back and identify more. You can't run a trial with too few sites.

Interviewer: And once a site passes selection?

Cra: Then the slow part. Contracting. You negotiate the clinical trial agreement and the budget — per-patient costs, all of it — with the site or the institution. That is almost always the long pole in startup. Legal on both sides, back and forth, it can eat months. Meanwhile, in parallel, the site's package is going to the IRB — the ethics committee — for approval. You cannot enroll a single patient until that's granted, full stop.

Interviewer: And there's paperwork beyond the contract and the IRB, I assume.

Cra: Oh, mountains. The essential documents. The Form 1572, the investigator CVs, financial disclosures, lab certifications and their normal ranges, the signed protocol — all of it gets collected and filed into the trial master file, the TMF. The TMF is the regulatory record of the trial. If it's not in the TMF, it didn't happen, that's the auditor's view. Collecting those docs is a gate too — you're not activating without them.

Interviewer: Then what?

Cra: Then the site initiation visit. The SIV. That's where I go in and actually train the site — walk the coordinators and the PI through the protocol, the EDC system, how to handle the investigational product, how to report adverse events, what counts as proper source documentation. After that the site is trained and, in theory, ready.

Interviewer: In theory?

Cra: [laughs] Right, because activation is its own checkpoint. Before a site gets the green light to enroll, I'm confirming everything lines up — CTA executed, IRB approved, essential docs complete, drug actually shipped to the site, SIV done. If any one of those is missing, the site sits on hold and you chase whatever's open. Only when it's all clear does the site go active. That's the green light.

Interviewer: OK, site's live. Now the patients.

Cra: Now recruitment. And this is the coordinator's world at the site — the CRC. They advertise, they review charts, they pre-screen people against the big obvious criteria. When they've got a candidate, before anything else happens, informed consent. The PI or coordinator sits the patient down, walks them through it, gets the signature. Nothing — no screening test, no blood draw, nothing — happens before consent. That's sacred.

Interviewer: And after consent?

Cra: Screening. You run the protocol's screening tests and you check them against inclusion and exclusion criteria. And here's a big loop — a lot of people screen fail. Their lab value's out of range, they're on a prohibited med, whatever. A screen fail means they're out, and you're back to recruiting more. This is where slow enrollment lives, this is the thing that keeps everybody up at night. You need X patients and the screen-fail rate is brutal.

Interviewer: And if they pass screening?

Cra: They get enrolled and randomized — usually through an IRT system that assigns them to a treatment arm. Then the investigational product gets dispensed, and the site has to log it meticulously — what was dispensed, what came back, what got destroyed. Drug accountability. We reconcile that at every monitoring visit; regulators care enormously about it.

Interviewer: So now the patient's in the trial. What's the rhythm?

Cra: Study visits, per the schedule of assessments. The patient comes in at the defined intervals, the site does the assessments, and the coordinator enters all the data into the EDC. That's the engine that runs for most of the trial — visits in, data in.

Interviewer: And your job during that stretch?

Cra: Monitoring. This is the heart of the CRA role. I visit the site — or remote-monitor — and I do source data verification. I'm checking what they typed into the EDC against the actual source documents. Does the EDC say the patient's blood pressure was one-thirty over eighty? Let me see the chart that says that. I'm also reviewing the consent's all in order, the drug accountability, the deviations. And it's a lot of travel, historically — that's a real grind of the job.

Interviewer: And when the data doesn't match?

Cra: Then queries. Any discrepancy gets raised as a query — data management or I flag it, and the site has to resolve it, correct the entry, explain it. And that's a loop, right? Query goes out, site fixes it, I re-verify, maybe it spawns another query. Query backlogs are a genuine plague. You can have hundreds of open queries on a big study and they all have to close eventually.

Interviewer: You mentioned adverse events earlier. Where do those fit?

Cra: So adverse events run alongside everything. Any AE gets recorded. But a serious one — an SAE — kicks off a totally separate, expedited safety track. That's not the routine data flow; the site reports it to the sponsor's safety group within twenty-four hours, and depending on what it is, it goes to regulators on a clock. That branch can fire at any point during treatment and it takes priority over everything. Same with protocol deviations — those get logged and reported on their own track, running concurrent with the monitoring.

Interviewer: OK. So how does the whole thing start to wind down?

Cra: First big milestone is last patient in — LPI. You've hit your enrollment target. And again, if you haven't hit it, you're looping back into recruitment, more sites, more screening. But once you're enrolled, you close recruitment. Then it's just treatment and follow-up until your last patient finishes their last visit — LPLV, last patient last visit. That's the moment the data stops coming in.

Interviewer: And then you can lock it?

Cra: Not yet — and this is the part people underestimate. Database cleaning. You have to resolve every open query, reconcile all the SAEs against the database, reconcile the drug, pull in external data like central labs. And it's a loop until it's genuinely clean — any open query sends you back. Only when data management is satisfied there's nothing outstanding do you lock the database. Hard lock. No more edits without a controlled, documented re-opening, which nobody wants to do.

Interviewer: And once it's locked?

Cra: Unblinding. The treatment assignments get revealed, and the locked, clean dataset gets handed to biostatistics for the analysis. And that's really the end of the operations arc — from there it's a statistics and reporting story, not mine.

Interviewer: So if you had to name the one thing people get wrong about all this — what is it?

Cra: They think the hard part is the science, or the lock at the end. It's not. The thing people get wrong is treating startup and enrollment as separate from the lock — like the clean database is a problem you solve at the end. It isn't. Every shortcut at the site initiation visit, every coordinator you didn't train well enough, every query you let pile up instead of resolving in the moment — all of that comes due at lock, all at once, usually against an immovable deadline. The clean lock is built — or lost — in month two, at the SIV, when you decided how disciplined the site was going to be. People sprint to activate the site and start enrolling because that's the visible milestone, and they pay for that haste for the entire rest of the trial. The teams that lock clean and on time are the ones who were boringly rigorous about source and queries from the very first patient. That's the whole game.

Interviewer: That's a great place to stop. Thanks.

Cra: Anytime. Beats writing the monitoring report.