The behind-the-curtain look at how a clinical research site actually runs. Written for people looking to enter the clinical trial world, for new coordinators learning the ropes, and open to anyone curious what happens at clinical research sites.

In this file:

·        Why the investigator and the sponsor both assess the same event, separately

·        What related, possibly related, and unlikely related actually mean on the form

·        Dechallenge and rechallenge, and why nobody rechallenges on purpose

·        Why an honest disagreement between two doctors is a feature, not a failure

The report Nani filed on Tuesday afternoon took her about twenty minutes. The box Dr. Reyes had to fill in on Wednesday morning was one field wide, and she sat with it for most of a coffee. Walter's headache, three days after his last dose of study drug. Resolved overnight. He had taken something over the counter for it. He had also had headaches on and off for years, and he takes a medication for something unrelated that lists headache on its own label. And Dr. Reyes did not know, and would not know for months, whether Walter had been taking the investigational drug or a placebo. She still had to answer the question.

At least two people assess causality on the same event, separately, and neither one replaces the other.

The first is the investigator. This is not a courtesy or a professional opinion offered in passing. Under 21 CFR 312.64(b), an investigator must immediately report a serious adverse event to the sponsor, and must include an assessment of whether there is a reasonable possibility that the drug caused the event. The regulation puts the judgment on the site, in writing, attached to the report.

The second is the sponsor. The sponsor decides what reaches FDA as an IND safety report, and the deciding concept is the suspected adverse reaction: an adverse event for which there is a reasonable possibility the drug caused it. The sponsor is looking at Walter's headache (even if they’re seeing it as subject 123-102’s AE) alongside every other event reported by every other site in the study, which is a view Dr. Reyes will never have.

Newcomers often assume the second assessment corrects the first. It does not. Both are recorded. Both travel with the case. The sponsor is not marking the investigator's homework, it is answering a different question with a different field of view.

One more thing worth being clear about, because it comes up in interviews. The coordinator does not make this call. Nani cannot assess causality, no matter how well she knows Walter, because causality assessment is a medical judgment delegated to a physician on the delegation log. What Nani controls is everything the physician will have to work from.

There is no single federal scale. The categories on the form come from the protocol, and protocols differ. Some use a binary: related, or not related. Many use a graded scale that runs something like unrelated, unlikely related, possibly related, probably related, definitely related. Read your own protocol before you assume you know the options, because a coordinator who learned a five-point scale at their last site will reach for it at a site that uses two.

What actually drives the choice is a short list of questions, asked in roughly this order.

·        Timing. Did the event happen in a window where the drug could plausibly be involved, given what is known about how it moves through the body?

·        Is this a known effect? The Investigator's Brochure carries what has been observed with this compound so far. An event already listed there sits differently than one nobody has seen.

·        Is there another explanation? This is where a concomitant medication, meaning anything the patient is taking alongside the study drug, and the patient's own history do their work.

·        What happened when dosing stopped or continued?

Run Walter through that and you get a genuinely mixed picture. The timing is loose. There is a concomitant medication that lists headache. There is a history of headaches predating the study by years. All of that pulls toward unlikely related. None of it rules the drug out.

So Dr. Reyes wrote possibly related, and that is the part worth slowing down on. Possibly related is not a hedge, and it is not an admission that nobody knows anything. It is a specific claim: a causal relationship cannot reasonably be excluded on the evidence available. New coordinators read it as weakness in the assessment. It is the opposite. Writing unlikely related when you mean possibly related is how a real signal gets buried under a hundred confident guesses.

And notice what Dr. Reyes did not have. She did not know Walter's arm. That is deliberate. If she knew he was on placebo she would write unrelated without thinking, and the safety database would end up shaped by what the investigator knew rather than by what happened to the patient. Blinded causality assessment is uncomfortable on purpose.

Two words you will hear constantly and rarely see defined.

Dechallenge is what happens when the drug is stopped or the dose reduced. If the event resolves, that is a positive dechallenge, and it supports a causal relationship.

Rechallenge is what happens if the drug is given again. If the event returns, that is the single strongest piece of causality evidence available in clinical research. It is also the one nobody goes looking for.

No one rechallenges on purpose. You do not deliberately re-expose a patient to something you suspect harmed them in order to firm up a data point. The rechallenges that appear in trial data are almost always incidental: a chronic dosing study where the next scheduled dose is itself a rechallenge, or a patient who resumed before anyone connected the two events.

Both tools have limits, and Walter is a good illustration of all of them. His headache resolved overnight, which looks like a dechallenge until you notice that nothing was dechallenged. He never stopped the drug, he was simply between doses. He took something over the counter, so there is a competing explanation for the resolution. And a headache is self-limiting anyway, which means it was going to resolve regardless of what caused it. Three days later, a new coordinator reading that chart could easily write down positive dechallenge and be entirely wrong. The record supports no such thing.

Note to File: The investigator's assessment is only ever as good as your intake. By the time Dr. Reyes opens the form, Walter is off the phone and back to his week, and everything she has is what somebody wrote down. So when you take the call, capture the boring things. Exact time of onset and resolution. What else he takes and when he last took it. What he took for the headache and how much. Whether this has happened before, and how often. What he was doing. You are not deciding causality, and you should not try to. You are building the only record the person who does decide will ever see.

It happens regularly, and it’s not a fight.

Dr. Reyes writes possibly related, because she is looking at one man she has met, with a headache history and a concomitant medication, and she cannot rule the drug out. The sponsor's medical monitor, reading the same report alongside every other event in the study, may land on unlikely related, because from where he sits Walter's headache looks like the ordinary background rate of headaches in men in their sixties.

Or it runs the other way, which is the version people forget. Three headaches reported in one week at three unrelated sites will get the sponsor's attention in a way no single investigator could possibly generate, because no single investigator can see the other two.

Neither assessment is deleted and neither overwrites the other. Both stay on the record. For the purpose of deciding what gets reported onward, the more cautious reading generally governs, which is the safe direction for the system to fail in.

That is the whole design, and it is worth understanding before you sit in your first safety call. The site sees one patient in depth. The sponsor sees every patient shallowly. Neither view is complete, so the system refuses to choose between them and keeps both.

Causality in a clinical trial is decided twice, independently. The investigator assesses it at the site, and under 21 CFR 312.64(b) that assessment is a regulatory requirement attached to the report, not an optional opinion. The sponsor assesses it again, with a view of every site in the study, to decide what reaches FDA. The categories on the form come from the protocol rather than from any single federal scale, and possibly related is a real answer meaning a causal link cannot reasonably be excluded, not a way of saying nobody knows. Dechallenge and rechallenge are the strongest causality evidence there is and the least available in practice, because nobody re-exposes a patient on purpose. And when the two assessments disagree, the disagreement is recorded rather than resolved, because neither party can see the whole study on their own. Dr. Reyes clicked possibly related, wrote two sentences explaining what else had been on the table, and sent it.

·        Previous issue: Issue 6, safety reporting

·        Next issue: Issue 8, the monitoring visit

·        The whole file: roottorx.com/t/open-site-file

·        Root to Rx: the newsletter and The Root Room, the patient-facing side of the same story, at roottorx.com

·        21 CFR 312.64 — Investigator reports (eCFR)

·        21 CFR 312.32 — IND safety reporting (eCFR)

·        FDA — E2A Clinical Safety Data Management: Definitions and Standards for Expedited Reporting

·        FDA — E6(R3) Good Clinical Practice, Guidance for Industry

•  •  •

The Open Site File is general education about how clinical research sites work. It is not regulatory, legal, or medical advice, and it is not tied to any specific study, sponsor, or employer. Study OLS-114 and all names are invented. Always follow your own site's SOPs, your protocol, and your PI and QA team. Characters are composites.

Filed by Justin Yamashita, MSc., Open Label Media

Next in the file: Issue 8, the monitoring visit, and the CRA who reads every word Nani wrote.