Critical-to-Quality Factors: Choosing What a Study Monitors
Critical-to-quality factors are the small number of things in a study that decide whether participants are safe and whether the results can be believed. Naming them is the first step of a risk-based approach, and the list is useful only if it is short enough that a team can act on every item.
The common failure is a list of forty. It gets assembled by walking the protocol section by section, adding anything that could conceivably go wrong, and circulating the result for comment. Every reviewer adds. Nobody removes. What comes out is the protocol restated as a risk register, which gives no basis for deciding where monitoring effort should go. This post is the detailed version of one section of the wider risk-based quality management plan.
What Makes a Factor Critical
ICH E8(R1) introduced the term and ICH E6(R3) builds the quality system on top of it. The definition is narrow on purpose. A factor is critical when a failure would affect the safety of participants or the reliability of the results. Not the convenience of the study team, not the preference of a sponsor reviewer, not the tidiness of the database.
That narrowness is the whole value. A study has hundreds of things that could go wrong and perhaps eight to twelve that would change what the study means. Monitoring designed around the first number is monitoring by volume. Monitoring designed around the second is what the guideline is asking for.
Four Tests That Cut the List
A candidate factor has to survive four questions. Most drafts lose half their entries at the second.
The first is whether a failure reaches safety or the reliability of the result. This removes the process complaints that accumulate on any list: a slow report, an awkward export, a form nobody likes.
The second is whether the factor can plausibly fail in this study. A theoretical risk with no mechanism in this protocol, with these sites and this population, is not a risk to plan around. Requiring someone to state how it would actually break here removes a surprising amount.
The third is whether a failure is recoverable. A missing laboratory value can often be repeated. A consent gap, a broken blind or a diary entry completed from memory cannot. That asymmetry deserves to be written down, because unrecoverable failures justify effort that a recoverable one does not.
The fourth is whether anyone would act. If a factor has no owner and no defined response, it is an observation rather than a control. Cutting these is what keeps the list usable through database lock.
Where the List Comes From
Candidate factors are already written down. They sit in the objectives and endpoints, the eligibility criteria, the safety assessments, the handling of investigational product, and the consent process. Reading those five areas with the four tests in hand produces a better list than a blank workshop, and it produces it faster.
Consent is the section worth treating separately. It is the only one where essentially no failure is recoverable, which is why it belongs on almost every study's list regardless of therapeutic area or design.
From Factor to Risk to Control
A named factor is not yet a plan. The row is complete when it carries a failure mechanism specific to this study, a control, and an owner. Without the mechanism, the control tends to be generic. Without the owner, nothing happens when the control fires.
Writing the failure mechanism is where most of the thinking happens, and it is the part reviewers skip. "Eligibility errors" is a category. "Screening chemistry transcribed from an outside laboratory report with no source in the site record" is a mechanism, and it points directly at a control that a monitor can actually run.
The controls themselves are ordinary: a central review, a targeted source check, an access review, a comparison across sites. The value comes from the fact that each one exists because a named factor could fail in a named way, which is also what makes the plan defensible later. Where the control depends on data flowing out of the electronic data capture system, the design decisions behind it are covered in why query volume rises late in a study.
Who Owns the Assessment
The assessment is cross-functional or it is wrong. Clinical operations knows what sites will struggle with. Data management knows which checks are possible in the system and which are wishful. The medical monitor knows which safety signals matter in this indication. Statistics knows what actually breaks the analysis.
Where a contract research organization runs the study, the assessment can be delegated but the judgment cannot. The sponsor decides what is critical, because that decision is a statement about the study's purpose rather than about monitoring logistics. A useful test at handover: can the sponsor explain, without reading from the plan, why each factor is on the list.
Keeping It Usable After Go-Live
A risk assessment produced once and filed is the most common way the whole approach fails. Enrollment patterns change, a site turns out to behave unlike the others, an amendment changes an endpoint. Each of those can add or retire a factor.
What keeps it alive is a scheduled review with a small number of defined triggers, and a record of what changed and why. Amendments are the clearest trigger, and the revalidation consequences of one are set out in handling a mid-study amendment without revalidating the whole build. Study management, oversight and the systems that carry the record are covered in the guide to which system owns which part of the study, and the platform side sits with Clinera CTMS and our clinical and R&D AI work.
References
ICH E8(R1), General Considerations for Clinical Studies, which introduces quality by design and critical-to-quality factors, adopted 6 October 2021. ICH E6(R3), Good Clinical Practice, Step 4 on 6 January 2025, which builds the quality management system on those factors. The FDA guidance on oversight of clinical investigations and its 2019 questions and answers on risk-based monitoring. TransCelerate BioPharma has published position papers and a risk assessment categorization tool that many sponsors use as a starting template.
Guideline texts are available at www.ich.org, www.ema.europa.eu and www.fda.gov. Check the current version before citing a date in a study document.
Frequently Asked Questions
How many critical-to-quality factors should a study have?
Few enough that every item has a named owner and a defined response. Most studies land somewhere between eight and twelve. The number is less important than the test: if a factor has no control attached and nobody would do anything differently when it fires, it is an observation and it should come off the list. A list of forty is the protocol restated, and it gives no basis for prioritizing monitoring effort.
What is the difference between a critical-to-quality factor and a risk?
The factor is the thing that matters, such as the integrity of the blind or the timing of the primary endpoint assessment. The risk is the specific way that factor could fail in this study, such as unblinded pharmacy staff being recorded in general study notes. One factor can carry several risks. Controls are written against the risks, not against the factor, which is why the failure mechanism has to be stated before the control is chosen.
Where do critical-to-quality factors come from in ICH guidance?
The concept is introduced in ICH E8(R1), General Considerations for Clinical Studies, as part of quality by design. ICH E6(R3) then builds the quality management expectation on top of it, asking that quality management be proportionate to the risks to participant safety and to the reliability of results, and that the reasoning be documented. The two are meant to be read together rather than separately.
Can we reuse the risk assessment from a previous study?
As a starting template, yes, and most sponsors do. As a finished document, no. The second of the four tests asks whether a factor can plausibly fail in this study, with this protocol, these sites and this population, and that answer is study specific. A copied assessment usually shows up at inspection as a list of generic risks with controls that do not match how the study is actually running.
Who should own the risk assessment when a CRO runs the study?
Execution can be delegated. The judgment about what is critical cannot, because it is a statement about the purpose of the study rather than about monitoring logistics. In practice the contract research organization drafts and maintains, and the sponsor reviews, challenges and approves. A reasonable test at handover is whether the sponsor can explain why each factor is on the list without reading from the plan.
How often should the assessment be reviewed?
On a schedule and on defined triggers, with a record of what changed and why. A quarterly review is common, and protocol amendments, the addition of a country or site type, a safety signal, and a change in enrollment pattern are the usual triggers. What matters at inspection is not the frequency but the evidence that the assessment influenced decisions after study start rather than being filed once and left.
