What Happens When Phase I Data Quality Breaks Down

Photo: National Institute of Allergy and Infectious Diseases

Phase I studies may involve relatively small numbers of participants compared with later-stage trials, but the data they generate can carry consequences far beyond the size of the study. Early findings inform dose selection, escalation decisions, pharmacokinetic modeling and safety assessments, helping determine how a drug progresses through the rest of development.

That makes data quality more than a matter of accurate database entry. When a study produces an unexpected result, sponsors need to know whether that finding reflects the investigational drug or the way the study was conducted. If that distinction becomes difficult to make, uncertainty can follow the program downstream.

The problem can begin with something that appears minor: a blood sample collected outside its scheduled window, an incomplete source record, an unresolved discrepancy or a protocol deviation. One isolated issue may be manageable. Repeated inconsistencies can make it increasingly difficult to determine whether differences in the dataset reflect biological variability or operational variability, an issue AXIS Clinicals explores in its discussion of Phase I data quality.

In pharmacokinetic studies, timing is particularly important. A blood sample collected outside its specified window can affect interpretation of measures such as maximum observed concentration and time to maximum concentration, as well as the broader concentration-time profile. Dosing errors, inconsistent fasting conditions, improper sample handling or deviations from food, water and activity restrictions can introduce additional sources of uncertainty.

The underlying issue is not simply whether the data were entered correctly after the fact. It is whether the study was executed consistently enough for the resulting dataset to answer the questions the protocol was designed to address.

When Discrepancies Follow the Data

Modern clinical research involves multiple systems and teams, which means a single data point can pass through several stages before it becomes part of a final analysis.

Consider a pharmacokinetic sample. Its collection time may be documented in source records, transferred into an electronic case report form, tracked by a laboratory and ultimately associated with an analytical result. If those records do not align, someone has to establish which information is accurate and why.

FDA’s guidance on electronic source data emphasizes reliability, quality, integrity and traceability from the original source through the electronic record. The agency also notes that source data should be attributable, legible, contemporaneous, original and accurate.

That is why contemporaneous documentation matters. A discrepancy identified while a participant is still at the research site may be investigated while the relevant personnel, records and circumstances are readily available. The same question discovered months later during database cleaning can be much harder to reconstruct.

ICH E6(R3) similarly emphasizes processes that preserve record integrity and traceability so that clinical trial information can be accurately reported, interpreted and verified.

When Data Quality Reaches A Decision Point

The consequences become more significant when questionable data reaches a development decision.

In a Single Ascending Dose or Multiple Ascending Dose study, emerging safety, tolerability and pharmacokinetic information can contribute to decisions about whether and how a program progresses to another dose or cohort. If adverse-event information is incomplete, laboratory data have not been reconciled or pharmacokinetic sampling times are unclear, the people reviewing the emerging evidence may have to make decisions with an incomplete picture.

That does not necessarily mean a study has failed. It can mean that additional review is required before the data can support a confident decision.

In some circumstances, sponsors may need sensitivity analyses or subject exclusions. In others, questions about data quality can trigger additional investigation or regulatory scrutiny. The common consequence is uncertainty: was an unusual result caused by the drug, by the participant, or by how the study was conducted?

For a Phase I program, answering that question may be critical to what happens next.

The Clinical And Laboratory Chain Has To Hold Together

Data quality also depends on what happens after a clinical observation is made.

Samples must be collected, processed, stored, transferred and analyzed according to defined procedures. Clinical records and laboratory results then need to be reconciled so that the final dataset accurately represents what happened during the study.

FDA’s Office of Study Integrity and Surveillance is specifically tasked with ensuring that data supporting regulatory decisions are reliable, including through inspections of pharmacokinetic, bioavailability/bioequivalence and other studies.

This regulatory focus illustrates why clinical operations, data management and bioanalysis cannot be treated as completely separate functions. A problem at one point in the chain can create questions somewhere else.

The issue becomes even more relevant as clinical research relies increasingly on electronic systems. FDA’s 2024 guidance states that electronic systems, records and signatures used in clinical investigations should be trustworthy, reliable and generally equivalent to paper records and handwritten signatures.

Electronic systems can make information easier to capture and review, but they cannot recreate an observation that was never collected correctly.

Database Cleaning Cannot Fix Everything

Modern EDC platforms, automated edit checks and electronic workflows can identify inconsistencies and help teams resolve queries. They are valuable safeguards, but they have limits.

A database can flag an implausible timestamp. It cannot determine with certainty what actually happened if the underlying source record is incomplete. It can identify a discrepancy between two records, but it cannot automatically reconstruct a mishandled sample or an undocumented deviation.

That distinction changes how sponsors should think about quality. The strongest datasets are not created during final database cleaning. They are built prospectively through realistic protocols, appropriate training, consistent execution and quality controls capable of identifying patterns while a study is still underway.

Building Quality Before The First Dose

Prevention therefore begins before participants are dosed.

Protocols need to be operationally realistic, particularly when they involve intensive pharmacokinetic sampling, multiple cohorts, dose escalation, long participant stays or food-effect conditions. Staff need clear training on procedures that are critical to data quality, and clinical and bioanalytical teams need workflows that account for how information and samples move between them.

For CROs, this makes infrastructure part of the data-quality equation. AXIS Clinicals, for example, operates integrated Phase I clinical and bioanalytical capabilities, with electronic data capture and connections between clinical operations and laboratory systems designed to support timely data visibility and reduce manual transcription. Its Phase I facility and adjacent bioanalytical laboratory are also structured to keep clinical and sample-processing workflows closely connected.

The value of that model is not simply speed. When the people, systems and laboratories generating different parts of a dataset are working from connected processes, discrepancies can be identified closer to their source rather than becoming questions during late-stage reconciliation.

That is increasingly important as sponsors seek faster development timelines without weakening the reliability of the evidence supporting each decision.

Ultimately, a completed Phase I study is not necessarily a useful Phase I study if its data leave fundamental questions unanswered.

The objective is not to eliminate every deviation or discrepancy. Clinical research is too complex for that. The objective is to create a system in which important problems are detected early, investigated while they can still be reconstructed and prevented from becoming larger sources of uncertainty.

For sponsors, that means treating data infrastructure with the same scrutiny given to recruitment, timelines and operational capacity. The value of Phase I data lies not in how much information a study produces, but in how confidently that information can be used to decide what comes next.