19.1 KRIs, Acceptable Ranges and Trial-Level Quality

KRIs, Acceptable Ranges and Trial-Level Quality

1. Adverse Event (AE) Underreporting:

   • Metric: Reported events per participant over a defined comparable exposure period; use person-time when exposure differs.
   • Operational Red Flag: Across 80 trial sites evaluating a novel cytotoxic therapeutic, the illustrative study-wide average is 4.2 AEs per subject over a comparable exposure period. Site 112, which has enrolled 18 subjects over 6 months, reports zero adverse events.
   • CRA Operational Interpretation: In an aggressive pharmacological study, reporting zero adverse events does not mean the site provides superior clinical care; it warrants prompt investigation for possible underreporting, absent investigator inquiry, failure to review concurrent medications, or lack of source data transcription. A proportionate follow-up, potentially an on-site visit, is warranted to execute comprehensive Source Data Review (SDR) across medical progress notes and nursing logs.

2. Concomitant Medication Underreporting:

   • Metric: Number of reported concomitant medications per participant.
   • Operational Red Flag: A geriatric patient with advanced diabetes and chronic kidney disease has zero concomitant medications recorded in the eCRF.
   • CRA Operational Interpretation: It may be unexpected for a multimorbid geriatric patient to receive no concomitant prescriptions, over-the-counter drugs, or nutritional supplements. The CRA must cross-reference pharmacy dispensing histories and primary care physician notes.

3. Protocol Deviation Velocity & Query Latency:

   • Metric: Rate of new protocol deviations per subject visit and median calendar days to resolve open EDC data queries.
   • Operational Red Flag: A site's query resolution latency exceeds 45 days (study benchmark is fewer than 5 days), accompanied by a high density of visit window deviations.
   • CRA Operational Interpretation: Indicates severe site staff turnover, lack of coordinator availability, or inadequate investigator oversight.

Quality Tolerance Limits (QTLs): Study-Level Risk Governance

A useful operational distinction is the difference between Key Risk Indicators (KRIs) and Quality Tolerance Limits (QTLs):

• Key Risk Indicators (KRIs): Often evaluated at the site level to identify individual outlier centers requiring targeted CRA visits, retraining, or corrective actions.
• Quality Tolerance Limits (QTLs): Evaluated at the trial level across the entire clinical study. QTLs represent predefined statistical boundaries that consider the medical and statistical characteristics of the trial design. They are designed to identify systematic issues that could compromise trial integrity or subject safety across the entire program.

Illustrative Trial-Level Ranges

These are hypothetical examples for understanding a metric; they are not ICH-prescribed limits or default acceptable rates. Each trial must justify its own ranges:

• Study Dropout / Lost-to-Follow-up Rate: Predefined QTL set at >15% overall participant withdrawal rate prior to primary endpoint completion.
• Primary Endpoint Missing Data: Predefined QTL set at >3% missing primary efficacy measurements across the study population.
• Major Protocol Deviation Rate in Primary Population: Predefined QTL set at >5% of randomized subjects experiencing a major deviation affecting the per-protocol analysis set.

Evaluating deviations from predefined acceptable ranges

E6(R3) calls for evaluation when predefined acceptable ranges relevant to critical-to-quality factors are exceeded. Determine whether the observation reflects an important systematic issue or another explanation. Investigate causes and implement proportionate remedial action where appropriate; do not assume every threshold crossing automatically invalidates the study or requires a programme-wide CAPA.

Document important issues and remedial actions, including relevant quality-management information in the clinical study report. An ICH guideline is not a universal statute assigning one legal consequence to every alert. Apply actual reporting requirements and assess participant protection and reliable results. A site KRI may require targeted follow-up even when a trial-wide range has not been crossed, while an overall signal can reveal a design or system problem affecting several sites.

Reading risk metrics with the right denominator

An illustrative trial has 200 participants expected to contribute a primary assessment at a data cut, and 8 assessments are missing. The missing-assessment proportion is 8/200 × 100 = 4%. The figure answers a defined operational question; it does not establish a universal acceptable limit. A threshold should have been justified from the study's scientific and operational characteristics.

If a site reports 2 missing assessments among 10 expected, its site proportion is 20%. Comparing 2 missing records with the trial's 8 without denominators would conceal the higher local rate. Investigate timing, participant status, source availability and whether the data were truly expected. An apparent missing assessment can represent late transmission, a wrongly assigned visit or genuine noncollection.

A metric crossing a predefined acceptable range should prompt evaluation of its importance and proportionate action. It is not automatic proof that the trial's conclusions are invalid. Conversely, a trial-wide proportion below the range can hide a serious local problem affecting participant protection. Use site and trial perspectives together, document meaningful issues and response, and reassess controls as evidence accumulates.

Do not change a threshold retrospectively just to make an alert disappear without a justified, controlled evaluation.

Official sources

Checked October 18, 2026: ICH E6(R3), Principles and Annex 1, ICH E8(R1).