Quality Tolerance Limits: Why Breaches Happen Early and How to Act Fast

Quality tolerance limits (QTLs) are trial-level thresholds on critical-to-quality (CtQ) factors — key measures where sustained deviation could affect safety or primary conclusions. They should be a small set of guardrails that trigger timely investigation and documented decisions within risk-based quality management (RBQM).

Since the QTL module was introduced in 2019, adoption has increased steadily in the CluePoints platform: ~580 studies have used QTLs, with 2.6 QTLs per study on average.

In this analysis, 48% of studies crossed a secondary (early-warning) limit, while 39% ultimately exceeded the tolerance limit, suggesting early warning can prompt action before issues escalate. Across time points (“snapshots”), only 11% of QTL-snapshots were above tolerance, indicating breaches occur in specific windows rather than continuously. Early in a trial, small denominators can temporarily push rates over a fixed threshold (e.g., 1–2 events in the first 10 patients), so snapshots should be interpreted as trends, not one-off alarms. Adding statistical context (e.g., confidence intervals) can reduce noise from a low sample size. Breaches were often short (median two snapshots above tolerance; P25–P75: 1–4).

Practical implications: First, expect some QTL breaches, especially early on when denominators are small; statistically supported breaches warrant rapid escalation, while absolute-only breaches may need quick validation and trend confirmation. Second, because breaches were brief, quick review and action can bring QTL below the threshold, preventing significant impacts to trial safety or end points.

To translate QTLs from a documentation exercise into a practical RBQM control, the following principles consistently help:

• Pick a few CtQ-focused QTLs: Prioritize trial-level, systemic risks that could impact safety or primary conclusions; use KRIs for metrics mainly aimed at comparing sites.
• Set defensible thresholds: Base limits on similar trial experience and clinical relevance, and add statistical context to reduce early-trial noise.
• Operationalize breach handling: Confirm the signal, assess scope and trend, decide and document actions (or no-action rationale) in a cross-functional team, then verify impact in subsequent reviews.
• Measure success by speed and impact: Track time-to-triage, time-to-decision, and whether the metric improves after intervention rather than aiming for “zero breaches.”

In conclusion, keep QTLs to a few CtQ guardrails, set thresholds you can justify and agree upfront who does what when a breach occurs. Expect some early breaches due to small numbers — use snapshot trends to decide whether to act or monitor. The win is fast triage, documented decisions and confirming improvement in the next reviews.

Sylviane de Viron, data and knowledge manager, and Frederic Blais, PhD, RBQM lead (strategic consulting team), both with CluePoints