FDA Findings: Manual Integration Without Justification

Chromatography data systems integrate peaks automatically, but no integration algorithm is perfect, so analysts sometimes adjust it. That is legitimate when it is scientific, documented and reviewed. It becomes a data integrity problem when it is used to move a result across a specification limit, applied inconsistently, or hidden from the reviewer. Three Warning Letters show the three versions.

Background: data integrity and ALCOA+ in practice and reviewing printouts instead of electronic data.

What FDA found

Company X1 — invisible to the reviewer

A contract testing laboratory. Under 21 CFR 211.194(a), the letter states that the firm manually integrated sample assay peaks, but the printed data reviewed by the quality unit did not show that manual integration had been performed, so the QU never reviewed it. Under 21 CFR 211.160(b), FDA notes that “manual integration is performed without a governing procedure”, and that procedures did not require evaluation of reprocessed integrations, aborted runs or audit trails.

Company X2 — no justification, inconsistent main peak, inhibited unknowns

Under 21 CFR 211.194(a), the letter describes “manual integration of HPLC assay peaks without scientific justification or procedural controls”, and inconsistent integration of the main peak in a stability batch. Under 21 CFR 211.192, unknown peaks in HPLC chromatograms were not integrated or investigated; peak integration was inhibited and integration parameters restricted. FDA states that inhibiting integration and restricting parameters must be scientifically justified and documented in the validated method. FDA requested a retrospective review of the firm’s manual integration practices.

Company X3 — timed events that changed the outcome

Cited under 21 CFR 211.160(a). According to the letter, an analyst changed the processing of an impurity peak in a tablet batch by “manually entering timed integration events into the processing methods”; had automatic integration been applied consistently, the result would not have met the release specification. FDA notes that the integration procedure did not explain when manual timed events may be used or how they should be reviewed, and did not require supervisory approval. FDA found the response inadequate because the firm did not commit to a retrospective review of all relevant chromatographic data.

Why this happens

Interpretation — the letters state findings, not root causes. These are common patterns, not claims about these firms.

• Weak methods. Poor separation or noisy baselines make manual intervention routine, and routine intervention is hard to review.
• No definition. If the procedure doesn’t define what counts as manual integration (moving baselines, timed events, inhibit windows, parameter changes), nobody knows what must be justified.
• Integration happens after the result is visible. Once the analyst can see that a result fails, every integration change is suspect — whatever the intention.
• Reviewers see only the final chromatogram. Without the original integration and the change history, the reviewer can’t judge whether the change was scientific.

What good looks like

Example: a fictional QC lab running HPLC assays and impurities on a CDS we’ll call ChromaLog.

1. An integration SOP, not just a manual-integration ban. Define integration parameters per method in the validated method; define what counts as manual intervention; state when it is permitted and what justification is required.
2. Same treatment for standards and samples. Integration is applied consistently across standards, samples and injections in a sequence. Changing one sample alone needs a strong documented reason.
3. Keep both versions. The system retains the original automatic integration and every reprocessed version, with the audit trail. FDA’s guidance (question 14) explains why saving only the final reprocessed result is not acceptable.
4. Flag it for the reviewer. Reports clearly mark manually integrated peaks; second-person review compares original and manual integration and checks the justification.
5. Fix the method, not the peaks. Trending manual interventions by method shows where method improvement — not more manual work — is the real fix.
6. Inhibit only with justification. Integration inhibit windows and minimum-area settings are justified in method validation and locked in the method, not set ad hoc.

Perspective: what FDA’s requests tell you

In X2 and X3, FDA asked for retrospective reviews of past integration — and in X3 it found the firm’s partial reprocessing insufficient. The cost of uncontrolled integration is therefore not just one investigation; it is re-reviewing years of chromatograms. For CSV teams, this is also a validation point: configure the CDS so that reprocessing is versioned, manual events are visible in reports, and reviewer roles can see them — and test that configuration in OQ.

Regulatory basis

• 21 CFR 211.194(a) — complete laboratory data, including all chromatograms.
• 21 CFR 211.160(a) and (b) — laboratory controls established by written procedures, followed, documented, and scientifically sound.
• 21 CFR 211.192 — thorough investigation of unexplained discrepancies.
• FDA guidance: Data Integrity and Compliance With Drug CGMP — Questions and Answers (December 2018), question 14 (saving only final reprocessed results) and question 7 (who should review audit trails). Guidance is non-binding; the regulations are.

EU angle

Check current versions on EudraLex before relying on specific clauses. EU GMP has no integration-specific clause; expectations follow from EudraLex Volume 4, Part I, Chapter 6 (Quality Control) and Annex 11 (audit trails, clause 9). Pharmacopoeial chromatography chapters address system suitability, not integration practice.

Takeaways

• Manual integration is allowed; undocumented, unjustified or invisible integration is not.
• Keep the original integration and every version.
• Treat standards and samples the same way.
• Trend manual interventions — they point to methods that need fixing.