Pharmaceutical process validation represents the critical leap from development to commercialization—the definitive demonstration that a process can reliably and consistently deliver product of the intended quality. A question frequently debated among industry practitioners has resurfaced: when one batch fails amid three consecutive process validation runs, do the preceding two batches retain their validation status? The answer demands rigorous technical judgment.
Part 1: Clarifying Regulatory Intent — What “Three Consecutive Successful Batches” Actually Requires
The Process Validation Inspection Guideline issued by the NMPA Center for Drug Inspection stipulates that traditional prospective validation shall normally comprise at least three consecutive successful batches at commercial scale. Two terms warrant precise interpretation:
“Consecutive” underscores the integrity of the validation sequence. It requires evidence of process performance within the normal production order—not a post hoc curation of “best results.” Selecting three passing batches from a four‑batch production sequence inherently breaches the continuity principle. EU GMP Annex 15, Clause 5.20, similarly acknowledges that at least three consecutive batches produced under routine conditions may be deemed sufficient, provided this does not conflict with Clause 5.19, which ties batch number to quality risk management. The regulatory focus, in other words, is not on assembling three qualified lot numbers but on an unbroken, representative record of routine manufacturing performance.
“Successful” extends beyond finished‑product release testing. Each batch must satisfy all pre‑approved Process Performance Qualification (PPQ) acceptance criteria. In practice, a batch may pass release testing yet fail at intensified PPQ sampling points or exhibit critical process parameter deviations. Such a batch may be commercially releasable but cannot count as a successful validation batch. As FDA emphasizes in Process Validation: General Principles and Practices, the PPQ report must summarize all non‑conformances and assess whether the process meets protocol conditions and remains in a state of control. Passing finished‑product tests alone cannot substitute for evidence of process reproducibility.
Part 2: Do Prior Batches Remain Valid After an Intermediate Failure?
When a failure occurs mid‑sequence, two extreme positions often emerge: one demands discarding all prior batches and starting over; the other dismisses the failure as immaterial so long as finished products pass release. Both are flawed. The issue must be examined on two levels.
Data from completed batches remain valid. Manufacturing records, test results, and in‑process observations from these batches are integral to the validation report and must not be deleted or suppressed because a later batch failed. PIC/S PI 006‑4 (revised July 2026, effective 1 October 2026) reinforces this: validation reports must enumerate all validation batches and provide a holistic assessment of all deviations. Historical batch data, therefore, are not merely retained—they become critical inputs to deviation investigation and risk assessment.
However, the evidence chain of “consecutive success” is broken. If investigation links the failure to process design, material attributes, or control‑strategy deficiencies, the sequence of consecutive successful runs terminates at the failing batch. PIC/S PI 006‑4 notably re‑frames the traditional “three consecutive batches” as a transitional benchmark, stressing that batch count should be scientifically and risk‑justified. That said, the guidance is not yet in force. Under China’s current GMP framework, three consecutive successful batches remain the baseline expectation; enterprises should not unilaterally reduce validation batch numbers.
Exceptions exist. Where investigation conclusively attributes failure to a well‑bounded, non‑process root cause—for example, a laboratory instrument malfunction producing erroneous results—the enterprise may, on scientific grounds and following risk assessment, determine whether the batch retains validation value. But “external cause” must not become a convenient label. Power interruptions, equipment alarms, and operator errors are, in many cases, precisely the sources of variability that routine manufacturing is expected to control.
Part 3: Decision Framework — Six Core Principles
The decision after a failure should not devolve into a numbers debate—”add one batch” versus “repeat all three.” It requires systematic evaluation across six dimensions:
1. Is the failure process‑related? If the root cause points to inadequate process parameter ranges, material‑attribute variability exceeding process capability, or control‑strategy gaps, the failure directly undermines the validation’s core premise. Per Clause 6.3.3 of the NMPA Process Validation Inspection Guideline, quality excursions occurring within established parameter ranges trigger change control and require re‑validation.
2. Is the root cause established with sufficient evidence? “Likely operator error” or “has not recurred” do not constitute acceptable root‑cause conclusions. Proceeding with production without a definitive root cause reduces validation to a “run‑until‑pass” exercise—leaving the underlying risk intact.
3. Do corrective actions alter the process state? Measures that merely restore baseline conditions (e.g., recalibrating an instrument) may leave process design unchanged. But adjustments to parameter ranges, material specifications, or control logic fall under change control. Batches before and after such changes represent distinct process states and must not be aggregated for validation counting.
4. What variability does the failure reveal? The batch position matters, but more critical is whether the failure exposes unanticipated within‑batch, between‑batch, or unit‑operation‑specific variability—this dictates the depth of additional investigation required.
5. How robust is the remaining evidence base? Development‑stage knowledge, engineering‑batch data, and statistical trends from completed validation batches should be integrated to assess whether available data adequately support process reproducibility. FDA no longer prescribes a fixed PPQ batch count; it requires data sufficient to demonstrate process control, with batch number commensurate with risk and complexity.
6. Is the final report transparent and defensible? All investigation steps and decision rationale must be fully documented. The report cannot merely state “deviation closed”; it must explain why the original validation conclusion remains valid—or why it no longer does.
Part 4: Pitfall Alert — Three High‑Risk Practices
The following practices carry significant compliance exposure:
Retrospective reclassification. Batches initiated and tested under an approved validation protocol cannot be retroactively relabeled as “engineering” or “trial” batches because results were unfavorable. FDA requires validation batches to be executed using the approved commercial process; non‑conforming data must not be excluded without scientific justification.
Substituting volume for investigation. Repeatedly manufacturing batches until three passing lots are accumulated—without completing root‑cause analysis. Validation exists to prove the process performs consistently under routine conditions, not that it can occasionally succeed. No number of subsequent passing batches neutralizes an uninvestigated failure.
Confounding process states. Aggregating batches produced before and after a substantive process or control‑strategy change for validation counting. Once corrective actions materially modify the process, the question shifts from “how many batches to add” to “how to validate the revised process.” Batches from different process states must be evaluated separately.
Conclusion
A batch failure during three‑run process validation neither nullifies all prior work nor can it be resolved through post hoc data selection. The test is not how well the problem is concealed, but how rigorously it is addressed. Data from completed batches must be preserved and reported. Once the consecutive‑success evidence chain is broken, it cannot be repaired by cherry‑picking results.
The decision must always rest on scientific investigation and risk‑based judgment—not on mechanical batch counting.
