Out-of-specification test results, and how FDA expects labs to investigate them

In fiscal 2025, FDA investigators cited 21 CFR 211.192, the rule that requires drug makers to investigate failed batches and unexplained discrepancies, 236 times on Form 483 inspection reports at drug sites, according to the agency's fiscal 2025 inspection observation file². That was up from 173 citations in fiscal 2024³. Under its main citation heading alone, "Investigations of discrepancies, failures," the rule drew 164 citations, second only to a requirement that quality control units have written procedures and follow them.

A large share of those investigations start in the lab, with a test result that falls outside its approved limit. FDA calls that an out-of-specification, or OOS, result, and it has published guidance on handling one since 2006. A Federal Register notice⁶ announced the final version in October of that year, after a 1998 draft, and FDA issued Revision 1⁴ in May 2022.

The guidance matters because the easiest way to make a failing batch disappear is to test it again until it passes. FDA's guidance rejects that approach, and its inspectors look for it. This piece explains what counts as an OOS result, the two phases of an investigation, the statistical shortcuts FDA rejects, and where OOS findings surface in recalls and enforcement.

What counts as out of specification

FDA's guidance defines OOS results as "all test results that fall outside the specifications or acceptance criteria established in drug applications, drug master files (DMFs), official compendia, or by the manufacturer," including in-process tests. It covers chemistry-based testing of drugs regulated by FDA's drug center, and it applies to contract labs on the same terms. "The responsibility of a contract testing laboratory in meeting these requirements is equivalent to that of a manufacturing firm," the guidance says. A footnote adds that much of it "may be useful" for results that are out of trend but still in specification.

Three CGMP sections set the frame. Section 211.160⁸ requires "scientifically sound and appropriate" specifications and test procedures, with any deviation "recorded and justified." Section 211.165(f)⁹ says products failing their standards "shall be rejected." And section 211.192⁷ requires that any failure "be thoroughly investigated, whether or not the batch has already been distributed," that the investigation "extend to other batches" that may be affected, and that a written record include "the conclusions and followup."

Phase I, the laboratory check

Phase I asks whether the lab made a mistake. The analyst is told to keep the test preparations and tell a supervisor when a result looks wrong, and not to finish an analysis already known to be invalid "for the sole purpose of seeing what results can be obtained." The supervisor then reviews the method with the analyst, examines raw data such as chromatograms, checks the calculations and any unauthorized changes to automated calculations, confirms instrument performance and reference standards, and documents it all.

FDA sets a high bar for blaming the lab. "Laboratory error should be relatively rare," the guidance says, and frequent errors point to training, equipment or "careless work." A result can be invalidated only when there is clear evidence of a lab error. If the evidence is unclear, the investigation moves to Phase II.

Phase II, the full investigation

A full-scale investigation is run by the quality unit and reaches into manufacturing, process development, maintenance and engineering, and into every site involved if production is split among contract manufacturers. It may include retesting part of the original sample or resampling the batch. The retest plan has to be fixed in advance in a written procedure, with the maximum number of retests set before any results come in.

FDA names the practice it is trying to stop. "FDA inspections have revealed that some firms use a strategy of repeated testing until a passing result is obtained, then disregarding the OOS results without scientific justification," it says. "This practice of 'testing into compliance' is unscientific and objectionable under CGMP." If no lab or calculation error turns up, "there is no scientific basis for invalidating initial OOS results in favor of passing retest results."

Averaging gets the same treatment. Mixing the original failing result with retest results "is not appropriate because it hides variability," the guidance says, and it gives two examples for an assay with limits of 90 to 110 percent of label strength.

Source: FDA, Investigating OOS Test Results (2022), section IV.C. Both sets average to a passing 90 percent, and in both cases FDA says the individual results, "not the average," should decide the batch.

Statistical outlier tests are restricted too. For validated chemical tests the guidance says an outlier test "should not be used to invalidate the suspect result," citing a February 1993 federal district court opinion in New Jersey, and it says outlier tests have "no applicability" where variability is the thing being measured, such as content uniformity or dissolution.

How an investigation ends

A confirmed OOS result means the batch fails, and the inquiry becomes a batch failure investigation that must extend to related batches. An inconclusive investigation, where no cause is found but the result is not confirmed either, leaves the quality unit to weigh the original result fully. The guidance describes one case where release might be defended: an initial assay of 89.5 percent followed by seven retests between 98.8 and 99.1 percent, with no lab error found, no production problems and consistent results from other tests. Even then, it says, the quality unit "should always err on the side of caution."

Low results that pass also deserve a look. The guidance warns that an assay near the bottom of its range may signal improper formulation and may project to a failure before the expiration date.

Where OOS results show up

For approved products, an OOS result on a batch already distributed can trigger a Field Alert Report, which 21 CFR 314.81(b)(1)¹⁰ requires within three working days. The guidance says an initial report should go in unless the result is invalidated within three days. FDA received 4,059 initial reports of that kind in fiscal 2025, and 17 percent involved "out-of-specification and stability testing issues," according to its FY2025 quality report¹².

Some become recalls. FDA's 2022 recall guidance¹³ lists "out-of-specification testing results" among the warning signs that should prompt a firm to investigate a distributed product. Of 18,076 drug recall records since June 2012, 1,012 mention out of specification or OOS in their reason, including 136 from 2024 through Oct. 28, 2026, according to GxRadar's analysis of openFDA¹⁴. FDA's FY2024 quality report¹¹ said most failed-specification recalls that year were found in stability testing, with impurities above limits accounting for 46 percent and dissolution failures for 36 percent. Our pieces on nitrosamine impurities and the GxRadar recall tracker follow those categories.

Weak investigations also slow approvals. The FY2025 report found that "inadequate investigations" were one of three top inspection citations behind facility-related holds on drug applications, which end in complete response letters. Inspection findings can escalate to warning letters, and FDA issued 105 quality-related warning letters to human drug manufacturing sites in fiscal 2024, the most in five years, according to the FY2024 report. Our explainers on 483s and warning letters and data integrity cover what follows a citation.

FDA's inspection observations page¹ lists citation files through fiscal 2025. The fiscal 2026 file, covering inspections that ended by Sept. 30, 2026, had not been posted when this was written, and the guidance docket⁵ remains open for comment.