Survey: State of the Art – 100% Visual Inspection of Parenterals (2025)
This survey summarizes current industry practice and regulatory alignment for the 100 % visual inspection of parenteral products.
The survey among members of the ECA Visual Inspection Group, carried out in September/October 2025, provides a representative overview of how companies have implemented and currently operate the 100% visual inspection of parenteral products. A total of 115 professionals participated in the survey. The results reveal a broad consistency with the main regulatory and compendial requirements defined in EU GMP Annex 1 (8.30-8.33), USP <790> and <1790>, Ph. Eur. 2.9.20 and 5.17.2, as well as the FDA Draft Guidance Inspection of Injectable Products for Visible Particulates (December 2021).The findings also align with the recommendations of the new ECA Visual Inspection Guide Version 5.0, which integrates the regulatory expectations into a unified best practice framework.
Manual visual inspection (MVI) remains the predominant method, reported by 86 % of respondents, while 44 % employ semi-automated (SAVI) and 49 % fully automated (AVI) systems. This reflects a diverse mix of technologies in use, but also confirms that manual inspection continues to serve as the regulatory reference method, which is also needed when using an AVI system.
Manual inspection working conditions were also well aligned with GMP expectations. Most companies reported uninterrupted inspection sessions between 15 and 30 minutes (46 %) or between 30 and 60 minutes (42 %), followed by defined breaks. These values fit the intent of Annex 1 (8.31), which calls for frequent rest periods to avoid fatigue; and correspond to USP <1790> recommendations on ergonomics and operator performance.
Figure 1: Which type of inspection do you perform? (Multiple answers possible)
Recommendation
Bernried, Germany10/11 March 2026
Low Endotoxin Recovery/Masking - Hands-on Laboratory Training Course
Figure 2: What is your maximum uninterrupted manual inspection time?
The same clause requires annual re-qualification of inspectors, which is achieved by 86 % of respondents - an excellent level of compliance. Only small minorities reported longer (or shorter) intervals, which are still acceptable if justified by risk assessment and performance trending. Inspector qualification should, as described in Annex 1 and USP <1790>, include eyesight checks, use of defect libraries covering worst-case conditions, and evaluation of both sensitivity and false-reject rate.
Regarding the use of magnification during manual inspection, 22 % of respondents indicated that magnifiers are used, whereas 78 % inspect without any optical aid. This result corresponds well with the expectation that manual inspection should be performed without magnification, as defined in USP <790> and Ph. Eur. 2.9.20, which specify visual examination by the unaided eye under controlled illumination.
When asked how many objects are inspected simultaneously, more than half of those surveyed (59 %) stated that they allow 1–2 objects, which is also in line with good practice. However, there is no pharmacopoeial limitation on the number of samples that may be inspected simultaneously. AQL data may be used to justify the simultaneous inspection of more than two samples. The advantage of inspecting several containers at the same time is that it is easy to detect over- or underfilling.
The majority of companies have also set clear limits on the number of visual inspections of a batch as part of the manufacturing. In response to the question "How many times can a lot be 100 % inspected as part of manufacturing?", 22 % allow only a single inspection without re-inspection, 50 % permit one re-inspection, 23 % two re-inspections, 2 % more than two, and 3 % report no defined limit. These figures are very interesting. Not allowing any re-inspections (22 %) is very strict. 73 % allow one or two re-inspections during manufacturing, which is in line with current regulations (the FDA tends to take the position of allowing only one re-inspection). However, 5 % allow more than two re-inspections or have no limit at all on the number of 100 % visual inspections. This could be argued by describing visual inspection not as a test but as a pure manufacturing step. However, if there is a regularly high number of re-inspections, questions about the validity of the preceding manufacturing processes will probably have to be answered in a GMP inspection.
Figure 3: How many samples are inspected simultaneously (in manual inspection)?
The majority of companies (50 %) permit one re-inspection of a lot after an AQL failure, while 21 % allow two re-inspections and a small proportion (3 %) report no defined limit. Such open-ended re-inspection practices represent a compliance risk, as Annex 1 requires investigation and trending of unusual defect levels and USP permits second inspections only under validated and controlled conditions. The new ECA Guide 5.0 proposes a procedure to follow when particles are found in the AQL test, recommending a second 100 % inspection only after documented root-cause analysis, corrective action, and with a tightened AQL acceptance criterion.
Nearly all participants (97 %) perform an Acceptable Quality Level (AQL) test after the 100 % inspection step, which is precisely the approach described in USP. The responsibility for performing the AQL test lies primarily with QA/QC departments (56 %), while 42 % indicated that production personnel perform AQL testing as part of the manufacturing. This distribution shows that more firms treat the AQL as a quality control activity rather than an in-process production check.
Recommendation
Bernried, Germany12/13 March 2026
Monocyte Activation Test (MAT) - Hands-on Laboratory Training Course
Concerning particle classification, 52 % of respondents classify visible particles as critical defects, 44 % as major, and only 4 % as minor. This distribution shows that most companies adopt a conservative approach, treating visible particles as critical. USP and describe visible particulates as potentially critical to patient safety, although the pharmacopoeia also suggests classifying them as major defects for batch-level acceptance decisions (for non-extrinsic particles). The survey therefore shows slightly stricter internal classification practice than the compendial minimum.
Figure 4: How many times can a lot be 100 % inspected as part of manufacturing?
When asked about validation criteria for automated systems, 63 % of respondents base acceptance on performance equal to or better than human inspection, while 26 % apply fixed detection-rate targets for different defect criticalities. Both approaches are supported by USP , which recognizes manual inspection as the reference and allows quantitative performance metrics such as probability of detection (PoD). Comparison with human inspection results therefore remains the so-called gold or reference standard.
The use of functional test kits in routine inspection also reflects mature practice: 34 % of respondents perform tests before and after each batch, 28 % either before or after, and 20 % every working day. These results meet the requirement of Annex 1 (8.32), which calls for challenge testing of automated systems prior to start-up and at regular intervals. They also correspond to the recommendations in USP and ECA 5.0, both emphasizing the importance of continuous performance verification for automated inspection equipment.
Figure 5: What is the maximum number of times a lot can be inspected with regards to the AQL result?
It was also asked how often revalidation of automated visual inspection (AVI) systems takes place. The results show that most companies perform revalidation within 6 to 12 months (66 %), which corresponds to an annual cycle and is well aligned with current regulatory expectations. Only 1 % reported a revalidation interval of less than six months, typically in cases where process or other changes require an early review. A smaller group performs revalidation every two years (14 %), while 19 % indicated intervals longer than two years. The survey therefore demonstrates that two-thirds of companies follow an annual or shorter revalidation frequency.
Artificial intelligence is not yet in routine use for visual inspection; only 1 % of respondents apply AI-based algorithms today, while 27 % plan to do so in the near future. This cautious adoption is consistent with the current regulatory landscape. Although no pharmacopoeia or GMP requirement explicitly addresses AI, its use falls under the same validation principles as any automated classifier: equivalence to manual reference, dataset representativeness, ongoing monitoring for drift, and documented change control. The ECA Guide 5.0 and USP both emphasize that any novel technology must demonstrate equal or improved detection capability and maintained process control.
Figure 6: How often do you use function test kits during routine inspection?
Figure 7: What is your revalidation interval for automated inspection programs?
Overall, the survey demonstrates strong industry alignment with the main regulatory and compendial requirements. Manual inspection continues to play a central role as reference method, while automated systems are increasingly validated to equivalent or better performance. AQL testing and trending are widely established, and most companies follow recognized best practices in functional challenge testing and operator qualification. The areas where improvement is still needed include the consistent definition of re-inspection limits after AQL failures, and the quantitative evaluation of automated system performance by defect class. By addressing these points, companies can ensure full alignment with EU GMP Annex 1, USP and, Ph. Eur. 2.9.20 and 5.17.2, and the practical recommendations consolidated in the ECA Visual Inspection Guide Version 5.0.
About the Author:
Dr Robert Eicher is Operations Director and organises and conducts courses and conferences on behalf of the ECA Academy around pharmaceutical technology.
