ECA Visual Inspection Working Group

Visual Inspection of medicinal

products for parenteral use

Good Practice Guide

Version 3.2 1 (16)

ECA Visual Inspection Working Group

Acknowledgement

This ECA Good Practice Paper was developed by the Steering Committee oft he ECA Visual
Inspection Working Group

Authors

• Dr Tobias Posset (Roche Diagnostics)

• Dr Helmut Gaus (Winsol, formerly Boehringer Ingelheim)
• Martin Dearden (M&F Quality Pharma Solutions Ltd)
• Dr Martin Becker (IDT Biologika)
• Al Goodwin (Amgen)
• Dr Robert Eicher (Concept Heidelberg)

Review

• Klaus Feuerhelm (GMP Inspector, Germany)

• Dr Bernd Renger (Past Chair of the European QP Association; Bernd Renger
Consulting)

Non-liability: While every effort has been made to assure the accuracy oft he content,
Concept Heidelberg or the ECA Foundation cannot be held liable for any errors or omissions

Version 3.2 2 (16)

ECA Visual Inspection Working Group

1. Scope
This paper aims to highlight best practices for carrying out 100 % visual inspection of
medicinal products for parenteral use in the pharmaceutical industry. It should be
seen as additional to and complementary to the monographs of the different
Pharmacopoeias. Visual inspection of medicinal products for parenteral use should
detect any readily identifiable visible container defect and ensure constant quality of
the product in terms of absence of particular matter and/or turbidity, and correct or
uniform appearance of a lyo cake.

Modifications from the procedures and figures proposed in this paper are possible
any time. However, following the proposed procedures and figures may lead to safer
inspection processes and may avoid discussions in GMP audits and inspections, as the
described approach reflects current practice and has shown its suitability during
many years of industrial operation and GMP inspections. The CCI (Container Closure
Integrity) topic is out of scope (see ECA position paper “Container Closure Integrity
testing of medicinal products for parenteral use”).

2. Manual inspection

2.1 Workplace
The premises were the manual visual inspection takes place should be suitable
for carrying out this operation. Besides the common GMP requirements for
manufacturing or quality control areas, the following inspection conditions are of
substantial importance:
Illumination: The intensity of the illumination at the inspection point should have
at least 2000 lux. For Blow-Fill-Seal Products an illumination of 10.000 lux is
recommended. The total uncertainty of the lux-meter should be considered. The
color reproduction, using the CRI index, should preferably have an RA value > 90%
and not less then 80%.

Illumination should be regularly qualified and part of the preventive maintenance

programme. An appropriate interval for checking the illumination is 6 months. It
is recommended that the technical measurement used to determine the light
intensity is performed at a prefixed point, which should be very close to the
inspection point of the operator.
The ambient illumination must not interfere with the illumination of the
workplace and should be turned down during inspection, if possible. Reflecting
surfaces should be avoided.

Version 3.2 3 (16)

ECA Visual Inspection Working Group

Ambient conditions:
Ambient conditions have a high impact on the quality of this operation.
Temperature should be room temperature and not exceed 25 °C in the summer if
not otherwise justified. The relative humidity and air velocity should be
controlled and ensure comfortable working conditions. The noise level should
therefore be below 55dB.

2.2 Personnel
Personnel involved in the visual inspection should undergo regularly eye tests.
The optometrist should focus on the ability to discriminate small differences in
uniform structures, e.g. open/closed circles. Also color vision of the inspector
should be tested.

Personnel performing visual inspection must be qualified, comprising initial

qualification and periodic requalification.

Initial qualification should follow a predefined schedule, starting with the

introduction of the new employees to training kits. These training kits should
contain all kind of defects and should be updated constantly with new evolving
defects out of production. These training kits should be specific for the dosage
form. In case of unstable defects or defects which are not available photos of
defects or artificial defects may be used.

A bracketing approach for similar products is acceptable. Following the training

via training sets there should be a side to side training with an experienced
operator. The trainee should have the chance to ask questions with the
experienced operator performing a parallel 100 % inspection of the inspected
species of the trainee.

Following successful completion of the side by side training the initial

qualification should be performed with a qualification set. This set should focus
on critical and major defects and also contain a limited number of minor defects.
The number specimen to be inspected should represent the number of objects
which are inspected in the routine process from one eye break to the next eye
break.

The European GMP requirements with regard to the visual inspection are
included in EU GMP Guide, Annex <1> “Manufacture of Sterile Products” in the

Version 3.2 4 (16)

ECA Visual Inspection Working Group

section “Finishing of sterile products”. All Annex 1 expectations are covered

within this guidance document.

Even if current regulations cover many topics concerning visual inspection,

situations were the expectations of the inspector differ from to the scientific
understanding of the pharmaceutical company may occur.

Expectation 1:
There is sometimes the expectation for operators to wear corrective lenses when
undergoing visual inspection qualification, if they are worn during daily life.
Corrective lenses are used to shift an observed object into the focus point of the
human eye. It is a known fact that individuals may shift an observed object into
their (individual) eye focus point simply by varying the distance to the object,
without the use of corrective lenses. So operators that need corrective lenses in
everyday life might be able to perform visual inspection without.
The requested medical check examines e.g. three-dimensional sight, colour
identification, stigmatic corrections and most importantly reflecting the normal
“daily life” visual capability. This ensures that the operator can see defects.
The medical check is not related to the competence of performing visual
inspection activity.
This is checked during the qualification of the operator and whether he has to
wear his glasses is determined in this process.

Expectation 2:
There is sometimes the expectation for operators to be qualified taking into
account different stress factors, including fatigue at the end of shift. It can be
doubted that a fatigue effect can be simulated in an examination situation with
the impact for the operator, whether he/she can keep his/her job. With this there
is always nervousness and some people have even examination anxiety which
leads to release of Adrenalin in the human body, which counters a fatigue effect.
Additionally when performing initial qualification the fatigue at the end of shift
can’t be simulated anyway, as the operators are not yet allowed to perform a
“normal” visual inspection workday.
As Annex 1 requests that operators performing visual inspection should be
following a work and rest regime specifically designed to minimize distractions,
there should be no fatigue effects at end of shift, which would impact the validity
of the visual inspection.

Keeping this in mind, it should be considered to address these two

topics/aspects/expectations by a Quality Risk Management approach.

Version 3.2 5 (16)

ECA Visual Inspection Working Group

Acceptance criteria for the qualification should be predefined. In the initial

qualification all critical defects and a predefined level for major defects should be
found. Also, a limit for the False Rejects (rejection of samples without defects)
should be predefined (e.g. 5-10%).
Routine requalification must be performed at least every 12 months. After a
second failure the operator must undergo a repeated eye test and a subsequent
new qualification with the training kit and a subsequent new qualification via a
qualification kit.

The qualification of semi-automated inspection processes/operators should be

performed on the same basis and rationale like the manual inspection (e.g eye
breaks and limits for defects that need to be found within the qualification set).

2.3 Operation
During the 100 % inspection each object should be inspected for at least 5
seconds against a white background and an additional 5 seconds against a black
background. Times may be shorter when using a semi-automatic system. The
objects should be slightly twisted or slowly rotated to swirl up particles whilst
avoiding the formation of air bubbles.

The post inspection recovery time for visual Inspection staff is of essential
importance. The maximum time for continuous inspection activity between
breaks and the total maximum inspection time for a shift/workday must be
defined.

A current good practice is 20 minutes of inspection, followed by a break of at

least 5 minutes continued for a total maximum duration of not longer than 4
hours. Maximum uninterrupted inspection activity should not exceed 40 minutes.

3. Automated inspection
3.1 Qualification / Validation
The central aspect of qualification/validation of a fully automated inspection
system is the verification that the automated system is at least as good as the
“gold standard” human inspector (without magnification) with regards to defect
detection rates. Qualification and validation can be done consecutively or
combined. The performance qualification of the system can be rated as validation
of the process if it is carried out product-specific. In the qualification phases

Version 3.2 6 (16)

ECA Visual Inspection Working Group

before the PQ, format-specific test sets can also be used. Bracketing concepts are
possible. In this case product specific qualification sets have to be used. During
machine qualification the inspection of a qualification set should be repeated at
least 10 times. The detection rate should be compared to the results of a
qualified manual inspection and should be at least as good as manual inspection
based on defect categories. It is worth mentioning that an automated inspection
machine is not capable of detecting and removing certain categories of cosmetic
defects.

NOTE: The detection of particles follows a probabilistic detection principle

depending on the size and potential physical movement of the particle in the
drug product solution being inspected. The detection rate of particles needs to be
viewed following the principle of the so called "Knapp-statistics". Julius Knapp
demonstrated that a ratio of not more than 25% defect vials within a set of non-
defect vials (good vials) gives the best statistical return for the accurate statistical
evaluation of the particle detection rate.
A man-machine comparison for introduction of a new or automated system of
inspection is carried out by multiple (10X) detection rates of a test set of vials
containing particles,(reject vials). Success is when results from the 10 repeats by
the machine and the resulting test statistics are equally as good as when the 10
times repeated vial inspection is performed manually. This is the so called Knapp-
test.

3.2 Routine Operation

During routine operation the actual performance of the automated system
should be demonstrated to be within the acceptable range of the pre-defined
(during PQ) defect detection limits, before and after the inspection process by
using a test set containing the range of defects that has been used in the
qualification. An abridged or reduced function set may also be used.

3.3 Requalification
Requalification of an automated inspection system should be ideally carried out
annually, or every two years at the latest. This must be done by evaluation of the
changes and deviations that occurred during the period of operation. This review
must include a statistical trend analysis of the performance data obtained during
routine inspection and system suitability determinations using the function sets
before and after every machine use.

Version 3.2 7 (16)

ECA Visual Inspection Working Group

3.4 Revalidation
An automated visual inspection machine must be considered a critical system.
Therefore, a periodic revalidation of the operation should be carried out, e.g.
ever 3-5 years. Revalidation (as any validation) is product specific. Bracketing
approaches for the revalidation are possible.

Different approaches may be used:

3.4.1 Revalidation can be done by repeating the verification of the human
inspection versus the inspection results of the machine, for example by
manual re-inspection of an automatically inspected batch or part of the
batch (e.g. 5000 vials). The acceptance criteria are the same as during the
initial validation.
3.4.2 Revalidation may also be done by a continuous revalidation approach
using the AQL results. The AQL is a manual inspection of a representative
sample performed for every inspected batch (man-machine comparison),
constituting a repeated verification of human inspection versus the
automated inspection machine on batch by batch level and thus be
considered a revalidation.
Result and evaluation should be documented in a revalidation report.

4. Inspection of lyophilized product

Lyophilized drug products also have to undergo a 100% visual inspection. However,
as particles can be detected only on the surface of the product cake and not in the
cake itself, an additional test on particulate matter has to be performed. For this, a
number of samples are reconstituted and visually inspected for particulates. The
result has to be within the predefined particle limits. As this supplemental testing is
destructive in nature, it is acceptable that only a limited number of samples are
inspected. Samples from the batch are taken according to a sampling plan (e.g. S-3
and/or S-4 according to DIN ISO 2859).

5. Defect Classes
There should be at least two product-specific defect classes defined. Defining more
defect classes may be appropriate, e.g.:
Critical defects: may cause a lack of sterility, container integrity or cause harm to
patients
Major defects: may alter the content or the function of the product

Further defect classes can be:

Version 3.2 8 (16)

ECA Visual Inspection Working Group

Minor defects: Defects that do not affect patient health or product functionality

If a company performs additional visual inspection for specific cosmetic appearance

defects for products intended to be placed to culturally sensitive markets it is
recommended not to include them into the GMP defect classification.

6. Evaluation of defect classes and trending

6.1 Manual, semi-automated and fully automated (non-inline) inspection
Based on the trend analysis of the production process a limit for each defect class
should be defined. There should be limits for individual defects and for the sum of all
defects within a defect class. Yields should also be monitored.
Limits should be defined on process history (overall maximum reject rate, rate per
defect / particle category) and should reflect and reference the process capability
index (CpK) for the process step.

Typical action limits for individual defects are for example:

Critical defects: 0.5 % to 1 % *
Major defects: 1 % to 3 %
Minor defects 3 % to 5 %
*There are critical defects (e.g. turbidity) with an acceptance limit of 0.

Typical action limits can also be set on specific defects (e.g. particles) instead of using
the categories critical, major and minor.
Setting alert limits may be reasonable e.g. to identify a production specific problem
as early as possible.

According to ASTM E2587-2 (Standard Practice for Use of Control Charts in Statistical
Process Control) control limits for trending should be calculated from process data
and not be based on fixed specification limits. However, the evaluation should
optimally be based on 30 representative data points of the process. This would
require preliminary control limits to be defined based on experience prior to the first
calculation based on measured data points.

The measures to be taken for batches exceeding these limits are to be predefined.
The re-inspection of a batch due to a limit violation should be investigated. If a re-
inspection is performed the method used for re-inspection must not necessarily be
the same method used for the previous inspection. This assessment should be part of
the investigation. For example: batches exceeding one time the acceptance limit
should trigger a failure investigation. Batches exceeding the acceptance limit two

Version 3.2 9 (16)

ECA Visual Inspection Working Group

times should be additionally 100 % re-inspected. Re-inspection should never be

performed without a previous investigation why limits have been exceeded. Re-
inspection should be performed independent of the initial technique used - manual,
semi-automatic or automatic visual inspection, but should be limited to not more
than 2 re-inspections. Limitations of the inspection system should also be taken into
consideration. If an automated system fails in identifying certain defects in the first
run, it is likely to fail also in a second run.

Re-inspection of rejected containers is not recommended and must not be

performed without justification based on a thorough investigation.

NOTE:
“Grey Channel” for fully automated inspection systems: Defining a “grey eject
channel” in automated inspection may be useful to separate containers for which the
inspection result is not clear. From a technical point of view, a “grey channel” is also
meaningful in case of e.g. machine stops, when it is uncertain whether a vial has been
fully inspected or not.

Examples for the Grey Channel:

1. Air bubbles: Air-bubbles might be the reason for an unclear inspection result.
Camera systems cannot distinguish between particulate matter and air-bubbles.
Therefore, re-inspection of containers from the grey channel after a certain
holding time of the product vials in order to reduce the air-bubbles is allowed but
should not be performed more than once.
2. Other defects in the grey channel: From a technical point of view some small
defects (e.g. small scratches) should be re-inspected manually if the camera
system is not capable of detecting the defect. The feasibility is shown during the
validation.

Objects in the grey channel should only be re-inspected one time. Trending is done
over the whole batch after the objects have been classified as good or defects.

6.2 Automated inspection: In-line

An in-line fully automated inspection system should also trend and analyse the
production process. Due to the in-line process and the fact that camera systems
cannot trend defect categories a trending system has to be established on the
technical level e.g. side-wall defect, crimping-defect, shoulder defect etc. Limits of
these cameras and their inspection scope have to be established according the ASTM
E2587-2 standard. This gives a reference to the process capability index (CpK) for the
in-line controlled process.

Version 3.2 10 (16)

ECA Visual Inspection Working Group

The measures to be taken during the batch production exceeding these limits need to
be predefined and should lead to actions during the batch production process. Herein
a re-inspection of the batch is not needed due to the immediate corrective action.

Re-inspection of rejected containers is not recommended and must not be

performed without judgment done within in an investigation.

7. Batch release
For the release decision two criteria need to be evaluated:
1. Trending analysis of the 100% batch inspection (see chapter 6) and
2. The result of the AQL manual inspection.

The results of the 100% visual inspection, done as part of the manufacturing process,
are an integral part of the batch documentation, specifying the types of defects
found (fiber, turbidity, crack, etc.) as well as a classification of the defect such as
critical, major or minor. Acceptance criteria must be pre-defined for these defect
classes of all defects found during the 100% inspection process (see above). For the
automated in-line inspection these classifications needs to be done on a technical
level (side-wall defect, crimping defect,…) and cannot be done using the classification
such as critical, major or minor (see above).

For the AQL manual inspection a randomized sampling of the 100% inspected batch
should be performed according to a pre-determined AQL procedure. AQL manual
inspection can be carried out by production staff under a quality oversight or the
quality unit.

Version 3.2 11 (16)

ECA Visual Inspection Working Group

For release of the batch, the minimum AQL should be

Critical: 0 defects within the AQL samples
Major: ≤ 0.65 AQL
Minor: ≤ 4.0 AQL

If an AQL limit is exceeded, the whole batch may be re-inspected 100% followed by a
second AQL manual inspection (see 6.). This process can be repeated, but must be
justified by an investigation and should not be performed more than 2 times in
maximum. It is recommended to have tighter AQL limits for the repeated AQL
testing. Tightening of AQL limits is possible by two ways: Stay on the AQL sample size
and use the next (tighter) AQL limit- OR: Stay on the AQL limit and use the next
(higher) AQL sample size.

According the USP<790> particles are allowed to be within the AQL manual
inspection. This is connected with AQL limits for particulates greater than 0.01
knowing that the AQL manual inspection is a “statistical tool” where a 100%
assurance of having no visible particles within the batch is never given.
Unfortunately, within the USP no statement regarding the criticality of particles is
given. With this one should investigate the particle(s) found. Any particulate defects
found in the AQL manual inspection must be evaluated to determine whether the
particulate is characteristic of the filling process. This evaluation should include the
nature, the size and the detection rate of the particle within the given Drug Product.
If the particulate defect is not characteristic of the filling process, failure of the AQL
may result. If the particulate defect is at the border of visibility, having a low and
local defined detection rate, a re-inspection of the affected batch may not be
necessary.

The number of AQL manual inspection steps should be evident in the batch
documentation.

AQL manual inspection can be carried out in a separate area, without constraint in
time for the personnel doing the visual inspection.

The results of the 100% visual inspection and the results of the AQL testing are part
of the finished product assessment that should embrace all relevant factors of the
batch. After assessment by quality control or quality assurance these data should be
made available in an easily readable format to the Qualified Person, responsible for
the batch certification and release decision.

Version 3.2 12 (16)

ECA Visual Inspection Working Group

8. Concerns regarding distributed product

When concerns regarding particulate matter of product already distributed to the
market arise, it is recommended to inspect a number of samples following a practical
DIN ISO 2859 compliant sampling plan. Visual inspection for particulates is a
probabilistic process, therefore non-critical defect sampling plans with a limit of zero
are not practical as an assessment tool and should be avoided. The inspection needs
to be performed under the same conditions the routine inspection was performed.
The relative batch and sample size will determine the acceptance criteria, if these
criteria are met by inspection of these samples, the batch can be considered
essentially free of particles. It is good practice to isolate the particle(s) of the
complaint sample and to determine the size and to characterize the nature of the
particle(s) within the investigation.

Version 3.2 13 (16)

ECA Visual Inspection Working Group

9. Definitions

AQL Acceptable Quality Limit, using ANSI/ASQ Z1.4 or ISO 2859-1,

general inspection level II

CRI-Index: The Color Rendering Index (CRI) is a unit of measure that defines
how well colors are rendered by different illumination conditions
in comparison to an ideal or natural light source. The Ra value uses
only the first eight from the 14 test colors of the DIN 6169. Light
sources have different Ra values, e.g. a white LED 80-95;
fluorescent lamps 50-90.

Function set: A function test kit (system suitability test kit) used before and
after the inspection of each batch to demonstrate the
functionality of the fully automated inspection system. It may
contain an abridged set of more apparent defects such as big
particles, cracked or empty containers.

Particle: A particle in the context of this paper means a readily visible

particle with a diameter or span of 150 µm or bigger that allows
reliable detection (70% probability or higher) in a clear solution.
Smaller, for example colored or shiny particles may be visible
down to a size of 50 µm or smaller and thus have also to be
counted to the visible particles.

Qualification set: A set of product specific containers with real product, with 10-20%
but less than 30% of the containers containing randomly
distributed defects. All known defects should be contained in the
test set. Any new defects should be added to the test set after
they have been identified. Units of tests sets for the manual
inspection containing a defect should be marked or encoded
invisibly. Obvious and clear readable numbers or letters should
not be used. The sets should be cleaned after usage and routinely
checked for defects at least every 6 months. There should be a
logbook for each test set.

Training set: A test set used for the training of the operators of the manual
inspection, to teach the possible defects. Similar to the
qualification set but contains only containers with defects.

Version 3.2 14 (16)

ECA Visual Inspection Working Group

Visual Inspection: The process of sorting unacceptable units from acceptable units by
human visual inspectors and/or through qualified equipment.

Fully automated visual inspection:

The process of sorting unacceptable units from acceptable units
by equipment (camera system). Detection and handling of units
to be inspected is performed by equipment.

Fully automated (non-inline) visual inspection:

The process of sorting unacceptable units from acceptable units
by equipment (camera system) after the batch production has
been finished.

In-line fully automated visual inspection:

The process of sorting unacceptable units from acceptable units
by equipment (camera system) parallel or at the same time of
the batch production.

Version 3.2 15 (16)

ECA Visual Inspection Working Group

10. References
• DIN/ISO 2859-1/ANSI/ASQ 71.4
• USP <790> and <1790>
• Ph. Eur. 2.9.20
• Ph. Eur. 5.17.2
• ASTM E2587-12
EU GMP Annex 1

11. Change History

Version Changes

1.0 New Document

2.0 • Editorial adjustments
• Inclusion of a false reject rate into the qualification of human inspectors
• Clearer description of revalidation (chapter 3.4)
• Inclusion of new chapter 4: Inspection of lyophilized product
• Addition of the ASTM E2587-2 standard (Standard Practice for Use of Control
Charts in Statistical Process Control) in chapter 6 (Trending)
• More detailed description of the batch release concerning the visual
inspection data (chapter 7)
• Harmonization of the AQL limits with USP in chapter 7
• Explanation on how to tighten AQL limits in chapter 7
• Number of possible re-inspections adjusted to 2 (Chapter 6 & 7)
• Inclusion of new chapter 8: Concerns regarding distributed product
3.0 • Editorial Adjustments
• Usage of artificial test sets amended in chapter 2.2
• Test of the color vision of human inspectors amended in chapter 2.2
• Usage of format specific sets possible in qualification phases before PQ
(chapter 3.1)
• Precision section 7 “Batch Release” including visible particles being allowed
within the AQL manual inspection according USP<790>
• Reference list added as chapter 10
3.1 • Adding CCI statement (being out of scope for this document)
• Precision of minimum AQL level for batch release
• Deletion of the requirement for “sum of defects”
3.2 • Statement added regarding the Annex 1 changes:
1. manual inspection qualification condition under worst case conditions
(section 2.2)
2. use of glasses/corrective lenses within manual inspection (section 2.2)

Version 3.2 16 (16)