Gravimetric Sample Preparation

White Paper
Reducing Sample Size and OOS Errors

Contents

1 What is gravimetric sample preparation?

2 Source and impact of OOS results

3 GWP® – The Weighing Standard

3.1 Science based compliance

3.2 Measurement uncertainty and minimum weight

3.3 Safety factor to compensate for variability

4 Potential errors in sample preparation

4.1 Types of laboratory variability

4.2 Use of volumetric glassware

4.3 Mixing

4.4 Labelling

5 Gravimetric sample preparation

5.1 Gravimetric sample preparation equipment

5.2 Effect on diluent addition

5.3 Effect on process security (data management)

5.4 Effect on process efficiency

5.5 How can a gravimetric approach reduce the sample volume?

5.6 Volumetric vs. gravimetric comparison

6 Conclusion

7 References
 1 What is gravimetric sample preparation?
White Paper

The definition of gravimetric sample preparation is weighing not only the

solid but also the solvent to enable a specific concentration to be prepared
accurately and precisely.

Gravimetric sample preparation offers a revolutionary alternative to the tra-

ditional way we prepare samples and standards. It dramatically reduces
the variability in the samples and minimizes the potential for out-of-speci-
fication results.

2 Source and impact of Out-of-Specification results

Out-of-Specification (OOS) results have had a significant impact in the pharmaceutical industry for many years,
but in particular since a court ruling in 1993. In this case, the court ruled in favor of Barr Labs which upheld their
view that an OOS result does not necessarily constitute a batch failure (1). They felt that an OOS result should be
investigated to determine if there are other causes such as a laboratory error. However, the court did not like the
way Barr Labs was conducting their laboratory investigations. As a consequence, the FDA updated their guide-
lines concerning how to handle an OOS result and how to perform a proper investigation in October 2006 (2).
Since then, the FDA has issued a significant number of 483 observations concerning poor investigations.
­Furthermore, in the FDA guidance concerning OOS investigations, the FDA states that:

“Laboratory errors should be relatively rare. Frequent errors suggest a problem that might
be due to inadequate training of analysts, poorly maintained or improperly calibrated
­equipment, or careless work.”

However, as there are a significant number of FDA 483 observations relating to poor investigations, the incidence
of laboratory errors may not be as rare as it should. Unfortunately, there is no published data which shows that
for every OOS result generated there were many more minor errors that didn’t lead to an OOS result. These errors
may have been classified as a “Note to Record”, or simply noted in the laboratory notebook as an error. Many
companies don’t investigate these errors even though they are probably symptoms of potentially more serious
­issues with the analysis method or process.

An article was published some years ago in LC/GC magazine concerning OOS results (3). The article ­discussed
two aspects of laboratory work: firstly, what are the sources of OOS errors and secondly, what activities con-
sume laboratory personnel’s time in an analytical workflow (Full Time Equivalent (FTE) spend in the laboratory).

2 METTLER TOLEDO White Paper

 Time Spent in Lab Sources of OOS Errors
Sample Introduction (6%)
Data Management (27%) Contamination (4%)
Chromatography (7%)
Columns (11%)
Integration (6%)

Collection (6%) Instrument (8%)

Analysis (6%)
Calibration (9%)
Operator (19%)

Sample Processing (61%) Sample Processing (30%)

Figure 1: Sources of OOS results and time spent in laboratory.

The results from this survey indicated that the two largest sources of OOS results are sample processing followed
by human error and the most time-consuming (manual) task is sample processing (Figure 1). Even though no
follow-up survey has been published since, it appears that these results are still valid today. In fact, the instru-
mentation, data systems, and columns have improved significantly during the last 10 years, whilst the sample
processing has remained essentially the same. Since these other developments have saved a significant amount
of time in the workflow, it is fair to assume that the sample processing aspect of the laboratory work is likely to
account for even more than 61% of the FTE.

Figure 2 illustrates the accepted formal process for how to investigate OOS results:

Out-of-Specification Result

Laboratory Investigation
Checklist Approach: Investigation of standards used,
analytical techniques, etc.

No Yes
Lab Error Identified ?

Retesting Error in Sampling

Procedure and Resample

No
Variation or Error in Manu- Retesting According to SOP Repeat Test to Substitute
facturing or Sampling or Protocol the OOS Result

Confirm Variation Additional OOS

in Manufacturing Result occur No further OOS results
Process

Stop Production Release

Figure 2: Formal process for an OOS investigation.

White Paper METTLER TOLEDO 3

 A great deal of effort is expended when a sample generates an OOS result and there have been numerous arti-
White Paper
cles published on how this process should work. Obvious OOS results can take three days of work but serious
ones can take months of work to resolve. The cost can easily run into many thousands or tens of thousands of
dollars. Given the large impact that an OOS result has on a pharmaceutical company, the best course of action
should be to put every effort into avoiding them in the first place.

Besides trying to determine the root cause, the other significant issue seems to be the mounting Corrective and
Preventative Actions or CAPAs that a company generates over a number of years as a result of these laboratory
investigations. These CAPAs typically cause procedural changes to SOPs and other documents and over time
they become unmanageable and difficult to follow which has the potential to cause even more issues.

The overriding problem with CAPAs is that, in the vast majority of cases, the assumption tends to be made that
it is an isolated incident and so only a specific item in a workflow or process is addressed. In other cases, there
is a tendency to blame a single employee or a simple laboratory error. In some cases, this simple error may be
the only thing that that needs to be addressed but in many, if not most cases, it is the whole process or workflow
that needs to addressed instead of one step. This is especially true of sample preparation in the laboratory.

3 GWP® – The Weighing Standard

3.1 Science based compliance

Before we elaborate further on the sample preparation process and its individual steps, we would like to highlight
the scientific background of one of the most important steps of the whole process – weighing. Weighing is a key
activity in most laboratories but it isn’t always sufficiently understood and its complexity is often underestimated.
As the quality of weighing strongly influences the quality of the whole sample and standard preparation process,
USP specifically requires in its General Chapter <41> highly accurate weighing results used for quantitative
analysis (4).

“Repeatability is satisfactory, if two times the standard deviation of the weighed value,
divided by the desired smallest net weight, does not exceed 0.10%.”

Such a stringent requirement is not implemented for other instruments, where quite often the analytical develop-
ment group sets the method requirements.

State-of-the-art strategies for adhering to consistently accurate and reliable weighing processes comprise of sci-
entific methodologies on balance selection and testing (5, 6, 7). The weighing standard GWP takes these prin-
ciples into account, and furthermore addresses typical misconceptions on weighing which are very widespread
in the industry.
One of them is that many users believe “what you see is what you get”. What do we mean by that? Here is an
example: A user weighs a standard on a semi-micro balance and gets a reading of 50.13 mg which he believes
is the true amount of material that he was weighing. However, this reading might not exactly reflect the amount
weighed, in other words, the amount weighed might differ slightly from the indication. This is due to the so-called
measurement uncertainty which is inherent for every instrument you might think of.

4 METTLER TOLEDO White Paper

 3.2 Measurement uncertainty and minimum weight

Measurement uncertainty of instruments is determined by calibration, and the results issued in appropriate cali-
bration certificates. In general, measurement uncertainty of weighing systems can be approximated by a straight
line – the higher the load on the balance, the larger the (absolute) measurement uncertainty becomes, as shown
in Figure 3.

Relative measurement uncertainty [%] Absolute measurement

(= Absolute measurement uncertainty / weight) uncertainty [mg]
Uncertainty U [mg or %]

Weighing accuracy [%]

Accuracy limit: Weight [g] Max

Minimum sample weight

Figure 3: Measurement Uncertainty: Absolute (green line) and relative (blue line) measurement uncertainty of a weighing instrument.
The accuracy limit of the balance, the so-called minimum weight, is the intersection point between relative measurement uncertainty
and the required weighing accuracy.

Looking at the relative measurement uncertainty, which is the absolute measurement uncertainty divided by
the load, and usually indicated in per cent, we clearly see that the smaller the load is, the larger the relative
measurement uncertainty becomes. If you weigh at the very low end of the balance‘s measurement range, the
relative uncertainty can become so high that the weighing result cannot be trusted anymore. It is good practice
to define accuracy (tolerance) requirements for every weighing process. For quantitative analysis this is even
stipulated by USP General Chapter <41>. Weighing in the red area as indicated in the figure will result in inaccu-
rate measurements, as here the measurement uncertainty of the instrument is larger than the required accuracy
of the weighing process. Consequently, there is a specific accuracy limit for every weighing instrument – the
so-called minimum sample weight, or short, minimum weight, and you have to weigh at least this amount of
sample in order to have a sufficiently small uncertainty that satisfies the specific weighing accuracy requirement.

While measurement uncertainty is described in much detail in the respective literature (8, 9), we want to
emphasize that for weighing small loads on analytical and microbalances – and samples and standards usu-
ally are small loads as compared to the capacity of the balance – the dominant contribution factor to weighing
uncertainty stems from repeatability (expressed as the standard deviation of a series of weighings). This is also
reflected in USP General Chapter <41> as discussed above.

Even though adherence to this USP requirement seems to be straightforward, many companies still have issues
with the correct interpretation. While environmental influences and operator variability which contribute to inde-
terminate errors and consequently to possible changes or fluctuations of the reading of a weighing device are
discussed later, another misconception which is prevalent in the industry is briefly discussed now. Many compa-
nies wrongly believe that the weight of the tare also accounts for the adherence to the minimum weight. In other
words, if the tare weighs more than the minimum weight, any sample quantity can be added and USP<41> is
automatically fulfilled. This would mean that with a large enough tare container you might believe that you could
even weigh a microgram on a 5-place balance and still comply with the uncertainty requirement of 0.1%. Such
an extreme example clearly shows us that this widespread misinterpretation indeed does not make any sense.
For this reason USP has attempted to clarify this issue in the latest revision of General Chapter <1251> (10):

“The minimum weight applies to the sample weight, not to the tare or gross weight.”

White Paper METTLER TOLEDO 5

 3.3 Safety factor to compensate for variability
White Paper

The minimum weight of balances is furthermore not constant, but varies over time. This is due to changing
environmental conditions that affect the performance of the instrument, as for example vibrations or draft. The
operator himself also adds variability to the minimum weight, as different people might weigh differently or with
a different skill level on the balance. In order to ensure that you always weigh above the minimum weight as
determined at calibration (at a particular time with particular environmental conditions by a particular qualified
person), it is highly recommended to apply a safety factor, see Figure 4.

Measurement uncertainty [%]

Relative measurement uncertainty [%]

Accuracy limit:
“Minimum weight”

Weighing tolerance [%]

Safe weighing range

Load

Figure 4: Safety Factor: Variability of the relative measurement uncertainty due to changing environmental conditions and influences
introduced by the operator. Weighing in the green area guarantees adherence to the weighing accuracy requirements (application of a
safety factor).

The safety factor describes that you would only weigh sufficiently above the minimum weight as determined at
calibration. For manual weighing, a safety factor of 2 is commonly used, provided you have reasonably stable
environmental conditions and trained operators. For very critical applications or a very unstable environment, an
even higher safety factor is recommended.

Later in this paper we will elaborate more in detail about typical minimum weight values and recommended
safety factors for automated gravimetric sample preparation as compared to manual weighing (see 5.5 How can
a gravimetric approach reduce the sample volume?).

4 Potential errors in sample preparation

4.1 Types of laboratory variability (errors)

To be able to deal with variability in the laboratory, it is important to first understand the types of variability and
where they occur. Variability in the data generated comes from two sources, determinate and indeterminate
errors. A determinate error has a definite direction and magnitude and has an assignable cause, their cause can
be determined. Determinate error is also called systematic error. Determinate errors can (theoretically) be elimi-
nated through instrument adjustments. Indeterminate errors are also called random errors, or noise. Indetermi-
nate errors can be minimized but cannot be eliminated. Examples of these types of errors are described in more
detail in 5.6 Volumetric vs. gravimetric comparison.

The largest cause of indeterminate errors in the laboratory is from manual operations where the human element
is involved. Figures 1 highlights the issue that sample processing and human operations are the biggest source
of laboratory errors.

6 METTLER TOLEDO White Paper

For a typical simple sample preparation workflow, Figure 5 demonstrates the number of steps involved.

Material and equipment

• Gathering materials – volumetric glassware, standards, reagents and samples

• Check balance cleanliness, minimum weight requirements, calibration date
• Check other instruments for calibration / qualification dates and record
• Prepare diluents – combining solvents in correct proportions, pH adjusting, etc.
• Record manufacture, lot number, expiration dates for all materials
• Labeling – hand written or generated

Weighing and labeling

• Tare weigh container

• Perform weighing operation by:
– carefully adding material into weighing container
– closing door and allowing balance to settle
– repeat until target weighing obtained
• Record weight and carefully transfer powder to volumetric
• Reweigh container, perform net weight calculation, dispose of container
• Label all in a compliant manner, safety and GMP

Sonicating and filling to the line

• Fill with specific amount of diluent, mix / sonicate and cool to room temperature
• Fill to the line with diluent
• Successive dilution, if required, fill to the line and mix
• Record data and label volumetric flasks
• Transfer to vials and label vials
• Repeat steps for each preparation

Analyze and clean up

• Place vials in instrument and perform analysis

• Volumetric flasks (after results are verified):
– rinse with solvent followed by water
– remove any labelling with solvent
• Have volumetric flasks and pipettes transported and washed
• Retrieve clean volumetric flasks and pipettes, place in storage
• Re-order ones that didn’t make it back from dishwasher or were damaged

Figure 5: Simple sample preparation workflow.

The steps are grouped into four distinct activities. The first concerns gathering materials and ensuring the equip-
ment is clean and calibrated. There are a number of steps at the beginning that would not necessarily cause an
OOS result but would certainly feature in both GxP and safety audits. Resolving these audit issues can consume
a significant amount of time and effort and should also be avoided as they may lead to future OOS results.

The second activity involves weighing and labeling steps. Typically, these are manual time consuming opera-
tions and the weighing steps can contribute to OOS results. An additional difficulty is that these steps are manu-
ally intensive, so it can potentially be very difficult to determine the root cause of this operation.

White Paper METTLER TOLEDO 7

 Following weighing, the diluent must be added. This is followed by dissolution and any successive dilutions that
White Paper
may be required. The weighing, sonication and filling to the line steps are repeated for each standard and sample.
Dilution steps are a big potential source of errors, and consequently this activity will be discussed in more detail
in the section below.

Finally, the samples are analysed and the materials and equipment are tidied up. This requires disposal of
unused solutions, rinsing of flasks and pipettes, and other resupply steps.

Therefore, a simple sample preparation takes ten or more steps with an additional ten miscellaneous steps. If
two standards and a sample were prepared, approximately 40 steps would need to be carried out. A 40 step
process has a significant number of areas where problems could potentially occur. Furthermore, some of these
steps can be expanded and a detailed analysis with more complex steps, including operations such as extrac-
tion and filtering, might result in a process with one hundred or more steps. Given this number of manual steps
where indeterminate errors occur, it is surprising that OOS results are not even more frequent. Fortunately, many
but not all of these errors are found before the final results are obtained but they do significantly impact the pro-
ductivity of the laboratory operation and the overall quality of the data.

4.2 Use of volumetric glassware for diluent addition

Some of the key steps in sample preparations involve the use of volumetric glassware.
The production process which created flasks with accuracies similar to those in use today has been in operation
for the last 75 years. It is astonishing to consider that sample preparation methods have not advanced signifi-
cantly during the last century. Especially when compared to the advances in instrumentation and software that
have provided dramatic improvements in analysis and data processing.

What are some of the errors that are associated with volumetric glassware?

4.2.1 Failure rates

A paper published by Coleman and Harris from NIST in 2005 (11) states the failure rates of new glassware to
meet Class A specifications have been found to be as high as 50%. This finding may not be too surprising since
there are a variety of glassware suppliers, some of them having very cut rate pricing. This simply illustrates
the importance of proper evaluation of glassware before selecting a vendor, and the conclusion that a supplier
should not be chosen on price alone.

4.2.2 Calibration temperature

A volumetric flask is typically calibrated for use at 20 °C. Any temperature change of the solution which results in
a volume change can cause errors if the working temperature is significantly away from the volumetric calibra-
tion point of 20 °C.
These temperature deviations may be caused by endothermic or exothermic mixing of solvents. In addition, a
sonicator which is often used to aid dissolution of solids, can cause a significant increase in the temperature of
the solutions. If care is not taken to return the flask and contents back to 20 °C, errors can be introduced.
Additional information regarding the operating temperature ranges for Class A glassware can be found in a UKAS
publication on traceability of volumetric apparatus (12).

4.2.3 Contamination risks

A significant problem with reusing any item in the laboratory such as volumetric glassware is contamination
from other products or reagents. It is very difficult, if not impossible, to qualify a glassware washing system in
development due to the number and variability of the products tested. Therefore it is very important to always
pre-rinse and post rinse the glassware with the appropriate solvents to minimize this. Unfortunately, this leads to
solvent waste and is time consuming.
There have been a number of OOS investigations where the analyst has forgotten to pre-rinse his volumetric
flasks. In fact, this implies that another analyst must have forgotten to perform a post rinsing operation too.

8 METTLER TOLEDO White Paper

 The problem created by this repeating issue is how to justify a CAPA that says retraining is addressing the prob-
lem when in fact is doesn’t. How do you know which analyst didn’t do the post-rinsing, do you retrain all ana-
lysts? If people continue to forget, what are the next steps? Do you spend hundreds of thousands of dollars on a
system to try and remove the contaminants from the glassware? Some companies have. But does it really make
sense to spend that much money on 100 year old technology?

4.2.4 Revalidation
Coleman and Harris also suggested in their paper (11) that the calibration of the glassware should be verified at
least every 10 years. This could potentially be a very expensive process considering that the number of volumet-
ric flasks in a typical pharmaceutical analytical or QA/QC department can be very large. It could potentially be
cheaper to throw them all away and start again.

4.2.5 Tolerances
In Table 1, the published NIST relative percent errors associated with each size of volumetric glassware are listed.
In each case as the size (volume) of the glassware decreases, the error risk (tolerance) increases significantly.

Pipettes Flasks
Volume (mL) Relative % error Volume (mL) Relative % error
1 0.60 5 0.40
2 0.30 10 0.20
3 0.33 25 0.12
4 0.25 50 0.10
5 0.20 100 0.08
10 0.20 200 0.05

Table 1: Relative percent errors for Class A glassware.

Any errors associated with the Class A glassware that does not meet the specification would be even larger, as
mentioned previously in 4.2.1 concerning the high failure rates.

Aside from the significant increase in the relative percent error, the smaller glassware is also very technique
dependent when it comes to manual manipulations. For example, in one company, it was found that many of the
analysts could not properly use a pipette below 2 mL and errors as high as 5% were noted in some cases (13).

4.2.6 Cost
The cost of using volumetric glassware is an issue that is often taken for granted. A great deal of effort goes
into keeping glassware organized and stored in a laboratory. Everyone who has worked in the lab has probably
been charged with ordering and putting away the clean glassware at some point in their career. This is costing
a proportion of a FTE. The pre and post rinsing on a company wide basis, assuming a very conservative 25 mL
use per flask and 10,000 sample preparations, might be costing the company $10,000 or more each year at a
$40/liter average solvent cost. These annually recurring costs can add up to a substantial amount, especially
when you include the rising cost of waste disposal. Additional costs can be incurred by lab services groups that
transport the flasks to and from the washing facility; and attrition due to breakage and damage, which results in
approx. 10% loss each year at a cost of about $20 per flask.

4.3 Mixing

Most samples are sonicated to expedite the breakup of tablets, capsules, or powders. Sonication can cause OOS
results when there is a lack of robustness in the method. The lack of robustness arises from the improper use
of the sonicator and whether or not the instrument is tuned properly. Most sonicators have the following instruc-
tions on them:

White Paper METTLER TOLEDO 9

 • Do not place parts or containers directly on the bottom of the cleaning tank: use a try or wire to suspend items.
White Paper
• Do not allow the solution to drop more that 3/8 inch below the operating level line with the cleaner on.

However, from industry feedback it seems that few people follow those instructions. The pictures in Figure 6
­illustrate the difference between a tuned sonicator on the left and an untuned sonicator on the right.

Figure 6: Foils from a tuned and untuned sonicator.

The untuned system has most of its energy focused in the middle of the bath, where you can see the large hole
in the foil (on the right). Therefore, the energy of the system can vary significantly depending on the placement of
the sample into the bath.

Many methods need to have better instructions for the final mixing step. Most methods only state to mix well
without realizing that a volumetric flask is an extremely poor mixing vessel that requires it to be inverted a num-
ber of times to ensure proper mixing.

4.4 Labeling

Labeling can potentially cause OOS results due to label mix ups but the most significant issues with labeling
are usually identified at safety and GxP audit times. Regardless of what a labeling SOP in the company states,
when flasks in laboratories are examined, the labeling content usually ranges from the absolute minimum to
the very detailed. Of course, all of these permanent marker labels must be removed before sending them out to
be washed and that necessitates the use of methanol or acetone to wipe down the flasks which takes time and
wastes more solvent.

5 Gravimetric sample preparation

It is universally accepted that a gravimetric measurement is intrinsically more accurate that a volumetric mea-
surement. In fact, pipettes and volumetric measuring equipment are calibrated using gravimetric methods. So
why are people still weighing solids and powders on a weighing paper, transferring them into volumetric flasks,
and subjectively reading the meniscus to prepare a specific concentration?

Gravimetric sample preparation, as described in USP General Chapter 1251, involves automated weighing and
dispensing of the solid and of the diluent and can reduce laboratory errors and increase laboratory efficiency (14).
Using this method, weighing papers, weighing boats and volumetric flasks are no longer necessary.

10 METTLER TOLEDO White Paper

 5.1 Gravimetric sample preparation equipment

There are a number of systems available which support the technique of gravimetric sample preparation.
These are described in more detail in this section and listed in Table 2.
In all of the systems, a gravimetric approach is used to deliver both powders and liquids into a small, dispos-
able target container (vial) which is positioned on an analytical or a micro-analytical balance (see Figure 7).

Figure 7: Gravimetric dispensing system: XPR Automatic Balance.

The user has to define the concentration required and the target amount of solution to be prepared. The software
calculates the target amount of solid to dispense and then, according to the actual amount of solid dispensed,
delivers the appropriate amount of diluent to prepare an extremely accurate and precise concentration. Serial
dilutions can be reached efficiently, without multiple dilution series, as the XPR Automatic Balance can achieve a
dilution factor of even 10,000 in a single step.

Individual dosing heads are used for each substance. The powder dosing heads are disposable and designed
for use with a single substance only. The solvent dispensing heads are also exclusively used for a single solvent,
and no valves or switching or washing of lines is employed. This means that all potential sources of cross con-
tamination risk are eliminated.

The dosing enables automated powder and liquid dispensing processes. The appropriate powder dispensing
head has to be selected and manually inserted into the system. Each dosing head is identified by RFID chip for
process security. This head has to be manually removed and replaced by the appropriate liquid dosing head at
the “Add diluent“ step. All data is recorded electronically and can be automatically printed onto labels.

White Paper METTLER TOLEDO 11

 Sample preparation Volumetric Gravimetric
White Paper
Configuration • XPR205 Analytical • XPR205 (Automatic • XPR205 (Automatic • XPR226DRQ (Automatic
Balance Balance) Balance) Balance)
• Volumetric flask • Automated liquid • Automated powder • Automated powder and
dispensing and liquid dispensing liquid dispensing
Sample dispensed Manual Manual Automated or manual* Automated or manual*
(with spatula) (with spatula)
Calculation of amount Manual Automatic (based on Automatic (based on Automatic (based on
of diluent required actual sample weight) actual sample weight) actual sample weight)
Diluent dispensed Manual (with pipette) Automated Automated Automated
Diluent units mL g g g
Concentration units mg/mL mg/g mg/g mg/g
USP Minimum weight** 14 mg 14 mg 10 mg (automated) 7 mg (automated)
(0.10%, k = 2, 5% load) 14 mg (manual) 10 mg (manual)
Minimum weight** 1.4 mg 1.4 mg 1.0 mg (automated) 0.7 mg (automated)
(U=1.0%, k = 2, 5% load) 1.4 mg (manual) 1.0 mg (manual)

* = Automated powder dispensing for free-flowing powders. Manual dispensing for sticky powders, pastes, gels, liquid samples, etc.
** = typical value

Table 2: Gravimetric dispensing system specifications.

These automated gravimetric systems are being adopted by analytical and QA/QC laboratories in the pharma-
ceutical industry.

5.2 Effect on diluent addition

We have already discussed how the addition of the diluent by volumetric measurement introduces a manifold of
indeterminate handling errors, such as reading the meniscus incorrectly or using the glassware at temperatures
where thermal expansion causes the limit of error to be exceeded. Gravimetric liquid dispensing avoids these
non-quantifiable handling errors, and furthermore weighing liquids at gram levels is very accurate because it
results in a completely negligible measurement uncertainty contribution of this process step. The amount of
diluent is typically far above the minimum weight of the balance, where the hyperbolic shape of the relative mea-
surement uncertainty curve flattens out to almost zero (see Figure 3).

With gravimetric sample preparation the exact amount of substance dispensed (whether dispensed manually by
spatula or using an automated dosing head) is recorded and used to accurately calculate the amount of solvent
to weigh in to the container. Any under or overshoot in powder weighing doesn’t require you to waste time add-
ing a tiny amount more or scooping material off the weighing paper with your spatula. The automated liquid
dispensing compensates for this and delivers the correct amount of diluent to achieve the required concentration.
The sample can then be sonicated and used without the need to be concerned about temperature and mixing.

5.3 Effect on process security (data management)

In terms of data management there are also distinct advantages of automated gravimetric sample preparation
in comparison with the manual volumetric approach. The manual approach requires hand transcription which
has a high error-risk, and it relies on the diligence of each individual analyst. It is simply not possible to digitally
record which size of volumetric flask was used automatically.
With an automated approach the data transcription is automated. All samples and solvents are identified by RFID
(radio frequency identification) to eliminate the possibility of weighing the wrong sample. All weighed samples
are documented electronically (target weights, actual weights and concentrations achieved) and the data is fully
traceable.

12 METTLER TOLEDO White Paper

 Labels with pre-defined fields can be printed automatically for immediate application to the vial containing the
prepared solution. This addresses the issues with accuracy and consistency of labeling which were discussed in
the context of the manual approach.

5.4 Effect on process efficiency

The simplified gravimetric sample preparation workflow is shown in Figure 8. This workflow has approximately
30% fewer steps than the volumetric workflow, described in Figure 3, which is a significant reduction. This
makes the process much more efficient and results in a significant amount of time saved in the sample prepara-
tion workflow. More importantly, the steps that have greatest potential to cause OOS results have been eliminated.

Material and equipment

• Gathering materials – standards, reagents and samples

• Check balance cleanliness, minimum weight requirements, calibration date
• Check other instruments for calibration / qualification dates and record
• Prepare diluents – combining solvents in correct proportions, pH adjusting, etc.
• Labeling handled by system using RFID technology

Weighing and labeling

• Place powder dosing head on instrument, deliver target amount

• Place solvent dosing head on instrument, deliver target amount
• Information and labels automatically recorded

Sonicating and filling to the line

• Sonicate, no filling to the line required

• Successive dilution, if required, using solvent dosing head
• Information and labels handled automatically
• Transfer to vials and label vials
• Repeat steps for each preparation

Analyze and clean up

• Place vials in instrument and perform analysis

• Dispose of containers

Figure 8: New simplified sample prep workflow using a gravimetric approach.

5.5 How can a gravimetric approach reduce the sample volume?

Let us now look at the typical USP minimum weight and the recommended safety factor for automated gravimet-
ric dispensing systems as compared to manual weighing systems. Provided the same weighing module is used
in both instruments, generally the minimum weight for the automated system is significantly lower as compared
to the equivalent conventional weighing system. One main reason is that environmental effects – especially
drafts and temperature differences between balance and sample – are more efficiently prevented when using
automated dispensing. Furthermore, the variability introduced by the operator is completely removed.

White Paper METTLER TOLEDO 13

 The exclusion of the operator variability and the efficient compensation of environmental effects allows for apply-
White Paper
ing a smaller safety factor for automated weighing systems, typically 1.5 instead of 2 within a controlled labo-
ratory environment with trained operators. Consequently, smaller sample sizes can be chosen for automated
weighing. For example 50 mg is weighed manually on a XPR analytical balance, while the same concentration
can by achieved by weighing as little as 5 mg on the XPR Automatic Balance.
Besides of the smaller minimum weight and the smaller safety factor, an additional benefit of gravimetric sample
preparation is that the analyst is not constrained to make a volume based on the size of volumetric flask avail-
able. These two factors combined mean that smaller amounts of sample can be used, smaller volumes of solu-
tions can be prepared, less solvent is consumed and there is less waste to dispose of. The automated nature of
the process also makes it safer for the analyst.

XPR205 XPR226DRQ
analytical balance Automatic Balance
(Minimum weight (Minimum weight
× safety factor) × safety factor)
= 14 mg × 2 = 28 mg = 7 mg × 1.5 = 10.5 mg
Solvent volume

5.3 ×
substance
& solvent
& waste

Concentration prepared

Figure 9: Comparison of volumetric and gravimetric sample preparation on the amount of solvent used.

Being constrained to Class A volumetric glassware forces an analyst to use much more substance than neces-
sary because they are limited to the capacity of the flasks available. The amount of solvent used when preparing
samples using a manual volumetric approach is indicated by the red line in Figure 8. The sharp vertical drops in
the red line on the graph indicate the four discrete points at which the minimum weight of the substance matches
an available volumetric flask size (10 mL, 25 mL, 50 mL, 100 mL) and thus the only concentrations at which
the actual minimum weight can be used to achieve the desired concentration. In all other cases, the amount of
sample must be rounded up to match the next size of volumetric flask available.

Automated gravimetric sample preparation is not limited to these discrete intervals, as the smooth green curve
indicates. With this method, the minimum amount of sample (10.5 mg) can be weighed at every point on the
curve, and the corresponding amount of solvent can be added, to achieve the required concentration.
At the point indicated by the blue arrow (which corresponds to a concentration of 1.1 mg/mL), over five times
less substance and solvent is used to achieve the same concentration gravimetrically. This is due to an equal
contribution from both of the factors mentioned previously: the minimum amount of sample that can be weighed
on the XPR Automatic Balance is just over 2.5 times smaller; also, the total sample volume has been rounded up
to the next available size of volumetric flask (50 mL).

For example, Table 3 illustrates the preparation of three concentrations of solution manually (volumetrically)
compared to automatically (gravimetrically) using either (a) an XPR205 balance and an 50 mL or 100 mL volu-
metric flask; (b) a XPR205 Automatic Balance for automated sample preparation; (c) a XPR226DRQ Automatic
Balance for automated sample preparation. The amounts of substance and solvent consumed for each method
are listing along with the savings.

14 METTLER TOLEDO White Paper

 Sample preparation Manual volumetric Automated gravimetric
Configuration • XPR205 analytical • XPR205 (Automatic • XPR205 (Automatic • XPR226DRQ
balance Balance) Balance) (Automatic Balance)
• Volumetric flask • Automated liquid • Automated powder and • Automated powder and
dispensing liquid dispensing liquid dispensing
USP Minimum weight* 14 mg 14 mg 10 mg (automated) 7 mg (automated)
(0.10%, k = 2, 5% load)
Recommended safety 2 2 1.5 1.5
factor **
Lowest sample weighed out 28 mg 28 mg 15 mg 10.5 mg
Concentration prepared 0.5 1.0 0.5 1.0 0.5 1.0 0.5 1.0
examples (mg/mL or mg/g)
Amount of solid dispensed 50 50 28 28 15 15 10.5 10.5
(mg)
Amount of diluent required 100 50 56 28 30 15 21 10.5
(mL or g)
Substance savings compared to manual method: 44% 70% 79%

* = typical value
** = The safety factor quoted is that recommended for stable environments and trained operators. For unstable environmental conditions
or insufficiently trained operators higher safety factors should be used.

Table 3: Substance saving based on automated sample preparation methods.

So, if a 1.0 mg/mL solution is prepared manually in a volumetric flask, 50 mL of solvent and 50 mg of
­substance are consumed. When the solution is prepared using an automated gravimetric method with an
XPE206DR-PL system, 10.5 mL of solvent and 10.5 mg of substance are sufficient (assuming a solvent density
of 1 g/mL).
A saving of 79% in both substance and solvent can be realized whilst remaining compliant with USP General
Chapter <41>.

5.6 Volumetric vs. gravimetric comparison

To directly compare the manual volumetric and the automated gravimetric methods, let’s look at a simple prepa-
ration comparing the two techniques. If the method requires a 0.25 mg/mL concentration then using a volumet-
ric system, one would use a 200 mL volumetric flask and weigh out 50 mg of material. Table 4 shows some of
the determinate and indeterminate errors that may be found in this simple procedure.

Step Volumetric Gravimetric

Determinate errors Indeterminate errors Determinate errors Indeterminate errors
200 mL container 0.05% Uncalibrated N/A – volumetric flasks are not required
Weigh 50 mg sample – 0.1% balance – 0.1% balance
Others are accounted for Others are accounted for
using a safety factor of 2 using a safety factor of
or higher 1.5, if automated
(in a controlled lab envi-
ronment)
Sample transfer Re-weighing container Powder transfer N/A – transfer of sample to volumetric flask
is not required
Fill to mark – Reading meniscus and N/A – no subjective reading of meniscus required
temperature effects with gravimetric method

Table 4: Comparing errors in volumetric and gravimetric processes.

As you can see in Table 4, with gravimetric sample preparation the number of determinate errors is reduced and
the indeterminate errors which tend to be much larger than the determinate ones are essentially eliminated or
accounted for.

White Paper METTLER TOLEDO 15

 5.6.1 Example of a standard preparation via automated vs. manual method
White Paper
The data in Table 5 describes an experiment carried out by a pharmaceutical company who were evaluating
gravimetric standard preparation for an SOP. Using the automated system they wanted to make 16.5 g total
amount of standard at a concentration of 0.603 mg/g. Nine replicates of minimum 10 mg of API were dispensed
into 20 mL brown glass vials. Following the solid dispense the solvent (an 80:20 acetonitrile: water mixture)
was added gravimetrically to achieve the target concentration based on the exact weight of API dispensed. The
suspension was dissolved and then a 2 µL sample was injected into the HPLC. In addition, injector precision and
variability in the analytical instrumentation was determined by comparison of 10 repeat injections of the same
solution – solution 1.

Solution Area Injector

Dose API Solvent added ­concentration ­correlated to Accuracy
# Solution (mg) (g) (mg/g) Area 0.6 mg/g Solution 1 Area
1 10.105 16.7481 0.60299 2596.8833 2584.00634 Inj. 1 2593.06665
2 10.320 17.1048 0.60298 2595.35474 2582.52818 Inj. 2 2604.43384
3 10.140 16.8063 0.60298 2597.53027 2584.69296 Inj. 3 2604.12378
4 10.125 16.7815 0.60298 2595.5564 2582.72885 Inj. 4 2604.89429
5 10.250 16.9885 0.60299 2597.12427 2584.24611 Inj. 5 2607.99683
6 10.200 16.9058 0.60298 2596.04517 2583.2152 Inj. 6 2605.37231
7 10.130 16.7895 0.60299 2602.42212 2589.51769 Inj. 7 2602.29541
8 10.040 16.6408 0.60297 2609.64868 2596.79455 Inj. 8 2608.33862
9 10.275 17.0297 0.60299 2604.74341 2591.82747 Inj. 9 2593.23218
– – – – – – Inj. 10 2601.73145
Mean 10.176 16.866 0.603 2599.479 2586.617 Mean 2602.549
σ 0.09092503 0.150659114 7.07107E-06 5.000822667 4.983629696 σ 5.37642343
RSD (%) 0.894 0.893 0.001 0.19 0.19 RSD (%) 0.21

Table 5: Automated gravimetric sample preparation data.

For the nine individual samples, the weight of solid dispensed is shown in column 2. The weight of solvent dis-
pensed is shown in column 3 and the concentrations achieved are shown in column 4. Accurate solvent weigh-
ing compensates for any variation in the amount of solid dispensed (which is already within a narrow range –
10.040 and 10.320 mg), to achieve an accurate concentration of 0.603 mg/g in every case. The repeatability of
the concentrations is excellent, as indicated by the low standard deviation, RSD = 0.001% (pale blue highlighted
cell). When the peak area is correlated after analysis, the RSD across these nine solutions is 0.19% (dark blue
highlighted cell).

The two columns at the right hand side of Table 5 show ten injections of the same solution, with the aim of
eliminating the contribution of the analytical system to the overall variability. The RSD of these samples is 0.21%
(green highlighted cell). So, in conclusion, automated standard preparation with solvent weighing shows no sig-
nificant variability in the solution concentrations, that can’t be attributed to the analytical instrumentation itself.

When the same experiment was repeated by preparing the individual standards manually, the results in table
6 were obtained. The RSD on the correlated areas was 0.57% (compared to 0.19% for automated). This IS
significant compared to the accuracy of the analytical instrumentation, and means that the variability can be
almost only attributed to the manual preparation process. As the weighing process and the HPLC analysis are
independent from each other, the associated %RSD are not added arithmetically but quadratically in order to
determine the %RSD of the peak area. As the %RSD of the peak area is 0.57% (sample weighed manually),
and the %RSD for the overall variability of the analytical instrumentation is 0.21%, the calculated %RSD for the
manual sample preparation is 0.54%, which fits nicely to the experimentally determined %RSD of 0.60% of
the manual weighing process. In other words, for this experiment the limiting factor for the overall accuracy of
the HPLC analysis was the manual sample preparation, whereas for automated sample preparation the limiting
factor for the overall accuracy is the HPLC itself.

16 METTLER TOLEDO White Paper

Manual dose
# Blend dosed Flask size Conc. achieved
mg mL mg/mL
1 12.920 20 0.646
2 12.960 20 0.648
3 12.860 20 0.643
4 12.840 20 0.642
5 13.060 20 0.653
6 12.980 20 0.649
7 12.860 20 0.643
8 12.800 20 0.640
9 12.860 20 0.643
10 12.920 20 0.646
Mean 12.906 20 0.645
σ 0.078 0.0 0.004
% RSD 0.60% 0.0 0.57%

Table 6: Manual (volumetric) sample preparation.

5.6.2 Summary of volumetric vs. gravimetric considerations

There are several key elements to consider in the comparison between manual volumetric and automated
gravimetric methods of sample preparation. These elements are summarized in Table 7:

Manual volumetric Automated gravimetric

Substance saving
Diluent addition accuracy
Process and data security
Uses far more substance than necessary due to:
• Higher minimum weight
• Higher safety factor
• Scaling up to appropriate size of volumetric
flask
Manual volumetric method has many potential
error risks:
• Meniscus reading is subjective
• Temperature deviation of contents compared to
calibrated temperature
• Cross-contamination risk
• Incorrect pour / drain times
• Dishwasher damage
• Up to 50% failure rates to comply with Class A
specifications
• Recalibration recommended every 10 years
• Requires hand transcription
• Relies on diligence of analyst
• Smaller minimum weight = smaller sample
sizes
• Dispense directly into target vessel and avoid
loss on transfer
• No sample spillage or overshooting target
weight
• Eliminates need to prepare duplicate samples
due to OOS weighing errors
Automated gravimetric method is highly accurate:
• Amount of liquid needed is calculated automati-
cally based on the actual weight of solid which:
• Eliminates manual calculation errors
• Eliminates human error in choice of pipette
• Eliminates human error in choice of flask
• The amount of solvent dispensed gravi-
metrically is far above the minimum weight
therefore measurement uncertainty of this
step is negligible
• Automated data transcription
• Labels/documentation generated automatically
• Integrated RFID chip stores and tracks sub-
stance information and eliminates risk of
sample confusion
• Data is electronically recorded and fully trace-
able

White Paper METTLER TOLEDO 17

 Manual volumetric Automated gravimetric
White Paper
Potential for cross- Volumetric flasks are re-used. They have to be • Disposable dosing head is used for a single
contamination rinsed before and after each use. substance only which eliminates cross con-
tamination
• The dosing head is stored in a transport con-
tainer which protects it from the atmosphere
• Each solvent has a unique liquid dosing head.
There are no valves or washing of lines is nec-
essary
• Disposable vial used to prepare samples
User safety Manual weighing of solids and sample prepara- Automated dispensing reduces user exposure:
tion involves close contact for user. • The powder is contained in the d­ osing head
• The dosing head is a closed system
• Only the required amount of powder is released
from the dosing head
• An optional safety enclosure is available

Table 7: Comparison of volumetric vs. gravimetric methods.

6 Conclusion
The most important measure to guarantee accurate weighing – and consequently to avoid the possibility of OOS
due to weighing – is the determination of the minimum weight of the balances. Consequently, it is important
to always weigh above the minimum weight in order to comply with the respective accuracy requirements. For
automated dispensing systems, the minimum weight is significantly smaller as compared to manual weighing.
It is good practice to apply a safety factor in order to compensate the variability of the minimum weight due to
different operators and changing environmental conditions, however, the safety factor can be chosen significantly
smaller for automated weighing systems as environmental effects are reduced and the variability introduced by
the operator is completely removed.

Reducing the occurrence of OOS results in the laboratory requires close attention to the details of where errors
can occur, a critical evaluation of the overall process workflow, and a concerted effort to change those practices
that lead to OOS results or errors in the data. New technologies must be brought into the laboratory to finally
improve the data quality that is being generated by laboratories around the world. In addition, most companies
want and need to achieve higher productivity with the same or less resources. This efficiency cannot occur with-
out a fundamental change in the way sample preparation is currently performed which has had little improve-
ment for the best part of a century and still accounts for more than 60% of our time spent in the laboratory.

Gravimetric sample preparation, which is defined as automated weighing and dispensing of both the solid and
the solvent, is based on the weighing standard GWP®. Furthermore, it has several important benefits in a mod-
ern laboratory: it improves user safety due to reduced exposure risk; it improves process safety via electronic
recording and tracking of data; and it can save substance and solvent due to reduced minimum weight and
elimination of volumetric flasks.

However, the most important benefit of gravimetric sample preparation is its innovative ability to enhance the
accuracy and drastically reduce the variability in sample processing steps, which is the major source of out-
of-specification results. Some areas of high error risk in the process, such as those involving use of volumetric
flasks, are eliminated completely. Gravimetric sample preparation has the effect of reducing laboratory errors and
increasing laboratory efficiency.

18 METTLER TOLEDO White Paper

 7 References
1. United States of America v. Barr Laboratories, Inc., 812 F Supp 458 (DNJ 1993).
2. U.S. Food and Drug Administration. Guidance for Industry: Investigating out-of-specification (OOS) test
results for pharmaceutical production. FDA. Available at: www.fda.gov/downloads/Drugs/GuidanceCompli-
anceRegulatoryInformation/Guidances/UCM070287.pdf.
3. Majors, R.E. LC / GC Magazine, 1991, 1997, 2002
4. General Chapter <41> “Weights and Balances”, US Pharmacopeia USP34 – NF29, Rockville, Maryland,
2013, Online-Edition
5. GWP® - Good Weighing Practice™ – A Risk-Based Approach to Select and Test Weighing Instruments,
White Paper, Mettler-Toledo AG, Greifensee, Switzerland, Nov. 2013.
6. Reichmuth A., Fritsch K., Good Weighing Practices in the Pharmaceutical Industry – Risk-Based Qualifica-
tion and Life Cycle Management of Weighing Systems, Pharmaceutical Engineering, Volume 29, Number 6,
ISPE, Tampa FL, USA, 2009.
7. Fritsch K., Quenot J.-L., Good Weighing Practices – Avoid OOS Results with Proper Weighing, Pharmaceuti-
cal Formulation & Quality, Volume 14, Number 1, Wiley-Blackwell, Hoboken, NJ, USA, 2012.
8. Evaluation of Measurement Data – Guide to the Expression of Uncertainty in Measurement (GUM), JCGM
100:2008, Bureau International des Poids et Mesures, Sèvres, France, 2008. Available at www.bipm.org.
9. Guidelines on the Calibration of Non-Automatic Weighing Systems, EURAMET cg-18, Version 3.0, European
Association of National Metrology Institutes, Braunschweig, Germany, 2011. Available at www.euramet.org.
10. General Chapter <1251> "Weighing on an Analytical Balance", Second Supplement to USP36-NF31,
June 2013, Rockville, Maryland, Online-Edition.
11. “Important Technical Guidance on Glassware”, Tom Coleman and Georgia Harris, NIST, Aug. 2005.
12. "Traceability: Volumetric apparatus", LAB15 Guidance, Edition 2, UKAS, 2009. Available at:
http://www.ukas.com/library/Technical-Information/Pubs-Technical-Articles/Pubs-List/LAB15.pdf
13. Dr Charles Ray, former Associate Director of Analytical R&D at Bristol-Myers Squibb Co., personal
experience from managing Analytical R&D teams in leading pharma companies.
14. Fritsch, K., Ratcliff, J., Ray Ch., Reducing Variability and Out-of-Specification Results by Implementing High
Quality Gravimetric Sample Preparation (GSP), Pharmaceutical Engineering, Volume 32, Number 1, ISPE,
Tampa FL, USA, 2012.

White Paper METTLER TOLEDO 19

 www.mt.com/xpr-automatic
For more information

METTLER TOLEDO Group

Laboratory Weighing
Local contact: www.mt.com/contacts

Subject to technical changes

© 03/2021 METTLER TOLEDO. All rights reserved
30086085C
Group MarCom RITM697833 MC / MA