Method Validation (Week 3)

Question 1

Why we need method validation?

Answer 1

To show methods are fit for purpose
Its a requirement of accreditation to ISO 17025 and 15189
Provides confidence to customers that it can generate accurate and reliable data

Question 2

Difference between validation and verification?

Answer 2

Validation - carried out on new methods or on incorporate changes

Verification - undertaken by a lab thats adopting a method that has been previously validated, or when a change has been made

Question 3

Method Validation process

Answer 3

1) Define analytical requirement - what tests are you looking to develope
2) Develop/identify candidate method - look at other previous work
3) Plan validation experiments
4) Carry out experiments
5) use data to assess fitness for purpose
6) Analytical requirements met? Yes = Validation report, No = back to 2)

Question 4

Defining analytical requirements

Answer 4

Customer requirements
Typical Tests:
- Qualitative - identify something
- Quantitative - how much is there
- Limit - could be purities, need to identify limit
What will happen to results? - high standards
What decisions made based on results?

Question 5

Parameters to consider?

Question 6

Validation of a HPLC analysis (case scene 1)

Answer 6

• A laboratory is looking to become
  accredited for the analysis of caffeine,
  saccharin, aspartame and benzoic acid in
  fruit juices, squashes and other soft
  drinks by HPLC.
• A candidate method has already identified
  and developed to provide the appropriate
  chromatographic conditions.

Question 7

Specificity (selectivity)

Answer 7

How discriminating is method? - can your method see what you’re trying to test for
Effect of other components on the analysis of the component of interest - make sure nothing else is effecting results - other compounds
Determined by adding materials that may be encountered in samples - CRMs

Question 8

Sensitivity

Answer 8

How much response to an analyte changes when conc changes - peak area
Shown by slope - sensitive analyses have steep slope for graph
For sensitive methods - big changes

Question 9

Question?

Answer 9

Ready-to-drink samples are going to be analysed undiluted
except for the addition of an internal standard.
Dilute-to-taste samples are going to be diluted by five times,
including the addition of an internal standard.

What other sample preparation may be needed?
May need to filter drinks - remove solid bits - will clog hplc columns
Degas the drinks

Question 10

Limit of detection (LoD) or DL

Answer 10

The smallest amount of analyte that can be detected and
identified with an acceptable degree of uncertainty, but not
necessarily quantified.
We are looking for concentration which gives a response that is 3 × the response
for blank values.

Question 11

LoD - graph

Answer 11

Baseline noise - squiggly line
Red line - LoD - 3x bigger the baseline peak
Blue line - LOQ - 10x bigger peak

Question 12

Determining loD and LoQ

Answer 12

Analyse a series of blanks (10 x analyses).
Calculate standard deviation of response (spectroscopic
techniques) or mean noise level (chromatographic techniques).
Determine detection/quantitation limit response as below

Question 13

Determining LoD and LoQ - 2

Answer 13

To determine response of standard, analyse a standard with a
concentration that will give a response close to the calculated
detection/quantitation limit responses.
Calculate LoD and LoQ using these equations:

Question 14

Quantitative analysis - analytical range

Answer 14

Determines how much of compound we actually have

The specified calibration range for an analysis depends on the
intended use of the method.
Typically ±20% of the expected test concentration(s).
Major component: covers 80-120% of test concentration.
Impurity: typically reporting level to 120% of test concentration.

Question 15

Quantitative analysis – linear range

Answer 15

The ability of a method to 
obtain results which are 
directly proportional to 
concentration.
Determine with linear
regression for line of best fit.
Confirm linearity using 
residuals.

Calibration curve - straight line
R2 - closer to 1 is better

Question 16

Quantitative analysis – linear range 2

Answer 16

Different methods have 
different ranges over 
which they are linear.
ICP has a large dynamic 
range.
LC-MS has a very small 
dynamic range.

Slightly s shaped graph - linear range in the middle

Question 17

Quantitative analysis – working range

Answer 17

Linear range – has an 
acceptable level of 
uncertainty.
The working range 
typically goes from the 
LoQ to a little beyond the 
linear range, up to where 
the sensitivity decreases.

Working range starts just after LoQ

Question 18

Accuracy?

Answer 18

Accuracy is the closeness of agreement between the value found
for the analyte and the true or accepted value

Question 19

Trueness?

Answer 19

Trueness is the closeness of agreement between a large number
of analyte values and the true value or accepted value.

Question 20

Bias?

Answer 20

Bias is a measure of the total systematic error in a method.

Question 21

Suggest appropriate concentrations for each additive standard for
the preparation of five point calibration graphs? Question should be before accuracy trueness cards

Answer 21

S - 80, C - 100, B -160

20% added - should be 120% above specification
S - 96
C - 120
B - 192

4 dilutions
S -
C -
B -

Question 22

Determining truness and bias

Answer 22

Trueness/bias can be determined in two ways:
Repeat analysis of a certified reference material (CRM).
In general, at least 7 replicates should be analysed.
Determination of recoveries from spiked samples:
- Analyse samples before and after spiking (adding) a known amount of
standard.
- For a method with no bias, the difference in the results is the amount of standard that was added (100% recovery)

Question 23

To check how well the analysis worked on actual samples, orange
juice was spiked with known amounts of additive standards.
2 cm3 of each of the additive spiked solutions was added to a
50 cm3 volumetric flask, plus an internal standard, and the
solutions made up to the mark with orange juice.
The recovery is calculated using the formula:
Recovery =
measured concentration × 100
/ actual concentration

Answer in book

Question 24

Precision?

Answer 24

The closeness of agreement between a series of independent
measurements obtained from multiple sampling of the same
homogeneous sample under the prescribed conditions.
Precision reflects random errors which occur in the method.
Precision is calculated by determining the standard deviation of
test results from repeat measurements.

Question 25

Expressing precision

Answer 25

An absolute measure of
precision = standard deviation (s or s.d)

or

variance (s2)
A relative measure of precision =
coefficient of variation (CV)
or
relative standard deviation (RSD)

Question 26

Precision

Answer 26

Repeatability – same analyst & equipment; limited time span.
Assess by: minimum of 6 replicates at 100% of test concentration, or 3
replicates at each of 3 concentrations (9 replicates in all).

Intermediate Precision – within-laboratory variations: different
days, analysts and equipment.

Reproducibility – different laboratories, analysts, samples and
equipment; same method; extended time span.

Question 27

The repeatability (r) of the method was checked by six injections
of appropriate additive standards. The concentrations measured
will be given in the next slide.
Determine the repeatability using the formula:
r = 3 x s
(where s is the standard deviation.)

Question 28

Uncertainty of measurement?

Answer 28

Uncertainty of measurement is a quantitative measure of the
limits within which the true value of the analyte is expected to
lie, to a specified level of confidence.
Values should typical lie within 2 × standard deviations of the
true value, i.e. 95% confidence level.
ISO 17025 and ISO 15189 requires that uncertainty is stated
with test results.

Uncertainty not only considers repeatability and reproducibility,
it takes into account many factors that could affect the result,
such as systematic effects which may cause bias.

Question 29

Uncertainty of measurement

Answer 29

Takes into account all random and systematic errors.
e.g. a glass pipette:
- The pipette will have a tolerance, i.e. ± 0.03 ml – this is a systematic error (it
will always be the same).
- Every time the operator uses the pipette, the position of the meniscus on the
line will vary slightly – random error.
- Therefore, there will be small variations in each volume of liquid dispensed.
The combination of these two errors will lead to the overall uncertainty in the
volume measured.