Analytical Principals and Methods

- A Clinical Chemists View

Performance Characteristics of
Biomedical Analyses
Reliability Criteria

a. Sensitivity – can it measure low enough levels ?

b. Specificity – can it distinguish the analyte from other

substances ?

c. Bias (Accuracy) – does it give the correct value ?

d. Precision – does it always give the same answer ?

 Precision and Bias

High accuracy Poor precision, Poor accuracy,

and precision high accuracy high precision
═ Bias
Performance Characteristics of
Biochemical Analyses
Practicability Criteria
a. Speed – turn around time

b. Cost

c. Analytical Skill

d. Dependability – equipment reliability

e. Safety
Methodological Aims
• Analytical Accuracy
Measure of the agreement between a measured quantity
and the true value

• Analytical Imprecision
Measure of the agreement between replicates

These will vary depending on the analyte, methodology etc.

Methodological Aims (Cont.)

• The analytical aims will vary according to the clinical

need.

• For following the course of a disease or monitoring

therapy, imprecision (reproducibility) is important.

• For diagnosis, accuracy is important as the measured

value is compared with reference data.
Methodological Aims (Cont.)
• In screening for a disease, accuracy and imprecision
must be acceptable, if not :

- May generate further unnecessary tests

- May produce false positives
- May produce false negatives
- May cause analyses to be repeated
Methodological Aims (Cont.)

• Internal and external quality control mechanisms help

assess accuracy and imprecision on a batch to batch,
day to day, month to month basis.

• QC samples containing known amounts of analyte must

be routinely analysed and results reported.
Quality Control Material

• Similar matrix to patients samples

• Have a reasonable degree of stability

• Available in a range of analyte concentrations

• Exhibit minimal between-sample variability

• May have assigned or approximate values

Internal QC
• 3 analytical ranges: H,M,L (within analytical range)

• Introduce samples into run on x number of occasions e.g. every

20 samples

• Record data → calculate mean+/- S.D.

• Compare with assigned value or derived value

• Act on results e.g. accept or reject

Internal QC (Cont.)
• May also have 2 or more analysers performing the
same analyses, hence must be able to compare :

1. Patient samples
2. Internal QC data
3. External QA results
External QC

• Organised by various external bodies e.g. NEQAS

• Process is voluntary, but necessary for accreditation

• Regular dispatch of samples from these centres

• Analyses required and returns within a set time period

• Reports issues, grouped according to method / instrumentation

 External QC (Cont.)
• Indication of continuing performance given

• Warnings issued to consistently poor performers. If trend continues,

then advisory panel notified

Action limits
T+/- 3SDs

Measured
variable Target value

Warning limits
T +/- 2SDs
Time

Shewart Chart
Levey Jennings QC Chart
Westgard Rules
• 1 control exceeds = +/- 3 SDs

• 2 consecutive controls exceed = +/- 2 SDs

• 1 control exceeds the + 2 SD limit and a second control

exceeds the - 2 SD limit.

• 4 consecutive controls exceed +/- 1 SD

Reference Ranges

• Reference ranges
• established by measuring the concentration of a
particular analyte in a normal healthy population and
from the calculation of the mean value and S.D.
Defining the Reference Range

‘Normal’ Subjects

n
Mean

-2 SD 2SD

Test Value
Reference Ranges (Cont.)
• Usually, defined as the mean +/- 2SDs as shown previously.

• But this excludes 5% of the healthy population.

• 1 in 20 will have abnormal values.

• May require separate reference ranges for certain analytes

• age, sex, ethnic differences.

• A hospitalised population may have a different reference range !

Reference values
• Dependent on:
• Selection of subjects
• Assessment of the state of health
• Characteristics of population, age and sex
• Specimen collection and storage
• Analytical technique performance characteristics
• Data handling techniques
Definitions (IFCC)
• A reference individual is an individual selected
using defined criteria
• A reference population consists of all possible
reference individuals
• A reference value is the value obtained by
measurement of a particular quantity on a
reference individual
Factors affecting reference
values
• Endogenous factors
• age (Igs, phosphate)
• sex (oestradiol/testosterone)
• body mass
• circadian (eg cortisol / testost)
• menstrual (LH / FSH / Oest / Prog),
• seasonal (vitamin D),
• pregnancy (hCG, urea, creatinine)
Cortisol
1000

900

800

700

600

Cortisol
500
(mmol/L)
400

300

200

100

0
0 4 8 12 16 20 24

Time of Day (hours)

Factors affecting reference
values
• Preanalytical factors
• Food intake (fasting/nonfasting/trigs)
• alcohol intake
• posture (renin/aldosterone)
• Immobilization
• previous medical and surgical care
• stress (cortisol)
• exercise (growth hormone)
• drug administration
Factors affecting reference
values
• Laboratory factors
• Specimen collection
• transport
• storage
• e.g serum/plasma
• Analytical technique – magnitude of imprecision

• Genetic factors
• Ethnicity
• E.g. increased dyslipidaemia in Asians who settle in the
US
Reference ranges
• Two approaches:

• Select a small number of individuals. Those who pass

selection criteria have samples collected and analysed under
controlled conditions. A reference range is generated.

• Select a large number of people and collect samples from all

of them. After analysis the data is examined and exclusion
criteria applied.
Problems with reference ranges
• Reference values cover 95% of population

• 1 in 20 (5%) ‘normals’ will have a test result outside

the reference interval

• 2.5% individuals will have results lower than the

lower ref limit and 2.5% will have results higher than
the higher ref limit
Problems with reference ranges

• No of tests % chance of a result

outside ref range

• 1 5
• 2 10
• 3 14
• 4 19
• 5 23
• 10 40
Problems with reference ranges
• No of tests • No of tests results
outside ref. range

• 2-7 • 1
• 8-16 • 2
• 17-28 • 3
• 29-40 • 4
• 41-52 • 5
Problems with reference ranges
• Biological variation of individuals around their
homeostatic setting points

• Multiple results for an individual for a single test

may be -
• Always within the ref range
• Both within and outside the interval
• Always outside the interval
Is test value abnormal?
• Compare with reference range

• Does result differ significantly from previous result?

• Difference between 2 results on same patient is
unlikely (1/20) to be due to analytical variation if >2.7
x analytical SD.
• Significant differences could be due to biological
variation eg ALP
Other ranges which aid interpretation
• Action Limits
• cholesterol
• Paracetamol

• Therapeutic Ranges - for drugs

• lithium
• digoxin

Used to trigger therapeutic / investigative actions

 Action Limits for Cholesterol
Relative risk of CHD
500
7.8

400 6.5

300
Relative Ideal
Risk 5.2
200

100

0
0 2.5 5 7.5 10 12.5

Cholesterol (mmol/L)
 Action Limits for Paracetamol
1000

Severe Liver
Damage Likely

Liver
Damage
Possible -
Paracetamol Treatment
(mmol/L)
100
Too Early
Desirable
to Tell

No Treatment
Needed

10
0 4 8 12 16

Time since ingestion (hours)

 The Ideal Diagnostic Test
‘Normal’ Diseased

n
Test Value

[Analyte]
No false positives or negatives
Ideal Tests
• Rarely available in routine practice

• High sensitivity and specificity rarely coexist

• Increased sensitivity traded for decreased specificity

and vice versa
Specificity Vs Sensitivity
– the necessary compromise
‘Normal’ Diseased

n
Test Value

[Analyte]
False Negatives False Positives
Sensitivity
• Measure of the incidence of positive results in
patients with the disease
• TRUE POSITIVES

TP
TP + FN
Specificity
• Is a measure of the incidence of negative results
in people who do not have the disease
• TRUE NEGATIVES

TN
FP + TN
Positive Predictive Value
• Proportion of patients with a positive test who
are in fact correctly diagnosed.

True Positives
Total Positive Tests
Negative Predictive value.
• Proportion of patients with a negative test who
are correctly diagnosed.

True negative patients

Total negative tests
ROC curves
• Receiver operator curves.

• Plot of sensitivity against 1- specificity

• Best test is that in the upper left hand corner.
 ROC Plot

Sensitivity

1-Specificity
ROC Plot

True
Positives

True negatives
References
• Clinical Investigation and Statistics in Lab. Medicine
R.Jones & B.Payne ISBN 0902429213
• Interpretation of Clin Chem Lab Data. C.Fraser ISBN
0632015799
• Tietz 3rd Edition Chapter 13 pg 320-335
• www.aacc.org
• J. Automatic Chem. Vol 6 No 3 1984 122-141
• Minimum Acceptability Performance Evaluation of Clin
Chem Methods NCCLS Document EP10-P Vol 6 No 3