Statistical analysis

Question 1

Technical/analytical evaluations to be performed when a new test is to be introduced to the routine lab

Answer 1

1. Estimation of assay imprecision
2. Comparison of new vs old assay
3. Assessment of allowable measurement range - estimation of upper and lower limits of quantification
4. Evaluation of interferences and carry over if applicable
5. Assess whether it meets lab accreditation requirements and/or other regulatory standards.

Question 2

Define CVa

Answer 2

Analytical coefficient of variation

Question 3

Define CVg

Answer 3

Between subject coefficient of variation

Question 4

Define CVi

Answer 4

Within-subject coefficient of variation

Question 5

Define CVrb

Answer 5

Sample-related random bias coefficient of variation

Question 6

Two types of biological variation?

Answer 6

Within subject and between subject

Question 7

What is within-subject variation?

Answer 7

Random fluctuations around a homeostatic point

Question 8

How is biological variation determined?

Answer 8

Using a biological variation study

Question 9

How is a biological variation study performed?

Answer 9

Recruit study subjects
Take serial samples at intervals consistent with the intended use of the tests
Analyse the samples in duplicate
Perform an Analysis of Variance (ANOVA)

Question 10

Variations that affect the test results of an individual person over time?

Answer 10

1. Pre-analytical variation (CVp)
2. Analytical variation (CVa)
3. Within subject variation (CVi)

Question 10

Variations that affect the test results of an individual person over time?

Answer 10

1. Pre-analytical variation (CVp)
2. Analytical variation (CVa)
3. Within subject variation (CVi)

Question 11

What is pre-analytical variation?

Answer 11

Variation due to differences in patient preparation between tests, sample collection and conditions that the collected sample experiences after collection but prior to measurement itself

Question 12

What is reference change value?

Answer 12

Helps us understand if the change in a patient’s serial result is statisically significant. Determined from Z-score, CVa and CVi. Assumes CVp negligible.

Question 13

Formula for RCV

Answer 13

Question 14

Formula for index of individuality

Answer 14

CVi/CVg

Question 15

What is the index of individuality used for?

Answer 15

Determining if population based reference intervals are useful for a patient. If the II is <0.6, population based reference intervals are not useful (within subject variation is very low and a change within the reference interval may still be significant). If the II is >1.4 then population based reference intervals are helpful.

Question 16

The index of individuality for troponin I is 0.29. What is the implication of this for interpreting troponin I results?

Answer 16

The low index of individuality means that within an individual, troponin levels do not show much fluctuation. The utility of population based reference intervals for determining the significance of a troponin I value is low. Instead, serial measurements will be of importance to determine if a change is significant or not.

Question 17

How is total method error calculated?

Answer 17

From bias and standard deviation established by method comparison experiments. TME = Bias + 1.65SD

Question 18

How are total allowable error goals determined

Answer 18

Legal criteria
Targets set by providers of proficiency testing/EQA schemes
Using biological variation data

Question 19

How is CV calculated?

Answer 19

CV% = SD/Mean x 100%

Question 20

Biological variation database provided by?

Answer 20

European Federation of Clinical Chemistry and Laboratory Medicine

Question 21

How should desirable imprecision (CVa) be calculated?

Answer 21

Less than 0.5CVi

Question 22

How should desirable bias be calculated?

Answer 22

Bias should be less than 0.25 total biological variation

Question 23

How is total biological variation calculated?

Answer 23

Square root of (CVi^2 +CVg^2)

Question 24

How is total allowable error calculated?

Answer 24

TME + Bias
= 1.65ximprecision goal + bias goal
= 1.65 x 0.5CVi + SqRt(CVi^2+CVg^2)