W13L17&18 - Reference Intervals Flashcards
What is a Reference Interval?
A range of values for a given analyte
Usually bound by two limits, and upper and lower limit
Commonly used for comparison with an observed value from a patient under investigation
Numerous types exist
Reference Intervals - ‘Stakeholders’
Manufacturers of reagents - often provide reference intervals with their assays/reagents
Clinical laboratories - often provide reference intervals with results
Clinicians - interpret results to make decisions about patient health
Patients - understand their health status and the monitoring of their health status
Situations for Describing/Defining a Reference Interval
Emergence of a new assay or biomarker for which the reference interval has not yet been determined
Expanding the use of an established reference interval
An existing reference interval that needs to be re-defined with a new reagent/method/instrument
Subject Based Reference Values
Refer to the levels of analytes obtained from the patient when they are healthy and used for comparison when the patient is being investigated for a disease
Population Based Reference Intervals
Most common type of reference interval
Values derived from studies on subjects from a population ‘without disease’
Subjects without signs of diseases selected to form reference population
Selection of Reference Individuals
Selection of reference individuals must avoid characteristics that will act on the analyte(s) of interest
- e.g. alcohol consumption, smoking, obesity etc.
Partitioning criteria are characteristics of the selected reference individual that divide the reference sample into significant subclasses
- e.g. age, sex, ethnicity, blood group etc.
Preanalytical Considerations for Subject Preparation and Specimen Collection
Subject preparation: - diet - sampling time - physical activity - stress - rest before collection Specimen collection - environmental conditions during collection - time - specimen type - collection site - blood flow - equipment - technique
Outliers
A value that stands unexpectedly far from most of the reference values
An outlier may be due to variability in the measurement or it may indicate experimental error
Visual inspection of a histogram is one way to identify outliers
Statistical tests that identify outliers include:
- Dixon’s Q test
- Grubbs’ test
- Chauvenet’s criterion
Transference
Basically how can you transfer one confidence interval to our particular situation
Also needs to make sure that it is appropriate and reliable
For the reference interval to be transferred it has to meet some prerequisites:
- have similar precision
- have similar known interferences
- use the same or comparable standards/calibrators
- provides values that are acceptably comparable
Validation of Transference - Approaches to Validation
Subjective assessment
Using small number of reference intervals
Using large number of reference intervals
Validation of Transference - Subjective Assessment
Acceptability of the transfer is assessed to see if it is appropriate to use in the lab due to the similarity between published reference values and your own situation
Validation of Transference - Using Small Number of Reference Intervals
Uses small number of samples from population to compare the system that has been published with your own system
20 individuals recommended
Analytical and preanalytical factors of the original study must be consistent with the receiving labs operation
E.g. With the 20 people, put them into your system and seeing if everyone falls between the published reference intervals, if they do then proven similarity in the systems
Validation of Transference - Using Large Numbers of Reference Intervals
Basically the same as using small numbers except with more people
60 individuals recommended
Used for more extensive reference interval transference study for analytes whose reference intervals are critically important for interpretation
Analytical and preanalytical factors of the original study must be consistent with the receiving labs operation
How to find upper and lower limits for non-parametric data?
Lower limit: r1 = 2.5th percentile = 0.025(N+1)
Upper limit: r1 = 97.5th percentile = 0.975(N+1)
E.g. for a study with 120 people:
0.025 x 121 = 3.025 = 3
0.975 x 121 = 117.975 = 118
Therefore the limits will be the values from rank 3 and rank 118
How to find the upper and lower limit for parametric data?
Lower limit = mean - 2x SD
Upper limit = mean + 2x SD
Therefore reference interval = mean+- 2x SD
