Clinical laboratory tests play an important role in diagnosing disease, monitoring treatment, and assessing patients’ health status.
Correct interpretation of laboratory results depends not only on knowing the measured value. The reference interval used for comparison, the test’s diagnostic performance, and the characteristics of the measurement method must also be considered.
Reference intervals and diagnostic cut-offs are often confused, yet they serve different purposes.
What is a reference interval?
A reference interval is defined from the distribution of laboratory values in a specified reference population.
A common clinical laboratory approach uses the central 95% of values from appropriately selected reference individuals.
For example, a reference interval for a serum biochemical parameter covers most values among reference individuals selected under defined conditions.
A result outside the reference interval does not necessarily indicate disease. Likewise, a result within the interval does not completely exclude disease.
Reference intervals are statistical tools that support clinical assessment.
1. Why can reference intervals differ by laboratory?
A test’s reference interval can be affected by:
- Demographic characteristics of the reference population
- Age and sex distribution
- The analytical method used
- Measurement system and calibration
- Sample collection and processing conditions
- Selection of reference individuals
Therefore, reference intervals taken from another laboratory or manufacturer should be assessed for suitability in the laboratory that will use them.
Verification becomes especially important when methods, patient populations, or sampling conditions differ.
2. What does verification of reference intervals mean?
Reference interval verification evaluates whether an existing interval is suitable for use in a given laboratory and target population.
This is not the same as establishing a new reference interval from scratch.
A laboratory can assess acceptability of an existing interval with an appropriate verification approach.
If verification finds the interval unsuitable, possible causes should be reviewed and, if needed, laboratory-specific intervals may be established.
3. How are reference intervals calculated?
Establishment studies generally use direct and indirect methods.
Direct methods
In direct methods, the reference population is selected by predefined criteria and measurements from these individuals are used.
This approach offers a controlled process for defining reference individuals, but recruiting enough suitable people can require time and resources.
Indirect methods
Indirect methods often use routine test results stored in laboratory information systems.
Because these datasets also include patient results, the reference distribution must be estimated with statistical methods.
A key advantage is use of existing laboratory data. Patient mix, data quality, and method assumptions can still affect results.
Careful method selection and evaluation of the resulting intervals are therefore important.
4. How do reference intervals differ from diagnostic cut-offs?
These concepts are different.
A reference interval reflects the measurement distribution in a defined reference population.
A diagnostic cut-off may be used to classify a disease, clinical condition, or risk level.
For example, a biomarker’s diagnostic cut-off can be assessed using results from diseased and non-diseased groups.
A laboratory test’s reference limit and clinical decision limit therefore do not need to be the same.
5. How is diagnostic performance evaluated?
A laboratory test’s ability to distinguish disease is assessed with several statistical metrics.
Sensitivity: The proportion of individuals with disease identified as positive by the test.
Specificity: The proportion of individuals without disease identified as negative by the test.
Positive predictive value: Among those with a positive result, the proportion who truly have disease.
Negative predictive value: Among those with a negative result, the proportion who truly do not have disease.
Unlike sensitivity and specificity, predictive values are also affected by disease prevalence in the evaluated population.
6. What is ROC analysis for?
ROC (Receiver Operating Characteristic) analysis evaluates the relationship between sensitivity and specificity of a diagnostic test across cut-offs.
The area under the ROC curve (AUC) summarizes the test’s ability to separate diseased and non-diseased individuals.
A high AUC does not mean the test is the best decision tool in every clinical setting.
Intended clinical use, the impact of false positives and false negatives, and characteristics of the patient group must also be considered.
Cut-off selection should be judged with clinical consequences, not statistical optimization alone.
7. Why is laboratory data analytics increasingly important?
Large volumes of routine test data now accumulate in laboratory information systems.
When evaluated appropriately, these data can support:
- Review of reference intervals,
- Comparison of laboratory methods,
- Assessment of biomarker diagnostic performance,
- Investigation of factors such as age and sex,
- Development of clinical prediction models.
Having large amounts of data alone does not produce reliable results. Data quality, patient selection, and appropriate statistical methods must be considered.
Conclusion
Reference intervals, diagnostic cut-offs, and performance metrics are different but related concepts that support clinical interpretation of laboratory results.
Assessing whether existing intervals are suitable, establishing laboratory-specific intervals when needed, and analyzing diagnostic performance with appropriate methods help use laboratory data more effectively.
Laboratory Data Analytics with Data4Care
Data4Care provides scientific and analytical support for reference interval analyses, ROC analysis, diagnostic performance evaluation, and laboratory method comparison.
For solutions suited to your laboratory data, explore our Laboratory Data Analytics & Diagnostic Performance service.