Verification and Validation of Reference Intervals in Clinical Laboratories

 Why is validation and Verification of Clinical Lab Reference Interval Required? 

Introduction

Understanding Reference Intervals:

Reference intervals are the normal ranges for laboratory test results in a healthy population. They are essential for interpreting test results and determining if a patient's values are within or outside the expected range.

Verification and validation of reference intervals are crucial processes in clinical laboratories, each serving distinct purposes in ensuring the accuracy and reliability of laboratory test results.

Verification of Reference Intervals: 

Verification refers to the process of confirming that a pre-established reference interval is applicable to a specific laboratory's population and testing methods. This is typically done when a laboratory adopts reference intervals from external sources, such as manufacturers or other laboratories. The verification process can be accomplished using a smaller sample size, often as few as 20 reference individuals from the local population.

Validation of Reference Intervals: 

Validation, on the other hand, involves establishing new reference intervals based on data collected specifically for a laboratory's patient population. This process is more comprehensive and resource-intensive, requiring, a minimum of 120 individuals is typically needed to establish statistically valid reference intervals, ensuring that the sample adequately represents the target population. 

Key Differences: 

Aspect	Verification	Validation
Purpose	Confirm applicability of existing intervals	Establish new reference intervals
Sample Size	Typically 20 individuals	Minimum 120 individuals
Process Complexity	Less complex; relies on existing data both use same statistical procedures	More complex; requires new data collection
Outcome	Confirms that existing intervals are suitable	Creates new reference intervals tailored to a population

Verification and validation are crucial because:

• Populations vary: Reference intervals may differ based on age, sex, ethnicity, and health status.
• Clinical utility: Accurate intervals improve diagnosis and treatment outcomes.
• Regulatory requirements: Many regulatory bodies mandate interval validation.

Steps Involved in Verification and Validation Reference Intervals:

1. Define the Reference Population: Identify the specific population you want to represent (e.g., age, gender, ethnicity). Ensure the population is healthy and free from factors that could influence test results. For this example, 20 healthy adults aged 20–40 years in Addis Ababa, Ethiopia, without underlying health conditions.

2. Collect Samples: Collect a sufficient number of samples from the defined reference population.

Consider factors like sample type, collection method, and storage conditions.

3. Perform Testing: Analyze the samples using the same method and equipment as used for patient testing. Ensure proper quality control procedures are in place.

4. Statistical Analysis: Calculate the mean, standard deviation, and confidence intervals for the test results. Determine the appropriate statistical method based on the distribution of the data as shown below.

 

4.1. Reference Interval for Normally Distributed Data                                                      

• When to use: When the data is normally distributed (bell-shaped curve).
• Calculation: Reference interval = Mean ± 1.96 * Standard Deviation.

Below are 20 glucose readings from the healthy reference population (in mg/dL):

85, 90, 87, 92, 95, 93, 89, 88, 86, 91,  

94, 89, 92, 96, 97, 90, 88, 89, 93, 91   

Formula for Mean (μ):

$$\mathrm{<b>\mu </b>}=\frac{<b>\sum </b>\mathrm{<b>X</b>}}{\mathrm{<b> /n \; \; \; \; \; \; \; </b>}}$$

$$\frac{\mathrm{<span\; style="font-size:\; large;"><b>Where:</b> \; \; \; \; \; \; \; </span>}}{}$$

• X = individual glucose values
• $\mathrm{<b>n\; =\; number\; of\; samples</b>}$

$$\mu = \frac{85 + 90 + 87 + 92 + \ldots + 91}{20} = 90.3 \, \text{mg/dL}$$

      Therefore, Mean (μ): 90.3 mg/dL

Formula for Standard Deviation (σ):

​

 

Calculate deviations from the mean for each value.

Square the deviations and sum them.

Example for first value:
$(85 - 90.3)^2 = (-5.3)^2 = 28.09$
(Similarly calculate for all 20 values.)

After calculating all squared deviations:

Standard Deviation (σ): 3.18 mg/dL

 Calculate the Reference Interval:

4.2. Reference Interval for Non-normally Distributed Data

Validating a reference interval for glucose using percentile-based calculations (for non-Gaussian data) involves establishing the 2.5th and 97.5th percentiles as the lower and upper reference limits. Let’s walk through the steps with example data and calculations.

Step-by-Step Calculation for Reference Interval Validation

Sample Data (Glucose levels in mg/dL):

73, 85, 77, 92, 110, 68, 87, 95, 72, 106  
82, 79, 91, 102, 75, 108, 80, 98, 84, 66  

• Total Sample Size: 20
• Distribution: Non-Gaussian

Step 1: Sort the Data in Ascending Order

66, 68, 72, 73, 75, 77, 79, 80, 82, 84,  
85, 87, 91, 92, 95, 98, 102, 106, 108, 110  

Step 2: Identify the 2.5th and 97.5th Percentile Positions

For non-Gaussian distributions, the percentile method is used to identify the limits.

Formula to locate the percentile position:

Where:

• $P$ = desired percentile (e.g., 2.5 for lower limit, 97.5 for upper limit)
• n = total number of samples

2.5th Percentile (Lower Reference Limit):

Since the value is between Position 1 (66) and Position 2 (68), we interpolate:

Interpolated Value=66+(0.525×(68−66))=67.05

         So, the 2.5th percentile value = 67.05 mg/dL.

97.5th Percentile (Upper Reference Limit): 

Since this is between Position 20 (110) and Position 19 (108), we interpolate:

Interpolated Value=108+(0.475×(110−108))=108.95

So, the 97.5th percentile value = 108.95 mg/dL.

Step 3: Calculate the Reference Interval

The reference interval based on 2.5th and 97.5th percentiles is:

Reference Interval=[67.05mg/dL,108.95mg/dL]

Step 4: Interpretation

In this example, the validated reference interval for glucose is 67.05–108.95 mg/dL. This interval can now be used to assess whether patient glucose levels are within the normal range. If a patient's value falls outside this range, it may indicate hyperglycemia or hypoglycemia.

This method ensures a robust approach for non-Gaussian data, as it does not rely on mean or standard deviation but uses percentiles to accommodate skewed distributions.

Important Considerations:

Outliers: Identify and manage outliers to avoid distorting the reference interval.

Sample Size: A sufficiently large sample size (at least 20 for reference validation and 120 for establishing new intervals) is required for reliable results.

Clinical Utility: A patient’s glucose reading below 67.05 mg/dL could indicate hypoglycemia, while a value above 108.95 mg/dL may suggest early hyperglycemia, requiring further investigation.

If two or more results (≥10%) fall outside the reference interval being validated, the validation is considered unsuccessful. In such cases, the process should be repeated with a larger sample size to ensure reliability and accuracy.

----------------------------------------------------------------------------------------------------------

Summary: 

1. Gaussian Distribution:

• When to use: When the data is normally distributed (bell-shaped curve).
• Calculation: Reference interval = Mean ± 1.96 * Standard Deviation.

2. Non-Gaussian Distribution:

• Percentile Method:
  • Order the data in asceding order and determine the desired percentiles (most of the time, 2.5th and 97.5th).
  • Use these percentiles as the reference interval limits.

References 

1.https://en.wikipedia.org/wiki/Normal_distribution 

2. https://www.investopedia.com/terms/n/normaldistribution.asp 

3.https://simple.wikipedia.org/wiki/Normal_distribution 

4. https://www.sciencedirect.com/topics/biochemistry-genetics-and-molecular-biology/gaussian-distribution 

5. https://www.medcalc.org/manual/referenceinterval.php