Clsi Ep28 ~repack~ Jun 2026

Defining the Baseline: A Deep Dive into CLSI EP28 – Determination of Reference Intervals In the world of laboratory medicine, a single test result is rarely a definitive diagnosis on its own. A potassium level of 4.2 mmol/L is meaningless until you know that the normal range (reference interval) is approximately 3.6 to 5.2 mmol/L. But how are those ranges established? Who decides what "normal" means? The answer lies in a critical, yet often overlooked, document: CLSI EP28 . Officially titled "Defining, Establishing, and Verifying Reference Intervals in the Clinical Laboratory," CLSI EP28 is the global gold standard for how laboratories determine the range of test results expected for a healthy population. Without this guideline, patient results would be a chaotic sea of numbers with no context, leading to missed diagnoses, unnecessary treatments, and eroded trust in medical testing. This article provides an exhaustive exploration of CLSI EP28, its history, its statistical methodologies, its practical application in the modern lab, and the legal and accreditation risks of ignoring it.

What is CLSI EP28? The Scope and Purpose The Clinical and Laboratory Standards Institute (CLSI) is a non-profit organization that develops voluntary consensus standards for healthcare testing. While "voluntary" is the legal term, in practice, adherence to CLSI guidelines is mandatory for laboratories seeking accreditation from bodies like CAP (College of American Pathologists), COLA, or The Joint Commission. CLSI EP28 (formerly known as C28-A3 and C28-A2 before revisions) serves a singular, vital purpose: to provide a standardized protocol for the statistical determination of reference intervals. The Core Definitions To understand EP28, you must master its specific lexicon:

Reference Individual: A person selected based on well-defined criteria (e.g., no acute illness, not taking certain medications) used to define a reference interval. Reference Population: The entire hypothetical set of reference individuals (e.g., all healthy adult women aged 30-40). Reference Sample Group: The actual subset of the population you can physically test. Reference Value: The test result obtained from a reference individual. Reference Interval: The interval between, and including, two reference limits (typically the 2.5th percentile and the 97.5th percentile). This is what doctors call the "normal range."

Why "EP28" Matters The document standardizes everything from how many healthy subjects you need to test (sample size) to how to handle outliers and partition by sex or age. It specifically warns against two dangerous practices:

"Transference" without verification: Assuming a range from a reagent manufacturer or a textbook applies to your specific population. "Eyeballing" the data: Using internal, non-standardized data from sick patients to guess a normal range.

The Three Pillars of EP28: Establishment, Transference, and Verification One of the most confusing aspects of laboratory medicine is knowing which method to use. EP28 clarifies three distinct approaches based on a lab’s resources and patient population. 1. Reference Interval Establishment (De Novo) This is the "ground-up" method. You recruit 120+ healthy individuals (the document suggests a minimum of 120 per partition to achieve 90% confidence intervals around the limits), collect specimens, run the tests, and calculate the central 95% interval. When to use: New analytes, unique local populations with genetic diversity, or when published data is non-existent. The 120 Rule: EP28’s most famous recommendation is that for a non-parametric (distribution-free) method, you need at least 120 reference individuals to reliably estimate the 2.5th and 97.5th percentiles. Fewer than 120 leads to wide confidence intervals, meaning your "normal range" might be wildly inaccurate. 2. Reference Interval Transference This is the most common method in modern labs. Instead of finding 120 healthy people, you adopt a reference interval from a "trusted source"—a reagent manufacturer, a peer-reviewed study, or another laboratory. The Catch: EP28 states that transference is only valid if your laboratory’s methodology and patient population are identical to the source’s. You must perform a "verification" study (see below) to prove the transferred range works. 3. Reference Interval Verification This is the pragmatic compromise. If you transfer an interval from a manufacturer, EP28 requires you to verify it using a much smaller sample size—typically 20 healthy individuals from your own local population. The EP28 Verification Rule:

Collect samples from 20 healthy reference individuals. Run the test. If no more than 2 of the 20 results fall outside the manufacturer’s claimed reference interval, the interval is verified for your lab. If 3 or more fall outside, you must reject the transferred interval and perform a full establishment study (120+ subjects).

This small verification study is the single most common CAP inspection finding. Labs that skip this step are cited for non-compliance.

The Statistical Engine: How EP28 Handles the Numbers EP28 is not a statistics textbook, but it provides rigorous, step-by-step instructions. Here is the high-level workflow prescribed by the guideline. Step 1: Defining Exclusion Criteria (The "Healthy" Problem) What does "healthy" mean? EP28 demands explicit exclusion criteria. Typical exclusions include:

Acute or chronic illness (diabetes, hypertension, renal failure) Pregnancy or lactation Obesity (BMI > 30) Use of prescription or over-the-counter drugs (hormones, NSAIDs) Smoking or heavy alcohol use

Step 2: Partitioning (Do you need separate ranges for men/women/children?) Before calculating the interval, you must determine if the population is homogeneous. EP28 recommends using Harris and Boyd's method (standardized in the document) to test if two subgroups (e.g., male and female) require separate reference intervals. Example: Creatinine requires sex-specific intervals. Glucose does not. Step 3: Outlier Detection A single drunk patient or a hemolyzed sample can skew the 97.5th percentile. EP28 specifically endorses the Reed-Dixon and Tukey methods for outlier removal. The rule of thumb: remove no more than 2-3% of the data, or the interval becomes invalid. Step 4: Calculating the Interval EP28 prefers non-parametric (rank-based) methods over parametric (Gaussian) methods because most biologic data is not perfectly bell-shaped.

Non-parametric calculation: Rank all results from low to high. The lower reference limit is the value at position (0.025 * N + 0.5). The upper limit is the value at position (0.975 * N + 0.5). Confidence Intervals: The document also requires you to calculate 90% confidence intervals around these limits. If those confidence intervals are too wide, your sample size (N) is too small.