Skip to main content

Cytogenetics

Validate & Reference Ranges

Before a Fluorescence In Situ Hybridization (FISH) probe can be used for patient diagnosis, it must undergo a rigorous Validation (for Laboratory Developed Tests/ASRs) or Verification (for FDA-cleared probes) process. This procedure confirms that the probe performs as intended within the specific conditions of the laboratory. It is not sufficient to rely on the manufacturer’s package insert; the laboratory must prove that its laboratory scientist s, equipment, and reagents can generate accurate and reproducible results. This process culminates in the establishment of a Normal Cut-off Value (NCV), the statistical threshold that differentiates a true low-level abnormality (mosaicism) from technical background noise

Probe Validation Components

The validation process assesses four key analytical performance characteristics: Accuracy, Precision, Sensitivity, and Specificity

  • 1. Localization (Mapping)
    • Before testing interphase nuclei, the probe must be hybridized to Normal Metaphase Spreads
    • Purpose: To confirm the probe binds to the correct chromosomal locus and only that locus
    • Specificity Check: The analyst verifies that there is no cross-hybridization (non-specific binding) on other chromosomes. For example, a probe for BCR (22q11) should signal only on chromosome 22. If it also lights up chromosome 14, the specificity is compromised
    • Sensitivity Check: The analyst verifies that the probe hybridizes to both homologues (e.g., both copies of chromosome 22) in 100% of metaphases
  • 2. Accuracy
    • The laboratory must run the new probe on samples with known genetic constitutions
    • Positive Controls: Samples known to carry the specific abnormality (e.g., a known CML patient for a BCR:ABL1 probe). The result must match the known diagnosis (Correct Positive)
    • Negative Controls: Samples known to be normal for that locus. The result must be negative (Correct Negative)
  • 3. Precision (Reproducibility)
    • Intra-assay Precision: The same sample is run multiple times in the same run to ensure the result is consistent
    • Inter-assay Precision: The same sample is run on different days by different laboratory scientist s. This ensures that the result is not dependent on which laboratory scientist performs the assay or slight daily variations in temperature/humidity
  • 4. Analytical Sensitivity & Specificity
    • Analytical Sensitivity: The percentage of chromosome targets that generate a scoreable signal. (e.g., “In 200 cells, 98% showed the expected signals”)
    • Analytical Specificity: The percentage of signals that are in the correct location (verified by the metaphase mapping step)

Establishing Reference Ranges (The “Normal Database”)

FISH is not a perfect binary assay. Due to technical limitations - such as signal overlap in 3D nuclei, inefficient hybridization, or random signal loss - even genetically normal samples will occasionally show “abnormal” patterns (false positives). The laboratory must quantify this background noise

  • The Validation Cohort
    • To establish the baseline, the probe is hybridized to 20 distinct, biologically normal samples: (e.g., 20 normal bone marrows for a leukemia probe, or 20 normal peripheral bloods for a constitutional probe)
    • These samples must represent the specific tissue type intended for the assay (e.g., you cannot validate a bone marrow probe using only blood)
  • Scoring the Normals
    • For each of the 20 normal samples, two laboratory scientist s independently score a large number of nuclei (typically 50–200 each)
    • They record the number of cells showing “false positive” patterns
    • Example: In a normal sample stained for BCR:ABL1 (fusion), a laboratory scientist might find 2 cells out of 200 that appear to have a fusion simply because the red and green signals overlapped by chance in 3D space. This is the “background.”

Calculating the Normal Cut-off Value (NCV)

The NCV is the statistical line in the sand. Any result below this number is considered “Normal/Background”; any result above it is considered “Abnormal/True Clone.”

  • Method 1: Gaussian Distribution (Mean + 3 SD)
    • The lab calculates the Mean percentage of false-positive cells across the 20 normal samples and the Standard Deviation (SD)
    • Formula: \(\text{Cut-off} = \text{Mean} + (3 \times \text{SD})\)
    • Logic: Statistically, 99.7% of normal data points fall within 3 SDs of the mean. Therefore, anything above this limit is statistically unlikely to be normal
  • Method 2: Beta-Inverse Function (Binomial Distribution)
    • Because false positives in FISH are rare events (often zero), the data does not follow a bell curve (Gaussian), making the “Mean + 3 SD” method less accurate
    • The Beta-Inverse Function (Excel BETAINV): This is the preferred statistical method for CAP-accredited laboratories. It calculates the upper confidence limit (usually 95%) for a population based on the observed zero-bound data
    • Result: This produces a conservative, statistically valid cut-off even when most normal samples show 0% false positives
  • Typical Cut-off Ranges
    • Fusion Probes (e.g., PML:RARA): Background is low (overlaps are rare). Typical NCV: < 1.0% to 3.0%
    • Break-Apart Probes (e.g., KMT2A): Background is moderate (random separation). Typical NCV: < 3.0% to 6.0%
    • Deletion Probes (e.g., ATM): Background is high (random signal loss/truncation is common). Typical NCV: < 5.0% to 10.0%
    • Amplification Probes (e.g., HER2): Based on Ratios rather than percentages (e.g., Ratio < 2.0)

Ongoing Maintenance (New Lot Verification)

Validation is not a one-time event. Whenever the laboratory receives a new Lot Number of the same probe from the manufacturer, a “Mini-Validation” (Verification) is required

  • The Comparison: The new lot is run in parallel with the old (validated) lot on the same patient sample
  • The Criteria: The results must be concordant. If the old lot shows 45% abnormal cells, the new lot should show a comparable percentage
  • Background Check: The new lot is often run on one normal control to ensure the background noise has not increased due to manufacturing changes