Chromosomal localization may also be confi rmed by using both known positive and negative control samples. The use of pos- itive and negative control slides validates abnormal results found in single signal probe hybridizations. In any case, a cell should not be counted if the proper control signal pattern is not observed.
Fig. 2 Colocalization of home-brewed FISH probes from bacterial artifi cial chromosome (BAC) clones on normal metaphases. Top panels : arrows pointing to the anticipated signals on chromosomes. Bottom panels : arrows pointing to the anticipated signals on inverted DAPI stain
Probe sensitivity measures how frequently a probe hybridizes to its intended genomic target. It is calculated as the percentage of cor- rect targets detected out of the total number of intended targets presented. Minimum probe sensitivity should be 95 %. Probe specifi city measures how frequently a target detected by a probe is truly the intended genomic target by design. It is calculated as the percentage of correct targets detected out of the total number of targets detected. Minimum probe specifi city should be 98 %. Probe sensitivity and specifi city are different in different types of samples, e.g., suspensions or FFPE tissues. A brief overview of probe sensi- tivity and specifi city determination, based on ACMG and CLSI recommendations [ 8 , 10 ], is summarized in Fig. 3 . During the initial step of probe validation, probe sensitivity and specifi city should be measured for the type of sample on which the assay will be used, and the results should be documented for inspection purposes. If a probe does not achieve a minimum sensitivity of 95 % and specificity of 98 %, the preparation of the samples and/or the FISH protocol should be optimized. Probe sensitiv- ity and specifi city must be remeasured after optimization. If the above-mentioned probe sensitivity and specifi city requirements are not met after remeasurement, the FISH test should not be introduced into clinical use.
Any possible cross-hybridization inherent to a probe, as well as contaminations with other DNA sequences, should be identifi ed and documented. Examine metaphases using the signal colocaliza- tion method described in Subheading 2.3 should be simple and effective. Probes with signifi cant cross-hybridization or contamina- tion should not be used.
2.4 Probe Sensitivity and Specifi city
2.5 Cross- Hybridization and Contamination Issues
FDA-approved Probe
Procedure:
Manufacture's instruction Validation: >10 samples with at least
1 abnormal sample
Metaphase FISH probe for nonmosaic
Interphase FISH probe for nonmosaic
Interphase FISH probe for mosaic conditions
Validation: >5 metaphases with at
least 1 abnormal metaphase Requirement:
sensitivity > 95%;
specificity>98%
Validation: at least 50 nuclei from 5 samples counted by two
readers
Paraffin-embedded FISH probe for mosaic
conditions Non-FDA-approved
FISH Probe Validation: 10 normal peripheral
blood samples Examine 100 metaphases for
autosomal targets & 200 metaphases for sex chromosome targets initially. If sensitivity and sepecificity are both 100%, reduce
to 20 metaphases for autosomal targets and 40 metaphases for sex
chromosome targets.
Validation: >20 normal samples with at least
200 nuclei and 1 abnormal sample Requirement:
sensitivity > 95%;
specificity>98%
Validation: >5 normal samples with at least 50 nuclei, 1 abnormal
sample Requirement:
sensitivity > 95%;
specificity>98%
Fig. 3 FISH probe sensitivity and specifi city determination
Intra- and inter- assay reproducibility should be demonstrated and documented in at least three samples. Intra-assay reproducibility can be achieved by testing triplicates of three or more samples on a single day. Inter-assay reproducibility can be done by testing dupli- cates of 3 or more samples on 2 different days. Variation of the results should not exceed 5 %.
Establishing and optimizing a SOP is very important to FISH probe validation, and the SOP should be strictly followed when dealing with different types of samples with special preparations.
Qualifi ed, accredited clinical laboratories must establish SOPs based on the laboratory FISH procedure used in-house or the FISH manufacturer’s recommendations. SOPs vary depending on the source and type of the FISH probe and should include methods for the modifi cation of prepared slides, pretreatment of the slides, denaturation conditions, probe hybridization, and post- wash in order to maximize signal intensity and minimize back- ground noise. It is also necessary to check fl uorescent imaging equipment to make sure all components are working properly, since incompatible fl uorescent fi lter cubes, misaligned mercury light bulbs, and incorrect fl uorescent microscope settings can all contribute to weak signaling for FISH assays. In some cases, the design of the probe or the method of fl uorescent labeling contrib- utes to weak signals.
Because there are currently no standardized criteria for signal pat- tern inclusion and exclusion, accredited laboratories—whether using manual signal quantitation or FISH spot-counting software—
must set up criteria for evaluating signal patterns and frequencies when validating clinical FISH tests. Depending on the probe design, there is usually more than one signal pattern that is consid- ered abnormal and/or indicative of a chromosome abnormality.
For example, dual-color, dual-fusion probes can give either a single fusion signal or dual fusion signals, both of which could be counted as patterns positive for an abnormality. In a break-apart probe assay, either a separated red and green signal pattern or a single- color signal outside the gene of interest (caused by the deletion of the other color signal that is mapped to the region of the gene of interest) is considered positive. Twenty to 30 known- positive sam- ples by another gold-standard test should be used to screen for different positive signal patterns and their frequencies. Rare signal patterns seen in less than 10 % of the known-positive samples should be excluded from future signal quantitation.
Quantitation criteria should be established for each clinical FISH assay during validation. Such criteria include the eligibility of cells for signal quantitation and the defi nition of positive and negative signal patterns. Clustered cells with too much overlap that 2.6 Reproducibility
2.7 Establishment of FISH Assay SOPs
2.8 Inclusion and Exclusion of Signal Patterns
2.9 Establishment of Quantitation Criteria
interferes with signal quantitation should not be used. Validated positive and negative signal patterns should be specifi ed on a FISH signal quantitation record sheet. Similar criteria should be estab- lished for FISH assays quantitated by automated FISH spot coun- ters. The effectiveness of quantitation criteria could be evaluated by clinical sensitivity and specifi city measurements.
Analytical sensitivity and specifi city evaluate the relationship between abnormal cells and abnormal signal patterns regardless of presence or absence of disease. Analytical sensitivity is defi ned as the percent- age of cells with abnormal signal patterns that a FISH test reliably detects. Analytical specifi city is defi ned as the percentage of cells with normal signal patterns that a FISH test identifi es correctly.
Analytical sensitivity and specifi city should be no less than 95 % for a good FISH assay, although greater discriminatory power may be needed to distinguish mosaicism. The ACMG guidelines recom- mend the use of 200 unique genomic targets for analytical sensitivity and specifi city establishment. Such unique genomic targets are defi ned as the same sequence from each of the separable metaphase chromatids that will hybridize with the unique sequence probe. By this defi nition, a metaphase has four unique genomic targets when the target sequence is away from the centromere and clearly separa- ble. Fifty cells are required for this type of analysis. On the other hand, a metaphase has two unique genomic targets if the target sequence is located near the centromere that could result in two non-separable hybridization signals. Such analysis would require 100 cells instead (ACMG, E9.2.1.1). Cells should be from fi ve chro- mosomally characterized individuals (Choosing aneuploidy cell lines can maximize the number of targets). Pooling of cells from these fi ve individuals is acceptable as long as all cells have the same number of potential targets and comparable mitotic indices. If discordance is present, test the individual cell lines separately. All internal validation data should be fully documented.
3 Clinical Validation of FISH Assays
Standard control probes or samples should be used to detect errors or technical failures during the three phases of the clinical valida- tion process. The goal in the pre-analytic phase is to determine whether the correct probe was used; the goal in the analytic phase is to verify that the counting criteria are followed correctly; and the goal in the post-analytic phase is to determine whether the assay’s performance is satisfactory. Qualitative FISH tests are used in most nonmosaic situations, while quantitative FISH tests are designed to evaluate tumor burden or levels of mosaicism. Different control strategies should be considered for qualitative and quantitative FISH tests. Positive samples are useful as quality control indices 2.10 Establishment
of Analytical Sensitivity and Specifi city
3.1 Use of Controls
during the clinical validation phase, and untreated patient samples are preferred because the percentage of abnormal cells may drop below the defi ned threshold level after treatment. It is also helpful to use control samples with different concentrations of positive cells in order to ensure the analytical sensitivity of quantitative FISH tests. Commercial, chromosome-specifi c, centromeric, or telomeric probes are commonly used as control probes for home- brewed, loci-specifi c probes either on the same or on differ- ent chromosomes. A common source of control samples is extra sampling materials, such as positive cell pellets in fi xative and cell lines that are positive for the tested abnormalities [ 12 ]. Known positive or negative samples are confi rmed by other testing to be with or with- out a particular genetic abnormality. FISH probe validation should include at least one positive sample during the process of technical specifi cation assessment. Negative samples used for clinical validation should be confi rmed by other tests as being truly negative for the abnormality that the FISH probe is meant to detect.
Clinical sensitivity and specifi city evaluate the relationship between disease status and a positive FISH test. Clinical sensitivity is defi ned as the percentage of positive cases that a FISH test identifi es and that are confi rmed as positive by a gold standard test. Clinical spec- ifi city is defi ned as the percentage of negative cases that a FISH test identifi es and that are confi rmed to be free of disease by a gold standard test. Gold standard tests useful in clinical validations of FISH tests include pathologic evaluations or, in the case of many genetic disorders, G-band karyotyping . A gold standard test that is a validated alternative method of confi rming positive and negative cases—performed either in the same laboratory or in a reference laboratory—should be used on the same set of samples for com- parison. For example, if immunohistochemistry (IHC) staining is used as an alternative validation method, a positive FISH test should correspond to a positive IHC staining test. Criteria for concordance should be maintained as a part of the laboratory’s validation records, and the degree of concordance between the two methods should be documented. Any major changes to or modifi cations of SOPs warrant revalidation of the assay (Fig. 4 ).
Hajdinjak’s evaluation of clinical sensitivity and specifi city using the UroVysion FISH probe kit for bladder cancer provides a good model of FISH test validation [ 13 ]. In addition, clinical studies supporting the validity of FISH assays and published in peer- reviewed journals can be very helpful in this regard.
Determining a reference range for interphase FISH assays is neces- sary for both constitutional and acquired abnormalities. The refer- ence range is defi ned as the range of FISH test values in 95 % of individuals who are free of the genetic condition relevant to the FISH test [ 14 ]. Healthy individuals should be recruited and tested 3.2 Clinical
Sensitivity and Specifi city
3.3 Determination of Reference Ranges