Development of an Evidence-Based Algorithm to Guide IGH-Based Reflex Testing for Plasma Cell Neoplasms

FISH testing on bone marrow aspirates is routinely used in the genetic workup for patients with plasma cell neoplasms. However, characteristic low cellularity and low in vitro proliferation rates may limit the number of plasma cells available to comprehensively test recurrent and prognostically significant genetic markers. Plasma cell selection via CD138+ cell isolation or cytoplasmic Ig staining is often required to improve sensitivity, and still may yield insufficient or poor quality cells, and possibly false negative results. Our current Multiple Myeloma panel (MMP) is designed as a reflex test, where positive IGH break-apart (IGH-BAP) and negative t(11;14)/IGH-CCND1 fusion patterns direct subsequent testing for t(4;14)/IGH-FGFR3 and t(14;16)/IGH-MAF fusions. Included in the first tier panel are probes for common trisomies 9 and 15, 1q/CKS1B gain, and 17p/TP53 deletion. This strategy is cost-effective by prioritizing testing for the most common IGH translocation (11;14) first, but can be problematic with low cellularity and/or low percentage patterns. To improve this process, we retrospectively reviewed data from 294-reflexed MMPs encountered in our clinical laboratory in order to develop an evidence-based algorithm for reflex testing. IGH-BAP-positive (IGH-BAP+) signal patterns included RGF (rearrangement), RF (5' deletion), and GF (3' deletion). Positive results for t(4;14) or t(14;16) dual fusion (DF) probes were observed in 75 (25.5%) and 40 (13.9%) samples, respectively. The median differential positivity rate for BAP versus DF probes was 8.5% (range: 0-39.5%), demonstrating similar detection rates between these probe sets (r=0.94, Pearson correlation). Reflex testing was performed on 263 (89.5%) CD138+ enriched cell pellets. IGH-BAP positivity (IGH-BAP+) rates ranged from 3.5-98.0% in this cohort, where higher rates correlated with positive t(4;14) and t(14;16) results (median DF+ =78.5%; median DF- =50.8%; p=2.97E-9, t-Test). Of 68 CD138+ samples with low percentage IGH-BAP+ (≤30%), only seven (10.1%) cases were DF+, six of which had additional abnormal MMP signal patterns at a similar percentage to IGH-BAP, suggestive of a low-level plasma cell population in these samples. The single positive case showed the typical GRF pattern for IGH-BAP (12.0%; DF=15.0%). Of 61 DF- cases, IGH-BAP+ was the sole finding in 23 (37.7%) cases, 16 (26.2%) cases had additional patterns at a similar percentage to IGH-BAP, and 22 (36.0%) cases showed additional signal patterns at much higher percentages, >2.5 times that of IGH-BAP. These findings suggest reflex testing has limited value for CD138+ samples with low IGH-BAP+ rates, especially with IGH-BAP as the sole abnormality, as well as for reflex cases with additional patterns at a much higher percentage, as t(4;14) and t(14;16) should represent primary alterations in these samples. Across all cases in this dataset, the IGH-BAP RF or GF (partial deletion) pattern was observed in 10/110 (9.0%) DF+ cases compared to 75/178 (42.1%) DF- cases, indicating these patterns are less likely to represent true IGH rearrangements. The 5' IGH probe may be deleted due to normal VDJ recombination. Of the four 5' IGH deletion cases that were DF+, all cases showed a complex partial deletion pattern (e.g. 2R2F, 2RF, etc) whereas 44 (81.5%, n=54) DF- cases showed a simple RF pattern, suggesting reflex may have limited utility for the simple RF pattern. Of 31 unsorted samples, 19 (61.3%) were DF+, 14 of which (73.7%) had IGH-BAP+ ≤30%. There was no difference between median IGH-BAP+ percentages in this cohort (DF+=14.0%, DF-=17.5%; range=4.5-77.0%; p=0.236, t-Test). Most DF+ cases (89.5%, n=17) had other abnormal MMP signal patterns in addition to IGH-BAP; two cases showed IGH-BAP+ as a sole finding, both with GRF patterns. These findings indicate reflex testing should be performed for all unsorted samples with IGH-BAP pattern above established cut-offs unless additional MMP signal patterns are present at much higher percentage. We present an algorithm for PCN sample processing that incorporates sample type, signal patterns, and relative detection rates for IGH and other MMP probes in order to guide appropriate IGH reflex testing in the clinical laboratory setting. This method predicts a reduction of greater than 10-15 percent of reflexed cases based on retrospective analysis. Prospective results from implementation will also be presented.