Oncomine Dx Express Test (CE-IVD) for Pathologists

Analytical accuracy study

The analytical accuracy was evaluated with 151 clinical FFPE samples from six cancer types (breast, colorectal, glioma, melanoma, non-small cell lung, and thyroid cancer) using the Oncomine Dx Express Test and two NGS-based orthogonal reference assays. The concordance evaluation study was performed at two sites.

The positive percent agreement (PPA) and negative percent agreement (NPA) were defined as the proportion of variant-positive and variant-negative specimens, respectively, as determined by the reference methods that were also determined by the Oncomine Dx Express Test.

Analytical accuracy results are summarized in the following table:

Table 4: Oncomine Dx Express Test FFPE concordance

Limit of Blank study

The limit of blank was established by profiling 30 clinical FFPE samples confirmed to be variant-negative by a reference method. The study included: 2 replicates per sample, 2 reagent lots, and 11 tissue types (bladder, brain, breast, bile duct, colon, endometrium, lung, pancreas, prostate, skin, and thyroid).

For all 30 samples, the false-positive rate of the test was determined to be 0.75% for SNVs, 0% for indels, 0% for CNVs, and 0% for fusions. By definition of the Clinical and Laboratory Standards Institute (CLSI) EP17-A2, the limit of blank is zero.

Limit of Detection (LoD) study

The LoD was evaluated with 20 representative SNVs, indels, CNVs, and RNA fusions detected by the Oncomine Dx Express Test in clinical FFPE samples. The LoD is defined as the lowest variant level that can be detected at least 95% of the time.

Clinical specimens representing six cancer types (breast cancer, colorectal cancer, glioma, NSCLC, melanoma, and thyroid cancer) were used as the source of DNA and RNA. Variant-containing specimens were blended with wild-type samples, and the study included: 6 titration levels, 2 reagent lots, and 10 replicates per sample blend.

Based on a representative approach, the LoDs ranged from:

• 3.07% to 5.57% allelic frequencies for SNVs and indels (mean = 4.29% allelic frequency)
• 4.91 to 5.32 copies for CNVs
• 5.27 to 12.35 molecular counts (median = 8.85 molecular counts) and 7.87 to 207.5 reads for fusions

Precision study

The repeatability and reproducibility were evaluated using 20 representative DNA variants and RNA fusions in FFPE samples from six cancer types: breast cancer, colorectal cancer, glioma, melanoma, NSCLC, and thyroid cancer.

Three sites, with two operators and instruments per site, were used for the study. DNA and RNA was extracted from clinical FFPE samples, then blended with wild-type DNA or RNA into seven DNA blends and seven RNA blends. Two levels per blend were generated and distributed to sites and operators for testing.

The mean call rates excluding no-calls were 99.23%, 100%, and 99.69% for variant-positive SNVs/indels, CNVs, and fusions, respectively. The mean call rates excluding no-calls was 100% for wild-type DNA (negative-calls) and wild-type RNA.

The within-run repeatability call rate (no-calls excluded) was 98.35%, 100%, and 99.79% for DNA SNVs, DNA CNVs, and RNA variants, respectively.