Proficiency testing of PIK3CA mutations in HR+/HER2-breast cancer on liquid biopsy and tissue

This ring trial was conducted in two phases: firstly, an internal proficiency test with five participating laboratories was carried out to identify suitable samples and to define the consensus mutation status. The results of this internal proficiency test were used to assemble the sample cohort for the ensuing external proficiency test with in total 62 institutions (Fig. 1).

Internal proficiency test—tissue

For internal testing of split 1 tissue, eight samples with concordant results of three institutes were used (Suppl. Table 3). Due to limited material, two PIK3CA mutated samples were used in duplicates. Tissue samples of split 1 contained six PIK3CA mutated (p.H1047R, p.E542K) and four PIK3CA wild-type samples. The results of the clinical trial qPCR assay (therascreen) showed strong divergence to expected values. The PIK3CA p.Q546R hotspot mutation was detected in 11 of 12 cases (split 1), which led to a false-positive rate of 92%. Therefore, results of this assay were excluded from further evaluation.

Seven samples tested for split 2 showed concordant results for at least three methods including PIK3CA mutations p.Q546K, p.E542K and p.H1047R (Suppl. Table 4). These samples and three wild-type samples tested in split 1 were considered suitable for external proficiency testing (10 samples split 1, 10 samples split 2). Allelic fractions of PIK3CA mutations ranged from 16 to 43%.

Internal proficiency test—liquid biopsy

Testing of LB reference design I (long fragments) with NGS assays showed concordant results for PIK3CA and KRAS with unexpected low allelic fractions for the hybrid capture-based Avenio assay (min 0.1% to max 5.1% allelic fractions) (Suppl. Fig. 1).

One out of three institutes used the clinical trial assay (therascreen) showing false-positive PIK3CA p.Q546R mutations in three samples leading to a false-positive rate of 30%.

Liquid biopsy reference design II (short fragments) showed concordant results for LB-specific assays (OBcfDNA, therascreen, and AVENIO) with allelic fractions in the expected range (Suppl. Fig. 1). Regardless of the expected allelic fraction, PIK3CA p.H1047R and KRAS p.G12D mutations were not detectable with the tissue-specific CLv2 assay in design II.

External proficiency testing

In total, 62 institutions from four countries registered for the external proficiency test, including 51 participants from Germany, eight from Austria, two from Switzerland, and one from Italy. Twenty-eight were applied for split 1 (ten tissue and ten LB samples) and 29 for split 2 (10 tissue samples). Seven institutes, which participated in split 1, and six additional institutes also opted in for LB testing of design II.

External proficiency testing—tissue

Forty-three of fifty-four participants (80%, split 1 and split 2) submitted results and 43 participants (80%) successfully passed the trial (Table 1).

Table 1 Methods and results of external proficiency tissue test split 1 and 2; NGS, next generation sequencing; qPCR, quantitative polymerase chain reaction; ddPCR, digital droplet PCR

For split 1, one wild-type sample was not evaluable in 5 cases and false-positive in two cases, which constituted > 25% (7/27) of the participants. Therefore, participants were given full scores for this case, and it was excluded from further evaluation.

For split 2, one PIK3CA p.Q546K mutated case was reported as false-negative by > 25% of the participants (7/27). Therefore, according to predefined definitions for the evaluation phase, the case was also excluded from further consideration. All attendees were given full scores for this case, and it was excluded from further analysis. Altogether, PIK3CA mutation detection for tissue reached a sensitivity > 98% and specificity of 97%, kappa 0.92 (Table 2).

Table 2 Results from 54 participants (27 split 1, 27 split 2) of external proficiency testing. Data include results of 18 tissue samples (split 1: 6× PIK3CA mutated (mut), 3× wild-type (WT), split 2: 6× PIK3CA mut, 3× WT). Two samples (1× WT split 1, 1× PIK3CA mut split 2) were excluded due > 25% deviation from external vs. internal resultsExternal proficiency testing—liquid biopsy

Eighty-three percent of the participants (20/24) that analyzed LB design I and 31% (4/13) for design II passed the proficiency testing successfully. The sensitivity of PIK3CA mutation detection was higher for participants using design I (95.8% vs. 73.0%). The same was observed for the specificity (98.3% vs. 82.1%) and kappa value (0.94 vs 0.55) (Table 3).

Table 3 Results of liquid biopsy external proficiency testing from 24 (design I, mutated fragments around 490 bp length) and 13 participants (design II, mutated fragments around 167 bp length). Data include results of 10 liquid biopsy samples (5× PIK3CA mutated (mut), 5× wild-type (WT))

Four participants did not pass the LB testing part with reference design I. One used a tissue-specific NGS assay (CLv2, Thermo Fisher Scientific) and had a false-negative result regarding a PIK3CA p.E545K mutation with an allelic fraction of 3%. Additionally, one participant evaluated a p.H1047R PIK3CA mutated sample with an allelic fraction of 3% as false-negative and two wild-type samples as false-positive, reporting PIK3CA p.C420R (1.9% allelic fraction) and a hotspot mutation p.E542K (0.9% allelic fraction).

Nine participants did not pass the testing of reference design II, four of them used a LB-specific assay. Results of mutation detection with design II showed a higher false-positive rate (15.4% vs. 1.7%) and a higher false-negative rate (26.1% vs. 4.2%). Participants who passed the trial used either digital droplet PCR (ddPCR), PCR/mass spectrometry, or LB-specific NGS assays.

cfDNA concentration was available from 35 participants and for all ten LB test sets of design I and II (Fig. 2). Two participants did not report the concentration, one reported constantly 10 ng/μl for all cases, and one participant used spectroscopic measurement leading to significant higher concentrations of more than 6 ng/μl. These data were excluded from further analysis. The highest concentrations were achieved when using a column-based silica membrane kit (QIAamp circulating nucleic acid kit) for cfDNA extraction with median of 1.22 ng/μl (range 0.25–4.40).

Fig 2figure 2

Range of cfDNA concentrations related to extraction methods used for design I and II. Results include data from 33 participants. Two participants did not report concentrations (design I), one was incorrect and data of spectroscopic measurement was excluded because of high divergence. CNA, circulating nucleic acid

Depending on the method, allelic fractions were reported by 21/24 participants for design I and 11/13 for design II. PIK3CA p.E545K mutation reference with 3% allelic fraction was detected in a range of 1 to 5% (median 2.4%) for design I vs. 1 to 3% (median 2.5%) for design II (Fig. 3). The p.E545K 6% reference was detected in a range from 2 to 11% allelic fraction (median 4.8%) vs. 1 to 6% (median 4.0%) for design I and II, respectively. PIK3CA p.H1047R mutation reference with 3% allelic fraction was detected in a range of 0.3 to 7% (median 3.6%) with an outlier at 32% allele frequency, which was excluded from the analysis. Design II showed a smaller range of 2 to 7% (median 3.9%). The PIK3CA p.H1047R 6% reference was detected in a wide range from 1 to 13% allelic fraction (median 7.0%) for design I vs. 4 to 7% (median 7%) for design II. KRAS p.G12D 20% allelic fraction reference was reported in a much higher range of 3 to 29% (median 17.0%) for design I vs. 1 to 23% (median 9.2%) for design II. A constantly low allelic fraction of 1 to 2% for all ten LB design II samples was reported with tissue-specific OFA (Thermo Fisher Scientific). The highest values around 15% were achieved with LB-specific Oncomine Breast cfDNA assay (Thermo Fisher Scientific).

Fig 3figure 3

Reported allelic fraction for PIK3CA and KRAS mutations in liquid biopsy reference samples design I (white) and design II (grey)

The most commonly used method for determination of PIK3CA mutation status was NGS for both, tissue (34/54 participants) and LB (18/24 participants) analyses (Fig. 4).

Fig 4figure 4

Methods used for mutation detection in external proficiency testing. A Methods used for tissue testing (split 1 and 2) [n = 58]. B Methods used for liquid biopsy analysis (design I and II) [n = 38]; NGS, next generation sequencing; ddPCR, digital droplet PCR

The participants used a broad range of assays for PIK3CA mutation analyses (Fig. 5). Five participants evaluating LB design I who used customized assays, including NGS amplicon-based panel and primer for ddPCR, passed the trial. One participant who did not pass the proficiency test used amplicon-based NGS assays specified for tissue specimens. Two participants successfully applied the clinical trial assay (therascreen) used in the SOLAR-1 trial for molecular stratification of activating PIK3CA mutations. All participants who tested LB design II and used a non-specific LB assay did not pass the trial.

Fig 5figure 5

Most commonly used assays for liquid biopsy mutation testing (design I and II). AST, actionable solid tumor; OBcfDNA, Oncomine breast cfDNA; CHPv2, cancer hotspot panel version 2

留言 (0)

沒有登入
gif