Matthew Mackay, Kabir Manghnani, Adam J Hockenberry, Joshua Drews, James L. Chen, Rotem Ben-Shachar, Justin Guinney
Background: Next-generation sequencing of circulating tumor DNA (ctDNA) and solid-tissue
can identify clinically actionable genomic variants that may be used for both treatment selection
and disease surveillance. Due to differences in tumor biology and assay design, ctDNA and
solid biopsies may identify unique variants. Here, we investigate a real-world dataset of breast
cancer patients to determine whether clinically actionable variant detection is enhanced by dual
ctDNA and solid tissue testing.
Methods: We used the deidentified Tempus Lens database to retrospectively analyze stage IV
breast cancer patients with known hormonal subtype. Each patient had dual testing defined as
Tempus xF (ctDNA) and Tempus xT (tumor tissue)—which resulted in clinical reports for both
tests. Patients were further stratified according to the timing of ctDNA biopsy relative to tissue
biopsy. Concurrent dual testing was defined as samples collected ≤30 days apart and
longitudinal dual testing was defined as liquid >30 days after solid. Variants were included in
analyses if they met the limit of detection criteria of both assays. Clinical actionability was
defined by indication-matched OncoKB Level 1-3. Fisher exact test was used to calculate
Results: Of the 1,341 breast cancer patients with dual ctDNA and tissue sequencing, at least one
actionable variant was identified in 61% (n=823) of patients. In the subset of concurrent tested
patients (n=782), 60% (n=473) had one or more actionable findings: 54% (n=257/473) of
patients with actionable variants had perfectly concordant variants, 29% (n=136/473) had at
least one unique variant detected only by solid tumor testing, and 20% (n=93/473) had at least
one unique variant detected only by ctDNA testing. Similarly, in the longitudinal set (n=559),
63% (n=350) had one or more actionable findings: 34% (n=118/350) were concordant, 43%
(n=150/350) were unique to solid, and 27% (n=96/350) were unique to ctDNA.
When stratifying concurrent patients by OncoKB levels of evidence, 72% (n=98/136) of patients
with variants unique in solid had at least one level 1-2 variant, while 39% (n=53/136) contained
unique level 3 variants. Level 1-2 variants in PIK3CA were the most frequent variants seen
uniquely in solid tumors, occurring in 54% (n=73/136) of patients. In contrast, in patients with
unique ctDNA variants, 37% (n=34/93) of patients had at least one level 1-2 variants and 72%
(n=67/93) had level 3 variants. Level 3 variants in ESR1 were the most frequent variants seen
uniquely in ctDNA, occurring in 57% (n=53/93) of patients. The proportion of concurrent patients
with actionable variants found exclusively in ctDNA significantly differed by subtype (p=0.04):
Luminal A (22%) and Luminal B (23%) contained the most patients with unique ctDNA variants.
This ability to detect additional variants in ctDNA remained true even if profiling occurred over
time. Indeed, in patients with ESR1 variants tested with ctDNA > 1 year after tissue, 78%
(n=43/55) had ESR1 variants only detected in blood.
Conclusions: We show that dual testing in breast cancer patients improves the identification of
clinically actionable variants which may be missed by either ctDNA or solid tissue biopsy alone. Adoption of dual testing should be considered as standard practice to provide a comprehensive
view of actionable molecular alterations.
VIEW THE POSTER