T h e ne w e ngl a nd jou r na l o f m e dicine
`
`original article
`
`Analysis of Circulating Tumor DNA
`to Monitor Metastatic Breast Cancer
`Sarah-Jane Dawson, F.R.A.C.P., Ph.D., Dana W.Y. Tsui, Ph.D.,
`Muhammed Murtaza, M.B., B.S., Heather Biggs, M.A.,
`Oscar M. Rueda, Ph.D., Suet-Feung Chin, Ph.D., Mark J. Dunning, Ph.D.,
`Davina Gale, B.Sc., Tim Forshew, Ph.D., Betania Mahler-Araujo, M.D.,
`Sabrina Rajan, M.D., Sean Humphray, B.Sc., Jennifer Becq, Ph.D.,
`David Halsall, M.R.C.Path., Ph.D., Matthew Wallis, M.B., Ch.B.,
`David Bentley, D.Phil., Carlos Caldas, M.D., F.Med.Sci.,
`and Nitzan Rosenfeld, Ph.D.
`
`Abstr act
`
`Background
`The management of metastatic breast cancer requires monitoring of the tumor
`burden to determine the response to treatment, and improved biomarkers are needed.
`Biomarkers such as cancer antigen 15-3 (CA 15-3) and circulating tumor cells have
`been widely studied. However, circulating cell-free DNA carrying tumor-specific
`alterations (circulating tumor DNA) has not been extensively investigated or com-
`pared with other circulating biomarkers in breast cancer.
`
`Methods
`We compared the radiographic imaging of tumors with the assay of circulating tumor
`DNA, CA 15-3, and circulating tumor cells in 30 women with metastatic breast
`cancer who were receiving systemic therapy. We used targeted or whole-genome
`sequencing to identify somatic genomic alterations and designed personalized assays
`to quantify circulating tumor DNA in serially collected plasma specimens. CA 15-3
`levels and numbers of circulating tumor cells were measured at identical time
`points.
`
`Results
`Circulating tumor DNA was successfully detected in 29 of the 30 women (97%) in
`whom somatic genomic alterations were identified; CA 15-3 and circulating tumor
`cells were detected in 21 of 27 women (78%) and 26 of 30 women (87%), respec-
`tively. Circulating tumor DNA levels showed a greater dynamic range, and greater
`correlation with changes in tumor burden, than did CA 15-3 or circulating tumor cells.
`Among the measures tested, circulating tumor DNA provided the earliest measure
`of treatment response in 10 of 19 women (53%).
`
`Conclusions
`This proof-of-concept analysis showed that circulating tumor DNA is an informa-
`tive, inherently specific, and highly sensitive biomarker of metastatic breast cancer.
`(Funded by Cancer Research UK and others.)
`
`From the Department of Oncology, Univer-
`sity of Cambridge and Cancer Research
`UK Cambridge Institute, Li Ka Shing Cen-
`tre (S.-J.D., D.W.Y.T., M.M., O.M.R., S.-F.C.,
`M.J.D., D.G., T.F., C.C., N.R.), the Depart-
`ments of Histopathology (B.M.-A.), Radi-
`ology (S.R., M.W.), and Clinical Biochem-
`istry and Immunology (D.H.) and the
`Cambridge Breast Unit (S.-J.D., H.B.,
`B.M.-A., S.R., M.W., C.C.), Addenbrooke’s
`Hospital, Cambridge University Hospital
`National Health Service Foundation
`Trust and National Institute for Health
`Research Cambridge Biomedical Research
`Centre, and the Cambridge Experimental
`Cancer Medicine Centre (C.C.), Cambridge;
`and Illumina, Little Chesterford (S.H., J.B.,
`D.B.) — all in the United Kingdom; and
`the Peter MacCallum Cancer Centre, East
`Melbourne, VIC, Australia (S.-J.D.). Ad-
`dress reprint requests to Dr. Rosenfeld or
`Dr. Caldas at Cancer Research UK Cam-
`bridge Institute, University of Cambridge,
`Li Ka Shing Centre, Robinson Way, Cam-
`bridge, CB2 0RE, United Kingdom, or at
`carlos.caldas@cruk.cam.ac.uk.
`
`Drs. Dawson and Tsui and Drs. Caldas
`and Rosenfeld contributed equally to this
`article.
`
`This article was published on March 13,
`2013, at NEJM.org.
`
`NEnglJMed2013;368:1199-209.
`DOI:10.1056/NEJMoa1213261
`Copyright © 2013 Massachusetts Medical Society.
`
`n engl j med 368;13 nejm.org march 28, 2013
`
`
`
`3daws_oa1213261.indd 11993daws_oa1213261.indd 1199
`
`1199
`
`
`
`3/26/13 1:10 PM3/26/13 1:10 PM
`
`The New England Journal of Medicine
`
`Downloaded from nejm.org on January 25, 2022. For personal use only. No other uses without permission.
`
` Copyright © 2013 Massachusetts Medical Society. All rights reserved.
`
`00001
`
`EX1050
`
`

`

`T h e ne w e ngl a nd jou r na l o f m e dicine
`
`Breast cancer is the most common
`
`cancer and the leading cause of cancer-
`related death in women worldwide.1 Meta-
`static breast cancer remains an incurable disease
`but is treatable by means of serial administration
`of endocrine, cytotoxic, or biologic therapies. The
`monitoring of treatment response is essential to
`avoid continuing ineffective therapies, to prevent
`unnecessary side effects, and to determine the
`benefit of new therapeutics. Treatment response is
`generally assessed with the use of serial imaging,
`but radiographic measurements often fail to de-
`tect changes in tumor burden. Therefore, there is
`an urgent need for biomarkers that measure tu-
`mor burden with high sensitivity and specificity.
`Cancer antigen 15-3 (CA 15-3) is a serum bio-
`marker that is clinically useful in some patients
`with metastatic breast cancer but has a sensitivity
`of only 60 to 70%.2-4 The enumeration of circu-
`lating tumor cells has emerged as a promising
`biomarker. Although there are numerous meth-
`ods to detect circulating tumor cells in the re-
`search setting,5-7 the CellSearch System is the
`only test approved by the Food and Drug Admin-
`istration. The system has a sensitivity of ap-
`proximately 65% for detecting circulating tumor
`cells (≥1 cell per 7.5 ml of blood) in patients with
`metastatic breast cancer.8,9 Elevated levels of
`circulating tumor cells (defined as ≥5 cells per
`7.5 ml of blood) have been associated with a
`worse prognosis.8,10
`Circulating DNA fragments carrying tumor-
`specific sequence alterations (circulating tumor
`DNA) are found in the cell-free fraction of blood,
`representing a variable and generally small frac-
`tion of the total circulating DNA.11,12 Advances in
`sequencing technologies have enabled the rapid
`identification of somatic genomic alterations in
`individual tumors, and these can be used to
`design personalized assays for the monitoring of
`circulating tumor DNA. Studies have shown the
`feasibility of using circulating tumor DNA to
`monitor tumor dynamics in a limited number of
`patients with various solid cancers, but few cases
`of breast cancer have been analyzed.13-20 Here,
`we provide a direct comparison between circu-
`lating tumor DNA and other circulating bio-
`markers (CA 15-3 and circulating tumor cells)
`and medical imaging, the current standard of
`care, for the noninvasive monitoring of meta-
`static breast cancer.
`
`Methods
`
`Patients and Sample Collection
`We carried out a prospective, single-center study
`to compare the sensitivity of measuring circulat-
`ing tumor DNA, CA 15-3, and circulating tumor
`cells for monitoring tumor burden in patients with
`metastatic breast cancer (see the Supplementary
`Appendix, available with the full text of this ar-
`ticle at NEJM.org). The study was approved by the
`local institutional research ethics committee.
`Eligible patients were women with metastatic
`breast cancer currently undergoing active treat-
`ment. A total of 52 women were recruited, and
`30 had genomic alterations suitable for monitor-
`ing. All women provided written informed con-
`sent. Serial blood samples (30 ml each) were
`collected between April 2010 and April 2012 at
`intervals of 3 or more weeks. Computed tomog-
`raphy (CT) was performed and reviewed in a
`blinded fashion to document response to treat-
`ment according to the Response Evaluation Cri-
`teria in Solid Tumors (RECIST), version 1.1.21 All
`reagents and equipment used in the study were
`purchased.
`
`Identification of Somatic Genomic
`Alterations
`Sequencing was performed on DNA from breast-
`cancer specimens and matched normal tissue
`specimens, with the use of one or both of two
`methods: tagged-amplicon deep sequencing22
`for PIK3CA (encoding the phosphatidylinositol-
`4,5-bisphosphate 3-kinase, catalytic subunit alpha
`protein) and TP53 (encoding tumor protein p53)
`or paired-end whole-genome sequencing (see the
`Supplementary Appendix). Tagged-amplicon deep
`sequencing was done by means of the Fluidigm
`Access Array and sequencing on the Illumina
`GAIIx or HiSeq instruments. Paired-end se-
`quencing was done with the use of the Illumina
`HiSeq2000 instrument. Candidate mutations and
`structural variants were validated and confirmed
`to be somatic with the use of Sanger sequencing.
`
`Isolation and Quantification of Circulating
`Tumor DNA
`Blood samples that were collected in EDTA tubes
`were processed within 1 hour after collection and
`were centrifuged to separate the plasma from the
`peripheral-blood cells. DNA was extracted from
`
`1200
`
`n engl j med 368;13 nejm.org march 28, 2013
`
`
`
`3daws_oa1213261.indd 12003daws_oa1213261.indd 1200
`
`
`
`3/26/13 1:10 PM3/26/13 1:10 PM
`
`The New England Journal of Medicine
`
`Downloaded from nejm.org on January 25, 2022. For personal use only. No other uses without permission.
`
` Copyright © 2013 Massachusetts Medical Society. All rights reserved.
`
`00002
`
`

`

`Circulating Tumor DNA in Metastatic Breast Cancer
`
`aliquots (2 ml) of plasma with the use of the
`QIAamp circulating nucleic acid kit (Qiagen). To
`measure the DNA carrying specific somatic ge-
`nomic alterations in plasma, we carried out a mi-
`crofluidic digital polymerase-chain-reaction (PCR)
`assay17,23-25 (using the Fluidigm BioMark system)
`or direct plasma sequencing by means of tagged-
`amplicon deep sequencing22 (using the Fluidigm
`Access Array and sequencing on the Illumina
`HiSeq2500 instrument) (see the Supplementary
`Appendix).
`
`Assay of CA 15-3 and Circulating Tumor Cells
`We measured levels of CA 15-3 in aliquots (50 µl)
`of plasma by means of the ADVIA Centaur im-
`munoassay system (Siemens Healthcare). Blood
`samples were collected in CellSave Preservative
`Tubes (Veridex) and were processed within 96
`hours for the enumeration of circulating tumor
`cells with the use of the CellSearch System (Veri-
`dex). The counting of circulating tumor cells was
`performed in a manner blinded to the results of
`CT and assessments of CA 15-3 or circulating
`tumor DNA.
`
`Statistical Analysis
`To estimate the sensitivity of each of the circulat-
`ing biomarkers, we used a modified bootstrap-
`ping method.26 We randomly sampled the com-
`plete data set to obtain a new data set containing
`only one time point for each patient. This ran-
`dom sampling was repeated 1000 times to obtain
`1000 data sets, each containing independent ob-
`servations. For each data set, we calculated the
`sensitivity of each biomarker. The median sensi-
`tivity for each biomarker and the median differ-
`ence in sensitivity between two biomarkers —
`circulating tumor DNA versus either CA 15-3 or
`circulating tumor cells — was then calculated
`across the 1000 data sets. The percentile method
`was used to obtain 95% confidence intervals.
`Survival analysis was performed by fitting a
`different Cox regression model for each of the
`three variables of interest: circulating tumor
`DNA, circulating tumor cells, and CA 15-3. Each
`model was constructed with the use of the count-
`ing process notation (start, end, event),27 such
`that for each time period, the date of the visit
`was taken as the start, and the date before the
`next visit (or the date of last follow-up) was con-
`sidered the end. The predictors were modeled as
`
`time-dependent covariates that use splines to
`account for nonlinear relationships. Estimated
`survival curves were produced for different val-
`ues of the covariates at the first visit. Wald sta-
`tistic P values were reported for each model, and
`relative hazard plots were computed for each co-
`variate, showing the linear predictor relative to
`the mean value of the covariate (for details, see
`the Supplementary Appendix).
`
`R esults
`
`Identification of Somatic Genomic
`Alterations
`Clinical details, results of CT imaging, and serial
`whole-blood samples were collected prospective-
`ly from 52 women undergoing therapy for meta-
`static breast cancer (Fig. 1, and Table S1 in the
`Supplementary Appendix). DNA extracted from
`archival-tumor tissue samples was analyzed to
`identify somatic genomic alterations, with the
`use of two approaches. First, we used targeted
`deep sequencing to screen for point mutations in
`PIK3CA and TP53,28 which we identified in 25 of
`the 52 patients (Table S2 in the Supplementary
`Appendix). Second, we used whole-genome
`paired-end sequencing of tumor-tissue specimens
`and matched normal-tissue specimens in 9 of the
`52 patients. We identified somatic structural
`variants29 in 8 patients (Table S3 in the Supple-
`mentary Appendix), including 5 in whom no mu-
`tations were previously identified in PIK3CA or
`TP53, bringing the total number of patients with
`identified genomic alterations to 30 of 52 women
`(Fig. 1, and Fig. S1 in the Supplementary Appen-
`dix). In 3 patients, both mutations and structural
`variants were identified, enabling us to compare
`and contrast the use of point mutations13 and
`structural variants14,15 for serial monitoring of
`circulating tumor DNA. For 1 patient, we used
`whole-genome paired-end sequencing to identify
`multiple somatic mutations, enabling us to mon-
`itor multiple mutations in parallel in circulating
`tumor DNA (Table S2 in the Supplementary Ap-
`pendix).
`
`Quantification of Circulating Tumor DNA
`in Plasma
`In the 30 women with somatic mutations or
`structural variants, circulating tumor DNA was
`quantified in a total of 141 serial plasma samples
`
`n engl j med 368;13 nejm.org march 28, 2013
`
`1201
`
`
`
`3daws_oa1213261.indd 12013daws_oa1213261.indd 1201
`
`
`
`3/26/13 1:10 PM3/26/13 1:10 PM
`
`The New England Journal of Medicine
`
`Downloaded from nejm.org on January 25, 2022. For personal use only. No other uses without permission.
`
` Copyright © 2013 Massachusetts Medical Society. All rights reserved.
`
`00003
`
`

`

`T h e ne w e ngl a nd jou r na l o f m e dicine
`
`52 Women with metastatic breast cancer
`
`Tumor tissue
`
`Serial computed
`tomography
`
`Serial blood samples
`collected
`
`Identification of somatic
`genomic alterations
`
`Targeted sequencing of
`PIK3CA or TP53 mutations
`in all 52 women
`
`Whole-genome sequencing
`to identify mutations, SVs,
`or both in 9 of 52 women
`
`Serial blood samples
`analyzed
`
`25 Had mutations
`
`9 Had mutations or SVs
`
`30 Had mutations or SVs
`22 Had mutations only
`3 Had both mutations
`and SVs
`5 Had SVs only
`
`141 Samples from
`30 women underwent
`quantification of
`circulating tumor
`DNA
`
`126 Samples from
`30 women underwent
`enumeration of
`circulating tumor cells
`
`114 Samples from
`27 women underwent
`quantification
`of CA 15-3
`
`126 Samples underwent
`comparison of circulating
`tumor DNA vs. circu-
`lating tumor cells
`
`114 Samples underwent
`comparison of circulating
`tumor DNA vs. CA 15-3
`
`Figure1.EnrollmentofPatientsandCollectionofClinicalSamples.
`In the 30 women who were found to have somatic mutations, structural variants (SVs), or both, the genomic altera-
`tions were determined through targeted deep sequencing or whole-genome paired-end sequencing of tumor-tissue
`specimens and matched normal-tissue specimens. CA 15-3 denotes cancer antigen 15-3.
`
`by means of either digital PCR assay or tagged-
`amplicon deep sequencing.
`Digital PCR assay was performed in 97 plasma
`samples from 19 of the 30 patients to track both
`somatic mutations and structural variants. The
`sensitivity of digital PCR assay allowed for the
`detection of a mutant allele fraction of 0.1% or
`more (one mutant molecule in a background of
`1000 wild-type molecules) (Fig. S2 in the Supple-
`mentary Appendix).17 Circulating tumor DNA was
`detected in 18 of the 19 women and in 80 of the
`97 plasma samples (82%) analyzed.
`As a high-throughput alternative to digital
`PCR assay, the remaining 44 plasma samples
`from the remaining 11 patients were analyzed
`with the use of tagged-amplicon deep sequenc-
`ing.22 The sensitivity of tagged-amplicon deep
`sequencing allowed for the detection of a mutant
`
`allele fraction of 0.14% or more with a confidence
`margin of 0.95.22 Using this approach, circulating
`tumor DNA was identified in all 11 patients and
`in 35 of the 44 plasma samples (80%) analyzed.
`In a subset of plasma samples in which circu-
`lating tumor DNA was analyzed by both tech-
`niques, quantification of mutant allele fraction
`by means of either tagged-amplicon deep se-
`quencing or digital PCR assay showed excellent
`agreement (Fig. S3 in the Supplementary Appen-
`dix).22 Taken together, circulating tumor DNA
`was detected in 29 of the 30 women (97%) and
`in 115 of the 141 plasma samples (82%). The
`median quantity of circulating tumor DNA across
`all samples was 150 amplifiable copies per milli-
`liter of plasma (interquartile range, 9 to 720)
`(Table S4 in the Supplementary Appendix). The
`median mutant allele fraction was 4% (interquar-
`
`1202
`
`n engl j med 368;13 nejm.org march 28, 2013
`
`
`
`3daws_oa1213261.indd 12023daws_oa1213261.indd 1202
`
`
`
`3/26/13 1:10 PM3/26/13 1:10 PM
`
`The New England Journal of Medicine
`
`Downloaded from nejm.org on January 25, 2022. For personal use only. No other uses without permission.
`
` Copyright © 2013 Massachusetts Medical Society. All rights reserved.
`
`00004
`
`

`

`Circulating Tumor DNA in Metastatic Breast Cancer
`
`A
`
`Patient6
`
`B
`
`Patient4
`
`Deletion 1
`
`Deletion 2
`
`Deletion 3
`
`PIK3CA
`
`Progressive
`disease
`
`Stable
`disease
`
`Progressive
`disease
`
`PIK3CA
`ZFYVE21
`CD1A
`IQCA1
`MET
`KIAA0406
`
`Epirubicin
`
`Paclitaxel
`
`106
`
`105
`
`104
`
`103
`
`102
`
`101
`
`100
`
`ND
`
`ctDNA(copies/ml)
`
`0
`
`50
`
`100
`
`150
`Days
`
`200
`
`250
`
`300
`
`0
`
`100
`
`200
`
`300
`
`500
`
`600
`
`700
`
`400
`Days
`
`D
`
`Progressive
`disease
`
`Progressive
`disease
`
`Stable
`disease
`
`106
`
`105
`
`104
`
`103
`
`102
`
`101
`
`100
`
`ND
`
`ctDNA(copies/ml)
`
`C
`
`Exemestane
`
`Carbo-
`platin
`
`Paclitaxel
`
`TP53
`
`PIK3CA
`
`Patient16
`Stable
`disease
`
`106
`
`105
`
`104
`
`103
`
`102
`
`101
`
`100
`
`ND
`
`ctDNA(copies/ml)
`
`TP53
`
`PIK3CA
`
`Patient18
`Nonmeasurable disease
`(bone metastases)
`
`Exemes-
`tane
`
`Vinorelbine
`
`106
`
`105
`
`104
`
`103
`
`102
`
`101
`
`100
`
`ND
`
`ctDNA(copies/ml)
`
`0
`
`50
`
`100
`
`150
`
`200
`
`250
`
`0
`
`50
`
`100
`
`Days
`
`200
`
`250
`
`300
`
`150
`Days
`
`Figure2.MonitoringMultiplePointMutationsandStructuralVariantsinCirculatingDNA.
`Panels A, B, and C show plasma levels of circulating tumor DNA (ctDNA) for three patients (one per panel), quantified in parallel by
`means of a digital polymerase-chain-reaction (PCR) assay across multiple time points. In Panels B, C, and D, the use of endocrine or
`cytotoxic therapy is indicated by colored shading, and disease status at various times (as ascertained on computed tomography) is shown.
`Panel A shows three structural variants (deletions) and a point mutation in PIK3CA. The three deletions occurred in the setting of a
`complex rearrangement associated with amplification. Panel B shows six point mutations, all of which showed similar dynamic patterns.
`Panel C shows point mutations in PIK3CA and TP53; the TP53 mutation was dominant in the circulation as compared with the PIK3CA
`mutation. Panel D shows plasma levels of ctDNA for a fourth patient, with point mutations in PIK3CA and TP53 quantified by means of
`tagged-amplicon deep sequencing. The TP53 mutation was identified in plasma only, and levels remained elevated after paclitaxel chemo-
`therapy despite a fall in the PIK3CA mutation level in the presence of stable disease. ND denotes not detected.
`
`tile range, 1 to 14). The 1 patient in whom circu-
`lating tumor DNA was not detected (Patient 12)
`had a low burden of metastatic disease (small-
`volume mediastinal lymphadenopathy) and no
`evidence of disease progression during the study.
`Overall, levels of total plasma DNA were mea-
`sured in parallel and had limited informative
`content (Fig. S4 in the Supplementary Appendix).
`
`Concurrent Monitoring of Multiple Somatic
`genomic Alterations in plasma
`Plasma levels of either mutations or structural
`variants identified in the tumor tissue of the same
`patient (Fig. S1C in the Supplementary Appendix)
`showed a similar dynamic pattern (Fig. 2A, and
`Table S4 in the Supplementary Appendix). This
`confirmed the utility and comparability of both
`
`n engl j med 368;13 nejm.org march 28, 2013
`
`1203
`
`
`
`3daws_oa1213261.indd 12033daws_oa1213261.indd 1203
`
`
`
`3/26/13 1:10 PM3/26/13 1:10 PM
`
`The New England Journal of Medicine
`
`Downloaded from nejm.org on January 25, 2022. For personal use only. No other uses without permission.
`
` Copyright © 2013 Massachusetts Medical Society. All rights reserved.
`
`00005
`
`

`

`T h e ne w e ngl a nd jou r na l o f m e dicine
`
`approaches. In women with tumors in which the
`genomic location of the structural variants over-
`lapped with an amplified locus, such alterations
`were detected in the plasma at higher concentra-
`tions, confirming that the assay of circulating
`tumor DNA is quantitative (Fig. 2A, and Fig. S1B
`and Table S5 in the Supplementary Appendix).
`When multiple mutations were identified in
`tumor-tissue samples (Fig. S1C in the Supplemen-
`tary Appendix), they generally showed similar
`dynamic patterns in plasma (Fig. 2B, and Table
`S4 in the Supplementary Appendix). However, in
`some cases, we also observed evidence of clonal
`heterogeneity, whereby certain mutations domi-
`nated in the plasma (Fig. 2C, and Table S4 in the
`Supplementary Appendix). Tagged-amplicon deep
`sequencing also identified mutations in plasma
`that were not detected in archival-tumor DNA
`(Fig. S1C in the Supplementary Appendix).22 In
`these cases, the archival primary tissue had been
`collected more than 10 years previously, and the
`discordance may have reflected tumor evolu-
`tion.30,31 These mutations showed diverging pat-
`terns over the course of disease progression and
`treatment (Fig. 2D, and Table S4 in the Supple-
`mentary Appendix), as compared with the muta-
`tions identified in the tumor, suggesting that
`they originated from different subclones.
`
`Sensitivity of Circulating Tumor DNa, CA 15-3,
`and Circulating Tumor Cells
`Data comparing CA 15-3 values and circulating
`tumor DNA levels were available across 114 serial
`time points for 27 patients (Fig. 3A, and Table S4
`in the Supplementary Appendix). CA 15-3 levels
`were elevated (>32.4 U per milliliter) at one or
`more time points in 21 of the 27 women (78%)
`and in 71 of the 114 samples (62%). In contrast,
`circulating tumor DNA was detected in 26 of 27
`women (96%) and in 94 of 114 samples (82%). Of
`the 43 samples without elevated CA 15-3 levels,
`27 (63%) had measurable levels of circulating tu-
`mor DNA. Using a modified bootstrapping
`method, we showed improved sensitivity of cir-
`culating tumor DNA as compared with CA 15-3
`(85% vs. 59%), with a median difference in sensi-
`tivity of 26% (95% confidence interval [CI], 11 to
`37; P<0.002).
`Circulating tumor cells were quantified by
`means of the CellSearch System at 126 time points
`for all 30 women (Fig. 3B, and Table S4 in the
`Supplementary Appendix). Circulating tumor cells
`
`(≥1 cell per 7.5 ml of blood) were detected at one
`or more time points in 26 of the 30 women (87%),
`and elevated circulating tumor cells (≥5 cells per
`7.5 ml of blood) were identified in 18 of the 30
`women (60%). Of the 126 samples, 50 (40%) had
`no detected circulating tumor cells, and 76 (60%)
`had 1 or more cells per 7.5 ml, of which 46 (37%
`of all 126 samples) had 5 or more cells per 7.5 ml.
`In contrast, circulating tumor DNA was detected in
`29 of the 30 women (97%) and at 106 of 126 time
`points (84%). In the 50 samples in which no cir-
`culating tumor cells were detected, 33 (66%) had
`measurable levels of circulating tumor DNA. Ac-
`cording to the modified bootstrapping method,
`circulating tumor DNA had sensitivity superior to
`that of circulating tumor cells (90% vs. 67%), with
`a median difference in sensitivity of 27% (95% CI,
`13 to 37; P<0.002). At the median, the number of
`amplifiable copies of circulating tumor DNA was
`133 times the number of circulating tumor cells
`and had a greater dynamic range (Fig. 3B).
`
`CT and Circulating Biomarkers for Tumor
`Monitoring
`We compared the performance of circulating bio-
`markers with the performance of CT in 20 pa-
`tients with measurable disease (as defined by
`RECIST21) and for whom circulating biomarker
`data were available at 3 or more time points over
`a period of more than 100 days of follow-up (Fig.
`S5 in the Supplementary Appendix). Circulating
`tumor DNA was detected and showed serial
`changes in 19 of 20 women (95%) with fluctua-
`tions in circulating tumor DNA generally corre-
`lating with treatment responses seen on imaging
`(Fig. 4A, and Fig. S5 in the Supplementary Ap-
`pendix). Similar findings were noted for women
`with 5 or more circulating tumor cells per 7.5 ml
`of blood (10 of 20 patients [50%]) in which serial
`changes in circulating tumor cell counts were
`evident and corresponded with responses ascer-
`tained on CT (Fig. 4A). However, in the remain-
`ing 10 women with a maximal count of circulat-
`ing tumor cells of fewer than 5 cells per 7.5 ml of
`blood, the number of circulating tumor cells was
`uninformative (Fig. 4B and 4C, and Fig. S5 in the
`Supplementary Appendix).
`Similar to the findings regarding circulating
`tumor cells was the finding that women with
`high levels of CA 15-3 had fluctuations corre-
`sponding to responses on imaging but with a
`smaller dynamic range (Fig. 4A and 4B, and Fig.
`
`1204
`
`n engl j med 368;13 nejm.org march 28, 2013
`
`
`
`3daws_oa1213261.indd 12043daws_oa1213261.indd 1204
`
`
`
`3/26/13 1:10 PM3/26/13 1:10 PM
`
`The New England Journal of Medicine
`
`Downloaded from nejm.org on January 25, 2022. For personal use only. No other uses without permission.
`
` Copyright © 2013 Massachusetts Medical Society. All rights reserved.
`
`00006
`
`

`

`Circulating Tumor DNA in Metastatic Breast Cancer
`
`r2=0.36
`P<0.001
`
`104
`
`103
`
`102
`
`101
`
`ND
`
`CA15-3(U/ml)
`
`Total
`
`21
`6
`27
`
`71
`43
`114
`
`ND
`
`100 101 102 103 104 105
`ctDNA(copies/ml)
`
`106
`
`A CA15-3vs.ctDNA
`CA15-3
`
`ctDNA
`Detected Not detected
`
`011
`
`4
`16
`20
`
`Patients
`Elevated (>32.4 U/ml)
`Not elevated (≤32.4 U/ml)
`Total
`ctDNA sensitivity, 26/27 (96%)
`CA 15-3 sensitivity, 21/27 (78%)
`Samples
`Elevated (>32.4 U/ml)
`Not elevated (≤32.4 U/ml)
`Total
`ctDNA sensitivity, 94/114 (82%)
`CA 15-3 sensitivity, 71/114 (62%)
`
`21
`5
`26
`
`67
`27
`94
`
`B CTCvs.ctDNA
`CTC
`
`ctDNA
`Detected Not detected
`
`Total
`
`Patients
`Elevated (≥5)
`Detected (1–4)
`Not detected (0)
`Total
`ctDNA sensitivity, 29/30 (97%)
`CTC sensitivity (detected, >0), 26/30 (87%)
`CTC sensitivity (elevated, ≥5), 18/30 (60%)
`Samples
`Elevated (≥5)
`Detected (1–4)
`Not detected (0)
`Total
`ctDNA sensitivity, 106/126 (84%)
`CTC sensitivity (detected, >0), 76/126 (60%)
`CTC sensitivity (elevated, ≥5), 46/126 (37%)
`Median ratio of ctDNA copy numbers (per 3.75 ml of plasma)
`to number of CTCs (per 7.5 ml of whole blood)=133 (interquartile
`range, 51–528)
`
`18
`
`84
`
`30
`
`r2=0.61
`P<0.001
`
`100 101 102 103 104 105
`ND
`ctDNA(copies/3.75mlofplasma)
`
`106
`
`104
`
`103
`
`102
`
`101
`
`100
`
`ND
`
`No.ofCTCs(per7.5mlofblood)
`
`46
`30
`50
`126
`
`0101
`
`12
`
`17
`20
`
`18
`
`74
`
`29
`
`45
`28
`33
`106
`
`Figure3.ComparisonofCirculatingTumorDNA,CA15-3,andCirculatingTumorCellsasBlood-BasedBiomarkers.
`Panel A shows comparisons of CA 15-3 levels (U per milliliter of plasma) and circulating tumor DNA (ctDNA) levels
`(amplifiable copies per milliliter of plasma) across the maximal value analyzed for individual patients and across all
`samples analyzed for all patients. The green horizontal dashed line indicates the CA 15-3 threshold of 32.4 U per
`milliliter. The Spearman correlation coefficient (r) between CA 15-3 levels and ctDNA levels across all time points
`was 0.36 (P<0.001). Panel B shows comparisons of circulating tumor cell (CTC) numbers (per 7.5 ml of whole blood)
`and ctDNA numbers (amplifiable copies per 3.75 ml of plasma) across the maximal value analyzed for individual pa-
`tients and across all samples analyzed for all patients. Copy numbers of ctDNA were adjusted for direct comparison
`to the numbers of circulating tumor cells from an equivalent volume of whole blood (7.5 ml). The purple dashed line
`indicates the CTC threshold of 1 cell per 7.5 ml of blood, and the orange dashed line indicates the CTC threshold of
`5 cells per 7.5 ml of blood. The Spearman correlation coefficient (r) between quantified ctDNA levels and numbers
`of CTCs across all time points was 0.61 (P<0.001). ND denotes not detected.
`
`S5 in the Supplementary Appendix). In patients
`with levels of CA 15-3 of 50 U or less per milli-
`liter (8 of 19 patients [42%]), no consistent se-
`rial changes in CA 15-3 levels were seen (Fig. 4C,
`and Fig. S5 in the Supplementary Appendix).
`Progressive disease was documented on CT
`(as defined by RECIST) in 19 of 20 women during
`the follow-up period; CA 15-3 data were avail-
`
`able for 18 of these women (95%) (Fig. S5 in the
`Supplementary Appendix). Increases in circulat-
`ing tumor DNA levels reflected progressive dis-
`ease in 17 of the 19 women (89%). In these
`women, on average, circulating tumor DNA levels
`increased by a factor of 505 (range, 2 to 4457)
`from the nadir before the establishment of pro-
`gressive disease. The numbers of circulating tu-
`
`n engl j med 368;13 nejm.org march 28, 2013
`
`1205
`
`
`
`3daws_oa1213261.indd 12053daws_oa1213261.indd 1205
`
`
`
`3/26/13 1:10 PM3/26/13 1:10 PM
`
`The New England Journal of Medicine
`
`Downloaded from nejm.org on January 25, 2022. For personal use only. No other uses without permission.
`
` Copyright © 2013 Massachusetts Medical Society. All rights reserved.
`
`00007
`
`

`

`T h e ne w e ngl a nd jou r na l o f m e dicine
`
`ctDNA
`CA 15-3
`CTC
`
`A Patient17
`
`B Patient8
`
`SD
`
`PD
`
`PD
`
`Vinorelbine
`
`Epiru-
`bicin
`
`Carbo-
`platin
`
`106
`105
`104
`103
`102
`101
`100
`ND
`
`CirculatingBiomarker
`
`ctDNA
`CA 15-3
`CTC
`
`PD
`
`PR PD
`
`SD
`
`PD
`
`Cape-
`cita-
`bine
`
`Vino-
`relbine
`
`Epiru-
`bicin
`
`Epiru-
`bicin
`
`106
`105
`104
`103
`102
`101
`100
`ND
`
`CirculatingBiomarker
`
`0
`
`100
`
`200
`Days
`
`300
`
`400
`
`500
`
`0
`
`50
`
`100
`Days
`
`150
`
`200
`
`250
`
`C Patient20
`
`D Patient6
`
`PR
`
`SD
`
`SD
`
`PD
`
`SD
`
`Epirubicin
`
`Letro-
`zole
`
`Epiru-
`bicin
`
`SD
`
`SD
`
`SD
`
`SD
`
`PD
`
`Capecitabine
`
`Carbo-
`platin
`
`106
`
`105
`
`104
`
`103
`
`102
`
`101
`
`ctDNA
`CA 15-3
`CTC
`
`120
`
`ctDNA
`
`CTC
`
`100
`
`ND
`
`CirculatingBiomarker
`
`ctDNA
`CA 15-3
`CTC
`
`300
`
`400
`
`0
`
`100
`
`200
`
`300
`
`Days
`
`106
`105
`104
`103
`102
`101
`100
`ND
`
`CirculatingBiomarker
`
`0
`
`100
`
`200
`Days
`
`E QuantilesofctDNAandOverallSurvival
`
`F ctDNA,CTCs,andRelativeHazard
`CTCs(per7.5mlofwholeblood)
`40
`60
`80
`100
`
`20
`
`0
`
`3
`
`2
`
`1
`
`0
`
`−1
`
`0
`
`LogeRelativeHazard
`
`25% quant. ctDNA (10.35)
`
`50% quant. ctDNA (150.1)
`
`75% quant. ctDNA (710.11)
`
`1.0
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`ProbabilityofSurvival
`
`P<0.001
`
`0.0
`
`0
`
`200
`
`95% quant. ctDNA (8016.3)
`
`400
`Days
`
`600
`
`2000
`4000
`6000
`ctDNA(amplifiablecopiespermlofplasma)
`
`8000
`
`mor cells increased in 7 of the 19 women (37%),
`and CA 15-3 levels increased in 9 of 18 women
`(50%) (Fig. S5 in the Supplementary Appendix). In
`10 of the 19 patients (53%), levels of circulating
`tumor DNA increased at one or more consecutive
`time points, on average 5 months (range, 2 to 9)
`
`before the establishment of progressive disease
`by means of imaging (Fig. 4D, and Fig. S5 in the
`Supplementary Appendix). In 2 women (Patients
`9 and 22), increasing levels of circulating tumor
`DNA did not reflect the presence of progressive
`disease as assessed on CT (a detailed description
`
`1206
`
`n engl j med 368;13 nejm.org march 28, 2013
`
`
`
`3daws_oa1213261.indd 12063daws_oa1213261.indd 1206
`
`
`
`3/26/13 1:10 PM3/26/13 1:10 PM
`
`The New England Journal of Medicine
`
`Downloaded from nejm.org on January 25, 2022. For personal use only. No other uses without permission.
`
` Copyright © 2013 Massachusetts Medical Society. All rights reserved.
`
`00008
`
`

`

`Circulating Tumor DNA in Metastatic Breast Cancer
`
`Figure4(facingpage).ComparisonofCirculating
`BiomarkerstoMonitorTumorDynamicsandPredict
`Survival.
`Panels A, B, C, and D show serial circulating tumor
`DNA (ctDNA) levels (number of copies per milliliter
`of plasma), circulating tumor cell (CTC) numbers (per
`7.5 ml of whole blood), CA 15-3 levels (U per milliliter),
`and disease status as ascertained on computed tomog-
`raphy (vertical dashed lines) for four patients (one in
`each panel). Details of endocrine or cytotoxic therapy
`are indicated by colored shading. The orange dashed line
`indicates the threshold of 5 CTCs per 7.5 ml of whole
`blood. The green dashed line indicates the CA 15-3
`threshold of 32.4 U per milliliter. ND denotes not de-
`tected, PD progressive disease, PR partial response,
`and SD stable disease. Panel E shows the results of a
`Cox regression model, which identified an inverse rela-
`tionship between quantiles (quant.) of ctDNA (indicated
`in copies per milliliter of plasma) and overall survival,
`with increa