Circulating Tumor Cells in Breast Cancer Patients Treated by Neoadjuvant Chemotherapy: A Meta-analysis

Circulating Tumor Cells in Breast Cancer Patients Treated by Neoadjuvant Chemotherapy: A... Abstract Background We conducted a meta-analysis in nonmetastatic breast cancer patients treated by neoadjuvant chemotherapy (NCT) to assess the clinical validity of circulating tumor cell (CTC) detection as a prognostic marker. Methods We collected individual patient data from 21 studies in which CTC detection by CellSearch was performed in early breast cancer patients treated with NCT. The primary end point was overall survival, analyzed according to CTC detection, using Cox regression models stratified by study. Secondary end points included distant disease–free survival, locoregional relapse–free interval, and pathological complete response. All statistical tests were two-sided. Results Data from patients were collected before NCT (n = 1574) and before surgery (n = 1200). CTC detection revealed one or more CTCs in 25.2% of patients before NCT; this was associated with tumor size (P < .001). The number of CTCs detected had a detrimental and decremental impact on overall survival (P < .001), distant disease–free survival (P < .001), and locoregional relapse–free interval (P < .001), but not on pathological complete response. Patients with one, two, three to four, and five or more CTCs before NCT displayed hazard ratios of death of 1.09 (95% confidence interval [CI] = 0.65 to 1.69), 2.63 (95% CI = 1.42 to 4.54), 3.83 (95% CI = 2.08 to 6.66), and 6.25 (95% CI = 4.34 to 9.09), respectively. In 861 patients with full data available, adding CTC detection before NCT increased the prognostic ability of multivariable prognostic models for overall survival (P < .001), distant disease–free survival (P < .001), and locoregional relapse–free interval (P = .008). Conclusions CTC count is an independent and quantitative prognostic factor in early breast cancer patients treated by NCT. It complements current prognostic models based on tumor characteristics and response to therapy. Neoadjuvant chemotherapy (NCT) is a standard treatment for patients with large, nonmetastatic breast cancer and may allow breast-conserving surgery after tumor downsizing while decreasing the risk of subsequent relapse (1). In addition to these prominent clinical benefits, NCT provides a unique opportunity to study and quantify the cancer response to antineoplastic agents by assessment of the pathological response on the residual tumor at the time of surgery. Pathological complete response (pCR) rate has been thoroughly investigated as a potential surrogate of post-NCT survival, with heterogeneous results according to the various breast cancer subtypes (2,3). In this context, improving post-NCT survival prediction may lead to better evaluation of NCT efficacy and help to optimize adjuvant therapy. Neoadjuvant therapy has two main modes of action, that is, shrinkage of the primary tumor and eradication of blood-borne tumor cell dissemination (4). Surprisingly, the second mode is totally ignored by current NCT assessments, despite the fact that cancer metastasis is the main cause of death in breast cancer and other solid tumors. The precise enumeration of circulating tumor cells (CTCs) in the peripheral blood of cancer patients as surrogate of the dissemination process is now possible after years of international standardization (5–7) and has demonstrated its clinical validity at metastatic stage in many cancer types (8–10). Moreover, experimental studies have shown that CTCs can induce tumors after xenografting into immunodeficient mice (11–13). Here, we report the results of a meta-analysis bringing together individual patient data from the United States, Japan, and European countries. This large-scale multicenter meta-analysis investigated in nonmetastatic breast cancer patients treated by NCT whether CTC counts add statistically significant prognostic information on postneoadjuvant survival to the established assessment of primary tumor characteristics and pathological complete response. Methods Literature Review and Study Identification The study protocol was set up by the study secretariat, and a search of Medline and major oncology congress abstracts was performed in February 2015 to find eligible studies (Supplementary Methods 1, available online). Direct contact was then established with all CTC analysis centers and cooperative research groups deemed to have eligible data. Inclusion criteria were published and unpublished studies conducted in nonmetastatic breast cancer patients treated by NCT; studies in which CTC count was performed at least once before surgery by the CellSearch system; studies with available survival data; studies declared and approved by an appropriate ethics committee; and patient accrual in or after January 2003 and before September 2015. Studies in which CTC count impacted patient management were not eligible. Data Collection Each local investigator was responsible for collecting after informed consent and sharing individual anonymized data, which were centralized until April 28, 2016. The list of collected data is detailed in the study protocol (Supplementary Methods, available online). To take into account the heterogeneity between studies in terms of the number and timing of blood sampling and the duration of NCT, the protocol distinguished four different time points for CTC collection: five or fewer weeks prior to NCT initiation; one to eight weeks after NCT initiation; five or fewer weeks prior to surgery; one to 52 weeks after surgery. Data files were monitored manually for eligibility, and queries were sent to centers whenever necessary. Data were then merged into a centralized repository accessible only to the study secretariat and statisticians. No financial compensation was provided to individuals or participating centers. As per French law, this in silico study of anonymized data was declared to the Commission Nationale de l’Informatique et des Libertés (French data protection authority; CNIL number 1873422v0). Statistical Analysis The meta-analytical fixed effect model for time-to-event metadata used to obtain an overall hazard ratio was a Cox model stratified by study (14,15). The primary end point of this study was overall survival, defined as the time from NCT initiation to death from any cause. Distant disease–free survival was defined as the time from NCT initiation to distant recurrence or death from any cause, whichever came first. Locoregional relapse–free interval was defined as the time from NCT initiation to locoregional relapse; patients who died or patients with no documented evidence of locoregional relapse were censored. We explored the heterogeneity in regards to CTC count distribution across centers (Supplementary Methods, available online) in addition to the heterogeneity in regards to CTC detection on survival. Between-study heterogeneity in regards to CTC detection on survival was assessed using chi-square and I2 statistics. Results were obtained based on a Cox regression model, stratified by study, with the following covariates: CTC (using the two or more CTCs threshold) and interaction between CTC and study. I2 represents the proportion of total variation in study estimates that is due to heterogeneity rather than sampling error. I2 values vary between 0% and 100%; I2 values of less than 30% correspond to low heterogeneity. Fisher exact tests with Monte Carlo approximation and Kruskal-Wallis tests were used to investigate associations of population characteristics with CTC count. The landmark method was used to assess the prognostic effects of the last CTC count (before surgery) and pathological complete response, with the landmark set at surgery; in these analyses conducted in patients with no event prior to surgery, the abovementioned survival times were calculated from the date of surgery. The Kaplan-Meier method was used to estimate survival curves. Cubic restricted splines of 2 degrees of freedom were used to model potential nonlinear effects of continuous CTC in the Cox regression model. To assess the added value of CTC count to clinicopathological variables, likelihood ratio (LR) statistics in Cox regression models stratified by study were used to estimate the added value of CTCs to clinicopathological models, which included a list of prespecified variables known to be prognostic in this setting (age <50 vs >50 years, tumor size T1–2, T3–4, T4d, lymph node status cN0 vs cN1–3, tumor grade 1–2 vs 3, tumor subgroup hormone receptor–positive/human epidermal growth factor receptor–negative [HER2-], HER2+, triple-negative). To control for overfitting of clinicopathological models and to allow unbiased estimates of the added value of CTCs, the data set was randomly divided 500 times into a training and validation series using a previously described method (4). Clinicopathological models were fitted to the training series, and we calculated the average increases in LR statistic (χ² LR value and associated P value) and concordance index on the validation series with 95% confidence intervals (CIs) based on the percentiles of the 500 resamples. Logistic regression models were used to investigate associations with pCR status. P values of less than .05 were considered statistically significant, and all statistical tests were two-sided. SAS (version 9.3) and R (version 3.3.2) were used for statistical analyses. Results Included Data Data for 2156 patients from 16 centers and 21 studies were collected; this report focuses on 2030 patients with CTC detection before NCT and/or before surgery (Figure 1; Supplementary Figure 1 and Supplementary Table 1, available online). The blood volume screened for CTC was 7.5 mL for 1670 patients (77.5%, 19 studies), 15 mL for 460 patients (21.3%, one study: GueparQuinto), and 30 mL for 26 patients (1.2%, one study: UMMC2003-045). We took into account the absolute number of CTCs detected, as 1) there were no statistically significant differences in CTC detection rate between patients with one blood tube and those with more than one blood tube screened at each time point; and 2) as reported beyond, only modest evidence of heterogeneity was found in this meta-analysis. Figure 1. View largeDownload slide Study flow chart. Overall, 2030 patients had circulating tumor cells (CTCs) detected prior to neoadjuvant chemotherapy (NCT) start and/or prior to surgery. Because of the lower number of collected data, results with CTC detection at one to eight weeks after NCT start are not reported here. CTC = circulating tumor cells; DDFS = distant disease–free survival; LRFI = locoregional relapse–free interval; NCT = neoadjuvant chemotherapy; OS = overall survival. Figure 1. View largeDownload slide Study flow chart. Overall, 2030 patients had circulating tumor cells (CTCs) detected prior to neoadjuvant chemotherapy (NCT) start and/or prior to surgery. Because of the lower number of collected data, results with CTC detection at one to eight weeks after NCT start are not reported here. CTC = circulating tumor cells; DDFS = distant disease–free survival; LRFI = locoregional relapse–free interval; NCT = neoadjuvant chemotherapy; OS = overall survival. CTC Detection and Clinicopathological Characteristics Among the 1574 patients with CTC count available before NCT, CTC count ranged from 0 to 559 (median = 0 CTCs; third quartile = 1 CTC): 398 (25.2%), 199 (12.6%), and 93 (5.9%) patients had one or more CTCs, two or more CTCs, and five or more CTCs detected, respectively (Supplementary Tables 2–4, available online). Patient characteristics and their association with CTC detection before NCT are shown in Table 1 (and in Supplementary Tables 2–5, available online). A prominent association was observed between CTC count and tumor size before NCT (P < .001), as higher detection rates were observed in inflammatory (cT4d) breast cancers. This association was expected, and the study protocol was designed to distinguish inflammatory from noninflammatory breast cancers. After excluding T4d cancers from the analysis, no statistically significant associations were observed between CTC detection rates and baseline characteristics (with the exception of tumor size and five or more CTCs, P = .049) (Supplementary Tables 6–9, available online). In-depth modeling of CTC count distribution supported the homogeneity assumption across centers (Supplementary Figures 2 and 3, available online). Table 1. Patient characteristics and circulating tumor cell detection before neoadjuvant chemotherapy Characteristics No. of patients (%) ≥1 CTCs threshold Continuous CTC count P 0 CTCs No. (%) ≥1 CTCs No. (%) P Age, y 1574 (100) .24* .19†  ≤50 843 (53.6) 746 (88.5) 97 (11.5)  >50 731 (46.4) 629 (86.0) 102 (14.0) Tumor size 1547 (100) <.001‡ <.001†  cT1 122 (7.9) 99 (81.1) 23 (18.9)  cT2 770 (49.8) 598 (77.7) 172 (22.3)  cT3 343 (22.2) 260 (75.8) 83 (24.2)  cT4a–c 108 (7.0) 77 (71.3) 31 (28.7)  cT4d 204 (13.2) 120 (58.8) 84 (41.2) Lymph nodes 1567 (100) .05* .02†  cN0 656 (41.9) 507 (87.3) 149 (22.7)  cN1–3 911 (58.1) 664 (72.9) 247 (27.1) Tumor grade 1371 (100) .51* .43†  I 65 (4.7) 52 (80.0) 13 (20.0)  II 702 (51.2) 514 (73.2) 188 (26.8)  III 604 (44.1) 443 (73.3) 161 (26.7) Tumor subgroup 1570 (100) .23* .12†  HER2+ 365 (23.2) 277 (75.9) 88 (24.1)  HR+ HER2- 800 (51.0) 607 (75.9) 193 (24.1)  HR- HER2- 405 (25.8) 290 (71.6) 115 (28.4) Characteristics No. of patients (%) ≥1 CTCs threshold Continuous CTC count P 0 CTCs No. (%) ≥1 CTCs No. (%) P Age, y 1574 (100) .24* .19†  ≤50 843 (53.6) 746 (88.5) 97 (11.5)  >50 731 (46.4) 629 (86.0) 102 (14.0) Tumor size 1547 (100) <.001‡ <.001†  cT1 122 (7.9) 99 (81.1) 23 (18.9)  cT2 770 (49.8) 598 (77.7) 172 (22.3)  cT3 343 (22.2) 260 (75.8) 83 (24.2)  cT4a–c 108 (7.0) 77 (71.3) 31 (28.7)  cT4d 204 (13.2) 120 (58.8) 84 (41.2) Lymph nodes 1567 (100) .05* .02†  cN0 656 (41.9) 507 (87.3) 149 (22.7)  cN1–3 911 (58.1) 664 (72.9) 247 (27.1) Tumor grade 1371 (100) .51* .43†  I 65 (4.7) 52 (80.0) 13 (20.0)  II 702 (51.2) 514 (73.2) 188 (26.8)  III 604 (44.1) 443 (73.3) 161 (26.7) Tumor subgroup 1570 (100) .23* .12†  HER2+ 365 (23.2) 277 (75.9) 88 (24.1)  HR+ HER2- 800 (51.0) 607 (75.9) 193 (24.1)  HR- HER2- 405 (25.8) 290 (71.6) 115 (28.4) * P values were calculated using a two-sided Fisher exact test. CTC = circulating tumor cell; HER2 = human epidermal growth factor receptor 2; HR+ = hormone receptor positive; HR- = hormone receptor negative. † P values were calculated using a two-sided Kruskal-Wallis test. ‡ P values were calculated using a two-sided Fisher exact test with Monte Carlo sampling. Table 1. Patient characteristics and circulating tumor cell detection before neoadjuvant chemotherapy Characteristics No. of patients (%) ≥1 CTCs threshold Continuous CTC count P 0 CTCs No. (%) ≥1 CTCs No. (%) P Age, y 1574 (100) .24* .19†  ≤50 843 (53.6) 746 (88.5) 97 (11.5)  >50 731 (46.4) 629 (86.0) 102 (14.0) Tumor size 1547 (100) <.001‡ <.001†  cT1 122 (7.9) 99 (81.1) 23 (18.9)  cT2 770 (49.8) 598 (77.7) 172 (22.3)  cT3 343 (22.2) 260 (75.8) 83 (24.2)  cT4a–c 108 (7.0) 77 (71.3) 31 (28.7)  cT4d 204 (13.2) 120 (58.8) 84 (41.2) Lymph nodes 1567 (100) .05* .02†  cN0 656 (41.9) 507 (87.3) 149 (22.7)  cN1–3 911 (58.1) 664 (72.9) 247 (27.1) Tumor grade 1371 (100) .51* .43†  I 65 (4.7) 52 (80.0) 13 (20.0)  II 702 (51.2) 514 (73.2) 188 (26.8)  III 604 (44.1) 443 (73.3) 161 (26.7) Tumor subgroup 1570 (100) .23* .12†  HER2+ 365 (23.2) 277 (75.9) 88 (24.1)  HR+ HER2- 800 (51.0) 607 (75.9) 193 (24.1)  HR- HER2- 405 (25.8) 290 (71.6) 115 (28.4) Characteristics No. of patients (%) ≥1 CTCs threshold Continuous CTC count P 0 CTCs No. (%) ≥1 CTCs No. (%) P Age, y 1574 (100) .24* .19†  ≤50 843 (53.6) 746 (88.5) 97 (11.5)  >50 731 (46.4) 629 (86.0) 102 (14.0) Tumor size 1547 (100) <.001‡ <.001†  cT1 122 (7.9) 99 (81.1) 23 (18.9)  cT2 770 (49.8) 598 (77.7) 172 (22.3)  cT3 343 (22.2) 260 (75.8) 83 (24.2)  cT4a–c 108 (7.0) 77 (71.3) 31 (28.7)  cT4d 204 (13.2) 120 (58.8) 84 (41.2) Lymph nodes 1567 (100) .05* .02†  cN0 656 (41.9) 507 (87.3) 149 (22.7)  cN1–3 911 (58.1) 664 (72.9) 247 (27.1) Tumor grade 1371 (100) .51* .43†  I 65 (4.7) 52 (80.0) 13 (20.0)  II 702 (51.2) 514 (73.2) 188 (26.8)  III 604 (44.1) 443 (73.3) 161 (26.7) Tumor subgroup 1570 (100) .23* .12†  HER2+ 365 (23.2) 277 (75.9) 88 (24.1)  HR+ HER2- 800 (51.0) 607 (75.9) 193 (24.1)  HR- HER2- 405 (25.8) 290 (71.6) 115 (28.4) * P values were calculated using a two-sided Fisher exact test. CTC = circulating tumor cell; HER2 = human epidermal growth factor receptor 2; HR+ = hormone receptor positive; HR- = hormone receptor negative. † P values were calculated using a two-sided Kruskal-Wallis test. ‡ P values were calculated using a two-sided Fisher exact test with Monte Carlo sampling. Among 1200 patients with CTC count available before surgery, CTC count was lower than the baseline count (P < .001) and ranged from 0 to 51 (median = 0 CTCs; third quartile = 0 CTCs): 181 (15.1%), 64 (5.3%), and 12 (1.0%) patients had one or more CTCs, two or more CTCs, and five or more CTCs detected, respectively (Supplementary Tables 10–12, available online). No statistically significant association was observed between CTC detection before surgery and patient characteristics (Supplementary Table 10–13, available online). CTC Detection and Pathological Complete Response Patients achieving a pathological complete response (defined as ypT0/isN0 in >90% of patients) (Supplementary Table 1, available online) had a better overall survival (multivariable analysis; hazard ratio [HR] = 0.19, 95% CI = 0.10 to 0.31, P < .001), distant metastasis–free survival (HR = 0.25, 95% CI = 0.16 to 0.38, P < .001), and locoregional relapse–free interval (HR = 0.49, 95% CI = 0.29 to 0.80, P = .004). In the population of patients without inflammatory cancers, a slightly higher rate of pathological complete response was observed in patients with no CTC compared with patients with one or more CTCs before NCT (24.2%, 95% CI = 21.7 to 26.9 vs 17.4%, 95% CI = 13.3 to 22.2, P = .01) (Supplementary Table 14, available online). However, this association between pathological complete response and CTC detection was not observed at other cutoffs and time points (Supplementary Tables 14–17, available online). Multivariable logistic regression models did not find any statistically significant impact on pathological complete response of CTC detection before NCT or before surgery (data not shown). CTC Detection and Postneoadjuvant Outcome With a median follow-up of 59.1 (range = 0.9–125.7, interquartile range = 39.3–70.2) months, 301, 418, and 157 overall survival, distant disease–free survival, and locoregional relapse–free interval events were observed, respectively. CTC count before NCT was a statistically significant prognostic factor in univariate analyses for overall survival (P < .001), distant disease–free survival (P < .001), and locoregional relapse–free interval (P < .001) (Figure 2, A–C). Patients with one, two, three to four, and five or more CTCs before neoadjuvant chemotherapy displayed hazard ratios of death (95% CI) of 1.09 (0.65 to 1.69), 2.63 (1.42 to 4.54), 3.83 (2.08 to 6.66), and 6.25 (4.34 to 9.09), respectively (Supplementary Table 18, available online). Hazard ratios for distant disease–free survival and locoregional relapse–free interval are displayed in Supplementary Table 18 (available online). The observed prognostic impact of CTC count did not change after removing T4d tumors from the analyses (Supplementary Table 19, available online). Using the fewer than two vs two or more CTCs categorization, the overall pooled estimate of having two or more CTCs vs fewer than two CTCs was given by a hazard ratio (95% CI) of 4.36 (3.19 to 5.90) for overall survival, 3.89 (3.00 to 5.03) for distant distant–free survival, and 3.30 (2.12 to 5.03) for locoregional relapse–free interval (Supplementary Table 18, available online). CTC count before surgery was also a statistically significant prognostic factor for overall survival and distant disease–free survival (Figure 2, D–F; Supplementary Tables 20 and 21, available online). There was only modest evidence for between-study heterogeneity of the prognostic effect of two or more CTCs on overall (I2 = 0%), distant disease–free (I2 = 0%), and locoregional relapse–free interval (I2 = 34.1%) in the whole patient population (Supplementary Table 22, available online). Multivariable analyses further confirmed CTC count before NCT as an independent prognostic factor for survival (Table 2; Supplementary Tables 23 and 24, available online). Table 2. Multivariable survival analyses* Variable OS DDFS LRFI HR (95% CI) P† HR (95% CI) P† HR (95% CI) P† Age, y .04 .004 .08  ≤50 0.72 (0.53 to 0.99) 0.69 (0.53 to 0.89) 0.70 (0.47 to 1.04)  >50 1.00 (reference) 1.00 (reference) 1.00 (reference) Tumor size .001 .0003 .48  T1–T2 1.00 (reference) 1.00 (reference) 1.00 (reference)  T3–T4 1.60 (1.11 to 2.31) 1.56 (1.15 to 2.12) 1.16 (0.70 to 1.89)  T4d 2.85 (1.51 to 5.11) 2.65 (1.54 to 4.37) 1.69 (0.66 to 3.74) Lymph nodes <.001 <.001 .02  cN0 1.00 (reference) 1.00 (reference) 1.00 (reference)  cN1–3 1.94 (1.39 to 2.76) 1.63 (1.07 to 2.54) 1.63 (1.07 to 2.54) Tumor grade .72 .94 .05  1–2 1.00 (reference) 1.00 (reference) 1.00 (reference)  3 1.06 (0.76 to 1.50) 0.99 (0.75 to 1.31) 1.53 (0.99 to 2.38) Tumor subgroup <.001 <.001 <.001  HR+ HER2- 1.00 (reference) 1.00 (reference) 1.00 (reference)  HER2+ 1.29 (0.80 to 2.04) 1.05 (0.72 to 1.51) 2.31 (1.28 to 4.19)  HR- HER2- 3.92 (2.65 to 5.84) 2.14 (1.55 to 2.96) 4.35 (2.51 to 7.69) CTC count before NCT <.001 <.001 <.001  0–1 1.00 (reference) 1.00 (reference) 1.00 (reference)  ≥2 3.93 (2.81 to 5.45) 3.73 (2.82 to 4.90) 3.02 (1.88 to 4.75) Variable OS DDFS LRFI HR (95% CI) P† HR (95% CI) P† HR (95% CI) P† Age, y .04 .004 .08  ≤50 0.72 (0.53 to 0.99) 0.69 (0.53 to 0.89) 0.70 (0.47 to 1.04)  >50 1.00 (reference) 1.00 (reference) 1.00 (reference) Tumor size .001 .0003 .48  T1–T2 1.00 (reference) 1.00 (reference) 1.00 (reference)  T3–T4 1.60 (1.11 to 2.31) 1.56 (1.15 to 2.12) 1.16 (0.70 to 1.89)  T4d 2.85 (1.51 to 5.11) 2.65 (1.54 to 4.37) 1.69 (0.66 to 3.74) Lymph nodes <.001 <.001 .02  cN0 1.00 (reference) 1.00 (reference) 1.00 (reference)  cN1–3 1.94 (1.39 to 2.76) 1.63 (1.07 to 2.54) 1.63 (1.07 to 2.54) Tumor grade .72 .94 .05  1–2 1.00 (reference) 1.00 (reference) 1.00 (reference)  3 1.06 (0.76 to 1.50) 0.99 (0.75 to 1.31) 1.53 (0.99 to 2.38) Tumor subgroup <.001 <.001 <.001  HR+ HER2- 1.00 (reference) 1.00 (reference) 1.00 (reference)  HER2+ 1.29 (0.80 to 2.04) 1.05 (0.72 to 1.51) 2.31 (1.28 to 4.19)  HR- HER2- 3.92 (2.65 to 5.84) 2.14 (1.55 to 2.96) 4.35 (2.51 to 7.69) CTC count before NCT <.001 <.001 <.001  0–1 1.00 (reference) 1.00 (reference) 1.00 (reference)  ≥2 3.93 (2.81 to 5.45) 3.73 (2.82 to 4.90) 3.02 (1.88 to 4.75) * Multivariable analyses on overall survival (n = 1362 patients, n = 184 events), distant disease–free survival (n = 1362 patients, n = 267 events), and locoregional relapse–free survival (n = 1362 patients, n = 110 events). Multivariable analyses taking into account CTC and T4d interaction, as well as CTC as a continuous variable, are displayed in Supplementary Material 8 (available online). CTC = circulating tumor cell; DDFS = distant disease–free survival; HER2 = human epidermal growth factor receptor 2; HR+ = hormone receptor positive; HR- = hormone receptor negative; LRFI = locoregional relapse–free interval; NCT = neoadjuvant chemotherapy; OS = overall survival. † Two-sided P values correspond to Wald tests. Table 2. Multivariable survival analyses* Variable OS DDFS LRFI HR (95% CI) P† HR (95% CI) P† HR (95% CI) P† Age, y .04 .004 .08  ≤50 0.72 (0.53 to 0.99) 0.69 (0.53 to 0.89) 0.70 (0.47 to 1.04)  >50 1.00 (reference) 1.00 (reference) 1.00 (reference) Tumor size .001 .0003 .48  T1–T2 1.00 (reference) 1.00 (reference) 1.00 (reference)  T3–T4 1.60 (1.11 to 2.31) 1.56 (1.15 to 2.12) 1.16 (0.70 to 1.89)  T4d 2.85 (1.51 to 5.11) 2.65 (1.54 to 4.37) 1.69 (0.66 to 3.74) Lymph nodes <.001 <.001 .02  cN0 1.00 (reference) 1.00 (reference) 1.00 (reference)  cN1–3 1.94 (1.39 to 2.76) 1.63 (1.07 to 2.54) 1.63 (1.07 to 2.54) Tumor grade .72 .94 .05  1–2 1.00 (reference) 1.00 (reference) 1.00 (reference)  3 1.06 (0.76 to 1.50) 0.99 (0.75 to 1.31) 1.53 (0.99 to 2.38) Tumor subgroup <.001 <.001 <.001  HR+ HER2- 1.00 (reference) 1.00 (reference) 1.00 (reference)  HER2+ 1.29 (0.80 to 2.04) 1.05 (0.72 to 1.51) 2.31 (1.28 to 4.19)  HR- HER2- 3.92 (2.65 to 5.84) 2.14 (1.55 to 2.96) 4.35 (2.51 to 7.69) CTC count before NCT <.001 <.001 <.001  0–1 1.00 (reference) 1.00 (reference) 1.00 (reference)  ≥2 3.93 (2.81 to 5.45) 3.73 (2.82 to 4.90) 3.02 (1.88 to 4.75) Variable OS DDFS LRFI HR (95% CI) P† HR (95% CI) P† HR (95% CI) P† Age, y .04 .004 .08  ≤50 0.72 (0.53 to 0.99) 0.69 (0.53 to 0.89) 0.70 (0.47 to 1.04)  >50 1.00 (reference) 1.00 (reference) 1.00 (reference) Tumor size .001 .0003 .48  T1–T2 1.00 (reference) 1.00 (reference) 1.00 (reference)  T3–T4 1.60 (1.11 to 2.31) 1.56 (1.15 to 2.12) 1.16 (0.70 to 1.89)  T4d 2.85 (1.51 to 5.11) 2.65 (1.54 to 4.37) 1.69 (0.66 to 3.74) Lymph nodes <.001 <.001 .02  cN0 1.00 (reference) 1.00 (reference) 1.00 (reference)  cN1–3 1.94 (1.39 to 2.76) 1.63 (1.07 to 2.54) 1.63 (1.07 to 2.54) Tumor grade .72 .94 .05  1–2 1.00 (reference) 1.00 (reference) 1.00 (reference)  3 1.06 (0.76 to 1.50) 0.99 (0.75 to 1.31) 1.53 (0.99 to 2.38) Tumor subgroup <.001 <.001 <.001  HR+ HER2- 1.00 (reference) 1.00 (reference) 1.00 (reference)  HER2+ 1.29 (0.80 to 2.04) 1.05 (0.72 to 1.51) 2.31 (1.28 to 4.19)  HR- HER2- 3.92 (2.65 to 5.84) 2.14 (1.55 to 2.96) 4.35 (2.51 to 7.69) CTC count before NCT <.001 <.001 <.001  0–1 1.00 (reference) 1.00 (reference) 1.00 (reference)  ≥2 3.93 (2.81 to 5.45) 3.73 (2.82 to 4.90) 3.02 (1.88 to 4.75) * Multivariable analyses on overall survival (n = 1362 patients, n = 184 events), distant disease–free survival (n = 1362 patients, n = 267 events), and locoregional relapse–free survival (n = 1362 patients, n = 110 events). Multivariable analyses taking into account CTC and T4d interaction, as well as CTC as a continuous variable, are displayed in Supplementary Material 8 (available online). CTC = circulating tumor cell; DDFS = distant disease–free survival; HER2 = human epidermal growth factor receptor 2; HR+ = hormone receptor positive; HR- = hormone receptor negative; LRFI = locoregional relapse–free interval; NCT = neoadjuvant chemotherapy; OS = overall survival. † Two-sided P values correspond to Wald tests. Figure 2. View largeDownload slide Survival curves according to circulating tumor cell (CTC) count. Survival curves and the corresponding number of patients at risk are displayed according to CTC count before neoadjuvant chemotherapy (NCT): zero CTCs, one CTC, two CTCs, three or four CTCs, five or more CTCs. P values were obtained by stratified log-rank tests and were two-sided. A) Overall survival (OS) and CTC count before NCT (in months; P < .001). B) Distant disease–free survival (DDFS) and CTC count before NCT (in months; P < .001). C) Locoregional relapse–free interval (LRFI) and CTC count before NCT (in months; P < .001). D) OS and CTC count before surgery (in months; univariate stratified P = .002). E) DDFS and CTC count before surgery (in months; P < .001). F) LRFI and CTC count before surgery (in months; P = .51). CTC = circulating tumor cell; DDFS = distant disease–free survival; LRFI = locoregional relapse–free interval; OS = overall survival. Figure 2. View largeDownload slide Survival curves according to circulating tumor cell (CTC) count. Survival curves and the corresponding number of patients at risk are displayed according to CTC count before neoadjuvant chemotherapy (NCT): zero CTCs, one CTC, two CTCs, three or four CTCs, five or more CTCs. P values were obtained by stratified log-rank tests and were two-sided. A) Overall survival (OS) and CTC count before NCT (in months; P < .001). B) Distant disease–free survival (DDFS) and CTC count before NCT (in months; P < .001). C) Locoregional relapse–free interval (LRFI) and CTC count before NCT (in months; P < .001). D) OS and CTC count before surgery (in months; univariate stratified P = .002). E) DDFS and CTC count before surgery (in months; P < .001). F) LRFI and CTC count before surgery (in months; P = .51). CTC = circulating tumor cell; DDFS = distant disease–free survival; LRFI = locoregional relapse–free interval; OS = overall survival. CTC Detection and Prognostic Models The addition of baseline CTC count modeled as a binary variable (fewer than two vs two or more CTCs) to clinicopathological prognostic models (Table 2) resulted in statistically significant increases in prognostication of overall survival (P < .001), distant disease–free survival (P < .001), and locoregional relapse–free interval (P = .008) (Table 3); similar results were obtained when modeling CTC count as a continuous variable (splines) (Table 3). In 861 patients with full data available, further investigations showed that adding baseline CTC count to models including clinicopathological data and pathological complete response status also increased postneoadjuvant survival prognostication (Table 3). Regarding CTC detection during NCT, we found that it did not increase survival prognostication (Table 3). Table 3. Assessment of added prognostic information of circulating tumor cells before and during neoadjuvant chemotherapy* Model 1 Model 1 average c-index Model 2 Model 2 average c-index Average c-index increase, model 2 – model 1 (95% CI) ddf Average increase χ2 (95% CI) P† Models with CTC count before NCT  OS (n = 1362)   CP 0.736 Model 1 + CTCBC (< or ≥2 CTC) 0.767 0.031 (–0.003 to 0.063) 1 28.9 (13.3 to 44.1) <.001   CP 0.741 Model 1 + CTCBC (splines) 0.779 0.038 (0.004 to 0.070) 2 30.0 (14.7 to 48.5) <.001  DDFS (n = 1362)   CP 0.666 Model 1 + CTCBC (< or ≥2 CTC) 0.705 0.039 (0.011 to 0.070) 1 38.7 (21.9 to 58.6) <.001   CP 0.664 Model 1 + CTCBC (splines) 0.709 0.045 (0.016 to 0.073) 2 41.2 (24.2 to 63.7) <.001  LRFI (n = 1362)   CP 0.728 Model 1 + CTCBC (< or ≥2 CTC) 0.734 0.006 (–0.021 to 0.031) 1 8.8 (2.4 to 19.2) .008   CP 0.727 Model 1 + CTCBC (splines) 0.731 0.004 (–0.032 to 0.032) 2 9.7 (1.9 to 20.7) .003 Models with CTC count before NCT and pCR (landmark analyses)  OS (n = 1332)   CP+ CTCBC (< or ≥2 CTC) 0.771 Model 1 + pCR 0.809 0.037 (0.008 to 0.064) 1 26.7 (13.1 to 44.0) <.001   CP+ CTCBC (splines) 0.778 Model 1 + pCR 0.813 0.035 (0.002 to 0.065) 1 18.8 (7.9 to 36.1) <.001   CP + pCR 0.774 Model 1 + CTCBC (< or ≥2 CTC) 0.809 0.035 (0.002 to 0.063) 1 30.3 (14.7 to 48.8) <.001   CP + pCR 0.774 Model 1 + CTCBC (splines) 0.813 0.039 (0.004 to 0.072) 2 30.5 (14.4 to 47.8) <.001  DDFS (n = 1323)   CP+ CTCBC (< or ≥2 CTC) 0.698 Model 1 + pCR 0.746 0.048 (0.019 to 0.072) 1 27.9 (13.0 to 45.0) <.001   CP+ CTCBC (splines) 0.704 Model 1 + pCR 0.751 0.047 (0.016 to 0.070) 1 21.8 (10.0 to 37.3) <.001   CP + pCR 0.703 Model 1 + CTCBC (< or ≥2 CTC) 0.746 0.044 (0.012 to 0.068) 1 36.3 (19.7 to 55.4) <.001   CP + pCR 0.703 Model 1 + CTCBC (splines) 0.751 0.048 (0.018 to 0.074) 2 37.1 (21.1 to 57.1) <.001  LRFI (n = 1322)   CP+ CTCBC (< or ≥2 CTC) 0.726 Model 1 + pCR 0.728 0.002 (–0.042 to 0.031) 1 4.90 (0.30 to 13.1) .03   CP+ CTCBC (splines) 0.724 Model 1 + pCR 0.726 0.003 (–0.041 to 0.029) 1 2.6 (0.01 to 9.61) .10   CP + pCR 0.729 Model 1 + CTCBC (< or ≥2 CTC) 0.728 –0.001 (–0.031 to 0.024) 1 9.1 (1.74 to 18.7) .002   CP + pCR 0.729 Model 1 + CTCBC (splines) 0.726 –0.003 (–0.037 to 0.021) 2 8.7 (1.7 to 18.9) .02 Models with CTC count before NCT, pCR, and CTC count during NCT (landmark analyses)  OS (n = 861)   CP+CTCBC+pCR 0.804 Model 1 + CTCDC (< or ≥2 CTC) 0.806 0.002 (–0.016 to 0.013) 1 3.1 (0.0 to 11.4) .08   CP+CTCBC+pCR 0.798 Model 1 + CTCDC (splines) 0.796 –0.003 (–0.038 to 0.080) 2 5.0 (0.2 to 12.9) .02  DDFS (n = 856)   CP+CTCBC+pCR 0.758 Model 1 + CTCDC (< or ≥2 CTC) 0.760 0.002 (–0.013 to 0.014) 1 3.1 (0.0 to 9.7) .08   CP+CTCBC+pCR 0.756 Model 1 + CTCDC (splines) 0.755 –0.001 (–0.016 to 0.007) 2 2.7 (0.02 to 10.8) .09  LRFI (n = 855)   CP+CTCBC+pCR 0.717 Model 1 + CTCDC (< or ≥2 CTC) 0.707 –0.011 (–0.074 to 0.006) 1 0.9 (0.0 to 3.6) .34   CP+CTCBC+pCR 0.725 Model 1 + CTCDC (splines) 0.715 –0.010 (–0.047 to 0.006) 2 0.6 (0.0 to 2.8) .44 Model 1 Model 1 average c-index Model 2 Model 2 average c-index Average c-index increase, model 2 – model 1 (95% CI) ddf Average increase χ2 (95% CI) P† Models with CTC count before NCT  OS (n = 1362)   CP 0.736 Model 1 + CTCBC (< or ≥2 CTC) 0.767 0.031 (–0.003 to 0.063) 1 28.9 (13.3 to 44.1) <.001   CP 0.741 Model 1 + CTCBC (splines) 0.779 0.038 (0.004 to 0.070) 2 30.0 (14.7 to 48.5) <.001  DDFS (n = 1362)   CP 0.666 Model 1 + CTCBC (< or ≥2 CTC) 0.705 0.039 (0.011 to 0.070) 1 38.7 (21.9 to 58.6) <.001   CP 0.664 Model 1 + CTCBC (splines) 0.709 0.045 (0.016 to 0.073) 2 41.2 (24.2 to 63.7) <.001  LRFI (n = 1362)   CP 0.728 Model 1 + CTCBC (< or ≥2 CTC) 0.734 0.006 (–0.021 to 0.031) 1 8.8 (2.4 to 19.2) .008   CP 0.727 Model 1 + CTCBC (splines) 0.731 0.004 (–0.032 to 0.032) 2 9.7 (1.9 to 20.7) .003 Models with CTC count before NCT and pCR (landmark analyses)  OS (n = 1332)   CP+ CTCBC (< or ≥2 CTC) 0.771 Model 1 + pCR 0.809 0.037 (0.008 to 0.064) 1 26.7 (13.1 to 44.0) <.001   CP+ CTCBC (splines) 0.778 Model 1 + pCR 0.813 0.035 (0.002 to 0.065) 1 18.8 (7.9 to 36.1) <.001   CP + pCR 0.774 Model 1 + CTCBC (< or ≥2 CTC) 0.809 0.035 (0.002 to 0.063) 1 30.3 (14.7 to 48.8) <.001   CP + pCR 0.774 Model 1 + CTCBC (splines) 0.813 0.039 (0.004 to 0.072) 2 30.5 (14.4 to 47.8) <.001  DDFS (n = 1323)   CP+ CTCBC (< or ≥2 CTC) 0.698 Model 1 + pCR 0.746 0.048 (0.019 to 0.072) 1 27.9 (13.0 to 45.0) <.001   CP+ CTCBC (splines) 0.704 Model 1 + pCR 0.751 0.047 (0.016 to 0.070) 1 21.8 (10.0 to 37.3) <.001   CP + pCR 0.703 Model 1 + CTCBC (< or ≥2 CTC) 0.746 0.044 (0.012 to 0.068) 1 36.3 (19.7 to 55.4) <.001   CP + pCR 0.703 Model 1 + CTCBC (splines) 0.751 0.048 (0.018 to 0.074) 2 37.1 (21.1 to 57.1) <.001  LRFI (n = 1322)   CP+ CTCBC (< or ≥2 CTC) 0.726 Model 1 + pCR 0.728 0.002 (–0.042 to 0.031) 1 4.90 (0.30 to 13.1) .03   CP+ CTCBC (splines) 0.724 Model 1 + pCR 0.726 0.003 (–0.041 to 0.029) 1 2.6 (0.01 to 9.61) .10   CP + pCR 0.729 Model 1 + CTCBC (< or ≥2 CTC) 0.728 –0.001 (–0.031 to 0.024) 1 9.1 (1.74 to 18.7) .002   CP + pCR 0.729 Model 1 + CTCBC (splines) 0.726 –0.003 (–0.037 to 0.021) 2 8.7 (1.7 to 18.9) .02 Models with CTC count before NCT, pCR, and CTC count during NCT (landmark analyses)  OS (n = 861)   CP+CTCBC+pCR 0.804 Model 1 + CTCDC (< or ≥2 CTC) 0.806 0.002 (–0.016 to 0.013) 1 3.1 (0.0 to 11.4) .08   CP+CTCBC+pCR 0.798 Model 1 + CTCDC (splines) 0.796 –0.003 (–0.038 to 0.080) 2 5.0 (0.2 to 12.9) .02  DDFS (n = 856)   CP+CTCBC+pCR 0.758 Model 1 + CTCDC (< or ≥2 CTC) 0.760 0.002 (–0.013 to 0.014) 1 3.1 (0.0 to 9.7) .08   CP+CTCBC+pCR 0.756 Model 1 + CTCDC (splines) 0.755 –0.001 (–0.016 to 0.007) 2 2.7 (0.02 to 10.8) .09  LRFI (n = 855)   CP+CTCBC+pCR 0.717 Model 1 + CTCDC (< or ≥2 CTC) 0.707 –0.011 (–0.074 to 0.006) 1 0.9 (0.0 to 3.6) .34   CP+CTCBC+pCR 0.725 Model 1 + CTCDC (splines) 0.715 –0.010 (–0.047 to 0.006) 2 0.6 (0.0 to 2.8) .44 * For each analysis, the number of included patients corresponds to the number of patients with available data. Clinical and pathological characteristics were not statistically different among the different patient populations included in these analyses. CI = confidence interval; CP = baseline clinicopathological model; CTC = circulating tumor cells; CTCBC = CTC count before NCT; CTCDC = CTC count during NCT, using the highest value of CTC detection either after NCT start or prior to surgery; ddf = degree of freedom; DDFS = distant disease–free survival; LRFI = locoregional relapse–free interval; NCT = neoadjuvant chemotherapy; OS = overall survival; pCR = pathological complete response. † Two-sided likelihood ratio test P value corresponds to the average increase in χ2. Table 3. Assessment of added prognostic information of circulating tumor cells before and during neoadjuvant chemotherapy* Model 1 Model 1 average c-index Model 2 Model 2 average c-index Average c-index increase, model 2 – model 1 (95% CI) ddf Average increase χ2 (95% CI) P† Models with CTC count before NCT  OS (n = 1362)   CP 0.736 Model 1 + CTCBC (< or ≥2 CTC) 0.767 0.031 (–0.003 to 0.063) 1 28.9 (13.3 to 44.1) <.001   CP 0.741 Model 1 + CTCBC (splines) 0.779 0.038 (0.004 to 0.070) 2 30.0 (14.7 to 48.5) <.001  DDFS (n = 1362)   CP 0.666 Model 1 + CTCBC (< or ≥2 CTC) 0.705 0.039 (0.011 to 0.070) 1 38.7 (21.9 to 58.6) <.001   CP 0.664 Model 1 + CTCBC (splines) 0.709 0.045 (0.016 to 0.073) 2 41.2 (24.2 to 63.7) <.001  LRFI (n = 1362)   CP 0.728 Model 1 + CTCBC (< or ≥2 CTC) 0.734 0.006 (–0.021 to 0.031) 1 8.8 (2.4 to 19.2) .008   CP 0.727 Model 1 + CTCBC (splines) 0.731 0.004 (–0.032 to 0.032) 2 9.7 (1.9 to 20.7) .003 Models with CTC count before NCT and pCR (landmark analyses)  OS (n = 1332)   CP+ CTCBC (< or ≥2 CTC) 0.771 Model 1 + pCR 0.809 0.037 (0.008 to 0.064) 1 26.7 (13.1 to 44.0) <.001   CP+ CTCBC (splines) 0.778 Model 1 + pCR 0.813 0.035 (0.002 to 0.065) 1 18.8 (7.9 to 36.1) <.001   CP + pCR 0.774 Model 1 + CTCBC (< or ≥2 CTC) 0.809 0.035 (0.002 to 0.063) 1 30.3 (14.7 to 48.8) <.001   CP + pCR 0.774 Model 1 + CTCBC (splines) 0.813 0.039 (0.004 to 0.072) 2 30.5 (14.4 to 47.8) <.001  DDFS (n = 1323)   CP+ CTCBC (< or ≥2 CTC) 0.698 Model 1 + pCR 0.746 0.048 (0.019 to 0.072) 1 27.9 (13.0 to 45.0) <.001   CP+ CTCBC (splines) 0.704 Model 1 + pCR 0.751 0.047 (0.016 to 0.070) 1 21.8 (10.0 to 37.3) <.001   CP + pCR 0.703 Model 1 + CTCBC (< or ≥2 CTC) 0.746 0.044 (0.012 to 0.068) 1 36.3 (19.7 to 55.4) <.001   CP + pCR 0.703 Model 1 + CTCBC (splines) 0.751 0.048 (0.018 to 0.074) 2 37.1 (21.1 to 57.1) <.001  LRFI (n = 1322)   CP+ CTCBC (< or ≥2 CTC) 0.726 Model 1 + pCR 0.728 0.002 (–0.042 to 0.031) 1 4.90 (0.30 to 13.1) .03   CP+ CTCBC (splines) 0.724 Model 1 + pCR 0.726 0.003 (–0.041 to 0.029) 1 2.6 (0.01 to 9.61) .10   CP + pCR 0.729 Model 1 + CTCBC (< or ≥2 CTC) 0.728 –0.001 (–0.031 to 0.024) 1 9.1 (1.74 to 18.7) .002   CP + pCR 0.729 Model 1 + CTCBC (splines) 0.726 –0.003 (–0.037 to 0.021) 2 8.7 (1.7 to 18.9) .02 Models with CTC count before NCT, pCR, and CTC count during NCT (landmark analyses)  OS (n = 861)   CP+CTCBC+pCR 0.804 Model 1 + CTCDC (< or ≥2 CTC) 0.806 0.002 (–0.016 to 0.013) 1 3.1 (0.0 to 11.4) .08   CP+CTCBC+pCR 0.798 Model 1 + CTCDC (splines) 0.796 –0.003 (–0.038 to 0.080) 2 5.0 (0.2 to 12.9) .02  DDFS (n = 856)   CP+CTCBC+pCR 0.758 Model 1 + CTCDC (< or ≥2 CTC) 0.760 0.002 (–0.013 to 0.014) 1 3.1 (0.0 to 9.7) .08   CP+CTCBC+pCR 0.756 Model 1 + CTCDC (splines) 0.755 –0.001 (–0.016 to 0.007) 2 2.7 (0.02 to 10.8) .09  LRFI (n = 855)   CP+CTCBC+pCR 0.717 Model 1 + CTCDC (< or ≥2 CTC) 0.707 –0.011 (–0.074 to 0.006) 1 0.9 (0.0 to 3.6) .34   CP+CTCBC+pCR 0.725 Model 1 + CTCDC (splines) 0.715 –0.010 (–0.047 to 0.006) 2 0.6 (0.0 to 2.8) .44 Model 1 Model 1 average c-index Model 2 Model 2 average c-index Average c-index increase, model 2 – model 1 (95% CI) ddf Average increase χ2 (95% CI) P† Models with CTC count before NCT  OS (n = 1362)   CP 0.736 Model 1 + CTCBC (< or ≥2 CTC) 0.767 0.031 (–0.003 to 0.063) 1 28.9 (13.3 to 44.1) <.001   CP 0.741 Model 1 + CTCBC (splines) 0.779 0.038 (0.004 to 0.070) 2 30.0 (14.7 to 48.5) <.001  DDFS (n = 1362)   CP 0.666 Model 1 + CTCBC (< or ≥2 CTC) 0.705 0.039 (0.011 to 0.070) 1 38.7 (21.9 to 58.6) <.001   CP 0.664 Model 1 + CTCBC (splines) 0.709 0.045 (0.016 to 0.073) 2 41.2 (24.2 to 63.7) <.001  LRFI (n = 1362)   CP 0.728 Model 1 + CTCBC (< or ≥2 CTC) 0.734 0.006 (–0.021 to 0.031) 1 8.8 (2.4 to 19.2) .008   CP 0.727 Model 1 + CTCBC (splines) 0.731 0.004 (–0.032 to 0.032) 2 9.7 (1.9 to 20.7) .003 Models with CTC count before NCT and pCR (landmark analyses)  OS (n = 1332)   CP+ CTCBC (< or ≥2 CTC) 0.771 Model 1 + pCR 0.809 0.037 (0.008 to 0.064) 1 26.7 (13.1 to 44.0) <.001   CP+ CTCBC (splines) 0.778 Model 1 + pCR 0.813 0.035 (0.002 to 0.065) 1 18.8 (7.9 to 36.1) <.001   CP + pCR 0.774 Model 1 + CTCBC (< or ≥2 CTC) 0.809 0.035 (0.002 to 0.063) 1 30.3 (14.7 to 48.8) <.001   CP + pCR 0.774 Model 1 + CTCBC (splines) 0.813 0.039 (0.004 to 0.072) 2 30.5 (14.4 to 47.8) <.001  DDFS (n = 1323)   CP+ CTCBC (< or ≥2 CTC) 0.698 Model 1 + pCR 0.746 0.048 (0.019 to 0.072) 1 27.9 (13.0 to 45.0) <.001   CP+ CTCBC (splines) 0.704 Model 1 + pCR 0.751 0.047 (0.016 to 0.070) 1 21.8 (10.0 to 37.3) <.001   CP + pCR 0.703 Model 1 + CTCBC (< or ≥2 CTC) 0.746 0.044 (0.012 to 0.068) 1 36.3 (19.7 to 55.4) <.001   CP + pCR 0.703 Model 1 + CTCBC (splines) 0.751 0.048 (0.018 to 0.074) 2 37.1 (21.1 to 57.1) <.001  LRFI (n = 1322)   CP+ CTCBC (< or ≥2 CTC) 0.726 Model 1 + pCR 0.728 0.002 (–0.042 to 0.031) 1 4.90 (0.30 to 13.1) .03   CP+ CTCBC (splines) 0.724 Model 1 + pCR 0.726 0.003 (–0.041 to 0.029) 1 2.6 (0.01 to 9.61) .10   CP + pCR 0.729 Model 1 + CTCBC (< or ≥2 CTC) 0.728 –0.001 (–0.031 to 0.024) 1 9.1 (1.74 to 18.7) .002   CP + pCR 0.729 Model 1 + CTCBC (splines) 0.726 –0.003 (–0.037 to 0.021) 2 8.7 (1.7 to 18.9) .02 Models with CTC count before NCT, pCR, and CTC count during NCT (landmark analyses)  OS (n = 861)   CP+CTCBC+pCR 0.804 Model 1 + CTCDC (< or ≥2 CTC) 0.806 0.002 (–0.016 to 0.013) 1 3.1 (0.0 to 11.4) .08   CP+CTCBC+pCR 0.798 Model 1 + CTCDC (splines) 0.796 –0.003 (–0.038 to 0.080) 2 5.0 (0.2 to 12.9) .02  DDFS (n = 856)   CP+CTCBC+pCR 0.758 Model 1 + CTCDC (< or ≥2 CTC) 0.760 0.002 (–0.013 to 0.014) 1 3.1 (0.0 to 9.7) .08   CP+CTCBC+pCR 0.756 Model 1 + CTCDC (splines) 0.755 –0.001 (–0.016 to 0.007) 2 2.7 (0.02 to 10.8) .09  LRFI (n = 855)   CP+CTCBC+pCR 0.717 Model 1 + CTCDC (< or ≥2 CTC) 0.707 –0.011 (–0.074 to 0.006) 1 0.9 (0.0 to 3.6) .34   CP+CTCBC+pCR 0.725 Model 1 + CTCDC (splines) 0.715 –0.010 (–0.047 to 0.006) 2 0.6 (0.0 to 2.8) .44 * For each analysis, the number of included patients corresponds to the number of patients with available data. Clinical and pathological characteristics were not statistically different among the different patient populations included in these analyses. CI = confidence interval; CP = baseline clinicopathological model; CTC = circulating tumor cells; CTCBC = CTC count before NCT; CTCDC = CTC count during NCT, using the highest value of CTC detection either after NCT start or prior to surgery; ddf = degree of freedom; DDFS = distant disease–free survival; LRFI = locoregional relapse–free interval; NCT = neoadjuvant chemotherapy; OS = overall survival; pCR = pathological complete response. † Two-sided likelihood ratio test P value corresponds to the average increase in χ2. Discussion This international meta-analysis shows that hematogenous cancer cell dissemination is affecting the outcome of treated patients and that this process cannot be predicted precisely enough by the current assessment of the primary tumor characteristics and response. Besides the diagnosis of inflammatory (T4d) breast cancer, CTC count was independent from any other baseline clinical or pathological characteristics. We also found that, in contrast to many other known clinic-pathological prognostic factors, CTC detection (either before NCT or before surgery) was not associated with pathological complete response. We found that the best postneoadjuvant survival models included prespecified clinicopathological prognostic markers at baseline, CTC detection at baseline, and pathological complete response, demonstrating the relevance of such metastasis-associated biomarkers for outcome prediction. Although presurgery CTC detection demonstrated its prognostic significance, it did not improve the accuracy of an optimized prognostic model taking into account CTC detection at baseline. In addition to the expected occurrence of distant metastases, CTC counts were also correlated to locoregional relapse. Although we cannot exclude that this unexpected finding might be simply a “bystander” effect of identifying more aggressive tumors, it could also suggest that the spread of breast cancer cells in the body might be also associated with a “self-seeding” mechanism (ie, breast cancer cells recirculate from distant sites back to the primary site), which has been demonstrated so far only in experimental models (4) and with bone marrow disseminated tumor cells (16). CTC count was a quantitative marker, as each CTC detected added a quantum of poor prognosis. In our clinical validity analyses, survival curves displayed a statistically significant survival difference starting from two CTCs detected. Interestingly, the two other thresholds that were tested in the current meta-analysis, namely one or more and five or more CTCs, were also able to distinguish a high-risk population. Any future adjuvant trial based on CTC detection should have to define a threshold compatible with its target population size (five or more CTCs being a rare event), taking into account that one CTC was detected in some healthy individuals with the applied CTC assay (5) and that the low CTC count observed in nonmetastatic breast cancer patients follows a Poisson’s distribution of rare events—blurring the line between zero and one CTCs detected in a single tube of blood (17). The same reason might explain why the longitudinal follow-up of CTC count and changes appeared of lower clinical relevance than previously reported in metastatic breast cancer patients (8). The main limitation of this study is related to the technical aspects underlying CTC detection by the CellSearch technique. Biologically, this US Food and Drug Administration–cleared technique captures CTC by their epithelial cell surface antigens, the expression of which is decreased in some cancer cells during the epithelial-to-mesenchymal transition (EMT) (18). However, abundant evidence supporting the clinical validity of this CTC detection technique in breast cancer had been reported (19); moreover, tumor cells with an intermediate EMT phenotype detected by CellSearch appear to be metastasis-initiator cells (17), which might explain why small incremental increases in CTC counts were associated with increased risks of relapse in the present analysis. Taken together, while the validity of NCT trials is impaired by the limited correlation between primary tumor response and postneoadjuvant survival, the enumeration of CTC improves postneoadjuvant survival prediction. NCT being also used in patients with other tumor entities where blood-borne dissemination of cancer cells plays also an important role (eg, rectal cancer) (20), the current report might stimulate future clinical studies in NCT-treated patients and therefore has implications beyond breast cancer. Funding This work was supported by Janssen Diagnostics and Innovative Medicines Initiative (grant agreement No. 115749, project “Cancer-ID”). Notes Authors: François-Clément Bidard, Stefan Michiels, Sabine Riethdorf, Volkmar Mueller, Laura J. Esserman, Anthony Lucci, Bjørn Naume, Jun Horiguchi, Rafael Gisbert-Criado, Stefan Sleijfer, Masakazu Toi, Jose A. Garcia-Saenz, Andreas Hartkopf, Daniele Generali, Françoise Rothé, Jeffrey Smerage, Laura Muinelo-Romay, Justin Stebbing, Patrice Viens, Mark Jesus M. Magbanua, Carolyn S. Hall, Olav Engebraaten, Daisuke Takata, José Vidal-Martínez, Wendy Onstenk, Noriyoshi Fujisawa, Eduardo Diaz-Rubio, Florin-Andrei Taran, Maria Rosa Cappelletti, Michail Ignatiadis, Charlotte Proudhon, Denise M. Wolf, Jessica B. Bauldry, Elin Borgen, Rin Nagaoka, Vicente Carañana, Jaco Kraan, Marisa Maestro, Sara Yvonne Brucker, Karsten Weber, Fabien Reyal, Dominic Amara, Mandar G. Karhade, Randi Ruud Mathiesen, Hideaki Tokiniwa, Antonio Llombart-Cussac, Alessandra Meddis, Paul Blanche, Koenraad d'Hollander, Paul Cottu, John W. Park, Sibylle Loibl, Aurélien Latouche, Jean-Yves Pierga, Klaus Pantel Affiliations of authors: Institut Curie, UVSQ and Paris Saclay University, Saint Cloud, France (FCB); Gustave Roussy, Villejuif, France (SM); Universitätsklinikum Hamburg-Eppendorf, Hamburg, Germany (SR, VM, KP); University of California at San Francisco, San Francisco, CA (LJE, MJMM, DMW, DA, JWP); MD Anderson Cancer Center, Houston, TX (ALu, CSH, JBB, MGK); Oslo University Hospital, Oslo, Norway (BN, OE, EB, RRM); Institute for Clinical Medicine, University of Oslo, Oslo, Norway (BN, OE); Gunma University Hospital, Gunma, Japan (JH, DT, RN, HT); Hospital Arnau de Vilanova, Valencia, Spain (RGC, JVM, VC, ALC); Erasmus MC Cancer Institute, Rotterdam, the Netherlands (SS, WO, JK); Kyoto University Hospital, Kyoto, Japan (MT, NF); Hospital Clinico San Carlos, Universidad Complutense, Madrid, Spain (JAGS, EDR, EDR, MM); Women’s Health Center, University of Tuebingen, Tuebingen, Germany (AH, FAT, SYB); Women Cancer Centre, University of Trieste, ASST of Cremona, Cremona, Italy (DG, MRC); Institut Jules Bordet, Université Libre de Bruxelles, Brussels, Belgium (FR, MI); University of Michigan Comprehensive Cancer Center, Ann Arbor, MI (JS); Complejo Hospitalario Universitario, Santiago de Compostela, Spain (LMR); Imperial College, London, UK (JS); Institut Paoli Calmettes, Aix Marseille University, CNRS, Inserm, Marseilles, France (PV); Institut Curie, PSL Research University, Paris, France (CP, FR, PC, JYP); Paris Descartes University, Paris, France (JYP); German Breast Group, Neu-Isenburg, Germany (KW, SL); Institut Curie, Inserm U900, Saint Cloud, France (AM, ALa); LMBA, Université de Bretagne Sud, Vannes, France (PB); International Drug Development Institute, Louvain La Neuve, Belgium (Kd’H); Conservatoire National des Arts et Métiers, EA4629, Paris, France (ALa). The study sponsor played no role in the design of the study; the collection, analysis, and interpretation of the data; the writing of the manuscript; and the decision to submit the manuscript for publication. This study was funded in part by a research grant from Janssen Diagnostics to F. C. Bidard’s institution. References 1 King TA , Morrow M. Surgical issues in patients with breast cancer receiving neoadjuvant chemotherapy . Nat Rev Clin Oncol. 2015 ; 12 ( 6 ): 335 – 343 . Google Scholar CrossRef Search ADS PubMed 2 Cortazar P , Zhang L , Untch M et al. , Pathological complete response and long-term clinical benefit in breast cancer: The CTNeoBC pooled analysis . Lancet. 2014 ; 384 ( 9938 ): 164 – 172 . Google Scholar CrossRef Search ADS PubMed 3 DeMichele A , Yee D , Berry DA et al. , The neoadjuvant model is still the future for drug development in breast cancer . Clin Cancer Res. 2015 ; 21 ( 13 ): 2911 – 2915 . Google Scholar CrossRef Search ADS PubMed 4 Comen E , Norton L , Massagué J. Clinical implications of cancer self-seeding . Nat Rev Clin Oncol. 2011 ; 8 ( 6 ): 369 – 377 . Google Scholar CrossRef Search ADS PubMed 5 Allard WJ , Matera J , Miller MC et al. , Tumor cells circulate in the peripheral blood of all major carcinomas but not in healthy subjects or patients with nonmalignant diseases . Clin Cancer Res. 2004 ; 10 ( 20 ): 6897 – 6904 . Google Scholar CrossRef Search ADS PubMed 6 Coumans FAW , Ligthart ST , Uhr JW , Terstappen LWMM. Challenges in the enumeration and phenotyping of CTC . Clin Cancer Res. 2012 ; 18 ( 20 ): 5711 – 5718 . Google Scholar CrossRef Search ADS PubMed 7 Ignatiadis M , Riethdorf S , Bidard FC et al. , International study on inter-reader variability for circulating tumor cells in breast cancer . Breast Cancer Res. 2014 ; 16 ( 2 ): R43 . Google Scholar CrossRef Search ADS PubMed 8 Bidard FC , Peeters DJ , Fehm T et al. , Clinical validity of circulating tumour cells in patients with metastatic breast cancer: A pooled analysis of individual patient data . Lancet Oncol. 2014 ; 15 ( 4 ): 406 – 414 . Google Scholar CrossRef Search ADS PubMed 9 Lorente D , Olmos D , Mateo J et al. , Decline in circulating tumor cell count and treatment outcome in advanced prostate cancer . Eur Urol. 2016 ; 70 ( 6 ): 985 – 992 . Google Scholar CrossRef Search ADS PubMed 10 Krebs MG , Sloane R , Priest L et al. , Evaluation and prognostic significance of circulating tumor cells in patients with non-small-cell lung cancer . J Clin Oncol. 2011 ; 29 ( 12 ): 1556 – 1563 . Google Scholar CrossRef Search ADS PubMed 11 Morrow CJ , Trapani F , Metcalf RL et al. , Tumourigenic non-small-cell lung cancer mesenchymal circulating tumour cells: A clinical case study . Ann Oncol. 2016 ; 27 ( 6 ): 1155 – 1160 . Google Scholar CrossRef Search ADS PubMed 12 Baccelli I , Schneeweiss A , Riethdorf S et al. , Identification of a population of blood circulating tumor cells from breast cancer patients that initiates metastasis in a xenograft assay . Nat Biotechnol. 2013 ; 31 ( 6 ): 539 – 544 . Google Scholar CrossRef Search ADS PubMed 13 Cayrefourcq L , Mazard T , Joosse S et al. , Establishment and characterization of a cell line from human circulating colon cancer cells . Cancer Res. 2015 ; 75 ( 5 ): 892 – 901 . Google Scholar CrossRef Search ADS PubMed 14 Michiels S , Baujat B , Mahé C et al. , Random effects survival models gave a better understanding of heterogeneity in individual patient data meta-analyses . J Clin Epidemiol. 2005 ; 58 ( 3 ): 238 – 245 . Google Scholar CrossRef Search ADS PubMed 15 Tudur Smith C , Williamson PR. A comparison of methods for fixed effects meta-analysis of individual patient data with time to event outcomes . Clin Trials. 2007 ; 4 ( 6 ): 621 – 630 . Google Scholar CrossRef Search ADS PubMed 16 Bidard FC , Kirova YM , Vincent-Salomon A et al. , Disseminated tumor cells and the risk of locoregional recurrence in nonmetastatic breast cancer . Ann Oncol. 2009 ; 20 ( 11 ): 1836 – 1841 . Google Scholar CrossRef Search ADS PubMed 17 Andree KC , van Dalum G , Terstappen LWMM. Challenges in circulating tumor cell detection by the CellSearch system . Mol Oncol. 2016 ; 10 ( 3 ): 395 – 407 . Google Scholar CrossRef Search ADS PubMed 18 Bidard FC , Proudhon C , Pierga JY. Circulating tumor cells in breast cancer . Mol Oncol. 2016 ; 10 ( 3 ): 418 – 430 . Google Scholar CrossRef Search ADS PubMed 19 Alix-Panabières C , Mader S , Pantel K. Epithelial-mesenchymal plasticity in circulating tumor cells . J Mol Med. 2017 ; 95 ( 2 ): 133 – 142 . Google Scholar CrossRef Search ADS PubMed 20 Cabel L , Proudhon C , Gortais H et al. , Circulating tumor cells: clinical validity and utility . Int J Clin Oncol. 2017 ; 22 ( 3 ): 421 – 430 . Google Scholar CrossRef Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png JNCI: Journal of the National Cancer Institute Oxford University Press

Loading next page...
 
/lp/ou_press/circulating-tumor-cells-in-breast-cancer-patients-treated-by-N7ZPu1nvIh
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com
ISSN
0027-8874
eISSN
1460-2105
D.O.I.
10.1093/jnci/djy018
Publisher site
See Article on Publisher Site

Abstract

Abstract Background We conducted a meta-analysis in nonmetastatic breast cancer patients treated by neoadjuvant chemotherapy (NCT) to assess the clinical validity of circulating tumor cell (CTC) detection as a prognostic marker. Methods We collected individual patient data from 21 studies in which CTC detection by CellSearch was performed in early breast cancer patients treated with NCT. The primary end point was overall survival, analyzed according to CTC detection, using Cox regression models stratified by study. Secondary end points included distant disease–free survival, locoregional relapse–free interval, and pathological complete response. All statistical tests were two-sided. Results Data from patients were collected before NCT (n = 1574) and before surgery (n = 1200). CTC detection revealed one or more CTCs in 25.2% of patients before NCT; this was associated with tumor size (P < .001). The number of CTCs detected had a detrimental and decremental impact on overall survival (P < .001), distant disease–free survival (P < .001), and locoregional relapse–free interval (P < .001), but not on pathological complete response. Patients with one, two, three to four, and five or more CTCs before NCT displayed hazard ratios of death of 1.09 (95% confidence interval [CI] = 0.65 to 1.69), 2.63 (95% CI = 1.42 to 4.54), 3.83 (95% CI = 2.08 to 6.66), and 6.25 (95% CI = 4.34 to 9.09), respectively. In 861 patients with full data available, adding CTC detection before NCT increased the prognostic ability of multivariable prognostic models for overall survival (P < .001), distant disease–free survival (P < .001), and locoregional relapse–free interval (P = .008). Conclusions CTC count is an independent and quantitative prognostic factor in early breast cancer patients treated by NCT. It complements current prognostic models based on tumor characteristics and response to therapy. Neoadjuvant chemotherapy (NCT) is a standard treatment for patients with large, nonmetastatic breast cancer and may allow breast-conserving surgery after tumor downsizing while decreasing the risk of subsequent relapse (1). In addition to these prominent clinical benefits, NCT provides a unique opportunity to study and quantify the cancer response to antineoplastic agents by assessment of the pathological response on the residual tumor at the time of surgery. Pathological complete response (pCR) rate has been thoroughly investigated as a potential surrogate of post-NCT survival, with heterogeneous results according to the various breast cancer subtypes (2,3). In this context, improving post-NCT survival prediction may lead to better evaluation of NCT efficacy and help to optimize adjuvant therapy. Neoadjuvant therapy has two main modes of action, that is, shrinkage of the primary tumor and eradication of blood-borne tumor cell dissemination (4). Surprisingly, the second mode is totally ignored by current NCT assessments, despite the fact that cancer metastasis is the main cause of death in breast cancer and other solid tumors. The precise enumeration of circulating tumor cells (CTCs) in the peripheral blood of cancer patients as surrogate of the dissemination process is now possible after years of international standardization (5–7) and has demonstrated its clinical validity at metastatic stage in many cancer types (8–10). Moreover, experimental studies have shown that CTCs can induce tumors after xenografting into immunodeficient mice (11–13). Here, we report the results of a meta-analysis bringing together individual patient data from the United States, Japan, and European countries. This large-scale multicenter meta-analysis investigated in nonmetastatic breast cancer patients treated by NCT whether CTC counts add statistically significant prognostic information on postneoadjuvant survival to the established assessment of primary tumor characteristics and pathological complete response. Methods Literature Review and Study Identification The study protocol was set up by the study secretariat, and a search of Medline and major oncology congress abstracts was performed in February 2015 to find eligible studies (Supplementary Methods 1, available online). Direct contact was then established with all CTC analysis centers and cooperative research groups deemed to have eligible data. Inclusion criteria were published and unpublished studies conducted in nonmetastatic breast cancer patients treated by NCT; studies in which CTC count was performed at least once before surgery by the CellSearch system; studies with available survival data; studies declared and approved by an appropriate ethics committee; and patient accrual in or after January 2003 and before September 2015. Studies in which CTC count impacted patient management were not eligible. Data Collection Each local investigator was responsible for collecting after informed consent and sharing individual anonymized data, which were centralized until April 28, 2016. The list of collected data is detailed in the study protocol (Supplementary Methods, available online). To take into account the heterogeneity between studies in terms of the number and timing of blood sampling and the duration of NCT, the protocol distinguished four different time points for CTC collection: five or fewer weeks prior to NCT initiation; one to eight weeks after NCT initiation; five or fewer weeks prior to surgery; one to 52 weeks after surgery. Data files were monitored manually for eligibility, and queries were sent to centers whenever necessary. Data were then merged into a centralized repository accessible only to the study secretariat and statisticians. No financial compensation was provided to individuals or participating centers. As per French law, this in silico study of anonymized data was declared to the Commission Nationale de l’Informatique et des Libertés (French data protection authority; CNIL number 1873422v0). Statistical Analysis The meta-analytical fixed effect model for time-to-event metadata used to obtain an overall hazard ratio was a Cox model stratified by study (14,15). The primary end point of this study was overall survival, defined as the time from NCT initiation to death from any cause. Distant disease–free survival was defined as the time from NCT initiation to distant recurrence or death from any cause, whichever came first. Locoregional relapse–free interval was defined as the time from NCT initiation to locoregional relapse; patients who died or patients with no documented evidence of locoregional relapse were censored. We explored the heterogeneity in regards to CTC count distribution across centers (Supplementary Methods, available online) in addition to the heterogeneity in regards to CTC detection on survival. Between-study heterogeneity in regards to CTC detection on survival was assessed using chi-square and I2 statistics. Results were obtained based on a Cox regression model, stratified by study, with the following covariates: CTC (using the two or more CTCs threshold) and interaction between CTC and study. I2 represents the proportion of total variation in study estimates that is due to heterogeneity rather than sampling error. I2 values vary between 0% and 100%; I2 values of less than 30% correspond to low heterogeneity. Fisher exact tests with Monte Carlo approximation and Kruskal-Wallis tests were used to investigate associations of population characteristics with CTC count. The landmark method was used to assess the prognostic effects of the last CTC count (before surgery) and pathological complete response, with the landmark set at surgery; in these analyses conducted in patients with no event prior to surgery, the abovementioned survival times were calculated from the date of surgery. The Kaplan-Meier method was used to estimate survival curves. Cubic restricted splines of 2 degrees of freedom were used to model potential nonlinear effects of continuous CTC in the Cox regression model. To assess the added value of CTC count to clinicopathological variables, likelihood ratio (LR) statistics in Cox regression models stratified by study were used to estimate the added value of CTCs to clinicopathological models, which included a list of prespecified variables known to be prognostic in this setting (age <50 vs >50 years, tumor size T1–2, T3–4, T4d, lymph node status cN0 vs cN1–3, tumor grade 1–2 vs 3, tumor subgroup hormone receptor–positive/human epidermal growth factor receptor–negative [HER2-], HER2+, triple-negative). To control for overfitting of clinicopathological models and to allow unbiased estimates of the added value of CTCs, the data set was randomly divided 500 times into a training and validation series using a previously described method (4). Clinicopathological models were fitted to the training series, and we calculated the average increases in LR statistic (χ² LR value and associated P value) and concordance index on the validation series with 95% confidence intervals (CIs) based on the percentiles of the 500 resamples. Logistic regression models were used to investigate associations with pCR status. P values of less than .05 were considered statistically significant, and all statistical tests were two-sided. SAS (version 9.3) and R (version 3.3.2) were used for statistical analyses. Results Included Data Data for 2156 patients from 16 centers and 21 studies were collected; this report focuses on 2030 patients with CTC detection before NCT and/or before surgery (Figure 1; Supplementary Figure 1 and Supplementary Table 1, available online). The blood volume screened for CTC was 7.5 mL for 1670 patients (77.5%, 19 studies), 15 mL for 460 patients (21.3%, one study: GueparQuinto), and 30 mL for 26 patients (1.2%, one study: UMMC2003-045). We took into account the absolute number of CTCs detected, as 1) there were no statistically significant differences in CTC detection rate between patients with one blood tube and those with more than one blood tube screened at each time point; and 2) as reported beyond, only modest evidence of heterogeneity was found in this meta-analysis. Figure 1. View largeDownload slide Study flow chart. Overall, 2030 patients had circulating tumor cells (CTCs) detected prior to neoadjuvant chemotherapy (NCT) start and/or prior to surgery. Because of the lower number of collected data, results with CTC detection at one to eight weeks after NCT start are not reported here. CTC = circulating tumor cells; DDFS = distant disease–free survival; LRFI = locoregional relapse–free interval; NCT = neoadjuvant chemotherapy; OS = overall survival. Figure 1. View largeDownload slide Study flow chart. Overall, 2030 patients had circulating tumor cells (CTCs) detected prior to neoadjuvant chemotherapy (NCT) start and/or prior to surgery. Because of the lower number of collected data, results with CTC detection at one to eight weeks after NCT start are not reported here. CTC = circulating tumor cells; DDFS = distant disease–free survival; LRFI = locoregional relapse–free interval; NCT = neoadjuvant chemotherapy; OS = overall survival. CTC Detection and Clinicopathological Characteristics Among the 1574 patients with CTC count available before NCT, CTC count ranged from 0 to 559 (median = 0 CTCs; third quartile = 1 CTC): 398 (25.2%), 199 (12.6%), and 93 (5.9%) patients had one or more CTCs, two or more CTCs, and five or more CTCs detected, respectively (Supplementary Tables 2–4, available online). Patient characteristics and their association with CTC detection before NCT are shown in Table 1 (and in Supplementary Tables 2–5, available online). A prominent association was observed between CTC count and tumor size before NCT (P < .001), as higher detection rates were observed in inflammatory (cT4d) breast cancers. This association was expected, and the study protocol was designed to distinguish inflammatory from noninflammatory breast cancers. After excluding T4d cancers from the analysis, no statistically significant associations were observed between CTC detection rates and baseline characteristics (with the exception of tumor size and five or more CTCs, P = .049) (Supplementary Tables 6–9, available online). In-depth modeling of CTC count distribution supported the homogeneity assumption across centers (Supplementary Figures 2 and 3, available online). Table 1. Patient characteristics and circulating tumor cell detection before neoadjuvant chemotherapy Characteristics No. of patients (%) ≥1 CTCs threshold Continuous CTC count P 0 CTCs No. (%) ≥1 CTCs No. (%) P Age, y 1574 (100) .24* .19†  ≤50 843 (53.6) 746 (88.5) 97 (11.5)  >50 731 (46.4) 629 (86.0) 102 (14.0) Tumor size 1547 (100) <.001‡ <.001†  cT1 122 (7.9) 99 (81.1) 23 (18.9)  cT2 770 (49.8) 598 (77.7) 172 (22.3)  cT3 343 (22.2) 260 (75.8) 83 (24.2)  cT4a–c 108 (7.0) 77 (71.3) 31 (28.7)  cT4d 204 (13.2) 120 (58.8) 84 (41.2) Lymph nodes 1567 (100) .05* .02†  cN0 656 (41.9) 507 (87.3) 149 (22.7)  cN1–3 911 (58.1) 664 (72.9) 247 (27.1) Tumor grade 1371 (100) .51* .43†  I 65 (4.7) 52 (80.0) 13 (20.0)  II 702 (51.2) 514 (73.2) 188 (26.8)  III 604 (44.1) 443 (73.3) 161 (26.7) Tumor subgroup 1570 (100) .23* .12†  HER2+ 365 (23.2) 277 (75.9) 88 (24.1)  HR+ HER2- 800 (51.0) 607 (75.9) 193 (24.1)  HR- HER2- 405 (25.8) 290 (71.6) 115 (28.4) Characteristics No. of patients (%) ≥1 CTCs threshold Continuous CTC count P 0 CTCs No. (%) ≥1 CTCs No. (%) P Age, y 1574 (100) .24* .19†  ≤50 843 (53.6) 746 (88.5) 97 (11.5)  >50 731 (46.4) 629 (86.0) 102 (14.0) Tumor size 1547 (100) <.001‡ <.001†  cT1 122 (7.9) 99 (81.1) 23 (18.9)  cT2 770 (49.8) 598 (77.7) 172 (22.3)  cT3 343 (22.2) 260 (75.8) 83 (24.2)  cT4a–c 108 (7.0) 77 (71.3) 31 (28.7)  cT4d 204 (13.2) 120 (58.8) 84 (41.2) Lymph nodes 1567 (100) .05* .02†  cN0 656 (41.9) 507 (87.3) 149 (22.7)  cN1–3 911 (58.1) 664 (72.9) 247 (27.1) Tumor grade 1371 (100) .51* .43†  I 65 (4.7) 52 (80.0) 13 (20.0)  II 702 (51.2) 514 (73.2) 188 (26.8)  III 604 (44.1) 443 (73.3) 161 (26.7) Tumor subgroup 1570 (100) .23* .12†  HER2+ 365 (23.2) 277 (75.9) 88 (24.1)  HR+ HER2- 800 (51.0) 607 (75.9) 193 (24.1)  HR- HER2- 405 (25.8) 290 (71.6) 115 (28.4) * P values were calculated using a two-sided Fisher exact test. CTC = circulating tumor cell; HER2 = human epidermal growth factor receptor 2; HR+ = hormone receptor positive; HR- = hormone receptor negative. † P values were calculated using a two-sided Kruskal-Wallis test. ‡ P values were calculated using a two-sided Fisher exact test with Monte Carlo sampling. Table 1. Patient characteristics and circulating tumor cell detection before neoadjuvant chemotherapy Characteristics No. of patients (%) ≥1 CTCs threshold Continuous CTC count P 0 CTCs No. (%) ≥1 CTCs No. (%) P Age, y 1574 (100) .24* .19†  ≤50 843 (53.6) 746 (88.5) 97 (11.5)  >50 731 (46.4) 629 (86.0) 102 (14.0) Tumor size 1547 (100) <.001‡ <.001†  cT1 122 (7.9) 99 (81.1) 23 (18.9)  cT2 770 (49.8) 598 (77.7) 172 (22.3)  cT3 343 (22.2) 260 (75.8) 83 (24.2)  cT4a–c 108 (7.0) 77 (71.3) 31 (28.7)  cT4d 204 (13.2) 120 (58.8) 84 (41.2) Lymph nodes 1567 (100) .05* .02†  cN0 656 (41.9) 507 (87.3) 149 (22.7)  cN1–3 911 (58.1) 664 (72.9) 247 (27.1) Tumor grade 1371 (100) .51* .43†  I 65 (4.7) 52 (80.0) 13 (20.0)  II 702 (51.2) 514 (73.2) 188 (26.8)  III 604 (44.1) 443 (73.3) 161 (26.7) Tumor subgroup 1570 (100) .23* .12†  HER2+ 365 (23.2) 277 (75.9) 88 (24.1)  HR+ HER2- 800 (51.0) 607 (75.9) 193 (24.1)  HR- HER2- 405 (25.8) 290 (71.6) 115 (28.4) Characteristics No. of patients (%) ≥1 CTCs threshold Continuous CTC count P 0 CTCs No. (%) ≥1 CTCs No. (%) P Age, y 1574 (100) .24* .19†  ≤50 843 (53.6) 746 (88.5) 97 (11.5)  >50 731 (46.4) 629 (86.0) 102 (14.0) Tumor size 1547 (100) <.001‡ <.001†  cT1 122 (7.9) 99 (81.1) 23 (18.9)  cT2 770 (49.8) 598 (77.7) 172 (22.3)  cT3 343 (22.2) 260 (75.8) 83 (24.2)  cT4a–c 108 (7.0) 77 (71.3) 31 (28.7)  cT4d 204 (13.2) 120 (58.8) 84 (41.2) Lymph nodes 1567 (100) .05* .02†  cN0 656 (41.9) 507 (87.3) 149 (22.7)  cN1–3 911 (58.1) 664 (72.9) 247 (27.1) Tumor grade 1371 (100) .51* .43†  I 65 (4.7) 52 (80.0) 13 (20.0)  II 702 (51.2) 514 (73.2) 188 (26.8)  III 604 (44.1) 443 (73.3) 161 (26.7) Tumor subgroup 1570 (100) .23* .12†  HER2+ 365 (23.2) 277 (75.9) 88 (24.1)  HR+ HER2- 800 (51.0) 607 (75.9) 193 (24.1)  HR- HER2- 405 (25.8) 290 (71.6) 115 (28.4) * P values were calculated using a two-sided Fisher exact test. CTC = circulating tumor cell; HER2 = human epidermal growth factor receptor 2; HR+ = hormone receptor positive; HR- = hormone receptor negative. † P values were calculated using a two-sided Kruskal-Wallis test. ‡ P values were calculated using a two-sided Fisher exact test with Monte Carlo sampling. Among 1200 patients with CTC count available before surgery, CTC count was lower than the baseline count (P < .001) and ranged from 0 to 51 (median = 0 CTCs; third quartile = 0 CTCs): 181 (15.1%), 64 (5.3%), and 12 (1.0%) patients had one or more CTCs, two or more CTCs, and five or more CTCs detected, respectively (Supplementary Tables 10–12, available online). No statistically significant association was observed between CTC detection before surgery and patient characteristics (Supplementary Table 10–13, available online). CTC Detection and Pathological Complete Response Patients achieving a pathological complete response (defined as ypT0/isN0 in >90% of patients) (Supplementary Table 1, available online) had a better overall survival (multivariable analysis; hazard ratio [HR] = 0.19, 95% CI = 0.10 to 0.31, P < .001), distant metastasis–free survival (HR = 0.25, 95% CI = 0.16 to 0.38, P < .001), and locoregional relapse–free interval (HR = 0.49, 95% CI = 0.29 to 0.80, P = .004). In the population of patients without inflammatory cancers, a slightly higher rate of pathological complete response was observed in patients with no CTC compared with patients with one or more CTCs before NCT (24.2%, 95% CI = 21.7 to 26.9 vs 17.4%, 95% CI = 13.3 to 22.2, P = .01) (Supplementary Table 14, available online). However, this association between pathological complete response and CTC detection was not observed at other cutoffs and time points (Supplementary Tables 14–17, available online). Multivariable logistic regression models did not find any statistically significant impact on pathological complete response of CTC detection before NCT or before surgery (data not shown). CTC Detection and Postneoadjuvant Outcome With a median follow-up of 59.1 (range = 0.9–125.7, interquartile range = 39.3–70.2) months, 301, 418, and 157 overall survival, distant disease–free survival, and locoregional relapse–free interval events were observed, respectively. CTC count before NCT was a statistically significant prognostic factor in univariate analyses for overall survival (P < .001), distant disease–free survival (P < .001), and locoregional relapse–free interval (P < .001) (Figure 2, A–C). Patients with one, two, three to four, and five or more CTCs before neoadjuvant chemotherapy displayed hazard ratios of death (95% CI) of 1.09 (0.65 to 1.69), 2.63 (1.42 to 4.54), 3.83 (2.08 to 6.66), and 6.25 (4.34 to 9.09), respectively (Supplementary Table 18, available online). Hazard ratios for distant disease–free survival and locoregional relapse–free interval are displayed in Supplementary Table 18 (available online). The observed prognostic impact of CTC count did not change after removing T4d tumors from the analyses (Supplementary Table 19, available online). Using the fewer than two vs two or more CTCs categorization, the overall pooled estimate of having two or more CTCs vs fewer than two CTCs was given by a hazard ratio (95% CI) of 4.36 (3.19 to 5.90) for overall survival, 3.89 (3.00 to 5.03) for distant distant–free survival, and 3.30 (2.12 to 5.03) for locoregional relapse–free interval (Supplementary Table 18, available online). CTC count before surgery was also a statistically significant prognostic factor for overall survival and distant disease–free survival (Figure 2, D–F; Supplementary Tables 20 and 21, available online). There was only modest evidence for between-study heterogeneity of the prognostic effect of two or more CTCs on overall (I2 = 0%), distant disease–free (I2 = 0%), and locoregional relapse–free interval (I2 = 34.1%) in the whole patient population (Supplementary Table 22, available online). Multivariable analyses further confirmed CTC count before NCT as an independent prognostic factor for survival (Table 2; Supplementary Tables 23 and 24, available online). Table 2. Multivariable survival analyses* Variable OS DDFS LRFI HR (95% CI) P† HR (95% CI) P† HR (95% CI) P† Age, y .04 .004 .08  ≤50 0.72 (0.53 to 0.99) 0.69 (0.53 to 0.89) 0.70 (0.47 to 1.04)  >50 1.00 (reference) 1.00 (reference) 1.00 (reference) Tumor size .001 .0003 .48  T1–T2 1.00 (reference) 1.00 (reference) 1.00 (reference)  T3–T4 1.60 (1.11 to 2.31) 1.56 (1.15 to 2.12) 1.16 (0.70 to 1.89)  T4d 2.85 (1.51 to 5.11) 2.65 (1.54 to 4.37) 1.69 (0.66 to 3.74) Lymph nodes <.001 <.001 .02  cN0 1.00 (reference) 1.00 (reference) 1.00 (reference)  cN1–3 1.94 (1.39 to 2.76) 1.63 (1.07 to 2.54) 1.63 (1.07 to 2.54) Tumor grade .72 .94 .05  1–2 1.00 (reference) 1.00 (reference) 1.00 (reference)  3 1.06 (0.76 to 1.50) 0.99 (0.75 to 1.31) 1.53 (0.99 to 2.38) Tumor subgroup <.001 <.001 <.001  HR+ HER2- 1.00 (reference) 1.00 (reference) 1.00 (reference)  HER2+ 1.29 (0.80 to 2.04) 1.05 (0.72 to 1.51) 2.31 (1.28 to 4.19)  HR- HER2- 3.92 (2.65 to 5.84) 2.14 (1.55 to 2.96) 4.35 (2.51 to 7.69) CTC count before NCT <.001 <.001 <.001  0–1 1.00 (reference) 1.00 (reference) 1.00 (reference)  ≥2 3.93 (2.81 to 5.45) 3.73 (2.82 to 4.90) 3.02 (1.88 to 4.75) Variable OS DDFS LRFI HR (95% CI) P† HR (95% CI) P† HR (95% CI) P† Age, y .04 .004 .08  ≤50 0.72 (0.53 to 0.99) 0.69 (0.53 to 0.89) 0.70 (0.47 to 1.04)  >50 1.00 (reference) 1.00 (reference) 1.00 (reference) Tumor size .001 .0003 .48  T1–T2 1.00 (reference) 1.00 (reference) 1.00 (reference)  T3–T4 1.60 (1.11 to 2.31) 1.56 (1.15 to 2.12) 1.16 (0.70 to 1.89)  T4d 2.85 (1.51 to 5.11) 2.65 (1.54 to 4.37) 1.69 (0.66 to 3.74) Lymph nodes <.001 <.001 .02  cN0 1.00 (reference) 1.00 (reference) 1.00 (reference)  cN1–3 1.94 (1.39 to 2.76) 1.63 (1.07 to 2.54) 1.63 (1.07 to 2.54) Tumor grade .72 .94 .05  1–2 1.00 (reference) 1.00 (reference) 1.00 (reference)  3 1.06 (0.76 to 1.50) 0.99 (0.75 to 1.31) 1.53 (0.99 to 2.38) Tumor subgroup <.001 <.001 <.001  HR+ HER2- 1.00 (reference) 1.00 (reference) 1.00 (reference)  HER2+ 1.29 (0.80 to 2.04) 1.05 (0.72 to 1.51) 2.31 (1.28 to 4.19)  HR- HER2- 3.92 (2.65 to 5.84) 2.14 (1.55 to 2.96) 4.35 (2.51 to 7.69) CTC count before NCT <.001 <.001 <.001  0–1 1.00 (reference) 1.00 (reference) 1.00 (reference)  ≥2 3.93 (2.81 to 5.45) 3.73 (2.82 to 4.90) 3.02 (1.88 to 4.75) * Multivariable analyses on overall survival (n = 1362 patients, n = 184 events), distant disease–free survival (n = 1362 patients, n = 267 events), and locoregional relapse–free survival (n = 1362 patients, n = 110 events). Multivariable analyses taking into account CTC and T4d interaction, as well as CTC as a continuous variable, are displayed in Supplementary Material 8 (available online). CTC = circulating tumor cell; DDFS = distant disease–free survival; HER2 = human epidermal growth factor receptor 2; HR+ = hormone receptor positive; HR- = hormone receptor negative; LRFI = locoregional relapse–free interval; NCT = neoadjuvant chemotherapy; OS = overall survival. † Two-sided P values correspond to Wald tests. Table 2. Multivariable survival analyses* Variable OS DDFS LRFI HR (95% CI) P† HR (95% CI) P† HR (95% CI) P† Age, y .04 .004 .08  ≤50 0.72 (0.53 to 0.99) 0.69 (0.53 to 0.89) 0.70 (0.47 to 1.04)  >50 1.00 (reference) 1.00 (reference) 1.00 (reference) Tumor size .001 .0003 .48  T1–T2 1.00 (reference) 1.00 (reference) 1.00 (reference)  T3–T4 1.60 (1.11 to 2.31) 1.56 (1.15 to 2.12) 1.16 (0.70 to 1.89)  T4d 2.85 (1.51 to 5.11) 2.65 (1.54 to 4.37) 1.69 (0.66 to 3.74) Lymph nodes <.001 <.001 .02  cN0 1.00 (reference) 1.00 (reference) 1.00 (reference)  cN1–3 1.94 (1.39 to 2.76) 1.63 (1.07 to 2.54) 1.63 (1.07 to 2.54) Tumor grade .72 .94 .05  1–2 1.00 (reference) 1.00 (reference) 1.00 (reference)  3 1.06 (0.76 to 1.50) 0.99 (0.75 to 1.31) 1.53 (0.99 to 2.38) Tumor subgroup <.001 <.001 <.001  HR+ HER2- 1.00 (reference) 1.00 (reference) 1.00 (reference)  HER2+ 1.29 (0.80 to 2.04) 1.05 (0.72 to 1.51) 2.31 (1.28 to 4.19)  HR- HER2- 3.92 (2.65 to 5.84) 2.14 (1.55 to 2.96) 4.35 (2.51 to 7.69) CTC count before NCT <.001 <.001 <.001  0–1 1.00 (reference) 1.00 (reference) 1.00 (reference)  ≥2 3.93 (2.81 to 5.45) 3.73 (2.82 to 4.90) 3.02 (1.88 to 4.75) Variable OS DDFS LRFI HR (95% CI) P† HR (95% CI) P† HR (95% CI) P† Age, y .04 .004 .08  ≤50 0.72 (0.53 to 0.99) 0.69 (0.53 to 0.89) 0.70 (0.47 to 1.04)  >50 1.00 (reference) 1.00 (reference) 1.00 (reference) Tumor size .001 .0003 .48  T1–T2 1.00 (reference) 1.00 (reference) 1.00 (reference)  T3–T4 1.60 (1.11 to 2.31) 1.56 (1.15 to 2.12) 1.16 (0.70 to 1.89)  T4d 2.85 (1.51 to 5.11) 2.65 (1.54 to 4.37) 1.69 (0.66 to 3.74) Lymph nodes <.001 <.001 .02  cN0 1.00 (reference) 1.00 (reference) 1.00 (reference)  cN1–3 1.94 (1.39 to 2.76) 1.63 (1.07 to 2.54) 1.63 (1.07 to 2.54) Tumor grade .72 .94 .05  1–2 1.00 (reference) 1.00 (reference) 1.00 (reference)  3 1.06 (0.76 to 1.50) 0.99 (0.75 to 1.31) 1.53 (0.99 to 2.38) Tumor subgroup <.001 <.001 <.001  HR+ HER2- 1.00 (reference) 1.00 (reference) 1.00 (reference)  HER2+ 1.29 (0.80 to 2.04) 1.05 (0.72 to 1.51) 2.31 (1.28 to 4.19)  HR- HER2- 3.92 (2.65 to 5.84) 2.14 (1.55 to 2.96) 4.35 (2.51 to 7.69) CTC count before NCT <.001 <.001 <.001  0–1 1.00 (reference) 1.00 (reference) 1.00 (reference)  ≥2 3.93 (2.81 to 5.45) 3.73 (2.82 to 4.90) 3.02 (1.88 to 4.75) * Multivariable analyses on overall survival (n = 1362 patients, n = 184 events), distant disease–free survival (n = 1362 patients, n = 267 events), and locoregional relapse–free survival (n = 1362 patients, n = 110 events). Multivariable analyses taking into account CTC and T4d interaction, as well as CTC as a continuous variable, are displayed in Supplementary Material 8 (available online). CTC = circulating tumor cell; DDFS = distant disease–free survival; HER2 = human epidermal growth factor receptor 2; HR+ = hormone receptor positive; HR- = hormone receptor negative; LRFI = locoregional relapse–free interval; NCT = neoadjuvant chemotherapy; OS = overall survival. † Two-sided P values correspond to Wald tests. Figure 2. View largeDownload slide Survival curves according to circulating tumor cell (CTC) count. Survival curves and the corresponding number of patients at risk are displayed according to CTC count before neoadjuvant chemotherapy (NCT): zero CTCs, one CTC, two CTCs, three or four CTCs, five or more CTCs. P values were obtained by stratified log-rank tests and were two-sided. A) Overall survival (OS) and CTC count before NCT (in months; P < .001). B) Distant disease–free survival (DDFS) and CTC count before NCT (in months; P < .001). C) Locoregional relapse–free interval (LRFI) and CTC count before NCT (in months; P < .001). D) OS and CTC count before surgery (in months; univariate stratified P = .002). E) DDFS and CTC count before surgery (in months; P < .001). F) LRFI and CTC count before surgery (in months; P = .51). CTC = circulating tumor cell; DDFS = distant disease–free survival; LRFI = locoregional relapse–free interval; OS = overall survival. Figure 2. View largeDownload slide Survival curves according to circulating tumor cell (CTC) count. Survival curves and the corresponding number of patients at risk are displayed according to CTC count before neoadjuvant chemotherapy (NCT): zero CTCs, one CTC, two CTCs, three or four CTCs, five or more CTCs. P values were obtained by stratified log-rank tests and were two-sided. A) Overall survival (OS) and CTC count before NCT (in months; P < .001). B) Distant disease–free survival (DDFS) and CTC count before NCT (in months; P < .001). C) Locoregional relapse–free interval (LRFI) and CTC count before NCT (in months; P < .001). D) OS and CTC count before surgery (in months; univariate stratified P = .002). E) DDFS and CTC count before surgery (in months; P < .001). F) LRFI and CTC count before surgery (in months; P = .51). CTC = circulating tumor cell; DDFS = distant disease–free survival; LRFI = locoregional relapse–free interval; OS = overall survival. CTC Detection and Prognostic Models The addition of baseline CTC count modeled as a binary variable (fewer than two vs two or more CTCs) to clinicopathological prognostic models (Table 2) resulted in statistically significant increases in prognostication of overall survival (P < .001), distant disease–free survival (P < .001), and locoregional relapse–free interval (P = .008) (Table 3); similar results were obtained when modeling CTC count as a continuous variable (splines) (Table 3). In 861 patients with full data available, further investigations showed that adding baseline CTC count to models including clinicopathological data and pathological complete response status also increased postneoadjuvant survival prognostication (Table 3). Regarding CTC detection during NCT, we found that it did not increase survival prognostication (Table 3). Table 3. Assessment of added prognostic information of circulating tumor cells before and during neoadjuvant chemotherapy* Model 1 Model 1 average c-index Model 2 Model 2 average c-index Average c-index increase, model 2 – model 1 (95% CI) ddf Average increase χ2 (95% CI) P† Models with CTC count before NCT  OS (n = 1362)   CP 0.736 Model 1 + CTCBC (< or ≥2 CTC) 0.767 0.031 (–0.003 to 0.063) 1 28.9 (13.3 to 44.1) <.001   CP 0.741 Model 1 + CTCBC (splines) 0.779 0.038 (0.004 to 0.070) 2 30.0 (14.7 to 48.5) <.001  DDFS (n = 1362)   CP 0.666 Model 1 + CTCBC (< or ≥2 CTC) 0.705 0.039 (0.011 to 0.070) 1 38.7 (21.9 to 58.6) <.001   CP 0.664 Model 1 + CTCBC (splines) 0.709 0.045 (0.016 to 0.073) 2 41.2 (24.2 to 63.7) <.001  LRFI (n = 1362)   CP 0.728 Model 1 + CTCBC (< or ≥2 CTC) 0.734 0.006 (–0.021 to 0.031) 1 8.8 (2.4 to 19.2) .008   CP 0.727 Model 1 + CTCBC (splines) 0.731 0.004 (–0.032 to 0.032) 2 9.7 (1.9 to 20.7) .003 Models with CTC count before NCT and pCR (landmark analyses)  OS (n = 1332)   CP+ CTCBC (< or ≥2 CTC) 0.771 Model 1 + pCR 0.809 0.037 (0.008 to 0.064) 1 26.7 (13.1 to 44.0) <.001   CP+ CTCBC (splines) 0.778 Model 1 + pCR 0.813 0.035 (0.002 to 0.065) 1 18.8 (7.9 to 36.1) <.001   CP + pCR 0.774 Model 1 + CTCBC (< or ≥2 CTC) 0.809 0.035 (0.002 to 0.063) 1 30.3 (14.7 to 48.8) <.001   CP + pCR 0.774 Model 1 + CTCBC (splines) 0.813 0.039 (0.004 to 0.072) 2 30.5 (14.4 to 47.8) <.001  DDFS (n = 1323)   CP+ CTCBC (< or ≥2 CTC) 0.698 Model 1 + pCR 0.746 0.048 (0.019 to 0.072) 1 27.9 (13.0 to 45.0) <.001   CP+ CTCBC (splines) 0.704 Model 1 + pCR 0.751 0.047 (0.016 to 0.070) 1 21.8 (10.0 to 37.3) <.001   CP + pCR 0.703 Model 1 + CTCBC (< or ≥2 CTC) 0.746 0.044 (0.012 to 0.068) 1 36.3 (19.7 to 55.4) <.001   CP + pCR 0.703 Model 1 + CTCBC (splines) 0.751 0.048 (0.018 to 0.074) 2 37.1 (21.1 to 57.1) <.001  LRFI (n = 1322)   CP+ CTCBC (< or ≥2 CTC) 0.726 Model 1 + pCR 0.728 0.002 (–0.042 to 0.031) 1 4.90 (0.30 to 13.1) .03   CP+ CTCBC (splines) 0.724 Model 1 + pCR 0.726 0.003 (–0.041 to 0.029) 1 2.6 (0.01 to 9.61) .10   CP + pCR 0.729 Model 1 + CTCBC (< or ≥2 CTC) 0.728 –0.001 (–0.031 to 0.024) 1 9.1 (1.74 to 18.7) .002   CP + pCR 0.729 Model 1 + CTCBC (splines) 0.726 –0.003 (–0.037 to 0.021) 2 8.7 (1.7 to 18.9) .02 Models with CTC count before NCT, pCR, and CTC count during NCT (landmark analyses)  OS (n = 861)   CP+CTCBC+pCR 0.804 Model 1 + CTCDC (< or ≥2 CTC) 0.806 0.002 (–0.016 to 0.013) 1 3.1 (0.0 to 11.4) .08   CP+CTCBC+pCR 0.798 Model 1 + CTCDC (splines) 0.796 –0.003 (–0.038 to 0.080) 2 5.0 (0.2 to 12.9) .02  DDFS (n = 856)   CP+CTCBC+pCR 0.758 Model 1 + CTCDC (< or ≥2 CTC) 0.760 0.002 (–0.013 to 0.014) 1 3.1 (0.0 to 9.7) .08   CP+CTCBC+pCR 0.756 Model 1 + CTCDC (splines) 0.755 –0.001 (–0.016 to 0.007) 2 2.7 (0.02 to 10.8) .09  LRFI (n = 855)   CP+CTCBC+pCR 0.717 Model 1 + CTCDC (< or ≥2 CTC) 0.707 –0.011 (–0.074 to 0.006) 1 0.9 (0.0 to 3.6) .34   CP+CTCBC+pCR 0.725 Model 1 + CTCDC (splines) 0.715 –0.010 (–0.047 to 0.006) 2 0.6 (0.0 to 2.8) .44 Model 1 Model 1 average c-index Model 2 Model 2 average c-index Average c-index increase, model 2 – model 1 (95% CI) ddf Average increase χ2 (95% CI) P† Models with CTC count before NCT  OS (n = 1362)   CP 0.736 Model 1 + CTCBC (< or ≥2 CTC) 0.767 0.031 (–0.003 to 0.063) 1 28.9 (13.3 to 44.1) <.001   CP 0.741 Model 1 + CTCBC (splines) 0.779 0.038 (0.004 to 0.070) 2 30.0 (14.7 to 48.5) <.001  DDFS (n = 1362)   CP 0.666 Model 1 + CTCBC (< or ≥2 CTC) 0.705 0.039 (0.011 to 0.070) 1 38.7 (21.9 to 58.6) <.001   CP 0.664 Model 1 + CTCBC (splines) 0.709 0.045 (0.016 to 0.073) 2 41.2 (24.2 to 63.7) <.001  LRFI (n = 1362)   CP 0.728 Model 1 + CTCBC (< or ≥2 CTC) 0.734 0.006 (–0.021 to 0.031) 1 8.8 (2.4 to 19.2) .008   CP 0.727 Model 1 + CTCBC (splines) 0.731 0.004 (–0.032 to 0.032) 2 9.7 (1.9 to 20.7) .003 Models with CTC count before NCT and pCR (landmark analyses)  OS (n = 1332)   CP+ CTCBC (< or ≥2 CTC) 0.771 Model 1 + pCR 0.809 0.037 (0.008 to 0.064) 1 26.7 (13.1 to 44.0) <.001   CP+ CTCBC (splines) 0.778 Model 1 + pCR 0.813 0.035 (0.002 to 0.065) 1 18.8 (7.9 to 36.1) <.001   CP + pCR 0.774 Model 1 + CTCBC (< or ≥2 CTC) 0.809 0.035 (0.002 to 0.063) 1 30.3 (14.7 to 48.8) <.001   CP + pCR 0.774 Model 1 + CTCBC (splines) 0.813 0.039 (0.004 to 0.072) 2 30.5 (14.4 to 47.8) <.001  DDFS (n = 1323)   CP+ CTCBC (< or ≥2 CTC) 0.698 Model 1 + pCR 0.746 0.048 (0.019 to 0.072) 1 27.9 (13.0 to 45.0) <.001   CP+ CTCBC (splines) 0.704 Model 1 + pCR 0.751 0.047 (0.016 to 0.070) 1 21.8 (10.0 to 37.3) <.001   CP + pCR 0.703 Model 1 + CTCBC (< or ≥2 CTC) 0.746 0.044 (0.012 to 0.068) 1 36.3 (19.7 to 55.4) <.001   CP + pCR 0.703 Model 1 + CTCBC (splines) 0.751 0.048 (0.018 to 0.074) 2 37.1 (21.1 to 57.1) <.001  LRFI (n = 1322)   CP+ CTCBC (< or ≥2 CTC) 0.726 Model 1 + pCR 0.728 0.002 (–0.042 to 0.031) 1 4.90 (0.30 to 13.1) .03   CP+ CTCBC (splines) 0.724 Model 1 + pCR 0.726 0.003 (–0.041 to 0.029) 1 2.6 (0.01 to 9.61) .10   CP + pCR 0.729 Model 1 + CTCBC (< or ≥2 CTC) 0.728 –0.001 (–0.031 to 0.024) 1 9.1 (1.74 to 18.7) .002   CP + pCR 0.729 Model 1 + CTCBC (splines) 0.726 –0.003 (–0.037 to 0.021) 2 8.7 (1.7 to 18.9) .02 Models with CTC count before NCT, pCR, and CTC count during NCT (landmark analyses)  OS (n = 861)   CP+CTCBC+pCR 0.804 Model 1 + CTCDC (< or ≥2 CTC) 0.806 0.002 (–0.016 to 0.013) 1 3.1 (0.0 to 11.4) .08   CP+CTCBC+pCR 0.798 Model 1 + CTCDC (splines) 0.796 –0.003 (–0.038 to 0.080) 2 5.0 (0.2 to 12.9) .02  DDFS (n = 856)   CP+CTCBC+pCR 0.758 Model 1 + CTCDC (< or ≥2 CTC) 0.760 0.002 (–0.013 to 0.014) 1 3.1 (0.0 to 9.7) .08   CP+CTCBC+pCR 0.756 Model 1 + CTCDC (splines) 0.755 –0.001 (–0.016 to 0.007) 2 2.7 (0.02 to 10.8) .09  LRFI (n = 855)   CP+CTCBC+pCR 0.717 Model 1 + CTCDC (< or ≥2 CTC) 0.707 –0.011 (–0.074 to 0.006) 1 0.9 (0.0 to 3.6) .34   CP+CTCBC+pCR 0.725 Model 1 + CTCDC (splines) 0.715 –0.010 (–0.047 to 0.006) 2 0.6 (0.0 to 2.8) .44 * For each analysis, the number of included patients corresponds to the number of patients with available data. Clinical and pathological characteristics were not statistically different among the different patient populations included in these analyses. CI = confidence interval; CP = baseline clinicopathological model; CTC = circulating tumor cells; CTCBC = CTC count before NCT; CTCDC = CTC count during NCT, using the highest value of CTC detection either after NCT start or prior to surgery; ddf = degree of freedom; DDFS = distant disease–free survival; LRFI = locoregional relapse–free interval; NCT = neoadjuvant chemotherapy; OS = overall survival; pCR = pathological complete response. † Two-sided likelihood ratio test P value corresponds to the average increase in χ2. Table 3. Assessment of added prognostic information of circulating tumor cells before and during neoadjuvant chemotherapy* Model 1 Model 1 average c-index Model 2 Model 2 average c-index Average c-index increase, model 2 – model 1 (95% CI) ddf Average increase χ2 (95% CI) P† Models with CTC count before NCT  OS (n = 1362)   CP 0.736 Model 1 + CTCBC (< or ≥2 CTC) 0.767 0.031 (–0.003 to 0.063) 1 28.9 (13.3 to 44.1) <.001   CP 0.741 Model 1 + CTCBC (splines) 0.779 0.038 (0.004 to 0.070) 2 30.0 (14.7 to 48.5) <.001  DDFS (n = 1362)   CP 0.666 Model 1 + CTCBC (< or ≥2 CTC) 0.705 0.039 (0.011 to 0.070) 1 38.7 (21.9 to 58.6) <.001   CP 0.664 Model 1 + CTCBC (splines) 0.709 0.045 (0.016 to 0.073) 2 41.2 (24.2 to 63.7) <.001  LRFI (n = 1362)   CP 0.728 Model 1 + CTCBC (< or ≥2 CTC) 0.734 0.006 (–0.021 to 0.031) 1 8.8 (2.4 to 19.2) .008   CP 0.727 Model 1 + CTCBC (splines) 0.731 0.004 (–0.032 to 0.032) 2 9.7 (1.9 to 20.7) .003 Models with CTC count before NCT and pCR (landmark analyses)  OS (n = 1332)   CP+ CTCBC (< or ≥2 CTC) 0.771 Model 1 + pCR 0.809 0.037 (0.008 to 0.064) 1 26.7 (13.1 to 44.0) <.001   CP+ CTCBC (splines) 0.778 Model 1 + pCR 0.813 0.035 (0.002 to 0.065) 1 18.8 (7.9 to 36.1) <.001   CP + pCR 0.774 Model 1 + CTCBC (< or ≥2 CTC) 0.809 0.035 (0.002 to 0.063) 1 30.3 (14.7 to 48.8) <.001   CP + pCR 0.774 Model 1 + CTCBC (splines) 0.813 0.039 (0.004 to 0.072) 2 30.5 (14.4 to 47.8) <.001  DDFS (n = 1323)   CP+ CTCBC (< or ≥2 CTC) 0.698 Model 1 + pCR 0.746 0.048 (0.019 to 0.072) 1 27.9 (13.0 to 45.0) <.001   CP+ CTCBC (splines) 0.704 Model 1 + pCR 0.751 0.047 (0.016 to 0.070) 1 21.8 (10.0 to 37.3) <.001   CP + pCR 0.703 Model 1 + CTCBC (< or ≥2 CTC) 0.746 0.044 (0.012 to 0.068) 1 36.3 (19.7 to 55.4) <.001   CP + pCR 0.703 Model 1 + CTCBC (splines) 0.751 0.048 (0.018 to 0.074) 2 37.1 (21.1 to 57.1) <.001  LRFI (n = 1322)   CP+ CTCBC (< or ≥2 CTC) 0.726 Model 1 + pCR 0.728 0.002 (–0.042 to 0.031) 1 4.90 (0.30 to 13.1) .03   CP+ CTCBC (splines) 0.724 Model 1 + pCR 0.726 0.003 (–0.041 to 0.029) 1 2.6 (0.01 to 9.61) .10   CP + pCR 0.729 Model 1 + CTCBC (< or ≥2 CTC) 0.728 –0.001 (–0.031 to 0.024) 1 9.1 (1.74 to 18.7) .002   CP + pCR 0.729 Model 1 + CTCBC (splines) 0.726 –0.003 (–0.037 to 0.021) 2 8.7 (1.7 to 18.9) .02 Models with CTC count before NCT, pCR, and CTC count during NCT (landmark analyses)  OS (n = 861)   CP+CTCBC+pCR 0.804 Model 1 + CTCDC (< or ≥2 CTC) 0.806 0.002 (–0.016 to 0.013) 1 3.1 (0.0 to 11.4) .08   CP+CTCBC+pCR 0.798 Model 1 + CTCDC (splines) 0.796 –0.003 (–0.038 to 0.080) 2 5.0 (0.2 to 12.9) .02  DDFS (n = 856)   CP+CTCBC+pCR 0.758 Model 1 + CTCDC (< or ≥2 CTC) 0.760 0.002 (–0.013 to 0.014) 1 3.1 (0.0 to 9.7) .08   CP+CTCBC+pCR 0.756 Model 1 + CTCDC (splines) 0.755 –0.001 (–0.016 to 0.007) 2 2.7 (0.02 to 10.8) .09  LRFI (n = 855)   CP+CTCBC+pCR 0.717 Model 1 + CTCDC (< or ≥2 CTC) 0.707 –0.011 (–0.074 to 0.006) 1 0.9 (0.0 to 3.6) .34   CP+CTCBC+pCR 0.725 Model 1 + CTCDC (splines) 0.715 –0.010 (–0.047 to 0.006) 2 0.6 (0.0 to 2.8) .44 Model 1 Model 1 average c-index Model 2 Model 2 average c-index Average c-index increase, model 2 – model 1 (95% CI) ddf Average increase χ2 (95% CI) P† Models with CTC count before NCT  OS (n = 1362)   CP 0.736 Model 1 + CTCBC (< or ≥2 CTC) 0.767 0.031 (–0.003 to 0.063) 1 28.9 (13.3 to 44.1) <.001   CP 0.741 Model 1 + CTCBC (splines) 0.779 0.038 (0.004 to 0.070) 2 30.0 (14.7 to 48.5) <.001  DDFS (n = 1362)   CP 0.666 Model 1 + CTCBC (< or ≥2 CTC) 0.705 0.039 (0.011 to 0.070) 1 38.7 (21.9 to 58.6) <.001   CP 0.664 Model 1 + CTCBC (splines) 0.709 0.045 (0.016 to 0.073) 2 41.2 (24.2 to 63.7) <.001  LRFI (n = 1362)   CP 0.728 Model 1 + CTCBC (< or ≥2 CTC) 0.734 0.006 (–0.021 to 0.031) 1 8.8 (2.4 to 19.2) .008   CP 0.727 Model 1 + CTCBC (splines) 0.731 0.004 (–0.032 to 0.032) 2 9.7 (1.9 to 20.7) .003 Models with CTC count before NCT and pCR (landmark analyses)  OS (n = 1332)   CP+ CTCBC (< or ≥2 CTC) 0.771 Model 1 + pCR 0.809 0.037 (0.008 to 0.064) 1 26.7 (13.1 to 44.0) <.001   CP+ CTCBC (splines) 0.778 Model 1 + pCR 0.813 0.035 (0.002 to 0.065) 1 18.8 (7.9 to 36.1) <.001   CP + pCR 0.774 Model 1 + CTCBC (< or ≥2 CTC) 0.809 0.035 (0.002 to 0.063) 1 30.3 (14.7 to 48.8) <.001   CP + pCR 0.774 Model 1 + CTCBC (splines) 0.813 0.039 (0.004 to 0.072) 2 30.5 (14.4 to 47.8) <.001  DDFS (n = 1323)   CP+ CTCBC (< or ≥2 CTC) 0.698 Model 1 + pCR 0.746 0.048 (0.019 to 0.072) 1 27.9 (13.0 to 45.0) <.001   CP+ CTCBC (splines) 0.704 Model 1 + pCR 0.751 0.047 (0.016 to 0.070) 1 21.8 (10.0 to 37.3) <.001   CP + pCR 0.703 Model 1 + CTCBC (< or ≥2 CTC) 0.746 0.044 (0.012 to 0.068) 1 36.3 (19.7 to 55.4) <.001   CP + pCR 0.703 Model 1 + CTCBC (splines) 0.751 0.048 (0.018 to 0.074) 2 37.1 (21.1 to 57.1) <.001  LRFI (n = 1322)   CP+ CTCBC (< or ≥2 CTC) 0.726 Model 1 + pCR 0.728 0.002 (–0.042 to 0.031) 1 4.90 (0.30 to 13.1) .03   CP+ CTCBC (splines) 0.724 Model 1 + pCR 0.726 0.003 (–0.041 to 0.029) 1 2.6 (0.01 to 9.61) .10   CP + pCR 0.729 Model 1 + CTCBC (< or ≥2 CTC) 0.728 –0.001 (–0.031 to 0.024) 1 9.1 (1.74 to 18.7) .002   CP + pCR 0.729 Model 1 + CTCBC (splines) 0.726 –0.003 (–0.037 to 0.021) 2 8.7 (1.7 to 18.9) .02 Models with CTC count before NCT, pCR, and CTC count during NCT (landmark analyses)  OS (n = 861)   CP+CTCBC+pCR 0.804 Model 1 + CTCDC (< or ≥2 CTC) 0.806 0.002 (–0.016 to 0.013) 1 3.1 (0.0 to 11.4) .08   CP+CTCBC+pCR 0.798 Model 1 + CTCDC (splines) 0.796 –0.003 (–0.038 to 0.080) 2 5.0 (0.2 to 12.9) .02  DDFS (n = 856)   CP+CTCBC+pCR 0.758 Model 1 + CTCDC (< or ≥2 CTC) 0.760 0.002 (–0.013 to 0.014) 1 3.1 (0.0 to 9.7) .08   CP+CTCBC+pCR 0.756 Model 1 + CTCDC (splines) 0.755 –0.001 (–0.016 to 0.007) 2 2.7 (0.02 to 10.8) .09  LRFI (n = 855)   CP+CTCBC+pCR 0.717 Model 1 + CTCDC (< or ≥2 CTC) 0.707 –0.011 (–0.074 to 0.006) 1 0.9 (0.0 to 3.6) .34   CP+CTCBC+pCR 0.725 Model 1 + CTCDC (splines) 0.715 –0.010 (–0.047 to 0.006) 2 0.6 (0.0 to 2.8) .44 * For each analysis, the number of included patients corresponds to the number of patients with available data. Clinical and pathological characteristics were not statistically different among the different patient populations included in these analyses. CI = confidence interval; CP = baseline clinicopathological model; CTC = circulating tumor cells; CTCBC = CTC count before NCT; CTCDC = CTC count during NCT, using the highest value of CTC detection either after NCT start or prior to surgery; ddf = degree of freedom; DDFS = distant disease–free survival; LRFI = locoregional relapse–free interval; NCT = neoadjuvant chemotherapy; OS = overall survival; pCR = pathological complete response. † Two-sided likelihood ratio test P value corresponds to the average increase in χ2. Discussion This international meta-analysis shows that hematogenous cancer cell dissemination is affecting the outcome of treated patients and that this process cannot be predicted precisely enough by the current assessment of the primary tumor characteristics and response. Besides the diagnosis of inflammatory (T4d) breast cancer, CTC count was independent from any other baseline clinical or pathological characteristics. We also found that, in contrast to many other known clinic-pathological prognostic factors, CTC detection (either before NCT or before surgery) was not associated with pathological complete response. We found that the best postneoadjuvant survival models included prespecified clinicopathological prognostic markers at baseline, CTC detection at baseline, and pathological complete response, demonstrating the relevance of such metastasis-associated biomarkers for outcome prediction. Although presurgery CTC detection demonstrated its prognostic significance, it did not improve the accuracy of an optimized prognostic model taking into account CTC detection at baseline. In addition to the expected occurrence of distant metastases, CTC counts were also correlated to locoregional relapse. Although we cannot exclude that this unexpected finding might be simply a “bystander” effect of identifying more aggressive tumors, it could also suggest that the spread of breast cancer cells in the body might be also associated with a “self-seeding” mechanism (ie, breast cancer cells recirculate from distant sites back to the primary site), which has been demonstrated so far only in experimental models (4) and with bone marrow disseminated tumor cells (16). CTC count was a quantitative marker, as each CTC detected added a quantum of poor prognosis. In our clinical validity analyses, survival curves displayed a statistically significant survival difference starting from two CTCs detected. Interestingly, the two other thresholds that were tested in the current meta-analysis, namely one or more and five or more CTCs, were also able to distinguish a high-risk population. Any future adjuvant trial based on CTC detection should have to define a threshold compatible with its target population size (five or more CTCs being a rare event), taking into account that one CTC was detected in some healthy individuals with the applied CTC assay (5) and that the low CTC count observed in nonmetastatic breast cancer patients follows a Poisson’s distribution of rare events—blurring the line between zero and one CTCs detected in a single tube of blood (17). The same reason might explain why the longitudinal follow-up of CTC count and changes appeared of lower clinical relevance than previously reported in metastatic breast cancer patients (8). The main limitation of this study is related to the technical aspects underlying CTC detection by the CellSearch technique. Biologically, this US Food and Drug Administration–cleared technique captures CTC by their epithelial cell surface antigens, the expression of which is decreased in some cancer cells during the epithelial-to-mesenchymal transition (EMT) (18). However, abundant evidence supporting the clinical validity of this CTC detection technique in breast cancer had been reported (19); moreover, tumor cells with an intermediate EMT phenotype detected by CellSearch appear to be metastasis-initiator cells (17), which might explain why small incremental increases in CTC counts were associated with increased risks of relapse in the present analysis. Taken together, while the validity of NCT trials is impaired by the limited correlation between primary tumor response and postneoadjuvant survival, the enumeration of CTC improves postneoadjuvant survival prediction. NCT being also used in patients with other tumor entities where blood-borne dissemination of cancer cells plays also an important role (eg, rectal cancer) (20), the current report might stimulate future clinical studies in NCT-treated patients and therefore has implications beyond breast cancer. Funding This work was supported by Janssen Diagnostics and Innovative Medicines Initiative (grant agreement No. 115749, project “Cancer-ID”). Notes Authors: François-Clément Bidard, Stefan Michiels, Sabine Riethdorf, Volkmar Mueller, Laura J. Esserman, Anthony Lucci, Bjørn Naume, Jun Horiguchi, Rafael Gisbert-Criado, Stefan Sleijfer, Masakazu Toi, Jose A. Garcia-Saenz, Andreas Hartkopf, Daniele Generali, Françoise Rothé, Jeffrey Smerage, Laura Muinelo-Romay, Justin Stebbing, Patrice Viens, Mark Jesus M. Magbanua, Carolyn S. Hall, Olav Engebraaten, Daisuke Takata, José Vidal-Martínez, Wendy Onstenk, Noriyoshi Fujisawa, Eduardo Diaz-Rubio, Florin-Andrei Taran, Maria Rosa Cappelletti, Michail Ignatiadis, Charlotte Proudhon, Denise M. Wolf, Jessica B. Bauldry, Elin Borgen, Rin Nagaoka, Vicente Carañana, Jaco Kraan, Marisa Maestro, Sara Yvonne Brucker, Karsten Weber, Fabien Reyal, Dominic Amara, Mandar G. Karhade, Randi Ruud Mathiesen, Hideaki Tokiniwa, Antonio Llombart-Cussac, Alessandra Meddis, Paul Blanche, Koenraad d'Hollander, Paul Cottu, John W. Park, Sibylle Loibl, Aurélien Latouche, Jean-Yves Pierga, Klaus Pantel Affiliations of authors: Institut Curie, UVSQ and Paris Saclay University, Saint Cloud, France (FCB); Gustave Roussy, Villejuif, France (SM); Universitätsklinikum Hamburg-Eppendorf, Hamburg, Germany (SR, VM, KP); University of California at San Francisco, San Francisco, CA (LJE, MJMM, DMW, DA, JWP); MD Anderson Cancer Center, Houston, TX (ALu, CSH, JBB, MGK); Oslo University Hospital, Oslo, Norway (BN, OE, EB, RRM); Institute for Clinical Medicine, University of Oslo, Oslo, Norway (BN, OE); Gunma University Hospital, Gunma, Japan (JH, DT, RN, HT); Hospital Arnau de Vilanova, Valencia, Spain (RGC, JVM, VC, ALC); Erasmus MC Cancer Institute, Rotterdam, the Netherlands (SS, WO, JK); Kyoto University Hospital, Kyoto, Japan (MT, NF); Hospital Clinico San Carlos, Universidad Complutense, Madrid, Spain (JAGS, EDR, EDR, MM); Women’s Health Center, University of Tuebingen, Tuebingen, Germany (AH, FAT, SYB); Women Cancer Centre, University of Trieste, ASST of Cremona, Cremona, Italy (DG, MRC); Institut Jules Bordet, Université Libre de Bruxelles, Brussels, Belgium (FR, MI); University of Michigan Comprehensive Cancer Center, Ann Arbor, MI (JS); Complejo Hospitalario Universitario, Santiago de Compostela, Spain (LMR); Imperial College, London, UK (JS); Institut Paoli Calmettes, Aix Marseille University, CNRS, Inserm, Marseilles, France (PV); Institut Curie, PSL Research University, Paris, France (CP, FR, PC, JYP); Paris Descartes University, Paris, France (JYP); German Breast Group, Neu-Isenburg, Germany (KW, SL); Institut Curie, Inserm U900, Saint Cloud, France (AM, ALa); LMBA, Université de Bretagne Sud, Vannes, France (PB); International Drug Development Institute, Louvain La Neuve, Belgium (Kd’H); Conservatoire National des Arts et Métiers, EA4629, Paris, France (ALa). The study sponsor played no role in the design of the study; the collection, analysis, and interpretation of the data; the writing of the manuscript; and the decision to submit the manuscript for publication. This study was funded in part by a research grant from Janssen Diagnostics to F. C. Bidard’s institution. References 1 King TA , Morrow M. Surgical issues in patients with breast cancer receiving neoadjuvant chemotherapy . Nat Rev Clin Oncol. 2015 ; 12 ( 6 ): 335 – 343 . Google Scholar CrossRef Search ADS PubMed 2 Cortazar P , Zhang L , Untch M et al. , Pathological complete response and long-term clinical benefit in breast cancer: The CTNeoBC pooled analysis . Lancet. 2014 ; 384 ( 9938 ): 164 – 172 . Google Scholar CrossRef Search ADS PubMed 3 DeMichele A , Yee D , Berry DA et al. , The neoadjuvant model is still the future for drug development in breast cancer . Clin Cancer Res. 2015 ; 21 ( 13 ): 2911 – 2915 . Google Scholar CrossRef Search ADS PubMed 4 Comen E , Norton L , Massagué J. Clinical implications of cancer self-seeding . Nat Rev Clin Oncol. 2011 ; 8 ( 6 ): 369 – 377 . Google Scholar CrossRef Search ADS PubMed 5 Allard WJ , Matera J , Miller MC et al. , Tumor cells circulate in the peripheral blood of all major carcinomas but not in healthy subjects or patients with nonmalignant diseases . Clin Cancer Res. 2004 ; 10 ( 20 ): 6897 – 6904 . Google Scholar CrossRef Search ADS PubMed 6 Coumans FAW , Ligthart ST , Uhr JW , Terstappen LWMM. Challenges in the enumeration and phenotyping of CTC . Clin Cancer Res. 2012 ; 18 ( 20 ): 5711 – 5718 . Google Scholar CrossRef Search ADS PubMed 7 Ignatiadis M , Riethdorf S , Bidard FC et al. , International study on inter-reader variability for circulating tumor cells in breast cancer . Breast Cancer Res. 2014 ; 16 ( 2 ): R43 . Google Scholar CrossRef Search ADS PubMed 8 Bidard FC , Peeters DJ , Fehm T et al. , Clinical validity of circulating tumour cells in patients with metastatic breast cancer: A pooled analysis of individual patient data . Lancet Oncol. 2014 ; 15 ( 4 ): 406 – 414 . Google Scholar CrossRef Search ADS PubMed 9 Lorente D , Olmos D , Mateo J et al. , Decline in circulating tumor cell count and treatment outcome in advanced prostate cancer . Eur Urol. 2016 ; 70 ( 6 ): 985 – 992 . Google Scholar CrossRef Search ADS PubMed 10 Krebs MG , Sloane R , Priest L et al. , Evaluation and prognostic significance of circulating tumor cells in patients with non-small-cell lung cancer . J Clin Oncol. 2011 ; 29 ( 12 ): 1556 – 1563 . Google Scholar CrossRef Search ADS PubMed 11 Morrow CJ , Trapani F , Metcalf RL et al. , Tumourigenic non-small-cell lung cancer mesenchymal circulating tumour cells: A clinical case study . Ann Oncol. 2016 ; 27 ( 6 ): 1155 – 1160 . Google Scholar CrossRef Search ADS PubMed 12 Baccelli I , Schneeweiss A , Riethdorf S et al. , Identification of a population of blood circulating tumor cells from breast cancer patients that initiates metastasis in a xenograft assay . Nat Biotechnol. 2013 ; 31 ( 6 ): 539 – 544 . Google Scholar CrossRef Search ADS PubMed 13 Cayrefourcq L , Mazard T , Joosse S et al. , Establishment and characterization of a cell line from human circulating colon cancer cells . Cancer Res. 2015 ; 75 ( 5 ): 892 – 901 . Google Scholar CrossRef Search ADS PubMed 14 Michiels S , Baujat B , Mahé C et al. , Random effects survival models gave a better understanding of heterogeneity in individual patient data meta-analyses . J Clin Epidemiol. 2005 ; 58 ( 3 ): 238 – 245 . Google Scholar CrossRef Search ADS PubMed 15 Tudur Smith C , Williamson PR. A comparison of methods for fixed effects meta-analysis of individual patient data with time to event outcomes . Clin Trials. 2007 ; 4 ( 6 ): 621 – 630 . Google Scholar CrossRef Search ADS PubMed 16 Bidard FC , Kirova YM , Vincent-Salomon A et al. , Disseminated tumor cells and the risk of locoregional recurrence in nonmetastatic breast cancer . Ann Oncol. 2009 ; 20 ( 11 ): 1836 – 1841 . Google Scholar CrossRef Search ADS PubMed 17 Andree KC , van Dalum G , Terstappen LWMM. Challenges in circulating tumor cell detection by the CellSearch system . Mol Oncol. 2016 ; 10 ( 3 ): 395 – 407 . Google Scholar CrossRef Search ADS PubMed 18 Bidard FC , Proudhon C , Pierga JY. Circulating tumor cells in breast cancer . Mol Oncol. 2016 ; 10 ( 3 ): 418 – 430 . Google Scholar CrossRef Search ADS PubMed 19 Alix-Panabières C , Mader S , Pantel K. Epithelial-mesenchymal plasticity in circulating tumor cells . J Mol Med. 2017 ; 95 ( 2 ): 133 – 142 . Google Scholar CrossRef Search ADS PubMed 20 Cabel L , Proudhon C , Gortais H et al. , Circulating tumor cells: clinical validity and utility . Int J Clin Oncol. 2017 ; 22 ( 3 ): 421 – 430 . Google Scholar CrossRef Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

JNCI: Journal of the National Cancer InstituteOxford University Press

Published: Apr 12, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off