Prognosis (including both randomization arms) is depicted with Kaplan-Meier curves for total stromal tumor-infiltrating lymphocyte (TIL) counts (high indicates >5% median score value) (A) and for CD8+ cytotoxic TILs (positive indicates a mean of ≥3 stromal CD8+ TILs per 0.6-mm diameter tissue disc, assessed by immunohistochemical analysis on tissue microarrays or 0.6-mm discs digitally extracted from full-face sections) (B). Predictive capacity for PFS is stratified by treatment with trastuzumab vs lapatinib in CCTG MA.31 study patients with low (C) or high (D) levels of CD8+ stromal TILs in their primary tumor specimen.
eFigure 1. REMARK Diagram of the Original CCTG MA.31 Trial and Biomarker Study Sets
eFigure 2. Distribution of TIL Counts Across the CCTG MA.31 H&E Slide Study Set
eFigure 3. Overall Survival Stratified by Treatment
eFigure 4. Model of Association of Immune Infiltrates in the Primary Tumor Biopsy (CD8+ sTIL) With the CCTG MA.31 Metastatic HER2-Positive Population
eTable 1. Patient Characteristics in the CCTG MA.31 Immune Biomarker Study Population
eTable 2. Biomarker Expression in the CCTG MA.31 Study Population, Stratified by Disease Status
eTable 3. Estimates of Prognostic Hazard Ratio (HR) for High vs Low Level Immune Infiltrates in the CCTG MA.31 IHC Immune Biomarker Study Population
eTable 4. Exploratory Additional Stratified Full Multivariate Analysis for PFS
Customize your JAMA Network experience by selecting one or more topics from the list below.
Liu S, Chen B, Burugu S, et al. Role of Cytotoxic Tumor-Infiltrating Lymphocytes in Predicting Outcomes in Metastatic HER2-Positive Breast Cancer: A Secondary Analysis of a Randomized Clinical Trial. JAMA Oncol. 2017;3(11):e172085. doi:10.1001/jamaoncol.2017.2085
Do cytotoxic tumor-infiltrating lymphocytes predict patient response to antibody- (trastuzumab) vs small molecule–based (lapatinib) HER2 inhibitors?
In this secondary analysis of a randomized clinical trial, 427 patients with HER2-positive metastatic breast cancer randomized to trastuzumab vs lapatinib underwent assessment for lymphocyte biomarkers. A significant predictive effect on progression-free survival was observed for CD8+ cytotoxic stromal tumor-infiltrating lymphocytes.
A low level of preexisting stromal cytotoxic T-cell infiltration in primary tumor biopsy specimens predicts women who benefit from an antibody- vs a small molecule–based drug intervention for HER2-positive breast cancer, in the metastatic setting.
Accumulating evidence indicates that tumor-infiltrating lymphocytes (TILs) are associated with clinical outcomes and may predict the efficacy of chemotherapy and human epidermal growth factor receptor 2 (HER2, encoded by the gene ERBB2)–targeted therapy in patients with HER2-positive breast cancer.
To investigate the role of TILs, particularly cytotoxic CD8+ T cells, in the prediction of outcomes in patients with HER2-positive metastatic breast cancer randomized to an antibody-based (trastuzumab) vs a small molecule–based (lapatinib) anti-HER2 therapy.
Design, Setting, and Participants
The Canadian Cancer Trials Group MA.31 phase 3 clinical trial accrued patients from 21 countries and randomized 652 with HER2-positive metastatic breast cancer to receive trastuzumab or lapatinib, in combination with a taxane, from January 17, 2008, through December 1, 2011. Patients had received no prior chemotherapy or HER2-targeted therapy in the metastatic setting. The median follow-up was 21.5 months (interquartile range, 14.3-31.0). The tumor tissue collected for primary diagnosis was used in this ad hoc substudy. Sections were scored for TILs on hematoxylin-eosin (H&E)–stained sections, and immunohistochemical analysis was performed to assess CD8, FOXP3, CD56, and programmed cell death protein 1 (PD-1) expression on stromal (sTILs) and intratumoral TILs. Data were analyzed from July 15, 2015, through July 27, 2016.
Treatment with trastuzumab or lapatinib in combination with taxane chemotherapy (paclitaxel or docetaxel) for 24 weeks.
Main Outcomes and Measures
Prognostic effects of biomarkers were evaluated for progression-free survival by stratified univariate log-rank test with Kaplan-Meier curves and by multivariate Cox proportional hazards regression; predictive effects were examined with a test of interaction between treatment allocation and biomarker classification.
Of the 647 treated women (mean [SD] age, 55.0 [10.8] years), 614 had tumor tissue samples scored for H&E sTILs and 427 for CD8 biomarker assessments. Overall H&E sTIL counts of greater than 5% (high) were present in 215 cases (35%) but did not show significant prognostic or predictive effects. Univariate stratified analyses detected a significant predictive effect on risk for progression with lapatinib compared with trastuzumab among patients with low CD8+ sTIL counts (observed hazard ratio, 2.94; 95% CI, 1.40-6.17; P = .003) and among those with high CD8+ sTIL counts (observed hazard ratio, 1.36; 95% CI, 1.05-1.75; P = .02), confirmed in stepwise multivariate analysis (interaction P = .04). Other immunohistochemistry biomarkers were not associated with prognostic or predictive effects.
Conclusions and Relevance
In this secondary analysis of a phase 3 randomized clinical trial, a low level of preexisting cytotoxic sTILs predicted the most benefit from an antibody- vs a small molecule–based drug against the same target.
clinicaltrials.gov Identifier: NCT00667251
Trastuzumab, a monoclonal antibody that functions not only by inhibiting the human epidermal growth factor receptor 2 (HER2, encoded by the gene ERBB2) signaling pathway but also by inducing antitumor immune responses,1-6 has significantly improved outcomes in women with HER2-positive breast tumors.7-10 The success of trastuzumab led to development of alternative HER2-targeted therapies. Small molecule tyrosine kinase inhibitors of HER2, such as lapatinib, are easier to produce and administer but do not induce antibody-mediated antitumor immune responses.
The presence of tumor-infiltrating lymphocytes (TILs),11 particularly CD8+ cytotoxic T cells,12 has been associated with improved prognosis among patients with HER2-positive breast cancer.13-16 The TIL count predicts a pathologic complete response to neoadjuvant chemotherapy,17-19 trastuzumab, and lapatinib.13-16 In the Finland Herceptin (FinHER) trial,17-19 an increased TIL count predicted better distant disease-free survival in patients with HER2-positive early breast cancer who were randomized to adjuvant trastuzumab. However, in a study including larger numbers of HER2-positive early breast cancers,17-19 high TIL levels predicted a lack of benefit from adding trastuzumab to adjuvant chemotherapy. Fewer studies have assessed metastatic breast cancer in this context.20
Although the percentage of stromal TILs measured on hematoxylin-eosin (H&E)–stained full-face sections is a convenient biomarker of net immune response,19,21 overall TIL counts do not discern the specific subtypes of lymphocytes present, which include effector and regulatory populations, implying different types of immune reactions. CD8+ TILs, representing the major antitumor effector population, can be assessed reliably by immunohistochemistry (IHC) on whole sections and tissue microarrays (TMAs).12-14,22,23 To date, however, few studies have assessed CD8+ TIL counts in phase 3 clinical trial series, in large part owing to difficulties in obtaining tissue for IHC from enough patients to power biomarker analyses.
The Canadian Cancer Trials Group (CCTG) MA.31 trial was the first phase 3 clinical trial, to our knowledge, to randomize women with metastatic HER2-positive breast cancer to trastuzumab vs lapatinib as first-line therapy (in combination with a taxane). The final analysis revealed trastuzumab to be superior to lapatinib (hazard ratio [HR], 0.73; P = .001) for the primary end point of progression-free survival (PFS).24 We designed this formal prospective-retrospective correlative study on tissue specimens from the CCTG MA.31 trial to evaluate the clinical significance of lymphocytic infiltrates in metastatic breast cancer. We hypothesized that preexisting cytotoxic immune responses in tumor biopsy tissue might predict which patients would benefit most from treatment with an immune-inducing antibody therapy compared with a small molecule inhibitor directed against the same target.
The CCTG MA.31 trial, conducted from January 17, 2008, through December 1, 2011, recruited 652 patients with HER2-positive metastatic breast cancer from 21 countries.24 Patients were randomized 1:1 to treatment with trastuzumab or lapatinib, in combination with taxane chemotherapy for 24 weeks, followed by the same HER2-targeted monotherapy until progression (a copy of the trial protocol is found in Supplement 1). Treatment included intravenous trastuzumab combined with a taxane for 24 weeks, followed by trastuzumab, 6 mg/kg once every 3 weeks, or oral lapatinib, 1250 mg/d, combined with taxane for 24 weeks, followed by lapatinib, 1500 mg/d. The following 2 taxane choices were allowed: intravenous paclitaxel, 80 mg/m2 once a week, or intravenous docetaxel, 75 mg/m2 once every 3 weeks, as described.24 Median follow-up was 21.5 months (interquartile range, 14.3-31.0 months).
Hematoxylin-eosin–stained slides and formalin-fixed paraffin-embedded unstained TMA cores or whole sections from the primary tumor were provided by the CCTG Tumour Tissue Data Repository for this correlative study, which was approved by the British Columbia Cancer Agency research ethics board. The protocol included an optional substudy for biopsy of metastatic lesions after randomization and before treatment and at the time of progressive disease, but very few samples were obtained and none were used in this study. Written informed consent was obtained for correlative study tissue collection and for assignment to treatment.
REMARK principles25 and guidelines for use of archived trial specimens for biomarker evaluation26 were followed in the design of this formal prospective, retrospective study. The primary end point of the CCTG MA.31 trial and of this correlative substudy was PFS, defined as the time from random assignment to disease progression per the Response Evaluation Criteria in Solid Tumors (RECIST), version 1.0, or to death from any cause.24 Overall survival was a secondary end point.
The prespecified primary aim of this study was to evaluate the predictive value of cytotoxic CD8+ TILs for PFS on trastuzumab vs lapatinib. We chose CD8 as the primary biomarker because it defines what is likely to be the most relevant subpopulation of antitumor immune cells and because CD8 IHC demonstrated consistent strong prognostic capacity among patients with HER2-positive disease on TMA from studies totaling 8978 patients.12 We could therefore use established CD8+ TIL classification cut points on TMA cores as defined in previously published work with equivalent, analytically validated techniques for IHC staining and interpretation.13 Prespecified secondary aims were to assess the prognosis and the predictive and prognostic effects of H&E-stained TIL counts27 and expression of FOXP3 (forkhead box P3 protein, marking T-regulatory cells),28 PD-1 (programmed cell death protein 1, marking check point–expressing T cells), and CD56 (marking natural killer cells).
Immunohistochemical staining for CD8 and FOXP3 was performed as previously described.13,28 Anti–PD-1 monoclonal mouse antibody (prediluted clone NAT105; Cell Marque) and anti-CD56 monoclonal mouse antibody (clone 123C3.D5 [dilution 1:200]; Cell Marque) were used according to the manufacturer’s instructions. Immunohistochemical analysis was performed using unstained sections from duplicate 0.6-mm core TMAs representing 367 cases for which blocks from primary excision specimens had been provided to the CCTG Tumour Tissue Data Repository. For 152 additional cases, only full-face unstained slides were available. Scoring of biomarkers was performed from digitally scanned images by one of us (D.G.) who had no access to the clinical data. Stromal TILs (sTILs) and intratumoral TILs (iTILs) were defined as previously described13,28: positive sTILs are biomarker-stained lymphocytes in the peritumoral stroma not in direct contact with carcinoma cells, whereas positive iTILs directly contact invasive carcinoma cells. For cases in which only whole sections were available, an algorithm was applied whereby the pathologist selected (at a low magnification power) areas of viable carcinoma to defined centers of 0.6-mm virtual cores. Mean biomarker scores were confirmed to show no significant differences between TMA cores and whole-section images analyzed using this method. The means of CD8+ TIL counts from replicate cores were calculated to give the following values per 0.6-mm core, matching the tumor area used to establish cutoffs for CD8+ and FOXP3+ TILs that were validated in a previously published study13 and were used for the primary analysis: sTIL of at least 3 vs less than 3 and iTIL of at least 1 vs less than 1. Because the numbers of TILs positive for PD-1 and CD56 were low and cutoffs had not been previously established on equivalent large TMA cohorts, we used any positive expression as an exploratory cutoff (>0 vs 0).
We assessed TILs on H&E-stained whole sections using the method originally described by Denkert et al17 that was standardized by the international TILs working group27,29 and externally validated.30 In brief, TILs were scored as the percentage of intratumoral stromal area infiltrated by lymphocytes, interpreted visually in the context of reference images. Of the available H&E-stained slides, 337 (55.2%) were from surgical excisions and 273 (44.8%) from core needle biopsies. To mitigate against any measurement bias that this method might create, we scored TILs on surgical excision specimens by digitally scanning and then virtually extracting a 1-mm strip image equivalent to a core needle biopsy.
Data were analyzed from July 15, 2015, through July 27, 2016. The CCTG central office independently executed a submitted prespecified statistical analysis plan, using the treated population to examine the predictive and prognostic effects of the investigated biomarkers. Fisher exact tests were used to assess for imbalances by allocation to treatment and stratification factors in those included or not included in the correlative study population. Kaplan-Meier curves were used to display survival distributions, and stratified log-rank testing was used to assess differences in PFS between biomarker-defined patient groups. Stratified Cox proportional hazards regression was used to estimate the HRs by biomarker status. We conducted tests for proportional hazards assumptions for overall survival and PFS in the multivariate models and found no evidence against the proportional hazards assumption for Cox models.
Multivariate analysis was conducted with stratified stepwise forward Cox models, in which patient baseline clinical and tumor characteristics (age, Eastern Cooperative Oncology Group performance status, disease status [metastatic disease at primary diagnosis vs recurrence after intended curative therapy], prior [neo]adjuvant anthracycline treatment, and estrogen and progesterone receptor status) were added in the model separately (P < .05 by likelihood ratio test). Predictive effects were evaluated with a test of interaction between treatment groups and biomarker classification at predefined cut points. All tests were 2 sided at a significance level of .05. Cases with missing values were removed from analyses, and only cases with complete data were considered. Missing value comparison analysis was performed to ensure that exclusion of cases with missing data did not introduce any bias. The TIL count was treated as a categorical variable for predictive analyses (using the median count as a cutoff [≤5% vs >5%] to maximize power for interaction testing); prognostic analyses tested TIL count as a categorical and as a continuous variable.
Among the 647 CCTG MA.31 trial patients who received their assigned treatment (mean [SD] age, 55.0 [10.8] years), IHC results could be obtained for 427 for CD8, 398 for FOXP3, 394 for PD-1, and 372 for CD56 (eFigure 1 in the Supplement). The H&E-stained slides were assessable for TIL counts for 614 patients. Clinical characteristics for the CCTG MA.31, H&E-stained TIL, and CD8 IHC study populations are shown in eTable 1 in the Supplement, along with the proportion of cases positive for each biomarker. We found no imbalances by assignment to treatment. The TIL counts and CD8 and CD56 biomarker expression did not differ by disease status, although FOXP3+ and PD-1+ sTIL counts were higher among patients who presented with metastatic disease (eTable 2 in the Supplement).
Among the 614 tumors assessable for overall H&E-stained TILs, the mean (SD) score (using the international TIL Working Group method27,29) was 9.2% (12.1%; median, 5%; IQR, 1%-10%). In this set of primary breast tumor tissue specimens from patients who developed metastatic disease, only 14 (2.3%) had TIL scores of at least 50%, a value defining lymphocyte-predominant breast cancer in other studies.30-32 Overall H&E sTIL counts of greater than 5% (high) were present in 215 cases (35%) but did not show significant prognostic or predictive effects. The TIL score distribution is shown in eFigure 2 in the Supplement. Median absolute counts for the IHC biomarkers (per 0.6-mm diameter tumor circle; 0.28-mm2 tumor area) were 13 CD8+ sTILs, 2 CD8+ iTILs, 6 FOXP3+ sTIL, and zero for PD1 and CD56; the percentage of cases considered to be high expressers for these immuno-oncology biomarkers based on the predefined cut points is shown in eTable 1 in the Supplement. The CD56+ infiltrates incorporating the natural killer cell population were very rarely identified and were excluded from subsequent survival analyses.
In univariate prognostic analyses, none of the assessed biomarkers (CD8 [HR, 0.91; 95% CI, 0.67-1.24], FOXP3 [HR, 0.91; 95% CI, 0.70-1.18], PD-1 [HR, 0.90; 95% CI, 062-1.30], or TILs as assessed on H&E-stained sections) showed a significant effect on PFS in the full study cohort (Table 1 and Figure, A and B) using the predefined cut points. This finding was also true within the hormone receptor–positive and –negative subsets (eTable 3 in the Supplement), as well as for the secondary end point of overall survival. In exploratory analyses, no significant prognostic effect was identified using IHC biomarker assessments of intratumoral as opposed to sTILs, nor was a significant prognostic effect observed using TIL counts as a univariate continuous variable. When stratified by disease status at presentation, no significant prognostic associations were seen for TIL count or CD8, PD1, or FOXP3 sTIL count (eTable 3 in the Supplement); FOXP3 expression assessed as iTILs was associated with improved survival only among patients who presented with stage IV disease.
Univariate analysis (Table 2) demonstrated a higher risk for disease progression among patients with low numbers of CD8+ cytotoxic sTILs who were treated with lapatinib compared with trastuzumab (HR, 2.94; 95% CI, 1.40-6.17; P = .003) (Figure, C) than was evident among patients with high numbers of CD8+ sTILs (HR, 1.36; 95% CI, 1.05-1.75; P = .02) (Figure, D). This differential effect was confirmed in the prespecified stratified stepwise multivariate Cox regression analysis (stratified HR, 0.52; 95% CI, 0.28-0.98; interaction P = .04) (Table 3), indicating that having low levels of CD8+ sTILs in the original biopsy specimen predicts inferior response to lapatinib relative to trastuzumab in the metastatic setting. A significant interaction was maintained in an exploratory, full multivariate model (eTable 4 in the Supplement) including age, performance status, disease status at presentation, centrally reviewed hormone receptor status, and weight (P = .04). Similar trends by CD8 status and treatment were observed for the secondary end point of overall survival (eFigure 3 in the Supplement), but the interaction test was not significant in the prespecified multivariate model (interaction HR, 0.56; 95% CI, 0.22-1.48; P = .24). Similarly, when CD8 scoring was interpreted based on iTILs rather than sTILs, the same trends in PFS and overall survival were noted but were not significant (eFigure 3 in the Supplement). The other tested biomarkers, including TIL counts, did not show significant predictive effects.
In the CCTG MA.31 phase 3 clinical trial of women with metastatic breast cancer randomized to antibody- vs small molecule–based anti-HER2 targeted therapy, we found that low levels of CD8+ infiltrates (<3 per 0.6-mm TMA core) in the original biopsy tissue predict increased benefit from immunostimulatory trastuzumab compared with tyrosine kinase inhibitor lapatinib in the first-line metastatic treatment. Trastuzumab was superior to lapatinib in all groups, but the degree to which lapatinib was inferior was greatest among women with low levels of preexisting cytotoxic immune infiltrates.
The most relevant comparator is the recently published analysis of TIL counts in 678 metastatic HER2-positive breast cancer specimens from the CLEOPATRA (Clinical Evaluation of Pertuzumab and Trastuzumab) trial, which randomized women to pertuzumab vs placebo to support trastuzumab and docetaxel as first-line therapy.20 Because both arms received antibody-based therapy, it is perhaps not surprising that no predictive effect was identified. In the CLEOPATRA trial, higher numbers of TILs were associated with better prognosis, albeit only for the secondary end point of overall survival (HR, 0.89; 95% CI, 0.83-0.96) and not for PFS (HR, 0.95; 95% CI, 0.90-1.00). Our results, which did not identify a significant prognostic effect in 614 patients from the CCTG MA.31 trial, do not necessarily disagree with those of the CLEOPATRA trial, because the 95% CIs around the HRs overlap and our study was underpowered for assessment of overall survival, with less than half the number of events as for PFS.24 Both studies found that the magnitude of the prognostic benefit of H&E-stained TIL counts in metastatic breast cancer is smaller than what has been reported in early-stage disease.33 The CCTG MA.31 and CLEOPATRA trials also revealed that mean TIL counts (eFigure 2 in the Supplement) are lower in populations of patients who develop metastases than mean values reported in populations with early-stage breast cancer.30-32
Accumulating evidence suggests that preexisting TILs may predict a better response to cytotoxic chemotherapy in early-stage, hormone receptor–negative breast cancer.17,18,23,31 In the GeparSixto randomized clinical trial,18 TIL counts and immune response messenger RNA biomarkers predicted pathologic complete response to a neoadjuvant anthracycline combined with a taxane, with or without carboplatin chemotherapy. Results concerning prediction for anti-HER2 targeted therapy in early breast cancer have been less consistent. In a prospective, retrospective study of the phase 3 FinHER trial,19 which randomized 232 patients with HER2-positive early breast cancer to chemotherapy with or without 9 weeks of trastuzumab, continuous TIL counts (each 10% increase) were significantly associated with improved distant disease-free survival in patients treated with trastuzumab. A secondary analysis of the NeoALTTO (Neoadjuvant Lapatinib and/or Trastuzumab Treatment Optimization) trial34,35 found that, although prognostic, neither H&E-stained TILs nor an immune signature derived from RNA sequencing had any significant predictive effect for neoadjuvant trastuzumab vs lapatinib vs their combination in HER2-positive early breast cancer. Furthermore, a study of 1282 tumors from the N9831 trial36,37 of patients with HER2-positive early breast cancer assigned to chemotherapy with or without 52 weeks of trastuzumab therapy found that TIL counts (and a 14-gene signature associated with immune response) were prognostic for relapse-free survival only in the chemotherapy arm (HR, 0.23; 95% CI, 0.07-0.73); outcomes in the trastuzumab arm were the same regardless of TIL count (HR, 1.01; 95% CI, 0.89-1.15).
One possible explanation for our observations in the CCTG MA.31 trial concerns the likely differences in the immunogenic state of tumors before and after metastasis and the association that this has with biomarker detection in primary specimens and anti-HER2 treatments given after metastases develop (eFigure 4 in the Supplement). During the development and progression of invasive cancers, neoplastic cells are selected to evade immunologic detection, induce tolerance, or overwhelm antitumor immune responses.38 Primary HER2-positive breast cancers can exist in the following states: (1) immunogenic but not immunoevasive, (2) immunogenic and immunoevasive, or (3) not immunogenic. Tumors in state 1 have the best prognosis and respond particularly well to adjuvant therapy. Because state 1 tumors are positive for TILs, CD8, and related biomarkers, they drive observed associations between immune infiltrates and good prognosis in early breast cancer. However, this association means that populations with metastatic disease (CCTG MA.31) do not include many state 1 tumors. Because most patients in the CCTG MA.31 trial were positive for CD8+ TILs, they likely represent state 2 tumors. The immunoevasive primary phenotype of state 2 tumors is maintained in the metastatic setting; thus, anti-HER2 therapies must function based on their ability to inhibit HER2 signaling pathways, with limited capacity for benefit via immune induction. In this context, trastuzumab has only a small therapeutic advantage over lapatinib. In contrast, patients in the CCTG MA.31 trial with low CD8 sTIL counts likely represent state 3 tumors, some of which may retain an incompletely immunoevasive phenotype. Trastuzumab not only inhibits HER2 signaling but also enhances antibody-dependent cell-mediated cytotoxicity and primes antitumor-adaptive T-cell responses.5,6 State 3 HER2-positive tumors might therefore derive a greater therapeutic benefit from trastuzumab compared with lapatinib in the metastatic setting, a potential explanation for the observed predictive effect of CD8+ sTILs in the CCTG MA.31 trial.
Lapatinib now has limited indications in breast cancer owing to trials showing its inferiority to trastuzumab, its higher rate of toxic effects,39 and the availability of newer anti-HER2 agents.40 However, small molecule–based drugs retain potential advantages (oral delivery and lower production costs). A newer small molecule inhibitors of HER2 is neratinib.41 Other examples of oncogenic changes alternatively targetable by antibody or small molecule strategies include FGFR1,42EGFR,43IGF1R,44 and MET.45 Our findings suggest that investigating whether low preexisting immune infiltrate levels identify a group of patients better treated with antibody-based therapies may be worthwhile, not only for breast cancer but also for other types of cancer.
Although we used predefined biomarker cutoffs and a prespecified statistical plan to find a significant interaction for our primary hypothesis, our study has important limitations. First, as a prospective, retrospective correlative study, it cannot provide level 1 evidence; novel findings need to be validated in similar trials. Second, owing to logistical limitations, the materials available for IHC represent a combination of TMA and full-face sections, and the full-face H&E-stained sections used for overall TIL assessments were from a mix of surgical excision and core needle biopsy specimens. To make measurements consistent, we used digital images to evaluate the biomarkers, extracting TMA-equivalent circles for IHC assessments of full-face sections and needle biopsy–equivalent strips to score TILs on surgical excision specimens. Biomarkers had to be assessed on primary tumor tissues; tissue samples from the metastatic lesions were not available for analysis. Third, the CCTG MA.31 trial includes a mix of patients with recurrent and metastatic disease at initial diagnostic presentation. We included this variable in a multivariate analysis to adjust for its potential confounding effect.
In the CCTG MA.31 trial of HER2-positive metastatic breast cancer, a low level of cytotoxic T-cell infiltration in the primary tumor’s stroma was predictive of which women receive the most benefit from trastuzumab therapy compared with treatment with lapatinib. Implementation of standardized evaluation of cytotoxic T-cell infiltration may be valuable for predicting response to antibody-based therapies.
Corresponding Author: Torsten O. Nielsen, MD, PhD, FRCPC, Department of Pathology and Laboratory Medicine, Genetic Pathology Evaluation Centre, JPN Room 1401, 855 W 12th Ave, Vancouver, BC V5Z 1M9, Canada (firstname.lastname@example.org).
Accepted for Publication: May 23, 2017.
Published Online: July 27, 2017. doi:10.1001/jamaoncol.2017.2085
Author Contributions: Drs Chen and Parulekar had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Liu, Chen, Kos, Parulekar, Gelmon, Nielsen.
Acquisition, analysis, or interpretation of data: Liu, Chen, Burugu, Leung, Gao, Virk, Parulekar, Shepherd, Nielsen.
Drafting of the manuscript: Liu, Chen, Burugu, Leung, Gao, Virk, Nielsen.
Critical revision of the manuscript for important intellectual content: Liu, Chen, Kos, Parulekar, Shepherd, Gelmon, Nielsen.
Statistical analysis: Liu, Chen, Parulekar.
Obtained funding: Liu, Chen, Kos, Nielsen,
Administrative, technical, or material support: Liu, Burugu, Gao, Virk, Parulekar, Shepherd, Gelmon, Nielsen.
Study supervision: Liu, Nielsen.
Conflict of Interest Disclosures: Dr Gelmon reports serving in an advisory role for Novartis, Pfizer, AstraZeneca, and Genentech. Dr Nielsen reports holding a proprietary interest in the Prosigna intrinsic subtype classifier (not used in this study) and receiving consultancy fees from NanoString Technologies. No other disclosures were reported.
Funding/Support: This study was supported by grant 16094 from the Canadian Breast Cancer Foundation (biomarker studies) and by a studentship from the Fonds de recherche du Québec–Santé (Ms Burugu).
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Create a personal account or sign in to: