Association Between CD8+ T-Cell Infiltration and Breast Cancer Survival
We used data from three observational studies of newly diagnosed breast cancer [Study of Epidemiology and Risk Factors in Cancer Heredity (SEARCH),N = 4079; the British Columbia Cancer Agency (BCCA),N = 4520; the Nottingham Tenovus Primary Breast Cancer Series (NBCS),N = 1842] and one randomised, controlled trial [the National Epirubicin Adjuvant Trial (NEAT),N = 1998 composed of both NEAT (n = 1684) and BR9/601 (n = 314)] of breast cancer. Analyses of T-cell data from two of these studies have been published previously. All participating studies were approved by the relevant research ethics committee. SEARCH is a prospective population-based study of women diagnosed with breast cancer in East Anglia, England. The BCCA study comprised women diagnosed with breast cancer between 1986 and 1992 in British Columbia and referred to BCCA for consideration of adjuvant therapy. The NBCS comprises patients diagnosed and treated at Nottingham City Hospital between 1987 and 1998. Details of the National Epirubicin Adjuvant Trial and BR9601 trial (here referred to collectively as NEAT) have been published previously. Briefly, this was a phase III trial in which patients were randomised on a 1 : 1 basis to receive cyclophosphamide, methotrexate and fluorouracil (CMF) or epirubicin in addition to CMF (E-CMF). Results of this trial were first published in 2006. Additional details are provided in the supplementary Methods and in Tables S1 and S2, available at Annals of Oncology online http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1.
Immunohistochemistry (IHC) was conducted for CD8 and FOXP3 proteins at host institutions. Details of scoring systems and cut points for positivity are provided in supplementary Table S3, available at Annals of Oncologyonline http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1. Tissue microarrays (TMAs) were used to analyse large numbers of tumour samples simultaneously, each represented by a single 0.6-mm tissue core. Additional details of IHC scoring are provided in supplementary Methods, available at Annals of Oncology online http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1. Absolute numbers of immunoreactive tumour-infiltrating lymphocytes were counted and classified as 'intratumoral' (iT) if seen in direct contact with tumour cells and 'stromal' (S) if they were not in direct contact with tumour cells. Tumours were classified into different molecular subtypes as previously described (supplementary Table S4, available at Annals of Oncology online http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1). T-cell counts were dichotomised for statistical analyses using a pre-specified cut point of zero versus any more than zero immunoreactive lymphocytes. This cut point was chosen because discrimination between tissue completely devoid of positive lymphocytes and tissue containing any positive lymphocytes is likely to be reliable. Information on FOXP3+ T lymphocytes was available for the SEARCH, NBCS and NEAT studies only.
Cox regression models stratified by study were used to test for associations with breast cancer-specific survival (BCSS). Follow-up time was truncated at 10 years. Women with estrogen receptor (ER)-positive and ER-negative breast cancer were analysed separately because of differences in their patterns of short and long-term survival. Late entry for the SEARCH study was accounted for by left truncation of survival time data. Variables that showed a time-dependent association with survival, and therefore violated the Cox proportional hazards assumption, were modelled by using an extended Cox model to include a coefficient (T) which varied linearly as a function of the logarithm of time. Variables significantly associated with BCSS on univariate analysis were also evaluated in multivariate analysis. Data on hormone therapy was not available for the NEAT study hence multivariate models excluding this study and including hormone therapy as a covariate are presented in the supplementary material, available at Annals of Oncology online http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1. Cochran's Q-test was used to test for heterogeneity of the prognostic effect of T-cell status according to different patient and tumour subgroups and of differential benefit of anthracyclines according to T-cell status in the NEAT trial. To determine whether cytotoxic and regulatory T lymphocytes contributed complementary prognostic value and, therefore, whether their prognostic accuracy could be improved by accounting for it, an interaction term between the variables was included in exploratory Cox regression analyses. Multiple imputation was used to adjust for the bias of missing data. This is a statistical technique which resolves missing values by predicting their probable value based on the complete data using a multivariate regression model. The variability between imputed (predicted) values is accounted for by producing multiple datasets. We imputed 50 datasets including all 12 439 patients. Survival estimates based on these data were computed per dataset and combined to account for between- and within-dataset variation (additional details are provided in supplementary Methods, available at Annals of Oncology online http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1). The distributions of imputed versus observed values for all variables included in the model are illustrated in supplementary Figure S1, available at Annals of Oncology online http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1. The necessity of adjusting for missing data is illustrated by supplementary Figure S2, available at Annals of Oncology online http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1 which depicts significant differences in survival in subgroups of patients according to whether data were missing for CD8, FOXP3, ER and human epidermal growth factor receptor 2 (human epidermal growth factor receptor 2 (HER2)). Relative survival estimates derived from imputed data are presented in the main report and estimates from the complete case analysis detailed in the supplementary material, available at Annals of Oncology online http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1. Absolute survival estimates are based on complete data. Statistical methods are further detailed in the supplementary material, available at Annals of Oncology online http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1. Analyses are reported in accordance with REMARK guidelines. All analyses were conducted using Intercooled Stata version 11.2 (Stata Corp., College Station, TX). Data analysis was conducted by Ali. The Stata code used for all survival analyses can be made available upon request from the corresponding author.
Methods
Ethics Statement and Study Populations
We used data from three observational studies of newly diagnosed breast cancer [Study of Epidemiology and Risk Factors in Cancer Heredity (SEARCH),N = 4079; the British Columbia Cancer Agency (BCCA),N = 4520; the Nottingham Tenovus Primary Breast Cancer Series (NBCS),N = 1842] and one randomised, controlled trial [the National Epirubicin Adjuvant Trial (NEAT),N = 1998 composed of both NEAT (n = 1684) and BR9/601 (n = 314)] of breast cancer. Analyses of T-cell data from two of these studies have been published previously. All participating studies were approved by the relevant research ethics committee. SEARCH is a prospective population-based study of women diagnosed with breast cancer in East Anglia, England. The BCCA study comprised women diagnosed with breast cancer between 1986 and 1992 in British Columbia and referred to BCCA for consideration of adjuvant therapy. The NBCS comprises patients diagnosed and treated at Nottingham City Hospital between 1987 and 1998. Details of the National Epirubicin Adjuvant Trial and BR9601 trial (here referred to collectively as NEAT) have been published previously. Briefly, this was a phase III trial in which patients were randomised on a 1 : 1 basis to receive cyclophosphamide, methotrexate and fluorouracil (CMF) or epirubicin in addition to CMF (E-CMF). Results of this trial were first published in 2006. Additional details are provided in the supplementary Methods and in Tables S1 and S2, available at Annals of Oncology online http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1.
Immunohistochemistry and Scoring
Immunohistochemistry (IHC) was conducted for CD8 and FOXP3 proteins at host institutions. Details of scoring systems and cut points for positivity are provided in supplementary Table S3, available at Annals of Oncologyonline http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1. Tissue microarrays (TMAs) were used to analyse large numbers of tumour samples simultaneously, each represented by a single 0.6-mm tissue core. Additional details of IHC scoring are provided in supplementary Methods, available at Annals of Oncology online http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1. Absolute numbers of immunoreactive tumour-infiltrating lymphocytes were counted and classified as 'intratumoral' (iT) if seen in direct contact with tumour cells and 'stromal' (S) if they were not in direct contact with tumour cells. Tumours were classified into different molecular subtypes as previously described (supplementary Table S4, available at Annals of Oncology online http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1). T-cell counts were dichotomised for statistical analyses using a pre-specified cut point of zero versus any more than zero immunoreactive lymphocytes. This cut point was chosen because discrimination between tissue completely devoid of positive lymphocytes and tissue containing any positive lymphocytes is likely to be reliable. Information on FOXP3+ T lymphocytes was available for the SEARCH, NBCS and NEAT studies only.
Statistical Analyses
Cox regression models stratified by study were used to test for associations with breast cancer-specific survival (BCSS). Follow-up time was truncated at 10 years. Women with estrogen receptor (ER)-positive and ER-negative breast cancer were analysed separately because of differences in their patterns of short and long-term survival. Late entry for the SEARCH study was accounted for by left truncation of survival time data. Variables that showed a time-dependent association with survival, and therefore violated the Cox proportional hazards assumption, were modelled by using an extended Cox model to include a coefficient (T) which varied linearly as a function of the logarithm of time. Variables significantly associated with BCSS on univariate analysis were also evaluated in multivariate analysis. Data on hormone therapy was not available for the NEAT study hence multivariate models excluding this study and including hormone therapy as a covariate are presented in the supplementary material, available at Annals of Oncology online http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1. Cochran's Q-test was used to test for heterogeneity of the prognostic effect of T-cell status according to different patient and tumour subgroups and of differential benefit of anthracyclines according to T-cell status in the NEAT trial. To determine whether cytotoxic and regulatory T lymphocytes contributed complementary prognostic value and, therefore, whether their prognostic accuracy could be improved by accounting for it, an interaction term between the variables was included in exploratory Cox regression analyses. Multiple imputation was used to adjust for the bias of missing data. This is a statistical technique which resolves missing values by predicting their probable value based on the complete data using a multivariate regression model. The variability between imputed (predicted) values is accounted for by producing multiple datasets. We imputed 50 datasets including all 12 439 patients. Survival estimates based on these data were computed per dataset and combined to account for between- and within-dataset variation (additional details are provided in supplementary Methods, available at Annals of Oncology online http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1). The distributions of imputed versus observed values for all variables included in the model are illustrated in supplementary Figure S1, available at Annals of Oncology online http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1. The necessity of adjusting for missing data is illustrated by supplementary Figure S2, available at Annals of Oncology online http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1 which depicts significant differences in survival in subgroups of patients according to whether data were missing for CD8, FOXP3, ER and human epidermal growth factor receptor 2 (human epidermal growth factor receptor 2 (HER2)). Relative survival estimates derived from imputed data are presented in the main report and estimates from the complete case analysis detailed in the supplementary material, available at Annals of Oncology online http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1. Absolute survival estimates are based on complete data. Statistical methods are further detailed in the supplementary material, available at Annals of Oncology online http://annonc.oxfordjournals.org/content/25/8/1536/suppl/DC1. Analyses are reported in accordance with REMARK guidelines. All analyses were conducted using Intercooled Stata version 11.2 (Stata Corp., College Station, TX). Data analysis was conducted by Ali. The Stata code used for all survival analyses can be made available upon request from the corresponding author.
SHARE
![Pediatric Supportive Care (PDQ®): Supportive care - Patient Information [NCI]-Overview](https://cdnus.oss-us-west-1.aliyuncs.com/libs/images/NewsJpg/cate_6/072_215x160.jpg)
