Extended Data Fig. 9: Parity, breastfeeding and TIL associations in human breast cancer subtypes.
From: Parity and lactation induce T-cell-mediated breast cancer protection
a, Breastfeeding cohort multivariate analysis with clinical prognostic factors shown (n = 270). OS events were censored at 15 years follow up and adjusted for breastfeeding status (BF), age at diagnosis (Age.dx), years since last live birth to BC diagnosis (time since last birth), and treatment with chemotherapy (treatment_chemo) or mastectomy (treatment_mastectomy). Hazard ratios are presented with 95% confidence intervals, shown as squares and horizontal lines, respectively. b, PB-TRM signature correlation with tumour infiltrating lymphocyte (TIL) counts in TNBC (Basal subtype, n = 136) from the TCGA dataset. R value indicate Spearman’s correlation coefficient (two-sided p value from correlation test). c, Enrichment of the PB-TRM signature in the indicated BC subtypes from the Metabric dataset, Basal (n = 329), HER2 (n = 240), LumA (n = 718) and LumB (n = 488). Kruskal Wallis test, p-value shown is unadjusted, d, Kaplan–Meier survival analysis indicates disease-free survival and e, overall survival from (n = 329) primary basal-like/TNBCs with prognostic separation according to PB-TRM signature from parous normal breast tissue. Log-rank, p-values and hazard ratios are shown. f, Intratumoural T cell density in HER2 and Luminal BC subtypes determined by immunohistochemistry for CD8+ and CD3+ in MyBrCa dataset, comparing nulliparous (N) and parous (P) women with differing breastfeeding histories prior to cancer diagnosis (bf: any recorded breastfeeding activity). CD8+ and CD3+ T cell density quantified as ratio of stain-positive pixels to all pixels within tumour margins, represented as a percentage. Modelled with beta regression adjusting for covariates of age at diagnosis and tumour grade. Unadjusted p values with the AME for each group compared to N are presented. Box plots: horizontal bar shows median, hinges represent IQR, and whiskers extend to the most extreme dot point within 1.5xIQR, and points beyond show outliers. Individual points show data from each case, coloured according to the status shown on the x-axis. Exact p values are shown.
