Tumour-infiltrating lymphocytes and response to neoadjuvant letrozole in patients with early oestrogen receptor-positive breast cancer: analysis from a nationwide phase II DBCG trial

Patients were treated with neoadjuvant letrozole for 4 months prior to curative intended surgery as part of a clinical phase II study conducted by the Danish Breast Cancer Group (DBCG) between 2009 and 2012. The study is registered on clinicaltrials.​gov (NCT00908531). The original study design and clinical results have been previously published [13]. Endpoints were clinical and pathological outcome. A total of 119 patients were registered to receive neoadjuvant letrozole. Eligible patients had histological confirmed, invasive, ER+, HER2-negative, operable breast cancer. They met the following criteria: tumour size ≥ 1 cm, ≥ 60 years at entry, Eastern Cooperative Oncology Group score 0–2 and Charlson comorbidity index 0–2. Patients with prior cytotoxic treatment including aromatase inhibitors and prior malignant disease were not eligible. Patients were registered in the DBCG database and updated prospectively. Four patients were excluded prior to study initiation, two cases due to HER2 positivity at central testing and two patients withdrew consent. An additional three were excluded from the intention-to-treat population. One hundred six patients had paired tissue samples from biopsy and surgery and were included in the present study (supplementary Fig. 1).

ER, PGR, HER2 and Ki67 were assessed centrally using international standards [14‐16]. ER and Ki67 were recorded as continuous variables. Tumours were considered ER positive when nuclear staining was equal to or higher than 10%. TILs were assessed by the use of the guidelines of the International Immuno-Oncology Biomarker Working Group on Breast Cancer [17]. Pre- and posttreatment stromal TILs were quantified on haematoxylin and eosin-stained slides, from biopsies and surgical specimens respectively, as percentage infiltration of mononuclear cells. TILs were assessed both continuously and categorised into groups: low (0–9%), intermediate (10–59%) and high (60–100%). Changes in TILs were calculated as the difference between pre- and posttreatment TIL counts and categorised as decreased when TILs dropped ≥ 10 percentage point, as increased when TILs elevated ≥ 10 percentage point or as no change (less than 10 percentage point decrease or increase).

Endpoint was to determine the association between changes in TILs during neoadjuvant endocrine treatment and achieving a pathological response. Pathological response was defined as a decrease in tumour cells ≥ 30% according to a modified Miller-Payne grading scale used by the DBCG [8]. On the modified scale response, grade 1 equals no invasive cells present in the tumour bed, pathological complete response (pCR). Grade 2 more than 90% loss of tumour cells and grade 3 between 30 and 90% reduction in tumour cells were considered partial response. Grade 4 is defined as less than 30% loss of tumour cells and was considered no response. Exploratively pathological response was also assessed on the RCB index scoring RD both continuously and categorically using the guidelines presented by the BIG-NABCG collaboration [9, 10]. In brief, the RCB index combines the bidimensional diameter of the primary tumour with the percentage of invasive cells in the tumour, corrected for the percentages of in situ carcinoma, and the number of positive lymph nodes including the diameter of the biggest lymph node metastasis in a generalised linear model. The higher RCB index, and the corresponding RCB-class, the more extensive residual disease load. For the RCB index calculations in this study, we used the online Residual Cancer Burden Calculator provided by the MD Andersson Cancer Center [18]. As an additional exploratory analysis, we assessed response after the PEPI scoring system. The PEPI score combines pathological tumour size, ER, Ki67 and node status in a score from 0 to 12 where a low PEPI score indicates excellent prognosis.

Patients’ basic characteristics are summarised in Table 1. Median age was 67 years (range 60–87 years). All patients had pretreatment TIL assessment, with a baseline median of 5%, range 0–50%. Apart from Ki67, there were no statistically significant differences between patient with low and intermediate levels of pretreatment TILs in the different clinicopathological subgroups (Table 1).

Assessment of pathological response according to Miller-Payne was available for 102 patients (96%). Of them, 58 (57%) had pathological response including one patient with pCR. Forty-four (43%) had no response. Assessment of RCB was possible for all 106 patients: range RCB index 0–3.649, and when grouped into RCB-classes: one patient had RCB-class 0 (pCR), six (6%) had RCB-class I, 90 (83%) had RCB-class II and 12 (11%) had RCB-class III. Assessment of the PEPI score was possible for 104 patients, with 45 (43%) patients obtaining a PEPI score of 0 and 59 (57%) a PEPI score between 1 and 6 (mean score 1.7) (supplementary Table 1).

When divided into responders and non-responders, the increase in concentration was significant among non-responders with a mean increase of 10.6 (p < 0.0001) compared to responders (mean increase 3.2, p = 0.10); these data are presented in Table 2 and Fig. 1.

All clinicopathological factors listed in Table 1 are tested for association to pathological response. Tumour size below 2 cm, a ductal subtype, low posttreatment Ki67 and stable or decreasing TILs were predictive of pathological response in the univariate analysis. When tested in multivariable analyses, only TILs and subtype remained significant for pathological response. In the final model, OR estimates for pathological response were 0.71 (95% CI 0.53–0.96; p = 0.02) per 10% absolute increase in TILs and 0.37 (95% CI 0.15–0.91; p = 0.03) for non-ductal subtype as presented in Table 3. For the explorative endpoint, RCB TILs were significantly predictive for response with OR = 0.56 (95% CI 0.40–0.78; p 0.0007) per 10% absolute increase in TILs and posttreatment Ki67 was borderline significant with OR 0.77 (CI 0.60–1.00; p = 0.05) per 10% increase in Ki67 index. There was no significant association between TILs and achievement of a PEPI score of 0: OR 0.85 (CI 0.66–1.10; p = 0.22) per 10% increase in TILs.

Ki67 levels paralleled pretreatment TIL levels. The correlation between pretreatment TILs and Ki67 was moderate (Pearson 0.4; p = 0.0002); however, the association grew stronger posttreatment (Pearson 0.5; p < 0.0001). TILs were weakly correlated to ER (Pearson − 0.2; p = 0.02 pretreatment and − 0.3; p = 0.0006 posttreatment) and not significantly correlated to PgR (Pearson − 0.04; p = 0.70 pretreatment and − 0.2; p = 0.21 posttreatment).