Patient characteristics

A total of 220 patients with axillary metastases who underwent NACT were included in the study. Over time, statistically significant increases were observed in the proportion of NACT patients with unifocal cancers (78 to 90%, p = 0.01) and grade 3 cancers (28 to 47%, p = 0.007). Otherwise, there were no statistically significant differences in the cohort characteristics between the two treatment periods (Table 1). Of note, amongst the high TILs group, 96% (64/66) and 76% (35/46) of patients had pre-TILhigh and post-TILshigh within the 10–49% category, respectively with only few allocated into the higher TILs category (over 50%).

Table 1 This showed overall cohort and tumor characteristics for the whole study period 2013–2020

Axillary metastases were diagnosed through fine needle aspiration in 74% (162/220), core biopsy in 14% (n = 30/220), SLNB before NACT in 8% (17/220), and clinical examination only in 5% (11/220) of the patients. Amongst the eleven patients with suspicious lymph nodes based on clinical examinations, seven were found to have axillary metastases after NACT. For the four patients, whose axillary metastases were not histopathologically detected after NACT, three had no evidence of regression of previously suspected metastases whereas one had lymph nodes with evidence of tumor regression.

TILs in relation to pathological tumor responses

Proportions of patients with tumors containing TILslow and TILshigh were compared with the respective pCR rates in the axilla, breast, and combined (Table 2).

Table 2 Table showed associations of pre-TILs, post-TILs and ΔTILs with the of axillary, breast and combined (both breast and axilla) pCR status after neoadjuvant chemotherapy

The presence of pre-TILshigh was statistically significantly associated with higher pCR rates compared to pre-TILslow in the axilla (58.2 vs. 35.9%, p = 0.002), breast (56.7 vs. 31.4%, p < 0.001), and combined (49.3 vs. 25.5%, p < 0.001). Post-TILshigh was conversely associated with lower pCR rates compared to post-TILslow in the axilla (26.1 vs. 47.1%, p = 0.01), breast (17.4 vs. 44.8%, p = 0.01), and combined (13.0 vs. 37.9%, p < 0.001). Additionally, a ΔTILsdecrease after NACT was significantly associated with the highest pCR rates compared to ΔTILsno change or ΔTILsincrease in the axilla (65.3 vs. 39.9% and 17.9%, p < 0.001), breast (69.4 vs. 33.6% and 14.3%, p < 0.001), and combined (59.2 vs. 28.7% and 7.1%, p < 0.001) (Table 2).

TILs as a predictor of pathological response

In the univariable analyses, the presence of pre-TILshigh was a statistically significant predictor for pCR in the axilla (OR 2.48; 95% CI 1.38–4.46; p = 0.002), breast (OR 2.87; 95% CI 1.59–5.18; p < 0.001) and combined (OR 2.84; 95% CI 1.56–5.18; p < 0.001). On the contrary, post-TILshigh was a predictor of non-pCR in the axilla (OR 0.40; 95% CI 0.19–0.82; p = 0.01), breast (OR 0.26; 95% CI 0.11–0.59; p = 0.001) and combined (OR:0.25; 95% CI:0.10–0.61; p = 0.003). The presence of ΔTILsdecrease after NACT was associated with a higher likelihood of pCR in the axilla (OR 2.84; 95% CI 1.44–5.99; p = 0.003), breast (OR 4.49; 95% CI 2.23–9.03; p < 0.001) and combined (OR 3.61; 95% CI 1.84–7.08; p < 0.001), compared with ΔTILsno change. On contrary, ΔTILsincrease was associated with a lower likelihood of pCR in all three groups (Table 3).

Table 3 Univariable regression analyses of pre-TILs, post-TILs and ΔTILs in relation to axillary, breast and combined (both breast and axilla) pCR status after neoadjuvant chemotherapy

Other factors such as age (OR 0.97; 95% CI 0.95–1.00; p = 0.03) and BMI (OR 0.94; 95% CI 0.89–0.99; p = 0.04) were statistically significantly associated with lower likelihood of axillary pCR but not with breast or combined pCR. Tumor subtypes including HER2 + /luminal (OR 3.92; 95% CI 1.71–9.03; p = 0.001), HER2 + /non-luminal (OR 24.1; 95% CI 9.31–62.2; p < 0.001), and TNBC (OR 4.95; 95% CI 2.11–11.6; p < 0.001) were statistically significantly associated with higher likelihood of axillary pCR. These same tumor subtypes were also statistically significantly associated with higher likelihood for breast and combined pCR (Table 3).

Separate multivariable analyses were conducted for each of the three TILs categories, adjusted for the factors significant in univariable analysis (age, BMI, and tumor subtype). The presence of pre-TILshigh was an independent predictor for pCR in the axilla (OR 2.03; 95% CI 1.02–4.05; p = 0.04), breast (OR 2.71; 95% CI 1.31–5.62; p = 0.007) and combined (OR 2.70; 95% CI 1.33–5.48; p = 0.006). Conversely, post-TILshigh was independently predictive of non-pCR in the axilla (OR 0.33; 95% CI 0.14–0.76; p = 0.009), breast (OR 0.17; 95% CI 0.07–0.44; p < 0.001) and combined (OR 0.20; 95% CI 0.08–0.55; p = 0.002). The presence of ΔTILsdecrease after NACT was independently predictive of pCR in the breast (OR 4.02; 95% CI 1.71–9.42; p = 0.001) and combined (OR 2.99; 95% CI 1.36–6.54; p = 0.006) but did not reach statistical significance for axillary response (OR 2.05; 95% CI 0.94–4.47; p = 0.07). Conversely, ΔTILsincrease was an independent predictor of non-pCR in the axilla (OR 0.25; 95% CI 0.08–0.79; p = 0.02), breast (OR 0.22; 95% CI 0.07–0.74; p = 0.02) and combined (OR 0.14; 95% CI 0.03–0.67; p = 0.01) (Table 4). Another independent predictor was tumor subtype for pCR in the axilla, breast and combined pCR, irrespective of TILs (Table 4).

Table 4 Multivariable analyses of TILs-based separately on pre-TILs, post-TILs and ΔTILs, all of which adjusted for clinicopathological factors like age, BMI and tumor subtypesPerformance characteristics of TILs

In predicting axillary pCR, performance characteristics of pre-TILs, post-TILs and ΔTILs were, respectively, as follows: 42% vs. 13% vs 34% for sensitivity, 78% vs. 73% vs. 87% for specificity, 58% vs. 26% vs. 65% for PPV and 64% vs. 53% vs. 64% for NPV. For predicting breast pCR, performance characteristics of pre-TILs, post-TILs and ΔTILs were, respectively, 44% vs. 9% vs. 40% for sensitivity, 78% vs. 72% vs. 89% for specificity, 43% vs. 17% vs. 69% for PPV and 68% vs. 55% vs. 69% for NPV. And to predict combined pCR, performance characteristics of pre-TILs, post-TILs and ΔTILs were, respectively 46% vs. 8% vs. 40% for sensitivity, 77% vs. 73% vs. 87% for specificity, 49% vs. 13% vs. 59% for PPV and 75% vs. 62% vs. 75% for NPV. All performance characteristics were statistically significant except for post-TILs in predicting axillary pCR (Table 5).

Table 5 To assess the diagnostic accuracy of TILs as a predictor of pCR, performance characteristics of pre-TILs, post-TILs and ΔTILs as a predictor were assessed based on sensitivity, specificity, positive predictive value, and negative predictive value using receiver operator characteristics and area-under-curve analysesAxillary pCR and TILs in relation to survival outcome

Since all three categories of TILs were statistically significantly related to pCR in the axilla as well as the breast, Cox regression analyses were conducted based on only pre-TILs and axillary pCR, aligning with the study aims. Additionally, regression analyses were performed using different combinations of pre-TILs and axillary pCR as a single factor and these were grouped into three categories, namely, axillary pCR with pre-TILshigh, heterogeneous combinations and no axillary pCR with pre-TILslow (Table 6).

Table 6 Cox regression analysis of axillary pCR and pre-TILs in relation to breast cancer-free interval and overall survival outcome

The average duration of follow-up was 4.9 years (range: 0.8 to 9.7 years). During the follow-up period, 4.1% (9/220) developed breast-only recurrences, while 5.5% (12/217) had axillary-only recurrences. Additionally, 22.7% (50/220) developed distant metastases only, of which 82% (41/50) had pre-TILslow tumors. The overall mortality was 18.2% (40/220), including three patients who died from causes unrelated to breast cancer. Of patients who died of breast cancers, 76% (28/37) had low TILs and 24% (9/37) had high TILs, both before and after NACT, respectively. However, noticeably, in the high TILs groups, there were no deaths amongst those patients who had large amount of TILs (> 50%) before and after NACT.

The presence of pre-TILshigh was an independent prognostic factor for BCFI (HR 0.46; 95% CI 0.22–0.96; p = 0.04) but not for OS (HR 0.73; 95% CI 0.35–1.55; p = 0.42). On the other hand, axillary pCR was both a prognostic factor for BCFI (HR 0.28; 95% CI 0.12–0.63; p = 0.002) and OS (HR 0.28; 95% CI 0.12–0.68; p = 0.004). When considering the combination of axillary pCR wih pre-TILshigh as a single factor, it was an independent prognostic factor for OS (HR 0.09; 95% CI 0.12–0.72; p = 0.02) compared to the combination of axillary non-pCR with pre-TILslow (Table 5 and Fig. 2). In this study cohort, TNBC subtype and NACT completion were the other two factors statistically significantly associated with BCFI and OS (Table 6).

Fig. 2

Breast cancer-free interval and overall survival based on axillary pCR (A1 & A2), pre-TILs (B1 & B2) and axillary pCR combined with pre-TILs (C1 & C2). Heterogeneous combinations defined as various combinations of axillary pCR and pre-TILs other than category-1 and -3 for example axillary pCR & low pre-TILs. Statistical Kaplan Meir survival analysis was used with log-rank test p < 0.05 defined as statistically significant. Abbreviations: pCR: complete pathological response, pre-TILs: tumor-infiltrating lymphocytes before neoadjuvant chemotherapy