Background: neoMono randomized patients (pts) with triple negative breast cancer (TNBC) to receive (Arm A) or not to receive (Arm B) 2-week Atezolizumab monotherapy prior to neoadjuvant 24-week Atezolizumab + chemotherapy. In prespecified interim analysis, no difference regarding pathological complete response (pCR) was observed among all pts. However, in regard to pCR rates, pts with PD-L1 positive TNBC seem to benefit from addition of the monotherapy window. Here we analyze the association between central TILs and Ki-67 at baseline (BL) / after 2 weeks (2w) and pCR. Methods: Ki-67 expression (% positive tumor cells) was analyzed using immunohistochemistry. Analysis of stromal TILs was performed according to the International TILs Working Group. Differences from BL at the 2w time point were denoted as Ki-67-diff and TIL-diff, respectively. Associations between pCR and BL/2w measurements were analyzed using logistic regression. Akaike information criterion (AIC) was used to evaluate goodness-of-fit between models. Since the a priori chosen biomarker thresholds ( < = 30% [Ki-67] and < = 60% [TILs]) resulted in very low subgroup sizes, k-means clustering analysis was performed and association of the cluster labels with pCR was then investigated. Results: The analysis included 50 (Arm A) and 51 (Arm B) pts. The threshold of Ki-67 ( < = 30%) was reached in 98% (BL) and 92% (2w), that of TILs ( < = 60%) in 11% (BL) and 18% (2w), respectively. PD-L1 (BL) was < = 1 in 26%. In separate logistic regression models (including BL and change), both BL-Ki-67 and BL-TILs were significantly though moderately associated with pCR. A combined model using both biomarkers yielded the best AIC. BL-Ki-67 and BL-TILs, as well as TIL-diff were highly significantly though again moderately associated with pCR. Cluster analysis I (using BL-Ki-67, Ki-67-diff, and PD-L1) yielded cluster 1 (medium BL-Ki-67 and subsequent increase) with a pCR rate of 50% compared to 77% in cluster 2 (high BL-Ki-67 and decreasing Ki-67). Similarly, cluster analysis II (using BL-TILS, TILS-diff and PD-L1) yielded cluster 3 (low BL-TILs/non-essential TILs change) with a low pCR rate of 55% compared to 84.4% and 85.7% for clusters 4 (low BL-TILs /substantial increase) and 5 (high BL-TILs/non-essential change), respectively. In a combined regression model of all clusters above, clusters 2 and 4 had the highest pCR probability. Conclusions: Our results demonstrate that while BL-Ki-67 and BL-TILs are highly informative of pCR probability, Ki-67-diff and even more so TIL-diff added further significant information. Analysis of the full neoMono dataset will allow us to investigate this association further and to potentially develop a predictive algorithm regarding pCR based on baseline and dynamic biomarkers.