Baseline immune signature score of Tregs × HLA-DR+CD4+ T cells × PD1+CD8+ T cells predicts outcome to immunotherapy in cancer patients

Abstract

BackgroundThe use of immunotherapy (IT) is rapidly increasing across different tumor entities. PD-L1 expression is primarily used for therapy evaluation. The disadvantages of PD-L1 status are spatial and temporal heterogeneity as well as tumor type-dependent variation of predictive value. To optimize patient selection for IT, new prediction markers for therapy success are needed. Based on the systemic efficacy of IT, we dissected the immune signature of peripheral blood as an easily accessible predictive biomarker for therapeutic success.MethodsWe conducted a retrospective clinical study of 62 cancer patients treated with IT. We assessed peripheral immune cell counts before the start of IT via flow cytometry. The predictive value for therapy response of developed immune signature scores was tested by ROC curve analyses and scores were correlated with time to progression (TTP).ResultsHigh score values of “Tregs ÷ (CD4+/CD8+ ratio)” (Score A) and high score values of “Tregs × HLA-DR+CD4+ T cells × PD1+CD8+ T cells” (Score B) significantly correlated with response at first staging (p = 0.001; p < 0.001). At the optimal cutoff point, Score A correctly predicted 79.1% and Score B correctly predicted 89.3% of the staging results (sensitivity: 86.2%, 90.0%; specificity: 64.3%, 87.5%). A high Score A and Score B statistically correlated with prolonged median TTP (6.13 vs. 2.17 months, p = 0.025; 6.43 vs. 1.83 months, p = 0.016). Cox regression analyses for TTP showed a risk reduction of 55.7% (HR = 0.44, p = 0.029) for Score A and an adjusted risk reduction of 73.2% (HR = 0.27, p = 0.016) for Score B.ConclusionThe two identified immune signature scores showed high predictive value for therapy response as well as for prolonged TTP in a pan-cancer patient population. Our scores are easy to determine by using peripheral blood and flow cytometry, apply to different cancer entities, and allow an outcome prediction before the start of IT.