Hyperprogressive Disease During PD-1/PD-L1 Blockade in Patients With Non-Small-Cell Lung Cancer

C. G. Kim; K. H. Kim; K.-H. Pyo; C.-F. Xin; M. H. Hong; B.-C. Ahn; Y. Kim; S. J. Choi; H. I. Yoon; J. G. Lee; C. Y. Lee; S. Y. Park; S.-H. Park; B. C. Cho; H. S. Shim; E.-C. Shin; H. R. Kim


Ann Oncol. 2019;30(7):1104-1113. 

In This Article


In this study, HPD as defined according to both TGK and TGR was observed in 18.9% (45/237) of NSCLC patients treated with PD-1/PD-L1 inhibitors. HPD was associated with worse PFS and OS. Although clinicopathologic variables failed to identify subpopulations with HPD, lower frequencies of effector/memory subtypes (CCR7CD45RA) in CD8+ T cells and higher frequencies of severely exhausted cells (TIGIT+) in tumor-reactive PD-1+CD8+ T cells predicted HPD.

The rate of HPD in our study is similar to that of previous studies (range 9.2%–29.4%) that included patients with various cancer types.[10,11] Recently, Ferrara et al.[13] reported that 13.8% of patients with NSCLC experienced HPD due to PD-1/PD-L1 blockade. As defining HPD based on TGK or TGR requires three consecutive CT scans, there is a possibility of underestimating the HPD rate. However, because our investigation along with other previous studies suggest that HPD is a relatively common pattern of progression due to PD-1/PD-L1 blockade, treating physicians should be more watchful in detecting HPD.

Because there is no consensus on the accurate definition of HPD, we applied and compared different definitions in our patient cohort. The concordance rate of TTF and TGK or TGR in defining HPD was relatively low, whereas that of TGK and TGR was high. The discordance of TTF and TGK or TGR in defining HPD may be due to the contrasting manner in estimating the parameters; TGK and TGR are based on the size of the target lesions, while TTF is based on the duration to treatment failure. This suggests that defining HPD based on tumor growth dynamics rather than TTF can be more reliable. However, defining HPD based on TTF could be clinically useful in some patients in whom TGK or TGR cannot be evaluated due to the lack of proper target lesions on radiologic examinations or lack of radiologic evaluation during the reference period. Furthermore, we also observed a heteroscedasticity phenomenon, where the absolute differences in the ratio of simple or log-transformed TGK and TGR between the reference and the experimental period were augmented in patients with smaller baseline SLD (supplementary Figure S15A and B, available at Annals of Oncology online). Collectively, these findings reveal the complementary nature of each parameter in defining HPD.

Supplementary Figure S15.

Heteroscedasticity phenomenon between TGK ratio (ratio of TGK between the reference and the experimental periods) and TGR ratio (ratio of TGR between the reference and the experimental periods) is prominent in patients with smaller SLD.(A) Correlation between SLD of target lesions and absolute differences in TGK ratio and TGR ratio. (B) Correlation between SLD of target lesions and absolute differences in log2(ratio of TGK+1) and log2(ratio of TGR+1). The red line and the gray zone represent the Lowessfit with 95% CI.

In our cohort, HPD was not associated with advanced age and EGFR mutation. The discrepancies in predictive factors for HPD among studies can be attributed to several reasons including differences in tumor biology according to anatomic locations, patient characteristics, immunotherapy agents, various definitions of HPD, and patterns of clinical practice by treating physicians between studies and regions. Therefore, further studies are needed to identify better biomarkers for HPD. To this end, the biologic basis of HPD needs to be further understood with patient-derived peripheral blood and tumor tissue. In this study, we investigated whether the development of HPD was associated with parameters that were previously known to be related with treatment responses of anti-PD-1/PD-L1 treatment, including PD-L1 expression, tumor mutation burden, and major histocompatibility complex expression. However, these variables were not specific for HPD.

Du et al.[21] recently presented a case study suggesting cancer cell-intrinsic PD-1 expression to be a potential mechanism by which PD-1 blockade promotes tumor growth. Lo Russo et al.[22] also reported that reprogramming of tumor-associated macrophage upon Fc receptor engagement by antibodies of IgG4 isotype boosts tumor growth in a humanized mouse model. However, these studies hardly explain HPD induced by PD-L1 blockade, and HPD developed after treatment with blocking antibodies other than IgG4 isotype. Because PD-1/PD-L1 blockade restores antitumor immunity by removing inhibitory signals for T lymphocytes,[5] we focused our analysis on phenotyping CD8+ T lymphocytes from the peripheral blood before PD-1/PD-L1 blockade. Pre-existing antitumor immunity is essential for the spontaneous regression of tumor and predicts patient response to immune checkpoint blockade.[23,24] Thus, we measured the pre-established antitumor immunity by analyzing the frequencies of effector/memory subtypes in CD8+ T lymphocytes. Lower frequencies of effector/memory subtypes predict HPD, highlighting the importance of pre-existing antitumor immune responses in preventing deleterious outcomes from PD-1/PD-L1 blockade. In addition, we also characterized PD-1+CD8+ T lymphocytes, in which tumor-reactive T cells are enriched.[20] T-cell exhaustion is progressive, and responsiveness to reinvigoration by PD-1/PD-L1 blockade is limited to the less exhausted subsets.[25] Hence, we focused on the expression of other immune checkpoints that are hallmarks of severe exhaustion in PD-1+CD8+ T cells. We found that the frequency of severely exhausted CD8+ T cells (TIGIT+PD-1+CD8+ T cells) is higher in patients with HPD than that in non-HPD patients, indicating that the depth of T-cell exhaustion is associated with the development of HPD after PD-1/PD-L1 blockade. To our knowledge, this is the first study that explains HPD in terms of adaptive antitumor immunity, shedding light on the critical points in the basic biology of PD-1/PD-L1 blockade-mediated T-cell reinvigoration, particularly in terms of pre-existing antitumor immunity and severity of T-cell exhaustion. We also substantiated the findings from immunophenotyping of peripheral blood CD8+ T lymphocytes through mechanistic studies focusing on activation-induced cell death. Intriguingly, CCR7CD45RA+, TIGIT+PD1+ CD8+ T cells were prone to apoptosis, suggesting that anti-PD-1 treatment may result in not only reinvigoration, but also depletion of some subsets of CD8+ T cells such as CCR7CD45RA+, TIGIT+PD1+ CD8+ T cells by enhancing activation-induced cell death.

In conclusion, our study demonstrated that PD-1/PD-L1 blockade caused HPD in a number of patients with NSCLC, and this disease flair-up is linked to poor outcomes. To our knowledge, this is the first study to document the incidence of HPD in NSCLC patients within the Asian population with the second largest sample size. We also found for the first time that the degree of pre-existing antitumor immunity and depth of T-cell exhaustion measured in peripheral CD8+ T lymphocytes are associated with HPD. These results suggest that phenotyping of CD8+ T lymphocytes before PD-1/PD-L1 blockade has clinical implications in predicting HPD. In particular, the use of anti-PD-1/PD-L1 monotherapy should be carefully considered in patients with CD8+ T lymphocytes exerting suboptimal antitumor immunity or severely exhausted phenotypes. Furthermore, prospective validation of these biomarkers is essential to guide the proper use of PD-1/PD-L1 blockade and prevent adverse treatment outcomes.