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

Disclosures

Ann Oncol. 2019;30(7):1104-1113. 

In This Article

Results

Patient Characteristics

A total of 263 patients were included in the analysis (Table 1). The median age was 63 years, and majority were men (191/263, 72.6%), smokers (168/263, 63.9%), and had relatively good performance status [ECOG 0 or 1: 203/263 (77.2%)]. The number of previous treatment varied between patients and ranged from 0 to 8 lines. PD-L1 immunohistochemistry was carried out in most patients (221/263, 84.0%) and was positive in more than two-thirds of the patients (152/221, 68.8%). EGFR mutation (31/263, 11.8%), ALK rearrangement (3/263, 1.1%), and ROS1 rearrangement (5/263, 1.9%) were detected only in patients with adenocarcinoma. The majority of patients received PD-1 blockade (246/263, 93.5%). Partial response (PR), stable disease (SD), and PD were achieved in 52 (19.8%), 112 (42.6%), and 99 (37.6%) patients, respectively.

Incidence of HPD According to Different Definitions

To evaluate the incidence of HPD, we analyzed the dynamics of tumor growth by comparing CT scans from the reference period and the experimental period as well as the time from the treatment initiation to discontinuation. As shown in Figure 1, deleterious effects by immune checkpoint blockade, represented by accelerated tumor growth and early failure to the treatment, were observed in some cases. We then compared the incidence of HPD based on the various definitions. In our cohort, 55 (20.9%), 54 (20.5%), and 98 (37.3%) patients experienced HPD as defined according to TGK, TGR, and TTF, respectively. The concordance rate of each criterion was higher between TGK and TGR (supplementary Figure S1A, available at Annals of Oncology online), implying that the definition based on the tumor growth dynamics may be interchangeable and may be used more universally than the definition based on TTF in NSCLC patients. Even if the definition of TTF was changed (<1 or 3 months), the trend remained unchanged (supplementary Figure S1B and C, available at Annals of Oncology online). When we analyzed the variations of TGK (Figure 2A) and TGR (Figure 2B) between the reference and the experimental periods across patients determined to achieve PD as their best response, the similarity of the two criteria was also evident (supplementary Figure S2, available at Annals of Oncology online). To exclude the possibility of overestimating the incidence of HPD, HPD was limited to only cases meeting both definitions based on TGK and TGR in further analysis. Patients who fulfilled both criteria also experienced more rapid progression based on the RECIST 1.1 criteria (Figure 2C). When the analysis was expanded to the total cohort, there was an inverse correlation between TGK or TGR during the reference period and the treatment response according to RECIST 1.1 at the first evaluation (supplementary Figure S3A and B, available at Annals of Oncology online).

Figure 1.

A case of HPD during PD-1/PD-L1 blockade. A 65-year-old female patient with metastatic NSCLC with EGFR mutation received PD-1 inhibitor as the fourth line of treatment. CT (A) and radiography (B) scans before (−8 weeks), at baseline, and at first evaluation (+8 weeks) are displayed in chronologic order. The SLD of the target lesions was 36, 44, and 78 mm on CT scans before, at baseline, and at first evaluation, respectively (C). She died shortly after treatment discontinuation due to disease progression accompanied with massive hematolymphangitic metastasis.

Figure 2.

Analysis of tumor growth dynamics between the reference and the experimental periods in patients with PD as the best response. (A) Comparison of TGK between the reference and the experimental periods. (B) Comparison of TGR between the reference and the experimental periods. Because the TGK and TGR are not normally distributed, log2 (x + 1) transformation was used. (C) Change in SLD of the target lesions according to RECIST 1.1 criteria between the reference and the experimental periods.

Supplementary Figure S1.

Concordance rate between different definitions of HPD. (A) Concordance rate between TGK, TGR, and TTF <2 months. (B) Concordance rate between TGK, TGR, and TTF <1 month. (C) Concordance rate between TGK, TGR, and TTF <3 months. Hierarchical clustering results were displayed as a dendrogram. The numbers indicate the concordance rate between each definition.

Supplementary Figure S2.

Comparison of the ratio of TGK and TGR between the reference and the experimental periods.Linear regression model was used to calculate R2and P value.

Supplementary Figure S3.

Treatment response according to RECIST 1.1 criteria during the experimental period is inversely correlated with TGK and TGR during the reference period.(A) Comparison of the TGK during the reference period and change of target lesion during the experimental period. (B) Comparison of the TGR during the reference period and change of target lesion during the experimental period. The red line and the gray zone represent the Lowessfit with 95% CI.

Association Between HPD, Outcomes, and Clinicopathologic Variables

We then investigated the predictive value of HPD for survival outcomes. To this end, we categorized responses to the treatment according to the following classes: PR, SD, PD without HPD, and PD with HPD (HPD). There was clear separation of PFS duration between patients with different responses (Figure 3A, overall log-rank P < 0.001). The median PFS of patients with PD without HPD was 48 days, whereas that of patients with HPD was 19 days (HR 4.619; 95% CI 2.868–7.440). Similar trends were obtained for OS (Figure 3B, overall log-rank P < 0.001). The median OS of patients with PD without HPD was 205 days, whereas that of patients with HPD was 50 days (HR 5.079; 95% CI 3.138–8.226). In addition, patients with HPD lost the opportunity for subsequent treatment (supplementary Figure S4A and B, available at Annals of Oncology online).

Figure 3.

PFS (A) and OS (B) according to response categories.

Supplementary Figure S4.

HPD is associated with decreased probability of receiving subsequent treatments. (A) Lines of subsequent treatments in patients with non-HPD and HPD. (B) Lines of subsequent treatments in patients with PR/SD, PD without HPD, and HPD. The analysis was performed in patients who died (N=147).

Next, we compared the variables widely used in routine clinical practice between patients with non-HPD (PR, SD, and PD without HPD) and with HPD. Although a previous report suggested that advanced age (>65 years) was associated with HPD,[10] we could not find any relationship between age and HPD (Table 1). In addition, patients with HPD were likely to have much more metastatic sites, liver metastasis, and elevated level of LDH (Table 1), while these variables were not different between PD with and without HPD (supplementary Table S1, available at Annals of Oncology online). Prognostic scores (RMH, GRIM, and LIPI score) were also higher in HPD than those in non-HPD (Table 1; supplementary Table S2, available at Annals of Oncology online). However, they could not discriminate between PD with and without HPD (supplementary Table S1, supplementary Table S2 and supplementary Table S3, available at Annals of Oncology online). Specifically, in contrast to a previous finding,[12] EGFR alteration represented by oncogenic EGFR mutation was not associated with HPD. In addition, tumor burden at baseline estimated by the SLD of the target lesions was not associated with HPD (supplementary Figure S5A and B, available at Annals of Oncology online). Furthermore, tumor mutation burden and the expression of major histocompatibility complex class I in tumor were not specific for HPD (supplementary Figure S6A, B and Table S4, available at Annals of Oncology online).

Supplementary Figure S5.

HPD is not associated with baseline tumor burden. (A) SLD of target lesions between non-HPD and HPD patients at baseline. (B) SLD of target lesions in patients with PR/SD, PD without HPD, and HPD at baseline.

Supplementary Figure S6.

HPD is not associated with tumor mutation burden (TMB). (A) TMB of tumor tissue between non-HPD and HPD patients at baseline. (B) TMB of tumor tissue of patients with PR/SD, PD without HPD, and HPD at baseline. Tumor mutation burden was calculated based on next generation sequencing data of Illumina TST 170 platform.

Biomarker Analysis With Peripheral CD8+ T Lymphocytes

To identify the potential biomarkers of HPD, we focused on CD8+ T lymphocytes considering their major role in antitumor immunity. To further categorize peripheral CD8+ T lymphocytes, we analyzed the frequency of effector/memory cells using two well-known markers, namely, CCR7 and CD45RA.[19] Intriguingly, frequencies of effector/memory subtype (CCR7CD45RA) in total CD8+ T cells were higher in patients with non-HPD than those in patients with HPD (Figure 4A and B). In addition, we analyzed the frequencies of PD-1+ cells in total CD8+ T lymphocytes, given that PD-1+CD8+ T lymphocytes from the peripheral blood are representatives of tumor-reactive populations.[20] The frequencies of PD-1+ cells in CD8+ T cells were not different between patients with non-HPD and with HPD (supplementary Figure S7A and B, available at Annals of Oncology online). To further elucidate the exhaustion status of PD-1+CD8+ T cells based on other immune checkpoints, we analyzed the frequency of TIGIT+CD8+ T cells (severely exhausted tumor-reactive CD8+ T cells) in PD-1+CD8+ T cells. Intriguingly, elevated frequencies of severely exhausted tumor-reactive CD8+ T cells predicted HPD (Figure 4C and D). In addition, decreased frequency of effector/memory CD8+ T cells and increased frequency of severely exhausted tumor-reactive CD8+ T cells were largely limited to the patients with HPD rather than patients with PD without HPD (Figure 4E and F). This indicates that HPD is a unique biologic process distinct from simple disease progression. Furthermore, the performance power of each biomarker was relatively fair (supplementary Figure S8A and B, available at Annals of Oncology online). To explore whether the above identified biomarkers can predict survival outcomes, we computed these two variables in the survival analysis. The results showed that these two biomarkers independently predict prognosis in terms of PFS (Figure 5A and B) and OS (Figure 5C and D). In addition, the combination of these two biomarkers had prognostic implications (supplementary Figure S9A and B, available at Annals of Oncology online). Phenotypically, expression of CD28 was higher in CCR7CD45RA and TIGITPD1+ CD8+ T cells compared with their counterparts (supplementary Figure S10A and B, available at Annals of Oncology online). Mechanistically, CCR7CD45RA+, TIGIT+PD1+ CD8+ T cells were vulnerable to activation-induced cell death during antigen recognition compared with their counterparts (supplementary Figures S11A, B and S12A, B, available at Annals of Oncology online). In addition, CCR7CD45RA+, TIGIT+PD1+ CD8+ T cells showed impaired functionality during antigen recognition (supplementary Figures S11C and S12C, available at Annals of Oncology online). Moreover, the expression of other immune checkpoints such as CTLA-4 was higher in TIGIT+PD-1+CD8+ T cells than TIGITPD-1+CD8+ T cells (supplementary Figure S12D, available at Annals of Oncology online). Intriguingly, the patients without CD8+T-cell infiltration to tumor (negative CD8+ tumor-infiltrating lymphocytes) had significantly lower frequency of CCR7CD45RA among CD8+ T cells and higher frequency of TIGIT+PD-1+ among PD-1+CD8+ T cells from peripheral blood, which were identified as predictive factors for HPD (supplementary Figure S13A and B, available at Annals of Oncology online). However, SLD at baseline was not correlated with the frequency of CCR7CD45RA among CD8+ T cells and frequency of TIGIT+PD-1+ among PD-1+CD8+ T cells from peripheral blood (supplementary Figure S14A and B, available at Annals of Oncology online).

Figure 4.

HPD is associated with a lower frequency of effector/memory subsets and a higher frequency of severely exhausted populations in CD8+ T lymphocytes. (A) Frequency of CCR7CD45RA T cells among the total CD8+ T cells between non-HPD and HPD patients. (B) Representative figures for the frequency of CCR7CD45RA T cells among the total CD8+ T cells. (C) Frequency of TIGIT+ T cells among PD-1+CD8+ T cells between non-HPD and HPD patients. (D) Representative figures for the frequency of TIGIT+ T cells among the PD-1+CD8+ T cells. (E) Frequency of CCR7CD45RA T cells among the total CD8+ T cells in patients with PR/SD, PD without HPD, and HPD. (F) Frequency of TIGIT+ T cells among the PD-1+CD8+ T cells in patients with PR/SD, PD without HPD, and HPD.

Figure 5.

Lower frequency of effector/memory subsets and higher frequency of severely exhausted populations in CD8+ T lymphocytes are associated with worse outcome. (A, B) PFS according to the frequency of CCR7CD45RA T cells among the total CD8+ T cells (Effector/memoryhigh versus Effector/memorylow) and the frequency of TIGIT+ T cells among the PD-1+CD8+ T cells (TIGIThighversus TIGITlow). (C, D) OS according to the frequency of CCR7CD45RA T cells among the total CD8+T cells (Effector/memoryhigh versus Effector/memorylow) and the frequency of TIGIT+ T cells among the PD-1+CD8+ T cells (TIGIThigh versus TIGITlow). Cut-off values were determined by analyzing receiver-operating characteristic curves.

Supplementary Figure S7.

Frequency of PD-1+T cells among total CD8+T cells. (A) Frequency of PD-1+T cells among total CD8+T cells in patients with non-HPD and HPD. (B) Frequency of PD-1+T cells among total CD8+T cells in patients with PR/SD, PD without HPD, and HPD.

Supplementary Figure S8.

Performance power of frequency of CCR7-CD45RA-T cells among the total CD8+T cells (A) and frequency of TIGIT+T cells among the PD-1+CD8+T cells (B) in predicting HPD.AUC, area under the curve; PPV, positive predictive value; NPV, negative predictive value.

Supplementary Figure S9.

Lower frequency of effector/memory subsets and higher frequency of severely exhausted populations in CD8+T lymphocytes are associated with worse outcome.(A) PFS and (B) OS according to the combination of frequency of CCR7-CD45RA-T cells among the total CD8+T cells (Effector/memoryhighvs. Effector/memorylow) and the frequency of TIGIT+T cells among the PD-1+CD8+T cells (TIGIThighvs. TIGITlow). Cutoff values were determined by analyzing receiver-operating characteristic curves.

Supplementary Figure S10.

Expression of CD28 in CD8+T cells.(A) Frequency of CD28+T cells in CCR7-CD45RA-CD8+T cells and CCR7-CD45RA+CD8+T cells. (B) Frequency of CD28+T cells in PD-1+TIGIT-CD8+T cells and PD-1+TIGIT+CD8+T cells. Pair-wise t-test was performed.

Supplementary Figure S11.

Characterization of CCR7-CD45RA-CD8+T cells.(A) Frequency of apoptotic cells (AnnexinV+) in CCR7-CD45RA-and CCR7-CD45RA+CD8+T cells upon T-cell receptor stimulation in the presence of anti-PD-1 antibody. (B) Representative figures for apoptotic cells in CCR7-CD45RA-and CCR7-CD45RA+CD8+T cells upon T-cell receptor stimulation in the presence of anti-PD-1 antibody. (C) Production of effector molecules (IFN-γ, TNF, granzymeB, and perforin) by CCR7-CD45RA-and CCR7-CD45RA+CD8+T cells upon T-cell receptor stimulation in the presence of anti-PD-1 antibody. Pair-wise t-test was performed.

Supplementary Figure S12.

Characterization of PD-1+TIGIT+CD8+T cells.(A) Frequency of apoptotic cells (AnnexinV+) in PD-1+TIGIT-and PD-1+TIGIT+CD8+T cells upon T-cell receptor stimulation in the presence of anti-PD-1 antibody. (B) Representative figures for apoptotic cells in PD-1+TIGIT-and PD-1+TIGIT+CD8+T cells upon T-cell receptor stimulation in the presence of anti-PD-1 antibody. (C) Production of effector molecules (IFN-γ, TNF, granzymeB, and perforin) by PD-1+TIGIT-and PD-1+TIGIT+CD8+T cells upon T-cell receptor stimulation in the presence of anti-PD-1 antibody. (D) Frequency of CTLA-4+cells among PD-1+TIGIT-and PD-1+TIGIT+CD8+T cells. Pair-wise t-test was performed.

Supplementary Figure S13.

Frequency of CCR7-CD45RA-T cells among the total CD8+T cells (A) and frequency of TIGIT+T cells among PD-1+CD8+T cells (B) from peripheral blood according to the positivity of CD8+T cell infiltration to tumor tissue.

Supplementary Figure S14.

Baseline tumor burden was not correlated with the frequency of CCR7-CD45RA-T cells among total CD8+T cells and the frequency of TIGIT+T cells among PD-1+CD8+T cells.(A) Correlation between SLD and the frequency of CCR7-CD45RA-T cells among total CD8+T cells. (B) Correlation between SLD and the frequency of TIGIT+T cells among PD-1+CD8+T cells. Pearson correlation analysis was performed.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....