DOI: https://doi.org/10.21203/rs.3.rs-1929731/v1
The PACIFIC study showed that patients with NSCLC after radio-chemotherapy derived a benefit in PFS and OS when treated with durvalumab, a PD-L1-inhibitor, vs placebo. In a post-hoc analysis this effect was limited to patients with a PD-L1 expression of > 1%, while patients < 1% did not benefit from durvalumab, partly because the outcome in the observational control arm was surprisingly favorable. Thus, it could be speculated that lack of PD-L1 expression confers a favorable outcome in stage III NSCLC.
We retrospectively analyzed PFS and OS in a group of 99 patients with stage III NSCLC homogeneously treated and not progressing after radio-chemotherapy. In all patients, PD-L1 expression was evaluated and outcome in the groups of PD-L1 < 1% and PD-L1 ≥ 1% was compared.
Clinical data, PD-L1 expression, predictive blood-markers and outcome were retrospectively captured from primary cases of the certified lung cancer center Oldenburg. Statistical analyses using log-rank test were performed.
The median OS of the patients with an expression of PD-L1 < 1% was 20 months (CI 10.5–29.5) and with an expression ≥ 1% 28 months (CI 16.5–39.2), (p = 0.734). The median PFS of the patients with an expression of PD-L1 < 1% was 9 months (CI 6.3–11.6) and with an expression ≥ 1% 12 months (CI 9.8–14.2), not statistically significant (p = 0.112). The blood parameters had no significant impact on the OS.
In this retrospective cohort of stage III patients, the assumption that lack of PD-L1 expression represents a favorable prognostic factor after radio-chemotherapy vs. PD-L1 expression > 1% was not confirmed.
The therapy of NSCLC has been revolutionized using immune checkpoint inhibitors (ICI). They are now established in the metastatic setting in first line as single agent or in combination with platin-based chemotherapy and have conferred both increased response rates, progress-free survival (PFS) and Overall survival (OS) compared to combination chemotherapy alone (1–4). PD-L1 in the metastatic setting has been established as a predictive factor for response, PFS and OS on ICI and is the only biomarker for stratification of ICI therapies, mainly due to licensing reasons (5). Data of PD-L1 expression regarding their prognostic impact are sparse. When comparing the control arms of the licensing studies such as Keynote 189, Impower 150 and Keynote 407 for the different PD-L1 expression subgroups (< 1%, ≥ 1–49%, ≥ 50%) no obvious differences in the outcome can be discerned suggesting that in the metastatic setting the prognostic impact of PD-L1 expression most likely is limited (6–8). In the PACIFIC study, patients with stage III not eligible for surgery were included and were treated with radio-chemotherapy with a radiotherapy of at least 54 Grey (Gy) (9). Patients who did not progress after radio-chemotherapy and that did not show any signs of radiation-induced pneumonitis were randomized to receive durvalumab vs. placebo. A statistically significant benefit for PFS and OS was shown for patients receiving durvalumab in the ITT population. The updated 5-year OS for Durvalumab was 47.5 months (38.4–52.6) versus 29.1 (22.1–35.1) for placebo (10). The OS HR was 0.71. In a post hoc analysis, the benefit for OS was statistically significant in the group with an expression of PD-L1 of ≥ 1%, but not in the group with PD-L1 expression of < 1%. One explanation why the PD-L1 < 1% group might not derive a benefit from durvalumab consolidation could be that the control arm of this cohort was significantly better than the control arm of the PD-L1 ≥ 1% group. Therefore, the hypothesis was generated that lack of PD-L1 expression could represent a favorable prognostic factor in NSCLC after radio-chemotherapy. To test this hypothesis, a retrospective cohort study of patients treated with radio-chemotherapy similar to the Pacific trial and in whom PD-L1 expression was known, was performed. The outcome (PFS and OS) was analyzed according to PD-L1 status (≥ 1% vs. <1%).
The study situation is limited. The study from K. Gennen et al. included 31 patients. They found that PD-L1 expression < 1% on tumor cells was associated with improved OS, PFS and local control in patients treated with concurrent CRT (11). In contrast to these results, the study from T. Tokito et al. with 74 patients in stage III NSCLC showed that PD-L1 expression was not correlated with PFS and OS(12). This has been confirmed by the study of A. Tufman et al. In their study 78 patients with stage III NSCLC were included and PD-L1 expression did not correlate with PFS following RTCT (13). Our cohort with 99 patients is largest cohort ever analyzed regarding the prognostic impact of PD-L1 expression after radio-chemotherapy.
Additional potentially prognostic serum markers (CRP and albumin) were evaluated. In a Chinese cohort, these parameters were deemed prognostically significant for PFS and OS in a population of patients including all stages (14). Therefore, our cohort was used to evaluate the prognostic significance of CRP and albumin after radio-chemotherapy in stage III.
To match the criteria of the Pacific study eligible patients for this retrospective analysis were primary cases of the lung cancer center Oldenburg, Pius-Hospital, had histologically or cytologically diagnosed NSCLC stage III by staging procedures including PET-CT and MRI of the brain. Staging was done according to the Staging Manual in Thoracic Oncology, version 8, of the International Association for the Study of Lung Cancer. The patients had received two or more cycles (defined according to local practice) of platinum-based chemotherapy (containing etoposide, vinblastine, vinorelbine, a taxane, or pemetrexed) concurrently with definitive radiation therapy (54 to 66Gy). The mean dose to the lung was less than 20Gy, the V20 (the volume of lung parenchyma that received 20 Gy or more) was less than 35%, or both. Additional inclusion criteria were no disease progression and no pneumonitis after radio-chemotherapy, and an age of 18 years or older. Patients who didn’t sign the patient informed consent of the Pius-Hospital were not included. Clinical characteristics captured included age, gender, histology, smoking status, ECOG and type of first line treatment. Smoking status was defined as follows: ex -heavy smoker: more than 10 packyears and having quit smoking for more than 10 years, current heavy smokes: current smokers with more than 10 PY, light smokers: less than 10 PY and having quit more than 10 years, never smokers: less than 100 cigarettes in a lifetime.
This retrospective study included 134 patients who were treated in the Pius Hospital from 2011 to 2018. The patients were selected from the lung cancer database of the Pius Hospital. In the end, 99 patients were included in the analysis, 35 patients could not be evaluated retrospectively for PD-L1 expression as not enough tumor material was left. PD-L1 expression on tumor cells was carried out with the anti-PD-L1 antibody SP263 on the Ventana platform at the Hematopathology Hamburg, to mimic the methods of the PACIFIC study. CRP and albumin values were used if they were analyzed no more than 14 days before the start of radio-chemotherapy.
Primary end points were progression-free survival and overall survival. PFS was defined as the first day after last radio-chemotherapy to the date of the first documented event of tumor progression or death in the absence of disease progression. Overall survival was defined as the time from the first day after last radio-chemotherapy until death from any cause.
Relevant clinical characteristics included age, gender, smoking-status, tumor histology, ECOG performance status, as well as the blood values Albumin and CRP.
The study protocol and amendments were approved by the ethics committee of Oldenburg University (2019-080), and the study was performed in accordance with the Declaration of Helsinki. Moreover, the names of the patients and all other confidential information are subject to medical confidentiality and the provisions of the Federal Data Protection Act (GDPR). Sample data were only passed on in anonymized form. Third parties did not have access to the original documents.
Statistical analysis was carried out using the Kaplan-Meier method and the log-rank test. A p-value of 0.05 was set as the significance threshold for each statistical test. Furthermore, based on survival curves of the pacific study we did a power analysis. Therefore, we used survival rates of both group with a follow up time of 45 month, considering that differences in the PACIFIC-study showed first after 15 months. The power of the study with a follow-up period of at least 21 month was 0.83.
Table 1 shows the patient characteristics. A total of 99 patients with NSCLC stage III were included in the retrospective cohort study. The median age was 61 years with a range between 41 and 82 years. Gender distribution was 64.6% (n = 64) men and 35.4% (n = 35) women. With respect to smoking status, patients were mainly former heavy smokers (48.5%, n = 48) and active smokers (38.4%, n = 38). Most of the patients had an ECOG of 0 or 1 (78.8%; n = 78). The distribution of non-squamous and squamous carcinomas was similar to the PACIFIC trial with almost 1:1 squamous vs. non-squamous. All chemotherapy was platinum-based, most frequently a combination of cisplatin with vinorelbine was used (58.7%, n = 58), followed by cisplatin-monotherapy (20.2%, n = 20). The distribution of PD-L1 values was 34.3% <1 (n=34) and 65.7% ≥1 (n=65).
Table 1 Baseline Characteristics, Stratification Factors
Variable | Total (n) | % |
Patients | 99 | 100 |
Observation period* Alive Dead |
32 67 |
32.3 67.7 |
Age (years), median (range) | 61 (52.8 – 69.2) | |
Sex Male Female |
64 35 |
64.6 35.4 |
Pack Years | 40 | |
Smoking Status Light Smoker Ex-Heavy Smoker Current Smoker Never Smoker Unknown |
2 48 38 4 7 |
2.0 48.5 38.4 4.0 87.1 |
Tumor histology type Nonsquamous Squamous |
51 48 |
51.5 48.5 |
Chemotherapy Cisplatin Cisplatin + Pemetrexed Cisplatin + Vinorelbine Cisplatin + Etoposid Cisplatin + Paclitaxel Carboplatin |
20 7 58 2 3 9 |
20.2 7.1 58.7 2.0 3.0 9.1 |
ECOG status 0 1 2 Missing |
33 45 4 17 |
33.3 45.5 4.0 17.2 |
Mean Albumin | 3.3 | |
Mean CRP | 3.0 | |
Mean Gy | 63.2 | |
PD-L1 status <1% ≥1% |
34 65 |
34.3 65.7 |
*Observation period: 01.01.20211 – 15.03.2020; ECOG, Eastern Cooperative Oncology Groups; Gy, Grey; PD-L1, Programmed death-ligand 1.
Figure 1 shows the PFS of the retrospective cohort according to the PD-L1 status. 99 cases were included in this evaluation.
The median PFS of patients with a PD-L1 expression of <1% (34.3%, n = 34) was 19 months (CI 6.4-11.6) vs. 12 months (CI 9.8-14.2) for patients with a PD-L1 expression of ≥1% (65.7%, n = 65). The log rank test showed no statistically significant difference with a p-value of 0.112. The tail of the PFS curve was higher for patients with PD-L1 expression > 1% vs. <1%.
Figure 2 shows the survival curve of 99 NSCLC patients according to the PD-L1 status. The median OS of patients with a PD-L1 expression of <1% (n = 34) was 20 months (CI 10.5-29.5) and 27.9 months (CI 16.5-39.3) for those with PD-L1 expression of ≥1% (n = 65). The log rank test showed no statistically significant difference with a p-value of 0.734.
A CRP value was determined pretherapeutic in 79 patients. The distribution was as follows: 22 patients had values below 5.61 mg/L, in 9 patients the value was between 5.61 – 8.58 mg/L and 48 patients had a value of >8.58mg/L. The median survival time for CRP values of <5.61 mg/L was 30 months (CI 23.9 – 36.5) and for values above 5.61 mg/L it was 14 months (CI 7.7– 21.1). Although there were numerical differences, these were not statistically significant in the log rank test with a p-value of 0.098.
Of the 99 patients, the albumin value was determined in 64 before the first radio-chemotherapy. The distribution was 65.6% (n = 37) < 35g/L and 34.4% (n = 26) ≥35g/L. The median overall survival in the group with an albumin value of <35g/L was 28 months (CI 16.8 - 39.3). In the group with a value of ≥35g/L, the median overall survival was 23.9 months (CI 10.3 - 37.4). The log rank test showed no statistically significant difference with a p-value of 0.868. Figure 4 (Appendix) shows the OS as a function of the albumin values. Most of the patients had a quotient of > 0.22 (n = 40), followed by a quotient <0.14 (n = 22). The longest median survival was 27.8 months (CI 14.7 – 40.9) for those who had the quotient <0.22. In the group > 0.22, the median survival time was 23.9 months (CI 6.3– 41.6). The log rank test showed no statistically significant difference with a p-value of 0.759 (Fig 5 Appendix).
In the current retrospective analysis, we tried to simulate the control group of the PACIFIC study and test for differences in the OS relative to the PD-L1 expression. No differences in the OS between PD-L1 < 1% and ≥ 1% were found.
The population enrolled in the PACIFIC trial and the retrospective population analyzed in the current study were very similar. In the PACIFIC study, 70.1% (n = 500) of the patients were male and 29.9% (n = 213) were female (9). In our cohort, the ratio was similar with 64.6% (n = 64) male and 35.4% (n = 35) female patients.
Also, no differences were seen in the median age, which was 64 years in the PACIFIC study and 61 years in this retrospective analysis. Furthermore, the histology’s were similar with 45.7% (n = 326) of the patients having squamous cell carcinoma and 54.3% (n = 387) non-squamous carcinoma, mainly adenocarcinoma in the PACIFIC and 48.5% (n = 48) squamous and 51.5% (n = 51) non-squamous cell carcinoma in our retrospective cohort.
In the PACIFIC trial, most patients in the control arm had an ECOG 1 (51.45%) as in our retrospective study (45.5%, n = 45).
The distribution of PD-L1 values in the PACIFIC study was 67.2% (n = 303) ≥ 1% and 32.8% (n = 148) < 1%. In our retrospective cohort, distribution of PD-L1 expression was similar with 65.7% (n = 65) ≥ 1% and 34.3% (n = 34) < 1%. These results are also in concordance with the distribution of PD-L1 expression values <1% from 30–35% and PD-L1 values ≥1% from 65–70% found in the literature as well as in a German prospective cohort, CRISP (15–18). This results in a high level of consistency in PD-L1 expression in the patient groups examined, which also speaks for the stability of the detection method for PD-L1 expression (11–13). The methodology to evaluate PD-L1 expression was identical to the one used in the PACIFIC study.
The number of patients in whom the PD-L1 status was available, was higher in the PACIFIC study (451) as compared to our analysis (99). In the PACIFIC study, smoking status was divided into current smoker (16.4%, n = 117), former smoker (74.6%, n = 74.6) and never smoker (9%, n = 64). In our retrospective cohort, significantly more patients were current smokers (38.4%, n = 38). 48.5% (n = 48) of them had previously smoked and only 4% (n = 4) never smoked.
The PACIFIC study included patients regardless of PD-L1 expression who did not progress after radio-chemotherapy with a platinum combination therapy and a dose of at least 54 Gy and who did not have radiation pneumonitis. A significant number of patients in the PACIFIC study received carboplatin (301 pts) while cisplatin combinations were administered in 395 pts. In our cohort except for 9 patients receiving carboplatin, all patients were treated with cisplatin, however 20% only received cisplatin monotherapy. Inclusion of the patients in the PACIFIC trial was not stratified for PD-L1 expression, PD-L1 expression was studied in a subpopulation of patients that had given optional archival tissue specimens for analysis. The PACIFIC study had two co-primary endpoints, PFS and OS. Both were highly positive for the total patient population. The study is important for the patients with NSCLC, because it could be shown for the first time that in the curative situation an immune checkpoint inhibitor improved survival with a statistically highly significant increase. The last 5-year analysis showed an OS improvement of 40% in the PD-L1 ≥ 1% cohort. Post-hoc, the approval authorities required the analysis of patients with PD-L1 expression < 1% and ≥ 1%. For this purpose, the patients were subsequently examined for PD-L1 expression, which was successful in 63% of the patients. In this so-called biomarker group, as in the ITT population, a statistically significant PFS and OS advantage with durvalumab was shown. The subgroup analysis of the patients with a PD-L1 expression of ≥ 1% showed an advantage for durvalumab that was highly statistically significant. Analysis of the population with expression < 1% showed that no OS benefit was achieved with durvalumab. Strikingly the OS was also very high in the control group (i.e., the group without durvalumab), similar to that in the PD-L1 ≥ 1% group that had received durvalumab and significantly higher than the OS in the group with a PD-L1 expression ≥ 1% treated with placebo (control group). Hence the consideration that after radio-chemotherapy the initial PD-L1 expression of < 1% has a favorable prognostic value.
In our retrospective analysis, all patients were tested for PD-L1. The survival time analysis showed that the PD-L1 status has no significant effect on the OS in the largest ever studied cohort of patients treated with definitive radio-chemotherapy in stage III. There was also no significant difference when assessing the influence of PD-L1 expression on PFS. Thus, the hypothesis that the lack of PD-L1 expression is associated with a favorable prognosis after radio-chemotherapy could not be confirmed in the present patient collective.
These results are consistent with two other studies that have retrospectively studied the impact of PD-L1 expression on PFS and OS after radiochemotherapy in smaller cohorts. In the studies by Tokito and Tufman, no differences were seen in the PFS and OS regarding PD-L1 expression in 74 pts (Tokito) and 78 pts (Tufman). In another study by Gennen including only 31 patients, a statistically significant difference was found, however the number of patients was very small suggesting an accidental finding.
In the PACIFIC study, the PD-L1 analysis was a post-hoc analysis, i.e., the patients were not stratified based on this analysis, so that there were maybe imbalances in the two groups. However, based on this post hoc analysis, the EMA has limited the approval for durvalumab to patients with a PD-L1 expression of > 1%, which has been scientifically criticized because of the post hoc analysis. Our data together with the published data show that most likely PD-L1 expression is not prognostic, therefore suggesting that the lack of difference in the PD-L1 < 1% group in the PACIFIC trial might be an accidental finding.
All patients who were treated in the Pius Hospital between 2011 and 2018 and who met the inclusion criteria were included in this retrospective study. Staging was uniformly performed according to the recommendations of the S3 guidelines and the Oncopedia guidelines, in place in Germany i.e., all patients had to undergo a PET-CT and an MRI to exclude distant metastases. Patient selection for definitive radio-chemotherapy was performed throughout the years by the identical interdisciplinary team of pneumologists, radiation-oncologists, oncologists, thoracic surgeons, nuclear medicine specialists and pathologists suggesting a high consistency of treatment and uniformness of the patient cohort. The data collection was almost complete, standardized and of high quality, so that very reliable statements about the patient characteristics could be obtained. The determination of the PD-L1 status was performed in a ring certified pathology laboratory with the identical technique used in the PACIFIC study. The agreement with the distribution of PD-L1 values in the literature also means that the test is accurate and meaningful. In addition, the patient collective is very similar to that of the PACIFIC study.
A weakness of the data analysis is the small study population compared to the prospective randomized trial. In addition, not all clinical characteristics could be evaluated in a few patients because not all data had been collected. This was particularly difficult when evaluating the smoking status. In nine out of 99 cases, no smoking status is documented. Same was true for blood values albumin and CRP that were not available in all patients within a period of 14 days before start of chemoradiotherapy.
387 patients of all stages were included in the “post-diagnostic C-reactive protein and albumin predict survival in Chinese patients with non-small-cell lung cancer: a prospective cohort study” (14). Gender distribution with 62.3% male and 37.7% female patients was similar to our cohort (65.3% male, 34.7% female). In their study (14) it was found that an increased CRP (> 8.58 mg/L), a decreased albumin (< 35g/L) and a CRP/Albumin quotient of > 0.22 had a negative prognostic impact for OS and PFS. However no statistically significant impact on OS was seen in the non-metastatic setting. However, in other studies, a prognostic impact of CRP and albumin was shown also in the non-metastatic setting (19–22).
In our cohort, we could show that patients with a CRP value > 8.58 mg/L had the lowest probability of survival, while CRP values <5.61 mg/L are associated with a longer OS. This difference could be shown numerically, but was not significant, presumably due to the small number of cases. Thus, our patient collective supports the data of the Chinese study to the extent that a numerically lower OS can be found in patients with NSCLC stage III with high CRP.
The aim of the study was to test the assumption that in patients with NSCLC in stage III after definitive radio-chemotherapy PD-L1 expression had a prognostic impact on OS. For this purpose, a retrospective analysis of the largest group published yet, uniformly treated in stage III with radio-chemotherapy and in whom PD-L1 expression was known, with respect to OS and PFS was performed. Furthermore, albumin, CRP and the ratio of the two biomarkers were also analyzed.
We found no impact of PD-L1 expression on PFS or OS in this retrospective cohort. Furthermore, neither CRP, nor albumin nor the ratio of the two blood markers had an impact on PFS and OS. A larger analysis using other patient data sources could further validate our findings
Not applicable.
Jan Wagner: Writing-Original draft preparation, Data curation, Investigation. Julia Roeper: Methodology, Writing-Reviewing and Editing. Lukas Heukamp: Molecular genetic analysis. Markus Falk: Molecular genetic analysis. Kay Willborn: Resources, Writing – Review & Editing. Frank Griesinger: Resources, Hypothesis creating, Writing-Reviewing and Editing, Funding acquisition, Supervision.
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
Not applicable.
The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). The study was approved by ethic committee of the University of Oldenburg (2019-080). Written informed consent was obtained from all participants.
Not applicable.
Jan Wagner: There are no competing interests and no conflict of interest in relation to this study.
Julia Roeper: Honoraria: Roche, AstraZeneca, Boehringer Ingelheim
Lukas Heukamp:
Honoraria: Astra Zeneca, Boehringer Ingelheim, Novartis, Pfizer, Roche, MDS, BMS, Siemens
Ad Boards: Astra Zeneca, Boehringer Ingelheim, Novartis, Pfizer, Roche, MDS, BMS, Siemens
Markus Falk:
Ad Boards: Boehringer, Pfizer
Honoraria: Roche, AstraZeneca, Boerhinger Ingelheim
Kay Willborn: There are no competing interests and no conflict of interest in relation to this study.
Frank Griesinger:
Scientific support:
ASTRA, Boehringer, BMS, Celgene, Lilly, MSD, Novartis, Pfizer, Roche, Takeda, Siemens, Sanofi
Speaker’s appointments:
ASTRA, Boehringer, BMS, Celgene, Lilly, MSD, Novartis, Pfizer, Roche, Takeda, Ariad, Abbvie, Siemens, Tesaro/GSK, Amgen, Sanofi
Ad Boards:
ASTRA, Boehringer, BMS, Celgene, Lilly, MSD, Novartis, Pfizer, Roche, Takeda, Ariad, Abbvie, Tesaro/GSK, Siemens, Tesaro, Amgen, Sanofi
no competing interests and no conflict of interest in relation to this study.