Real-World Safety and Outcome of First-Line Pembrolizumab Monotherapy for Metastatic NSCLC with PDL-1 Expression ≥ 50%: A National Italian Multicentric Cohort (“PEMBROREAL” Study)

Simple Summary Pembrolizumab monotherapy remains the first-line standard of care for patients with metastatic NSCLC and a PD-L1 tumor proportion score ≥ 50%. However, although other real-world studies have been published, long-term follow-up data on progression-free survival, overall survival, treatment response, and safety in a large cohort of patients are still lacking. We evaluated these outcomes in a cohort of 880 patients treated with first-line pembrolizumab monotherapy in 16 Italian centers. We also analyzed the prognostic impact of variables such as age, sex, histology, PD-L1 expression, ECOG, habitual smoking, and the presence of brain metastases. Median PFS and OS were 8.6 and 25.5 months, respectively, consistent with the results of Keynote-024. According to univariate analysis, it was determined that PD-L1 expression, ECOG, and habitual smoking had an impact on PFS, while age, sex, ECOG, histology, and habitual smoking had an impact on OS. However, results from univariate analysis should be considered with caution. Abstract Results from the phase III Keynote-024 clinical trial established pembrolizumab monotherapy as the first-line standard of care for patients with metastatic NSCLC who have PD-L1 expression ≥ 50%, EGFR, and ALK wild-type tumors. However, given the differences between patients treated in routine clinical practice and those treated in a clinical trial, real-world data are needed to confirm the treatment benefit in standard practice. Given the lack of data on large cohorts of patients with long follow-ups, we designed an observational retrospective study of patients with metastatic NSCLC who were treated with pembrolizumab, starting from its reimbursement eligibility until December 2020. The primary endpoints were PFS and OS, determined using the Kaplan–Meier method. Response and safety were also evaluated. We followed 880 patients (median follow-up: 35.1 months) until February 2022. Median PFS and OS were 8.6 months (95% CI: 7.6–10.0) and 25.5 months (95% CI: 21.8–31.6), respectively. We also found that ECOG PS, PD-L1 expression, and habitual smoking were prognostic factors for PFS, while age, sex, ECOG PS, habitual smoking and histology had an impact on OS. Multivariable analysis confirms the prognostic role of PD-L1 for PFS and of ECOG for both PFS and OS. 39.9% of patients reported an adverse event, but only 6.3% of patients discontinued therapy due to toxicity. Our results suggest a long-term benefit of pembrolizumab in the first-line setting, as well as a safety profile consistent with the results of Keynote-024. Many collected variables appear to influence clinical outcome, but results from these exploratory unadjusted analyses should be interpreted with caution.


Introduction
In 2023, about 44,000 new cases of lung cancer are expected in Italy: it is the second most common neoplasm in men (15%) and the third most common in women (6%).In 2022, about 35,700 deaths from lung cancer were recorded in Italy: 23,600 in men and 12,100 in women.Lung cancer is the first cause of cancer death in men and the second in women.The 5-year survival rate of lung cancer patients in Italy is 16% for men and 23% for women, which is negatively affected by the high proportion of patients diagnosed with advanced-stage disease [1].
Nevertheless, there have been significant advancements in therapeutic options for patients with metastatic non-small cell lung cancer (NSCLC) in recent years, which provides hope for improving survival rates.Developments in molecular testing and the availability of new therapies targeting specific molecular alterations have significantly impacted patient management.Clinicians are now able to select therapies that are most likely to improve individual clinical outcomes, taking into account patients' characteristics and tumor histology.
For patients with metastatic NSCLC without driver mutation of EGFR, ALK, and ROS1, drugs that act on the PD-1/PDL-1 axis are predominantly used in the first-line setting with or without platinum chemotherapy.This happened since pembrolizumab monotherapy proved its superiority in terms of progression-free survival (PFS) (10.3 mo. vs. 6.0 mo., HR 0.50; 95% CI, 0.37 to 0.68; p < 0.001) and overall survival (OS) (30.0 mo. vs. 14.2 mo.; HR, 0.63; 95% CI, 0.47 to 0.86; p-value = 0.002) against platinum-based chemotherapy in Keynote-042 and become available and reimbursed by the national health system (NHS) in June 2017 for patients with NSCLC with a tumor proportional score (TPS) of PDL-1 ≥ 50% [9,10].Following the results of Keynote-189 and Keynote-407, pembrolizumab has become available, in combination with chemotherapy selected based on tumor histology, for non-squamous and squamous cancers since December 2019 and December 2020, respectively [11,12].However, the Italian Regulatory Agency (AIFA) limited reimbursement for chemo-immunotherapy to patients with PD-L1 expression < 50%, even though its superiority over chemotherapy alone was demonstrated regardless of PD-L1 expression.Considering this limitation, singleagent immunotherapy is still the best available option for previously untreated patients with metastatic NSCLC with a tumor proportional score (TPS) of PDL-1 ≥ 50% and without EGFR, ALK, and ROS1 mutation [13].
The survival benefit shown by pembrolizumab is also confirmed by long-term data, which are not yet available for the other two alternatives, Atezolizumab and Cemiplimab, which have been reimbursed by the NHS for the same therapeutic indication since June and August 2022, respectively.Nevertheless, the role of real-world observational studies is very important to confirm that the benefits seen in clinical trials are also seen in the population treated in everyday clinical practice.
Velcheti et al. recently published the 3-year follow-up real-world results of their observational study in US oncology practice.The study has confirmed the results of the Keynote-024 clinical trials in terms of real-world time on treatment (rwToT), which is a surrogate indicator associated with survival in NSCLC studies [14].
However, given the differences in the population that could be treated in Italy for regulatory reasons, the availability of different therapeutic options for subsequent lines that could change patient survival, and the importance of outcome indicators such as progression and death of patients treated in real-life, we designed a multicenter observational study to evaluate the long-term effectiveness and safety outcome of pembrolizumab monotherapy (the PEMBROREAL study).
This work aims to show the 3-year effectiveness and safety results from a cohort of patients treated in sixteen institutions in Italy following the standard clinical practice.

Study Design
PEMBROREAL is a retrospective, multicenter study conducted in Italy.The study involved a cohort of patients who received at least one dose of pembrolizumab monotherapy in the first-line treatment of metastatic NSCLC without EGFR and ALK driver mutation.The medical records of adult patients with NSCLC who were consecutively treated at one of the sixteen study centers were reviewed by study investigators from each participating center, according to the eligibility criteria established by AIFA.For each enrolled patient, chart extractions were performed on two predetermined dates: one in January 2021 and the other in February 2022.The first data extraction aimed to guarantee that participating centers met high standards in collecting the data.In contrast to the eligibility criteria for Keynote-024, it was decided that patients with NSCLC who have not received prior treatment for metastatic disease will also have access to pembrolizumab monotherapy if they have ECOG PS 2, active central nervous system (CNS) metastases, and a life expectancy of less than three months.To be considered for inclusion, patients who began treatment with pembrolizumab between June 2017 and December 2020 were required to have provided informed consent for their medical records to be accessed.Patients who could not be reached despite all reasonable efforts or who had passed away were included, in accordance with national regulations.Patients who received pembrolizumab through an interventional clinical trial, compassionate-use program, or off-label were not eligible for inclusion.The study protocol (protocol approval ID: 6346/2020 I.5/187) was approved by the Ethics Committees of all participating institutions.The study was conducted in accordance with the principles of the 1964 Declaration of Helsinki and its subsequent amendments.

Assessment
The study aimed to evaluate the effectiveness, safety, and activity of pembrolizumab monotherapy in previously untreated metastatic NSCLC patients, based on the reimbursement criteria set by AIFA.The aim was to assess the consistency of the results obtained in Keynote-042 in real-world clinical practice.The study evaluated two main outcomes: real-world PFS and OS.PFS was calculated starting from the date of first administration of pembrolizumab (index date) until the date of progression as determined by the treating physician, death (if no progression), or the end of follow-up, whichever occurred first.OS was calculated from the index date until death or the end of follow-up, whichever occurred first.Additionally, we assessed the overall response rate (ORR), calculated as the proportion of patients who achieved a complete or partial response out of the total number of patients.Similarly, the disease control rate (DCR) was determined by the proportion of patients who achieved a complete or partial response or disease stability out of the total number of patients.Duration of response (DoR) is measured from the date of the first documented response until disease progression or death (if no progression).During the follow-up period, patients who did not exhibit any signs of progression were censored.The identification and management of adverse events (AEs) were carried out by the treating physician.The grade of AE was determined by the reporter using the Common Terminology Criteria for Adverse Events, version 4. Due to the real-world nature of PEMBROREAL, physicians may determine progression and response to therapy through assessments or by following the Response Evaluation Criteria in Solid Tumours (RECIST) version 1.1, depending on local practice.
In addition, key secondary endpoints include rwPFS and OS for subgroups of interest, as well as demographic, patient, and disease characteristics.

Statistical Analyses
No formal sample size calculation was performed.Categorical variables were presented as absolute and relative frequency, whereas for continuous variables, data were presented as median with minimum and maximum values.Patients lost to follow-up (i.e., alive at last visit or contact before database cut-off) were censored from rwPFS and rwOS data (i.e., alive at last visit or contact before database cut-off).Median and landmark rates were calculated using the Kaplan-Meier method, and the corresponding 95% confidence interval was calculated using Greenwood's method.The potential prognostic role of the investigator-identified factors was assessed using the log-rank test.A multivariable Cox regression model was used for the assessment of independent prognostic factors, and the hazard ratios (HR) were calculated.Proportional hazard assumptions were tested using Schoenfeld residuals.In case of violation, a time-dependent effect was also calculated and added to the main effect.
Analyses were based on the total population involved in the study.A complete case analysis was conducted without any imputation, due to the retrospective and pragmatic nature of the study, as well as the unlikelihood of finding a reliable pattern derived from other variables in a multiple imputation context; subgroup analyses (i.e., stratified by disease, therapy, or other demographic or prognostic variables) were performed as indicated by the investigators at the participating centers.A p-value < 0.05 was considered statisti-cally significant.Statistical analyses were performed using STATA/MP 15.0 for Windows (StataCorpLP, College Station, TX, USA).

Patients and Treatment Characteristics
Between 25 June 2017 and 31 December 2020, a total of 880 eligible patients received their first pembrolizumab dose and were included in the analysis.The median follow-up duration in the full analysis set was 35.1 months (ranging from 0.1 to 56.0).Only 39 patients (4.4%) were lost to follow-up, and none of the enrolled patients were excluded from the final analysis.The median age of patients in the full analysis set was 69.9 years (ranging from 37.9 to 90.2) at their first administration of pembrolizumab; 229 (26.0%) were aged above 75 years.The majority of patients (70.2%) were male and had a PS of 0 or 1 (91.9%) at their first administration of pembrolizumab.The most common histological type was lung adenocarcinoma, with 673 (76.5%) patients enrolled, while 158 (18.0%) patients had squamous cell carcinoma.In addition to the two most prevalent histological types, our study included a small number of patients with NSCLC not otherwise specified (NOS), adenosquamous carcinoma, and large-cell lung cancer.For one patient, the tumor histology was unknown.
The TPSs of PD-L1 were immunohistochemically evaluated by a validated 22C3 IHC laboratory test for all patients.All patients enrolled had PD-L1% expression ≥ 50%.For most of them (527 of the 880 patients (59.9%)), the exact expression of PD-L1 was available in the clinical documentation.Of these 527 patients, 120 (22.8%) had high expression of PD-1 (≥90%).A total of 134 patients were diagnosed with brain metastasis (15.7%).For 29 patients included in the study, the presence or absence of brain metastasis was not documented.Out of the 880 patients included in the study, 542 (61.6%) had known smoker status, with the majority being either current (n = 186; 34.3%) or former smokers (n = 251; 46.3%).EGFR and ALK mutations were tested for 788 and 790 of 880 included patients, respectively, using next-generation sequencing (NGS) for the first mentioned mutation and immunohistochemical validated Ventana ALK (D5F3) and rabbit monoclonal antibody for the second one.The result was negative for 100.0% of the tested patients.Of note, all patients with unknown EFGR and ALK status, 92 and 90, respectively, had a squamous NSCLC (Table 1).Giving the rarity of these two mutations and the low efficacy of target treatments in this histological subtype cell tumor, searching for these driver-gene alterations is not mandatory considering current guidelines and the eligibility criteria for NHS reimbursement.The study did not collect data on other driver mutations, such as KRAS, BRAF, RET, and ROS1, which may have affected OS and PFS.It is worth noting that participating centers evaluated these mutations at their discretion.
Additionally, the median total treatment time, including dose interruptions, was 189 days.Notably, 34.7% of patients received pembrolizumab for more than 12 months.Patients received a median of 9.0 infusions of pembrolizumab (range: 1-75).It is worth noting that 8.1% of patients had a temporary treatment interruption of more than three weeks for any reason.

Reasons for Discontinuing Pembrolizumab
The most common reasons for discontinuing pembrolizumab treatment were disease progression (53.9%) and death (13.6%).A small percentage of patients (6.3%) stopped treatment due to unacceptable toxicity.At the time of data cut-off, 16.0% of patients were still receiving pembrolizumab treatment (refer to Supplementary Table S1).

Reasons for Discontinuing Pembrolizumab
The most common reasons for discontinuing pembrolizumab treatment were disease progression (53.9%) and death (13.6%).A small percentage of patients (6.3%) stopped treatment due to unacceptable toxicity.At the time of data cut-off, 16.0% of patients were still receiving pembrolizumab treatment (refer to Supplementary Table S1).
The multivariate analysis is shown in Supplementary Table S5: among all the prognostic factors that affected RwOS, the only variable that confirmed this effect was ECOG PS, and the effect was time-dependent.

Analysis of Treatment Response
In the study, it was found that out of 880 patients, 611 (69.4%) reported a response to treatment.Among these patients, 179 (20.3%) had a partial (17.6%) or complete response rate (2.7%) as the best response to treatment (refer to Supplementary Table S2).At the first evaluation, 25.9% of patients had experienced disease progression.The disease control rate (DCR) for the included patients was 79.2%.The median duration of response (DoR) for patients with a documented complete or partial response was 27.1 months (95% CI: 22.0-33.8).
A total of 67 patients reported grade 3-4 AEs, of which 14 were gastrointestinal disease, 13 were skin and subcutaneous tissue disorders, 13 were general disorders and administration site conditions, and 13 were related to alteration of diagnostic exams (Table 4).
It is worth noting that most of the reported adverse events did not require any management measures such as drug administration suspension, reduction, definitive interruption, hospitalization, or specific pharmacologic treatment.Among the patients who experienced an adverse event (AE) that required specific measures, it was observed that 20.1% (177 cases) were managed with pharmacologic treatment, while 8.1% (71 cases) required temporary treatment interruption.It was found that only 5.3% (47 patients) definitively discontinued treatment due to toxicity (see Supplementary Table S3).

Discussion
In this large multicenter, real-life observational study, we considered 880 patients with metastatic NSCLC and a PD-L1 expression ≥ 50%, EGFR, and ALK wild-type tumors who were treated with first-line pembrolizumab monotherapy.With a median follow-up time of 35.1 months, we reported a median PFS of 8.6 months and a median OS of 25.5 months.These results are consistent with findings from Keynote-024, in which PFS was 7.7 months (CI 95%: 6.1-10.2) and OS was 26.3 months (CI 95%: 18.3-40.4).However, we observed that patients were somewhat older in our cohort of patients than in the experimental arm of Keynote-024 (median age 69.9 vs. 65.5) and included proportionately more women (40.3% vs. 29.8%).
In addition, we included 8.1% of patients with ECOG PS 2 with less than three months of life expectancy and with active brain metastases who were treated in routine clinical practice but would be excluded from enrollment in a clinical trial.Moreover, one of the limitations of Keynote-024 was that only a few patients who never smoked were enrolled (3.2%), whereas in the real-world setting, this rate was higher among patients whose smoking habits were known (19.4%).
To the best of our knowledge, many other real-world studies that have been published had shorter follow-ups, focused on surrogate endpoints, or used results that referred to few patients.We have compiled a table of all previously published real-world studies of the same treatment for similar populations.(Table 5) [14][15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30].Compared with the results of other previously published real-world studies in the same setting, OS in the PEMBROREAL study was longer, based on a longer follow-up, and derived from the observation of more patients than those in the majority of other reported studies.PFS was similar to that previously observed in the reported studies, except for that of Jiménez Galán [16], in which it should be noted that almost one third of the patients enrolled had an ECOG PS ≥ 2. A possible explanation of the longer OS in our recent study could be the availability of newer second-line therapies, especially for patients with mutations of other driver genes (e.g., KRAS, BRAF, RET, MET), that could be responsible for a longer survival for these patients.On the other hand, the consistency of the PFS data could be explained by the fact that the availability of this outcome takes less time if compared to OS, and therefore, data from other studies are sufficiently robust, despite the limited follow-up of some of them.Univariate Cox regression confirmed that PD-L1 expression, ECOG PS, and smoker status significantly influence PFS.Other variables studied, such as age ≥ 75 years at diagnosis, sex, histological type, and presence of brain metastases, were not statistically significant in determining PFS.However, a slightly favorable trend for patients without brain metastases was observed for this latter variable.
According to previously published papers [15,16,31,32], clinical outcomes are significantly improved in NSCLCs with a PD-L1 expression of ≥90% and ECOG PS < 2. The prognostic role of ECOG PS was explored in many other studies, and despite the limitations due to the subjective nature of this parameter-considering that two physicians may classify the same patient with a different performance status-it is an important variable influencing patient outcome.In addition, we know from the work of Facchinetti et al. that the worst outcomes were found for patients for whom a high ECOG PS value was determined by disease burden rather than for patients for whom it was determined by comorbidities [33].However, it was not possible to differentiate between these two types of patients in our study.
In NSCLC patients receiving immunotherapy, smoking status has previously been shown to be associated with clinical outcomes [29].A recent study of patients with NSCLC treated with different drugs active on the PD-1/PD-L1 axis across multiple lines showed a non-statistically significant trend towards an improved PFS in heavy and light smokers compared to those who never smoked [34].The authors found that patients who smoked heavily had a higher tumor mutational burden (TMB) in their tumors, which makes cancer cells easier for the immune system to recognize; this was an explanation for the difference in the clinical outcome.Notably, the difference between patients who never smoked and former or current smokers was not confirmed by the multivariable model, so this finding on the prognostic role of smoking habit should be considered with caution.
Interesting results were obtained from the univariate Cox regression analysis of the variables that influence OS.Not surprisingly, a significantly better OS was found for patients with age at diagnosis < 75 years and for patients with lower value of ECOG PS.We considered more interesting the role of sex in determining OS, since, in our results, female patients had a better prognosis than male patients.Another interesting variable of prognostic significance was tumor histology, since patients with a non-squamous tumor type had a better outcome than patients with a squamous tumor type.Finally, the prognostic value determined for smoker status was inverted if we considered OS compared to PFS.If the outcome for PFS was better for patients who were former or current smokers, the outcome for OS was better for patients who had never smoked.It is important to note that these findings are exploratory and unadjusted; therefore, they should be interpreted with caution.
Research has consistently shown that sex is a prognostic factor in lung cancer, even before the availability of immune checkpoint inhibitors.A recent study conducted on a large Australian cohort confirmed that women have a significantly longer survival rate after a lung cancer diagnosis compared to men [35].The study also found that men were typically older at diagnosis, less likely to be non-smokers, and had more comorbidities.Some authors have suggested that lung cancer in women may have a different natural history, which could be related to immunological differences as well as patient factors [36,37].
Regarding results about the influence of cancer histology on OS, these were consistent with findings from the real-world study conducted by Tambo et al. [26].In this study, patients with squamous NSCLC had a greater number of metastatic sites compared with patients with adenocarcinoma, and these findings were correlated with a longer OS.Unfortunately, there are no published data relative to the effect of histologic type on the outcome of patients treated with immunotherapy; however, squamous cell carcinoma is a well-known poor prognostic factor.
Considering the opposite effect of smoker status on OS, these results are consistent with findings of the meta-analyses conducted by Popat et al. [38] and could be explained by the fact that, according to the previously published report, patients who were never smokers were more likely to be women and be diagnosed with non-squamous tumor histology [39].
Treatment response seems to be inconsistent across different studies.In our cohort of patients, ORR results were lower if compared with results from Keynote-024 (ORR: 44.8% vs. 20.3%).However, these differences may be explained by differences in response assessment between an experimental clinical trial, where the response is rigorously assessed and centrally reviewed, and retrospective observational studies, where reports of tumor response are individually reported by a single clinician and collected by chart extraction, with concerns about missing reports.
Safety assessment revealed that pembrolizumab monotherapy was generally well tolerated.Comparing the results of our observational study with those of patients treated in the experimental arm of Keynote-024, the rate of patients with at least one adverse event was halved in patients treated in current clinical practice (39.9 vs. 73.4).Grade 3 and 4 AE were uncommon in current clinical practice, occurring only in 67 out of 880 (7.6%) patients.Interestingly, in the majority of cases, toxicity management do not require specific intervention, and situations that do require specific management can be easily managed with pharmacological therapy or temporary withholding of pembrolizumab.
Most concerns regarding safety evaluation in our study are related to the phenomenon of under-reporting of toxicity by treating physicians, especially in the case of low-grade AEs or for those considered as not clinically relevant.
Strengths of this study include the large multicenter cohort with the longest follow-up ever published.These considerations provide consistency to our outcome results compared with previous publications with a limited follow-up or with few patients.
However, it is important to note that retrospective data evaluation has limitations such as the potential for missing or inaccurately recorded data.For example, smoking status was unknown in more than 23% of patients, and for 40% of patients, we only know that they had a TPS of PD-L1 ≥ 50%, but we did not know the exact value of their score.However, we have no reason to think that missing data for some of the variables considered for univariate or multivariate analysis could be unbalanced between the subgroups.
Another limit is related to outcome assessments.Patients included in our study were treated within the valid standard of care outside a clinical trial with varying imaging intervals, thereby potentially biasing PFS.Additionally, progression can be determined based on radiological or clinical evidence.It may be worthwhile to conduct further analysis to explore the potential impact of this limitation on PFS.
Finally, considering that 13 of the 16 institutions selected to participate in our study had an internationally recognized role in cancer research and treatment, this could be considered a selection bias in the choice of patients, as these patients are more likely to receive better care than patients treated at less prestigious hospital sites, resulting in better outcomes.
Despite these limitations, our results are consistent with those of other retrospective studies and with the results of Keynote-024, and they help to identify patients who may best benefit from treatment with pembrolizumab monotherapy.

Conclusions
In conclusion, the results of our retrospective observational study of pembrolizumab monotherapy in first-line metastatic NSCLC with PD-L1 ≥ 50% are consistent with previously reported real-world studies and the results of Keynote-024.This is true even though the distribution of patient characteristics in routine clinical practice is slightly different from that in the clinical trial and even though patients in clinical trials are selected by strict selection criteria, including some related to performance status, presence of active brain metastases, and life expectancy.
In addition, we identified several prognostic factors that influence patient outcomes.This is of interest for patients, who should be informed about the benefit of a particular therapy for patients with their characteristics; for physicians, who can identify which

Figure 1 .
Figure 1.Progression-free survival of patients, shown as Kaplan-Meier distribution.Figure 1. Progression-free survival of patients, shown as Kaplan-Meier distribution.

Figure 1 .
Figure 1.Progression-free survival of patients, shown as Kaplan-Meier distribution.Figure 1. Progression-free survival of patients, shown as Kaplan-Meier distribution.

Figure 2 .
Figure 2. Overall survival of patients, shown as Kaplan-Meier distribution.Dashed lines represent 12-and 24-month landmark analyses.

Figure 2 .
Figure 2. Overall survival of patients, shown as Kaplan-Meier distribution.Dashed lines represent 12-and 24-month landmark analyses.

Table 1 .
Patients' demographic and disease characteristics.
Note: Percentages reported in the table were calculated using the number of patients with available data (for each variable).

Table 2 .
Univariable analysis of real-world progression-free survival.

Table 3 .
Univariable analysis of real-world overall survival.
NR = not reached; NE = not estimable from statistical software.