Comparative Efficacy and Safety of Immunotherapy Alone and in Combination With Chemotherapy for Advanced Non-small Cell Lung Cancer

There is a lack of direct cross-comparison studies in clinical trials between immunotherapy alone and combination treatment, especially in Non-Small Cell Lung Cancer (NSCLC) patients with high PD-L1 expression. To determine if anti-PD-(L)1 antibody combined with chemotherapy is more efficient than immune checkpoint inhibitor (ICI) monotherapy for advanced NSCLC patients in the real-world data. We retrospectively collected 325 patients with advanced NSCLC treated with ICI alone with or without chemotherapy from 11th July 2016 to 26th May 2020 to investigate which treatment scenario is the most efficient, and how clinical factors impact response. Patients with advanced NSCLC were treated with ICI monotherapy (178/325, 54.8%) or in combination with chemotherapy (147/325, 45.2%). The objective response rate and disease control rate were higher in the combination group than the monotherapy group. Patients (including those with distant metastasis) treated with chemo-immunotherapy were associated with a significantly longer median PFS and OS compared with the monotherapy group, irrespective of the PD-L1 expression level and previous treatment lines. No significant increase in the risk of immune-related adverse events (irAEs) was found after combination with chemotherapy (50.6 vs. 57.8%). IrAEs predicted better PFS of immunotherapy in the monotherapy group, especially for patients with late irAEs (after ≥4 cycles). Collectively, we demonstrated that ICI monotherapy plus chemotherapy might have better anti-tumor activity and an acceptable side-effect profile regardless of PD-L1 level or previous treatment lines. Both regimens were well-tolerated and cost-effective, the more efficient is usually recommended.


INTRODUCTION
The advent of immune checkpoint inhibitors (ICIs) has radically changed the therapy paradigm in advanced NSCLC over the past 5 years. A remarkable improvement in the management of metastatic NSCLC occurred in 2015, when nivolumab was approved for the treatment of patients with progressive disease during or after a platinum-doublet treatment (1). Both anti-programmed death 1 (PD-1) and anti-programmed death ligand-1 (PD-L1) antibodies have demonstrated their benefits in comparison with standard chemotherapy (2)(3)(4)(5).
Due to the encouraging results from clinical trials, the U.S. Food and Drug Administration (FDA) granted approval for ICIs as monotherapy in advanced NSCLC. What's more, pembrolizumab is recommended as the first-line treatment in oncogene-negative tumors with high (Tumor Proportion Score, TPS ≥50%; category 1) or low PD-L1 expression (1% ≤ TPS <50%; category 2B); and atezolizumab or pembrolizumab combined with carboplatin-based doublet as the front-line treatment is also approved (category 1) (6). Chemotherapy or immunotherapy alone (no previous ICI treatment) is preferred as the second-line treatment for PS 0-2. Nevertheless, deciding between therapeutics remains a challenge today.
The objective of this retrospective study is to investigate the efficacy and safety of ICI monotherapy or in combination with chemotherapy for advanced NSCLC patients in the real-world.

Treatments
Patients were treated with anti-PD-(L)1 alone (n = 178) or combined with chemotherapy (n = 147), and it was their first exposure to ICIs. Treatment was given until disease progression, severe toxicity, or death. Assessments of progression occurred every two cycles until disease progression as per RECIST v1.1 (tumor assessments of nivolumab occurred every three cycles).

Outcomes
All patients were followed up for survival until death, or loss-to follow-up (4/325, 1.2%) from 11th July 2016 to 26th May 2020. Progression-free survival (PFS) and Overall survival (OS) were measured as the time between start of treatment and documented disease progression or death owing to any cause (PFS) or to the latter (OS). Time to treatment failure (TTF) was assessed from immunotherapy to cessation of ICI treatment for any reason. Disease control rate (DCR) refers to the proportion of patients with complete response (CR), partial response (PR) or stable disease (SD) for at least 6 months. Objective response rate (ORR) was defined as the proportion of patients with CR or PR for at least 6 months. Duration of response (DOR), defined as initial CR or PR to progressive disease (PD) or death.
Adverse events (AEs) were graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events (NCI-CTCAE), version 4, and were classified according to their characteristics: treatment-related AEs (trAEs) and immunotherapy-related AEs (irAEs) (7)(8)(9)(10). Assessments were done by at least three independent medical professionals.
In addition, progression in no-target lesions was quantified based on four progression items: pre-existing lesions, new intrathoracic metastasis, new extrathoracic metastasis, or new malignant effusion (11,12). Score 1 point for each progression item and add up the total. The final score will show the level of tumor burden.

Statistical Analyses
Associations between variables and PFS or OS were analyzed using Kaplan-Meier survival curves, the log-rank test, and univariate or multivariate Cox regression models. Multivariate hazard ratios (HRs) were estimated with a stratified Cox regression model, and 95% confidence Intervals (CIs) were calculated with the Brookmeyer-Crowley method. Subgroup analyses were done with unstratified HRs estimated from a cox proportional hazards model. Analyses were carried out using IBM SPSS Statistics 21.0 software. Categorical variables were compared in the same platform by the Fisher's exact or chi-square test. A two-tailed p < 0.05 was considered significant.   0.430, 95% CI: 0.319-0.579, log-rank p <1 * 10 (−6) ] ( Figure 1A). The median OS for combination treatment has not been reached [(combination vs. immunotherapy) HR: 0.296, 95% CI: 0.171-0.511, log-rank P = 4 * 10 (−6) ] ( Figure 1B). Reasons for drug withdrawal are mainly progression disease ( Figure 1C). However, there was no significant difference in the scores of tumor burden between them (P = 0.284; Figure 1D; Table 2). The predominant sites for disease progression after ICIs were shown in Figure 1F, there was no significant difference between those two groups except soft tissue (P = 0.016) (baseline metastases were demonstrated in Figure 1E, no significant difference was found). Subgroup analyses revealed that almost all subgroups were significantly associated with improved PFS in combination group, except for EGFR mutation, previous treatment line = 1 and baseline liver, adrenal gland, or lymph node metastasis (Supplementary Figure 1A). Similar results were demonstrated in OS, whereas patients with brain, liver, distant lymph node or adrenal gland metastases, 1 or ≥2 previous treatment lines, PD-L1 TPS expression <25%, EGFR mutation and overweight were not associated with better OS in the combination group (Supplementary Figure 1B). Furthermore, multivariate analysis demonstrated that the PD-L1 ≥25% or unknown, 0, 1, or ≥2 previous treatment lines and obesity were related with better PFS or OS of patients with advanced-stage NSCLC (Supplementary Tables 2, 3; Supplementary Figure 2). Conclusively, the efficacy observed in the combination group was better and not due to higher PD-L1 expression or less previous treatment.

Distinct Baseline Metastases Have Differential Outcomes
Given that metastasis is the dominant lethal event in NSCLC patients, most immunotherapy trials set stringent requirements for the eligible participants. Our study showed that, in patients with baseline bone metastases, the combination group provided an survival benefit when compared with monotherapy The incidence of brain metastasis is apparently high in advanced NSCLC. The current standard regimen is becoming, early local therapies before or in conjunction with ICIs (13,14). In order to evaluate the efficacy of immunotherapy, patients with target brain metastases (15) [without receiving whole-brain radiotherapy (WBRT) or stereotactic radiosurgery (SBS)] before immunotherapy were analyzed. The basic characteristics are shown in Table 3 and Figure 3A. Of patients with ICI alone, 2/11 (18.2%) had brain metastasis responses (CR; Figure 3B). The confirmed central nervous system (CNS) responses were durable (at data cutoff, responses had lasted 35.27, 4.23 months, respectively), but one patient discontinued ICI due to progression of pulmonary lesions. 2/11 (18.2%) patients had SD and 6/11 (54.5%) patients had PD in the CNS. 1/11 (9.1%) patient was unconfirmed in the CNS due to sudden death caused by rapid systemic progression. In contrast, all patients received combination treatment had brain lesion responses. The best response was CR in 1/3 (33.3%) patient. 2/3 (67.7%) patients had PR in the brain, although one patient had PD within half a year. Another patient with PR remained on treatment at data cutoff. In conclusion, combination ICI with chemotherapy has demonstrated a survival benefit for metastatic NSCLC patients.

Adverse Events and Outcomes
Notable trAEs that were in a higher incidence rate in the combination vs. immunotherapy group (≥1.5 times) were elevated transaminase or bilirubin, myelosuppression, electrolyte disturbance, ECG abnormalities, myalgia, dysfunction of intestine, constipation, nausea, and vomiting, hyperglycemia, pyrexia, alopecia, elevated creatinine and peripheral neuropathy ( Table 4, Supplementary Table 5). Whereas, the incidence of hypothyroidism (≥1.5 times) was higher in immunotherapy arm. With these exceptions, the occurrence rates of dash, fatigue, hyperthyroidism, pneumonitis, decreased appetite, xerostomia, diarrhea, and hypertension were similar. Serious AEs (SAEs; stages 3-4) were reported in 10 (5.6%) patients from the immunotherapy group and 56 (38.1%) patients from the combination group (Supplementary Table 4). The SAEs were primarily related to myelosuppression in the Combination arm, of which the incidence rate was 32.7% (48/147). Of the SAEs that resulted in treatment delay (n = 17, 9.6%; n = 31, 21.1%, respectively), 12 (6.7%) and 20 (13.6%) patients were considered related to ICIs, respectively. Moreover, 12 (6.7%) patients in immunotherapy arm (7 interstitial lung disease cases, three   (B) Best brain metastasis response in assessable patients. The lower dashed line represents the −30% cut-off that defines an objective response. And the upper dashed line represents 20% cut-off that defines progression disease. CR, complete response; PR partial response; SD, stable disease; PD, progressive disease. a One patient had developed progression of lung tumors and withdrew from immunotherapy despite 100% shrinkage of brain metastasis. b One patient had progressive disease despite <20% enlargement due to the development of new brain metastasis.
hypothyroidism cases, one ICI-related encephalitis and one fatigue case) and 5 (3.4%) patients in combination arm (two interstitial lung disease cases, two dash cases, and one ICI-related myocarditis case) withdrew from treatment due to SAEs. Up to now, no treatment-related death was reported. Similar results were obtained in irAEs (Supplementary Table 6). All AEs were assessed by at least three independent medical professionals. Overall, no significant increase in the risk of irAEs was found after combination treatment.
In our study, outcomes of patients with and without early irAEs (16) were shown in Supplementary Figure 3. The analysis showed that the development of early irAEs was significantly associated with increased PFS in immunotherapy arm [log-rank P = 0.053; multivariate HR (95% CI), 0.621 (0.411-0.941), P = 0.024], which were consistent with previous studies (16,17). However, similar trend was not found in combination group (Supplementary Figures 3, 4).

DISCUSSION
In this real-world study, we assembled a cohort of 325 patients with advanced NSCLC treated with ICI to retrospectively investigate which treatment scenario is the most efficient: ICI monotherapy or in combination with chemotherapy, and how clinical factors impact response and survival of those patients. The combination group demonstrated promising anti-tumor activity and an acceptable side-effect profile regardless of PD-L1 level or previous treatment lines. Both regimens were welltolerated, the more efficient is usually recommended. Currently, there are many controversial issues regarding immunotherapy in the real-world practice. Firstly, the lack of direct cross-comparison studies in clinical trials between ICI monotherapy and chemo-immunotherapy. Secondly, the low detection rate and positive predictive value (19,20) of PD-L1 or tumor mutation burden (TMB) in clinical practice of immunotherapy. Thirdly, ICI alone have demonstrated minimal benefit in liver metastases [a common metastasis and a negative prognostic indicator for lung cancer (21,22)]. Some combination regimens were investigated in various randomized phase III studies (23,24), but the final conclusion is still pending. Lastly, severe irAEs require high-dose intravenous steroids and even temporary or permanent discontinuation of ICIs (25). But the occurrence of irAEs was related to better outcomes in NSCLC subjects treated with ICIs (16,17). How to balance between safety and efficacy of ICIs in the clinical practice? Our study has partially answered those issues.
Preclinical data have emerged suggesting that chemotherapy can significantly enhance the efficacy of certain forms of cancer immunotherapy (26). But the exact mechanism is still unclear. For one thing, studies have indicated that local chemotherapy combined with anti-PD-1 antibody facilitates an antitumor immune response and improves survival (p < 0.001) in glioblastoma, but addition of systemic chemotherapy to anti-PD-1 treatment resulted in systemic and intratumoral lymphodepletion, with decreased immune memory in long-term survivors (27); and the toxicity of some chemotherapeutic agents to immune cells limits the extent of immune stimulation and can lead to immunosuppression (28). For another, other researchers have demonstrated that certain conventional chemotherapies may have positive effects on tumor immunity: chemotherapyinduced immunogenic cell death activates innate immune responses and elicits a tumor-specific adaptive immune response (29,30); it can directly block immunosuppressive pathways in the tumor microenvironment (TME) (31,32).
Emerging data from clinical trials demonstrated that chemo-immunotherapy combination had better efficacy than ICI alone in certain clinical scenarios. We summarized these results in Supplementary Table 7. The survival benefits of combination treatment were significantly higher than ICI monotherapy. In the high PD-L1 cohort, the combinations of chemotherapy with ICIs were overall the better treatments regarding PFS; in the low PD-L1 arm (including negative), all combinations with chemotherapy examined (pembrolizumab, atezolizumab, nivolumab) showed superior outcomes to immunotherapy-alone regarding PFS and OS. Similar results were shown in our study, especially for patients with unknown PD-L1 status (195/325, 60%), which had longer PFS and OS in the combination group compared with the ICI monotherapy. When talking for the patients with PD-L1 ≥50%, it requires more prospective or external cohort data to further confirm whether combination treatment would be better as first-line treatment or not (Supplementary Table 2).
For most subgroups, the magnitude of treatment efficacy was greater in the combination arm than ICI-alone arm. Interestingly, we found that immunotherapy/chemotherapy seems to benefit more men than women [(median PFS, moths) monotherapy arm 6.27, 2.23, respectively, combination arm 15. 53, 9.47, respectively]. However, different studies have different conclusions (33,34). In addition, patients who smoke responded better to ICI combination therapies than nonsmokers, which is consistent with previous studies (35,36). And fundamental research had revealed that higher level of aryl hydrocarbon receptor (AhR) may play a role (37). BMI ≥25 was correlated with better outcomes of patients treated with ICI monotherapy according to previous studies (38), and our results also found that patients in the Combination arm had a similar trend (Supplementary Figure 2). Overweight could be considered a tumorigenic immune-dysfunction that could be effectively reversed by ICIs (39,40). ICI combination treatment can improve outcomes in metastatic NSCLC patients according to our subgroup analyses, but a larger cohort of patients is necessary.
Safety is another primary concern. Our results revealed that patients with irAEs had longer PFS in ICI alone group, while no evidence showed that the occurrence of toxicity in   the combination arm for patients is related to the efficacy. Previous studies have demonstrated that patients treated with immunotherapy with early irAEs had better outcomes (16,17), nevertheless, other researches showed that the response of immunotherapy had no relationship with the side effects (41). Therefore, more research is needed on the relationship between efficacy and toxicity. However, rational combination of chemotherapy and immunotherapy faces many challenges (42): the requirement for an accurate predictive biomarker of efficacy; optimization efficacy, safety and tolerability through appropriate drug ratios, dosing and scheduling; and possible combination approaches for patients who had low response rates after immunotherapy (such as EGFR mutations, liver metastasis and so on). The proportion of NSCLC patients with PD-L1 ≥50% was around 30% (43). In the real world, accurate expression level of PD-L1 is largely unknown (60%, our result). And the dynamics of PD-L1 expression may limit its use as a tissue-based predictive biomarker (44,45). TMB also has many limitations (46,47). To address the first challenge, we should make further efforts to deepen the mechanism study and careful design of biomarker exploratory studies. Recently, the number of ICIs and clinical trials of ICI combination treatment have growth rapidly. Different doses (36), sequencing possibilities (48,49), combination therapy regimens (23,50), and inclusion criteria (PD-L1 level, driver genes, PS ECOG and so on) (5,9,51) have different outcomes. Under the premise of ensuring safety, it is not an easy task to decide which regimen is best. Undoubtedly, it is unwise to simply increase the number of drugs in combination therapy. Comprehensive analysis of our study and clinical trials, the approach of ICI plus chemotherapy might have better anti-tumor activity than ICI monotherapy regardless of PD-L1 level in advanced NSCLC patients.
Our study has several limitations. Firstly, this research was a single-center retrospective study with a limited sample, inevitably, the AEs could be underreported due to the retrospective nature. The prospective or external cohort validations were needed to verify in the future. Secondly, there is 60% patients in this study without PD-L1 expression data. It is well-known that PD-L1 plays a crucial role in the progression of tumor by altering status of immune surveillance. KEYNOTE-024 study showed that pembrolizumab was associated with significantly longer PFS and OS, and is preferred as 1L treatment in advanced NSCLC with PD-L1 ≥50%. However, the ability of PD-L1 level to predict efficacy of immunotherapy in NSCLC is controversial (52,53). TMB, neutrophil-to-lymphocyte ratio, MSI and some other biomarkers have also been reported to be predictive. Therefore, PD-L1 combined with other clinical biomarkers could benefit to predict the outcomes of immunotherapy. Thirdly, patients in the immunotherapy alone group at later line were relatively more, and patients in the immunotherapy combination group at first-line were relatively more, while this is the characteristic and real record of the real-world research. To balance the discrepancy and the multiple confounding factors, we did the cox proportional analysis in overall patients taking account into the number of treatment lines as a confounding factor. The results show that chemo-immunotherapy group had better PFS and OS in the multivariate analysis. Further univariate analysis among the first-line, second-line or later line, respectively repeated that combination treatment had better survival in each group (Supplementary Table 2). More prospective and external cohort validations are required to further confirm our results.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the Ethics Committee of Shanghai Chest Hospital. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

AUTHOR CONTRIBUTIONS
XN and ZC: administrative support. XW, XN, and ZC: provision of study materials or patients. XW and XN: collection and assembly of data. XW, NA, and YS: data analysis and interpretation. All authors: conception, design, drafting manuscript, and final approval of manuscript.