2.1 Study design and participants

The trial was conducted at 35 different hospitals across China between December 2016 and December 2019. Eligible patients were aged between 18 and 75 years with histologically or cytologically confirmed stage IIIB-IV LSCC. The approaches to obtain histological specimens included computer tomography-guided percutaneous lung puncture biopsy and transbronchial lung biopsy. For patients with cytological specimens, p40 or p63 by immunohistochemistry (IHC) were used to confirm the diagnosis and was a standard way of diagnosis at all the participating institutions. LSCC was staged according to the 2009 International Association for the Study of Lung Cancer (IASLC) lung cancer staging project for tumor-node-metastasis (TNM) classification. Patients who had received no prior chemotherapy, biotherapy, immunotherapy, or who experienced recurrence and metastasis more than 6 months after the last adjuvant chemotherapy (excluding gemcitabine and paclitaxel), with an Eastern Cooperative Oncology Group performance status (ECOG PS) of 0 or 1 were eligible for enrollment. All patients had at least one measurable lesion according to version 1.1 of the Response Evaluation Criteria in Solid Tumors (RECIST 1.1) [23]. The details of the study design and patient eligibility criteria are available in https://clinicaltrials.gov (NCT02996214).

The whole blood samples of patients enrolled in this clinical trial were collected at different time points, including baseline when patients were randomized, the time when patients were evaluated for efficacy, and the time when the disease has progressed. The whole blood samples collected were centrifuged for two hours to isolate plasma for preliminary biomarker analysis.

This trial was conducted in accordance with the Declaration of Helsinki. The independent ethics committee or institutional review board of each participating study center approved the protocol and all amendments. All the participants provided written informed consent before enrollment in this trial.

2.2 Randomization and treatment

Block randomization was used wherein a computer-generated randomization sequence in blocks was prepared using the Statistical Analysis System software version 9.4 (SAS 9.4) PROC PLAN. The participants were randomized in a 1:1 ratio to receive paclitaxel liposome for injection (Lipusu^®,175 mg/m², day 1, results from a phase II study recommended a Lipusu of 175 mg/m² triweekly as a component of the combination chemotherapy [24]) plus cisplatin (75 mg/m², day1) (LP group), or gemcitabine (1000 mg/m², day 1 and 8) plus cisplatin (75 mg/m², day 1) (GP group) for 4-6 cycles at the discretion of investigators, until unacceptable toxicities, disease progression, or withdrawal of consent. Dose adjustments were allowed as per label for resolution of the adverse events. Specifically, for Lipusu, dose reduction was allowed if grade 4 hematologic adverse events or ≥ grade 3 non-hematologic toxicity occurred (first reduction to 155 mg/m² triweekly, second reduction to 135 mg/m² triweekly, and only two-dose reductions were allowed during treatment course). Randomization was stratified by bodyweight reduction (≥5% vs. <5% within 6 months), age (≥ 65 vs. < 65 years), and TNM stage (IIIB vs. IV).

2.3 Endpoints and assessments

The primary endpoint was progression-free survival (PFS), calculated from the date of randomization to the date of disease progression or death from any cause. The secondary endpoints included ORR, which was the proportion of patients who achieved complete response (CR) or partial response (PR) according to RECIST 1.1; disease control rate (DCR), which was the proportion of patients who achieved stable disease (SD), CR or PR, and; overall survival (OS), which was calculated from the date of randomization to the date of death from any cause.

Tumor assessment was performed every 2 cycles (42 ± 3 days) following the RECIST 1.1 criteria [23]. The response confirmation occurred 4 weeks or more after an initial measurement and every other cycle thereafter. After disease progression, the follow-up was conducted every 90 days by telephone until imaging examination proved progressive disease (PD) or death or lost during follow-up. AEs were followed up to within 21 ± 3 days after the last administration of the study drugs.

2.4 Safety

The safety was assessed weekly based mainly on the occurrence, frequency, and severity of AEs by using the National Cancer Institute-Common Terminology Criteria for Adverse Events (NCI CTCAE) version 4.0 and coded to a preferred term using the Medical Dictionary for Regulatory Activities (MedDRA). The safety events included AEs and severe AEs (SAEs). For all AEs, wherever necessary, patients were withdrawn from the study.

2.5 Detection of the levels of cytokines

Plasma samples from 92 LSCC patients in the LP group were obtained at baseline, the first efficacy evaluation was carried out after administration of the two cycles of LP regimen and disease progression. To isolate plasma, ethylenediaminetetraacetic acid (EDTA)-anticoagulated whole blood samples were centrifuged (1000 × g, 15 min). A total of 45 cytokines were simultaneously measured in plasma samples using the Cytokine/Chemokine/Growth Factor 45-Plex Human ProcartaPlex Panel (Thermo Fisher Scientific, Inc., Waltham, MA, USA). The concentration of the cytokines used was the expression level of the cytokines. To define potential “cytokine signatures” in the plasma, the 45 peripheral immune factors were classified into different subgroups according to their biological characterization (e.g., growth factors, chemokines, interferons, interleukins, and colony stimulatory factors). After log2 transformation of the concentration value, a signature score was calculated by averaging the various included cytokines. Univariate Cox regression analysis was used to screen for potential signatures associated with ORR and PFS. A detailed list of all detected cytokines is provided in Supplementary Table S1.

2.6 Statistical analyses

The sample size was based on a reported PFS of 5 months for locally advanced or metastatic LSCC patients receiving first-line gemcitabine plus cisplatin, and the expected PFS of patients receiving paclitaxel liposome plus cisplatin was 6.5 months. For the enrolment period of 18 months and study period of 36 months, to obtain a statistically significant result of mid-term analysis, a total of 466 PFS events were needed. By assuming two-sided α = 0.05, β = 20% and a dropout rate of at least 20%, the total sample size was calculated to be 536.

Statistical analyses were pre-specified and followed the intention-to-treat (ITT) principle which implied that all participants randomized should be included in the statistical analysis and analyzed according to the group they were originally assigned, regardless of the treatment regimen administered to them. The full analysis set (FAS) included all patients who received at least one dose of the study medication, had a baseline evaluation and at least one post-baseline evaluation. The per protocol set (PPS) included patients who met the eligibility criteria and did not show any major violation of the study protocol, completed the treatment, underwent efficacy evaluation and showed good compliance. Patients who did not receive full evaluation for efficacy were censored. A mid-term and a final analysis were conducted using the O'Brien-Fleming alpha spending function approach. The significance levels assigned to the mid-term and final analysis were α1 = 0.005 and α2 = 0.048, respectively.

For the cytokine study, the Wilcoxon rank-sum test and Wilcoxon signed-rank test were performed to determine the possible differences between two independent or paired subgroups of cytokines, respectively. Univariate Cox regression analysis was used to screen the cytokine signatures associated with ORR and PFS. PFS analysis was performed using the Kaplan-Meier method and log-rank test.

The safety set included all patients who received at least one dose of the study drug, had at least one follow-up safety assessment and were analyzed mainly using descriptive statistics.

All statistical analyses were conducted using the SAS software package, version 9.4 (SAS Institute Inc., Cary, NC), GraphPad Prism 8.0 and the SPSS statistical software, version 24.0 (SPSS Inc., Chicago, IL, USA). All the tests were two-tailed with a level of significance set at α = 0.05 except for testing the primary endpoint.

3.1 Patient demographic and baseline characteristics

The study flowchart has been shown in Figure 1. Five hundred forty patients underwent screening for eligibility and were randomized to receive LP regimen (n = 268) or GP regimen (n = 272). Three patients from each group did not receive the study medication and were excluded from the study. The FAS included 265 patients in the LP group and 269 patients in the GP group. One hundred seventy-seven (33.2%) patients had stage IIIb LSCC and 357 (66.9%) patients had stage IV LSCC. The demographic and baseline variables were well-balanced between the two different treatment groups (Table 1).

3.2 Treatment characteristics

Patients in the LP and GP group received a mean (± standard deviation) of 4.4 ± 1.8 and 4.5 ± 1.7 cycles of chemotherapy, respectively. 73.2% of patients in the LP group and 74.0% of patients in the GP group completed four or more cycles of chemotherapy. Significant dose reductions were required in 33.2% of patients in the LP group and 27.1% of patients in the GP group. Furthermore, 11.3% of patients in the LP group and 26.4% of patients in the GP group experienced treatment interruptions, respectively (Table 2).

3.3 Efficacy results

The last date of follow-up was on December 10, 2019, and the median duration of follow-up was 15.4 months (range: 0.2 to 34.5 months). One patient was lost during follow-up. 237 patients in the LP group and 253 patients in the GP group were included in the PPS. In the PPS, the median PFS was 5.2 months (95% confidence interval [CI], 4.7-5.6) in the LP group and 5.5 months (95% CI, 5.1-5.8) in the GP group (HR: 1.03, P = 0.742) (Figure 2A). The median OS in the LP group was 14.6 months (95% CI 12.4-15.8) and was 12.5 months in the GP group (95% CI 10.8-14.6) (HR: 0.86, P = 0.215) (Figure 2B). There were no significant differences between the two groups in terms of PFS (P = 0.742) and OS (P = 0.215) in the PPS. Across most subgroups that were analyzed, the LP group exhibited a similar PFS compared with the GP group (Figure 2C).

In the PPS, there were no significant differences between the LP group and the GP group regarding ORR (LP: 41.8% vs. GP: 45.9%, P = 0.412), and DCR (LP: 90.3% vs. GP: 88.1%, P = 0.443) (Table 3). The results of PPS were found to be generally consistent with those of FAS (Supplementary Table S2).

3.4 Safety results

In the current study, 534 patients were included in the safety set. AEs of any grade were reported in 97.7% of patients in the LP group and 99.6% of patients in the GP group (Table 4). SAEs occurred in 31.3% of patients in the LP group and 31.6% of patients in the GP group. A significantly lower proportion of patients in the LP group experienced AEs leading to treatment interruptions (10.9% vs. 26.4%, P < 0.001) or treatment termination (14.3% vs. 23.1%, P = 0.011). Though numerically more patients in the LP group experienced AEs leading to dose reductions, no statistical difference was found between the two groups (29.8% vs. 22.7%, P = 0.063).

The most frequently observed AE of any grade in the LP and GP group was anemia (71.3% vs. 80.7%, P = 0.015), followed by leucopenia (46.8% vs. 60.6%, P = 0.002) and neutropenia (43.0% vs. 55.0%, P = 0.006) (Supplementary Table S3). The most frequent grade ≥ 3 AE in the LP group was neutropenia (35.5%) followed by leucopenia (23.4%) and bone marrow suppression (15.5%) (Table 4). The most frequent grade ≥ 3 AE in the GP group was anemia (31.2%) followed by neutropenia (28.3%) and leucopenia (19.0%). The incidence of grade ≥ 3 anemia and thrombocytopenia in the LP group were significantly lower than in the GP group (Table 4).

3.5 The association between plasma cytokines and clinical outcomes of LSCC patients and observed dynamic changes in plasma cytokines in the LP group

To explore the association between the plasma cytokines levels and clinical outcomes of LSCC patients receiving LP treatment, we examined a panel of 45 cytokines in 86 patients randomly chosen from the LP group. Patient demographics, baseline features, and efficacy outcomes are shown in Supplementary Table S4. Figure 3A shows the expression levels of 45 cytokines in patients before initiation of LP therapy. Twenty-seven cytokines including TNF-α, interferon-gamma (IFN-γ), and IL-12p70 were differentially expressed in 45 patients who achieved PR and 41 patients who had SD (n = 27) or PD (n = 14) during the first efficacy evaluation (P < 0.05) (Figure 3B and Supplementary Figure S1). Univariate Cox regression analysis further showed that 15 cytokines including TNF-α, IFN-γ, IL-6, and IL-8 were significantly correlated with PFS (P < 0.05) (Figure 3C), with lower cytokine levels indicating better PFS (P < 0.05) (Supplementary Figure S2). Cytokines can function as effective but complex immune mediators. An essential feature of cytokine signaling is the degree of redundancy, in which multiple cytokines may display overlapping activities [17]. We, therefore, hypothesized that a group of cytokines, known as peripheral immune signature, might function synergistically. To better describe the function of these cytokines, the 45 cytokines were classified into different signatures according to their biological characterization, and corresponding signature scores were calculated. The scores of tumor necrosis factor signature and interferon signature in the LSCC patients before initiation of LP therapy were found to be negatively correlated with PFS; with lower cytokine signature scores indicating better PFS (Figure 4C).

To explore the potential of a subsequent combination of immunotherapy in the LP group, we examined the plasma cytokine levels in 57 LSCC patients with the paired samples of baseline and first efficacy evaluation. The plasma levels of eotaxin (CCL11) were significantly elevated (P = 0.005) while BDNF (P = 0.002), IL-8 (CXCL8) (P = 0.004) and IP-10 (CXCL10) (P = 0.003) were noticeably reduced after the two cycles of LP therapy versus baseline (Figure 4A). Moreover, subgroup analysis further showed that the levels of plasma CCL11 (P = 0.004) and SCF (P = 0.018) were significantly elevated while brain-derived neurotrophic factor (BDNF) (P = 0.006), IL-8 (P = 0.005) and CXCL10 (P = 0.002) were noticeably declined after the two cycles of LP therapy versus baseline in 39 patients who achieved PR (Figure 4B). Meanwhile, the levels of plasma VEGF-A were significantly elevated after two cycles of LP therapy versus baseline in 18 patients who had SD or PD (P = 0.043) (Figure 4B). The various dynamic changes in plasma cytokine levels by PFS achieved are shown in Figure 4C.