Article Text


Anti-PD-1/anti-PD-L1 immunotherapy versus docetaxel for previously treated advanced non-small cell lung cancer: a systematic review and meta-analysis of randomised clinical trials
  1. Allan Ramos-Esquivel1,
  2. Alicia van der Laat2,
  3. Raquel Rojas-Vigott2,
  4. Melissa Juárez1,
  5. Luis Corrales-Rodríguez3
  1. 1Departamento de Oncología Médica, Hospital San Juan de Dios, Universidad de Costa Rica, San José, Costa Rica
  2. 2Departamento de Oncología Médica, Hospital México, San José, Costa Rica
  3. 3Department of Medical Oncology, Centro de Investigación y Manejo del Cáncer CIMCA, San José, Costa Rica
  1. Correspondence to Dr Allan Ramos-Esquivel; allan.ramos{at}


Background To compare the efficacy and toxicity of anti-programmed cell death receptor 1 (PD-1) and anti-programmed cell death ligand 1 (PD-L1) versus docetaxel in previously treated patients with advanced non-small cell lung cancer (NSCLC).

Materials and methods Phase III randomised clinical trials (RCTs) were identified after systematic review of databases and conference proceedings. A random-effect model was used to determine the pooled HR for overall survival (OS), progression-free survival (PFS) and duration of response. The pooled OR for overall response and treatment-related side effects were calculated using the inverse-variance method. Heterogeneity was measured using the τ2 and I2 statistics.

Results After the systematic review, we included four phase III RCTs (n=2737) in this meta-analysis. The use of anti-PD-1/anti-PD-L1 agents (atezolizumab, nivolumab and pembrolizumab) was associated with better OS in comparison with docetaxel alone (HR: 0.69; 95% CI 0.63 to 0.75; p<0.00001). Similarly, the PFS and duration of response was significantly longer for patients receiving immunotherapy (HR: 0.85; 95% CI 0.75 to 0.96; p=0.007 and HR:0.32; 95% CI 0.24 to 0.43; p<0.00001, respectively) versus single agent chemotherapy. The overall response rate was also higher for patients who received any anti-PD-1/anti-PD-L1 therapy in comparison with docetaxel (OR: 1.77; 95% CI 1.26 to 2.50; p=0.001). Regarding treatment-related side effects grade 3 or higher, patients who received immunotherapy experienced less events than patients allocated to docetaxel (OR: 0.19; 95% CI 0.12 to 0.30; p<0.00001)

Conclusion The use of anti-PD-1/anti-PD-L1 therapy in patients with progressive advanced NSCLC is significantly better than the use of docetaxel in terms of OS, PFS, duration of response and overall response rate.

  • non-small-cell lung cancer
  • atezolizumab
  • pembrolizumab
  • nivolumab
  • meta-analysis

This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See:

Statistics from

Key questions

What is already known about this subject?

  • Recent trials have acknowledged the survival benefit of checkpoint inhibitors in patients with advanced and progressive non-small cell lung cancer (NSCLC).

What does this study add?

  • This systematic review and meta-analysis resumes the available data on immunotherapy from randomised clinical trials and summarises the overall efficacy and safety of anti-programmed cell death receptor 1  (PD-1)/anti-programmed cell death ligand 1 (PD-L1) agents.

How might this impact on clinical practice?

  • Our findings supports the current clinical evidence regarding the overall survival benefit derived from anti-PD-1/anti-PD-L1 agents in comparison with docetaxel for previously treated patients with advanced NSCLC.


Lung cancer is the leading cause of cancer-related death worldwide, estimated to be responsible of 1.59 millions deaths in 2012.1 2 Non-small cell lung cancer (NSCLC) accounts for approximately 85% of cases, and the majority of patients are diagnosed with locally advanced or metastatic disease.3 Despite the significant therapeutic advances with the introduction of molecularly targeted therapies, antiangiogenics and new chemotherapeutic agents, the prognosis of the majority of these patients is still poor.1 Relapses are frequent after first-line chemotherapy or molecular-targeted agents against epidermal growth factor receptor (EGFR) or echinoderm microtubule-associated protein-like 4 and anaplastic lymphoma kinase fusion gene (EML4-alk), and most patients will not experience sustained disease control with second-line agents.4 For relapsing patients, docetaxel was approved as a second-line treatment based on two phase III trials.5

It has been recently acknowledged that cancer cells can induce immune tolerance in the tumour microenvironment by blocking coinhibitory signals of cytotoxic T-cells, also known as immune checkpoints.6 The programmed cell death receptor 1 (PD-1)/programmed cell death ligand 1 (PD-L1) interaction inhibits T-cell response, promotes differentiation of CD4 T cells into T regulatory cells, induces apoptosis of tumour-specific T cells and causes T cell resistance.6 Besides, cancer cells can also change the tumour microenvironment to induce an immunosuppressive state through expression of inhibitory cytokines and recruitment of regulatory T lymphocytes and myeloid-derived suppressor cells.7

Antibodies against PD-1 (nivolumab and pembrolizumab) and PD-L1 (atezolizumab) have demonstrated activity against chemotherapy-refractory NSCLC tumours in several phase III trials.8–11 These recent trials have questioned the paradigm of treatment for previously treated metastatic NSCLC. For this reason, we aimed to determine the overall efficacy and safety of these agents versus docetaxel.

Materials and methods

Search strategy and study selection

The Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) statement for reporting systematic reviews was followed (checklist available as online supplementary file 1).12 First, four authors (AR-E, AvdL, MJ and RR-V) independently scrutinised the titles and abstracts retrieved by a search strategy in electronic databases (MEDLINE, Embase and The Cochrane Central Register of Controlled Trials) from May 2007 to 1 May 2017 (online supplementary file 2). The search was done in May 2017. Proceedings of the American Society of Clinical Oncology annual meeting, International Association for the Study of Lung Cancer World Conference on Lung Cancer annual meeting and European Society of Medical Oncology annual meeting were also searched from 2012 to 2017 for relevant abstracts. In case of reports of the same trial, we included only the most recent results (corresponding to longer follow-up). Then, the authors examined full-text articles of potential eligible studies according to the eligibility criteria. Disagreements were resolved in discussion with another author (LC-R). Data extraction tables were designed specifically for this review to aid data collection. Data from relevant studies were extracted and included information on trial design, participants, interventions and outcomes.

Supplementary file 1

Supplementary file 2

Eligibility criteria

We decided to include published and unpublished phase III randomised controlled trials (RCTs) that enrolled patients with advanced non-small cell lung cancer (NSCLC). We included the reported comparisons of chemotherapy versus any anti-PD-1/anti-PD-L1 agent used as monotherapy. We excluded trials with incomplete data and those studies published in non-English languages. In case of finding a trial with more than two comparisons, we decided to count each intervention separately.


The primary outcome was overall survival (OS), calculated from the date of randomisation to the date of death. Secondary outcomes included: (1) progression-free survival (PFS): defined by the RECIST 1.1. criteria13; (2) objective response rate: defined as the percentage of patients with complete or partial response as per RECIST 1.1; and (3) duration of response, defined as time from first evidence of partial and complete response until disease progression or death.

We also evaluated the safety of each drug in all patients who received at least one dose of the study treatment. Adverse drug reactions were graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events, version 4.0.

Quality assessment

The risk of bias was evaluated by three reviewers using the Cochrane Collaboration Tool,12 including: sufficient sequence generation, adequate allocation concealment, blinding, incomplete outcome data addressed and freedom from selective reporting. Publication bias was visually examined in a funnel plot. The risk of bias was categorised as ‘low risk’, ‘high risk’ or ‘unclear risk’.

Data collection and statistical analysis

For time-to-event outcomes, the treatment efficacy was measured using the HR with its corresponding 95% CI. For the association of the risk of overall response, we employed the OR and its corresponding 95% CI. We used a random-effect model for the efficacy measures according to the DerSimonian-Laird method. The pooled HR and pooled risk ratio were calculated according to the inverse-variance method, as described by Parmar et al.14

Heterogeneity was determined by the τ2 and I2 statistics. Data analysis was performed using RevMan V.5.3 software.


Study selection

Through the search strategy, we identified four trials8–11 that enrolled 2737 patients with progressive disease for advanced non-small cell lung cancer who were treated with an anti-PD-1/anti-PD-L1 antibody or docetaxel (the PRISMA flow diagram for study inclusion is available as online supplementary file S1).

Description of studies and patients

Table 1 resumes the main characteristics of each trial. One trial explored the effect of pembrolizumab in two different doses. Therefore, we show the results for each specific dose. The primary outcome was OS in all trials. Of note, the majority of patients had wild-type EGFR non-squamous NSCLC. All patients had measurable disease per RECIST 1.1 criteria.

Table 1

General characteristics of the included trials


Table 2 summarises the main outcomes of each trial.

Table 2

Overall results of the included trials

Overall survival

As shown in table 2 and figure 1A, there was an OS improvement of anti-PD-1/anti-PD-L1 therapy in comparison with docetaxel (pooled HR=069; 95% CI 0.63 to 0.75). There was no evidence of significant heterogeneity among the included trials regarding this outcome (τ2: 0.00; I2: 0%; p=0.53).

Figure 1

Forest plot of HRs for overall survival (A), progression-free survival (B), duration of response (C) and ORs for overall response (D).

Progression-free survival

Patients with immunotherapy experienced less progression events than those patients receiving docetaxel (pooled HR: 0.85; 95% CI 0.75 to 0.96). There was evidence of moderate heterogeneity among the included trials for this specific outcome (τ2: 0.01; I2: 54%; p=0.07) (figure 1B).

Duration of response

The duration of response was significantly longer for patients receiving immunotherapy in comparison to docetaxel (pooled HR: 0.32; 95% CI 0.24 to 0.43). For this outcome we detected no significant heterogeneity (τ2: 0.00; I2: 0%; p=0.96) (figure 1C).

Overall response

The odds of overall response significantly increased with use of anti-PD-1/anti-PD-L1 therapy versus docetaxel alone (pooled OR: 1.77; 95% CI 1.26 to 2.50). There was significant heterogeneity found among the included trials (τ2: 0.09; I2: 61%; p=0.04) (figure 1D).

Treatment-related side effects

Regarding adverse drug reactions grade 3 or higher, use of docetaxel significantly increased the burden of therapy, with 42.2% of patients experiencing any treatment-related side effect, particularly neutropaenia and febrile neutropaenia (figure 2).

Figure 2

Forest plot of ORs for treatment-related side effects.

Subgroup analyses

Figure 3 shows the OS assessment according to EGFR mutation status and PD-L1 expression. We found heterogeneity between these two groups. Patients with wild-type EGFR were more likely to obtain an OS benefit from immunotherapy in contrast to patients with any EGFR mutation (p=0.005). Similarly, patients with high PD-L1 expression on immunohistochemistry had better OS than their counterparts (p=0.0001). Of note, the definition of high PD-L1 expression varied in each trial.

Figure 3

Subgroup analysis of OS according to EGFR mutation status and PD-L1 expression. EGFR, epidermal growth factor receptor; OS, overall survival; PD-L1, programmed cell death ligand 1.

Risk of bias

The risk of bias assessment is presented as online supplementary file S2. All included trials were open-label with high risk of performance and detection bias. Selection bias was likely to occur in one trial due to unmask allocation.9 We did not detect evidence of substantial publication bias in the funnel plot analysis (online supplementary file S3).

Supplementary file 3


This systematic review and meta-analysis resumes the available data from published phase III RCT regarding the OS benefit of anti-PD-1/anti-PD-L1 agents versus docetaxel in patients with progressive advanced NSCLC. In summary, patients receiving these checkpoint inhibitors live longer and had better responses than patients allocated to docetaxel. Although this meta-analysis includes three different drugs with similar mechanism of action (atezolizumab, nivolumab and pembrolizumab), we did not detect any significant heterogeneity in the OS assessment. However, these trials share some differences in the inclusion criteria. Specifically, the KEYNOTE-010 trial included patients with a cut-off of at least 1% PD-L1 positive staining.8 In contrast, use of atezolizumab and nivolumab was associated with an OS improvement regardless of the PD-L1 status.9–11 Besides, the OAK Trial allowed the inclusion of patients with PD-L1 expression on tumour cells or tumour-infiltrating immune cells.9 It has been previously demonstrated that PD-L1 is expressed in the cytoplasmic membrane of T-lymphocytes, macrophages and dendritic cells, and this expression may confer prognostic information.15 Furthermore, PD-L1 determination was different in each trial using different antibody clones and detection methods. Recent studies have determined that PD-L1 expression in NSCLC can be discordant due to different antibody affinities, limited specificity or distinct epitopes.16 These different methods could contribute to some of the heterogeneity found in our analysis. Still, efforts have been done to determine if a specific antibody confers a difference in the selection of any specific checkpoint inhibitor.17

Our findings confirm previous results showing that patients with high expression of PD-1/PD-L1 on immunohistochemistry derived the greatest benefit from immunotherapy in terms of OS. For example, the POPLAR study showed that the increased improvement in OS was associated with increased PD-L1 expression with use of atezolizumab in patients with previously treated NSCLC.18 This association between PD-L1 expression and response to therapy has been described in several clinical trials, including NSCLC and patients with melanoma,19 renal cell carcinoma,20 bladder carcinoma21 and gastric cancer.22 However, some other trials have not shown consistently shown between PD-L1 expression and the efficacy of checkpoint inhibitors23 24

Although the PD-1/PD-L1 expression seems to be a useful biomarker, there are multiple challenges regarding its use as a predictive marker of immunotherapy efficacy. First, the expression of these proteins is a dynamic and heterogeneous process.25 Besides, previous studies have described some discrepancies in the expression of these proteins between the primary tumour and the metastatic site.26 Furthermore, other inflammatory cells can also express these proteins and may alter the pathologist’s interpretation.27 Other authors argue that different patterns of staining confer different prognosis.26 Despite these controversies, our data support current evidence that PD-L1 expression could be a useful biomarker to predict better responses to PD-1/PD-L1 inhibitors.

Our findings also showed that patients with EGFR wild-type NSCLC derived better OS improvement with anti-PD-1/anti-PD-L1 than tumours harbouring common EGFR mutations. Of note, the total proportion of patients with EGFR mutations was low (11.2%), and we cannot exclude the possibility of a type II error due to the small size of this particular subgroup in the whole sample. Besides, some of the included RCTs did not assess the EGFR status in the recruited patients. Despite these caveats, our findings are in line with previous studies showing that EGFR-mutated NSCLC has low mutation burden in comparison with wild-type EGFR tumours,28 and this low mutation burden can relate to a decreased sensitivity to checkpoint inhibitors. The mechanisms behind this association may be related to the overexpression of cytoplasmic neoantigens formed as consequence of somatic mutations that can be recognised by effector T-cells.29 30 Further analyses are warranted to determine the efficacy of immunotherapy in patients with common EGFR mutations as well as in patients harbouring ALK rearrangements.

The assessment of the secondary outcomes of this meta-analysis also favoured the immunotherapy arm in terms of PFS, overall response, duration of response and fewer side effects than chemotherapy with docetaxel. The overall HR for PFS was mainly derived from data from the CheckMate 017 Trial that showed a significant benefit in terms of progression for patients who received nivolumab in squamous NSCLC.10 Since treatment with checkpoint inhibitors was allowed beyond disease progression in all included trials, we considered that this finding is not clinically relevant.

However, we detected some heterogeneity between trials in the assessment of these secondary outcomes. Some sources of heterogeneity can be attributed to differences in the proportion of patients with PD-1/PD-L1 expression across the included trials. As previously described, the rate of response is highly dependent on PD-L1 positivity.23 24

Regarding side effects, trials assessing the efficacy of nivolumab in squamous and non-squamous NSCLC reported less grade 3 or higher toxicity (from 7% to 10%) in the experimental arm than trials using pembrolizumab or atezolizumab. Conversely, the included RCTs with nivolumab reported more toxicity from docetaxel use than the two remaining trials. Therefore, we suggest that the OR of treatment-related side effect is quite low among trials using nivolumab in comparison with RCTs using pembrolizumab and atezolizumab. However, the rate of discontinuation was very similar among the included trials, and the pattern of side effects was very similar among anti PD-1/PD-L1 therapies.

Our findings must be interpreted cautiously since all included studies have high risk of bias, especially performance and detection bias. These systematic errors could overestimate or underestimate the association measures of each trial. Besides, the aforementioned heterogeneity found in the secondary outcomes deserves further study to determine if there are efficacy differences among atezolizumab, nivolumab and pembrolizumab. However, our findings confirm the class effect of anti-PD-1/anti-PD-L1 therapy as a mainstay of treatment for patients with progressive advanced NSCLC.


In conclusion, our findings show an OS improvement of anti-PD-1/anti-PD-L1 therapy versus docetaxel in previously treated advanced NSCLC. PFS, overall response and duration of response also favoured the use of checkpoint inhibitors when compared with chemotherapy. In terms of side effects, even though the toxicity profile is different between checkpoint inhibitors and chemotherapy, grade 3 or higher adverse events were more likely seen with docetaxel .


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
View Abstract


  • Contributors All authors had full access to all the data in the study and take responsibility for the integrity and accuracy of the data analysis. All authors contributed to the study concept and design. All authors read and approved the final version of this manuscript.

  • Funding This work was partially supported by Roche (no grant number is applicable).

  • Competing interests None declared.

  • Patient consent None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.