The effects and safety of PD‐1/PD‐L1 inhibitors on head and neck cancer: A systematic review and meta‐analysis

Abstract Background Inhibitors of programmed cell death‐1 (PD‐1) and its ligand (PD‐L1) have been increasingly used in head and neck cancer therapy and reported to improve the outcomes with an acceptable safety profile. This systematic review and meta‐analysis was conducted to assess the benefit and risk of PD‐1/PD‐L1 inhibitors in patients with head and neck cancer. Method The PubMed, Cochrane Library, EMBASE and Web of Science databases were systematically searched to find potentially eligible studies up to May 30, 2019. Primary outcomes were overall survival (OS), progression‐free survival (PFS), objective response rate (ORR), disease control rate (DCR) and adverse events. Results Overall, this analysis consisted of nine eligible studies, with two randomized controlled trials and seven single arm trials. In the treatment of recurrent or metastatic head and neck cancer, PD‐1 inhibitors showed significantly lower relative risk of death than standard‐of‐care therapy (odds ratio [OR] = 0.60, 95% confidence interval [CI]: 0.44‐0.82, I 2 = 0%, P = .001). Programmed cell death‐1 inhibitors also decreased the risk of disease progression, however, there was no statistically significant difference of PFS between the treatments (OR = 0.69, 95% CI: 0.48‐1.01, I 2 = 0%, P = .05). Subgroup analysis showed that human papillomavirus (HPV) positive patients had higher response rates than HPV negative patients in PD‐1/PD‐L1 inhibitors‐treated population (ORR: 18.8% vs 12.2%; DCR: 42.8% vs 34.4%). The most common any‐grade and grade ≥3 treatment‐related adverse events were fatigue (14.7%, 95% CI: 12.3%‐17.1%) and aspartate aminotransferase increased (1.6%, 95% CI: 0.3%‐2.9%), respectively. Conclusion Programmed cell death‐1 inhibitors prolonged OS in comparison with standard‐of‐care therapy in recurrent or metastatic head and neck cancer patients. Human papillomavirus positive patients were superior to HPV negative patients in the treatment of PD‐1/PD‐L1 inhibitors. More phase III randomized controlled trials are warranted to confirm our findings.


| Search strategy and study selection
This meta-analysis was conducted followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses guideline (PRISMA). 18 The search was done in the electronic databases PubMed, Cochrane Library, Web of Science, and EMBASE to identify all relevant records until 30 May 2019. Additionally, search terms included "nivolumab or pembrolizumab or cemiplimab or atezolizumab or durvalumab or avelumab or PD-1 inhibitor or PD-L1 inhibitor", "head and neck cancer or head and neck or head and neck neoplasm or head and neck tumor or head and neck carcinoma", and "trial or clinical trial or randomized clinical trial or randomized controlled trial". We also manually searched the references of relevant published studies and reviews articles for more eligible trials. The search results were uploaded into a citation database program for further review (EndNote: http://endno te.com/).
Studies eligible for inclusion met all of the following criteria: (a) prospective clinical trials in patients with head and neck cancer; (b) for single arm trials, participants were treated with only an anti-PD-1/PD-L1 agent; (c) for controlled clinical trials, participants in the control arm were treated with standard-of-care therapy; (d) antitumor activity and safety data was included; (e) trials were published in English. We excluded conference abstracts due to the absence of safety data and the increase in heterogeneity. For multiple publications that were identified reporting on the same clinical study, the one with the most complete publication data was eligible. Any discrepancies were resolved by discussion.

| Data extraction
Detailed reviews of full-text articles regarding study design, baseline characteristics, outcomes, and toxicities were performed by BW and CF independently. The trial name, publication year, study design, number of patients, number of male patients, PD-L1 expression, HPV status, anti-PD-1/PD-L1 agent used, dosing schedule, OS, PFS, objective response rate (ORR), disease control rate (DCR), median duration of response, and safety data (number of any-grade and grade ≥3 treatment related adverse events) reporting in the articles and Supplementary materials were obtained from each included study.

| Statistical analysis
Data from randomized controlled trials (OS and PFS) and PD-1/PD-L1 arm (ORR and DCR) was assessed by odds ratio (OR) and 95% confidence interval (CI). The subgroup analyses of OS were assessed by hazard ratio (HR) and 95% CI because of the absence of individual data. The abovementioned meta-analyses were conducted using RevMan version 5.3 software (Cochrane Collaboration's Information Management System). If the heterogeneity test showed no statistical significance (P ≥ .10, I 2 ≤ 50%), a fixed-effects model was used. Otherwise, a random-effects model was applied; P < .05 was considered as statistically significant outcomes.
Pooled effect sizes of ORR, DCR, and treatment-related adverse events were done using STATA statistical software, version 14.0. Random-effects models were applied in these analyses due to the single arm data. Statistical heterogeneity among the studies was tested by the Cochran Q chi-square test and I 2 statistic percentages. I 2 < 50% or P > .10 was defined as low heterogeneity, otherwise was high heterogeneity. Egger's test was displayed to evaluate latent publication bias for small-study effects.
The characteristics of the nine eligible studies were described in Table 1, and the summary of outcomes was presented in Table 2.

| Overall survival
Data regarding OS was available from two of nine studies, 21,23 including 487 patients in the PD-1/PD-L1 inhibitor group and 369 patients in the standard-of-care therapy group. Forest plots showed that the PD-1/PD-L1 inhibitor group had a 40% lower risk of death compared to the standard-ofcare therapy group (OR = 0.60, 95% CI: 0.44-0.82, I 2 = 0%, P = .001) (Figure 2A).

| Progression-free survival
Progression-free survival data was also available from the two studies in OS analysis. Forest plots showed that the PD-1/ PD-L1 inhibitor group had a 31% lower risk of disease progression compared to the standard-of-care therapy group, but the difference was not statistically significant (OR = 0.69, 95% CI: 0.48-1.01, I 2 = 0%, P = .05) ( Figure 2B).

PD-L1 negative subgroups
In the pooled analysis of OS in the subgroup of PD-L1 positive patients, PD-1 inhibitors-treated patients had a risk of death that was 31% lower than the risk among patients assigned to standard-of-care therapy. The HR for death among patients treated with PD-1 inhibitors vs standard-of-care therapy was 0.69 (95% CI: 0.56-0.85, I 2 = 29%, P < .001) ( Figure  3A), whereas in the subgroup of PD-L1 negative patients, the HR was 1.08 (95% CI: 0.77-1.52, I 2 = 7%, P = .66) ( Figure  3B).

| Objective response rate by single arm analysis
The ORR data were available from nine studies. Pooled ORR of overall patients was 14.2% (95% CI: 11.2%-17.2%). Subgroup analysis indicated that the pooled ORR of HPVpositive group was 18.8% (95% CI: 12.9%-24.6%) vs HPVnegative group was 12.2% (95% CI: 8.6%-15.7%) ( Table 3). In six trials included both HPV-positive and HPV-negative groups, we analyzed the OR for ORR in PD-1/PD-L1 arm, involving 192 patients in the HPV-positive group and 319 patients in the HPV-negative group. Forest plots showed that patients in the HPV-positive group had a 56% higher chance to achieve ORR in comparison with the HPV-negative group. However, there was no statistically significant difference (OR = 1.56, 95% CI: 0.93-2.61, I 2 = 1%, P = .09) ( Figure 4A).

| Publication bias
For OS and PFS, publication bias was not assessed due to the small number of studies analyzed. In the results of Egger's test based on the overall and subgroup analyses of ORR (P = .246 > 0.1) and DCR (P = .914 > 0.1), publication bias was not observed. Above all, this meta-analysis was at low risk for reporting bias.

| DISCUSSION
Most head and neck cancer patients presented with stage III-IV disease at diagnosis. Over half of the patients recurred loco-regionally or distantly and median OS was less than a  year. 2,28,29 The poor outcome emphasizes the need for better treatment strategies. Moreover, head and neck cancer is an immunosuppressive tumor, with a lower lymphocyte count than those found in healthy participants. 30 Programmed cell death-1 ligand was found to be expressed in up to 60% of patients with head and neck cancer, leading to impaired natural killer cell activity and poor antigen-presenting function. 30,31 Development of anti-PD-1/PD-L1 agents might be a novel way to modify the standard treatment of head and neck cancer.  Thus, we conducted this meta-analysis to estimate the antitumor activity and safety of PD-1/PD-L1 inhibitors in patients with head and neck cancer. The present meta-analysis quantitatively integrated the data of published clinical trials.
In the eligible studies, only CheckMate 141 and KEYNOTE-040 were randomized and standard-of-care therapy controlled clinical trials. OS and PFS data were extracted for further analyses. However, we noticed that PD-1 inhibitors prolonged OS in head and neck cancer patients, but shortened PFS compared with standard-of-care therapy, regardless of pretreatments. The difference might be attributed to the inclusion of both PD-L1 positive and PD-L1 negative patients. [32][33][34] In a phase III study, KEYNOTE-040, the survival results also showed that pembrolizumab provided a clinically meaningful   35 Immune checkpoint inhibition with monoclonal antibodies against cytotoxic T lymphocyte antigen-4 and PD-1/PD-L1 by combination treatments became another new option in head and neck cancer. 36 The ongoing EAGLE study is comparing the efficacy and safety of durvalumab as monotherapy or in combination with tremelimumab to standard-of-care in head and neck cancer. 37 In CheckMate 141 trial, treatment with nivolumab resulted in longer OS than standard therapy, regardless of PD-L1 expression or HPV status. KEYNOTE-055 and PCD4989g studies also concluded that no association between HPV status and response to pembrolizumab or atezolizumab was detected. However, HAWK study revealed that HPV positive patients with recurrent or metastatic head and neck cancer had a numerically higher response rate and survival than HPV negative patients.
In the pooled analysis of ORR and DCR, HPV positive patients had a higher ORR (18.8% vs 12.2%) and a higher DCR (42.8% vs 34.4%) than HPV negative patients. However, in the single arm analysis, PD-1/PD-L1 inhibitors treated HPV positive patients showed a higher ORR, the difference was not statistically significant. Even though there was no significant difference, HPV status could also be a potential predictive biomarker in the immune checkpoint inhibitors treatments for patients with recurrent or metastatic head and neck cancer. This hypothesis will need to be tested to explain the responses observed in the subgroup based on HPV status.
All selected clinical trials reported a manageable safety profile. Nonetheless, from the standpoint of patient counseling, several results of treatment-related adverse events should be paid more attention. Approximately, 61.9% head and neck cancer patients treated with PD-1/PD-L1 inhibitors in clinical trials experienced at least one adverse event of any grade, and 12.8% head and neck cancer patients had at least one grade ≥3 adverse event. Moreover, PD-1 inhibitors had a higher incidence of any-grade and grade ≥3 adverse events than PD-L1 inhibitors

| Limitations
This meta-analysis has several limitations. First, there were only two randomized controlled trials limiting the analysis in our study. Second, the analyses of ORR and DCR between HPV positive and HPV negative head and neck cancer patients were single arm researches. Despite these limitations, this meta-analysis is a meaningful study of the estimates the benefits and risk of PD-1/PD-L1 antagonists.
In conclusion, PD-1/PD-L1 inhibitors appeared to be effective for treating recurrent or metastatic head and neck cancer. Moreover, our study of anti-PD-1 therapy indicated a superior survival in patients with PD-L1 positive recurrent or metastatic head and neck cancer. At the same time, careful monitoring of the treatment related adverse events of anti-PD-1/PD-L1 agents will be needed. More phase III randomized clinical trials are warranted to confirm our findings.