The efficacy and safety of combination of PD-1 and CTLA-4 inhibitors: a meta-analysis

Background Recently, a series of clinical trials showed that combination of anti-programmed cell death-1 (α-PD-1) and anti-cytotoxic T-lymphocyte-associated protein 4 (α-CTLA-4) could effectively eliminate tumor. However, in comparison with widely adopted mono-immune checkpoint inhibitors, chemotherapy, and targeted therapy, the advantage of combination therapy of α-PD-1 and α-CTLA-4 in response rate and prognosis is controversial especially considering probably increased treatment related adverse event. Thus, we conducted this meta-analysis to explore the efficacy and safety of combination treatment of α-PD-1 and α-CTLA-4. Methods This meta-analysis involved 8 clinical trials. In most trials, the primary endpoint was objective response rate (ORR). Thus we calculated risk ratio (RR) and 95% confidence interval (CI) to compare ORR of patients undergoing different treatment strategies. Moreover, the co-primary endpoints in few trials included progression-free survival and overall survival. Hazard ratio (HR) with 95% CI were employed to weigh the influence of different treatments on prognosis of patients. Subgroup analysis was conducted in patients with high and low expression of PD-L1. Lastly, the safety of combination therapy was evaluated by comparing treatment related adverse events among various treatment groups. Results Our results showed that ORR was significantly higher in patients receiving α-PD-1 plus α-CTLA-4 compared with α-PD-1 (RR 1.31, 95% CI 1.16–1.48) or α-CTLA-4 monotherapy (RR 2.11, 95% CI 1.84–2.43), chemotherapy and targeted therapy (RR 1.41, 95% CI 1.26–1.58). α-PD-1 plus α-CTLA-4 treated patients had a great advantage on monotherapy, chemotherapy and targeted therapy treated patients in PFS. Notably, no significant alteration in total adverse event rate was observed in α-PD-1 plus α-CTLA-4 treated patients. Results of subgroup analysis showed that combination therapy could enhance anti-tumor response in comparison with other treatments, especially for low PD-L1 expression patients undergoing nivolumab treatment (ORR: RR 1.35, 95% CI 1.11–1.65). Conclusion Combination treatment of α-PD-1 and α-CTLA-4 is a feasible strategy with enhanced efficacy and acceptable adverse event. Moreover, for some low PD-L1 expression patients, α-CTLA-4 might decrease the risk of resistance to α-PD-1 and demonstrate the synergistic anti-tumor effect.


Background
Cancer-immunity cycle model was established in 2013 to describe a series of stepwise events regulating anti-tumor immune response [1]. In this model, immune checkpoints act as inhibitory modulators and help cancer cell escape immune surveillance [2,3]. As a vital immune checkpoint molecule, cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) is constitutively expressed by regulatory T cells (Tregs) but transiently expressed by conventional T cells post activation [4][5][6]. Apart from T cell receptor (TCR) recognizing antigen peptide-major histocompatibility complex, CD28 binding to CD80 or CD86 is an essential co-stimulatory signal for T cells activation. CTLA-4 is a competitive antagonist for CD28-CD80/86 binding which further impedes priming and activation of T cells [7].
In theory, anti-PD-1 (α-PD-1) plus anti-CTLA-4 (α-CTLA-4) treatment simultaneously block two inhibitory signaling pathways of anti-tumor immune response [12]. However, in some clinical trials, no significant advantage was observed in therapeutic effect parameters such as objective response rate (ORR), progressionfree survival (PFS), and overall survival (OS) for patients undergoing α-PD-1 plus α-CTLA-4 treatment, especially in comparison with α-PD-1 monotherapy treated patients [13]. Besides, combination therapy might increase the risk of treatment related adverse event, causing treatment discontinuation [14]. Therefore, the combination therapy might not absolutely bring benefit to patients.
To comprehensively compare the efficacy and safety of combination therapy of α-PD-1 and α-CTLA-4 with monotherapy, chemotherapy, and targeted therapy, we reviewed the relevant clinical trials and conducted this meta-analysis. Moreover, given the crucial role of PD-L1 expression in immune checkpoint inhibitor treatment, we performed a subgroup analysis to evaluate efficacy difference among different treatments in the context of high or low PD-L1 expression [2].

Study design and systematic review protocol
This meta-analysis was designed based on Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [15].

Participants, interventions, comparators
Randomized controlled trials included in the metaanalysis all consisted of one treatment arm (α-PD-1 plus α-CTLA-4) and one or two control arms such as α-PD-1 or α-CTLA-4 monotherapy, chemotherapy, and targeted therapy. ORR, PFS, and OS were primary parameters to evaluate efficacy of treatment. Response Evaluation Criteria in Solid Tumors [RECIST] 1.1 was adopted to measure treatment outcome. The safety of treatments was estimated by probability of any grade and 3-4 grade adverse event. The assessment of adverse event was according to National Cancer Institute Common Terminology Criteria for Adverse Events (version 4.0).

Data sources, studies selections and data extraction
All studies included in the meta-analysis met the following criteria: (1) randomized controlled trial; (2) efficacy and/or safety of α-PD-1 plus α-CTLA-4 therapy was investigated; (3) patients in control arm received other treatments except for combination treatment mentioned above; (4) efficacy and safety data were available in the paper. Studies were excluded according to the standards as: (1) non-randomized controlled trial; (2) efficacy and safety data were not available in the paper; (3) patients in control arm underwent combination treatment as well, aiming to explore the influence of dose on treatment effect; (4) article contained updated data in the followedup paper.
Two authors (Ming Yi and Shuang Qin) independently selected studies met the inclusion criteria. All screened studies were filtered by title and abstract firstly. Then, uncertain studies were assessed by fulltext review. For all studies included in meta-analysis, we extracted data including study name, first author name, cancer type, treatment arm, control arms, number of patients assigned into every arm, efficacy data, and safety parameters. For few studies without available Hazard Ratio (HR) and 95% confidence interval (CI), HR value was estimated from Kaplan-Meier curve by Engauge-Digitizer software.
The Cochrane Collaboration's tool for assessing risk of bias was employed to assess each involved study [16]. Selection bias, performance bias, detection bias, attrition bias, as well as reporting bias were evaluated.

Data analysis
Comparison of efficacy between combination therapy and other treatments was conducted by RR of ORR, HR of PFS and OS. Safety of treatment was evaluated by RR of adverse event. Heterogeneity among treatment groups was assessed by Chi square-based Q statistic. If I 2 > 50% or p < 0.05, random-effect model was adopted [17]. Otherwise, fixed-effect model was employed. A funnel plot with 10 studies or less is often misinterpreted, so we did not assessed publication bias [18]. All data were performed by Stata software (version 12.0).
Treatment related adverse event in combination therapy. We assessed the safety of combination therapy by rate of any grade adverse event and grade 3-4 adverse event. On the whole, there was no significant difference in total adverse event rate between combination therapy and other treatments (pooled RR 1.35, 95% CI 0.93-1.97) (Fig. 4). Subgroup analysis showed that neither α-PD-1 monotherapy (pooled RR 2.02, 95% CI 0.98-4.17) nor α-CTLA-4 monotherapy (pooled RR 1.78, 95%        Efficacy of combination therapy in different PD-L1 expression statuses. In view of that PD-L1 was adopted as a predominant biomarker to screen patients for α-PD-1 treatment, we compared the efficacy of combination therapy with other treatments. In comparisons of combination therapy vs. α-CTLA-4 monotherapy and vs. chemotherapy or targeted therapy, combined therapy showed significant advantages in ORR, PFS, and OS regardless of PD-L1 expression status (Figs. 5, 6, 7). Notably, the advantages of combination therapy over α-PD-1 monotherapy were not consistent in different PD-L1 expression statuses. For patients with high PD-L1 expression (determined as ≥ 1% in the most studies), there was no significant difference between combination therapy and α-PD-1 monotherapy in ORR (pooled RR 1. 26  ( Fig. 5b). Moreover, combination therapy treated patients tended to have a better outcome even though without reaching statistical significance (pooled HR of PFS 0.52, 95% CI 0.22-1.21) than α-PD-1 monotherapy (Fig. 6b). Despite limited data, results of CheckMate 067 showed the advantage of combination therapy over α-PD-1 monotherapy in OS (HR 0.70, 95% CI 0.50-0.98) (Fig. 7b).

Risk of bias
According to Cochrane Collaboration's tool for assessing risk of bias, we assessed the quality of all involved randomized controlled trials ( Table 2). As a whole, 8 studies included in the meta-analysis were high-quality random controlled trials with most information at low risk of bias. Besides, we conducted a sensitivity analysis by Stata software 12.0 with metainf command. The pooled values did not changed significantly in the condition that any one study was omitted (Additional file 1: Figure S1).

Discussion
Immune checkpoint inhibitors relieve inhibitory tumor immune microenvironment and restore T cells activity from exhausted status. Reactivated T cells could effectively recognize tumor cell-derived neoantigen and subsequently kill tumor cell. However, in clinical practice,  the application of immune checkpoint inhibitor is limited by unsatisfactory response rate. Patients undergoing α-CTLA-4 or α-PD-1 monotherapy are prone to primary or adaptive resistance to immune checkpoint inhibitor. However, based on our meta-analysis, primary resistance could be overcome by α-CTLA-4 plus α-PD-1 treatment.
The primary reason is that simultaneously blocked two inhibitory signaling pathways have a synergistic effect for anti-tumor immunity [19]. CTLA-4 mainly targets interaction between antigen presentation cells (APCs) and naïve T cells which interferes the expansion of T cells epitopes. Therefore, α-CTLA-4 broadens repertoire of TCR and enhances recognition of tumor associated antigen and neoantigen. Nevertheless, due to inhibitory tumor immune microenvironment, emergence of tumor specific T cells is a necessary but not sufficient condition for tumor elimination. Accompanied with accumulated tumor infiltrating lymphocytes, upregulated PD-L1 expression by inflammatory signals such as interferon-γ means deficient immune surveillance even though formation of "hot tumor". Based on that, we speculated that α-PD-1 could substantially reduce primary resistance. Moreover, we supposed that decreased probability of adaptive resistance contributed to the improved prognosis as well. Immunoediting during the cancer progression is an important reason for adaptive resistance [20]. Broaden epitopes resulting from combination therapy could reduce the failure of recognition subclonal tumor cell-derived antigen, which provides durable and potent tumor-killing activity. Despite all of this, the main concern of oncologists about combination therapy is the magnified risk of adverse event [21]. Our meta-analysis showed that there was no significant difference between patients received combination therapy and monotherapy in total adverse event rate. Even though monotherapy had fewer highgrade adverse event rate than combination therapy, we believed that adverse event of combination therapy was acceptable especially compared with chemotherapy or targeted therapy.
Given that PD-L1 expression strongly relates with efficacy of α-PD-1 treatment. Therefore, we investigated efficacy of combination therapy in different PD-L1 expression statuses. Notably, in the context of high PD-L1 expression, the advantage of combination therapy in efficacy over α-PD-1 monotherapy is not significant. However, in the condition of low PD-L1 expression, outcome of combination therapy was obviously better than α-PD-1 monotherapy. Our analysis suggested that for patients with high PD-L1 expression, α-PD-1 monotherapy would be a better option for minimized adverse event and medical cost. On the other hand, for patients with low PD-L1 expression, α-CTLA-4 treatment could increase patients' sensitivity to α-PD-1 treatment. It is obvious that the combination therapy had better treatment effect than ipilimumab, chemotherapy, and targeted therapy in any PD-L1 expression condition.
In fact, other than α-CTLA-4, many interventions such as radiotherapy, oncolytic virus, and cancer vaccine have been adopted to enhance efficacy of α-PD-1 therapy [22,23]. Due to interdependence between different antitumor immune stepwise events, enhanced neoantigen release, recognition, and priming/activation tumor-associated antigen or neoantigen specific T cells all indirectly promote downstream tumor-killing activity [24]. The complexity of tumor immune microenvironment suggests it is hard to completely reverse inhibitory microenvironment by a single-target therapy [25]. Therefore, combination therapy would be a promising strategy and deserves further attention.
The application of immune checkpoint inhibitors is changing the landscape of cancer therapeutics [26,27]. In the meanwhile, from an economic point of view, it was reported the total healthcare cost of patients receiving nivolumab plus ipilimumab treatment was lower than patients undergoing nivolumab monotherapy or ipilimumab monotherapy [28]. This cost advantage of combination therapy was attributed to the lower non-drug cost due to decreased hospitalization rates after initiation treatment [28]. Some limitations still existed in our meta-analysis. Firstly, we resolved most heterogeneity by subgroup. However, we classified chemotherapy and sunitinib as one class named chemotherapy or targeted therapy, for distinguishing them from immunotherapy. Actually, this classification resulted in some heterogeneity existed in some subgroup analyses. Secondly, limited by amount of available studies, we did not analyzed efficacy of combination treatment in each specific cancer type, so our results should be carefully interpreted. Finally, our metaanalysis just included English literatures which might cause potential selection bias.

Conclusions
Combination therapy of α-CTLA-4 and α-PD-1 had significant advantages in efficacy over monotherapy, chemotherapy and targeted therapy without significantly increased adverse event. For high PD-L1 expression patients, combination therapy did not show obviously enhanced efficacy than α-PD-1. However, for low PD-L1 expression patients, simultaneous administration of α-CTLA-4 and α-PD-1 would be an optimized strategy to acquire better clinical benefits by overcoming primary drug resistance.