Multiple chemotherapy-based combination therapy strategies for advanced lung cancer patients: a systematic review and network meta-analysis

Background: At present, the treatments for patients with advanced lung cancer focus on chemotherapy, targeted therapy, immunotherapy, or a combination of multiple treatments. Purpose: The main purpose of this study is to compare the various chemotherapy-based combination therapies and find the best one for patients with advanced lung cancer. Methods: Based on database (PubMed, EMBASE and Medline) for randomized controlled trials of advanced lung cancer with combination therapy from 2008 to 2020, we searched literatures with overall survival (OS), progression-free survival (PFS), objective response rate (ORR) and adverse as outcome indicators and established a Bayesian mesh meta-analysis for multiple treatment strategies. Then, we combined the results of four outcome indicators to find out the best chemotherapy-based combination therapy strategy for patients with advanced lung cancer, further, we tried to screen out the best drugs of which were commonly used now. Results: It contained a total of 51 studies, including five combination therapies: Chemotherapy/Chemotherapy plus placebo (CT), chemotherapy plus one targeted therapy drug (CT+T), chemotherapy plus two targeted therapy drugs (CT+T+T), chemotherapy combined with immunotherapy (CT+I) or chemotherapy combined with biotherapy (CT+B). In terms of four outcome indicators, CT+I showed the best therapeutic benefits. In the comparison of immunotherapy drugs, pembrolizumab showed the best effect. Conclusion: Our results showed that, among the multiple chemotherapy-based Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 31 May 2020 doi:10.20944/preprints202005.0478.v1


Introduction:
Lung cancer is the most common type of malignancy worldwide and is also the leading cause of cancer-related deaths (18.4% of the total number of cancer deaths) [1].
More than 50% of patients diagnosed at a distant metastasis, for which the median overall survival (OS) is less than 1 year and the 5-year survival rate is only 4% [2].
For decades, platinum-based chemotherapy has been the main treatment for most patients with advanced lung cancer. But with the emergence and development of targeted therapy and immunotherapy in recent years, patients have more choices. More and more clinical trials have confirmed that both targeted therapy [3,4] and immunotherapy [5][6][7] can bring survival benefits to patients in advanced stage. Therefore, the current status of coexistence of three treatment methods (chemotherapy, targeted therapy and immunotherapy) for patients with advanced lung cancer has been formed. However, the current status of treatment for patients with advanced lung cancer is still not optimistic. Targeted therapy only works well in patients with specific mutants [8], and even in the presence of mutants, most patients will experience various degrees of resistance in the coming months after using targeted therapy [9]. In addition, studies have shown that immunotherapy works only in a small number of patients, and many patients who initially responded quickly showed disease progression [10]. Therefore, it is imperative to find a more universal and effective treatment for patients with advanced lung cancer. In the clinical treatment of advanced lung cancer, combination therapy is an important way to avoid various drawbacks of using single drug. Some studies have shown that targeted therapy [11] [12] or immunotherapy [13] combined with chemotherapy can achieve better prognosis, but there is still controversy [14]. Therefore, this study compared the efficacy of several combination treatments (targeted therapy, immunotherapy or other treatments combined with chemotherapy).
Because this article involves multiple treatments and there is a lack of head-to-head research between some treatment strategies, traditional meta-analysis is not feasible.
Bayesian network meta-analysis (NMA) is proper to be used, which allows direct and indirect comparison of multiple treatment modalities [15]. And this method can combine all the data into a single analysis, which can avoid the selection bias and information loss that may exist when analyzing individual statistics separately [16]. In search of the best combination therapy, we performed a random-effect network metaanalysis of five treatments, chemotherapy or chemotherapy plus placebo (CT), chemotherapy plus one targeted drug (CT+T), chemotherapy plus two targeted drugs (CT+T+T), chemotherapy combined with immunotherapy (CT+I), and chemotherapy combined with biological therapy (CT+B, oncolytic virus or pseudomonas aeruginosa preparation).

Search strategy and selection criteria
We conducted a systematic literature search based on the guidelines of PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) [17]. All clinical studies involving multiple treatments for lung cancer from January 2008 to May reviewed all manuscripts and filtered them according to the following inclusion and exclusion criteria. Inclusion criteria: 1) Randomized controlled trial (RCT); 2) A multiarm study involving two or more of the above five treatment strategies (CT, CT+T, CT+T+T, CT+I, CT+B); 3) Subject for advanced lung cancer. Exclusion criteria: 1) Non-RCT; 2) One-arm study; 3) Non-chemotherapy-based combination therapy; 4) Comparison between the same treatment strategies; 5) Non-advanced lung cancer.

Data extraction and assessment for risk of bias
Two evaluators (Lecai Xiong, Yuquan Bai) independently screened the literatures, extracted the data, and cross-checked. If the results were not uniform, we consulted with the third evaluator (Hexiao Tang). And then disagreements were discussed and resolved with the consensus of all evaluators. We conducted a risk assessment of the final included studies through the Cochrane bias risk assessment methodology, and summarized all the treatment strategies included in the study into five categories: CT, CT+T, CT+T+T, CT+I, CT+B (one is chemotherapy combined with oncolytic virus [18] and another is chemotherapy combined with pseudomonas aeruginosa preparation [19], we collectively refer them as chemotherapy combined with biological therapy). The basic information was extracted, such as research authors, publication time, and treatment strategies in the trials. The median overall survival (OS) and progression-free survival (PFS) were included as primary outcome measures, treatment objective response rate (ORR) and adverse were used as secondary outcome measures. We only focused on grade 3-4 adverse because grade 12 had lesser clinical significance and was not consistently reported in the included trials.

Statistical process
OS, PFS, ORR and grade 3-4 adverse were used as the outcome measures. In order to balance the heterogeneity in different trials, we constructed a random effects model for meta-analysis. The network meta-analysis of OS and PFS used HRs and 95% CI data reported in the primary publications. When HRs were not reported, we estimated them from summary statistics with the method described by Tierney and colleagues [20]. We further calculated LnHR and its standard error (SE) according to the Cochrane handbook, then we corrected the LnHR and its SE in multi-arm trials [16]. In addition, there are comparisons between two or more of the same treatment strategies in some multi-arm trials [21][22][23][24]. The survival curve from the original manuscript was extracted and quantified by Engauge Digitizer 10.8 software, and then the individual patient data (IPD) was obtained through the "MASS", "splines", and "survival" R packages. By merging IPDs of the same treatment strategy, the new survival curves, HRs and its SE [25] were obtained. For the trails involving the same treatment strategy that cannot be combined, we randomly select one of the cohorts for analysis [26] [27] [28]. Finally, LnHR and its SE were used as initial data for network meta-analysis. For ORR and adverse, we calculated odds ratios (ORs) based on the number of total patients and the number of patients who responded or had an adverse reaction in each trial for metaanalysis, and merged data by direct superposition.
Bayesian network meta-analysis was done with "gemtc" package in R-3.4.3 [29][30][31]. We used non-informative uniform and normal prior distributions and three different sets of starting values to fit the model [15], yielding 20000 iterations (5000 per chain) to obtain the posterior distributions of model parameters [32]. In the OS and PFS groups, "cloglog" was used as the connection function, and in the ORR and the reverse group was "log".

Search results and assessment for the risk of bias
The literature search identified a total of 3673 articles, among which we removed 3481 articles according to the title and abstract. We reviewed the full texts of 192 studies.
Of these, studies of assessing two identical treatment strategies (n=37), nonchemotherapy-based combination therapy (n=34), no corresponding outcome indicator (n=22), non-RCT (n=14), assessing non-advanced lung cancer (n=12), assessing other treatments (n=12), one-arm study (n=10) were excluded. Finally, the remaining 51 studies were included in the analysis (Supplemental materials, Table A1). A flow chart of study selection is showed in Figure 1. We conducted a risk assessment of the included studies based on the Cochrane bias risk assessment methodology (Supplemental materials, Figure A1).

Characteristics of eligible studies
The including trials were all randomized controlled trials of Phase II/Phase III, a total of 23491 patients, all of them were patients with stage III-IV non-small cell lung cancer (NSCLC) or patients with extensive-stage small cell lung cancer (SCLC), and 42 trials were done in patients with NSCLC, 9 trials were done in patients with SCLC.
for ORR, and 20 trials for adverse were analyzed respectively (Table A1).

Data collection and calculation
The values of OS and PFS related LnHR and its SE were obtained as described in the method, merged IPDs and made correction in multi-arm trials ( Table 1). The ORR and adverse data were extracted from the included trials and summarized as Table A2.

Markov Chain Monte Carlo (MCMC)
We plotted trace plots, density plots, and Brooks-Gelman-Rubin diagnosis plots in R to assess the convergence and stability of the model. Each group of trace plots showed that each MCMC chain overlaps, and the visual perception does not recognize any one chain. All density plots showed a normal distribution and the bandwidth values approach zero. All above results prove that the model is very satisfactory [33]. In the Brooks-Gelman-Rubin diagnosis plots, the median value and the 97.5% value of the shrink factor both approached 1.0 and fit each other after 20,000 iterations, further showing that the model has good convergence [34]. The results can be found in supplemental materials ( Figure A2).

Differences in efficacy among five combination therapy strategies
We summarized the results of pairwise comparisons in the NMA as shown in Table   2. As shown in the found that the four combination therapy strategies are superior to the CT, and the best treatment strategies are still CT+T+T and CT+I in PFS subgroup analyses. But unlike the OS analysis, the progression-free survival of CT+T+T was better than CT+I, followed by CT+T and CT. In the analysis of secondary outcome indicators (ORR and adverse), the response rates from high to low were CT+I, CT+T+T, CT+T, CT+B, CT, and the incidence of 3-4 grade adverse reactions from high to low were CT+T+T, CT+B, CT+T, CT and CT+I respectively. In the Figure 3, we showed the forest plot of the OS analyses. The forest plot of PFS analyses presented in Figure A3.

Ranking of different treatment strategies
In Figure 4, the rankings of the five competing treatment strategies were summarized in terms of OS, PFS, ORR and adverse-with details provided in the supplemental materials, Table A3. CT+I and CT+T+T had similar ranking and was most likely to be ranked as the best or the second best in terms of OS, PFS, ORR. However, the incidence of adverse reaction in the CT+T+T group was significantly higher than in the CT+I group.
The two statistical analyses reached consistent results, both results indicated that CT+I was the best treatment strategy for chemotherapy-based combination therapy for advanced lung cancer.

Comparison of different immunotherapeutic drugs
To further explore the differences between the different drugs in the best combination therapy strategy (CT+I), we used OS and PFS as outcome indicators to compare them indirectly. There are 10 trails for chemotherapy combined with immunotherapy in this study, including 5662 patients. Three of them were chemotherapy plus pembrolizumab, four were chemotherapy plus ipilimumab, and one for each chemotherapy combined with TG4010, atezolizumab, durvalumab. There is no direct comparison between the drugs combined with chemotherapy. We connected them by chemotherapy alone or chemotherapy plus placebo and made indirect comparisons. Figure 5 and Figure A4 showed the forest plots for pairwise comparisons in the network of OS and PFS, and we summarized the results as Table 3. According to the results in the table, the effect of five immunotherapy drugs combined with chemotherapy are all better than chemotherapy or chemotherapy plus placebo, and chemotherapy combined with pembrolizumab is the best one (Compared with CT+ pembrolizumab, the HRs of OS of CT+ ipilimumab, CT+TG4010, CT+atezolizumab, CT+durvalumab are 1.6(95% CI, 1.1, 2.1), 1.3(95% CI, 0.80, 2.2), 1.2(95% CI, 0.74, 1.9), 1.3(95% CI, 0.79, 1.9) respectively).

Discussion
Before 2003, the treatment for patients with advanced lung cancer mainly relied on systemic chemotherapy. Although chemotherapy had a certain effect on the survival and quality of life for patients, the median survival time of patients after chemotherapy was still only 8-10 months [35]. With the development of targeted drugs, it brings a new hope to patients with advanced lung cancer. Currently, the most widely used targeted therapy is for patients with epidermal growth factor receptor (EGFR) gene mutations. Studies have shown that patients with tumors carrying EGFR mutants have a median survival time of more than 2 years after using EGFR inhibitors [9], better survival benefits to patients than chemotherapy alone. In recent years, another milestone in cancer treatment is the immunotherapy. Currently, there are two major immunotherapeutic agents, one is for the PD-1/PD-L1 pathway [36] and another is for the CTLA4 (Cytotoxic T-lymphocyte antigen 4) molecule [37]. Many studies have also shown that immunotherapy has many advantages over than traditional chemotherapy [6,23,38].
However, in-depth research on targeted therapy and immunotherapy also reveals the limitations of both. Firstly, it is the drug resistance of targeted therapy, due to longterm exposure to one targeted drug, patients will have various degrees of resistance [9].
In addition, targeted therapy is only effective in the patients with mutated genes, but such patients account for only a small fraction (about 25%) [8]. For immunotherapy, the obvious defect is the selectivity of the treatment population. Studies showed that only about 20% of patients benefit from monotherapy with immunosuppressive agents [39].
Currently, it mainly combats resistance by using drugs against secondary mutations or combination different kind of therapies [40]. Similarly, immunotherapy combined with other treatments had also shown better efficacy [41]. Combination therapy not only overcomes drug resistance, but also reduces the dose of both drugs, thereby minimizing the side effects and enhancing the therapeutic effect of a single strategy [42]. This research mainly analyzes the therapeutic effects of multiple treatments combined with chemotherapy. The efficacy among different combination treatments is still controversial [14]. Some studies have shown that low-dose erlotinib combined with cisplatin has synergistic effects in some cell lines [43]. However, more trials have shown that the combination of targeted therapy and chemotherapy not only has no survival benefits, but also aggravates adverse reactions [44,45]. Our study showed that chemotherapy combined with targeted therapy was superior to chemotherapy alone in terms of OS, PFS or ORR, but the incidence of adverse reactions did increase ( Figure   4). At present, the mechanism of combination therapy is still unclear and more research is needed to explore its effectiveness. Combination therapy with multiple targeted drugs aims at circumventing drug resistance through a so-called bypass signaling mechanism by targeting horizontal pathways, or vertical pathways, or both [46]. And some studies have confirmed that the combination of two targeted drugs can restore the sensitivity of tumor cells to drugs [47,48]. This study had the same conclusion as the previous researches, among the various combination therapies, the treatment effect of CT+T+T was second to CT+I in OS analyses ( However, the incidence of adverse reactions was the highest. Some studies have shown that chemotherapeutic drugs are likely to synergize or superimpose with immunological checkpoint inhibitors by boosting the immunosuppressive environment within tumor microenvironment [39]. The results of this study showed that CT+I was superior to other treatment strategies in terms of OS and ORR, and PFS was second only to CT+T+T. It is worth mentioning that the incidence of adverse reaction rate of CT+I was the lowest of the five treatment strategies in this study. The efficacy of immunotherapy has been confirmed in many clinical trials, and some immunotherapeutic drugs have been incorporated in the treatment strategy of advanced non-small cell lung cancer in first-and second-line setting improving the prognosis of these patients [49]. There are many types of drugs for immunotherapy, and this study contained five of them (pembrolizumab, ipilimumab, TG4010, atezolizumab, durvalumab). Due to the lack of head-to-head research, we conducted an indirect comparison with them through network analysis. The results showed that among the five immunotherapy drugs, pembrolizumab combined with chemotherapy had the best therapeutic effect (Table 3). Even not directly, but this is the first time to compare the efficacy of various immunotherapy drugs combined with chemotherapy in clinical trials, and it's very important for patients with advanced lung cancer to choose in the future treatment and for our understanding with immunotherapy.

Conclusion:
According to the results of this study, chemotherapy combined with immunotherapy is the best combination of multiple chemotherapy-based combination therapy strategies for patients with advanced lung cancer, and it is superior to other treatment strategies in terms of long-term survival and adverse reactions. Moreover, we found that pembrolizumab combined with chemotherapy has the best effect by comparing various immune drugs.
Due to the lack of direct comparative trials between different immunotherapy drugs combined with chemotherapy, we can only construct indirect comparisons for them.
There are many other combinations of combination therapy for advanced lung cancer, but this study is only based on chemotherapy-based combination therapies. Next, we will further explore the efficacy of other combination therapy strategies in patients with advanced lung cancer.