Chimeric Antigen Receptor T-cell (CAR T) Therapy for Hematologic and Solid Malignancies: Efficacy and Safety—A Systematic Review with Meta-Analysis

Chimeric antigen receptors T cells (CAR T) had been used for treating various tumor patients in clinic, and owned an incredible efficacy in part of malignancies. However, CAR T therapy remains controversial due to doubts about its efficacy and safety in the clinical treatment of various malignancies. A total of 997 tumor patients from 52 studies were included in this review. Eligible studies were searched and reviewed from the databases of PubMed, Web of Science, Wanfang and Clinicaltrials.gov. Then meta-analysis and subgroup analysis were used to investigate the overall response rate (ORR), complete response rate (CRR), common side effect rate (CSER) and relapse rate (RR) of CAR T therapy for patients in clinical researches, respectively. The results further confirmed that CAR T therapy had a higher response rate for hematologic malignancies. More importantly, CAR T therapy had a higher CSER in patients with hematologic malignancies, and it had a similar RR in patients with different malignancies. Cell cultured without the addition of IL-2 and total administration less than 108 cells were recommended. This study offers a reference for future research regarding the application in solid and hematologic malignancies, side effects and relapse, and even the production processes of CAR T cells.

Hematologic malignancies were mainly concentrated on lymphoma and leukemia, including acute lymphoblastic leukemia (ALL), splenic marginal zone lymphoma (SMZL), primary mediastinal large B-cell lymphoma (PMBCL), acute myeloid leukemia (AML), Hodgkin lymphoma (HL), mantle cell lymphoma (MCL), diffuse large B cell lymphomas (DLBCL) etc. Besides, 878 patients received chemotherapy-based lymphodepletion before CAR T therapy, 98 patients did not receive lymphodepletion, and 21 patients did not any information about lymphodepletion. Since no information on clinical response was provided for 156 patients, only 841 patients were eligible for the evaluation of response rate. And seven included studies did not offer any information about common side effects, thus 896 patients were eligible for the analysis of common side effects. Finally, 533 patients who achieved overall response after treatment were all included in the analysis of relapse.

Sources of Heterogeneity
The heterogeneity analysis results showed that heterogeneity also existed in other factors, such as cell culture with the addition of IL-2. Table 3 summarized the results of meta-regression analysis.
The result of meta-regression analysis between overall response and different factors showed that ORR presented a significant associations with "IL-2 addition" (p = 0.0176). And the result of meta-regression analysis between complete response and different factors showed that CRR had a significant associations with "total administration dose" (p = 0.0067). When "IL-2 addition" was entered as a variable to stratify the meta-analysis of ORR, the result showed that cell cultured without the addition of IL-2 had a higher ORR (67%, 95% CI: 53-79%) compared to cell cultured with IL-2 addition (37%, 95% CI: 20-58%) (Figure 5a). Then, when "Total administration dose" was used as a variable to stratify the meta-analysis of CRR, the result suggested that total administration dose less than 10 8 cells had a higher CRR (74%, 95% CI: 66-81%) compared to total administration dose more than 10 8 cells (26%, 95% CI: 16-40%) (Figure 5b).  for CRR and CI of the total administration dose more than 10 8 cells and less than 10 8 cells of each study and the overall.

Discussion
As one of the most promising treatments for tumors, two kinds of CAR T therapy (Tisagenlecleucel and Yescarta) had been approved by the Food and Drug Administration for patients with ALL and large B-cell lymphoma in 2017, which means that CAR T therapy had officially emerged as a regimen for tumor patients in the clinic. This decade, a lot of clinical trials regarding CAR T therapy were carried out around the world. As of April 19, 2018, the information from Clinicaltrails.gov showed that more than 270 CAR T-related studies were undergoing clinical trials in many different countries. Based on the clinical outcomes of CAR T therapy, several important factors were found. For example, if the patients with hematologic malignancies did not accept lymphodepletion before treatment with CAR T cells, they would suffer very serious CRS and even life-threatening events. However, there is still no systematic review of the efficacy and safety of CAR T therapy in patients with different tumors. Thus, this study fills this blank for the first time by using meta-analysis, subgroup analysis and meta-regression analysis.
By analyzing the ORR and CRR of CAR T therapy in patients with different tumors, we found that the response rate was significantly higher for patients with hematologic malignancies compared to patients with solid malignancies. Previous studies suggested that CAR T therapy still had many problems and challenges for patients with solid malignancies, such as antigen validation, tumor trafficking and infiltration, tumor heterogeneity, and an immunosuppressive microenvironment [27,66,67]. These problems and challenges led to poor response rate of patients with solid Figure 5. The forest plots of meta-regression analysis: (a) Forest plot for ORR and CI of cell culture with IL-2 addition and cell culture without IL-2 addition of each study and the overall; (b) Forest plot for CRR and CI of the total administration dose more than 10 8 cells and less than 10 8 cells of each study and the overall. Table 3. The result of meta-regression analysis about the association between the characteristic of cohorts and the difference rates included in this studies.

Discussion
As one of the most promising treatments for tumors, two kinds of CAR T therapy (Tisagenlecleucel and Yescarta) had been approved by the Food and Drug Administration for patients with ALL and large B-cell lymphoma in 2017, which means that CAR T therapy had officially emerged as a regimen for tumor patients in the clinic. This decade, a lot of clinical trials regarding CAR T therapy were carried out around the world. As of April 19, 2018, the information from Clinicaltrails.gov showed that more than 270 CAR T-related studies were undergoing clinical trials in many different countries. Based on the clinical outcomes of CAR T therapy, several important factors were found. For example, if the patients with hematologic malignancies did not accept lymphodepletion before treatment with CAR T cells, they would suffer very serious CRS and even life-threatening events. However, there is still no systematic review of the efficacy and safety of CAR T therapy in patients with different tumors. Thus, this study fills this blank for the first time by using meta-analysis, subgroup analysis and meta-regression analysis.
By analyzing the ORR and CRR of CAR T therapy in patients with different tumors, we found that the response rate was significantly higher for patients with hematologic malignancies compared to patients with solid malignancies. Previous studies suggested that CAR T therapy still had many problems and challenges for patients with solid malignancies, such as antigen validation, tumor trafficking and infiltration, tumor heterogeneity, and an immunosuppressive microenvironment [27,66,67]. These problems and challenges led to poor response rate of patients with solid malignancies, and our results further confirmed it. Besides, this study analyzed the ORR and CRR of CAR T therapy in patients with different B-cell malignancies by using meta-analysis and subgroup analysis. The result confirmed that CAR T therapy had a great response rate for patients with ALL, NHL and CLL, but it had poor response rate for patients with HL.
The common side effects that caused by CAR T cells included CRS, NS and TLS. The main symptom of CRS included fever, myalgia, headache, anorexia, nausea and vomiting, renal dysfunction, coagulopathy, hypotension, capillary, leak, and pulmonary edema. The main symptom of NS included confusion, B-cell aphasia, unresponsive-ness, and seizures. And the main symptom of TLS included hyperuricemia, hyperkalemia, hyperphosphatemia, and hypocalcemia [11]. The result regarding CSER of CAR T therapy in patients with solid and hematologic malignancies suggested that the CRS rate was significantly higher for patients with hematologic malignancies compared to patients with solid malignancies, and the NS rate was slightly higher for patients with hematologic malignancies compared to patients with solid malignancies. These results confirmed that the patients with hematologic malignancies could achieve a better response rate in treatment of CAR T therapy, but they also need to take a higher risk of common side effects at the same time. More importantly, the result about RR of CAR T therapy showed that the RR of patients with hematologic malignancies was similar to patients with solid malignancies, which means that no matter the patients with solid or hematologic malignancies, they had 25-30% RR in the end. This study firstly found that different types of tumors, different CAR T protocols and even different response rate had a similar RR in the end. This finding was attributed to the inclusion of various types of tumors in the same meta-analysis.
The result of meta-regression analysis indicated that the response rate of CAR T therapy had association with "cell culture with IL-2 addition" and "total administration". Previous study recommended that cytokines could be used in improving the expansion of first generation CARs, and it had potentially benefit for second and third generation CARs. IL-2 addition can promote the expansion of CAR T cells in vitro as a cytokine support. Zhang et al. [2] analyzed the sources of heterogeneity of CAR T therapy for patients with hematologic malignancies by using meta-regression analysis, and they found that cell cultured without IL-2 addition associated with better clinical response. However, the included studies of Zhang et al. were all published before 2015, and some studies had a same author, thus their results had some limitations. The results of this study further indicate that cell culture without IL-2 addition was associated with higher clinical response rates in patients with solid and hematologic malignancies. Then, our result also confirmed that response rate was not associated with different generation CARs, thus the need for the addition of IL-2 to culture cells remains to be studied and verified. "Total administration dose" and "persistence" in vivo were thought as two key factors for response rate and RR of CAR T therapy, respectively. Our results further confirmed that "total administration dose" had an association with response rate, but it suggested that total administration dose less than 10 8 cells led to a better response rate, which was contrary to our long-standing perception of CAR T therapy. Furthermore, previous study [39] had confirmed that too short persistence (less than two months) could make CAR T cells no longer effective in vivo and lead to relapse. However, our result indicated that RR was not associated with "persistence". That was the reason why we included such various CAR T protocols in the same meta-analysis, based on the meta-regression analysis of these different factors, we could identify potential factors that influenced the clinical outcomes of CAR T cells and offer a reference for the future research of CAR T therapy.
Although the subgroup analysis and meta-regression analysis were used to discuss the factors of heterogeneity in this study, this research still had some limitations. One of the most important limitations was that there are fewer patients with solid tumors, and the other limitations included ages, races, countries, etc. These limitations are inevitable and unovercomable in the present data or available data. However, the publishing time of the included studies were close (2013-2018), and the most studies were published at 2016 to 2018, then CAR T cells were almost autologous T cells. To evaluate the efficacy and safety, we tried to make a systematic analysis for CAR T therapy in all clinical studies. The publication bias main focus on index text and patient selection. Since our study included such various tumors, it was difficult to consider the identified reference standard as the best reference standard or correctly classify. CAR T therapy needs a process of production of T cells, and it resulted that some patients had not enough CAR T cells to start the treatment.

Search Strategy
Firstly we searched the meta-analysis regarding CAR T cells for patients with hematologic and solid malignancies. No same articles were found. Secondly, we searched studies and clinical data from the databases of PubMed, Web of Science, Wanfang and Clinicaltrials.gov. The search terms "chimeric antigen receptors", "CAR T" and "clinic" were used in the process of search. There were no language limitations. "Full text", "article", "clinical research" and "completed" were used to filter articles and clinical data. Finally, the reference lists of primary studies were reviewed by two authors. The latest search happened on April 19, 2018 [68].

Selection Criteria for Considering Studies for This Review
All patients from included studies suffered from different hematologic or solid malignancies, and they agreed to participate in an experimental study about the efficacy and safety of treatment with CAR T cells. We excluded the following studies: (1) non-clinical research, (2) research with no clinical outcomes, (3) multiple publications of a same study, (4) graduation theses, editorials, abstracts and letters, and (5) uncompleted studies [68].

Initial Review of Studies
The initial database was compiled, and all duplicate articles were eliminated. We screened these citations depending on the title, abstract and the relevant studies for inclusion based on the criteria identified previously. Only after assessment of the full-text articles by two authors, the studies were finally selected for inclusion in the review. Any disagreement was resolved by discussion between two authors [68].

Data Abstraction
The data of initial review were recorded on a standard data extraction form by both authors independently. The name of the first author, publication year, the number of patients, types of CAR T cells, vector, original T cell sources, gene transduction method, T cell activate method, T cell culture times, CAR T cell persistence times in vivo, the types of malignancies, the method of lymphodepletion, total administration doses, clinical response and the side effects associated with CAR T therapy were all collected by two authors in each study.

Assessment of Study Quality and Risk of Bias
The two authors independently assessed the quality of the studies (risk and bias) by using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool [69]. According to the QUADAS-2 user guidelines [70], items were modified for this study. In domain 1 (Patient selection), the item "Was a case-control design avoided?" was omitted. In domain 2 (Index test), the items "Were the index test results interpreted without knowledge of the results of the reference standard?" and "If a threshold was used, was it pre-specified?" were substituted for the item "Was the method for determining the outcomes of patients after administration described?" In domain 3 (Reference standard), the items "Were the reference standard results interpreted without knowledge of the results of the index test?" was omitted. In domain 4 (Flow and timing), the item "Was there an appropriate interval between index test and reference standard?" was omitted, and the item "Did all patients receive the same reference standard?" and "Were all patients included in the analysis?" were replaced by the item "Did all patients accept the treatment of CAR T cells?" Based on the QUADAS-2 guidelines, researches and data were assessed for each item according to the following rating scale: high risk of bias, low risk of bias, or unclear. Any disagreement was resolved by discussion between two authors.

Statistical Analysis
"Meta package" is a statistical tool for meta-analysis in R software, it can be used to enhance the functionality of R software in the meta-analysis. R software has a special "Meta package" (named "Metaprop") for rate meta-analysis. "Metaprop" realized some procedures of special binomial data, it allowed the computation of exact binomial and score test that based on CI, which means that the proportions of closing to 0 or 100% could be included from the meta-analysis, and it was used in this study. Then, the Logit transformation was used to compute the overall pooled response rate, CSER and RR.
Cochran's Q test and Higgins' I 2 statistics were used to make homogeneity test for eligible studies. A p-value ≤ 0.1 and/or I 2 ≥ 50% indicated significant heterogeneity, the data should be calculated by the random-effect model. Accordingly A p-value > 0.1 and/or I 2 < 50% indicated significant homogeneity, the data should be calculated by fixed-effect model. [68] The subgroup analysis of hematologic and solid malignancies was used in all result of meta-analysis. Potential heterogeneity factors of each analysis, including "Generation of CARs", "Vector", "Cell culture time", "Transfection method", "IL-2 addition", "Persistence", "Lymphodepletion" and "Total administration dose", were assessed by meta-regression analysis based on the Cox proportional hazards regression model. All analysis were performed by using Review Manager 5.3 (The Cochrane Collaboration, Copenhagen, Denmark), R software 3.5.0 (R Foundation for Statistical Computing, Vienna, Austria) and SPSS Statistics 22 (IBM, Armonk, NY, USA).

Conclusions
This study used meta-analysis to further confirm that CAR T therapy had a higher ORR and CRR in patients with hematologic malignancies compared to patients with solid malignancies, especially in patients with ALL, NHL and CLL. More importantly, this study confirmed that the patients with hematologic malignancies had a higher CSER, and patients with hematologic and solid malignancies had a similar RR in the end. The meta-regression analysis results suggest that cell cultured without IL-2 addition and total administration less than 10 8 cells were recommended for manufacture of CAR T cells and clinical treatment, respectively. These results may make CAR T therapy becoming cheaper, because the cost of IL-2 could be saved, and the cost of CAR T cells amplification could be appropriately reduced. The number of patients who were unable to be treated due to not having enough CAR T cells would also be decreased. We believe that CAR T therapy could play an increasingly important role in tumor treatment in the future.