Evaluation of effectiveness of granulocyte-macrophage colony-stimulating factor therapy to cancer patients after chemotherapy: a meta-analysis

The impact of granulocyte-macrophage colony stimulating factor (GM-CSF) on hematologic indexes and complications remains existing contradictory evidence in cancer patients after treatment of chemotherapy. Eligible studies up to March 2017 were searched and reviewed from PubMed and Wanfang databases. Totally 1043 cancer patients from 15 studies were included in our research. The result indicated that GM-CSF could significantly improve white blood cells count (SMD = 1.16, 95% CI: 0.71 – 1.61, Z = 5.03, P < 0.00001) and reduce the time to leukopenia recovery (SMD = -0.85, 95% CI: -1.16 – -0.54, Z = 5.38, P < 0.00001) in cancer patients after treatment of chemotherapy. It also could improve absolute neutrophil count (SMD = 1.11, 95% CI: 0.39 – 1.82, Z = 3.04, P = 0.002) and significantly shorten the time to neutropenia recovery (SMD = -1.47, 95% CI: -2.20 – -1.75, Z = 3.99, P < 0.0001). However, GM-CSF could not improve blood platelet (SMD = 0.46, 95% CI: -0.37 – -1.29, Z = 1.10, P = 0.27). And GM-CSF had significant connection with fever (RR = 3.44, 95% CI: 1.43 – 8.28, Z = 2.76, P = 0.006). The publication bias existed in the data of the impact of GM-CSF on blood platelet and complication. In conclusions, GM-CSF had an intimate association with some hematologic indexes and complications. Our study suggested that more hematological indexes and even more other indexes need to be observed in future studies.


INTRODUCTION
Chemotherapy has been widely used for treating different cancers for many years. Whereas chemotherapy can lead to many adverse drug reactions, such as hematologic toxicity, thromboembolism, and neurotoxicity. [1][2][3]. Hematologic toxicity is a common adverse reaction such as neutropenia and leukopenia. It not only delays the time of next therapy but also leads to the life-threatening events (such as severe infection, bleeding, copper deficiency and protein malnutrition) if decreased blood cells had not been managed timely [4,5]. The previous studies had confirmed that chemotherapyinduced neutropenia or leukopenia could reduce survival rates in patients with advanced cancer [3,6]. Thus, the hematologic indexes, including white blood cells (WBC) count, absolute neutrophil count (ANC), blood platelet (PLT) count and monocytes count, are important objective indexes in cancer patients after chemotherapy.
GM-CSF is the cytokine most extensively used as hemopoietin in the clinical practice, it can promote the activation, proliferation, and differentiation of myeloid www.oncotarget.com Oncotarget, 2018, Vol. 9, (No. 46), pp: 28226-28239 Meta-Analysis www.oncotarget.com precursor cells in the body. Beside GM-CSF plays an important role in the recruitment, development, and maturation of dendritic cells, which are necessary for the subsequent T helper cell type 1 and cytotoxic T lymphocyte activation. GM-CSF also can enhance the function of mature granulocytes and mononuclear phagocytes [5,[7][8][9]. In the 1990s, GM-CSF was approved by Food and Drug Administration (FDA) as a drug for treatment of older adults with acute myeloid leukemia, after induction chemotherapy, to shorten the time to neutrophil recovery and to reduce the incidence of life-threatening infections. Thus GM-CSF is often used to treat chemotherapy-induced neutropenia and leucopenia [8,10]. GM-CSF was also approved by FDA for myeloid reconstitution following allogeneic bone marrow transplantation (BMT), autologous BMT or peripheral blood stem cell transplantation. Furthermore, GM-CSF was approved for peripheral blood stem cell mobilization, BMT failure, and engraftment delay. GM-CSF also has many proinflammatory functions [11]. The trade name of GM-CSF includes sargramostim, leucomax (American name), "Te Li Er", "Ge Ning" (Chinese name), and molgramostim (Indian name).
GM-CSF has been used to treat chemotherapyincluded neutropenia and leucopenia for many years, and many different clinical responses and data are accumulated. However, these clinical data and responses, including the impact of GM-CSF on WBC count, ANC, BLT, and complications, still lack systematic analysis and evaluation. Hence this study aims to quantify the data from previous clinical studies via meta-analysis, for systematically analyzing and evaluating the clinical impact of GM-CSF on hematologic indexes and complications.

Complications
Because of individual differences, different patients often had different degrees of complications after treatment with GM-CSF, such as fever, local reactions, bone and muscle pain. Except for fever, most of the complications could be tolerated and do not lead to serious consequences. Ten studies [12, 14-19, 21, 23, 25, 26] provide the data of fever, and 805 patients were included.
Homogeneity test showed that there is no homogeneity in these studies (χ 2 = 34.98, P < 0.0001, I 2 = 74%, Figure 5). Hence random-effect model was used to calculate these

Publication bias
As shown in Figure 6, the funnel plots for evaluating the publication bias for the impact of GM-CSF on hematologic index and complications. Open circles represent studies included in the meta-analysis. The perpendicular in the center indicates the summary proportion. And the other two dotted lines represent the 95% CI. On the visual assessment of funnel plot, there is no evidence of publication bias was revealed on the association between GM-CSF with WBC count, the time to neutropenia recovery and leucopenia recovery ( Figure  6A, 6B and 6D). However, the publication bias existed on the studies regarding the impact of GM-CSF on ANC, PLT count, and complications.

DISCUSSION
Recent studies regarding chemotherapeutic agents show that targeted therapies have received  increasing attention [27][28][29][30]. However chemotherapyincluded hematologic toxicity still exists. It not only increases pain to the patient, but also reduce medication adherence, delay the treatment cycle, and even lead to life-threatening events. Thus it is important for the patient to control hematologic toxicity after treatment with chemotherapy.

Registration 23
Registration  Since GM-CSF and granulocyte colony stimulating factor (G-CSF) were approved by FDA in the 1990s, they were both widely used to treat various types of cancer patients with chemotherapy-induced neutropenia and leucopenia, and achieved an excellent result. Because of the outstanding clinical efficacy of GM-CSF and G-CSF, they have made hundreds of millions of dollars each year in sales. G-CSF, as a member of cytokine families, is often used to promote the production of granulocytes or antigen presenting cells. And the clinical efficacy of G-CSF has been systematically evaluated and analyzed in many studies [10,[37][38][39]. However, as a cytokine that has a similar efficacy to G-CSF, the clinical efficacy of GM-CSF is lack of systematic analysis and evaluation. Thus this study fills the gap. The result of our study confirmed that GM-CSF can help cancer patient after treatment of chemotherapy quickly improving WBC count and ANC. And GM-CSF has no association with PLT count, however, the result regarding GM-CSF and PLT from A. Le Cesne et al. [16] were that GM-CSF could significantly improve the PLT count in cancer patients. Besides, GM-CSF has a significant connection with fever.
Although the included data come from different countries, including Italia, Canada, France, America, and China. The drugs (GM-CSF) that they used were manufactured in different places and even have different trade names. The types of patients were all cancer patients after treatment of chemotherapy, administration route were all intravenous injection, administration time was all at least 24 hours after treatment of chemotherapy, and clinical observation index was all hematologic index. moreover, the clinical response rate in these studies was similar (73.7%-97%) except A. Le Cesne et al. [16] (21.3%). Besides, three studies were published by one first     [18,23,26], which means that Zhou was a specialist in this field. And these three studies used three different brands of GM-CSF to treat different patients at the same dosage and duration. It not only avoided the difference in the research system but also studied drug standardization of GM-CSF.
Our study also exists some limitations, the main limitations are sample size and the publication time of studies. According to the data needs of this study, the result of the search was so. The method and the duration of GM-CSF administration still keep the same. The publication bias is not robust. And the data of statistical analysis were expressed in median and range. That was the reason why we did this study. We tried to find the connections between GM-CSF and hematologic index by analyzing available data, which can help the doctor better observe the efficacy of GM-CSF at the treatment stage, rather than waiting until the end of treatment.
The potential heterogeneity come from patients, there are over 30 cancer types with the total patients number, including lung cancer, breast cancer, lymphoma etc. And they were treated with different chemotherapy. However, firstly when GM-CSF is clinically used for chemotherapy-induced neutropenia and leucopenia, it is a broad-spectrum drug that does not target a particular tumor. GM-CSF is used simultaneously with chemotherapy drugs. The rapid proliferation and differentiation of myeloid precursor cells are sensitive to chemotherapeutic drugs, thus affecting the efficacy of chemotherapy. Secondly, the mechanism of GM-CSF therapy is actually immunotherapy. Thus the heterogeneity in immunity of patients is more important than cancer types. Unfortunately the data included in our metaanalysis lack the immunity information of patients.
In conclusion, according to the results of this study, we found that an included study [16] shown a significant difference from the other studies in two results: the impact of GM-CSF on PLT count and the clinical response rate of GM-CSF, respectively. The most likely reason for this is sources of medication, however, we cannot rule out other possible causes. Thus we believed that only the index of WBC count and ANC were used to judge the outcome of GM-CSF treatment in previous studies, which may not completely reflect the efficacy. We suggested that hematological indexes need to be expanded in future studies. Further, the other indexes should also be observed, such as immunological indexes.

Search strategy
Firstly we searched the systematic reviews and meta-analysis on the role of GM-CSF in cancer patient after treatment of chemotherapy. No reviews or meta-analysis were found. Secondly, we searched the PubMed and Wanfang databases for studies reporting the use of GM-CSF in cancer patient after treatment of chemotherapy. And the following keywords were used in the process of search, respectively: GM-CSF cancer therapy clinic, GM-CSF white blood cell, sargramostim, leucomax and molgramostim. The languages of articles were limited in English and Chinese. "Full text" and "Human" were used to filter article. Finally, the reference lists of primary studies were reviewed by two authors. The latest search happened on March 9, 2017.

Selection criteria for considering studies for this review
The patients from included studies were diagnosed with various forms of cancer, and all of them received systemic chemotherapy. At the same time, they agreed to participate in an experimental study of the impact of GM-CSF on hematologic indexes and complications. Then we excluded the following studies: (1) editorials, abstract and letters, (2) studies in which the clinical outcome was not reported, (3) single-patient case report or short case series involving fewer than four patients (4) multiple publications from the same study to prevent erroneous patient count and (5) graduation theses.

Initial review of studies
The initial database was compiled, and all duplicate articles were eliminated. We screened these citations depending on 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.

Data abstraction
The data of initial review were recorded on a standard data extraction form by both the authors independently. The following items were extracted: (1) publication details (title, authors, years and other citation details, including geographic locale); (2) type of study (clinical trial or review); (3) treatment details, (chemotherapy or radiotherapy); (4) patients details (number, sex ratio, age, weight, KPS); (5) dosage and duration of GM-CSF; (6) sources of medication; and (7) clinical response and toxic or side effects, if any.

Assessment of study quality
The CONSORT 2010 checklist (25 items) was used to assess the reporting quality of eligible studies. According to this checklist, two authors assessed the reporting quality, respectively. Every 25 items had one common criterion, including "yes", "no" and "I". They www.oncotarget.com represented "adequately reported", "Not reported" and "partially reported, but insufficient", respectively. Besides, "unable to judge" represented if there was lacking information regarding the item. Any disagreement was resolved by discussion between the two authors.

Statistical analysis
Cochran's Q test and Higgins' I 2 statistics were used to making 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 [40]. Standard mean differences (SMD) and pooled risk ratios (RR) with 95% confidence intervals (95% CI) were used to evaluate the association of GM-CSF with hematologic indexes and complications, respectively. The Z test was used in effect comparison between experimental and control. A P-value < 0.05 in the Z test for pooled RR, or no overlap of the 95% CI with 1 was considered statistically significant. And the publication bias was assessed by visual inspection of funnel plots. All data analyses were performed by using Review Manager 5.3 (Cochrane Collaboration, London, UK) and adhered to the PRISMA guidelines.

CONFLICTS OF INTEREST
All the authors declare to have no conflicts of interest.