Evaluation of efficacy and safety for Brucea javanica oil emulsion in the control of the malignant pleural effusions via thoracic perfusion

We searched and identified relevant RCTs from the databases of MEDLINE/PubMed, EMBASE, Cochrance Library, Web of Science, and CNKI database (from January 2000 to April 2017). The key words applied in the search were as followed: “malignant pleural effusion”, “MPE”, “Brucea javanica oil emulsion”, “BJOE injection”, “BJOEI,” “BJOE,” “Yadanzi”, and “chemotherapy”, “Brucea javanica oil emulsion injection,” “Yadanzi injection,” and “Ya-dan-zi injection.” In addition, if we find that the references of the included studies are closely related to BJOE, we should further search and identify them. The retrieved studies were regarded as potential source and reviewed manually. Moreover, although the published year of these literatures were unlimited, only English and Chinese literatures were involved in this study.

The general data that we selected are as follows: (1) the publication date of each randomized controlled trial; (2) the number of patients included in each study and grouping; (3) the clinical and pathologic features of patients included each study, (4) the patterns of treatment intervention for treating MPE; (5) trials design and implementation. The data on outcomes in present meta-analysis included clinical efficacy, QOL, and adverse effects (AEs) according to World Health Organization (WHO) criteria and Response Evaluation Criteria in Solid Tumors (RECIST). The tumor response included complete response (CR), partial response (PR), stable disease (SD), and progressive disease (PD). The overall response rate (ORR) was defined as CR + PR/overall cases and disease control rate (DCR) was calculated as CR + PR+ SD /overall cases. Toxicity was graded from 0 to IV in severity on the basis of the WHO Recommendations. This meta-analysis only investigated the incidence of Grade II or above.

Inclusion criteria: (1) study design was confined to RCTs on comparing traditional chemotherapy drugs plus BJOE with chemotherapy drugs alone for treating the MPE; (2) study subjects with MPE must be diagnosed pathologically and (or) cytologically; (3) drugs must be administered by intraluminal injection; (4) outcome measures determined by WHO criteria or RECIST, improvement of QOL evaluated by Karnofsky score (KPS), and AEs assessed by WHO Recommendations for Grading of Acute and Subacute Toxicity must be showed and (6) the sample size of the study must be more than or equal to 60.

The following criteria were used for the literature exclusion: (1) animal experiments, review, and other irrelevant studies; (2) patient also received other medications; (3) non-RCTs studies; (4) no detailed data about ORR, DCR, evaluation of QOL, and AEs or no indicators for them; (5) investigations were supported by drug producers; (6) lack of comparable control group and (7) single-arm study.

The test design must meet the following rules: (1) RCTs of traditional chemotherapy drugs plus BJOE versus traditional chemotherapy drugs alone via intrapleural injection for controlling MPE; (2) the dosage of BJOE was determined by the suggestions of producers; (3) dosing interval: once a week; (3) number of times of administration: more than or equal to 2 times; and (3) observations on efficacy and safety: ORR, DCR, QOL, and AEs.

The criteria of assessment that provided by Cochrane Handbook was employed to evaluate the quality of included investigations. It contained the following Items: (1) sequence generation; (2) how to carry out blinding; (3) how to carry out allocation concealment; (4) how to perform outcome data selective; (5) a description of intention to treat and (6) other sources. According to the above criterion, the quality of trials was defined into three levels: low risk of bias, unclear risk of bias, and high risk of bias [7].

All of the data was calculated by 14.0 (Stata Corporation, TX, USA) software package and Review Manager 5.3 software. The odds ratio (OR) with 95% confidence intervals (CI) was applied to analyze the dichotomous data [6]. By calculating the Z-value of the chi-square test, the statistical p-value < 0.05 was considered to be significantly different. The fixed effect model and the random effect model are commonly used statistical models for meta-analysis. According to the presence or absence of heterogeneity, both were selected to measure the safety and efficacy of BJOE pleural perfusion in the treatment of MPE. The χ² statistic and the I² statistic tests were used to assess statistical heterogeneity among included studies [3]. A more common way to indicate the degree of heterogeneity is the statistical test, which is often described as the Cochran chi-square test. A p value is often cited as an indicator of the degree of variability in the study. If the P value is less than 0.05, no statistical difference is considered, suggesting that the heterogeneity is small. The I² value describes the percentage of variability in point estimates that is due to heterogeneity rather than sampling error, may be readily calculated from most published meta-analyses, and a closed form uncertainty interval is available. If the I² value is less than 50%, the heterogeneity of the study is considered acceptable. If no heterogeneity existed, the method of fixed effects model was adopted, or using the random effects model. To assess the impact of a single study on overall statistical performance, we removed each study from the estimated library one by one, to analyze the impact of each study on overall effectiveness [2]. Further, we employed Begg’s funnel plot and Egger’s test to test the publication bias [7]. The SPSS (version 19.0, Chicago, USA) software was employed to finish the statistics of varying variables. The statistical p-value < 0.05 was considered to be significantly different.

Originally, we conducted the systematic research from online database, and 96 potentially relevant references were yielded. Of them, 25 studies were discarded because the design and implementation of these studies are not eligible for our research analysis. After further screening and eligibility assessment, 29 trials were excluded because some of them were not RCTs and others did not belong to first-hand research data such as summary of meetings, medical reviews and newsletters. Remaining 42 studies seemed to meet the inclusion criterion, but 28 studies were deleted because of the following reasons: repeated data reporting, animal studies, statistical irregularities and too little sample size. Finally, 14 trials were selected as appropriate for inclusion in this meta-analysis. The flow chart showing the selection process was presented in Fig. 1A.

The 14 selected trials [13‐26] were all RCTs and conducted in China. The qualified 14 studies included a total of 1085 patients, the total number of samples included in the study was from 60 [13, 26] to 123 [16] patients. The volume of pleural effusion of all patients in the amount were all more than 1000 mL and patients’ age varied from 25 [17] to 86 [22] years. From these studies, lung cancer and breast cancer were the most common cause of MPE. A detailed database for meta-analysis on general characteristics was listed in Table 1.

We found that the number of males (569) was more than the females (546) in the BJOE combined group and control group, respectively. The design of 12 studies were that BJOE combined with cisplatin versus cisplatin alone through thoracic perfusion for treating MPE [13‐18, 20, 22‐26], one study was BJOE combined with bleomycin versus bleomycin alone [21], another was BJOE combined with oxaliplatin versus oxaliplatin alone [19]. The dosages of BJOE via thoracic perfusion and follow-up times for efficacy evaluation had a good consistency, which was shown in Table 2. All studies had a certain tumor diagnosis by pathology or pleural effusion cytology diagnosis, and KPS score of each patient greater than 50 points. Generally, the dosage of BJOE was administered at the range of 40-100 mg per one time and frequency of administration was two times at least, which were given by thoracic perfusion after drainage of pleural effusions. There was no significant difference between the two groups in the general data (p > 0.05), indicating that they had good comparability.

Two investigators of us independently reviewed and assessed the quality of each study according to the criteria shaped by the Cochrane Handbook, which was specialized in evaluating the systematic reviews of Interventions (Version 5.0.1) [3]. As shown in Table 3, we found that 8 of the 14 studies (57.1%) showed the low risk of bias [17, 20‐26] and that the remaining 6 investigations [13‐16, 18, 19] displayed the unclear risk of bias (42.9%) (Table 3, Fig. 1B, C). We conducted a heterogeneity analysis of included studies. The results showed that chi-squared was 1.61 (Degrees of freedom = 13; p = 1.000) and that the value of I-squared (variation in OR attributable to heterogeneity) showed as 0.0%. These results indicated that these included RCTs had very good homogeneity. Combining the clinical information of these studies, we believe that these studies have very good comparability. Based on no heterogeneity, we completed the subsequent statistical analysis using the fixed effects model.

As shown in Table 4, all of fourteen RCTs [13‐26] in this meta-analysis showed the data on comparison of ORR between traditional chemotherapy drugs plus BJOE versus traditional chemotherapy drugs alone via intrapleural injection for controlling MPE. Via the fixed effects model analysis, we found that odds ratio was 1.39 (95% CI 1.15 to 1.67; Z value = 3.46, p = 0.001), which suggested that the ORR of traditional chemotherapy drugs plus BJOE was remarkably higher than that of traditional chemotherapy drugs alone (Fig. 2A). In addition, fourteen studies [13‐26] showed the data about DCR and displayed that the BJOE combination arms and chemotherapeutic agents single group had the same DCR rate (odds ratio = 1.04, 95% CI 0.888 to 1.23; test for overall effect: Z = 0.44, p = 0.663).

As shown in Table 4, a total of 8 trials [14, 15, 17, 20, 21, 23, 25, 26] provided the data on comparing the QOL between the BJOE plus traditional chemotherapy drugs versus traditional chemotherapy drugs alone via intrapleural injection for controlling MPE. The QOL improvement was evaluated by the KPS score of patient. After treatment, the KPS score increased by ≥10 points was defined as improvement of QOL. We found that the improvement rate of BJOE combined perfusion group (262/324, 80.86%) was significantly higher than that of chemotherapy group alone (168/319, 52.66%). The results of meta-analysis showed that the odds ratio ranged from 1.17 to 1.99 and the pooled odds ratio in this analysis displayed a value of 1.56 (95% CI 1.21 to 2.00; Z = 3.49, p < 0.001), which suggested that BJOE combined with chemotherapeutic agents significantly improved the QOL of patients with MPE, as compared with single chemotherapeutic agents (Fig. 2B).

Nine [14, 15, 17, 18, 20, 23‐25] of 14 studies compared the AEs on traditional chemotherapy drugs plus BJOE versus traditional chemotherapy drugs alone via intrapleural injection for controlling MPE. As shown in Table 5, the most common AEs in combined group and single group were myelotoxicity (71/363, 19.5% versus 141/350, 40.3%), nausea/vomiting (55/392, 14.03% versus 104/386, 28.4%), liver and renal injury (7/124, 5.6% versus 10/124, 8.1%), chest pain (39/299, 13.04% versus 52/299, 17.39%) and fever (30/309, 9.7% versus 45/305, 14.75%).

Nine [14, 15, 17, 18, 20, 23‐25] of 14 studie compared the myelotoxicity on traditional chemotherapy drugs plus BJOE versus traditional chemotherapy drugs alone via intrapleural injection for controlling MPE, we found that the incidence rate of myelotoxicity in BJOE combined perfusion group was lower than that of traditional chemotherapy drugs alone group (odds ratio = 0.5, 95% CI 0. 36 to 0.70, p < 0.001) (Fig. 3A). Ten [14, 15, 17, 18, 20, 21, 23‐25] of 14 studie compared the gastrointestinal reactions, the incidence rate of nausea/vomiting in BJOE combined group also showed a significant decrease compared with traditional chemotherapy drugs alone group (odds ratio = 0.50, 95% CI 0. 35 to 0.72, p < 0.001) (Fig. 3B). In addition, three studies compared liver and renal injury (odds ratio = 0.70, 95% CI 0. 25 to 1.91, p = 0.483), eight studies compared the incidence of chest pain (odds ratio = 0.70, 95% CI 0. 44 to 1.11, p = 0.130), and seven studies compared the incidence of fever (odds ratio = 0.67, 95% CI 0. 41 to 1.10, p = 0.111). However, these results suggested that the incidence rate of these AEs did not have differences between both of two projects (p > 0.05) (Fig. 4A-C).

Through removing each study, a further meta-analysis was performed to compare with previous results of meta-analysis to explore whether the deleted study have an certain impact on the overall statistical effect [27]. Sensitivity analysis shows that excluding of any study could not change the overall statistical effect, nor could it affect the final statistical conclusion, with an OR pool oscillating between 1.08 and 1.63 (Fig. 5A). We also drew a funnel plot of included studies and noticed that the included studies are symmetrically distributed on both sides of the funnel (Fig. 5B). In addition, for comparing traditional chemotherapy drugs with chemotherapy drugs plus BJOE for controlling MPE, we performed the Egger’s test and the results were as the following: t = 1.75 with 13 d.f, p = 0.105 (Fig. 5C). We also performed a Begg’s test and the test results showed that Std. Dev. of Score was 18.27, p value was 0.112 (Fig. 5D). Come together, the funnel plot, Egger’s tests and the Begg’s test all suggested that publication biases did not have a significant influence on the results.