Real world analysis on renal impairment of elemene emulsion injection based on propensity score matching

LEI Chao, ZHAO Xiao-xiao, YU Xiao-kang, XIE Yan-ming, WANG Zhi-fei?

1. Institute of Basic Research in Clinical Medicine,China Academy of Chinese Medical Sciences,Beijing 100700,China

2. School of Statistics,Renmin University,Beijing 100872,China

Keywords:Real world Elemene emulsion injection Renal impairment Propensity score

ABSTRACT

1. Introduction

Elemene Emulsion Injection (EEI) is an emulsion injection made from Chinese traditional Chinese medicine wenyujin (wenzedoary turmeric) as raw material and extracted its effective anticancer components by modern science and technology. The drug is a natural anti-tumor drug, and its main components are β- Elemene[1,2]. EEI can play an anti-tumor role by inhibiting the growth,invasion and metastasis of cancer cells [3]. For example, EEI has a good therapeutic effect on lung cancer, breast cancer and prostate cancer [4-6]. However, with the wide application of EEI, the clinical problems and deficiencies are also highlighted. At present, most of the reports of adverse reactions to EEI are clinical case reports [7-9],and the main adverse reactions involved are phlebitis, fever, pain,etc[10]. There is little analysis based on large samples of real-world data, and there is a lack of research and report on the impact of realworld data on renal function.

At present, research based on the real world has been widely used to evaluate the safety of clinical application of Chinese patent medicines [11], so as to provide evidence of adverse events / adverse reactions of Chinese patent medicines after listing [12], so as to provide evidence reference for clinical decision-making and drug supervision.

In this study, based on the real-world data, we used the nested case study method to take the people who used EEI but did not have renal function injury as the control group to explore the correlation between combined drugs, combined diseases and renal function injury using EEI. Because the previous analysis data showed that there was no significant correlation between combined diseases and renal function injury, Therefore, this paper mainly explores the main influencing factors of EEI combined medication on renal function injury, in order to provide evidence support for rational and safe medication and clinical early prevention.

2. Materials and methods

2.1 Data sources

Based on 442 1 patients using Elemene Emulsion Injection in 21 hospitals across the country. The selection steps of patient data are as follows: first, blood creatinine and blood urea nitrogen are selected as physical and chemical indexes to judge renal function injury; Second, the physicochemical index value of the latest laboratory test within 7 days before drug administration is defined as "physicochemical index value before drug administration", and the physicochemical index value of the latest laboratory test after Drug Administration (which must also be carried out within 7 days after Drug Administration) is defined as "physicochemical index value after drug administration"; Third, compare the physical and chemical indexes before and after treatment; Fourth, the changes of physical and chemical indexes of renal function injury include the following two situations: patients who are normal before medication,abnormal after medication, or abnormal before medication and more abnormal after medication; The specific standard for judging renal function injury is that either blood creatinine or blood urea nitrogen is 1.1 times higher than the upper limit of normal [13,14]. According to this standard, a total of 19 patients with renal function injury were screened.

2.2 Data standardization processing

A unified method is adopted to standardize the extracted information for statistical analysis. Standardization of Western medicine diagnosis information and doctor's order information:extract the information of combined medication, convert the western medicine recorded as commodity name in the database into general name and merge the same items, merge the Chinese patent medicines with similar drug components but different dosage forms, and retain the original name of other Chinese patent medicines. 0.9% sodium chloride injection, 5% glucose injection, glucose sodium chloride injection and other solvents are not included in the analysis.

2.3 Case matching grouping

Patients with renal function injury were defined as "renal function injury group", and patients without renal function injury were defined as "normal renal function group". A 1:4 propensity score matching was conducted according to the same gender and age difference ± 5 years. Finally, the data of 95 patients were included,that is, 19 patients in the renal function injury group and 76 patients in the normal renal function group.

2.4 Data analysis

Statistical description of combined medication was carried out.At the same time, MCP (minimal concove penaly) algorithm was used to screen the key influencing factors [15], traditional logistic regression, propensity score weighted logistic regression and other methods to further accurately analyze the influencing factors. The statistical analysis software is SAS 9.2 and R 4.1.0. P＜0.05 was regarded as the difference with statistical significance.

3. Results

3.1 Main combined drugs

According to the order of the frequency of combined drugs, the top 20 combined drugs were included, including all combined drugs within 7 days before the use of EEI, as shown in Table 1; During the use of EEI, as shown in Table 2; And the combined medication(later) within 7 days after stopping the use of EEI, as shown in Table 3.

3.2 Traditional logistic regression results

Logistic regression analysis method was used to find the key drug combinations that lead to abnormal changes in renal function of patients with EEI. Considering that the number of independent variables is greater than the sample size, MCP method was used to screen the key variables before analysis. Finally, three important variables affecting abnormal changes of renal function were screened out by MCP algorithm.

Classical logistic regression was performed on the three important variables with abnormal renal function. The P values were less than 0.05, which was statistically significant.

3.3 Logistic regression results based on propensity score

Propensity score weighted logistic regression can synthesize multiple confounding variables into one variable, so as to effectively balance the distribution of covariates, eliminate the influence of confounding factors on the results, and make the research results more reliable [16]; GBM propensity score weighted logistic regression is a more flexible multiple nonparametric regression technique. Compared with the previous analysis method, its analysis results are closer to the clinical reality[17]. In order to further evaluate the key drug combination of abnormal changes in renal function in patients with EEI, we analyzed the above three variables by propensity score weighted logistic regression and GBM propensity score weighted logistic regression.

Table 1 Combination and medication of renal function injury group and control group within 7 d before elemene emulsion injection［n(%)］

Table 2 Combination and medication of renal function injury group and control group during the use of elemene emulsion injection［n(%)］

Table 4 Screening results of important variables based on MCP algorithm

Table 5 Traditional logistic regression results

According to table 6 and table 7, the P values of variable potassium chloride (before) and furosemide (during) in the logistic regression analysis of two propensity scores are less than 0.05, that is, the difference is statistically significant, indicating that the combined use of potassium chloride within 7 days before EEI or furosemide during EEI may be an important factor leading to renal function injury in patients with EEI. The p value of the variable furosemide (before) in the logistic regression analysis of the two propensity scores is greater than 0.05, that is, the difference is not statistically significant. It can not be considered that the combined use of furosemide within 7 days before the use of EEI is a possible important factor leading to renal function injury in patients with EEI.

Table 6 Weighted logistic regression results of propensity score

Table 7 Weighted logistic regression results of GBM propensity score

4. Discussion

According to the theory of traditional Chinese medicine, the kidney is the foundation of congenital essence. It is the foundation of yin and Yang of human viscera and the source of human growth,development and reproduction. In modern medicine, the kidney is not only a urinary organ, but also the main organ for the metabolism and excretion of drugs and poisons in the human body. It plays a key role in regulating the balance of metabolism of various substances in the human body[18]. Therefore, understanding and mastering the relevant factors of possible renal injury after medication is of great significance for clinical treatment and risk avoidance.

In the 1980s, foreign scholars Rosenbaum and Rubin proposed the propensity scoring method[19], which is most suitable for retrospective data analysis. In the absence of random distribution,it can effectively balance the confounding bias in non randomized controlled studies and make the research results closer to the clinical reality. Compared with traditional analysis methods, propensity scoring method can effectively reduce the difference between groups, and has the advantages of accommodating more covariates and lower data requirements[20]. In the past decade, the propensity scoring method has been applied to the analysis of clinical data in many medical studies[21-25], which is a powerful scientific research tool in real-world research. Therefore, we learned from these professional research methods and combined with statistical knowledge to explore the combination of drugs that may lead to renal function injury in patients with EEI.

In this study, the MCP variable selection method was used to screen the main factors that may lead to renal function injury, and then the classical logistic regression and propensity score logistic regression were used to further analyze the above main factors.In the analysis process, we took potassium chloride (before),furosemide (before) and furosemide (during) as the exposure group and the other two factors as the non exposure group, Constantly balance the relationship between the exposed group and the non exposed group, and finally come to the conclusion that the combined use of potassium chloride within 7 days before the use of EEI or the combined use of furosemide during the use of EEI (i.e. from the beginning of the doctor's order to the end of the doctor's order) will cause renal function damage.

This study is based on real clinical data, with a large amount of information and a wide range. The conclusions can provide researchers with some ideas and directions. However, there are also some limitations in the process of analysis and exploration.For example, although the number of class III and class A hospitals included is large, it still can not fully reflect the overall situation of the country, and some biases and confounding factors in the study are difficult to avoid. Therefore, clinicians still need to further verify in the follow-up, so as to guide clinical safe drug use more scientifically and reasonably and reduce the risk of renal function injury.

Author's contribution:

Lei Chao: data analysis and thesis writing; Zhao Xiaoxiao: data verification; Yu Xiaokang: data analysis; Wang Zhifei: project design and thesis modification; Xie Yanming: project design and review. All authors declare that there is no conflict of interest.