The search was conducted in accordance with the principles outlined in the Cochrane Handbook for Systematic Reviews of Interventions (21). Studies were identified by searching electronic databases and relevant websites. We carried out a comprehensive search to identify potential articles in PUBMED, MEDLINE database and EMBASE up to June 2022, by using the search terms: “Polo-like Kinase 1 Inhibitor” or “Plk1 inhibitor” or “Rigosertib” or “ON 01910.Na” or “Volasertib” or “BI 2536” or “BI 6727” or “GSK461364A” and “cancer” or “tumor” or “carcinoma” or “neoplasm”, limited to clinical trials. There was no limit on the language of publication. In order to ensure the completeness and quality of the results, relevant scientific meetings were retrieved and independent search was performed by using the Web of Science database, and unpublished trials were checked in the clinical trial registry (http://www.clinicaltrials.gov).

The search was complemented by additional sources, including relevant systematic reviews and the reference lists of included studies that were manually searched to identify additional potentially relevant studies. To be included, studies had to be RCTs, quasi-RCTs, and nonrandomized comparative studies, and had to report at least one outcome of interest (4). Secondary outcomes included AEs that were reported individually or collectively (i.e., based on severity grading systems). Single-arm trials and trials with Plk1 inhibitors used in both arms were accepted. Studies with no comparative elements were excluded. Trials with Plk1 inhibitors used to treat other diseases were also excluded. In all the included RCTs, Plk1 inhibitors were used alone or in combination with other chemotherapeutic agents as the treatment group, while placebo or other chemotherapeutic agents were used in the control group. Two investigators independently reviewed the articles to exclude irrelevant and overlapping studies. Disagreement was resolved by discussion, and where no agreement was reached, a third independent party acted as an arbiter.

A data extraction form was first developed specifically to collect information on study design, characteristics of participants, characteristics of interventions, and outcome measures. Two reviewers independently extracted data related to the prespecified outcomes. Primary study authors were asked to provide missing, ambiguous, or important additional data. Here, we collected the following information from all the included RCTs: first author’s surname, year of publication, number of participants, type of tumors, trial phase, type of inhibitors, median age, and male (%). In addition, the ORs and ES of the AEs, and the standard mean difference (SMD) of the pharmacokinetic parameter (22) (e.g., total body clearance rate, terminal half-life, apparent volume of distribution at steady state) with 95% CIs were extracted from most of the trials to evaluate the curative effect of Plk1 inhibitors. Information on AEs was also retrieved to calculate the safety of Plk1 inhibitors (23). A study was not included in the meta-analysis if there was evidence of severe bias or heterogeneity.

The quality of the selected studies was evaluated with a methodology checklist developed by the Scottish Inter-collegiate Guidelines Network (SIGN: http://www.sign.ac.uk/pdf/sign50.pdf). Because some studies selected were nonrandomized, we used the SIGN guideline for quality assessment of nonrandomized studies, which consisted of the following judgments: focused questions, selection bias of subjects, assessment of outcomes and exposure status, handling confounders, and provision of confidence intervals.

All analyses were performed with Stata version 13.0 (Stata Corp, College Station, Texas). For data analysis, descriptive statistics was used to summarize the data of baseline characteristics. A quantitative synthesis (i.e., meta-analysis) was performed only in RCTs, and only if methodologically appropriate. The random-effects model was legitimately used due to the anticipated clinical heterogeneity of participants and interventions. For time-to-event data such as PFS and OS, ES and 95% CIs directly obtained from studies were used to compare results by using the inverse variance method. A 2-tailed P value of less than 0.05 was judged as statistical significance. ORs and 95% CIs were used to assess the AEs between Plk1 inhibitors group and the control group. In addition, we extracted dichotomous data from all studies reporting number of patients with AEs and total participants, which were pooled for calculating ES with 95% CIs. For the continuous outcomes such as pharmacokinetic parameter, the SMD was calculated. The degree of heterogeneity was measured by the I² statistic, with I² < 25%, 25-75% and > 75% to represent low, moderate and high degree of inconsistency, respectively (24). Statistical heterogeneity was defined as an I² statistic value of more than 50% (24). In analyses, if the heterogeneity was low, then we needed to use a fixed-effect model, or else apply the random-effect model. We further performed a subgroup analysis based on tumor types and different kinds of Plk1 inhibitors (BI 2536, Rigosertib and Volasertib). Funnel plot and Egger’s regression asymmetry test were used to access the publication bias of literatures.

A Preferred Reporting Items for Meta-Analyses (PRIMAs) studies were used in this work. We initially identified 107 potential eligible studies through title and abstract screening. Among them, 76 studies were excluded since they were not relevant to our analysis, leaving 31 articles for full review. After assessing the full texts of these potentially relevant studies, 9 studies were excluded for the following reasons: 14 contained no relative outcomes and 12 were duplicate publication. Ultimately, 22 eligible clinical trials involving a total of 1256 patients were included for analyses. A flow diagram of the trial selection process was showed in Figure 1 . Data on trial details (e.g., first author, publication year, phase, type of tumors, type of inhibitors, number of patients, and patient characteristics), treatments, and outcomes were separately extracted into a spreadsheet. Survival data assessed by the independent review facility were preferably extracted to avoid potential assessment bias by investigators. We minimised issues arising from potential lack of transitivity by only including RCTs with strict patient allocation, and optimized balance to address all treatments for the same condition. Transitivity was evaluated by the use of descriptive statistics for study and population baselines, such as sample size, age and sex. Age of Participants recruited in different studies ranged from minimum of 17 years to maximum of 87 years, with a median age size of 61. Although the potential eligible studies from EMBASE were partly overlapped with previous publications, some of them provided elaborated incidence about AEs. They were not mentioned in previous meta-analysis-based articles, so these relevant studies were also included. We preferred the use of treatment related AEs, but if not specified as treatment related, we used all adverse events. ClinicalTrials.gov and other available sources were evaluated for the most recent and complete data. Our study included 6 phase II/III clinical trials and 16 phase I clinical trials, and the types of inhibitors involved BI 2536, Volasertib (BI 6727), Rigosertib (ON 01910.Na) and GSK461364A. In all 22 clinical trials, 1 used GSK461364A, 7 used BI 2536, 4 used Rigosertib (ON 01910.Na), and 10 used Volasertib (BI 6727). Moreover, 2 trials were about acute myeloid leukemia (AML), 1 trial was about locally advanced or metastatic urothelial cancer (LAMUC), 1 trial was about non-Hodgkin lymphoma, and 18 trials were about advanced solid tumors. The characteristics of 22 trials included were detailedly presented in Table 1 .

For the quality assessment of the studies, the bias risk assessment tool recommended by Cochrane is applied to evaluate the quality of all included studies and the risk of bias. The assessment will include random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective outcome reporting, and other sources of bias. The risk of high and low bias can be expressed as “high risk” and “low risk”, respectively. The information provided in the study is inaccurate or insufficient for the bias assessment, which can be expressed as “unclear risk”. The evaluation of the above content is independently evaluated by two researchers. If there are different opinions, the discussion will be conducted. If there are still differences, consult the third appraiser. Otherwise, consult with the Cochrane Professional Group. We used Funnel plot and Egger’s regression asymmetry test to access the publication bias of literatures. Arrangement of data points did reveal evidence of obvious asymmetry. It was further confirmed by Egger’s linear regression asymmetry test for each outcome, and the results still did show evidence of publication bias (t = 2.70, P = 0.016) ( Figure 3 and Table S1 ). Therefore, more studies will be needed to provide stronger evidence in the future.

Three investigators (Hartmut, Peter and B. H. O’Neil) were independently appraised the potential risk of bias of the RCTs by using the Cochrane Risk of Bias Methods at the study and outcome levels. Methodological quality summary presented the authors judgments about each methodological quality item for each included study. Potential risks of bias within individual trials were shown in Table 3 . The included RCTs were at high risk of bias and unclear methodological quality items. Selection bias, caused by inappropriate methods of generating random sequences, was the main source of potential bias in this meta-analysis.

A total of 22 eligible articles revealed treatment outcomes with Plk1 inhibitors. The type of AEs primarily involved the respiratory system, urinary system, skin system, nervous system, facial features, digestive system (liver function, stomach, nausea or vomiting, appetite, abdomen), hematopoietic system (anemia, neutropenia, leukopenia, thrombocytopenia), musculoskeletal system, and non-specific tissues and organs. In this study, the meta-analysis was used to combine the incidence of AEs in single group, and the heterogeneity analysis and summary analysis were displayed in the Table 4 and Figures S3–S18 . From these outcomes, it was clearly observed that the incidence of all AEs was not more than 20.2%. Of which, the incidence of AEs in the nervous system was shown to be the highest (ES, 0.202; 95% CIs, 0.161-0.244), followed by in blood system (ES, 0.190; 95% CIs, 0.178-0.201) and in digestive system (ES, 0.181; 95% CIs, 0.150-0.213). In particular, the incidence of AEs in the non-specific system was only 6.9% (ES, 0.069; 95% CIs, 0.046-0.093).

To evaluate the safety of Plk1 inhibitors, we analyzed the risk factor of any side-effects in the overall population with AEs. All included studies concerning the main comparisons of interventions were summarized in Tables 5 , 6 and Figures S19–S24 . In this study, the relationship between Plk1 inhibitors and type of AEs was presented here. After the intervention with the Plk1 inhibitors, the group of non-solid tumors presented the highest incidence of AEs in the blood system (ES, 0.426; 95% CIs, 0.117-0.736), while the group of solid tumors showed the lowest incidence of AEs in the digestive system (ES, 0.180; 95% CIs, 0.147-0.212). Furthermore, three Plk1 inhibitors were used primarily in the clinical trials, thus we also separately evaluated the safety of the BI 2536, Volasertib (BI 6727) and Rigosertib (ON 01910.Na) subgroups. Analysis of patients in the overall population showed that, compared with other systems, Rigosertib (ON 01910.Na) were associated with a decreased risk of AEs in digestive system (ES, 0.103; 95% CIs, 0.059-0.147), but Volasertib (BI 6727) increased risk of AEs in blood system (ES, 0.399; 95% CIs, 0.294-0.504).

Five eligible studies reported the pharmacokinetic parameters following one cycle of treatment. Among them, relevant studies that reported only mean value without standard deviation value were excluded from the pooled analysis. The overall estimate of SMD of pharmacokinetic parameters and 95% CIs from the individual studies was shown in Figure 4 and Table 7 . Due to the degree of heterogeneity (I² = 37.2%) remaining low and acceptable, the meta-analysis with the random-effects model was used to assess the outcome, and a random-effects model could be selected to perform heterogeneity tests for pooled results. As a result, the standard mean difference (SMD, 0.33; 95% CIs, -0.24-0.91) was presented through the meta-analysis, and assume that the test confirms that the SMD difference was not statistically significant (Z =1.13, P = 0.257). After using different doses of Plk1 inhibitors (the 100 mg low dosage cohort and the 200 mg high dosage cohort), there was no statistical difference in the total plasma clearance, terminal half-life and apparent volume of distribution at steady state.

We used Funnel plot and Egger’s regression asymmetry test to access the publication bias of literatures. Arrangement of data points did not reveal any evidence of obvious asymmetry. It was further confirmed by Egger’s linear regression asymmetry test for each outcome, and the results still did not show any evidence of publication bias (t = 0.85, P = 0.413) ( Figure 5 and Table S2 ).

Compared with the previous traditional meta-analysis, the use of indirect binary comparisons within this meta-analysis added additional information to the evidence on the efficacy, safety and pharmacokinetic efficacy of Plk1 inhibitors in the treatment of tumors, especially in determining the impact of Plk1 inhibitors on the treatment of tumors related to the nervous system, digestive system and blood circulation system. Our review included more trials and substantially more patients. Most importantly, for the first time, we refined the incidence of AEs to the respiratory system, urinary system, skin system, nervous system, facial features, digestive system (liver function, stomach, nausea or vomiting, appetite, abdomen), blood circulation system (anemia, neutropenia, leukopenia, thrombocytopenia), musculoskeletal system and nonspecific tissues and organs. Accordingly, it can be said that it is a relatively comprehensive systematic review and meta-analysis so far. In this review, we not only provide strategies for the selection of Plk1 inhibitors during clinical cancer treatment, but also make reasonable recommendations for the selection of dosages for their pharmacokinetic efficacy.

The strength of our study is that we simultaneously used a meta-analysis to compare the outcomes of four different types of Plk1 inhibitors in solid versus non-solid tumor patients. Namely, the efficacy of Plk1 inhibitors was evaluated by PFS and OS. The safety of Plk1 inhibitors was assessed by the incidence of AEs in different systems, tissues or organs throughout the body. Subgroup analysis was conducted to evaluate not only the incidence of AEs in the nervous system, blood circulation system and digestive system during the treatment of solid tumors and non-solid tumors with Plk1 inhibitors, but also the incidence of AEs in the treatment of tumors with three Plk1 inhibitors (BI 2536, Volasertib (BI 6727) and Rigosertib (ON 01910.Na). Additionally, we compared the SMD of pharmacokinetic parameters (total plasma clearance, terminal half-life and apparent volume of distribution at steady state) to assess the pharmacokinetic effects of Plk1 inhibitors. More importantly, we included RCTs without language restrictions to avoid bias.