Dataset for meta-analysis and metaregression of published outcomes data on definitive radiotherapy with or without chemotherapy for locally advanced cervical cancer in patients with contraindications to cisplatin
Description
We performed a systematic review and meta-analysis to examine the outcomes of patients with locally advanced cervical cancer with contraindications to cisplatin, that were managed with definitive radiotherapy with or without chemotherapy or other co-interventions. We then performed a metaregression to examine the effects of clinical and treatment variables on outcomes. We identified eligible studies and extracted relevant data using a standardized data extraction template. Table 1 summarizes, per eligible study, the characteristics of the population or subgroup (including nature of contraindication to cisplatin), and the intervention groups (radiotherapy, brachytherapy and/or chemotherapy regimens). Table 2 summarizes our assessment of risk of bias for each study using templates that were based on the CASP Critical Appraisal Checklists. Table 3 and 4 summarize the outcomes (tumor response, survival, compliance rates, and toxicity) for each intervention group. Table 5 consolidates numerical data from Table 1, 3 and 4, suitable for meta-analyses and meta-regression. Table 6 is a composite database that contains the main template used in the preparation of the other datasets for the meta-analyses and meta-regression, given the variables and outcomes of interest.
Files
Steps to reproduce
1. Do a systematic search to identify relevant studies to your research question and objective. 2. Screen each study by examining the title and abstract; review the full text to determine final eligibility. 3. Extract data using a standardized and piloted data-extraction template. 4. Tabulate the data. For proportional meta-analyses, outcome rates will be necessary. For comparative meta-analyses, risk ratios, odds ratios, or hazard ratios will be necessary. 5. For the meta-analyses, use MetaXL v5.3. The software allows for both proportional and comparative meta-analyses. 6. When pooling outcomes from different study designs, use the quality effects model. This model allows assigning quality index (Qi) score to each study. 7. In case of heterogeneity, MetaXL allows for sub-group and sensitivity analyses. 8. In case of an anlyses that include >10 studies, MetaXL allows for generation of both funnel plots and Doi plots to evaluate publication bias. 9. For the meta-regression, use the MetaXL to generate metaregression data. The metaregression data could then be entered into Stata v16.1. 10. Stata allows for both categorical and continuous meta-regression, depending on the nature of the variable of interest.