Abstract
Objective
To assess the conclusiveness of Cochrane reviews in the field of gynaecological cancer.
Methods
The Cochrane Library was searched for reviews regarding gynaecological cancer published between 1 January 2000 and 1 November 2014. Data were extracted from each paper and the conclusiveness of each review was assessed.
Results
The study included 66 reviews, 41 (62.1%) of which were conclusive. Of these, 58 included randomized controlled trials (RCTs), 37 (63.8%) of which were conclusive. Conclusive reviews of RCTs included significantly more patients than inconclusive reviews, but there was no difference in the number of included studies. Of the eight reviews of nonrandomized studies, four (50.0%) were conclusive. The majority of reviews recognized the need for additional studies.
Conclusions
In the field of gynaecological cancer, reviews are more likely to be conclusive when they include RCTs, as well as large numbers of patients.
Introduction
According to the Oxford Centre for Evidence Based Medicine, systematic reviews of randomized clinical trials (RCTs) can provide level 1 clinical evidence. 1 The Cochrane Collaboration is an educational not-for-profit organization, with >28 000 members in >120 countries. 2 Evidence-based medicine has been greatly assisted by Cochrane reviews, which are free from any commercial sponsorship or conflicts of interest, and are internationally accepted as the highest quality systematic reviews in human health care and health policy. 2 Generally, systematic reviews analyse available RCTs and/or nonrandomized studies, providing recommendations regarding what works, what may work, and whether there is insufficient evidence to make a judgement. Due to insufficient numbers or quality of studies, not all Cochrane reviews provide conclusive recommendations.
The aim of the present study was to establish the clinical conclusiveness of Cochrane reviews in the field of gynaecological cancer. We tested the hypotheses that (i) most reviews are inconclusive; (ii) the ability to reach a conclusion is determined by the number of studies and the number of patients; (iii) most reviews require additional studies; (iv) conclusiveness may be influenced by year of publication.3–5
Materials and methods
Search methods
Titles, abstract and keywords of articles in the gynaecological cancer section of the Cochrane Library and Archive were searched using the terms fallopian tube, ovary, uterine body, uterine cervix, vagina, vulva and trophoblast. The search included all reviews published between 1 January 2000 and 1 November 2014 (date of search), with the exception of those where the reference list did not include any articles relevant to our study.3–5
Data extraction
Data regarding type of conclusion (conclusive vs inconclusive), number of RCTs that fulfilled Cochrane review criteria, number of nonrandomized studies that fulfilled Cochrane review criteria, number of enrolled patients, and the need for additional studies (including number of studies, sample size, specific subgroups and rare side-effects.) were extracted independently by two investigators (Y.C. and A.W.). Disagreements were resolved by discussion.3–5
Data analysis
Cochrane reviews were classified as either conclusive (either one intervention was better than another, or one intervention was similar to another) or inconclusive (either the quality of the studies was inadequate or there were insufficient data).3–5
Statistical analyses
Results were expressed as mean ± SD, or median (range). Reviews were stratified according to whether they included RCTs or nonrandomised studies. Differences between conclusive and inconclusive reviews were analysed using Kruskal–Wallis test. Linear regression analysis was used to evaluate the correlation between percentage of conclusive reviews and year of publication. P-values < 0.05 were considered statistically significant.
Results
The search strategy retrieved 81 Cochrane reviews related to gynaecological cancer. Of these, 15 were excluded because the reference lists did not include any relevant articles. The study included 66 reviews, 41 of which were conclusive.
Study parameters in Cochrane reviews of randomized controlled trials in the field of gynaecological cancer, stratified according to review conclusiveness.
Data presented as mean ± SD and (median [range]).
Kruskal–Wallis test.
NS, not statistically significant (P ≥ 0.05).
Study parameters in Cochrane reviews of nonrandomized studies in the field of gynaecological cancer, stratified according to review conclusiveness.
Data presented as mean ± SD and (median [range]).
Kruskal–Wallis test.
NS, not statistically significant (P ≥ 0.05).
There was no correlation between the percentage of conclusive reviews and year of publication (Figure 1).
Linear regression analysis of the correlation between the percentage of conclusive Cochrane reviews (CRs) in the field of gynaecological cancer, and year of publication; no correlation was observed.
Discussion
The present study found that over half of all Cochrane reviews in the field of gynaecological cancer were conclusive. These findings are inconsistent with the hypothesis that the majority of reviews are inconclusive.3–5 Reviews of RCTs were more likely to be conclusive than those of nonrandomized studies (63.8% vs 50.0%).
Based on our findings, a specific review cannot always provide gynaecological oncologists with a definite clinical recommendation. Reviews of RCTs, however, are likely to be conclusive and provide practical answers regarding what works (or what may work) in a particular clinical situation. 4 Although it is possible to make a decision based on conclusive reviews, the quality of included RCTs and the review itself should be assessed.6,7 The methodological quality of review articles is assessed using the AMSTAR tool. 8 This checklist comprises eleven items including a priori design, duplicate study selection and data extraction, and the use of publication status as an inclusion criteria. The quality of RCTs in a specific Cochrane review is assessed using the Cochrane Collaboration tool, which includes selection bias (random sequence generation and allocation concealment), performance bias (blinding of participants and personnel), detection bias (blinding of outcome assessment), attrition bias (incomplete outcome data), reporting bias (selective reporting of outcomes), and any other possible sources of bias.7,9–12
In the present study, both the number of included articles and the number of patients enrolled influenced the conclusiveness of reviews based on nonrandomized studies. In contrast, the number of patients alone affected the conclusiveness of reviews of RCTs, suggesting that an increase in patient numbers should be sufficient to improve the conclusiveness of these reviews. Although the majority of reviews recognized the need for additional studies, the reasons given were not always detailed: many articles included simple statements such as “not enough studies” or “sample size too small”.
The present study has some limitations. Due to heterogeneity, we did not search MEDLINE® and EMBASE® for systematic reviews. In addition, the limited number of review articles for each type of cancer meant that it was not possible to perform subcategory analyses.
In conclusion, Cochrane reviews are recognized as being of the highest standard, and cover the most popular topics in evidence-based health care.13,14 In the field of gynaecological cancer, reviews are more likely to be conclusive when they include RCTs, as well as large numbers of patients.
Footnotes
Declaration of conflicting interest
The authors declare that there are no conflicts of interest.
Funding
This work were supported by a grant from the National Natural Science Foundation of China (Grant No. 81370703) and a grant from the Ministry of Science and Technology of China (Grant No. 2012BAI32B05).