Trial Sequential Analysis (TSA), one kind of cumulative meta-analysis, is a method which introduces sequential analysis into traditional meta-analysis to avoid random errors (false positive or false negative outcomes) that occurred during repeated updates when traditional meta-analysis is performing. It is also applied to calculate required information size (RIS) of a firm conclusion. This study aims to summarize the proposal, fundamental theory, application software, and current limitation of TSA, and to clarify the advantages of TSA on the basis of detailed examples, in order to attract more attention of researchers and promote the methodological development of meta-analysis in China.
The robustness of results of statistical analysis would be altered on the condition of repeated update of traditional meta-analysis and cumulative meta-analysis. In addition, the cumulative meta-analysis lacks estimation of the sample size. While trail sequential analysis (TSA), which introduces group sequential analysis in meta-analysis, can adjust the random error and ultimately estimate the required sample size of the systematic review or meta-analysis. TSA is performed in TSA software. In the present study, we aimed to introduce how to use the TSA software for performing meta-analysis.
ObjectiveTo systematically evaluate the outcomes of drug-eluting balloon (DEB) in treating coronary artery in-stent restenosis (ISR) by using meta-analysis and trial sequential analysis (TSA). MethodsWe searched PubMed, EMbase, The Cochrane Library（Issue 4, 2016）, CNKI, CBM, VIP and WanFang Data to collect randomized controlled trials (RCTs) regarding the treatment of ISR by DEB from inception to April 2016. After two reviewers independently screened citations, extracted data and assessed the bias risk of included studies, we carried out meta-analysis and TSA analysis by using RevMan 5.3 version and TSA v0.9 respectively. ResultsA total of 10 RCTs involving 1909 patients were included. Seven-hundred and forty-seven patients were included with regard to the comparison between DEB and POBA, 1162 patients were recruited to compare DEB and drug-eluting stents (DES). The results of meta-analysis revealed that DEB was associated with decreased mortality (OR=0.36, 95%CI 0.14 to 0.93, P=0.04), compared with that of plain old balloon angioplasty (POBA). And TSA showed that cumulative Z-curve strode the conventional threshold value but not the TSA threshold value which suggested a false positive result of meta-analysis. In comparison with that of POBA, DEB had a lower incidence of target lesion revascularization (TLR) (OR=0.16, 95%CI 0.07 to 0.38, P＜0.01). And the result of TSA displayed that the cumulative Z-curve strode both the conventional and TSA threshold value which validated the result of meta-analysis. Besides, the results of meta-analysis showed that there were no significant differences in mortality (OR=0.84, 95%CI 0.41 to 1.72, P=0.63) and TLR (OR=1.55, 95%CI 0.76 to 3.16, P=0.22) between DEB and DES. However, the result of TSA revealed that the cumulative Z-curve did not strode both the conventional and TSA threshold value, and the included sample size less was than required information size which suggested that the reliability of the meta-analysis needed more studies to confirm. While the subgroup analysis of EES revealed that DEB had a higher incidence of TLR than that of DEB (OR=3.37, 95%CI 1.59 to 7.15, P＜0.01). And the result of TSA displayed that the cumulative Z-curve strode both the conventional and TSA threshold value which validated the result of meta-analysis. ConclusionCurrent evidence shows, EES is superior to DEB in decreasing the incidence of TLR in patients with ISR, while DEB is superior to POBA. However, the comparison of DEB and other strategies on reducing of mortality in patients with ISR still needed more studies to prove.
Cumulative meta-analysis could help researchers to justify the effectiveness of the intervention and whether the obtained evidence is sufficient. However, the process of the meta-analysis does not adjust the repeated testing of the null hypothesis and neither quantifies the statistical power. The sequential meta-analysis has solved the aforementioned problems and has been widely used in the clinical practice and decision-making. Currently several methods of sequential meta-analysis have been proposed and these methods differ from each other. Of which, the methodology of trial sequential (TSA) is well developed and corresponding performance is relatively easy; the methodology of double-triangular test of Whitehead is lagged than TSA and its performance is relatively difficult; the approach of semi-Bayes refers to the theory of Bayes and it's very difficult to generalize. Our paper aimed to give a brief introduction of the methodology of the sequential meta-analysis.
ObjectiveTo systematically review the association between angiotension-converting enzyme (ACE) gene insertion/deletion (I/D) polymorphism and osteoarthritis (OA) by using meta-analysis and trial sequential analysis (TSA). MethodsThe PubMed, EMbase, CNKI, CBM, VIP, and WanFang Data were searched up to October 12th, 2016 for case-control or cohort studies on the correlation between ACE I/D polymorphism and OA risk. Two reviewers independently screened literature, extracted data and assessed the risk of bias of included studies. Then, meta-analysis and TSA analysis were performed using Stata 13.1 software and TSA v0.9 soft ware. ResultsA total of six case-control studies involving 1 165 OA patients and 1 029 controls were included. The results of meta-analysis showed that the ACE I/D was associated with OA risk (DD+DI vs. II: OR=1.72, 95%CI 1.02 to 2.90, P=0.04; DI vs. II: OR=1.65, 95%CI 1.06 to 2.56, P=0.03). Subgroup analysis of ethnicity showed that, in Caucasians, the ACE I/D was associated with OA risk (DD vs. DI+II: OR=2.10, 95%CI 1.54 to 2.85, P<0.01; DD+DI vs. II: OR=3.11, 95%CI 2.20 to 4.39, P<0.01; DD vs. II: OR=4.01, 95%CI 2.68 to 6.00, P<0.01; DI vs. II: OR=2.65, 95%CI 1.06 to 2.56, P<0.01; D vs. I: OR=2.11, 95%CI 1.72 to 2.58, P=0.73). And TSA showed that all of the cumulative Z-curve strode the conventional and TSA threshold value which suggested the result of the association between ACE I/D polymorphism and OA in Caucasians was very reliable. However, the association did not exist in Asians (DD vs. DI+II: OR=0.80, 95%CI 0.60 to 1.07, P=0.13; DD+DI vs. II: OR=1.08, 95%CI 0.87 to 1.35, P=0.49; DD vs. II: OR=0.86, 95%CI 0.62 to 1.20, P=0.38; DI vs. II: OR=1.18, 95%CI 0.93 to 1.50, P=0.19; D vs. I: OR=0.93, 95%CI 0.83 to 1.14, P=0.73). And the results of TSA displayed that all of the cumulative Z-curve did not strode both TSA threshold value and required information size line excepting for DD vs. DI+II genetic model which suggested that the sample-size in Asians was insufficient. ConclusionsThe ACE D allele maybe a risk factor for OA in Caucasians. However, the association between ACE I/D polymorphism and OA risk in Asians still need more studies to prove.
Trial sequential analysis (TSA) could be performed in both TSA software and Stata software. The implementation process of TSA in Stata needs the command of "metacumbounds" of Stata combines with the packages of "foreign" and "ldbounds" of R software. This paper briefly introduces how to implement TSA using Stata software.
Trial sequential analysis (TSA) can identify inclusive results of apparently conclusive of meta-analyses by providing require information size and monitoring boundary. Certain methods of calculating information size are existed. Our objective was to give a brief introduction of four methods to help readers to better perform TSA in making meta-analyses.
The sample size of a meta-analysis should not be less than a single randomized controlled trial. Trial sequential analysis (TSA) can provide required information size and monitoring boundary to justify the conclusion of meta-analysis. However, the TSA software is only suitable for binary and continuous data, and it cannot analyze the time-to-event data. This paper aimed to introduce how to analyze the time-to-event data using TSA approach.
The assumption of fixed-effects model is based on that the true effect of the each trial is same. However, the assumption of random-effects model is based on that the true effect of included trials is normal distributed. The total variance is equal to the sum of within-trial variance and between-trial variance under the random-effects model. There are many estimators of the between-trial variance. The aim of this paper is to give a brief introduction of the estimators of between-trial variance in trial sequential analysis for random-effects model.
Objective To detect the false-positive results of cumulative meta-analyses of Cochrane Urology Group with the trial sequential analysis (TSA). Methods The systematic reviews of Urology Group of The Cochrane Library were searched to collect meta-analyses with positive results. Two researchers independently screened literature and extracted data of included meta-analyses. Then, TSA was performed using TSA software version 0.9 beta. Results A total of 11 meta-analyses were included. The results of TSA showed that, 8 of 11 (72.7%) meta-analyses were potentially false-positive results for failing to surpass the trial sequential monitoring boundary and to reach the required information size. Conclusion TSA can help researchers to identify the false-positive results of meta-analyses.