Trial sequential analysis: Quality improvement for meta-analysis

Systematic review and meta-analysis provide a comprehensive insight into similar published papers. Nineteen systematic reviews were published in the Indian Journal of Anaesthesia from July 2023 to June 2024.[1-19] Of these systematic reviews, meta-analysis was reported in 12 articles. Of these, five studies reported trial sequential analysis (TSA).[1,4,14,17,19] Two out of five TSAs confirmed the results of meta-analyses, and the required sample size was reached in these two meta-analyses. These TSAs suggest that the result is definitive and that other randomised controlled trials are unlikely to modify the effect on the outcomes.[1,4] On the contrary, in three TSAs, the trials did not achieve the required sample size. Thus, TSAs suggest that the result is not definitive, although the pooled effect is statistically significant. Further studies are necessary to be conclusive.[14,17,19] Meta-analysis sometimes suffers from random errors (type-I and type-II errors), which may lead to misleading results and impact the reliability of the study. TSA is a statistical method used to enhance the reliability of meta-analysis by addressing these challenges. TSA monitors the cumulative evidence sequentially and helps researchers figure out when they have gathered sufficient evidence to make a reliable conclusion.[20] TSA takes into consideration acquired information size (AIS) (total number of participants included so far in the trials under review) and required information size (RIS) (target number of participants needed to make reliable conclusions). If the AIS reaches RIS, the study can be considered statistically strong enough to make a reliable conclusion. TSA uses sequential monitoring boundaries to establish whether an intervention is beneficial (boundary for benefit) or harmful (boundary of harm) and whether we can stop early as a result collected is statistically valid, and further analysis will only waste time and money (boundary of futility). TSA also utilises the Z curve. The Z curve indicates the addition of trials. If the Z curve crosses the benefit of the boundary, we can say that the result is statistically significant. If it crosses the benefit of harm, we can say that the result is insignificant. If the Z curve crosses RIS, we can say an adequate sample size exists [Figure 1].[21,22]Figure 1: Trial sequential analysis graphTSA introduces systematic checks throughout the process of accumulating data in a meta-analysis. TSA introduces stricter significance thresholds than meta-analysis. By setting benefit and harm boundaries, TSA only confirms an intervention effect when the evidence is strong enough to cross these boundaries. Similarly, calculating the RIS prevents premature conclusions about the lack of effect (false negative) or its effectiveness (false positive) by ensuring sufficient data is collected to reject/accept the effect of the intervention confidently. TSA helps to identify when sufficient evidence has been gathered to support a conclusion, preventing unnecessary further trials, which can be costly and time-consuming. Once the Z curve crosses the benefit of boundary or harm, we can conclude that the evidence is strong enough to say that the intervention is effective, and no further participants need to be enroled. This saves the time and resources of running additional studies that would provide redundant information. TSA prevents duplicative studies and ensures that resources are directed towards more impactful research. For example, once the TSA concludes an intervention’s effectiveness, the researchers and funding agencies can reallocate resources to other unmet medical needs rather than repeating similar studies.[20-22] However, like any other statistical procedure, TSA also has some limitations. Firstly, TSA is a complex statistical method that requires a deep application of statistical principles. If TSA parameters, alpha levels, beta spending functions, RIS, etc., are not properly understood and applied, TSA studies can lead to incorrect conclusions. Secondly, TSA sometimes can be too conservative, leading to premature termination of trials before the full potential of treatment is explored. TSA uses strict boundaries, which could result in beneficial treatments being abandoned too soon. For example, in a meta-analysis, where the benefit boundary is set very high, the trial may not reach the threshold for benefit even though the treatment shows promise. As a result, the trial may stop early without fully exploring the treatment’s potential. Lastly, TSA works best when the studies in a meta-analysis are relatively similar in design, populations, and outcomes. When there is high heterogeneity, that is, large differences between the studies, the estimates might become unreliable, and larger sample sizes are often required.[20-22] To conclude, systematic reviews and meta-analyses may lead to spurious results if they include a limited number of studies and participants. TSA may act as an additional tool to improve the quality of meta-analysis. TSA shows information on the precision and uncertainty of the meta-analysis results. TSA guides researchers if further trials are necessary, thus preventing the waste of effort and other resources. However, TSA is relatively complex and prone to misinterpretation. TSA can be a powerful tool for researchers and policymakers as it strengthens the conclusions of meta-analyses.