Bayesian methodologies to use historical data in the analysis of clinical trials — Pubrica

In brief

Historical data has been offered as prior knowledge in the analysis of the current trial using Bayesian techniques. The meta-analytic-predictive previous is the most promising technique for estimating parameters for between-trial heterogeneity, as it offers the best trade-off of power, accuracy, and type I error. Randomized controlled studies for acute myeloid leukaemia make up the driving data set.

Introduction

Clinical trials are rarely conducted in isolation. Data from prior studies with a similar setup are often accessible. Such previous data may contain information relevant to the present trial’s research questions. Incorporating this historical data into the current trial’s analysis might enhance the precision of the estimations, increasing statistical power for hypothesis testing and lowering sample sizes.

When using historical data in the analysis of a clinical trial, one must consider the possibility of trial heterogeneity. The experimental therapy varies from trial to trial, but the treatment in the control arm is generally consistent; thus, only the control arms of prior trials are suitable for inclusion in the current trial’s analysis. Heterogeneity among historical trials, as well as between the present trial and the historical trials, can be caused by differences in patient populations or other trial-specific factors. The historical data should not be used if there are significant differences between the past trials and the current trial.

Designing clinical trials

Historical data from past clinical studies are usually considered while developing clinical trials. For example, to calculate the variance and clinically relevant effect size for a sample size calculation or gather data on recruitment rates and population sizes. However, it appears that historical data is rarely employed in clinical trial analyses.

One way to include historical data into a clinical trial’s analysis is to use historical controls to replace or complement current controls. Numerous assumptions must be valid for the historical control group when using this method (s). The historical control group(s) must, for example, (1) have received the same precisely specified standard therapy as the randomized controls in the current trial, and (2) have participated in a prior clinical study with the same subject inclusion and exclusion criteria. Additional specifications may be found in.

Potential benefits of using Bayesian methods

There are several Bayesian approaches for incorporating historical control data from single research into trial data analysis. Power prior [1], Hierarchical Power prior [2], Modified Power prior [3], and comparable prior [4] are some of them. While the most basic pooling technique assumes that the historical controls are equivalent to the current study’s randomized controls, additional ways downweight previous data. Size reduction via prior information

While the power prior and commensurate prior may be used to include historical control data from numerous clinical trials into a new study’s analysis, there are also two meta-analytical methods. The retrospective Meta analytical combination (MAC) analysis combines previous and current data to produce a meta-analysis. This may be a non-Bayesian analysis. The prospective Meta analytical-predictive (MAP) analysis does a meta-analysis of historical data to create a MAP prior, which is then combined with current data using the Bayesian rule. The gold standard is the meta-analytic previous (MAP), utilized in numerous published researches.

Historical control data

In the analysis of clinical trials, we can enhance statistical power or lower the needed sample size by incorporating historical control data. With increased small-population trials (e.g., rare disease studies, paediatric studies, studies in difficult-to-recruit therapeutic populations) and challenges in meeting evidential standards, methods to re-use patient data from previous clinical trials are expected to become more effective popular in the coming years.

Specific statistical and computational expertise

The Bayesian method frequently necessitates statistical skill in Bayesian computing and analysis. MCMC and other special computational techniques are commonly used to

· Examine trial data

· Validate model assumptions

· Assess prior probabilities

· Run simulations to assess the probability of different outcomes and

· Determine sample size.

The increased precision on device performance that can be obtained by incorporating prior information, or the benefits of a flexible Bayesian trial design in the absence of prior knowledge, may offset the technical and statistical costs involved in successfully designing, conducting, and analyzing a Bayesian trial (e.g., smaller expected sample size resulting from interim analysis).

References

1. vanRosmalen J, Dejardin D, van Norden Y, Löwenberg B, Lesaffre E. Including historical data in the analysis of clinical trials: Is it worth the effort? Stat Methods Med Res. 2018 Oct;27(10):3167–3182. doi: 10.1177/0962280217694506. Epub 2017 Feb 21. PMID: 28322129; PMCID: PMC6176344.

2. Neuenschwander B, Branson M, Spiegelhalter DJ. A note on the power prior. Stat Med 2009; 28: 3562–3566.

3. Duan Y, Ye K, Smith EP. Evaluating water quality using power priors to incorporate historical information. Environmetrics 2006;