Incorporating individual historical controls and aggregate treatment effect estimates into a Bayesian survival trial: a simulation study

Variation in Outcomes Associated With Blunt Splenic Injury Management

Introduction The management of blunt splenic injury has evolved to include splenic artery embolization in addition to non-surgical management, and splenic surgery. Though research has been conducted examining outcomes between management approaches, the inferential findings are often limited by single-site study designs and small sample sizes. However, results from large-scale prior studies can inform inference if a non-frequentist (Bayesian) framework is used. Therefore, the purpose of this study was to examine mortality and length of stay associated with blunt splenic injury management using both frequentist and Bayesian methods. Methods A total of 56 patients presenting with blunt splenic injury were included in this retrospective, single-center, quantitative study from January 1, 2021 to December 31, 2022 to inform both methodological approaches. Findings from a national retrospective sample of 117,743 patients presenting with blunt splenic injury between 2007 and 2015 were included in the prior distribution for the Bayesian estimates to provide sufficient statistical power and improve internal validity and generalizability of findings. Results Mortality rates and hospital mean length of stay were not significantly different between blunt splenic injury management approaches of non-operative management (n=43), surgery (n=7), and splenic artery embolization (n=6) using a frequentist approach (9.3%, 0%, and 0%, P=.52; and 10.8 (15.8), 10.8 (4.7), and 4.6 (1.8), P=.86, respectively). Bayesian 95% highest density interval (HDI) estimates of the likelihood of mortality ([0.02; 0.18], [-6.4e-23; 0.3], and [-2.2e-22; 0.3]) and hospital mean length of stay ([7.7; 8.3], [11.0; 12.3], and [8.7; 10.2]) provided reduced uncertainty in point and dispersion estimates. Conclusions The inclusion of findings from large high-quality studies provides increased certainty in estimates from smaller studies. Posterior estimates can inform predictive models for testing in future studies.

Estimation of treatment effects in early-phase randomized clinical trials involving external control data

There are good reasons to perform a randomized controlled trial (RCT) even in early phases of clinical development. However, the low sample sizes in those settings lead to high variability of the treatment effect estimate. The variability could be reduced by adding external control data if available. For the common setting of suitable subject-level control group data only available from one external (clinical trial or real-world) data source, we evaluate different analysis options for estimating the treatment effect via hazard ratios. The impact of the external control data is usually guided by the level of similarity with the current RCT data. Such level of similarity can be determined via outcome and/or baseline covariate data comparisons. We provide an overview over existing methods, propose a novel option for a combined assessment of outcome and baseline data, and compare a selected set of approaches in a simulation study under varying assumptions regarding observable and unobservable confounder distributions using a time-to-event model. Our various simulation scenarios also reflect the differences between external clinical trial and real-world data. Data combinations via simple outcome-based borrowing or simple propensity score weighting with baseline covariate data are not recommended. Analysis options which conflate outcome and baseline covariate data perform best in our simulation study.

Enriching single-arm clinical trials with external controls: possibilities and pitfalls

For the past decade, it has become commonplace to provide rapid answers and early patient access to innovative treatments in the absence of randomized clinical trials (RCT), with benefits estimated from single-arm trials. This trend is important in oncology, notably when assessing new targeted therapies. Some of those uncontrolled trials further include an external/synthetic control group as an innovative way to provide an indirect comparison with a pertinent control group. We aimed to provide some guidelines as a comprehensive tool for (1) the critical appraisal of those comparisons or (2) for performing a single-arm trial. We used the example of ciltacabtagene autoleucel for the treatment of adult patients with relapsed or refractory multiple myeloma after 3 or more treatment lines as an illustrative example. We propose a 3-step guidance. The first step includes the definition of an estimand, which encompasses the treatment effect and the targeted population (whole population or restricted to single-arm trial or external controls), reflecting a clinical question. The second step relies on the adequate selection of external controls from previous RCTs or real-world data from patient cohorts, registries, or electronic patient files. The third step consists of choosing the statistical approach targeting the treatment effect defined above and depends on the available data (individual-level data or aggregated external data). The validity of the treatment effect derived from indirect comparisons heavily depends on careful methodological considerations included in the proposed 3-step procedure. Because the level of evidence of a well-conducted RCT cannot be guaranteed, the evaluation is more important than in standard settings.

Challenges of using external data in clinical trials- an illustration in patients with COVID-19

BACKGROUND

To improve the efficiency of clinical trials, leveraging external data on control and/or treatment effects, which is almost always available, appears to be a promising approach.

METHODS

We used data from the experimental arm of the Covidicus trial evaluating high-dose dexamethasone in severely ill and mechanically ventilated COVID-19 patients, using published data from the Recovery trial as external data, to estimate the 28-day mortality rate. Primary approaches to deal with external data were applied.

RESULTS

Estimates ranged from 0.241 ignoring the external data up to 0.294 using hierarchical Bayesian models. Some evidence of differences in mortality rates between the Covidicus and Recovery trials were observed, with an matched adjusted odds ratio of death in the Covidicus arm of 0.41 compared to the Recovery arm.

CONCLUSIONS

These indirect comparisons appear sensitive to the method used. None of those approaches appear robust enough to overcome randomized clinical trial data.

TRIAL REGISTRATION

Covidicus Trial: NCT04344730, First Posted: 14/04/2020; Recovery trial: NCT04381936.

Informed Bayesian survival analysis

BACKGROUND

We provide an overview of Bayesian estimation, hypothesis testing, and model-averaging and illustrate how they benefit parametric survival analysis. We contrast the Bayesian framework to the currently dominant frequentist approach and highlight advantages, such as seamless incorporation of historical data, continuous monitoring of evidence, and incorporating uncertainty about the true data generating process.

METHODS

We illustrate the application of the outlined Bayesian approaches on an example data set, retrospective re-analyzing a colon cancer trial. We assess the performance of Bayesian parametric survival analysis and maximum likelihood survival models with AIC/BIC model selection in fixed-n and sequential designs with a simulation study.

RESULTS

In the retrospective re-analysis of the example data set, the Bayesian framework provided evidence for the absence of a positive treatment effect of adding Cetuximab to FOLFOX6 regimen on disease-free survival in patients with resected stage III colon cancer. Furthermore, the Bayesian sequential analysis would have terminated the trial 10.3 months earlier than the standard frequentist analysis. In a simulation study with sequential designs, the Bayesian framework on average reached a decision in almost half the time required by the frequentist counterparts, while maintaining the same power, and an appropriate false-positive rate. Under model misspecification, the Bayesian framework resulted in higher false-negative rate compared to the frequentist counterparts, which resulted in a higher proportion of undecided trials. In fixed-n designs, the Bayesian framework showed slightly higher power, slightly elevated error rates, and lower bias and RMSE when estimating treatment effects in small samples. We found no noticeable differences for survival predictions. We have made the analytic approach readily available to other researchers in the RoBSA R package.

CONCLUSIONS

The outlined Bayesian framework provides several benefits when applied to parametric survival analyses. It uses data more efficiently, is capable of considerably shortening the length of clinical trials, and provides a richer set of inferences.

Using horseshoe prior for incorporating multiple historical control data in randomized controlled trials

Meta-analytic approaches and power priors are often used to incorporate historical controls into the analysis of a current randomized controlled trial. In this study, we propose a method for incorporating multiple historical controls based on a horseshoe prior, which is a type of global-local shrinkage prior. The method assumes that historical controls follow the same distribution as the current control. In the case in which only a few historical controls are heterogeneous, we consider them to follow a potentially biased distribution from the distribution of the current control. We analyze two clinical trial examples with binary and time-to-event endpoints and conduct simulation studies to compare the performance of the proposed and existing methods. In the analysis of the clinical trial example, the posterior standard deviation of the treatment effect is decreased by the proposed method by considering the bias between the current control and heterogeneous historical control. In the scenarios in which the current and historical controls follow the same distribution, the statistical power using the proposed method is higher than that using existing methods. The proposed method is advantageous when few or no heterogeneous historical controls are expected.

Using historical control data in bioassays for regulatory toxicology

Historical control data (HCD) consist of pooled control group responses from bioassays. These data must be collected and are often used or reported in regulatory toxicology studies for multiple purposes: as quality assurance for the test system, to help identify toxicological effects and their effect-size relevance and to address the statistical multiple comparison problem. The current manuscript reviews the various classical and potential new approaches for using HCD. Issues in current practice are identified and recommendations for improved use and discussion are provided. Furthermore, stakeholders are invited to discuss whether it is necessary to consider uncertainty when using HCD formally and statistically in toxicological discussions and whether binary inclusion/exclusion criteria for HCD should be revised to a tiered information contribution to assessments. Overall, the critical value of HCD in toxicological bioassays is highlighted when used in a weight-of-evidence assessment.

Hierarchical Bayesian modelling of disease progression to inform clinical trial design in centronuclear myopathy

Background

Centronuclear myopathies are severe rare congenital diseases. The clinical variability and genetic heterogeneity of these myopathies result in major challenges in clinical trial design. Alternative strategies to large placebo-controlled trials that have been used in other rare diseases (e.g., the use of surrogate markers or of historical controls) have limitations that Bayesian statistics may address. Here we present a Bayesian model that uses each patient’s own natural history study data to predict progression in the absence of treatment. This prospective multicentre natural history evaluated 4-year follow-up data from 59 patients carrying mutations in the MTM1 or DNM2 genes.

Methods

Our approach focused on evaluation of forced expiratory volume in 1 s (FEV1) in 6- to 18-year-old children. A patient was defined as a responder if an improvement was observed after treatment and the predictive probability of such improvement in absence of intervention was less than 0.01. An FEV1 response was considered clinically relevant if it corresponded to an increase of more than 8%.

Results

The key endpoint of a clinical trial using this model is the rate of response. The power of the study is based on the posterior probability that the rate of response observed is greater than the rate of response that would be observed in the absence of treatment predicted based on the individual patient’s previous natural history. In order to appropriately control for Type 1 error, the threshold probability by which the difference in response rates exceeds zero was adapted to 91%, ensuring a 5% overall Type 1 error rate for the trial.

Conclusions

Bayesian statistical analysis of natural history data allowed us to reliably simulate the evolution of symptoms for individual patients over time and to probabilistically compare these simulated trajectories to actual observed post-treatment outcomes. The proposed model adequately predicted the natural evolution of patients over the duration of the study and will facilitate a sufficiently powerful trial design that can cope with the disease’s rarity. Further research and ongoing dialog with regulatory authorities are needed to allow for more applications of Bayesian statistics in orphan disease research.

Recent advances in understanding osteosarcoma and emerging therapies

Osteosarcoma is the most common bone cancer in adolescents and young adults, but it is a rare cancer with no improvement in patient survival in the last four decades. The main problem of this bone tumor is its evolution toward lung metastatic disease, despite the current treatment strategy (chemotherapy and surgery). To further improve survival, there is a strong need for new therapies that control osteosarcoma cells with metastatic potential and their favoring tumor microenvironment (ME) from the diagnosis. However, the complexity and heterogeneity of those tumor cell genomic/epigenetic and biology, the diversity of tumor ME where it develops, the sparsity of appropriate preclinical models, and the heterogeneity of therapeutic trials have rendered the task difficult. No tumor- or ME-targeted drugs are routinely available in front-line treatment. This article presents up-to-date information from preclinical and clinical studies that were recently published or presented in recent meetings which we hope might help change the osteosarcoma treatment landscape and patient survival in the near future.

Latent process model of the 6-minute walk test in Duchenne muscular dystrophy : A Bayesian approach to quantifying rare disease progression

Duchenne muscular dystrophy (DMD) is a rare X-linked genetic pediatric disease characterized by a lack of functional dystrophin production in the body, resulting in muscle deterioration. Lower body muscle weakness progresses to non-ambulation typically by early teenage years, followed by upper body muscle deterioration and ultimately death by the late twenties. The objective of this study was to enhance the quantitative understanding of DMD disease progression through nonlinear mixed effects modeling of the population mean and variability of the 6-min walk test (6MWT) clinical endpoint. An indirect response model with a latent process was fit to digitized literature data using full Bayesian estimation. The modeling data set consisted of 22 healthy controls and 218 DMD patients from one interventional and four observational trials. The model reasonably described the central tendency and population variability of the 6MWT in healthy subjects and DMD patients. An exploratory categorical covariate analysis indicated that there was no apparent effect of corticosteroid administration on DMD disease progression. The population predicted 6MWT began to rise at 1.32 years of age, plateauing at 654 meters (m) at 17.2 years of age for the healthy population. The DMD trajectory reached a maximum of 411 m at 8.90 years before declining and falling below 1 m at age 18.0. The model has potential to be used as a Bayesian estimation and posterior simulation tool to make informed model-based drug development decisions that incorporate prior knowledge with new data.