George Busby, Allelica CSO & Co-Founder
Randomized clinical trials are the gold standard approach for testing the efficacy and safety of novel therapeutics in real world populations and are essential for bringing drugs safely to market. However, a number of variables make clinical trials challenging. These include patient recruitment and retention, lengthy timelines, and complex protocols. Additionally, regulatory delays, monitoring, and data management and validation can all play a part in making the path to a successful trial outcome difficult to navigate.
In short, clinical trials are expensive and risky. Pharmaceutical companies have a lot to gain by making their drug trials quicker, cheaper, and more effective.
Fortunately, when it comes to super-powering your clinical trial for success, Allelica has a range of products that allow pharmaceutical companies to streamline their trials and accelerate towards successful outcomes. This means that efficacy can be demonstrated at a lower cost — and more quickly — bringing drugs to market faster, thereby realizing the financial benefits of novel therapeutics more rapidly.
Selecting the right patients with population stratification
A key component of clinical trial cost is the number of individuals tested in a trial. Reducing this number enables pharmaceutical companies to lower the cost of a trial. But lowering the number of individuals used in a trial risks the trial becoming underpowered and increases the chances that the trial finds no effect of the drug on the outcome, even if such an effect exists.
Wouldn’t it be great to lower the number of individuals in a trial AND increase the chances of success?
Patient recruitment can be improved by selecting participants based on specific clinical or molecular characteristics. One such method is prognostic enrichment which increases statistical power by increasing the number of patients in a trial who are likely to go on to get disease.
Examples include the clinical trials of the Moderna SARS-CoV2 vaccine, where participants were enrolled who were known to be at high risk of COVD-19 because they lived in areas where the virus was spreading quickly. The overall effect of this type of enrichment is to drive down clinical trial costs by requiring fewer participants to show equivalent statistical power of an intervention.
In a similar way, in trials for drugs for common diseases, we can use biomarkers to identify people who are at higher genetic risk. Assessing biomarkers as an inclusion criterion for trial enrollment means that individuals can be stratified such that only those who are more likely to respond to the drug under trial are included. Equally, if a biomarker has been associated with non-response, these individuals can be excluded to ensure that precious resources are not wasted on individuals on whom the drug will never work.
Stratification of patient populations prior to enrollment is already a common practice. Trial protocols comprise key inclusion/exclusion criteria for trial participants. Allelica’s products allow trial managers to utilize a novel and powerful data source for patient enrollment: genomic screening using Polygenic Risk Scores (PRSs), which measure the influence of common genetic variation on disease and trait outcomes.
Allelica offers two solutions to accelerate the use of prognostic enrichment for clinical trials. Our PREDICT software module can be used to compute hundreds of ancestry-aware PRSs for many common diseases and our DISCOVER software module can be used to build and validate new PRSs when a suitable PRS is unavailable. Predictive PRSs are able to stratify populations, identifying those individuals who are more likely to get disease. For example, Allelica’s breast cancer mortality PRS can predict a woman’s 5 year mortality after breast cancer diagnosis, identifying those women who are more likely to die during the trial, potentially due to more aggressive cancer subtype or other intrinsic factors captured by the PRS.
Demonstrating better drug efficacy
A second approach to increasing trial efficiency is to enrol participants who are more likely to have a greater benefit from the intervention being tested. For example, during a trial for the lung cancer therapeutic gefitinib, molecular profiling of cancer DNA enabled researchers to identify individuals whose lung cancer had specific mutations in the drug’s target receptor gene, and these cancers were shown to respond much better to treatment. As such, tumor profiling is now a common approach to developing oncologic drugs because drugs can be developed to target disease that results from specific mutations.
More effective drugs command higher prices, so any approach that can increase the demonstrated efficacy of drugs will lead to a greater return on investment.
In a retrospective analysis of statin trial data, individuals with a high PRS for Coronary Artery Disease (CAD) were shown to gain a greater effect from the statin than those with lower PRS (Figure B below). By concentrating the trial on these individuals, it is estimated that the same positive effect of the statin could have been estimated with a 90% reduction in clinical trial size.
Adjusting for unmeasured confounders
Clinical trials can be de-risked and failure rates reduced through covariate adjustment. This approach aims to more accurately assess the effect of an intervention by controlling for relevant covariates, improving quantitative evidence for a drug’s effect and for assessing the eligibility of individuals for the trial.
When performing a trial for an oncology treatment, for example, participants may die during the trial or have increased side effects for reasons that have nothing to do with the trial intervention itself. If individuals in the trial will have died earlier anyway, then estimates of the effect of the drug may potentially be diluted and incorrectly estimated. One way to overcome this is to adjust for appropriate prognostic covariates when analysing trial data. As a simple example, the likelihood of death from any cause increases with age so you would want to use information on participant age in any subsequent analysis of the efficacy of a trial’s intervention.
Genetic data can be used to generate prognostic biomarkers that can be used in a covariate adjustment framework. For example, PRSs for disease mortality or for any of the over 1000 traits, diseases and biomarkers available in Allelica’s PRS catalogue can be used to control for underlying variation that can affect intervention effect estimation. Machine Learning Models that include these PRSs as well as clinical variables — and the interactions among them — can be built to extract signal from noise when retrospectively analysing clinical trial datasets. In addition, it is possible to build ad hoc models using multiple modalities and machine learning to build novel predictive covariates and that can subsequently be used to control for effects that may confound the ability of the trial to show true effects of a therapeutic.
Rescuing clinical trials
Around 50% of clinical trials fail at Phase 3, many because of lack of efficacy. These failures have led to billions of lost dollars as well as loss of potentially viable drugs. However, with the right preparation, all is not lost. If cheap genetic data is collected on all trial participants at enrolment, this information can be used retrospectively to rescue failed clinical trials. For example, applying a PRS to the full set of trial participants can identify those who are genetically unable to respond to the drug. Removing these people from downstream analyses can then increase the chances that the therapeutic can be demonstrated to be effective.
Repurposing drugs
Large genomics dataset can be mined with exposure and biomarker PRSs to discover new signals that can lead to repurposing of licensed drugs. Pharmaceutical companies can mine their internal data lakes comprising genetic and matched clinical datasets to extract novel associations between biomarkers and drug performance, leading to new insights on the biological etiology of disease and the identification of drug targets. For example, an analysis of PRSs for cardiometabolic disease and blood plasma proteins showed an association that was able to predict future myocardial infarction and type 2 diabetes events as a result of high protein levels. The causal relationship between the proteins and disease risk uncovered the potential for drugs that were already available to modify protein levels to be used to reduce cardiovascular disease and diabetes risk.
The FOURIER clinical trial randomized 27,564 patients with cardiovascular disease to a placebo or evolucumab, a cholesterol-lowering therapy, and followed patients for a median of 2.2 years. This trial design was based on a power calculation that predicted an event rate of ~6.4% in the control arm and a relative risk reduction (RRR) of 15%. Fahed and colleagues used these data to model power calculations using polygenic score enrichment under either of two models.
(A) With prognostic enrichment (increasing event rates beyond the 6.4% in the original trial), a polygenic score enrichment improves statistical power to detect a benefit despite a fixed effect size (relative risk reduction of 15%)
(B) with predictive enrichment (increasing effect size of intervention beyond the 15% RRR in the original trial), a polygenic score enrichment improves power with a fixed event rate in the placebo arm of 6.4%. The dashed line in both panels denotes 90% power to detect a statistical benefit, a threshold commonly used in trial design. Using polygenic scores to enrich clinical trials could markedly improve power and reduce the number of participants needed by increasing event rates (“prognostic enrichment”) and/or increasing the effect size (“predictive enrichment”).
Your complete PRS-powered clinical trial workflow is readily available from Allelica
From designing your clinical trial through to clinical data management, Allelica has a suite of products and services that can help you superpower your clinical trials. Our expert team is on hand to consult on clinical trial design that utliizes the power of PRSs to optimize your trial resources. Allelica’s PRS software can be used as an enrolment tool that enables genomics-based stratification to be used for trial inclusion and exclusion.
When it comes to Clinical Trial Assay (CTA) development, Allelica has unique capabilities and experience building clinical lab assets that can support your trial goals. We have a number of CLIA/CAP laboratory partners that deploy a suite of genomic tools, from targeted NGS panels, through microarrays to high coverage whole genome sequencing, meaning that we can provide genomics-powered clinical trial solutions, whatever your budget. Finally, all of our products are covered under Allelica’s ISO13485 and ISO9001 certificates, ensuring that our Quality Management Systems (QMS) adhere to the highest industry standards for safety, reliability, and effectiveness.
