Why Reducing Complexity Can Lead to Better Clinical Trials

Phesi has released its latest report focusing on how the complexity of clinical trials impacts return on investment (ROI). For example, overly complex trial protocols with numerous outcome measures can lead to longer timelines, higher costs, increased patient burden, and lower success rates.

Using its AI-driven Trial Accelerator platform, Phesi analyzed phase III clinical trial data across five top ranking indications: COVID-19, type 2 diabetes mellitus (T2DM), atopic dermatitis, non-small cell lung cancer, and cystic fibrosis. The analysis looked for trials that were complete and that had reported patient data in the selected time frame. In general, trials with fewer outcome measures reported better site enrollment and shorter completion times. ​

The report’s recommendations range from simplified trial designs, reduced outcome measures and costs, and enhanced data quality. ​To find out more, we spoke to Phesi Founder and President Gen Li.

How did you define “trial complexity” in the report? And how does this complexity impact clinical trial timelines and costs?
 

Trial complexity consists of many different elements in the clinical trial protocol. In essence, a clinical trial protocol describes the patient population that we are inspired to treat, and the data we plan to collect from them over a particular time frame, helping to define the efficacy and safety of the experimental medicine. 

Trial complexity exists in two categories: how we define the patients we are treating, and how much data we are collecting from these patients, such as treatment outcome measures. In the initial analysis that was the basis of this new report, we chose to focus on the element of outcome measures to demonstrate how just one area of unnecessary complexity significantly impacts trial success. 

Overly complex protocol designs prolong cycle times by months, adding expense and reducing return on investment. Our analysis shows that as the number of outcome measures in a trial protocol increases, the percentage of reported outcomes decreases. On average, more than a third (35 percent) of outcome measures were not reported. Trials with fewer-than-average outcome measures reported 94 percent of these in their results, while those with more than the median number of outcomes typically only reported 56 percent. The more outcome measures included in a protocol, the higher the percentage of measures reported in clinical trial results. These figures underline that the number of outcome measures can and should be optimized.

The report mentions over-collection of outcome measures. What factors typically lead to this excessive data collection?
 

Investigators can fall victim to data FOMO (“fear of missing out”), which leads them to want to collect as much data as possible while they have access to a patient. This results in trial protocols having many outcome measures that are redundant or duplicative. For example, trials often use several different but very similar physical performance status measures on the same patients, which places unnecessary pressure on both patients and investigator sites.

Collecting a robust amount of data is essential, but it must be the right data collected at the right time. To prevent over-collection, sponsors should harness insights from historical and existing trial designs to understand the impact of each measure and be sure they are using an optimized number of outcome measures.

How does trial complexity influence patient recruitment and retention?
 

There is a strong correlation between trial complexity and patient recruitment. Trials with a lower number of outcome measures tend to have better enrollment, as streamlined protocols reduce the burden on both patients and investigator sites. For example, in a comparison of type 2 diabetes trials, we found that a sponsor with an average of 25 outcome measures had lower site enrollment performance (10.2 patients per site versus 11.2) and a slower enrollment rate (0.46 patients per site per month versus 0.53) than a sponsor using 10 outcome measures on average.

Simplifying trial complexity and reducing patient burden are two sides of the same coin. By leveraging data from ongoing and historical trial protocols and amendments, investigators can gain a thorough understanding of the target patient population so they can select outcome measures with greater precision. 

Could you explain the “complexity score” and how it assists in optimizing trial protocols?
 

Trial Accelerator is an AI-driven platform that holds data on 4,000 diseases and more than 120 million patients. To obtain a complexity score, we analyze existing and historical trial protocols for the chosen indication, and define the media number of outcomes. 

The median number of outcome measures from past, similar clinical trials are those in the “green zone.” When the outcome measures are close to Q1 and Q3 of the median value, they are in the “amber zone.” Once a trial’s measures fall outside of Q1 and Q3 of the median, you enter the “red zone.” You want to avoid the red zone! When sponsors select reference protocols, it is important that they focus on those that successfully recruited patients and reported patient data. However, only 35 percent of trials registered on ClinicalTrials.gov published results within 24 months of study completion, and a significant proportion of trials – ranging from 25 percent to 50 percent – continue to be unpublished. Sponsors must base protocol design on the best available studies and the trials that have successfully reported patient data.

We can combine our approach of red, amber, and green zones with other elements of trial complexity. The aim is to give our clients a view of how aligned their study is with industry medians, and to help them understand what to avoid including in a protocol. 

Do you have examples of how reduced trial complexity can lead to cost savings?
 

In the current economic landscape, where the average R&D costs to progress a drug from discovery to launch exceed $2 billion, improving ROI is essential. Being more precise with outcome measures will make it easier to select investigator sites, recruit more patients quickly, and collect better quality data from those patients – all of which will lead to improved ROI.

Take one example from our analysis of a T2DM study. One trial planned 38 outcome measures in its protocol, an outlier for a median of 10 outcome measures. While originally planned to complete in 19 months, it overran by seven months. In reporting the results for this trial, only 19 of the 38 outcome measures were reported. During that seven months, the sponsor needlessly accumulated hundreds of thousands of dollars in site activation and maintenance costs.

What future trends do you foresee in clinical trial design?
 

The pharmaceutical industry has made significant strides in precision medicine, which has stratified patient populations. However, the same level of precision has not yet been fully applied to trial design, particularly in the selection of patients and investigator sites. The clinical development industry must address this gap – and data science is the key to doing that. 

As an industry, we have access to reams of data at our fingertips – but these data are not being used effectively to improve trial design and implementation. By leveraging predictive analytics, scenario modelling, and data-driven insights, sponsors can simulate the implementation of a clinical trial, eradicating the reliance on “gut feel” that defines a lot of decision making in clinical development today. This shift will lead to streamlined trial protocols, reduced complexity, and higher trial success rates – bringing true precision to clinical development.

Gen Li is a regular contributor to The Medicine Maker. You can read more of his contributions here.