Three reasons image analysis should be incorporated into your immunotherapy real-world evidence development strategy

Jan 9, 2017 8:00:00 AM

Clinical trials are extremely important to assess the safety and effectiveness of a new therapy or of a currently-available therapy in a new indication. However, there are a few drawbacks to clinical trials, such as:

Both BioPharma and regulators are starting to understand and recognize the value of collecting real world evidence (RWE) to complement registrational studies following the successful market authorization of a new technology. By collecting RWE, investigators are able to obtain valuable data on the remaining 94-97% of patients who do not enroll in clinical studies. This data may then be used to inform new clinical studies for new drugs or new indications, support competitive pricing and/or formulary placement, and identify areas of unmet needs for a subset of patients.

RWE_illustration_170105.jpg

Data sources for RWE often include electronic medical records, pharmacy data, and health insurance claims. However, more and more RWE researchers are integrating other information from social media, and unique to cancer, information about the tumor from genomic and other pathology data. Companies like Foundation Medicine and Guardant Health have partnered with firms like Flatiron Health and IMS Health to incorporate genomic data into databases that allow researchers to draw conclusions they may not otherwise be able to draw.

Nonetheless, even these exciting partnerships are missing one key piece of information:  phenomic data from the tumor and its immune contexture. Definiens’ image analysis and sophisticated analytic software and services allow researchers and BioPharma to accurately extract and quantify relevant data from the cancer tissue section into a format that is easily integrated with other data sets to yield valuable real-world insights.

Image analysis maximizes and optimizes tissue data

The reason that researchers are aggregating data from so many different resources is to maximize the amount of data in the hopes of identifying patterns and draw conclusions. A great deal of data can be extracted from each of the aforementioned data sources of RWE, and image analysis can be used to extract just as much contextual information from the tumor microenvironment, including the identification of cellular morphological features and the calculation of cellular spatial relationships and cell population densities. Cellular spatial relationships and densities have demonstrated to have predictive values in identifying patients who will experience a response to immunotherapies (see O1 Althammer S, et al from SITC 2016), and these relationships and densities (in addition to the morphological features) are difficult, if not impossible, to calculate using manual methods. Further, using advanced machine learning methods, researchers can be assured that the correlation of phenomic data with clinical outcomes will deliver predictive algorithms which are highly accurate and robust with respect to the noise of real world data sources.

Image analysis complements NGS data

The use of NGS to capture DNA and RNA information from tumors has ushered in a new era in medicine. Through this genetic information, BioPharma can better understand how and why targeted drugs work and better identify patients who will respond to these drugs. However, nature selects for phenotype, not genotype. Cancer-relevant phenotype plays out at the tissue level, where both environment and genetics shape the architecture of a tumor and its surroundings. The Tissue Phenome is the manifestation of these interacting influences and, unlike the genome, it includes information about all post-translational and epigenetic modifications above and beyond genes. Phenomic and genomic information together paint a complete picture of the tumor, and when paired together and combined with other RWE sources, only then can valuable holistic insights be developed.

Image analysis standardizes data

One of the hurdles of maximizing the efficiency of RWE data collection is the lack of standardization. Each data source often has a proprietary database where data is collected and stored, and these databases often lack standardization. For example, the electronic medical record industry is dominated by Epic, but some health systems will use Cerner or McKesson (among others), and each system has their own data collection, storage, and reporting structures and formats. 

Researchers often have very limited resources for RWE, even if they are industry-sponsored studies. Image analysis allows researchers to select which measures they are most interested in out of hundreds visually inaccessible options, resulting in the generation of a standardized data base that can be aligned with other data sources to make the process of analyzing all of the aggregated data as efficient as possible. This not only lends itself to standardization of information, but it also allows researchers to confidently compare data between patient populations and cancer indications.

When developing clinical trial designs, researchers are often limited by the number of patients to be enrolled, the timelines imposed by the study sponsors and regulators, and the number of outcomes measures that can reasonably be expected to measure. However, there are a number of questions that still remain after the conclusion of a clinical trial, and the use of RWE can help answer these questions through the generation of more data that is illustrative of the real world faster and cheaper relative to a clinical trial. Some of these questions include (specific to immune-oncology, but relevant for other areas):

  • What is unique about patients responding to monotherapy?
  • Why do some patients respond better to combination therapy?
  • Why do some patients have less durable responses?
  • What is unique about patients who don’t respond to therapy?
  • How does alternative dosing impact response and outcomes?
  • How does alternative dosing impact adverse reactions?
  • What are the costs associated with a given treatment strategy by patient (sub)population?

The answers to these questions can only be found by aggregating all of the relevant data and using sophisticated analytics to identify patterns and draw conclusions. Phenomic data of the tumor and its immune contexture is relevant data that is often missing from RWE studies. Manual methods had made capturing this data nearly impossible, but today’s advances in image analysis can capture this valuable data that will allow BioPharma to develop and market new therapies (or old therapies in new indications) to improve the lives of patients.

Matt Houliston, Definiens Inc.