Increasingly, researchers are considering the patient when assessing new therapies – “patient centricity.” This is great (obviously) as the patient can have crucial insight into how a therapy/treatment impacts her life. On the one hand, it’s important that a therapy is safe and effective. And yet, often that is not enough. If the patient is left with no quality of life, that is an important trade off to consider.

To address this issue, investigators have turned to Patient Reported Outcomes (PROs). These measure aspects such as “quality of life” and “amount of daily activity.” The idea is to quantify the qualitative and subjective measures that make life what it is. Therefore, many PROs are scales or questionnaires, that turn these measures into reported and computable data points.

However, while there are certainly PROs that cut across any disease area (e.g., general ones like “quality of life”), often the meatier and important PROs are disease specific. For instance, in Parkinson’s disease, a critical aspect is how much a person’s gait may be impacted, since that can affect daily activity. Or in head-and-neck cancer, as we will see, there are life aspects around eating in public or how a therapy affects one’s ability to speak. The neck area is important to these activities, and any impact there adds extra burden to someone’s life.

And yet, the specific PROs to Parkinson’s may not apply to Head-and-Neck cancer, and vice versa.

Pragmatically, such PROs are very important. They help investigators and researchers figure out what to ask patients during a clinical trial; areas of concern for doctors to communicate with patients; and help to hone in on the areas that truly affect ones quality of being, beyond safety and effectiveness that can meaningfully differentiate one treatment over another. Therefore, understanding these therapy specific PROs is an important activity.

It’s also a hard area to understand. PROs are often captured through “Patient Reported Outcome Measures” which are usually some sort of scale or index. Yet, there isn’t a single register of these PRO measures on a disease-by-disease basis, making it challenging to comprehensively understand the possibilities within a disease. To deal with this, researchers often resort to pulling the papers and reading them by hand – noting the PRO measures as they go.

Can AI help? We think so, and so we partnered with a customer to see.

Specifically, we looked at PRO measures related to Head and Neck Squamous Cell Carcinoma. This is a devastating disease and the treatments can end up affecting everyday life activities from eating to speaking to moving around.

Methods and Results

To start our analysis, we first searched the Evid Science platform for all articles related to “Head and Neck Squamous Cell Carcinoma” – This resulted in pulling nearly 125,000 results from around 18,000 papers. Some of the outcomes in these results were PROs, and some were not.

We then did a quick filtering based on some generic keywords for PROs – things like “activity” or “daily.” Our goal was to leverage what we knew about generic PROs in order to find the more disease-specific ones. This filtering pared down our initial set of around 125,000 outcomes to a more manageable set of 2,528 outcomes related to PRO measures. Of course, at this point it’s not unreasonable to simply read through this list (might take half a day?) and keep the ones you are interested in. However, this set is still a bit large and therefore analyzing the set to find the “common” themes is challenging…

So – and because we are an AI company – we added one more step. We took that filtered list and ran clustering over it (clustering is a machine-learning technique whereby similar items are grouped together into distinct “clusters”), in order to group together similar terms. Then we took the frequent terms mentioned in the clusters, and produced a set of disease-specific PRO focus areas. The 26 HNSCC-specific PRO areas are below (on the right):

What’s interesting is that we see these areas are well captured by the specific PROs in our filtered list. For instance, we see high-level PRO areas around Eating, Voice, and Performance. And in the filtered list, these are represented by outcomes such as the “Mean Performance Status Scale: Eating in Public,” or “Less voice disability.” This lets us clearly tie back the themes from the 26 measures to the specific PRO measures associated with them!

Now, not only do we have a good understanding of the features that people care about for the PROs specific to Head and Neck Squamous Cell Carcinoma, we also know which measures relate to which topic. This whole process is a push-of-a-button and is much, much faster than reading thousands of papers by hand.

Conclusion

Need help finding disease specific PROs for your specific area? Engaging in a more patient-centric approach to your clinical trails? Why don’t you give us a shout – we would love to hear from you.

Read Also