Searching for #clinicaltrials is not easy and how @theNCI is making it better

Disclosure: I work at the NCI, and the opinions on this blog are my own. My work at the NCI focuses on supporting people in cancer research training and is not related to clinical trials or the work on the NCI’s clinical trials database. As readers of this blog know, I am not a medical doctor, but rather a cancer researcher and survivor of breast cancer.

Finding clinical trials that are relevant to a given cancer patient is surprisingly challenging.

First, the more we understand about the biology of cancer, the more fragmented the picture becomes. The very terms personalized or precision medicine convey that the goal is to tailor therapy to the biology of a unique individual. So searching for breast cancer clinical trials will find many that are not relevant to any one patient. You have to add in type, stage and some molecular markers at least.

Second, clinical trials are complex as well. Often the therapies they offer have multiple elements, like drug combinations. Then they have inclusion and exclusion criteria, and may well be tailored to a very specific disease. But that disease may not be based on classical distinctions such as breast cancer or lung cancer, and may instead be based on the presence of a mutation that can appear in several cancers.

Third, the data on patients and the data on trials are not perfect. Here one challenge is that they are mostly not structured, meaning that information is mostly in free text that a human entered. As a result, a given piece of information can be represented in multiple different ways. Even the way you can write the name of a cancer center, or the name of a drug has variations. This makes developing searches hard.

Fourth, the search tools that exist are evolving, but in many cases can lead to unintended inclusions and omissions. Relevant to my attempts to find trials that specifically include men with breast cancer is that a search for trials that include breast cancer as a disease and that permit men to enroll yields many, many trials that should not be included. That is because they allow many cancers besides breast, and that is where the men enroll.

The NCI is doing some important work in this area, and just recently made upgrades to their clinical trials data and the API you can use to search it. I have been using the first version of their API for several months, and am excited with the new developments. Here is what is going on in response to the challenges above:

1. In relation to the complexity of cancers, NCI is of course doing a lot of research intramurally and supporting more across the nation and the world. Similarly, NCI is heavily involved in running and supporting cancer clinical trials.
2. In relation to the lack of structure in the data, the NCI is both developing tools and supporting manual curation of trial data. A key tool is the NCI Thesaurus, which provides a way to code in a unique way something that is represented in many different ways in clinical data. For example, there are 17 ways in which being positive for the estrogen receptor can be represented, and now thanks to the thesaurus they can all be represented by the single unique code C15492.
3. Regarding search tools, NCI has just upgraded its curated data on the clinical trials it supports and the way to search them. Very helpful are the upgrades to the API they publish – think of it as an information desk for algorithms and scripts. You can write a simple program that asks for very specific information from the NCI’s data and then presents it.

As a result of the refinements and improvements NCI has made I have now changed my own search strategy for current breast cancer clinical trials that enroll men. Using the new curated concept of ‘lead disease’ allows me to identify trials that are focused on breast cancer as a disease. Then, if they enroll both men and women the chance that this trials includes men with breast cancer is very high. This search approach finds fewer trials, but the chance that the trials it does find are relevant is higher. You will see this new strategy reflected on my page here listing trials, as well as in my tweets.

Please bear in mind that if you are looking for trials for metastatic breast cancer, our colleagues over at BreastCancerTrials.org have some good search tools as well.

As always, comments welcome.