Background

With the growing success of precision medicine (PM) therapies across multiple types of cancer, genomic profiling of tumors is becoming more commonplace in cancer care ^1–5. However, despite the promise of cancer PM, genomically-driven trials continue to have low participation rates ^6–10. Low clinician awareness of eligible trials, performance status, and patient attitudes all affect patient participation in genomically driven trials ^11–13. Several trial matching platforms have already been developed, but these platforms are proprietary and cannot easily be adopted by other institutions ^14,15. Thus, an open-source informatics system with proven clinical impact would be useful to support PM trial enrollment.

We developed MatchMiner, an open-source trial matching platform, at Dana-Farber Cancer Institute (DFCI) as a way for clinicians and trial staff to manage an increasing amount of patient and trial data for PM trials. With the introduction of MatchMiner, clinicians and trial staff now have a way to view available PM trial matches for patients in real time. Here, we summarize MatchMiner’s features and impact on patient care from our recent manuscript.

MatchMiner features

MatchMiner has several modes of clinical use: (1) patient-centric, where clinicians look up patients to view all trial matches for that particular patient, (2) trial-centric, where clinical trial teams identify patients for their particular PM trials of interest, and (3) trial search, where clinicians manually enter search criteria of interest to identify available trials based on external genomic reports. For patient-centric and trial search modes, we integrated MatchMiner into the DFCI electronic health record (EHR) for ease of use in the clinic (see Supplemental Fig. 5 from Klein et al., 2022). MatchMiner serves as a pre-screening tool since not all trial eligibility criteria are included in the matching process and there is no consideration about a patient’s readiness to participate in a trial. For data integrity and security standards, MatchMiner meets HIPAA requirements when deployed within an institutional firewall.

MatchMiner matches patients to trials via the MatchEngine, an algorithm that computes trial matches based on patient genomic and clinical data and PM trial eligibility criteria (Fig. 1). The MatchEngine accepts many different data inputs for patient-trial matching and can be adapted to data available at any institution (see Klein et al., 2022 for more information on data inputs). Daily trial open/closed status updates from a clinical trial management system can be coupled with MatchEngine runs for real time trial match updates.

Figure 1. MatchMiner overview of data flow into the MatchEngine. The algorithm integrates data inputs from patients and trials to create a patient match record. (Adapted from Klein et al., 2022)

Trial eligibility is encoded in a human-readable markup language called clinical trial markup language (CTML). Eligibility criteria including genomic, clinical, and demographic criteria, are encoded with nested Boolean logic that allows for inclusion and exclusion clauses. This format is compatible with many data types and allows for complex trial curations and precise trial matching. At DFCI, all trials are expertly curated and reviewed by a team of scientists before being entered into MatchMiner. Alternatively, an open-source curation platform (MatchMinerCurate) allows a user to pull in criteria from clinicaltrials.gov¹⁶ – this can be helpful for extracting basic trial information, but specific genomic and clinical eligibility may need to be entered manually.

MatchMiner’s impact on patient care

At DFCI, MatchMiner impacts patient care by providing PM trial options and fostering timely trial accrual. As a pre-screening tool for PM trial matches, MatchMiner shows clinicians all potential current PM trial options at their institute. This is useful as institutional trial portfolios are continually changing. Among patients who had cancer genomic profiling at DFCI, we found MatchMiner provided an average of 6 trial matches for each patient and 80% of patients matched to at least one trial. Thus, most genomically profiled DFCI patients had MatchMiner trials available for treatment consideration.

So far, we have identified 166 trial consents that were facilitated by MatchMiner. To characterize the impact of MatchMiner on trial enrollment, we measured the time from when a patient’s genomic report entered MatchMiner to the time when a patient consented to a PM clinical trial, and compared this time for our 166 MatchMiner consents to a control group of consents not facilitated by MatchMiner. We found MatchMiner decreased a patient’s time to consent to a PM trial by a median of 55 days. Timely trial enrollment is particularly important because patient performance status can decline rapidly ^12,13. In addition, recruiting for PM trials can be particularly difficult due to the abundance of genomic and trial eligibility data required for enrollment. We find MatchMiner may provide clinical impact by accelerating time to consent for PM trials. Indeed, we highlighted two case studies where MatchMiner positively impacted patient care. In both cases, clinical staff viewed the trial match in MatchMiner 2-3 days before trial consent. After trial enrollment, these patients had reductions in index lesions, demonstrating timely matching through MatchMiner can positively affect patient outcomes.

Lastly, we assessed MatchMiner’s broader impact on trial enrollment at DFCI by identifying the proportion of all PM trial consents that were attributed to MatchMiner. Out of 847 total consents for 354 PM trials that had been curated into MatchMiner, 20% (166) were facilitated by MatchMiner. PM trial participation for patients with profiled actionable variants has been historically low (10-15%)^6–10. Considering that MatchMiner appears to have increased PM trial consents by 24% (847/681=1.24), we therefore find the impact of MatchMiner to be promising for the future of PM trial recruitment. 

MatchMiner is open source

MatchMiner is open source and has already been adopted by other institutions. Data requirements are similar to those required for local installations of other popular open-source software platforms, such as the cBioPortal for Cancer Genomics ^17,18. Indeed, MatchMiner can also be integrated into other clinical systems (e.g. EHR), or used in a research context (e.g. MatchMiner-derived trial matches can be visualized within the cBioPortal). For those interested in setting up MatchMiner at their institution, the MatchMiner API, UI, and MatchEngine code is available at https://github.com/dfci/matchminer. To try out a demo instance of MatchMiner, visithttps://matchminer.gitbook.io/matchminer/deployment/deployment-with-docker. If you have any questions, our team at DFCI can be contacted at https://matchminer.org/. Link to publication: https://www.nature.com/articles/s41698-022-00312-5. 

References

1. Roper, N., Stensland, K. D., Hendricks, R. & Galsky, M. D. The landscape of precision cancer medicine clinical trials in the United States. Cancer Treat. Rev. 41, 385–390 (2015).
2. Ahmad, A. & Gadgeel, S. M. Lung Cancer and Personalized Medicine: Novel Therapies and Clinical Management. (Springer, 2015).
3. Halim, A. Biomarkers, Diagnostics and Precision Medicine in the Drug Industry: Critical Challenges, Limitations and Roadmaps for the Best Practices. (Academic Press, 2019).
4. Von Hoff, D. D. & Han, H. Precision Medicine in Cancer Therapy. (Springer, 2019).
5. Jürgensmeier, J. M., Eder, J. P. & Herbst, R. S. New strategies in personalized medicine for solid tumors: molecular markers and clinical trial designs. Clin. Cancer Res. 20, 4425–4435 (2014).
6. Zehir, A. et al. Mutational landscape of metastatic cancer revealed from prospective clinical sequencing of 10,000 patients. Nat. Med. (2017) doi:10.1038/nm.4333.
7. Tsimberidou, A.-M. et al. Personalized Medicine in a Phase I Clinical Trials Program: The MD Anderson Cancer Center Initiative. Clinical Cancer Research vol. 18 6373–6383 Preprint at https://doi.org/10.1158/1078-0432.ccr-12-1627 (2012).
8. Stockley, T. L. et al. Molecular profiling of advanced solid tumors and patient outcomes with genotype-matched clinical trials: the Princess Margaret IMPACT/COMPACT trial. Genome Med. 8, 109 (2016).
9. Meric-Bernstam, F. et al. Feasibility of Large-Scale Genomic Testing to Facilitate Enrollment Onto Genomically Matched Clinical Trials. J. Clin. Oncol. 33, 2753–2762 (2015).
10. Sholl, L. M. et al. Institutional implementation of clinical tumor profiling on an unselected cancer population. JCI Insight 1, e87062 (2016).
11. Ersek, J. L., Black, L. J., Thompson, M. A. & Kim, E. S. Implementing Precision Medicine Programs and Clinical Trials in the Community-Based Oncology Practice: Barriers and Best Practices. Am Soc Clin Oncol Educ Book 38, 188–196 (2018).
12. Lara, P. N., Jr et al. Prospective evaluation of cancer clinical trial accrual patterns: identifying potential barriers to enrollment. J. Clin. Oncol. 19, 1728–1733 (2001).
13. Galvin, R., Chung, C., Achenbach, E., Dziadkowiec, O. & Sen, S. Barriers to Clinical Trial Enrollment in Patients With Pancreatic Adenocarcinoma Eligible for Early-Phase Clinical Trials. Oncology 34, 407–412 (2020).
14. Eubank, M. H. et al. Automated eligibility screening and monitoring for genotype-driven precision oncology trials. J. Am. Med. Inform. Assoc. 23, 777–781 (2016).
15. Meric-Bernstam, F., Farhangfar, C., Mendelsohn, J. & Mills, G. B. Building a personalized medicine infrastructure at a major cancer center. J. Clin. Oncol. 31, 1849–1857 (2013).
16. matchminer-curate: MatchMiner Curation User Interface. (Github).
17. Cerami, E. et al. The cBio Cancer Genomics Portal: An Open Platform for Exploring Multidimensional Cancer Genomics Data. Cancer Discov. 2, 401–404 (2012).
18. Gao, J. et al. Abstract 5277: The cBioPortal for cancer genomics and its application in precision oncology. Cancer Res. 76, 5277–5277 (2016).