OHDSI Home | Forums | Wiki | Github

Collaboration opportunity with National Cancer Institute's CISNET


(Patrick Ryan) #1

Team:

This week, I have the honor to present about OHDSI at the National Cancer Institute’s Cancer Intervention and Surveillance Network (CISNET) annual meeting. I hadn’t known much about CISNET beyond what @Andrew had introduced me to a few years ago, but in learning more, I’m very impressed by their community of collaborators who are applying advanced mathematical models to inform clinical care and health policy decisions across 6 cancers: lung, breast, cervix, esophageal, colorectal, and prostate. More information about CISNET is available here, including open funding opportunities. The group seeks to fill the evidence gap from whats known from clinical trials and observational studies through simulation modeling which uses prior studies as inputs to produce outputs that extrapolate to scenarios not previously studied.

In my talk, I plan to highlight various activities within our community that may be of interest to CISNET, including: @rchen’s prior work with NCI to phenotype cancers and characterize treatment pathways, the Women of OHDSI’s study to predict breast cancer amongst those with negative mammagraphy (@aostropolets @MauraBeaton @krfeeney), the advances in standardizing data with the Oncology module (@mgurley @rimma @rtmill), and large-scale population-level estimation studies via LEGEND (@msuchard @schuemie @hripcsa). I also wanted to demonstrate the value of network analyses using our standardized analytics (@gregk @pavgra @Chris_Knoll), so I designed a clinical characterization study that covers the cancers of interest to NCI CISNET. If anyone would like to run this analyses on their data to test the feasibility of studying these cancers, I would appreciate it as it may facilitate future collaborations between our communities. Also, if anyone else has any other points of connection that you think I should introduce to the CISNET team, please let me know. I am excited by the opportunity we all have to improve health by empowering our communities to collaboratively generate the evidence that promotes better health decisions and better care.


(Thomas Falconer) #2

Hi @Patrick_Ryan, this sounds great! I’d be happy to run this characterization feasibility study on Columbia’s data.


(Sarah Seager) #3

Hi @Patrick_Ryan - well done on securing the spot at CISNET’s annual meeting. Here at IQVIA, we’d be happy to run your study- count us in! :slight_smile:


(Andrew Williams) #4

Patrick, thanks for making this connection! They are a great and very influential group. Three things you might cover in addition to the work you’ve listed are:
A) The geographic breadth of OHDSI data. They have had international data partners already, but I don’t think they’ve been on the scale and have the transparency about quality and the data granularity that the OHDSI community could potentially provide. Their target decisions are often national-level policies. Outcomes from policy-driven variations across countries in screening, diagnosis, and treatment will be very valuable to them.
2) The transparency of data quality and granularity of OHDSI data. You know the value of data quality transparency since you did so much to engineer it. The granularity of data extends beyond what’s already done in the oncology extension on standard representation of conditions, treatment episodes, and treatment regimes. It includes the genomic data extension that Chan and colleagues have developed and that he and Meera are leading the testing of in the Oncology Genomics subgroup. The robust capacity for reliable information extraction and normalization enabled by the work of the NLP workgroup are also relevant to the quality and granularity of the data and the phenotypes they’ll be able to reliably characterize. The ability to characterize the performance of phenotypes efficiently and accurately via pheValuator seems worth mentioning. The emerging capacity to specify devices and related regulations might be too immature to feature in your discussion, but might be worth mentioning as future areas of potential collaboration.
3) Predictive modeling and health economics work in the OHDSI community. The great work on standardizing the PLP approach fits with CISNET’s interest in standardizing across models. They have great sophistication in developing cost effectiveness models and costs are an important dimension of the policy decisions they are interested in informing. Though health economics isn’t the most developed set of OHDSI analytics, it is there, and collaboration with CISNET or one of it’s contributing modeling groups might be a great way to drive that work forward.


(Andrew Williams) #5

Our IRB needs a more detailed description of the goals of the analyses to approve it as a “prep to research” analysis. Can you put that into a couple sentences?


(Patrick Ryan) #6

The intent of this analysis is to assess the feasibility of the OHDSI network in defining cohorts for the 6 CISNET targets of interest, and to characterize the baseline characteristics of those cohorts to understand the available data that could be potentially used in future OHDSI-CISNET collaborations.


(ruth etzioni) #7

Thank you Patrick for a fantastic presentation and the seemingly boundless positive energy that you brought to the CISNET meeting today. Warmly – Ruth Etzioni, CISNET prostate group PI


(ruth etzioni) #8

Andrew, thanks for the complimentary comments about CISNET - I am the prostate group PI and am brimming with ideas for projects. Your comments embolden me to put them out there.
thank you


(Andrew Williams) #9

Welcome to OHDSI Ruth! I am so glad that Patrick introduced OHDSI to you (who better?) and that you see the potential for moving your work forward in the community. I can’t wait to hear the ideas that you want to pursue and hope we can collaborate.


(ruth etzioni) #10

First collaboration ideas for work with the CISNET cancer modeling group.

To everyone who has shown an interest in the Patrick’s post about speaking at CISNET (thank you Patrick!) I just wanted to respond to you specifically regarding some first ideas about work that would be of great interest to the CISNET consortium. I will post these more broadly but am just starting here.

The defining goal of the CISNET consortium was to explain and learn from population trends in cancer incidence and mortality.

In prostate cancer, we were tasked with quantifying the roles of PSA screening and primary treatment changes in prostate cancer mortality declines (mortality rates have by now gone down 50% since their peak in the early 90s). (this is the explaining trends bucket)

In the learning from trends bucket, we used a population version of statistical models for estimating cancer latency from individual-level screening histories to learn about overdiagnosis associated with incidence trends.

(Prostate cancer has a most interesting incidence trend - almost doubling after PSA screening was introduced, declining pretty rapidly, then increasing and declining again.)

We are working on renewing our CISNET prostate cancer grant and are seeking collaborations and data to help us understand:

(1) patterns of uptake of new treatments for newly-diagnosed metastatic prostate cancer and their associated costs

(2) trends in metastatic progression among prostate cancer cases, and

(3) patterns of cardiovascular disease (a key toxicity of systemic treatments for prostate cancer that suppress androgens) following adminstration of these treatments and in relation to prior cardiovascular comorbidity.
(4) trends in imaging among prostate cancer cases across the post-diagnosis continuum of care.

Our grant includes support for a co-ordinating center which will have a modest amount of funding to cover data and related collaborations each year and also for rapid response projects that address evolving issues in the field. We are developing plans for co-ordinating center partners who typically work with the group for up to two years of the five-year project.

I have other ideas for research projects beyond prostate cancer, relating to trends in patterns of care for ovarian cancer and metastatic melanoma but will include those in my broader post.

If anyone is interested in collaborating on any of these projects I’d love to hear from you. If you have interest in one of the other CISNET cancers (breast, lung, colorectal, cervical oesophagus), I can connect you with those groups as well.

thank you!


(Andrew Williams) #11

Ruth, These are great ideas! We are definitely interested!

As I suspect Patrick covered in his presentation, there is a very active Oncology workgroup in OHDSI. We are proving out a new extension to the OMOP CDM specifically designed to accommodate cancer-specific data structures and representations. That involves use cases such as yours which are valuable both for the evidence they generate and for the requirements that they specify. These requirement drive the work on data structures, vocabularies, analytic methods, etc.

We are currently finishing up the remaining tasks required to support initial proof of concept analyses - basic survival analyses that draw on tumor registry data. Rimma presented the oncology extension and POC analyses at this year’s OHDSI symposium.

Even for such basic analyses we want to use best practices and draw on the formidable analytic expertise in the community. And we are getting excellent guidance from George, Marc, Martijn, Adler, and others on the best practices for that. E.g whether and how to take an approach similar to the one used in LEGEND that exploits the wealth of covariates available in most OMOP instances to control bias from right censoring.

After we redo those POC analyses, we are eager to get to work on other network studies. Yours would certainly be among the ones we could take on and use to define the data, informatics, and analytic requirements.

So, there are at least a couple of ways for you to advance your work through engagement with OHDSI. Clearly there is value in what you’ve done already by posting your ideas to find data partners. There’s also a very exciting potential to do that in a way that helps advance the community’s capacity to accomplish that kind of study generally by building out reusable capacity for these analyses. That would involve by making some of your use cases ones that we design informatics and analytics around by joining some activities in the Oncology WG or contributing to discussion around best analytic practices.

It might also involve engagement with the newly combined Population-level Estimation and Patient-level Prediction workgroups that are devoted to innovating and evaluating best practices and developing the associated tooling for those.

If you are interested in finding out more about the different workstreams in the OHDSI oncology workgroup I’d be happy to communicate directly on a call or you can reach out to the marvelous Shilpa who is valiantly coordinating all the workgroup’s efforts. The conversations about best practices has been happening so far on a smaller email thread. Let me know if you would like to be included in that.


(ruth etzioni) #12

Andrew, thanks very much for taking the time to explain all this. I would love to join the Oncology Workgroup - how do I do that? Can we plan a call to discuss?


(Andrew Williams) #13

Wonderful. Please email me (awilliams15 at tuftsmedicalcenter.org) with times that are convenient for a call next week.


(Alex Butler) #14

Hello @retzioni , my name is Alex Butler and I’m a medical student at Columbia University working with @chunhua . Our work focuses on clinical trial eligibility criteria and we are currently exploring a number of new areas, including how these criteria change over time and what role they play in the real-world clinical space. Speaking personally, I worked for 2 years at Memorial Sloan Kettering in the Prostate Cancer Service before starting medical school so this is definitely an area of interest for me. We would be happy to speak about some ways to collaborate! Please feel free to reach out to us at amb2453 (at) cumc.columbia.edu and we can find some time to speak!


t