Eve Everywoman is a 45 year old WF that presents to her family doctor with a breast lump she noticed
on self exam. Dr. FP exams performs a general history and physical an exam. He notes a 3-4 cm right
breast mass and orders a mammogram and breast ultrasound. The mammogram is suspicious with
scattered microcalcifications and the ultrasound shows a solid mass. Eve is referred to the local
pathologist for a core biopsy of the mass.
The biopsy shows a lobular adenocarcinoma with Grade 2 histology. Immunohistology shows that the
tumor is ER - , PR – and HER2/neu positive.
Dr. FP discusses the results with Eve and they elect to have her visit the local medical center that has a
breast clinic for further management. Dr FP. Sends an electronic CDA referral to the breast clinic with all
of the pertinent data for Eve.
Eve visits the breast clinic and learns that there is an Transcend I-SPY 2 clinical trial that she is eligible for
and she elects to join the study. Eve is concerned that both her mother and 1 of her 2 sisters also has
breast cancer and worries about her daughter. At the clinic they perform a detailed family history that is
entered into the Transcend Pedigree form and order a BRCA profile and submit the results of that
history to the Hughes BRCA PRO web service.
The Hughes Risk App returns a probability result that shows Eve is at high risk for Ovarian Carcinoma as
well as breast cancer , and she is referred to the Gynecological Oncology Group at the medical center for
consultation. Eve also notifies her second sister to undergo BRCA screening as soon as possible.
The clinical trial uses the Hughes Risk APP BRCA PRO scores as another variable in the analysis of trial
response.
At the end of the trial, the Transcend data for Eve submitted anonymously to an outcomes analysis
warehouse for broader population studies.
So high level, a referral comes to transcend from a clinician that wants their patient enrolled in a clinical
trial. Since Transcend is adaptive, the outcomes assessment changes throughout the trial. We want to fix
a trial start risk status, so as soon as the referral data comes in and we get a pedigree model, we send
the data to the Hughes Risk App service and get a risk ratio returned to document current risk status at a
point in time. This gives us another benchmark to use as a measure of outcome change progress.
So we need to decide what payload modalities work through each exchange to give the necessary data
for decision making through the scenario.
The Hughes Risk app is fixed so that is done. It takes a V3 service payload. The Referral from the doc to
the clinical trial is fixed as well from our perspective and is our referral message. The Transcend to
Outcomes is the debatable one, but I think that should be a V3 service payload as well, capturing the
pedigree model as well as the baseline data from the trial and the serial outcome measures at each
intervention change.
What I would like to suggest as a scenario is that we provide a referral service into Transcend using our
work to date. That we add the Pedigree forms to the transcend system (dev work for Tolven). And that
we build a service interface to the Hughes Risk app.
Finally, we port the cumulative data from Transcend, including the pedigree data to an Outcomes
model.
This would require defining an outcomes service, but it would be scoped to only what Transcend can
provide, so is doable I think
This accomplishes several wins.
First, Transcend, which is concerned with Tumor Marker correlation, now has a way to look at patterns
in family history that do not follow the tumor marker expression, potentially leading to new avenues of
investigation for new markers.
Second, It adds a Bayesian risk assessment to the data in Transcend that gives an additional baseline risk
assessment beyond the ones measured now.
Third it gives us what we have said we want to do all along, which is an outcomes data set.
Finally, it makes a good news splash by going outside the NCI to involve the Harvard -Partners group as
well as the Transcend group.
It is likely that we can push some of this off to Mike Hogarth at UCD to add some functionality on the
Tolven side. That might help us get there faster.
We have the I-SPY2 data elements, Jean has modeled most of them in the Demo Pilot model, so we have
that piece done anyway.