Data Sharing Dilemmas
Michelle Mello, JD, PhD
Disclosures
Advisor to Verily Life Sciences, LLC on the design of a product
facilitating safe return to work and school during COVID-19
Collaborator on studies using MPOG data
Potential benefits of data & code sharing
Facilitate secondary analyses to accelerate science
Heighten the ability to evaluate and reproduce studies
Strengthen the culture of sharing and openness in science
Conserve scarce research funding
Increase opportunities for teaching and learning
Honor patients’ and research participants’ contributions
3
Data sharing dilemmas
4
Meaningful
access
Burdens and
incentives
Patient privacy
and consent
Accountability
5
Data sharing policies
Surgery journals in 2019:
• Among 82 journals publishing
trials, 5% required DS statement
and 45% encouraged one
• Statement need not express intent
to share IPD
• Policies only apply to clinical trials
Source: Lombard N et al., Trials 2020.
Funders in 2019:
• 38% of noncommercial funders
had a DS policy; 60% of those
required DS
• 41% of commercial funders had a
policy committing to share their
own data
• Policies only apply to clinical trials
Source: Gaba et al., PLoS ONE 2020.
How is “access” commonly interpreted?
Examined 487 clinical trial reports in
JAMA, Lancet, & NEJM 2018-2020
Dilemmas
7
Meaningful
access
Burdens and
incentives
Patient privacy
and consent
Accountability
Potential commercial harms
• Competitive harm: could help competitors develop competing products
or gain a marketing advantage
• Use in regulatory enforcement or litigation
• Patent issues:
• Early disclosure may be “prior art” that could jeopardize patentability
• May also trigger the patent period to begin running
• Arms competitors with information to beat the discloser to be first to file
• Cost burdens:
• Statistical and clinical staff to evaluate data requests
• Legal staff to draft and enforce DUAs
• Technical staff to prepare data
Who uses shared data?
• dbGaP: 82% not-for-profits
• European Medicines Agency, 2010-2012: 39% industry or
consultant,16% lawyers, 8% academics
• YODA (J&J data): 90% academic, 5% industry
Other incentives problems
• Fairness concerns
• Reduced incentives to contribute data
• Potential to be “scooped” on subsequent analyses
• The PITA factor
• Acknowledgment vs. coauthorship
10
Potential incentives
User fees
Institutional recognition
Individual authorship credit
11
Source: Bierer BE et al. N Engl J Med 2017;376:1684-1687.
An individual researcher (letters
A through F) may be designated
and credited as an author, a
data author, or both, depending
on the person’s contribution to
the data and analysis in the
published work.
DOI = digital object identifier MS
= manuscript
Data authorship
Dilemmas
13
Meaningful
access
Burdens and
incentives
Patient privacy
and consent
Accountability
14
Overall, how do you think the potential benefits of sharing anonymous, individual
clinical trial data weigh against the potential negative consequences?
Negatives strongly
outweigh the benefits
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Benefits strongly outweigh
the negatives
Neutral
8% had a negative view
Source: Mello MM et al., N Engl J Med 2018
15
How likely would you be to allow your anonymous, individual
clinical trial data to be used in the following ways?
0% 20% 40% 60% 80% 100%
To help scientists check the accuracy of research results
announced by other scientists or companies (by re-doing…
To help patients and groups of patients learn more about
health problems that affect them.
To do research on health problems that affect my family or
me.
To help get answers to scientific questions faster using
information that others have already gathered.
To do research that will help others.
To help lawyers prove their case in lawsuits claiming that
medical products are unsafe.
To learn more about diseases that only a small number of
people have (by combining data from many clinical trials).
Very likely Somewhat likely Neither likely nor unlikely Somewhat unlikely Very unlikely
Source: Mello MM et al., N Engl J Med 2018
0 5 10 15 20 25 30 35 40 45 50
The informed consent form should ask me for
specific permission to share my data, separate from
my overall consent to be in the clinical trial (and tell
me how the researchers will protect my identity).
The informed consent form should explain that my
data are going to be shared (and tell me how the
researchers will protect my identity).
As long as the researchers have good data security
protections, data sharing doesn’t need to be
discussed in the informed consent form.
Initial impression After probe
Source: Mello MM et al., N Engl J Med 2018
16
Which of the following best describes how you would feel about being asked for
permission to share your anonymous, individual
clinical trial data with people
outside of the trial?
PROBE: Suppose it’s true that when more people decide not to share their individual clinical trial data, the
scientific value of the remaining data is lower. Please think about whether that would change your answer.
27% of this group would
prefer research not be
conducted if getting
permission isn’t feasible
Source: Cho MK et al., Annals Intern Med 2015
Dilemmas
18
Meaningful
access
Burdens and
incentives
Patient privacy
and consent
Accountability
Accountability strategies
• Strong DUAs
• Incorporate patients into decisions about data uses and transfers
• “Minimum necessary” principle
• Use of “honest brokers”
19
21
Source:
ncats.nih.gov/n3c
Conclusions
22
Meaningful
access
Burdens and
incentives
Patient
privacy and
consent
Accountability
Current requirements
are minimal and
loosely enforced
While security issues
are real, patients
support sharing IPD.
For EHR data, need to
find ways to honor the
desire to give consent.
Incentives problems are
systemic, but MPOG
could adopt expansive
attribution policies.
Promising systems
exist, though
technological strategies
require resources.
