The Consent Clause That Outlived the Study
Photo: Alpha from Melbourne, Australia, CC BY-SA 2.0, via Wikimedia Commons
On a secure server at a large American university, a longitudinal cohort study that began in the late 1990s still holds decades of clinical measurements, biospecimens, and survey responses. The participants consented to one study, described in one protocol, with one set of risks. They did not consent to secondary research use, to linkage with genomic data, or to the machine-learning analyses that today's investigators propose. The science has moved. The paperwork has not.
This is not a fringe problem. Across the United States, cohort studies, clinical trial archives, and hospital biobanks hold millions of specimens and petabytes of linked health data. Much of it remains scientifically valuable precisely because it is longitudinal and hard to replace. But a substantial share sits behind consent documents drafted before anyone imagined routine data sharing, cloud analysis, or artificial intelligence. The data are not lost. They are legally unusable.
A Legal Freeze With Clinical Consequences
These holdings are not obscure. They underpin research on cardiovascular disease, diabetes, cancer, aging, and rare conditions. When access is blocked, the cost is not merely administrative. It is delayed evidence, duplicated recruitment, and studies that cannot be replicated because the original participants cannot be re-contacted.
The revised Common Rule, effective in 2018, created a broad consent pathway for future research. It did not automatically rewrite older agreements. HIPAA authorizations add another layer, and state law adds more. A consent form signed in 2002 may authorize a specific investigator, a specific institution, and a specific purpose. Reading that language as permission for open-ended reuse would be a stretch. Reading it narrowly can strand data that participants might gladly share if asked.
What Re-Consent Actually Costs
Re-consent is not a mailing. It requires locating participants who may have moved, changed names, or died. It requires staff time, IRB review, translations, accessible formats, and a way to document refusal. It requires re-contact that may itself feel intrusive. For a cohort of 50,000, even a modest per-participant cost becomes a program rather than a project.
The cost is also uneven. Younger participants may be easier to reach through digital channels. Older participants, rural participants, and participants with limited English proficiency may be harder to reach, which can bias the resulting sample. Institutions that lack dedicated research staff may simply decline to try. The data then remain in place, technically preserved and practically abandoned.
The IRB Postcode Problem
In the United States, IRBs interpret old language differently from one campus to the next. One board may treat de-identified genomic data as outside human-subjects rules. Another may demand fresh authorization for any secondary analysis. One may accept a waiver for minimal-risk research. Another may not. The result is a patchwork in which the same dataset can be open at one institution and sealed at another.
This inconsistency is not evidence of bad faith. It reflects genuine ambiguity in old forms and genuine disagreement about risk. But it creates a planning problem for multi-site studies and a fairness problem for researchers at less-resourced institutions. When access depends on local interpretation, science becomes a function of geography.
Can Broad Consent Travel Backward?
Ethically, retroactivity is not a technical fix. People who signed a narrow form made a specific bargain. A later institution cannot simply declare that the bargain was broader than it appeared. Broad consent is also not a blank check. It works best when paired with governance, transparency, and meaningful participant choice.
Several alternatives deserve more serious use. Tiered consent allows participants to choose among levels of sharing. Dynamic consent keeps a channel open for later preferences. Participant advisory boards can review secondary-use proposals. Waivers may be appropriate for minimal-risk research on de-identified data. None of these approaches is cost-free, but each is more honest than treating silence in an old form as permission.
A Path That Does Not Require Betrayal
A credible path starts with an audit. Institutions should inventory legacy consents, classify their permissions, and publish clear rules for what can be reused. They should encode consent terms in machine-readable form so that access decisions do not depend on a single reviewer's memory. They should fund re-consent where the science justifies it, rather than assuming that investigators will absorb the cost. And they should use secure enclaves and honest-broker models where data can be analyzed without being freely redistributed.
At BAM Dataset, the premise is that verified, well-documented data should be discoverable and reusable. That premise does not override consent. It demands better stewardship of the promises attached to data. The alternative is a growing archive of American health research that is preserved, expensive, and silent, not because participants refused to share, but because no one asked them in language that still fits the science.
