Using health technology as an example, a new article in Harvard Business Review gives data-based tips on how to succeed with regulators - and some may apply to payers as well.
The original article is Pahnke et al. in Strategic Management Journal: Strategic positioning and accelerating regulatoryclearance for new ventures, here. It's covered and summarized in Harvard Business Review by Blanding (11/18/2026) here. I got both open access without registration today.
###
AI CORNER
###
A new study of 239 medical device startups finds that their choice of FDA predicates is associated with the speed of 510(k) clearance. For a first device, a focused comparison to a familiar product in the same category appears helpful; for a second, the firm’s own cleared device may be the strongest reference. The findings illuminate how human categorization operates inside regulatory review, but multiple predicates may also signal a genuinely complex product. We consider what this research suggests for MolDX—and why a clear analogy alone cannot produce payer coverage.
_____
How to Make a New Product Legible to a Gatekeeper
A business-school study of FDA predicates, human cognition, and the harder analogy to diagnostic coverage
The first thing a company says about a novel product is often a comparison. It is “like” an established product, except for the feature that makes it worth buying. Such analogies help an investor decide what market to imagine. They can also help a regulator decide what work lies ahead. But these audiences have different jobs, and the same analogy need not work equally well for both.
In a new Strategic Management Journal paper, Emily Cox Pahnke, Tiona Zuzul, and Michael Howard ask a surprisingly concrete question: When a medical-device startup identifies earlier devices in its FDA 510(k) submission, does its choice of comparisons affect how long clearance takes? Their answer is yes, with a striking change between a firm’s first and second products. A shorter HBS Working Knowledge article presents the managerial version: make the comparison easy for a busy reviewer, and build a product sequence that can use its own earlier clearance. The underlying research article is more qualified than a slogan about “use one predicate.” sms.onlinelibrary.wiley.com
What exactly did Pahnke and colleagues study?
The authors assembled the first 510(k) clearances of 239 US venture-backed medical-device firms from 2001 through 2019 and examined subsequent clearances. They hand-coded the predicate devices named in FDA correspondence, using a broader collection of more than 60,000 cleared-device records to identify which earlier products were especially prominent within their FDA product-code categories. They combined this analysis with 55 interviews, including founders, device executives, regulatory specialists, investors, and regulators. The investors supplied context about startups and milestones; investor decisions were not the measured outcome. The outcome was the number of days between submission and FDA clearance. sms.onlinelibrary.wiley.com
This is a useful setting because a 510(k) applicant must identify an already marketed predicate and make a substantial-equivalence case. The firm has some choice about which device or devices to name. The reviewer has to understand whether the new device fits the applicable regulatory framework and whether its differences raise new questions. A familiar product analogy is thus part of a formal review task, not just marketing copy.
The researchers modeled time to clearance and adjusted for such factors as FDA review panel, venture funding, patenting, and aspects of the cited predicates. Their analysis describes associations among devices that ultimately received clearance. It does not randomize predicate choices, measure the fate of every unsuccessful submission, or establish that changing the words in an otherwise identical application would cause a specific number of days to disappear.
The first-device findings: fewer, closer, and familiar
Fewer references were associated with faster clearance
For a firm’s first device, each additional cited predicate was associated with an approximately 8% lower instantaneous rate of clearance in the authors’ Cox model (hazard ratio 0.92; p = .065). That is a borderline statistical result, and “8%” is a change in the modeled clearance hazard, not an 8% increase in the total number of review days. In the authors’ descriptive comparison, the median was 120 days with one predicate versus 133 days with eight. This suggests a cost to expanding the comparison set, but the data do not show that every company could simply drop seven predicates and save 13 days. sms.onlinelibrary.wiley.com
The category of the comparison mattered
Citing at least one external predicate with the same FDA product code as the applicant’s device was associated with faster first-device clearance (hazard ratio 1.46; p = .043). The reported median times were 102 versus 112 days. A reviewer can more readily use a familiar set of product-specific requirements when the new and old devices occupy the same regulatory category. “It is another one of these” gives the reviewer a starting framework; a comparison across categories may require explaining why a different framework applies. sms.onlinelibrary.wiley.com
An exemplar could help
The authors defined a category exemplar empirically: a device frequently cited as a predicate within its category. First devices citing such an exemplar had a higher modeled clearance rate (hazard ratio 1.65; p = .007), with reported medians of 90 versus 107 days. This was not necessarily the device with the most impressive clinical claims. It was a salient, frequently used reference that reviewers could plausibly recognize. Familiarity may make the comparison easier to verify, although the study cannot directly isolate familiarity from all the other properties of an often-cited device. sms.onlinelibrary.wiley.com
The second-device finding: your own history becomes an asset
For a firm’s second device, citing its own first cleared device was associated with much faster clearance than relying on external exemplars. The reported hazard ratio for self-citation was 7.8 (p < .001), and the descriptive medians were 86 versus 100 days. The large hazard ratio should not be read as “seven times fewer days”; it is a modeled rate of reaching clearance at a given time. It may also reflect how closely a second device resembles the first, not merely the persuasive power of the company’s name. sms.onlinelibrary.wiley.com
Still, the strategic logic is compelling. A prior clearance gives the reviewer a documented regulatory starting point, and the manufacturer knows both generations well enough to specify the differences. The paper’s dynamic insight is that the best reference can change as a company acquires its own record. A startup cannot cite a device it has not yet cleared; later it may be able to build a coherent family rather than begin each submission with somebody else’s exemplar.
Cognitive psychology, with a regulatory job attached
The paper is an application of familiar ideas about human categorization. We recognize a new thing by comparing it with known examples. A salient exemplar supplies a quick mental model. A same-category example helps us apply the right criteria. Too many competing examples can increase the work of reconciling differences. These are hypotheses about cognition, made consequential because an FDA reviewer must actually complete a comparison and may need to send follow-up questions.
The unexpected part is that attention is already assured. An investor, customer, or analyst may never look at your company at all. There, an analogy may need to attract attention, establish a recognizable category, and preserve some distinctiveness. FDA’s assigned reviewer must process the submitted device under a codified pathway. The useful analogy is the one that clarifies the evaluation task. The paper discusses investors as one kind of market audience and interviews them about startup pressures, but it does not show that investors favor one predicate, same-code devices, or self-citation. Those are FDA clearance findings.
A caution about the “too many predicates” story
There is a serious alternative explanation for the first result. Imagine a device with six material features for which the regulatory team believes it needs five earlier clearances to explain the relevant technology. Review may be slower because the device combines several functions and raises several regulatory questions. Its five references are then a symptom of the underlying complexity and the structure of the 510(k) substantial-equivalence task. The delay is not necessarily a psychological penalty imposed because someone wrote down five names.
The authors do include controls and additional tests for breadth across product codes and review panels; they do not find that their particular complexity measure explains away the result. That is reassuring, but a count of codes, panels, or patents cannot fully measure why a specific engineering feature required a second or fifth predicate. Nor can observational data supply the counterfactual submission for the identical device with a shorter predicate list. The practical advice is therefore choose the smallest defensible, coherent comparison set, not “delete predicates until only one remains.” FDA substantial equivalence and the device’s actual characteristics determine what is defensible. If a product truly spans several regulatory concepts, the company must solve that regulatory problem, not conceal it with elegant prose.
Where else might the idea apply?
Pahnke and colleagues mention other institutional gatekeepers and comparison regimes, including EPA registration, FAA replacement parts, generic drugs, and patent examination. Their empirical results, however, come from one FDA pathway: 510(k). A generic drug application prizes a different sort of sameness; a patent examiner cares about novelty as well as prior art. It would be a mistake to transfer the numerical hazard ratios to those settings. The portable question is: What comparison is the evaluator required to make, and which reference makes that work tractable?
For a startup, there are two additional applications. First, sequence the pipeline with regulatory continuity in mind: an initial product that can be reviewed cleanly might later become an internally controlled reference for related products. Second, distinguish the audience in every presentation. An investor pitch can emphasize the market’s size and the product’s difference; the regulatory submission must show how the device satisfies its pathway’s requirements. Those messages can both be true without being identical.
Now try the analogy on MolDX—and mark where it breaks
MolDX is a gatekeeper, but coverage is not a 510(k)
A molecular laboratory approaching MolDX also wants an evaluator to understand a new test. It may help to start with one familiar clinical use case and one close coverage precedent, then make the differences explicit. For example: “This assay informs the same treatment decision in this defined cancer population as the already covered class; here is the evidence that its result changes management and improves the relevant outcome.” That is more reviewable than a deck claiming the assay is simultaneously analogous to five different tests across screening, diagnosis, prognosis, therapy selection, and surveillance.
But a covered assay is not an FDA predicate. Another laboratory’s Z-code or LCD does not confer substantial equivalence, and MolDX need not cover the new test because it resembles a covered one. Coverage turns on the precise patient population, intended use, clinical validity, clinical utility, evidence standard, and medical-necessity language. A single assay can warrant different coverage decisions for breast, lung, and colorectal cancer. An analogy that collapses those indications may obscure the very decision MolDX must make.
Which lessons are most germane?
Start with a single decision, population, and precedent. Identify the clinical decision first: which patient, at which point in care, and what changes when the result is positive or negative? A close covered use case can orient the reviewer. State the boundary of that analogy immediately. If the evidence is strong for one cancer and preliminary for another, present separate indication-specific cases rather than asking the reviewer to infer a pan-cancer claim from a collection of publications.
Make the evidence map follow the decision. Put analytic validity, clinical validity, clinical utility, guidelines, and any relevant coverage criteria under the same intended use. Additional examples should answer particular objections, not multiply the claimed identity of the test. A familiar comparator may ease comprehension, but it cannot replace missing utility evidence.
Use the lab’s own prior coverage carefully. A covered earlier assay or indication may establish operational familiarity and a useful history. It is not an automatic sequel entitlement. Specify what is held constant and what has changed: assay, specimen, disease, stage, clinical action, and evidence. The dramatic second-device self-citation result is a hypothesis generator for this conversation, not a measured MolDX effect.
Find the actual decision mechanism. MolDX review may involve technical assessment, Z-code identification, an LCD or article, contractor processes, and claims implementation. A concise analogy can reduce a reader’s effort at one point in that chain. It cannot make all of those mechanisms operate on a startup’s preferred timetable.
The crucial difference: FDA must work your file; the payer may never take up your proposed decision
This is where my September 9 essay, “You Better Learn the Biggest Myth About Payor Coverage of Diagnostics,” matters. I argued that the comforting story—publish good papers, get a payer meeting, persuade the medical director, win unexpected coverage for this one test—is usually a poor business plan. Evidence matters, but publications do not implement themselves as policies. Coding, payment, coverage language, committee cycles, guideline incorporation, utilization management, and claims edits each have their own clock. discoveriesinhealthpolicy.com
The FDA reviewer in Pahnke’s study has a submitted 510(k) to review within an established pathway. A payer’s meeting with a lab does not create a comparable duty to issue a new coverage policy, much less on the lab’s schedule. Even at MolDX, where there is a defined assessment infrastructure, scientific clarity and formal policy change remain distinct events. That is why “make their job easier” is good advice inside an active review, but not an adequate theory of how to obtain coverage in the first place.
My practical translation is modest: present one well-bounded analogy at a time, so the evaluator can see the relevant clinical category, the evidence for that particular intended use, and the precise question requiring a decision. Then ask the harder institutional questions from the September essay: Who actually has to say yes? Through what policy and claims mechanism? On what timeline? Can the company survive until that mechanism moves? A clearer comparison can help a gatekeeper work. It cannot substitute for a gatekeeper’s mandate, a coverage pathway, or the evidence that pathway demands.
Sources: Emily Cox Pahnke, Tiona Zuzul, and Michael Howard, “Strategic positioning and accelerating regulatory clearance for new ventures,” Strategic Management Journal 47 (2026): 1730–1763; Michael Blanding, “The Product Positioning Tactic That Helps Regulators Say Yes,” HBS Working Knowledge, November 18, 2026; Bruce Quinn, “You Better Learn the Biggest Myth About Payor Coverage of Diagnostics,” Discoveries in Health Policy, September 9, 2026.
