For most quantitative IVD assays, the validation has to defend the assay's performance across an analytical range that spans several orders of magnitude. An estradiol assay's range stretches from pre-pubertal pediatric values below 10 pg/mL to gonadotropin-therapy values above 3,000 pg/mL — three orders of magnitude on the same instrument, with the IFU's intended-use language touching populations at every part of the curve. Similar shapes appear for testosterone, thyroid markers, vitamin D, hCG, the cardiac troponins, and most therapeutic drug monitoring panels.
The conventional approach to recruiting samples for these validations is to enroll "healthy adults" generically and accept the value distribution that results. The approach is operationally simple and it produces studies that finish on time. It also produces datasets clustered in the middle of the analytical range, with the high and low ends underpopulated and the validation unable to defend its performance at the extremes the IFU references. The agency notices. The submission gets pushed back on. The cohort that looked complete on enrollment day turns out to be incomplete on review.
The framework below describes the alternative we now use. It treats the analytical range as a design surface and the population segments that populate it as the recruitment plan. You map the values before you recruit. You recruit deliberately by population. You adjust in real time with the physicians who see the patients. The recruitment is harder. The validation it produces is meaningfully more defensible.
One scope note before the principles. This framework applies to analytes whose value distribution can be predicted from identifiable population segments — most hormones, therapeutic drug monitoring panels, vitamin and mineral assays, and any quantitative marker whose physiology tracks demographic or clinical strata. Some analytes do not work this way. Cardiac troponins after acute presentation, inflammatory markers during active disease, post-prandial glucose, hCG in suspected pregnancy — for these, the value an individual subject produces is driven by transient clinical state in ways no population map can predict in advance. When the population map will not hold, the framework below is the wrong tool, and you fall back on the power of numbers: enroll enough volume across the right clinical contexts that the value distribution emerges statistically rather than by design. Knowing which kind of analyte you have — mappable or not — is the first decision before any of the principles below apply.
Map values to populations before recruitment opens.
Each part of the analytical range corresponds to identifiable population segments. Build the map before you screen a single subject.
The first principle is the one most often skipped, because it requires work that has to happen before any subject is consented and that does not show up in the enrollment timeline as visible progress. The work is to draw, explicitly, the chart that maps the assay's analytical range to the populations whose physiology produces values in each part of it. The chart is the recruitment plan. Without it, recruitment is a guess about which populations will produce which values, and the guess is almost always wrong about the extremes.
For estradiol, the chart we built on a recent program looked like this. Pre-pubertal and pubertal children populate the low end of the range, roughly 1 to 10 pg/mL. Post-menopausal women and adult males populate the next band, roughly 10 to 50 pg/mL. Pre-menopausal women across the cycle populate the mid range, roughly 50 to 500 pg/mL. Women receiving gonadotropin therapy as part of assisted reproduction populate the high end, often 1,000 pg/mL and above. Each band is a different recruitment problem with different sites, different consent flows, different operational logistics. The chart names them all in advance, and the recruitment plan addresses each band as its own sub-cohort.
The mapping work is not the assay developer's work alone. It draws on published reference ranges, on the assay's own analytical validation data showing where precision falls off, on claims-data analysis identifying which clinical populations carry the relevant phenotypes, and on the IFU's intended-use language that names which populations the assay is going to claim performance for. Most of this material exists somewhere; the framework's contribution is to require that it be assembled into a single chart before recruitment opens, and that the chart be the document the recruitment plan is built against.
The estradiol study we ran built the population-to-value chart in the first two weeks of the program, before any site was activated. The chart identified five population bands across the analytical range, including a deliberate gap zone — bioactive concentrations within physiological reach but beyond the assay's minimum detectable dose — that the validation team agreed was not in scope for this submission but would need attention in a follow-on. Naming the gap was as important as naming the bands. Recruitment proceeded against the chart, and the resulting dataset covered the full claimed range with sub-cohort populations defensible against pre-submission feedback.
Before any site is qualified, build the chart. Analytical range on the X-axis, target sample count on the Y-axis, population segments labeled at the band each one populates, with explicit gaps where the science does not produce values. The chart becomes the recruitment plan. Sites get assigned to populations, not to volume targets.
Recruit deliberately, by population segment.
"Healthy adults" is not a recruitment target. The populations that fill the analytical range are specific, identifiable, and require their own operational designs.
The second principle is what the chart enables. Once each part of the analytical range is mapped to a population segment, the recruitment plan becomes a set of sub-recruitment plans — one per segment — each with its own qualified sites, its own consent flow, its own kit specification, and its own operational cadence. The pediatric band is recruited at sites that see pediatric patients. The gonadotropin-therapy band is recruited at fertility clinics. The post-menopausal band is recruited through OB-GYN or primary-care practices that already see those patients on cadence. None of these are interchangeable, and none of them are well-served by a generic "healthy volunteer" recruitment plan.
This is where the conventional approach most clearly breaks. A generic healthy-adult recruitment will, statistically, produce a value distribution dominated by the populations who consent most readily — typically pre-menopausal women in mid-cycle and post-menopausal women — and will under-populate the extremes that require dedicated recruitment infrastructure. The pediatric band rarely fills itself. The gonadotropin-therapy band requires fertility-clinic partnerships that the generic recruitment plan does not include. The validation that results passes the middle of its claimed range and fails the edges, and the failure is structural, not a matter of recruiting harder.
The deliberate alternative is to size each sub-cohort against the chart, qualify sites against the populations they actually see, and treat each sub-cohort as its own mini-study with its own SIV, monitoring cadence, and forecast. The cumulative recruitment is harder than the generic version. The cumulative dataset covers the analytical range the way the IFU claims it.
Build the recruitment plan as a set of sub-recruitment plans, one per population band on the chart. Each sub-plan names its target sites, its consent flow, its kit specification, and its enrollment forecast. Sub-plans get monitored independently — a pediatric site that is failing to enroll does not get reassigned to gonadotropin therapy.
Adjust in real time with the physicians.
The chart is the plan. The physicians who see the patients know what their practice can actually deliver. Recalibrate weekly with them, not against them.
The third principle is the one that distinguishes a framework that survives contact with the field from one that does not. The chart drawn in the planning phase is necessarily a forecast, and forecasts diverge from actuals within the first month of recruitment. The right response to the divergence is not to push harder against the original plan. It is to recalibrate the plan with the physicians at the active sites, who know more about what their practice can produce than the chart does.
What this looks like in practice is a weekly conversation with each site's PI and lead coordinator about the prior week's actuals against the population-level expectation. A pediatric site that was forecast for ten subjects per week and is producing four is signaling something — about its referral patterns, about the cohort overlap with its existing patient base, about the season, about the consent flow — and the signal is most efficiently extracted by asking the PI directly, not by escalating the deficit through a status meeting two weeks later. Physicians know what their practice can deliver. The chart does not. The recalibration is the framework's mechanism for closing the gap.
The deeper move this enables is the right kind of partnership with the site. Most CROs treat sites as recruitment vendors who deliver against a plan. The framework treats sites as operational partners whose physician judgment is part of the recruitment design, not just an input to it. Sites recalibrated weekly through real conversation produce more total subjects than sites managed against a fixed forecast that no one updates, because the conversation surfaces the operational knowledge the original chart could not access. The framework formalizes the conversation.
On the estradiol program, the original chart underestimated the gonadotropin-therapy band's recruitment rate at one site by roughly 50% — the partner fertility clinic produced subjects faster than the chart had assumed. The Monday call with the PI surfaced the over-performance in week three, and the cohort plan was revised to expand the band's target from that site and de-emphasize a parallel site that was under-performing. The total program timeline moved up, not back, because the recalibration captured the over-performance instead of letting it run unstructured.
A weekly call with each site's PI and lead coordinator, on a fixed cadence, against the prior week's actuals. The call is not a status meeting; it is a recalibration. Sites that need more capacity get it. Sites that are under-producing get diagnosed before the deficit accumulates.
What this framework rules out.
The three principles describe how value-bucket recruitment becomes a planned exercise rather than a probabilistic one for analytes whose distribution is mappable. They also rule out a few conventions worth naming.
They rule out applying this framework to analytes whose distribution is not population-predictable. For acute-phase markers, transient-state analytes, and any value driven by clinical events rather than demographic strata, the framework's central instrument — the population-to-value chart — cannot be drawn before recruitment opens. The right approach for those analytes is the inverse: enroll enough volume across the right clinical contexts that the distribution emerges from the data. Forcing a population map onto an unmappable analyte produces a recruitment plan that is wrong in ways the data will eventually expose.
They rule out "healthy adults" as a recruitment target for any quantitative IVD whose analytical range spans more than a single population's physiology. The label is operationally meaningless and produces datasets that fail at the extremes.
They rule out fixed forecasts that no one updates. Forecasts diverge from actuals within weeks. Frameworks that do not include a recalibration mechanism produce studies that miss their cohort distribution silently and discover it at submission.
They rule out the convention that physicians are recruitment vendors rather than operational partners. The physicians' judgment about what their practice can deliver is part of the recruitment design, and frameworks that do not include that judgment as a structural input produce studies that under-deliver against any plan more sophisticated than "enroll volume."
The framework is not closed. When the study outcome matters, you call RDI. Map the values, recruit by population, recalibrate with the physicians who see the patients. The validation that results is the one the IFU can claim.