Why Remediation So Often Fails
When a Phase III trial falls behind enrollment targets, the standard response is predictable: add sites, increase site management contact frequency, bring in a patient recruitment vendor, consider protocol amendments to broaden eligibility. These are legitimate interventions, and sometimes they produce results. But they often don't — not because they're inherently ineffective, but because they're applied to symptoms rather than root causes.
A trial running 40% below enrollment projections at month 12 is not necessarily experiencing a "recruitment problem." It may be experiencing a site activation problem, a protocol complexity problem, a patient identification problem, or a competitive landscape problem — and the correct intervention differs substantially for each. Treating all enrollment delays as equivalent, to be resolved by the same add-site-and-add-vendor response, produces the same cycle of delay, marginal intervention, and continued shortfall that characterizes too many Phase III programs.
What follows is a more precise taxonomy of what actually drives Phase III enrollment delays, drawn from patterns observed across oncology, CNS, metabolic disease, and rare disease therapeutic areas.
Root Cause 1: Site Selection Based on Historical Performance, Not Current Patient Population Fit
Site selection carries the highest leverage on enrollment outcomes and the longest lag between decision and consequence. A site selected for its strong historical enrollment performance in a prior oncology trial may have a dramatically different patient population for a new indication — and that mismatch won't become visible until eight to ten months into the enrollment period, long after the opportunity cost has accumulated.
The conventional site selection process relies heavily on investigator CVs, prior trial history, and site audit findings. These are reasonable proxies for operational quality but they are poor predictors of enrollment velocity for a specific protocol in a specific patient population window. A site with excellent Phase II breast cancer enrollment history may have a thin population of the specific biomarker-selected patients required for a targeted therapy trial. That information exists in the EHR — it's rarely interrogated at the site selection stage.
The operational consequence is that sponsors activate sites with poor patient-population fit, spend 6-12 months discovering this through low enrollment velocity, and then add sites — a process requiring another 3-6 months for activation, including IRB submissions and feasibility assessments. The total timeline cost of poor site selection on a Phase III program can be 12-18 months, and the information needed to prevent it is largely available before activation decisions are made.
Root Cause 2: Protocol Complexity That Concentrates Burden on Site Coordinators
Protocol complexity has increased substantially over the past two decades. The number of combined inclusion and exclusion criteria in Phase III protocols has grown considerably — with some analyses of recent oncology trials identifying 30 or more eligibility criteria. Each additional criterion increases coordinator screening time, raises screen failure probability, and creates more opportunities for protocol deviation.
The enrollment impact of high protocol complexity operates through two channels. First, coordinators at high-volume sites with multiple concurrent trials deprioritize complex protocols — the opportunity cost calculation is straightforward when a simpler protocol offers comparable coordination burden and compensation. Second, complex inclusion/exclusion logic is more prone to coordinator interpretation errors, producing screen failures that could have been prevented with clearer protocol language or better pre-screening support.
We're not saying complex protocols should be avoided — the scientific and regulatory rationale for specific eligibility criteria is often well-founded. We're saying that enrollment timeline projections for complex protocols need to account for this cost, and sites managing complex trials need better pre-screening tools to maintain throughput without overburdening coordinators.
Root Cause 3: Patient Identification That Depends Entirely on Passive Referral
In therapeutic areas with high patient visibility — conditions presenting acutely or managed routinely in specialist clinics — passive referral can sustain enrollment. In oncology, neurology, and rare disease, where eligible patients are distributed across health systems and may not be actively seeking treatment, passive referral consistently underperforms projections.
The structural problem with passive identification is that it depends on a physician encounter happening at the right time at a site that is actively enrolling. A patient with the relevant biomarker profile seen quarterly at a community practice that is not a trial site will never surface in the enrollment pipeline. EHR-based proactive identification addresses this by finding candidates in the record rather than waiting for them to present — but it requires data access infrastructure, de-identification pipelines, and matching algorithms that most sites do not maintain in-house.
Programs that do not deploy any proactive identification mechanism are, structurally, leaving a large fraction of their eligible patient population invisible to the enrollment process. The fraction varies significantly by therapeutic area and geographic market, but in rare disease and biomarker-selected oncology, it routinely exceeds 40% of the eligible population at any given site.
Root Cause 4: Screen Failure Rates Underestimated at Protocol Design
Screen failure rates are the most consequential and underestimated cost driver in Phase III enrollment planning. The industry-wide average screen failure rate runs approximately 35-45% across therapeutic areas, but in biomarker-selected oncology and rare disease programs, rates of 55-70% are not unusual and are not outliers.
When enrollment projections are built on optimistic screen failure assumptions — as they commonly are during protocol design — the downstream timeline impact is severe. A trial projecting 2 screens per enrolled patient that actually requires 3.5 screens per enrolled patient will miss its enrollment target by that corresponding factor unless site patient volume is substantially increased or screening efficiency improves. Neither happens quickly.
Pre-screening tools that apply eligibility logic to structured patient data before a patient enters the formal consenting and screening process offer the most direct lever on this root cause. Pre-screening doesn't eliminate failures requiring in-clinic evaluation, but it systematically filters obvious failures before they consume coordinator time and per-screen protocol costs.
Root Cause 5: Competitive Enrollment Pressure in Dense Trial Markets
In concentrated clinical trial markets — the Greater Boston area, Research Triangle, the Bay Area, and Houston Medical Center — multiple Phase II and Phase III programs in the same indication routinely run simultaneously, drawing from overlapping patient populations and the same pool of experienced sites and coordinators. Standard enrollment projections treat available patient population at a site as a standalone resource, which it is not.
A Phase III sponsor activating five Boston-area sites without analyzing what competing trials those sites are currently running — and what patient volume is already allocated — will consistently experience enrollment rates below standalone projections. Patients are shared; site coordinator bandwidth is shared; investigator attention is shared. These are resource constraints that affect enrollment velocity in ways that adding a sixth site cannot resolve.
Geographic diversification — deliberately selecting a portion of the site network in markets with lower competitive trial density for the relevant indication — can improve enrollment velocity without sacrificing the patient population depth that high-density markets provide. This requires competitive landscape analysis at the site selection stage, which most sponsors do not systematically perform.
The Compound Effect and the Path Out
These five root causes rarely operate in isolation. A Phase III program with suboptimal site selection, a complex 30-criteria protocol, passive-only patient identification, an optimistic 30% screen failure assumption against an actual 58% rate, and five sites in a competitive oncology market is not experiencing one enrollment problem. It is experiencing five compounding problems — and adding vendors or sites addresses none of them at their source.
The sponsors who consistently meet enrollment projections treat site selection as a data problem, protocol design as an operations problem, and patient identification as an informatics problem. Each of these is tractable. The clinical development timeline is largely determined by how early in the program lifecycle sponsors choose to treat them that way.
