Protocol complexity is one of those variables that clinical operations teams discuss qualitatively but rarely measure. A protocol is "complex" when coordinators say it is hard to screen for. A protocol is "simple" when candidates are easy to identify and consent rates are high. These assessments are real and they track actual enrollment outcomes, but without systematic analysis of what complexity consists of and how it maps to enrollment pace, the language stays vague.
Over the past year, we have been building a dataset from the protocols that sites running on Enrollvue have processed. Across 60-plus protocols in oncology, metabolic disease, and CNS indications, we have been coding specific structural features of each protocol and comparing them to the actual enrollment rates reported by the sites. What follows is an honest account of what we found, including where the correlations are strong, where they are weak, and what the analysis does not tell us.
What we measured
We coded protocols along several dimensions. Exclusion criterion count: the raw number of distinct exclusion criteria in the eligibility section. Free-text criterion density: the fraction of criteria that cannot be evaluated from structured EHR fields alone and require note reading. Temporal qualifier prevalence: the fraction of criteria that include time windows, prior therapy intervals, or look-back periods. Compound criterion rate: criteria containing multiple conditions joined by AND or OR logic, including carve-outs and exceptions.
We used sites' actual pre-screen-to-screen ratios as our primary enrollment rate proxy. Pre-screen-to-screen ratio measures how many patients a site reviews in pre-screening for each patient who makes it to formal screening. A ratio of 5:1 means five pre-screened candidates per one screen. A ratio of 12:1 means twelve. Higher ratios mean more coordinator time per enrolled patient and generally correspond to slower enrollment pace when site capacity is the binding constraint.
Exclusion criterion count: the clearest signal
The strongest correlation we found was between the total number of exclusion criteria and the pre-screen-to-screen ratio. Protocols with 15 or fewer exclusion criteria had a median pre-screen-to-screen ratio of around 4.5:1 across the sites in our dataset. Protocols with 25 to 35 exclusion criteria had a median ratio of 8.5:1 to 11:1. Protocols with more than 35 exclusion criteria reached ratios of 14:1 and above in several cases.
This is not a surprising finding, but having the numbers makes it concrete. Each additional exclusion criterion has an additive effect on the proportion of the candidate pool that gets disqualified, and on the time required to assess each candidate. Both effects compound with criterion count.
The relationship is not linear. Moving from 15 to 20 criteria has a smaller proportional effect than moving from 25 to 30. This may reflect a threshold effect: once a protocol has a certain complexity floor, adding further criteria applies to candidates who were already likely to fail screening on other grounds.
Free-text dependency: the operational multiplier
Free-text criterion density does not predict eligibility rate by itself, but it multiplies the time cost of pre-screening in a way that criterion count alone does not capture. A protocol with 20 exclusion criteria, 8 of which require reading clinical notes, demands significantly more coordinator time per candidate than a protocol with 25 exclusion criteria that are all resolvable from structured fields.
We found that protocols where more than 40 percent of criteria require free-text evaluation had pre-screening time costs roughly 2.5x those of protocols where fewer than 20 percent of criteria require note reading. The eligibility outcome may be similar, but the path to reaching that outcome is much longer when notes are involved.
This is part of why we think about the matching problem as a reading problem rather than a query problem. Reducing time in free-text review has a larger impact on operational throughput than reducing the number of structured-field queries a coordinator runs.
Temporal qualifiers: the hidden complexity layer
Protocols with high temporal qualifier prevalence showed elevated rates of screening failures relative to pre-screen-to-screen ratio. In other words, candidates made it through pre-screening and into formal screening, but then failed at the eligibility review stage when the temporal window was assessed more carefully.
This suggests that pre-screening tools, including manual review, tend to underweight temporal qualifiers. A coordinator reviewing whether a patient has received a specific therapy may confirm the therapy in the chart but not adequately assess whether it falls within the relevant look-back window. The candidate advances to screening, and the protocol deviation is caught at the site initiation visit review or by the medical monitor, adding a formal screening failure to the count.
From a protocol design standpoint, temporal qualifiers are often there for genuine safety reasons and cannot simply be removed. But protocols that use overlapping time windows with different reference dates for different criteria create an unusually high burden. Simplifying the temporal structure, where clinically feasible, has a meaningful effect on screening failure rates and coordinator error frequency.
What the analysis does not tell us
We want to be clear about the limits of this data. The protocols in our dataset are not a random sample. They are protocols that sites chose to run on Enrollvue, which means they skew toward protocols where the site expected pre-screening to be operationally demanding. Simpler protocols are underrepresented. The correlation between complexity metrics and enrollment rates is directionally robust, but the absolute thresholds we report are specific to this population.
We also did not control for therapeutic area prevalence. An oncology protocol with 30 exclusion criteria is operating against a fundamentally different candidate population than a metabolic disease protocol with 30 exclusion criteria. The absolute pre-screen ratios are not comparable across indications. What is comparable is the relative effect of complexity within an indication.
Sponsor and protocol teams who want to use complexity metrics as a design input should treat this analysis as directional guidance. The finding that exclusion criterion count above 25 correlates with materially worse enrollment ratios is useful for protocol design discussions. It should not be used as a hard cutoff.
Implications for protocol design and operational planning
For teams in the early stages of protocol development, these findings support a specific design principle: audit exclusion criteria not just for scientific necessity but for operational impact. Each criterion that requires note reading, that carries a temporal qualifier, or that creates a compound condition adds to the coordinator burden that sites will experience for the duration of the trial.
For teams planning site activation and coordinator staffing, complexity metrics provide a basis for more accurate enrollment forecasting. Sites with high coordinator-to-protocol ratios may handle complex protocols well. Sites where coordinators are already stretched thin will hit the complexity penalty harder. Matching sites to protocols based on operational fit, not just patient volume, is one of the underused levers in site selection.