The Recognition Gap: Why U.S. Clinical Trial Sites Get Overlooked

Between 2017 and 2022, the number of active U.S. cancer treatment trials grew by roughly 4% per capita annually. The number of trial sites, over that same period, fell by 3% per capita each year ^[1]. By 2022, 6,710 cancer studies were open for enrollment across just 1,836 sites, and 70% of U.S. counties reported no active cancer treatment trial at all ^[1]. Those counties represent 74% of the country's land area and contain 19% of Americans aged 55 and older ^[1], the demographic group most likely to receive a cancer diagnosis.

That gap, between where trials exist and where patients actually live and receive care, is not reducible to a single cause. Geographic concentration reflects several overlapping forces: insufficient research infrastructure in community settings, transportation and awareness barriers for patients, and a site selection process that may structurally favor familiar institutions over qualified but less-visible ones. This article focuses on that last factor. The result is a pattern of overlooked clinical trial sites: practices and hospitals that could enroll patients, have qualified investigators, and serve populations with genuine clinical need, but are not making it onto sponsor shortlists. Many never receive a feasibility questionnaire. Some have never been contacted at all.

This article examines why that pattern persists, what it costs the clinical research enterprise, and how the field is beginning to respond.

The Mismatch Between Where Patients Are and Where Trials Run

Approximately 85% of U.S. cancer patients receive care in community settings, not at the academic medical centers where the large majority of clinical trials are conducted ^[2]. That figure has been cited by the National Cancer Institute for years ^[2], and the American Association for Cancer Research reiterated it in its 2024 Cancer Progress Report ^[3]. Despite that, the trial site portfolio has continued to consolidate around a shrinking group of high-volume institutions.

The consequences are measurable. According to data published by the American Society of Clinical Oncology in 2024, over half of U.S. cancer patients do not have access to a trial at their primary care site ^[16]. Ninety percent of adult cancer patients do not participate in clinical trials at all ^[16]. That figure is frequently attributed to patient hesitancy or strict eligibility criteria, but site access is a substantial upstream cause. When a community oncology practice runs no trials, participation is not a choice the patient can make.

The geography compounds the access problem in ways that correlate directly with disease burden. A 2024 JCO Oncology Practice study by Kirkwood and colleagues found that 98% of U.S. counties with the highest cancer mortality rates had no active clinical trial in 2022 ^[16]. Ninety-three percent of the most socially vulnerable counties, as identified by the CDC's Social Vulnerability Index, were in the same position ^[16]. The communities that have the greatest need for experimental therapies are, in aggregate, the ones least likely to have them available.

The oncology data is the most thoroughly documented example of this pattern, and it is in oncology where the geographic and institutional concentration of trial sites has been most precisely quantified.

How Sites Get onto the Short List

When a sponsor or CRO begins building a candidate site list for a new study, the primary discovery channels are internal sponsor databases of previously used sites, CRO proprietary site networks, and commercially available investigator directories ^{[6, 19]}. Each of those channels carries a structural bias toward sites that have already run industry-sponsored trials.

A site that has never participated in a sponsor-activated trial is unlikely to appear in that sponsor's internal database. Internal databases are primarily built from operational history: sites that have been contracted, monitored, and closed in prior studies. Sites without that activation record must enter through other channels, such as CRO site networks or commercially available investigator directories. Those directories are often not kept current, which further narrows attention toward sites with recent, visible activity ^[19].

An industry analysis of site selection practices observed that well-known investigators and sites with long-term sponsor partnerships may receive preferential consideration in ways that are not grounded in protocol fit ^[11]. That favoritism "might lead to the selection of a site that is not well suited" for a specific study while simultaneously causing qualified but unfamiliar sites to be overlooked ^[11]. Some physician-led groups treat patients across multiple therapeutic areas but are excluded from trials outside of their best-known indication, not because they lack the capacity, but because their broader profile does not surface through the channels sponsors routinely use ^[6].

The academic medical center occupies a privileged position in this process for reasons that are partly legitimate. AMCs typically have established IRB processes built up over years of multi-sponsor research, experience accepting standardized master service agreements, and established compliance systems built through experience managing regulatory inspections ^[11]. Those attributes reduce administrative friction for operations teams managing multi-site studies. But administrative convenience and clinical performance are not the same thing. Selecting for convenience carries real population-level consequences when it means entire categories of sites, and the patient populations they serve, never enter the consideration set.

The Feasibility Questionnaire Problem

When a site does appear on an initial candidate list, the next filter is the Site Feasibility Questionnaire (SFQ). The SFQ is supposed to give sites an opportunity to demonstrate patient population fit, investigator qualifications, and infrastructure readiness. In practice, it has become another mechanism that advantages sites with dedicated research support operations.

Only about 65% of SFQs distributed by sponsors and CROs are completed and returned ^[5]. The reasons vary by site type: high administrative burden, redundant questions that duplicate information provided in prior studies with the same sponsor, and, for smaller practices, insufficient staff time during active clinical care hours. A 2024 task force convened by the Site Enablement League, representing sponsors, CROs, and research sites, found that the feasibility process lacks clear differentiation between its three main stages and that CROs frequently engage sites prematurely, before protocol documents are finalized, generating unnecessary burden on both sides ^[7].

The financial scale of this system is underappreciated. Based on modeled estimates from a 290-response industry survey, an analysis published in Applied Clinical Trials put the global investigative site community's spend at approximately $170 million in 2024 for completing feasibility assessments and site qualification visits for FDA-regulated industry-funded trials ^[8]. That investment translates to approximately 2,500 staff hours per investigative site per year, on average ^[8]. Sites with dedicated feasibility assessment teams complete the process in roughly two-thirds the average time ^[8]. That speed advantage does not translate directly into selection: the same analysis found that PI and coordinator participation predicted better win rates, not team size alone ^[8]. Sites with dedicated feasibility teams complete assessments faster, but the evidence associates PI and coordinator participation with higher reported win rates ^[8].

Academic medical centers spend proportionally less per qualification process than site networks do, yet achieve substantially higher selection win rates ^[8]. That outcome likely reflects several compounding factors, among them faster IRB turnaround, established contracting infrastructure, and investigator reputation, though the contribution of each is not separately quantified in the available evidence ^[8]. These proposed explanations should be understood as operational hypotheses rather than findings.

Eighty-three percent of research sites report wanting more study opportunities, yet the same survey found that sites rarely receive feedback from sponsors and CROs on why their site was not chosen ^[5]. Without that feedback, a site has limited information with which to identify and address whatever shortcomings caused it to be passed over. A process that could function as an onramp for new sites instead functions as a gate that few unfamiliar sites can locate, let alone open.

Geographic Concentration and the Counties Left Behind

The geographic distribution of U.S. clinical trials follows a recognizable pattern: concentration in major metropolitan centers, predominantly in the Northeast, mid-Atlantic, and coastal California, with large portions of the country receiving few or no active sites.

Geographic inequity extends well beyond individual counties. Industry analyses have noted that the Southern region of the United States holds a disproportionately small share of active clinical trial sites relative to its population ^[9]. The contrast is sharpest where disease burden is highest. In 2022, 98% of U.S. counties with the highest cancer mortality rates had no active cancer treatment trial, and 93% of the most socially vulnerable counties were in the same position ^[16].

A 2024 cohort study by Sekar and colleagues, published in JAMA Network Open, used ClinicalTrials.gov data linked to the CDC Social Vulnerability Index (SVI) to evaluate the association between county-level social determinants of health and cancer trial availability ^[21]. The most socially vulnerable counties were significantly less likely to have any trial (49.6% vs 70.0% of counties with at least one trial in the least vulnerable group; odds ratio 0.33) and had approximately 39% as many population-adjusted trials compared with the least vulnerable counties ^[21]. The gap in whether counties had any trial remained stable across the study's 15-year observation period, while the disparity in population-adjusted trial counts widened over time ^[21].

The consolidation trend documented by Kirkwood and colleagues reinforces the concern. Between 2017 and 2022, the per capita count of active cancer trials increased while the per capita count of trial sites fell ^[1]. More studies are running through fewer physical locations. That means the geographic gaps are not narrowing over time. They are concentrating. A smaller number of high-volume sites manage a growing share of the study portfolio, while sites outside those networks fall further behind in both sponsor relationship history and operational familiarity with large-scale trial conduct.

The point is not that high-volume academic centers should be excluded from trials. It is that when 9% of U.S. counties hold 100 or more active cancer trials and are home to 48% of the 55-and-older population ^[1], the patients in the remaining 91% of counties are receiving materially different access to the clinical research enterprise, for reasons that are structural rather than medical.

What the Enrollment Record Actually Shows

A reasonable defense of the current site selection process would be that sponsors and CROs select from experienced, high-performing sites because those sites deliver. The enrollment record complicates that argument.

Research from the Tufts Center for the Study of Drug Development, based on nearly 16,000 investigative sites across 151 Phase II and Phase III global studies, found that 37% of selected sites under-enrolled and 11% failed to enroll a single patient ^[4]. Among North American sites specifically, 13% enrolled zero patients ^[4]. This is not a small inefficiency at the margins. Nearly half of selected sites either miss their targets or contribute nothing to the study.

The cost profiles diverge sharply across site performance levels. A Phesi analysis published in Applied Clinical Trials estimated that a site enrolling one participant over 30 months accumulates approximately $130,000 in activation and management costs, against roughly $14,000 per enrolled participant at a high-performing site of similar duration ^[17]. That is more than a nine-fold difference in cost efficiency. A site that enrolls no participants at all generates the same overhead with no data contribution.

That pattern of enrollment shortfall is frequently attributed to patient hesitancy or eligibility criteria. But when nearly half of activated sites fail to meet their enrollment targets and 11% contribute no patients at all, site selection decisions may contribute. The 2024 PLOS ONE analysis showing that data-driven, real-world-informed site ranking outperformed historical baselines suggests that the pool of sites being considered may not be the best match for many protocols ^[14].

Regulatory Context

The regulatory framework has consistently pushed toward broader site engagement and greater participant diversity, with varying degrees of binding force.

Under the Food and Drug Omnibus Reform Act of 2022 (FDORA), Congress established a statutory requirement for sponsors of certain drug and device studies to submit Diversity Action Plans addressing the enrollment of participants from underrepresented populations ^[12]. FDA issued a draft guidance in June 2024 describing how those plans should be formatted, what they should contain, and when they must be submitted as part of IND applications or premarket submissions ^[12]. The draft guidance specified that compliance would take effect 180 days after the final guidance was published.

The draft guidance was initially removed from the FDA website in early 2025 following an executive order on diversity, equity, and inclusion ^[13]. It was subsequently restored under a federal court order, and as of June 2026, when this article was prepared, the draft remains accessible on FDA's website, accompanied by a court-order notice and FDA's standard nonbinding characterization ^[12]. The guidance has not yet been finalized; the 180-day compliance clock will not begin until a final version is published. The underlying statutory requirement under FDORA sections 3601 and 3602 remains in law regardless of the guidance's current state. Sponsors planning late-stage trial submissions should consult regulatory counsel on current obligations and monitor FDA's guidance calendar accordingly.

ICH E6(R3), which reached Step 4 completion at ICH in January 2025 and was formally adopted by FDA in September 2025 ^{[20, 22]}, updated Good Clinical Practice guidelines to emphasize risk-proportionate monitoring and quality management systems throughout the trial lifecycle. While its primary focus is on operational oversight rather than site discovery, the guideline's framework for evaluating site quality on a study-specific, risk-adjusted basis creates a foundation for assessing community and independent research sites on criteria other than institutional brand or prior trial volume.

A 2023 commentary in Clinical and Translational Science by Carter-Edwards and colleagues noted that diversity, equity, inclusion, and access must inform clinical trial design from the earliest stages through dissemination ^[10]. The authors specifically identified community health centers and rural health centers as settings where trial access needs to expand, and emphasized that doing so requires not just adding site types to shortlists, but actively building research infrastructure and relationships in those settings ^[10]. That distinction matters: regulatory intent toward diversity will not translate into outcome changes unless operational practices around site discovery change as well.

AI and the Site Discovery Problem

Traditional site identification was designed for a world where the candidate pool was small enough to evaluate through manual review. The current scope of clinical research has outgrown that approach. As of 2024, over 400,000 studies are registered on ClinicalTrials.gov ^[18], and the global investigative site community spans tens of thousands of locations across widely different institutional types, patient demographics, and research capabilities. Relationship-based lists and static investigator databases cannot adequately cover that full range of sites.

Machine learning approaches are beginning to address this. The 2024 PLOS ONE study by Hurtado-Chong and colleagues developed a model that integrated historical enrollment data with real-world patient claims data to rank candidate sites by predicted recruitment performance ^[14]. Evaluated in inflammatory bowel disease and multiple myeloma studies, the model outperformed both median historical enrollment baselines and site-level historical performance alone. The authors noted that new and diverse research sites need to be considered alongside historically strong performers to improve the geographic and demographic representativeness of trial populations ^[14].

The practical implication is that indication-level historical enrollment records combined with real-world patient claims data can provide a richer basis for ranking candidate sites than historical-enrollment baselines alone ^[14]. The PLOS ONE study evaluated ranking performance across sites already within the study population; it did not demonstrate identification of previously unknown sites absent from sponsor networks. Whether claims-based models can reliably surface sites with no prior trial history remains an open research question. The potential, however, is for site discovery to depend less on an institution's existing visibility in sponsor databases.

The limitations are real and should not be minimized. A 2024 scoping review in the Journal of the American Medical Informatics Association found that AI tools trained for specific indications often perform poorly when generalized to different populations or settings ^[15]. Training AI systems on data drawn from historically selected sites risks encoding the same geographic and institutional biases that characterize conventional selection. Incomplete training data can lead to problematic predictions, and over-reliance on algorithmic outputs without human review can propagate errors at scale ^[15].

The technology is not a substitute for institutional commitment to finding new sites. It is a tool that makes broader, more systematic discovery operationally feasible for teams that would otherwise be limited by the size of their existing networks.

How Kitsa Fits Into This Problem

For sponsors and CROs that want to extend site discovery beyond existing investigator databases, the core challenge is matching protocol requirements to patient population data across a wider range of sites than historical databases cover. KScout, Kitsa's site selection intelligence product, is designed for that protocol-to-population matching across academic, community, and independent research settings (vendor-stated design intent). According to Kitsa, the product is built to surface sites whose catchment areas and investigator profiles fit a study's target population, including sites with no prior sponsor relationship on record. As with any site intelligence tool, independent validation of performance across indications and site types is an appropriate consideration for sponsors evaluating the product.

Key Takeaways

• Seventy percent of U.S. counties had no active cancer treatment trial in 2022, yet approximately 85% of cancer patients receive care in community settings, not at the academic medical centers where most trials run ^{[1, 2]}.
• Site-selection processes often give substantial weight to prior activation and performance history. Sponsor databases and CRO networks tend to reflect previously activated sites, giving qualified sites with no prior sponsor relationship limited visibility in these channels ^{[6, 19]}.
• Eleven percent of sites that are selected fail to enroll a single patient, and 37% under-enroll ^[4]. Post-activation enrollment results show substantial inefficiency at the site level, though the causes are multi-factorial.
• Only 65% of Site Feasibility Questionnaires are returned, and sites rarely receive feedback from sponsors and CROs on why their site was not chosen ^[5]. This creates no path for new sites to understand or address the barriers to entry.
• The most socially vulnerable counties and those with the highest cancer mortality rates are consistently the least likely to host any active trial ^{[21, 16]}, compounding access deficits for the patient populations that need research options most.
• Machine learning models using real-world data have demonstrated better site ranking than historical-enrollment baselines alone in the specific indications studied ^[14].
• The statutory requirement for Diversity Action Plans under FDORA remains in law. FDA's June 2024 draft guidance was restored online under a court order and remains a draft, nonbinding document ^{[12, 13]}. Sponsors should monitor finalization and consult counsel on current obligations.

Frequently Asked Questions

Why are many clinical trials concentrated at academic medical centers?

+

Academic medical centers may offer established research infrastructure, contracting experience, and compliance systems that reduce certain forms of administrative uncertainty ^[11]. Those attributes can make AMCs easier to work with operationally, though they do not necessarily reflect superior clinical performance for a given protocol. Community practices, even when they serve a patient population that matches a study's eligibility criteria, often lack dedicated clinical research coordinators and the administrative infrastructure to complete startup requirements at the pace sponsors expect.

What is the recognition gap in clinical trial site selection?

+

The recognition gap is the disconnect between the universe of sites that could appropriately conduct a given trial and the much smaller set that sponsors and CROs actually evaluate. Sites without prior industry trial history may be less likely to appear in sponsor databases. Sites serving rural or underrepresented populations may not have the CRO relationships needed to receive a feasibility questionnaire. The gap is structural: it reflects how discovery processes are built, not the actual distribution of capable investigative sites in the U.S.

How do enrollment failures relate to site selection decisions?

+

Research from the Tufts Center for the Study of Drug Development found that 11% of activated Phase II and Phase III trial sites fail to enroll a single patient and 37% under-enroll ^[4]. Among North American sites specifically, 13% enrolled no patients ^[4]. The Tufts data documents the scale of the problem but does not identify the selection decisions that produced those outcomes. A separate 2024 PLOS ONE study showed that models incorporating real-world patient claims data outperformed historical-enrollment-only baselines in predicting site recruitment in the specific indications studied ^[14], suggesting that historical familiarity alone may be an insufficient basis for site selection.

What role does geography play in which sites receive trial opportunities?

+

A 2024 JCO Oncology Practice study found that 70% of U.S. counties had no active cancer treatment trial in 2022 ^[1]. The Southern region holds the largest population share of any U.S. Census region but has among the lowest rates of ongoing trials relative to population ^[9]. Ninety-eight percent of counties with the highest cancer mortality rates and 93% of the most socially vulnerable counties had no active trial ^[16]. The geographic distribution of trial sites tracks closely with existing social and economic disparities rather than with the distribution of disease burden.

What has FDA required about trial site diversity?

+

Under the Food and Drug Omnibus Reform Act of 2022 (FDORA), Congress established a requirement for sponsors of certain drug and device studies to submit Diversity Action Plans addressing enrollment from underrepresented populations ^[12]. FDA issued a draft guidance in June 2024 describing the required format and content of those plans. The guidance was initially removed from the FDA website in early 2025 following an executive order, but was subsequently restored under a federal court order ^[13]. As of June 2026, when this article was prepared, the draft remains accessible on FDA's website as a nonbinding, non-final document ^[12]. The underlying statutory obligation under FDORA remains in law. Sponsors with late-stage submissions in planning should monitor regulatory developments and consult legal counsel on current obligations.

Can AI meaningfully change which sites get selected?

+

Yes, with important qualifications. ML models that incorporate real-world data, including claims-derived indication-specific patient-population features and historical enrollment records, have demonstrated superior site ranking compared to historical-enrollment-only baselines in the specific indications studied ^[14]. Such approaches may, in principle, surface sites with limited prior sponsor visibility, though the PLOS ONE study evaluated ranking among known candidate sites and did not demonstrate discovery of previously unknown sites ^[14]. The risk is that systems trained on historically selected sites will encode the same geographic and institutional biases present in conventional selection. Data quality, model generalizability, and human oversight are all necessary components of any responsible AI-assisted site identification process ^[15].

References

1. [1]Kirkwood MK, et al. "State of Geographic Access to Cancer Treatment Trials in the United States: Are Studies Located Where Patients Live?" JCO Oncology Practice. 2025;21(3):427-437. https://ascopubs.org/doi/10.1200/OP.24.00261
2. [2]National Cancer Institute. "Bringing Research to the Community to Reduce Cancer Disparities." cancer.gov. 2017. https://www.cancer.gov/research/areas/disparities/chanita-hughes-halbert-clinical-trials-community-access
3. [3]American Association for Cancer Research. "Disparities in Clinical Research and Cancer Treatment." AACR Cancer Progress Report 2024. https://cancerprogressreport.aacr.org/disparities/cdpr24-contents/cdpr24-disparities-in-clinical-research-and-cancer-treatment/
4. [4]Tufts Center for the Study of Drug Development. "New Research from Tufts CSDD Characterizes Effectiveness and Variability of Patient Recruitment and Retention Practices." Fierce Biotech. January 2013. https://www.fiercebiotech.com/biotech/new-research-from-tufts-center-for-study-of-drug-development-characterizes-effectiveness
5. [5]Advarra. "Strategies for Successful Site Selection in Clinical Trials." advarra.com. 2022. https://www.advarra.com/blog/strategies-for-successful-site-selection-in-clinical-trials/
6. [6]Clinical Leader. "Everyone Talks About Site Selection, But Few Actually Get It Right." clinicalleader.com. November 2025. https://www.clinicalleader.com/doc/everyone-talks-about-site-selection-but-few-actually-get-it-right-0001
7. [7]Bruneau B, et al. "Redefining Feasibility in Clinical Trials: Collaborative Approaches for Improved Site Selection." Contemporary Clinical Trials Communications. 2024;40:101291. https://pmc.ncbi.nlm.nih.gov/articles/PMC11298845/
8. [8]Applied Clinical Trials Online. "Benchmarking the Investigative Site Qualification Process." appliedclinicaltrialsonline.com. 2025. https://www.appliedclinicaltrialsonline.com/view/benchmarking-the-investigative-site-qualification-process
9. [9]Clinical Trials Arena. "Clinical Trial Access in the US: Where Are Today's Trials Missing?" clinicaltrialsarena.com. September 2025. https://www.clinicaltrialsarena.com/sponsored/clinical-trial-access-in-the-us-where-are-today-trials-missing/
10. [10]Carter-Edwards L, Hidalgo B, Lewis-Hall F, et al. "Diversity, Equity, Inclusion, and Access Are Necessary for Clinical Trial Site Readiness." Clinical and Translational Science. 2023 Nov 30. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10877510/
11. [11]IntuitionLabs. "Clinical Trial Site Selection: Key Factors and Best Practices." intuitionlabs.ai. March 2026. https://intuitionlabs.ai/articles/clinical-trial-site-selection
12. [12]U.S. Food and Drug Administration. "Diversity Action Plans to Improve Enrollment of Participants from Underrepresented Populations in Clinical Studies." Draft Guidance for Industry. Docket No. FDA-2021-D-0789. June 2024. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/diversity-action-plans-improve-enrollment-participants-underrepresented-populations-clinical-studies
13. [13]AJMC. "FDA Quietly Removes Draft Guidance on Diversity in Clinical Trials Following Executive Order on DEI." ajmc.com. 2025. https://www.ajmc.com/view/fda-quietly-removes-draft-guidance-on-diversity-in-clinical-trials-following-executive-order-on-dei
14. [14]Hurtado-Chong A, et al. "Enhancing Site Selection Strategies in Clinical Trial Recruitment Using Real-World Data Modeling." PLOS ONE. 2024 Mar 11;19(3):e0300109. https://pmc.ncbi.nlm.nih.gov/articles/PMC10927105/
15. [15]Li X, et al. "Artificial Intelligence for Optimizing Recruitment and Retention in Clinical Trials: A Scoping Review." Journal of the American Medical Informatics Association. 2024;31(11):2749-2759. https://pmc.ncbi.nlm.nih.gov/articles/PMC11491624/
16. [16]American Society of Clinical Oncology. "Cancer Clinical Trials Continue to be Out of Reach for Many Patients." asco.org. October 2024. https://www.asco.org/news-initiatives/policy-news-analysis/cancer-clinical-trials-continue-be-out-reach-many-patients
17. [17]Applied Clinical Trials Online (Phesi). "Assessing Single Patient Investigator Sites in Cancer Clinical Trials." appliedclinicaltrialsonline.com. https://www.appliedclinicaltrialsonline.com/view/phesi-report-assessing-single-patient-investigator-sites-in-cancer-clinical-trials
18. [18]ClinicalTrials.gov. "ClinicalTrials.gov Statistics." U.S. National Library of Medicine. Accessed June 2026. https://www.clinicaltrials.gov/about-site/about-ct-gov/ctgov-stats
19. [19]Premier Research. "Avoiding 3 Common Pitfalls to Efficient Site Selection." premier-research.com. October 2024. https://premier-research.com/perspectives/avoiding-3-common-pitfalls-to-efficient-site-selection/
20. [20]International Council for Harmonisation. "ICH Harmonised Guideline: Good Clinical Practice E6(R3)." Step 4, January 6, 2025. https://database.ich.org/sites/default/files/ICH_E6(R3)_Step4_FinalGuideline_2025_0106.pdf
21. [21]Sekar RR, Herrel LA, Stensland KD. "Social Determinants of Health and the Availability of Cancer Clinical Trials in the United States." JAMA Network Open. 2024;7(5):e2410162. doi:10.1001/jamanetworkopen.2024.10162. https://pmc.ncbi.nlm.nih.gov/articles/PMC11077395/
22. [22]U.S. Food and Drug Administration. "E6(R3) Good Clinical Practice." Guidance for Industry. September 2025. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/e6r3-good-clinical-practice-gcp

Suggested internal links: kitsa.ai/kscout | Kitsa blog: Manual Screening Failures in Oncology | Kitsa blog: Becoming a Principal Investigator | Kitsa blog: AI-Assisted Patient Recruitment