College of American Pathologists

Please see attached comments from the College of American Pathologists (CAP) in response to the FDA’s Discussion Paper titled "Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion Paper and Request for Feedback"

October 19, 2026

Kyle A. Diamantas, J.D.
Acting Commissioner of Food and Drugs
US Food and Drug Administration (FDA)
10903 New Hampshire Avenue
Silver Spring, MD 20993

Re: Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion
Paper and Request for Feedback [Docket No. FDA-2026-N-7874]

Submitted via Electronic Submission to www.regulations.gov

Dear Acting Commissioner Diamantas:

The College of American Pathologists (CAP) appreciates the opportunity to comment on the Food
and Drug Administration’s (FDA) Center for Devices and Radiological Health (CDRH) discussion
paper: Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion
Paper and Request for Feedback. The CAP serves patients, pathologists, and the public by
fostering and advocating excellence in the practice of pathology and laboratory medicine worldwide.
Our members have extensive expertise in providing and directing laboratory services under the
Clinical Laboratory Improvement Amendments of 1988 (CLIA) regulations, which require
compliance with requirements through a quality system approach for overall operations and
administration of the clinical laboratory.

As physicians specialized in the diagnosis of disease through laboratory methods, pathologists
deliver high-quality diagnostic services to patients and other physicians. Pathologists place patient
safety and test validity foremost as they determine which tests to implement in the clinical
laboratory. The CAP’s comments in response to the FDA's questions are informed by its
commitment to ensuring that patients receive the right test, at the right time, and with the right
result. As a CLIA-deemed accrediting organization and as a leading provider of laboratory
proficiency testing/external quality assessment (PT/EQA), the CAP accredits over 8,500
laboratories and regularly introduces new PT/EQA programs to elevate the quality of laboratory
medicine. The CAP Accreditation Checklists require the laboratory director to ensure that the
performance specifications for new tests, instruments, and methods introduced to the laboratory
have been properly verified and validated prior to being used for patient testing, including devices
that use artificial intelligence (AI).

EXECUTIVE SUMMARY
The CAP supports a risk-based, device-specific, lifecycle approach to the regulation of GenAIenabled medical devices. The FDA should evaluate these devices based on their intended use, role
within the clinical workflow, potential effect on patient care, and the ability of qualified users to
identify and correct erroneous outputs. Premarket evaluation should be proportionate to risk,

College of American Pathologists
1001 G Street, NW, Suite 425W
Washington, DC 20001
202-354-7100
tailored to the individual device, and supported by clinically relevant benchmarking and confirmation
that reflect real-world populations, workflows, and failure modes. Postmarket monitoring should
supplement, rather than substitute for, appropriate premarket evidence. With pathology-specific
devices, appropriate pathologist oversight and manufacturer accountability remain essential to
patient safety.

The use of generative AI-enabled (GenAI) tools should assist and augment the diagnostic
interpretation of pathologists, providing an additional source of information to integrate into the
overall diagnostic report. GenAI systems cannot provide a medical diagnosis or be responsible for
decision-making in patient care. Physician review of AI-generated information is necessary prior to
incorporating the information into a diagnostic report. Pathologists with responsibility for oversight of
clinical and anatomic pathology laboratories must have a leadership role in the selection,
configuration, deployment, application, and monitoring of AI systems that will be involved in the preanalytical, analytical and post-analytical phases of laboratory workflow.

CAP SPECIFIC COMMENTS TO THE FDA DISCUSSION PAPER
The CAP offers responses to selected questions below. Because AI models are at the early stages
of pathology-specific applications, we believe that more experience is required before we can
respond to some of the FDA’s questions. For those questions we have not answered, the FDA
should proceed in a stepwise manner, beginning with pilot testing of GenAI-enabled approaches
and learning from that experience before broadening their application. More experience will help
inform the approach. The CAP always stands willing to work with the FDA and other government
agencies, industry, pathologists, and other stakeholders to support high quality laboratory
operations and the safe and effective use of GenAI in medical care.

Section IV: Considerations for the Assessment of Risk for GenAI-Enabled Devices

1. Does the two-axis risk framework, organized around device activity and the
consequence of relying on an incorrect output, appropriately capture the dimensions
most relevant to the risk of a GenAI-enabled software function? If there are additional
dimensions—such as the reversibility of a resulting action, the availability of
downstream safeguards, the time pressure of the deployment setting, or the traceability
of the output (i.e., to primary source materials)—that should be represented in a risk
framework, please describe and provide examples of how they should be represented.

CAP Comment: The FDA's proposed two-axis risk framework is a good starting point;
additional factors, however, should be addressed to more fully appraise risk. This is
evident because Figure 1 included in the discussion paper displays exclusively a linear
risk relationship, but some types of risk may not be linear. The FDA should also consider
how the GenAI component functions or what task/action it accomplishes within the
overall device function. Different types of tasks/actions should have different levels of
risk. For example, a pathology-specific GenAI function with a low level of risk is one that
reviews the results of a preliminary test with a patient’s clinical factors and uses an

College of American Pathologists
1001 G Street, NW, Suite 425W
Washington, DC 20001
202-354-7100
 algorithm to suggest reflex or additional tests for the physician to order. For a
pathologist, this type of function would be relatively low risk because they would be
working within their specialty-specific area of expertise. Another low-risk example would
be to provide secondary review of microscopic slides that a pathologist identifies as
unusual. The GenAI, in this use case, would be acting as a quality tool to help ensure
consistency across a group of pathologists.

Conversely, a GenAI algorithm would be operating at increased risk if it reviews
information from a diverse array of pathology tests and renders conclusions about a
clinical state such as sepsis or acidosis. The complexity of information or size of the
dataset should also factor into the level of risk.

In addition, the CAP affirms the FDA’s suggestions that the reversibility of a resulting
action and availability of downstream safeguards should be factored into the risk level.
These factors would help understand the degree to which use of the GenAI will directly
affect patient care or treatment decisions. Using these additional risk modifiers could
help the FDA distinguish limited or low-consequence clinical tasks from those that have
greater magnitude of potential patient harm.

2. CDRH seeks input on how to account for the spectrum of GenAI-enabled informational
functions that vary in the degree to which they direct a user to a particular action (i.e.,
between “non-directive” and “action-directing”). What characteristics of an output—
such as its wording, specificity, personalization, or context—could be considered as
modifiers of the risk associated with an informational function, after accounting for the
device’s overall functionality and intended use? What additional information would
provide manufacturers with sufficient clarity and predictability around risk assessment
for informational functions while recognizing that directiveness may exist along a
continuum rather than as a binary distinction?

CAP Comment: The FDA should consider intended use and deployment context as
important modifiers of risk. The same GenAI-enabled model may present markedly
different risks depending on its intended use and role within the clinical workflow. A
function used to support an intermediate workflow step may warrant a lower level of risk
than the same function when its output is incorporated into a diagnostic interpretation,
pathology report, or treatment decision.

In assessing the risk of informational functions, the FDA should consider both the
output's content and its intended audience and integration into the clinical workflow.
Safeguards such as clear designation of the intended recipient and appropriate
documentation within the medical record may help reduce the risk of GenAI-enabled
informational functions.

3. CDRH seeks input on whether and when a GenAI-enabled function that results in

College of American Pathologists
1001 G Street, NW, Suite 425W
Washington, DC 20001
202-354-7100
 delivery of clinical information to patients, as opposed to HCPs, could present different
or higher risks, while also recognizing the potential benefits associated with improved
patient empowerment, engagement, and access to clinical information. What device
characteristics, output features, or safeguards might mitigate risks that could arise when
a user lacks the domain knowledge to independently evaluate an output, without
unnecessarily underestimating patient capability?

CAP Comment: The same AI output may be interpreted very differently depending on
the user’s training, ability to detect errors, and authority to act on the recommendation.
Informational output directed to patients likely presents different risks than those directed
to qualified healthcare professionals. Information directed at specialists may also have
an additional contextual significance. Compared to healthcare professionals, there is
greater variability among patients in their ability to appropriately interpret healthcare
information, so additional information may be necessary to account for individuals with
low-, medium-, and high-levels of information. Providing additional information for
patients would help ensure they have the context to help them know when to be
appropriately cautious and consult a medical professional. Moreover, patient-facing
GenAI devices should be specifically trained to not provide guidance responses to
patients if the question asked by patients specifically requests moral judgement calls.

4. CDRH seeks input on whether and how the distinction between generalist and specialist
physicians might inform the assessment of risk for HCP-facing GenAI-enabled functions.
Under what circumstances, if any, might risk be affected when an HCP who lacks the
relevant clinical specialist knowledge receives information that falls within a specialist
area of practice? What device characteristics or safeguards might mitigate such risks?

CAP Comment: The risk of adverse decision-making associated with pathology-related
GenAI-enabled functions would likely increase when used by a generalist who does not
have sufficient information to independently assess the underlying inputs, assumptions,
limitations, or validity of the output. Pathology and laboratory medicine functions involve
specialized knowledge in the areas of laboratory testing, diagnostic interpretation, and
other services that are specific to the clinical laboratory. Pathologists help recognize and
anticipate specialty-specific care challenges so that hospitals can employ the necessary
safeguards for these high-risk environments. Involving pathologists in the development
of relevant GenAI model development and testing will help the developers anticipate
these scenarios. The FDA should therefore consider both the expertise of the intended
user and the specific pathology-related function being performed when assessing risk.

5. For multi-turn conversational GenAI-enabled devices that may migrate from providing
“non-directive” information to “action-directing” information over the course of an
exchange, how should risk be assessed across realistic conversational trajectories?
How could the intended use of such a device be characterized when its behavior is
emergent across a conversation?

College of American Pathologists
1001 G Street, NW, Suite 425W
Washington, DC 20001
202-354-7100
 CAP Comment: For multi-turn conversational GenAI-enabled devices, risk should be
assessed across the entirety of a realistic interaction, rather than based on individual
outputs in isolation. A response that appears low risk when evaluated independently
may contribute to a higher-risk outcome when considered in the context of an ongoing
exchange. Risks may emerge through the accumulation of context over multiple
interactions, increasingly directive recommendations, misunderstanding of user-provided
information, or a failure to recognize circumstances that warrant referral or escalation to
a healthcare professional.

6. For GenAI-enabled care escalation functions, CDRH is considering how the evaluation
may account for both under-escalation and over-escalation. How could manufacturers
characterize and weigh these two directions of error, given that they may not be
commensurable and that acceptable trade-offs may vary by clinical context?

CAP Comment: We encourage the FDA and GenAI developers to collaborate with
practicing clinicians to help ensure that the broader range of clinical scenarios are
considered. Under CLIA, pathologists operating as laboratory directors have a legal
responsibility to ensure that any test used has validity – through initial validation and
ongoing calibration verification. We are concerned that some GenAI companies are
developing pathology-specific devices without consulting pathologists. Laboratory
directors will not use a device if they have no way to determine if the device is
functioning as intended. Familiarity with CLIA-related regulatory obligations also places
pathologists in a singular position to help characterize acceptable trade-offs in escalation
function development. By virtue of their medical education, clinical training, and specialty
certification, pathologists are best qualified to help manufacturers with this manner of
GenAI characterization.

Section V: A Competency-Based Approach for Premarket Evaluation of GenAI-Enabled
Devices

7. Is the competency-based approach described above, i.e., device benchmarking followed
by clinical confirmation, a useful and appropriate framework for evaluating GenAIenabled devices?

CAP Comment: The CAP generally agrees with the FDA’s suggestion of a competencybased framework consisting of device benchmarking followed by clinical confirmation as
an appropriate approach for evaluating GenAI-enabled devices. The framework is
conceptually similar to proficiency testing performed by pathologists and laboratory
professionals. Pathology and laboratory medicine encompass a broad range of
functions. As a result, benchmarking and clinical confirmation should be tailored to the
specific device, intended use, and risk profile of the GenAI-enabled function being
evaluated. The CAP believes that benchmarking should be based on clinically relevant

College of American Pathologists
1001 G Street, NW, Suite 425W
Washington, DC 20001
202-354-7100
 scenarios with known outcomes and should reflect the real-world conditions in which a
device is expected to operate. Evaluation approaches that rely heavily on contrived
cases or narrowly constructed test scenarios may fail to capture the breadth of variability
and irregularities encountered in clinical practice. In addition, benchmarking results
should be accompanied by appropriate measures of uncertainty, such as confidence
intervals, to facilitate meaningful interpretation of device performance.

In designing a competency-based approach, the FDA should recognize that
performance measures developed for human evaluation may not be directly
transferrable to AI systems. Tasks that are difficult for trained clinicians may be relatively
easy for an AI-enabled device; tasks that appear straightforward to clinicians may
present challenges for AI systems. Consequently, benchmarking methodologies should
be designed to evaluate the capabilities, limitations, and failure modes of AI-enabled
devices rather than relying solely on evaluation criteria developed for human
performance.

8. How might the two-axis risk framework described in Section IV be considered within the
competency-based approach described above to help determine the level of evidence
needed for a premarket submission?

CAP Comment: As noted above, we find the two-axis risk framework does not
completely encompass the full range of risks inherent in the pathology setting. We
recommend that the FDA expand the model to incorporate the task/action of the GenAI
within the model, complexity and variability of inputs, reversibility of resulting actions,
availability of downstream safeguards, intended use, and anticipated recipients of the
output in its clinical context.

9. Would a benchmarking structure such as the one described above be likely to provide
adequate evidence of clinical knowledge, analytic capabilities, safety behavior,
communication, and generalizability to support a reasonable assurance of safety and
effectiveness? Are there elements that are missing, redundant, or inappropriately
categorized? Are there externally developed standards that could be leveraged?

CAP Comment: Benchmarking should evaluate performance across progressively more
complex use cases rather than relying solely on a single endpoint assessment. Given
the lack of experience with GenAI in pathology, proceeding in a stepwise manner is
appropriate. Benchmark results should be interpreted cautiously and supplemented with
clinically grounded evidence that reflects intended populations, workflows, edge cases,
and failure modes. Because pathology use cases span nearly the entire array of
medicine, multiple benchmarks will be required. A benchmark employed for one specific
GenAI system is unlikely to be appropriate for a different pathology use case.
Developing a benchmark will likely require significant resources – including the
involvement of practicing clinicians. Different types of incentives may need to be

College of American Pathologists
1001 G Street, NW, Suite 425W
Washington, DC 20001
202-354-7100
 considered.

Once deployed, GenAI systems will require ongoing monitoring for drift, unexpected
failure patterns, bias, safety signals, and the impact of updates. Representative samples
should be evaluated repeatedly through multiple stages of analysis, with outputs
assessed and compared at each step of the process. Repeated evaluation of the same
sample under increasingly complex conditions may provide greater insight into the
consistency, reliability, and predictability of device performance.

10. CDRH recognizes that publicly available benchmarking assets may be subject to data
contamination, saturation, and limited real-world representativeness. How should a
sponsor establish that performance on a given benchmark predicts safe and effective
real-world behavior for the device’s intended use? What evidence should support the
construct validity of a benchmark used to gate device evaluation, and what role should
sponsor-developed benchmarks play given potential concerns around independence
and optimization to the test?

CAP Comment: Benchmarking assets should be frozen throughout a testing period to
ensure consistent and reproducible evaluation. Data should be added to or removed
from benchmarking assets only through a controlled, documented process.

Construct validity may be akin to the clinical confirmation process that is employed for
PT/EQA. Developers and clinicians may be able to collaborate on solutions that would
benefit both parties. Sponsors should be expected to work with clinical specialists who
have experience in the relevant field to demonstrate the clinical relevance of
benchmarks. As sensitivities change, benchmarking should also change. An appropriate
degree of skepticism and independent evaluation of benchmarks will help ensure that
clinically relevant contexts are considered.

Clinical knowledge and understanding are constantly evolving with scientific advances.
A good example of this is the updates that the FDA made in 2025 for antimicrobial
susceptibility testing. The FDA used standards published by the Clinical Laboratory
Standards Institute, which were based on real-world experience in clinical laboratories.1
Because antimicrobial resistance affects nearly 3 million Americans annually, this type of
collaboration yields meaningful results for the population. A process could also be used
to detect rare disease occurrences or outcomes.

11. CDRH is considering that clinical confirmation for a GenAI-enabled device might not
require a prospective clinical study in every case and has described above a range of

1 Humphries RM, Simner PJ. Major updates to FDA-recognized Clinical and Laboratory Standards

Institute breakpoints are a win for the fight against antimicrobial resistance. J Clin Microbiol. 2025
May 14;63(5):e0010625. doi: 10.1128/jcm.00106-25. Epub 2025 Apr 14. PMID: 40227045; PMCID:
PMC12077184.
College of American Pathologists
1001 G Street, NW, Suite 425W
Washington, DC 20001
202-354-7100
 approaches of increasing rigor and patient exposure. How might a sponsor select and
justify a confirmation approach tailored to a device’s intended use and proportionate to
the device’s risk profile? Are there device types or risk profiles for which one or more of
these approaches would be insufficient or inappropriate? How might the anticipated
distribution of real-world inputs be taken into consideration? Are there other methods of
clinical confirmation that might help inform the evaluation of GenAI-enabled devices?

CAP Comment: The appropriate level of clinical confirmation should be proportionate to
a GenAI-enabled device's intended use and risk profile. For example, lower-risk
functions may require evidence that the system performs reliably under controlled
conditions, while higher-risk applications may need stronger proof that the tool improves
or at least preserves safety, accuracy, workflow integrity, and patient outcomes in real
clinical environments.

14. For open-ended device outputs, how should performance comparators and acceptance
criteria be selected? When a panel of qualified clinicians serves as the comparator, how
should the applicable standard (for example, the standard of care versus the
performance of a median clinician in practice) be defined and justified? Should
generalist or specialist physicians be used as a performance standard? When should
human-AI team performance, rather than the device operating alone, serve as the basis
for evaluation?

CAP Comment: The evaluation of open-ended device outputs that have diagnostic,
prognostic, or therapeutic implications should be directly tied to factual accuracy relative
to authoritative sources, such as peer-reviewed literature, guidance documents, etc.
Acceptance criteria should include assessment and mitigation of hallucinations or
unsupported claims. The device should be trained not to provide guidance responses to
patients if the question specifically requests moral judgement calls.

For pathology-related functions, pathologist specialty expertise provides the most
appropriate benchmark for evaluating device performance. Performance comparators
and acceptance criteria should be selected based on the device's intended use and the
level of care the device is intended to support. The appropriate comparator would likely
differ depending on whether the device is intended for use by a generalist physician,
specialist physician, another healthcare professional, or a patient.

15. Are there ways in which performance might be assessed relative to the care, technology,
or course of action likely to occur in the absence of the device, rather than to the
comparators described in this section? What approaches might be used to identify and
justify a comparator such as unaided clinical judgment, delayed specialist review, or no
intervention?

CAP Comment: The CAP does not believe that unaided clinical judgment, delayed

College of American Pathologists
1001 G Street, NW, Suite 425W
Washington, DC 20001
202-354-7100
 specialist review, or no intervention should serve as comparators. Performance
assessment should be based on meaningful standards that reflect the intended clinical
use of the device and current best practices rather than on the absence of care or
delayed access to care. Comparators should be selected to support a rigorous
evaluation of safe and effective performance for its intended use.

16. Should independent third parties be involved in some or all aspects of a competencybased approach, including device benchmarking and clinical confirmation? If so, in what
ways might qualified, independent third-party participation contribute to this approach,
and what qualifications and independence criteria should apply? Are there aspects of
the assessment for which third-party involvement would be impractical or inadvisable?
What safeguards or program design features would be critical to prevent third-party
participation from limiting competition or preventing innovation?

CAP Comment: The CAP believes that qualified independent third parties should play a
vital role in competency-based evaluation. The CLIA laboratory accreditor program
offers experience that the FDA should consider. CLIA requires that organizations must
have nonprofit status to apply for designation by the Centers for Medicare and Medicaid
Services (CMS). In addition, designated organizations must provide information about
financial interests and ongoing, transparent information about their performance to
maintain their status. The third-party process should maintain FDA authority, while also
requiring that the FDA provide adequate information about decisions and the ability to
appeal. The FDA should also be fully appraised of prior third-party reviews and any
corrective actions suggested and whether they were completed.

The CAP agrees with the FDA’s statements that “it may be appropriate to expect
benchmark testing to assess clinical knowledge and analysis across other relevant
clinical domains” and “it may be appropriate to expect benchmark datasets to be
representative of the deployment contexts in which the device is expected to be used.”2
Requiring third-party review to be specific to the device under evaluation (rather than
broad categories of devices) will facilitate deployment-relevant expertise. In addition, the
role of the independent third party could vary depending on level of risk posed by the
device.

The CAP does not, however, support the FDA’s suggestion that expert adjudicators
could be large-language models. We need greater experience with these models before
drawing this conclusion.

Section VI: Postmarket Monitoring for GenAI-Enabled Devices

18. CDRH is considering whether it may be appropriate to accept greater premarket

2 US Food and Drug Administration. Considerations for the Regulation of Generative AI-Enabled

Medical Devices: Discussion Paper and Request for Feedback. Page 14.
College of American Pathologists
1001 G Street, NW, Suite 425W
Washington, DC 20001
202-354-7100
 uncertainty regarding a GenAI-enabled device’s benefit-risk profile through greater
reliance on postmarket monitoring. Under what conditions might such an approach be
appropriate, and what characteristics of a monitoring program would need to be in place
to justify reduced premarket evidence? Are there device types or risk profiles for which
this approach would not be appropriate?

CAP Comment: The CAP firmly believes that greater reliance on postmarket monitoring
is not an acceptable trade-off for greater premarket uncertainty. Allowing greater
premarket uncertainty would likely lead to clinician reluctance to deploy the devices, as
well as insurer reluctance to provide coverage. In addition, pathologists would still be
expected to meet the CLIA standard for validation prior to use in the clinical laboratory,
regardless of the FDA’s approach. It is also unclear whether the public would
understand the differences between GenAI devices with acceptance of greater
premarket uncertainty versus devices with sufficient premarket certainty.

GenAI devices should demonstrate safety and efficacy in a premarket setting – like all
other devices. The regulatory framework needs clear lines of responsibility for validation,
labeling, monitoring, updates, and risk mitigation. The risks of rapidly expanding to other
uses before adequately understanding the benefits and limitations are too great. Three
factors influence our skepticism: 1) the large number of postmarket studies that fail to be
completed, 2) the difficulty of removing a product on the market once it has been
approved, and 3) the experience of frontier AI models under training that exceeded their
safeguards without immediate detection. Accepting greater premarket uncertainty holds
the very real risk that people’s health and safety will be in jeopardy; exchanging potential
safety and efficacy is unacceptable with only a promise or anticipation of success.

19. Please comment on the potential approaches to postmarket performance evaluation,
including periodic re-benchmarking, sample-based clinician review, and performance
degradation monitoring. What additional approaches should CDRH consider, and how
should the cadence and triggering events for reassessment be determined?

CAP Comment: Postmarket performance evaluation will require ongoing surveillance to
track model performance, user behavior, unexpected failure patterns, emerging safety
signals, bias across patient groups, and the effects of updates or environmental changes
on clinical reliability. As highlighted in the CAP’s response to question #10, the CLIA
requirements for PT/EQA offer a helpful model for monitoring postmarket safety. The
CAP would be interested in exploring this further with the FDA and device developers.

21. What roles might clinicians, healthcare institutions, professional societies, standardssetting bodies, and other stakeholders appropriately play in postmarket monitoring, and
how can these roles be structured without diffusing manufacturer accountability?

CAP Comment: Postmarket monitoring requires clinician and patient feedback.

College of American Pathologists
1001 G Street, NW, Suite 425W
Washington, DC 20001
202-354-7100
 Involvement of clinicians who were not involved in developing the GenAI will help ensure
that a broader array of clinical contexts is considered. A model for soliciting this clinician
and patient feedback is needed.

Section VII: Other Topics

26. Are there additional considerations that inform the premarket and postmarket evaluation
of agentic GenAI-enabled devices, beyond those applicable to nonagentic GenAIenabled devices? How should the elevated risk associated with autonomous multi-step
action, tool use, and reduced opportunity for human review be reflected in acceptance
criteria and oversight?

CAP Comment: Agentic AI raises additional considerations because it may plan, call
tools, modify workflows, or initiate actions. Multi-step autonomous action, tool use, and
reduced human review warrants stricter acceptance criteria and oversight. When a
system can plan, call external tools, modify workflows, or execute clinical-adjacent
actions, the FDA should consider not only output accuracy but also task boundaries,
error recovery, auditability, permissions, escalation rules, the impact and risk
categorization of intermediate steps, and fail-safe mechanisms.

Thank you for the opportunity to submit these comments. The CAP looks forward to working with the
FDA. Please direct questions on these comments to Han Tran at htran@cap.org.

Sincerely,

The College of American Pathologists

College of American Pathologists
1001 G Street, NW, Suite 425W
Washington, DC 20001
202-354-7100