eduKateSG · Why Science?
Spend monitoring attention where a trial can most seriously fail
Use FDA’s April 2023 final Q&A through critical data and processes, documented risk assessment, central and on-site monitoring, thresholds, escalation and learning.
Monitoring is a quality-control activity that checks whether a clinical investigation is being carried out as planned. A risk-based approach does not mean doing less work. It means concentrating methods and frequency on the errors that matter most for participant protection and reliable results.
FDA’s final April 2023 questions-and-answers guidance expands its August 2013 risk-based monitoring guidance. It applies across human drugs, biological products, medical devices and combination products and explains planning, monitoring plans, findings and communication.
Did you know? FDA says sponsors should monitor not only important risks predicted at the start, but also important new risks detected while the trial is running. The monitoring plan is therefore a controlled plan that can adapt to evidence.
For students, this is scientific attention management. You learn to identify critical variables, anticipate failure, choose sensitive checks and update the plan when observations challenge the original assumptions.
Chapter index
Chapters 1–9: critical data, processes, risk assessment and monitoring design
- 1. Monitoring protects participants and data
- 2. The April 2023 Q&A is final
- 3. Critical data and processes come first
- 4. Risk assessment should be documented
- 5. Some risks should be designed out
- 6. The plan should connect risks to methods
- 7. Centralised monitoring can see patterns across sites
- 8. On-site monitoring still has targeted value
- 9. Remote monitoring needs access and privacy controls
Chapters 10–17: central review, thresholds, escalation, documentation and adaptation
- 10. Source-data review should be proportionate
- 11. Thresholds turn observations into action
- 12. Findings need timely, traceable communication
- 13. Escalation paths should be explicit
- 14. Corrective action should address causes
- 15. The plan should adapt as the trial evolves
- 16. Cross-study learning strengthens future trials
- 17. A complete monitoring receipt shows why attention moved
FDA describes monitoring as a quality-control tool for determining whether study activities are being carried out as planned. Sponsor oversight should help protect participant rights, safety and welfare and preserve data integrity across the investigation.
For monitoring protects participants and data, risk is a map of where attention creates the most protection. FDA describes monitoring as a quality-control tool for determining whether study activities are being carried out as planned. Sponsor oversight should help protect participant rights, safety and welfare and preserve data integrity across the investigation. The map remains honest only when this limit stays visible: Monitoring supports oversight; it does not transfer the investigator’s clinical responsibilities to the sponsor.
Try this
Build a risk-to-control card for monitoring protects participants and data. List the participant-protection and data-integrity questions that monitoring must answer for one protocol. Add the evidence that would show the control is failing. Keep this limit: Monitoring supports oversight; it does not transfer the investigator’s clinical responsibilities to the sponsor.
A strong science note separates observation from interpretation. List the participant-protection and data-integrity questions that monitoring must answer for one protocol. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.
Evidence checkpoint
List the participant-protection and data-integrity questions that monitoring must answer for one protocol. Keep this boundary visible: Monitoring supports oversight; it does not transfer the investigator’s clinical responsibilities to the sponsor.
FDA identifies the guidance as final, dated April 2023, under docket FDA-2019-D-0362. It expands the 2013 final guidance and includes recommendations for monitoring human drug, biological, medical-device and combination-product investigations.
A monitoring decision about the april 2023 q&a is final should connect signal, consequence and action. FDA identifies the guidance as final, dated April 2023, under docket FDA-2019-D-0362. It expands the 2013 final guidance and includes recommendations for monitoring human drug, biological, medical-device and combination-product investigations. Keep this guardrail: The Q&A should be read with applicable regulations, the 2013 guidance and product-specific requirements.
Try this
Run a monitoring tabletop for the april 2023 q&a is final. Create a source receipt with status, date, docket, scope and the earlier guidance it expands. Assign a detection time, escalation owner and closure test. Preserve this boundary: The Q&A should be read with applicable regulations, the 2013 guidance and product-specific requirements.
A strong science note separates observation from interpretation. Create a source receipt with status, date, docket, scope and the earlier guidance it expands. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.
Evidence checkpoint
Create a source receipt with status, date, docket, scope and the earlier guidance it expands. Keep this boundary visible: The Q&A should be read with applicable regulations, the 2013 guidance and product-specific requirements.
At protocol design, sponsors should identify data and processes critical to participant protection and data integrity. Examples may include eligibility, informed consent, treatment assignment, safety reporting, primary endpoints and investigational-product accountability depending on the trial.
The science in critical data and processes come first appears when the team changes course for a documented reason. At protocol design, sponsors should identify data and processes critical to participant protection and data integrity. Examples may include eligibility, informed consent, treatment assignment, safety reporting, primary endpoints and investigational-product accountability depending on the trial. The conclusion stops here: A standard checklist cannot identify what is critical for every protocol.
Try this
Explain critical data and processes come first using critical factor, signal, threshold and response. Mark each protocol process as critical, important or routine and justify the consequence of failure. Let a classmate challenge whether the response is timely. Respect this caution: A standard checklist cannot identify what is critical for every protocol.
A strong science note separates observation from interpretation. Mark each protocol process as critical, important or routine and justify the consequence of failure. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.
Evidence checkpoint
Mark each protocol process as critical, important or routine and justify the consequence of failure. Keep this boundary visible: A standard checklist cannot identify what is critical for every protocol.
FDA recommends evaluating potential causes, detectability and severity of risks to critical data and participant protections. Sponsors should document methods, conclusions and how the assessment shaped the protocol, plans and monitoring. FDA may request initial and updated assessments during inspection.
For risk assessment should be documented, risk is a map of where attention creates the most protection. FDA recommends evaluating potential causes, detectability and severity of risks to critical data and participant protections. Sponsors should document methods, conclusions and how the assessment shaped the protocol, plans and monitoring. FDA may request initial and updated assessments during inspection. The map remains honest only when this limit stays visible: A risk score without its reasoning is difficult to challenge, update or inspect.
Try this
Build a risk-to-control card for risk assessment should be documented. Create a risk register with cause, consequence, detectability, existing control, residual risk and owner. Add the evidence that would show the control is failing. Keep this limit: A risk score without its reasoning is difficult to challenge, update or inspect.
A strong science note separates observation from interpretation. Create a risk register with cause, consequence, detectability, existing control, residual risk and owner. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.
Evidence checkpoint
Create a risk register with cause, consequence, detectability, existing control, residual risk and owner. Keep this boundary visible: A risk score without its reasoning is difficult to challenge, update or inspect.
Before relying on monitoring to catch errors, the study team can simplify the protocol, remove unnecessary procedures, improve definitions or redesign workflows. Building quality into the design prevents predictable problems and reduces dependence on downstream correction.
A monitoring decision about some risks should be designed out should connect signal, consequence and action. Before relying on monitoring to catch errors, the study team can simplify the protocol, remove unnecessary procedures, improve definitions or redesign workflows. Building quality into the design prevents predictable problems and reduces dependence on downstream correction. Keep this guardrail: Monitoring cannot reliably compensate for a protocol that is too complex or ambiguous to follow.
Try this
Run a monitoring tabletop for some risks should be designed out. Choose one high-risk protocol step and redesign it to remove a handoff, duplicate entry or unclear decision. Assign a detection time, escalation owner and closure test. Preserve this boundary: Monitoring cannot reliably compensate for a protocol that is too complex or ambiguous to follow.
A strong science note separates observation from interpretation. Choose one high-risk protocol step and redesign it to remove a handoff, duplicate entry or unclear decision. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.
Evidence checkpoint
Choose one high-risk protocol step and redesign it to remove a handoff, duplicate entry or unclear decision. Keep this boundary visible: Monitoring cannot reliably compensate for a protocol that is too complex or ambiguous to follow.
A monitoring plan should explain identified risks and how monitoring methods will detect, manage, mitigate or eliminate them. It should define responsibilities, timing, scope, communication and escalation. The connection between risk and method makes monitoring scientifically defensible.
The science in the plan should connect risks to methods appears when the team changes course for a documented reason. A monitoring plan should explain identified risks and how monitoring methods will detect, manage, mitigate or eliminate them. It should define responsibilities, timing, scope, communication and escalation. The connection between risk and method makes monitoring scientifically defensible. The conclusion stops here: A schedule of visits is not a risk-based plan unless it explains what each activity is meant to control.
Try this
Explain the plan should connect risks to methods using critical factor, signal, threshold and response. Pair every critical risk with a signal, monitoring method, frequency, threshold and response. Let a classmate challenge whether the response is timely. Respect this caution: A schedule of visits is not a risk-based plan unless it explains what each activity is meant to control.
A strong science note separates observation from interpretation. Pair every critical risk with a signal, monitoring method, frequency, threshold and response. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.
Evidence checkpoint
Pair every critical risk with a signal, monitoring method, frequency, threshold and response. Keep this boundary visible: A schedule of visits is not a risk-based plan unless it explains what each activity is meant to control.
CHAPTER 7 OF 17 · Plan the controls
7. Centralised monitoring can see patterns across sites
Central review can use accumulated data to identify missingness, unusual distributions, outliers, protocol deviations, inconsistent reporting or site-level patterns. Cross-site comparison may reveal signals that a monitor reviewing one participant record at a time cannot see.
For centralised monitoring can see patterns across sites, risk is a map of where attention creates the most protection. Central review can use accumulated data to identify missingness, unusual distributions, outliers, protocol deviations, inconsistent reporting or site-level patterns. Cross-site comparison may reveal signals that a monitor reviewing one participant record at a time cannot see. The map remains honest only when this limit stays visible: A statistical signal is a prompt for investigation, not proof of misconduct or error.
Try this
Build a risk-to-control card for centralised monitoring can see patterns across sites. Design a central dashboard with three signals and a rule for verifying each before escalation. Add the evidence that would show the control is failing. Keep this limit: A statistical signal is a prompt for investigation, not proof of misconduct or error.
A strong science note separates observation from interpretation. Design a central dashboard with three signals and a rule for verifying each before escalation. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.
Evidence checkpoint
Design a central dashboard with three signals and a rule for verifying each before escalation. Keep this boundary visible: A statistical signal is a prompt for investigation, not proof of misconduct or error.
Visits can assess source records, facilities, investigational-product handling, delegation, consent practice and local workflows when remote information is insufficient. A risk-based plan may vary visit frequency and depth by study, site and emerging signal.
A monitoring decision about on-site monitoring still has targeted value should connect signal, consequence and action. Visits can assess source records, facilities, investigational-product handling, delegation, consent practice and local workflows when remote information is insufficient. A risk-based plan may vary visit frequency and depth by study, site and emerging signal. Keep this guardrail: Risk-based monitoring does not require eliminating on-site visits or source-data review.
Try this
Run a monitoring tabletop for on-site monitoring still has targeted value. Choose which questions need physical observation, which can be reviewed remotely and why. Assign a detection time, escalation owner and closure test. Preserve this boundary: Risk-based monitoring does not require eliminating on-site visits or source-data review.
A strong science note separates observation from interpretation. Choose which questions need physical observation, which can be reviewed remotely and why. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.
Evidence checkpoint
Choose which questions need physical observation, which can be reviewed remotely and why. Keep this boundary visible: Risk-based monitoring does not require eliminating on-site visits or source-data review.
CHAPTER 9 OF 17 · Watch the signals
9. Remote monitoring needs access and privacy controls
Remote review can improve timeliness but requires authorised access, secure systems, privacy protection and clarity about what records monitors may see or copy. Sites and sponsors should plan technical support and alternatives when systems or local policies limit access.
The science in remote monitoring needs access and privacy controls appears when the team changes course for a documented reason. Remote review can improve timeliness but requires authorised access, secure systems, privacy protection and clarity about what records monitors may see or copy. Sites and sponsors should plan technical support and alternatives when systems or local policies limit access. The conclusion stops here: Convenient access should not expose unrelated clinical information or weaken site control of source records.
Try this
Explain remote monitoring needs access and privacy controls using critical factor, signal, threshold and response. Create a remote-access matrix for role, permitted records, session controls, export limits and incident response. Let a classmate challenge whether the response is timely. Respect this caution: Convenient access should not expose unrelated clinical information or weaken site control of source records.
A strong science note separates observation from interpretation. Create a remote-access matrix for role, permitted records, session controls, export limits and incident response. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.
Evidence checkpoint
Create a remote-access matrix for role, permitted records, session controls, export limits and incident response. Keep this boundary visible: Convenient access should not expose unrelated clinical information or weaken site control of source records.
FDA regulations do not prescribe one universal amount of source-data verification. Sponsors can target review according to criticality, risk, system controls and monitoring findings while ensuring reliable trial conduct and reporting.
For source-data review should be proportionate, risk is a map of where attention creates the most protection. FDA regulations do not prescribe one universal amount of source-data verification. Sponsors can target review according to criticality, risk, system controls and monitoring findings while ensuring reliable trial conduct and reporting. The map remains honest only when this limit stays visible: Reducing routine verification requires evidence that other controls can detect important errors.
Try this
Build a risk-to-control card for source-data review should be proportionate. Compare full verification, targeted verification and central checks for one critical and one noncritical field. Add the evidence that would show the control is failing. Keep this limit: Reducing routine verification requires evidence that other controls can detect important errors.
A strong science note separates observation from interpretation. Compare full verification, targeted verification and central checks for one critical and one noncritical field. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.
Evidence checkpoint
Compare full verification, targeted verification and central checks for one critical and one noncritical field. Keep this boundary visible: Reducing routine verification requires evidence that other controls can detect important errors.
Plans should define when an error rate, missingness pattern, delay or other metric triggers investigation, added monitoring, retraining or escalation. Thresholds may be absolute, relative or trend-based and should reflect clinical and statistical importance.
A monitoring decision about thresholds turn observations into action should connect signal, consequence and action. Plans should define when an error rate, missingness pattern, delay or other metric triggers investigation, added monitoring, retraining or escalation. Thresholds may be absolute, relative or trend-based and should reflect clinical and statistical importance. Keep this guardrail: A threshold can focus attention but should not excuse a serious event just below a numerical cutoff.
Try this
Run a monitoring tabletop for thresholds turn observations into action. Set a threshold for one signal and add an override for high-severity single events. Assign a detection time, escalation owner and closure test. Preserve this boundary: A threshold can focus attention but should not excuse a serious event just below a numerical cutoff.
A strong science note separates observation from interpretation. Set a threshold for one signal and add an override for high-severity single events. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.
Evidence checkpoint
Set a threshold for one signal and add an override for high-severity single events. Keep this boundary visible: A threshold can focus attention but should not excuse a serious event just below a numerical cutoff.
CHAPTER 12 OF 17 · Watch the signals
12. Findings need timely, traceable communication
Important monitoring results should reach the people able to act, such as investigators, sponsor leadership, medical monitors, data teams or quality personnel. Reports should distinguish observation, risk assessment, requested action, owner and due date.
The science in findings need timely, traceable communication appears when the team changes course for a documented reason. Important monitoring results should reach the people able to act, such as investigators, sponsor leadership, medical monitors, data teams or quality personnel. Reports should distinguish observation, risk assessment, requested action, owner and due date. The conclusion stops here: A report that arrives after the affected decision point cannot protect participants in real time.
Try this
Explain findings need timely, traceable communication using critical factor, signal, threshold and response. Map one finding from detection through notification, decision, action, verification and closure. Let a classmate challenge whether the response is timely. Respect this caution: A report that arrives after the affected decision point cannot protect participants in real time.
A strong science note separates observation from interpretation. Map one finding from detection through notification, decision, action, verification and closure. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.
Evidence checkpoint
Map one finding from detection through notification, decision, action, verification and closure. Keep this boundary visible: A report that arrives after the affected decision point cannot protect participants in real time.
The monitoring plan should say when and to whom important issues are escalated. Urgency may depend on participant danger, data impact, recurrence, breadth and whether local correction is effective. Defined paths prevent a serious issue from waiting inside routine correspondence.
For escalation paths should be explicit, risk is a map of where attention creates the most protection. The monitoring plan should say when and to whom important issues are escalated. Urgency may depend on participant danger, data impact, recurrence, breadth and whether local correction is effective. Defined paths prevent a serious issue from waiting inside routine correspondence. The map remains honest only when this limit stays visible: Escalation should be proportionate but must not be delayed to protect performance metrics.
Try this
Build a risk-to-control card for escalation paths should be explicit. Build a three-level escalation ladder with examples, time limits and decision owners. Add the evidence that would show the control is failing. Keep this limit: Escalation should be proportionate but must not be delayed to protect performance metrics.
A strong science note separates observation from interpretation. Build a three-level escalation ladder with examples, time limits and decision owners. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.
Evidence checkpoint
Build a three-level escalation ladder with examples, time limits and decision owners. Keep this boundary visible: Escalation should be proportionate but must not be delayed to protect performance metrics.
Repeated deviations or errors may require more than reminding staff. Root-cause analysis can examine protocol design, training, workload, interface problems, incentives and oversight. Corrective and preventive actions should be specific, owned, dated and checked for effectiveness.
A monitoring decision about corrective action should address causes should connect signal, consequence and action. Repeated deviations or errors may require more than reminding staff. Root-cause analysis can examine protocol design, training, workload, interface problems, incentives and oversight. Corrective and preventive actions should be specific, owned, dated and checked for effectiveness. Keep this guardrail: Closing an action item is not evidence that the underlying risk has fallen.
Try this
Run a monitoring tabletop for corrective action should address causes. Write a root-cause tree and an effectiveness check tied to the original monitoring signal. Assign a detection time, escalation owner and closure test. Preserve this boundary: Closing an action item is not evidence that the underlying risk has fallen.
A strong science note separates observation from interpretation. Write a root-cause tree and an effectiveness check tied to the original monitoring signal. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.
Evidence checkpoint
Write a root-cause tree and an effectiveness check tied to the original monitoring signal. Keep this boundary visible: Closing an action item is not evidence that the underlying risk has fallen.
FDA recommends reevaluating risk and revising monitoring plans when important issues appear. New sites, recruitment changes, protocol amendments, technology updates or safety information can shift priorities. Controlled revisions should record evidence, rationale, approvals and communication.
The science in the plan should adapt as the trial evolves appears when the team changes course for a documented reason. FDA recommends reevaluating risk and revising monitoring plans when important issues appear. New sites, recruitment changes, protocol amendments, technology updates or safety information can shift priorities. Controlled revisions should record evidence, rationale, approvals and communication. The conclusion stops here: Adaptation should respond to evidence rather than hide an unsuccessful original plan.
Try this
Explain the plan should adapt as the trial evolves using critical factor, signal, threshold and response. Create a change receipt showing trigger, updated risk, plan revision and affected sites or systems. Let a classmate challenge whether the response is timely. Respect this caution: Adaptation should respond to evidence rather than hide an unsuccessful original plan.
A strong science note separates observation from interpretation. Create a change receipt showing trigger, updated risk, plan revision and affected sites or systems. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.
Evidence checkpoint
Create a change receipt showing trigger, updated risk, plan revision and affected sites or systems. Keep this boundary visible: Adaptation should respond to evidence rather than hide an unsuccessful original plan.
CHAPTER 16 OF 17 · Respond and learn
16. Cross-study learning strengthens future trials
Information from one investigation can inform other ongoing or future studies and product-development programmes. Recurrent findings may reveal a common vendor weakness, confusing endpoint, unrealistic protocol or inadequate control. Learning should be documented and shared without losing study-specific context.
For cross-study learning strengthens future trials, risk is a map of where attention creates the most protection. Information from one investigation can inform other ongoing or future studies and product-development programmes. Recurrent findings may reveal a common vendor weakness, confusing endpoint, unrealistic protocol or inadequate control. Learning should be documented and shared without losing study-specific context. The map remains honest only when this limit stays visible: A lesson from one design or population may not transfer unchanged to another.
Try this
Build a risk-to-control card for cross-study learning strengthens future trials. Turn one finding into a reusable lesson with scope conditions and evidence required before reuse. Add the evidence that would show the control is failing. Keep this limit: A lesson from one design or population may not transfer unchanged to another.
A strong science note separates observation from interpretation. Turn one finding into a reusable lesson with scope conditions and evidence required before reuse. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.
Evidence checkpoint
Turn one finding into a reusable lesson with scope conditions and evidence required before reuse. Keep this boundary visible: A lesson from one design or population may not transfer unchanged to another.
CHAPTER 17 OF 17 · Respond and learn
17. A complete monitoring receipt shows why attention moved
The receipt records critical factors, risk method, assumptions, monitoring plan, central and on-site activities, thresholds, findings, escalation, corrective action, plan changes and lessons. It lets a reviewer see why resources were directed to certain risks and whether action was timely.
A monitoring decision about a complete monitoring receipt shows why attention moved should connect signal, consequence and action. The receipt records critical factors, risk method, assumptions, monitoring plan, central and on-site activities, thresholds, findings, escalation, corrective action, plan changes and lessons. It lets a reviewer see why resources were directed to certain risks and whether action was timely. Keep this guardrail: A complete monitoring file cannot repair participant harm or unreliable data that were allowed to continue.
Try this
Run a monitoring tabletop for a complete monitoring receipt shows why attention moved. Ask an independent reviewer to trace one critical risk from protocol design to final resolution. Assign a detection time, escalation owner and closure test. Preserve this boundary: A complete monitoring file cannot repair participant harm or unreliable data that were allowed to continue.
A strong science note separates observation from interpretation. Ask an independent reviewer to trace one critical risk from protocol design to final resolution. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.
Evidence checkpoint
Ask an independent reviewer to trace one critical risk from protocol design to final resolution. Keep this boundary visible: A complete monitoring file cannot repair participant harm or unreliable data that were allowed to continue.
Risk-based monitoring map
| Question | Evidence to capture | Boundary to retain |
|---|---|---|
| What is critical? | Participant protections, endpoints, safety and key processes | Criticality is protocol-specific |
| What could fail? | Causes, detectability, severity and residual risk | A score needs documented reasoning |
| How will failure be seen? | Central signals, site review, thresholds and trends | A signal is not proof |
| What happens next? | Communication, escalation, CAPA, plan revision and learning | Closure requires effectiveness evidence |
Science learning becomes useful when a student can show where a claim came from, why it applies and where it stops. Continue with the eduKate Science Learning Hub, compare choices in the Education Hub, and connect these habits to career pathways built from subject capabilities.
Risk-based monitoring is optimistic because it treats quality as something a study can learn and improve while it runs. The cheerful discipline is to watch the right signals early, act openly and keep participant protection at the centre of every monitoring choice.
Primary sources and verification date
Source status checked 26 October 2026. FDA guidance describes the agency’s current thinking and does not replace applicable statutes, regulations or product-specific communication with FDA.
- FDA: A Risk-Based Approach to Monitoring of Clinical Investigations — Questions and Answers — Final guidance, April 2023; docket FDA-2019-D-0362.
- FDA: Full risk-based monitoring Q&A PDF — Official recommendations on risk assessment, plans, findings and communication.
- FDA: Oversight of Clinical Investigations — A Risk-Based Approach to Monitoring — Final foundational guidance, August 2013.