Clinical Trial Data Management: From Collection to Database Lock

Clinical trial data management is the systematic process of capturing, cleaning, validating, and locking clinical research data before statistical analysis and regulatory submission. This article explains the complete data management lifecycle, from case report form design through database lock, with practical guidance for students, researchers, and life-science professionals who need to understand or implement these processes.

The data management workflow determines whether a clinical trial produces reliable, auditable evidence. Poor data handling creates risks that range from delayed regulatory approval to invalid study conclusions. Understanding the full pathway from source data collection to the final locked database helps researchers plan resources, anticipate common problems, and maintain data integrity throughout the study lifecycle.

At a Glance: Clinical Data Management Workflow

The table above summarizes the four major phases of clinical data management. Each phase requires specific documentation and quality controls before progression to the next stage.

The Clinical Trial Lifecycle and Data Management Position

Clinical trials follow a structured lifecycle that begins with protocol development and ends with regulatory submission and publication. Data management occupies a central position in this lifecycle because every downstream activity depends on the quality and completeness of the collected data.

The trial lifecycle typically includes protocol design, ethics and regulatory approval, site initiation, patient enrollment, treatment administration, follow-up visits, data analysis, and results reporting. Data management activities run parallel to clinical operations instead of occurring as a single discrete phase. Database build and validation happen during study startup, data cleaning continues throughout the treatment and follow-up periods, and database lock occurs after the last patient completes their final visit.

Understanding where data management fits in the broader lifecycle helps research teams allocate appropriate resources and timelines. A study that completes enrollment quickly but has poor data quality will still face lengthy cleaning periods before analysis can begin. Conversely, a well-designed database with clear collection standards can compress the time between the last patient visit and database lock.

Core Principles of Clinical Data Management

Data Integrity and the ALCOA Framework

Data integrity rests on principles that regulatory agencies and research organizations apply to all clinical trial records. The ALCOA framework describes the qualities that make data trustworthy: attributable, legible, contemporaneous, original, and accurate. These principles apply to every data point collected during a clinical trial, regardless of whether the data is captured on paper or in an electronic system.

Attributable data can be traced to the individual who recorded it. Legible data can be read and understood by anyone who needs to review it. Contemporaneous data is recorded at the time the observation occurs instead of reconstructed later. Original data represents the first recording of an observation. Accurate data correctly reflects the actual observation or measurement.

The ALCOA framework has been extended in recent years to include additional qualities such as complete, consistent, enduring, and available. These extensions address the challenges of electronic data capture and long-term data retention. Research organizations should document their data integrity policies and train all staff who handle trial data.

Source Data and Source Document Verification

Source data is the original record of a clinical observation or measurement. Source documents include medical records, laboratory reports, patient diaries, and any other original documentation where data first appears. The case report form, whether paper or electronic, is not considered a source document unless it is the original location where the data is recorded.

Source document verification is the process of comparing data entered into the case report form against the source documents to confirm accuracy. This verification is typically performed by clinical monitors during site visits. The verification process identifies transcription errors, missing data, and discrepancies between the source and the case report form.

Research teams should define which documents serve as the source for each data element before enrollment begins. For example, blood pressure measurements might be sourced from the clinic vital signs record, while adverse events might be sourced from the medical chart. Clear source document definitions prevent confusion during monitoring and auditing.

The Role of Standard Operating Procedures

Standard operating procedures provide the institutional framework for consistent data management practices. These procedures document how the organization handles data collection, entry, cleaning, coding, and locking. Regulatory inspectors expect to see documented procedures that staff actually follow.

Organizations should maintain standard operating procedures for database build and validation, data entry conventions, query management, medical coding, and database lock. Each procedure should describe the responsible roles, required steps, and documentation expectations. Staff training records should demonstrate that all personnel understand and follow the applicable procedures.

Case Report Form Design and Database Build

Principles of Effective CRF Design

The case report form is the primary instrument for collecting clinical trial data. A well-designed CRF captures all required protocol data elements while minimizing ambiguity and entry errors. Poor CRF design creates downstream cleaning burdens and can compromise data quality.

Effective CRF design begins with a thorough review of the protocol to identify every data element required for the study objectives. Each data element should be mapped to a specific CRF field with clear instructions for completion. Fields should be grouped logically by visit and assessment type to match the clinical workflow.

Question wording should be unambiguous and use standard medical terminology. Response options should be exhaustive and mutually exclusive. Free-text fields should be limited to data that cannot be captured through structured responses, because free text is difficult to analyze and prone to inconsistency.

The CRF should be reviewed by multiple stakeholders, including the protocol author, statistician, data manager, and site coordinators. Pilot testing on a small number of mock cases can identify confusing instructions or missing fields before the database goes live.

Electronic Data Capture Systems and ePRO

Electronic data capture systems have largely replaced paper CRFs in modern clinical trials. These systems provide built-in validation checks, audit trails, and real-time data visibility that are difficult to achieve with paper. Electronic systems also support direct data entry at the point of care, reducing transcription errors.

Electronic patient-reported outcome instruments allow study participants to enter their own symptom and quality-of-life data directly into the trial system. These instruments can be completed on tablets, smartphones, or computers at the clinic or at home. Electronic patient-reported outcomes reduce the burden on site staff and capture data closer to the actual patient experience.

The selection of an electronic data capture system should consider validation status, audit trail capabilities, user interface, and the ability to support the specific study design. Systems must comply with applicable regulatory requirements for electronic records and electronic signatures. The organization should validate the system before use and document the validation activities.

Edit Check Programming and Validation

Edit checks are automated validation rules that identify potentially erroneous or missing data at the time of entry. These checks range from simple range checks, such as verifying that a temperature value falls within a plausible range, to complex cross-field validations, such as confirming that a pregnancy test was performed for female participants of childbearing potential.

Edit checks should be designed to flag genuine errors without creating excessive false positives that burden site staff. Each check should have a clear specification that describes the condition, the error message displayed to the user, and the expected resolution. The data management team should review edit check specifications with the clinical team before programming begins.

Database validation testing confirms that the system functions as specified before any real patient data is entered. Validation typically includes user acceptance testing, where the data management team enters test data designed to trigger each edit check and verifies that the system responds correctly. The validation results should be documented and approved before the database is released for production use.

Data Collection and Entry Practices

Site Training and Data Entry Conventions

The quality of clinical trial data depends heavily on the training and practices of site staff who enter the data. Sites should receive comprehensive training on the protocol, the CRF completion guidelines, and the electronic data capture system before enrolling their first patient.

Data entry conventions should be documented and distributed to all sites. These conventions address common questions such as how to record missing values, how to handle out-of-range laboratory results, and how to document corrections. Consistent conventions reduce the number of queries generated during data cleaning.

Sites should enter data promptly after each patient visit while the information is fresh and source documents are available. Delayed data entry increases the risk of transcription errors and makes query resolution more difficult because the source information may be harder to locate.

Query Management and Resolution

A query is a formal request for clarification or correction of a data element. Queries are generated by edit checks, manual data review, or monitoring activities. Each query should clearly state the problem, reference the specific data element, and request a specific action from the site.

Query management requires a systematic approach to tracking, resolution, and closure. The data management team should monitor query aging to ensure that sites respond in a timely manner. Unresolved queries at the time of database lock can delay the entire study.

Sites should respond to queries by reviewing the source data and either confirming the original entry or providing a correction. The response should include sufficient explanation to document the resolution. All query activity is recorded in the audit trail to provide a complete history of data corrections.

Monitoring and Source Data Verification

Clinical monitors visit sites to verify that the trial is conducted according to the protocol and that the reported data is accurate and complete. Monitoring activities include source data verification, review of regulatory documents, and assessment of site compliance with the protocol and applicable regulations.

The monitoring plan defines the frequency and extent of monitoring visits based on the study risk assessment. High-risk studies or sites may require more frequent monitoring with a higher percentage of source data verification. Risk-based monitoring approaches focus verification activities on the data elements that are most critical to the study conclusions.

Monitoring findings are documented in visit reports and communicated to the site for corrective action. Significant findings may require additional training, re-verification of data, or in extreme cases, termination of the site from the study.

Data Cleaning and Quality Control

The Data Cleaning Process

Data cleaning is the systematic process of identifying and resolving errors, inconsistencies, and missing values in the clinical database. This process begins during data collection and continues until database lock. The goal is to produce a clean data set that accurately reflects the clinical observations.

The cleaning process combines automated edit checks with manual review by data managers. Automated checks identify out-of-range values, inconsistent dates, and missing required fields. Manual review examines the data for patterns that automated checks might miss, such as implausible changes between visits or inconsistencies across related data elements.

Data cleaning is an iterative process. Each round of cleaning generates queries that sites must resolve. After the queries are resolved, the data management team reviews the updated data to confirm that the corrections are appropriate and that no new issues have been introduced.

Data Quality Metrics and Tracking

Data management teams should track quality metrics throughout the study to identify problems early and measure cleaning progress. Common metrics include the number of open queries, query resolution time, data entry error rates, and the percentage of data elements that have been verified against source documents.

These metrics should be reviewed regularly by the data management team and reported to the study leadership. Trends in query volume can indicate training needs at specific sites or problems with specific CRF fields. Early identification of these issues allows corrective action before the problems become widespread.

The data cleaning status should be documented at regular intervals, typically through data cleaning status reports that summarize the number of patients enrolled, the completeness of data entry, and the number of open queries by site and by CRF form.

Medical Coding of Adverse Events and Medications

Medical coding is the process of standardizing clinical terminology using established dictionaries. Adverse events are coded using the Medical Dictionary for Regulatory Activities, and concomitant medications are coded using the World Health Organization Drug Dictionary. Standardized coding enables consistent analysis and reporting across the study.

Coding should be performed by trained coders who understand the clinical context and the coding conventions. Each adverse event term is mapped to the most appropriate dictionary term, and the coding decisions are documented for audit purposes. Coding discrepancies should be reviewed and resolved according to the coding conventions.

The coding process should begin early in the study and continue as new adverse events are reported. Coding completion is a prerequisite for database lock because the analysis data sets require coded terms.

The Database Lock Process

Database Lock Readiness Assessment

Database lock is the formal process of freezing the clinical database so that no further changes can be made. The lock occurs after all data has been collected, cleaned, coded, and verified. The database lock readiness assessment confirms that all required activities have been completed before the lock is authorized.

The readiness assessment should verify that all patients have completed their required visits, all data has been entered into the database, all queries have been resolved, all adverse events have been coded, and all serious adverse events have been reconciled with the safety database. The assessment should also confirm that all external data, such as laboratory results and central imaging readings, has been integrated into the database.

A database lock checklist provides a structured approach to the readiness assessment. Each item on the checklist should be verified and documented by the responsible individual. The completed checklist becomes part of the trial master file and provides evidence that the lock was performed appropriately.

The Lock Authorization Process

Database lock requires formal authorization from the study leadership. The data manager prepares a lock request that summarizes the data cleaning status and confirms that all readiness criteria have been met. The request is reviewed by the study team, including the statistician, medical monitor, and project manager.

The lock authorization should be documented in writing, typically through a database lock notification that identifies the date and time of the lock and the individuals who authorized it. After the lock, the database is made read-only, and no further data modifications are permitted.

Any changes to the locked database require a formal process for database unlock. The unlock process should be documented in the standard operating procedures and should require justification for the change, approval from the study leadership, and a complete audit trail of the modification.

Post-Lock Activities and Data Archival

After the database is locked, the data is exported for statistical analysis. The analysis data sets are created from the locked database and provided to the statistician. The locked database is archived along with all supporting documentation, including the CRF, edit check specifications, query logs, and lock documentation.

Data archival ensures that the trial data remains available for regulatory inspection, audit, and future research. The archival process should preserve the data in a format that can be read and understood in the future. The archive should include the data dictionary, the annotated CRF, and any software or system information needed to interpret the data.

The archived data should be stored in a secure location with controlled access. The retention period should comply with applicable regulatory requirements and institutional policies. Access to the archived data should be documented and limited to authorized individuals.

Electronic Trial Master File and Regulatory Context

The eTMF and Document Management

The electronic trial master file is the centralized repository for all essential documents generated during a clinical trial. These documents include the protocol, investigator brochures, ethics approvals, site contracts, monitoring reports, and data management documentation. The eTMF provides the complete documentary record of the trial.

Data management documentation that belongs in the eTMF includes the data management plan, CRF completion guidelines, edit check specifications, database validation report, query logs, and database lock documentation. These documents demonstrate that the data management activities were performed according to the approved procedures.

The eTMF should be maintained throughout the trial and finalized after the study is complete. Document management practices should ensure that documents are complete, current, and accessible to authorized personnel. The eTMF provides the evidence base for regulatory inspections and audits.

Regulatory Expectations for Data Management

Regulatory agencies expect clinical trial data to be accurate, complete, and verifiable. The data management practices described in this article align with the expectations articulated in regulatory guidance documents. Organizations should be familiar with the applicable regulations and guidance for their jurisdiction and study type.

The Bioanalytical Method Validation Guidance from the U.S. Food and Drug Administration addresses the validation of analytical methods used to measure drug concentrations in biological samples. While this guidance focuses on bioanalytical methods, it illustrates the broader regulatory expectation that all data supporting a marketing application must be generated under controlled, documented conditions.

The Laboratory Quality Management System Handbook from the World Health Organization describes the quality system elements that laboratories should implement to ensure reliable results. These elements include document control, records management, and internal audits, which parallel the quality controls required in clinical data management.

Good Documentation Practices

Good documentation practices apply to all records created during a clinical trial, whether paper or electronic. These practices require that records are legible, accurate, and complete. Corrections should be made in a way that preserves the original entry and documents the reason for the change.

For paper records, corrections should be made with a single line through the original entry, the correction written nearby, and the change initialed and dated. The original entry must remain legible. For electronic records, the audit trail automatically captures the original entry, the correction, the user who made the change, and the date and time.

Organizations should train all personnel on good documentation practices and monitor compliance through internal audits. Consistent application of these practices ensures that the trial records can withstand regulatory scrutiny.

Common Failure Patterns in Clinical Data Management

Incomplete or Inconsistent Data Entry

One of the most common failure patterns is incomplete data entry at the site level. Sites may miss required fields, enter data for the wrong visit, or fail to enter data in a timely manner. These problems generate queries, delay cleaning, and can compromise the completeness of the analysis data set.

Prevention strategies include comprehensive site training, clear CRF completion guidelines, and regular monitoring of data entry completeness. The data management team should track entry completeness by site and follow up promptly when sites fall behind.

Poor Query Resolution Practices

Sites sometimes respond to queries without actually reviewing the source data, or they provide responses that do not adequately explain the resolution. These practices create data quality problems because the correction may be based on guesswork instead of the actual source record.

Data management teams should review query responses for adequacy and return queries to the site when the response is insufficient. Training should emphasize that query responses must be based on source data review and should include enough detail to document the resolution.

Delayed Medical Coding

Medical coding that falls behind can delay database lock significantly. Coding backlogs accumulate when adverse events are not coded promptly or when coding questions are not resolved in a timely manner. The coding team should monitor coding status throughout the study and escalate issues that require medical review.

Inadequate Database Validation

Databases that are not thoroughly validated before production use can contain programming errors that affect data quality. These errors may not be detected until the data cleaning process, when they are more difficult and expensive to fix. Comprehensive validation testing with well-designed test cases reduces the risk of programming errors reaching the production database.

Limitations and Professional Escalation Criteria

Limitations of Automated Data Cleaning

Automated edit checks and data cleaning tools cannot identify all data quality issues. Some errors require clinical judgment to detect, such as an adverse event that is coded to the wrong term or a laboratory value that is implausible given the patient's clinical presentation. Data management teams should combine automated checks with manual review by experienced personnel.

Research on data cleaning technologies continues to evolve. The Research on instance-level data cleaning technology paper from the 2021 International Conference on Artificial Intelligence Big Data and Algorithms describes approaches to cleaning individual data records. The UniClean: A Multi-Signal Fusion Pipeline for Optimizing Data Cleaning Workflow paper from the 2025 International Conference on Data Engineering presents a pipeline that integrates multiple signals to optimize cleaning workflows. These technologies may improve efficiency, but they do not replace the need for human review of critical data.

When to Escalate Data Quality Issues

Data management teams should escalate data quality issues to the study leadership when the issues cannot be resolved through routine query management. Escalation criteria include persistent data entry problems at a specific site, unresolved discrepancies in critical safety data, and data integrity concerns that suggest possible misconduct.

The escalation process should be documented in the data management plan. Escalation typically involves the medical monitor, project manager, and sponsor representatives. In cases of suspected fraud or serious noncompliance, the issue may need to be reported to the institutional review board or regulatory authorities.

Professional Judgment in Data Management

Data management requires professional judgment in situations where the data does not clearly indicate the correct resolution. For example, a laboratory value that is outside the normal range but consistent with the patient's clinical condition may not require correction. The data manager should consult with the medical monitor when clinical judgment is needed to interpret the data.

Professional judgment also applies to decisions about whether to query a data element. Excessive queries burden sites and can delay the study, while insufficient queries risk leaving data quality problems unresolved. The data management team should calibrate their query practices based on the study risk profile and the specific data element.

Safety Data Management and Reconciliation

Adverse Event and Serious Adverse Event Handling

Adverse event data requires special attention in clinical data management because of its importance to patient safety. Adverse events are recorded on the CRF and coded using the Medical Dictionary for Regulatory Activities. Serious adverse events are also reported to the sponsor and regulatory authorities according to the protocol and applicable regulations.

The data management team must ensure that adverse event data is complete, accurate, and coded consistently. The medical monitor reviews adverse event data to assess safety signals and to ensure that events are classified appropriately.

Safety Database Reconciliation

Serious adverse events are typically recorded in both the clinical database and the safety database used for expedited reporting. Reconciliation is the process of comparing the two databases to ensure that all serious adverse events are consistently recorded in both systems.

Reconciliation should occur at regular intervals throughout the study and must be completed before database lock. The reconciliation process identifies discrepancies in event details, causality assessments, and outcome information. All discrepancies should be resolved and documented before the lock.

The Laboratory Biosafety Manual from the World Health Organization addresses the safe handling of biological materials in laboratory settings. While this manual focuses on laboratory safety instead of clinical data management, it illustrates the broader institutional commitment to safety that should inform all clinical research activities.

Records, Measurements, and Documentation

Essential Data Management Records

The data management team should maintain a complete set of records that document all data management activities. These records include the data management plan, CRF completion guidelines, edit check specifications, database validation documentation, query logs, coding decisions, and database lock documentation.

Each record should be version-controlled and stored in the eTMF. The records should demonstrate that the data management activities were planned, executed, and documented according to the approved procedures.

Measuring Data Management Performance

Data management performance can be measured through metrics that track efficiency and quality. Efficiency metrics include the time from last patient visit to database lock, query resolution time, and the number of cleaning cycles required. Quality metrics include the number of unresolved queries at lock, the rate of data entry errors, and the number of data corrections after lock.

These metrics should be tracked over time and compared across studies to identify opportunities for improvement. Organizations should review their data management metrics regularly and use the findings to refine their processes and procedures.

Practical Implementation Steps

Building a Data Management Plan

The data management plan is the foundational document that describes how data will be handled throughout the study. The plan should be written during the study startup phase and approved before the database is released for production use.

The plan should describe the data collection instruments, the database system, the edit check strategy, the query management process, the medical coding approach, and the database lock procedures. The plan should also identify the roles and responsibilities of the data management team and the timelines for key activities.

Conducting a Database Lock Readiness Review

The database lock readiness review is a structured assessment that confirms all data management activities are complete before the lock is authorized. The review should be conducted by the data manager and presented to the study leadership for approval.

The readiness review should verify that all patients have completed their required visits, all data has been entered and cleaned, all queries have been resolved, all coding is complete, and all external data has been integrated. The review should also confirm that the safety database reconciliation is complete and that all documentation is in order.

Preparing for Regulatory Inspection

Regulatory inspections may review the data management processes and records for a clinical trial. The inspection may focus on the data integrity controls, the query management process, the coding decisions, and the database lock documentation.

Organizations should be prepared for inspections by maintaining complete and organized records, training staff on the inspection process, and conducting internal audits before the inspection. The data management team should be able to explain their processes and demonstrate that they followed the approved procedures.

Frequently Asked Questions

What is the difference between source data and case report form data?

Source data is the original record where a clinical observation is first documented, such as the medical chart or laboratory report. Case report form data is the information transcribed into the study database. The case report form is not considered a source document unless it is the original location where the data is recorded. Source document verification compares the case report form entries against the source records to confirm accuracy.

How long does the data cleaning process typically take?

The duration of data cleaning depends on the study size, complexity, data quality, and the number of sites. A typical cleaning period ranges from several weeks to several months after the last patient visit. Studies with poor data quality or delayed site responses may require additional cleaning cycles. The data management plan should include a realistic timeline for cleaning activities.

What happens if a data error is found after database lock?

If a data error is found after database lock, the study team must follow the documented process for database unlock. The process requires justification for the change, approval from the study leadership, and a complete audit trail of the modification. Unlocking the database is a significant event that should be avoided through thorough cleaning before the lock.

What is the role of the data manager in a clinical trial?

The data manager is responsible for planning and executing the data management activities for the trial. This includes designing the case report form, building and validating the database, managing the query process, overseeing medical coding, and coordinating the database lock. The data manager works with the clinical team, statistician, and sites to ensure that the data is complete, accurate, and ready for analysis.

How does electronic data capture improve data quality?

Electronic data capture improves data quality through built-in validation checks that identify errors at the time of entry, audit trails that document all changes, and real-time visibility that allows data managers to monitor entry progress. Electronic systems also reduce transcription errors by allowing direct entry at the point of care and support electronic patient-reported outcomes that capture data directly from participants.

What is the purpose of medical coding in clinical trials?

Medical coding standardizes clinical terminology so that adverse events and medications can be analyzed consistently across the study. Adverse events are coded using the Medical Dictionary for Regulatory Activities, and medications are coded using the World Health Organization Drug Dictionary. Coding enables the study team to summarize safety data and identify patterns that might not be apparent from the raw terms.

What documents are needed for database lock?

The database lock requires documentation that all data management activities are complete. This includes the database lock checklist, the query resolution summary, the coding completion report, the safety database reconciliation report, and the lock authorization notification. These documents are archived in the trial master file and provide evidence that the lock was performed appropriately.

How should sites handle missing data in the case report form?

Sites should follow the CRF completion guidelines for handling missing data. In general, missing values should be documented instead of left blank, and the reason for the missing value should be recorded when possible. The data management team will generate queries for missing data to determine whether the data is truly unavailable or was simply not entered.

Related Articles

• Multiplex Real-Time RT-PCR for Differential Diagnosis of Canine Infectious Respiratory Disease Complex (CIRDC): Panel Design, Validation, and Clinical Utility
• qPCR Data Presentation: How to Report Results in Figures and Tables
• Cloud-Based Diagnostic Data Integration for Herd Health Management
• Multiplex RT-qPCR for Differential Diagnosis of Canine Respiratory Pathogens: Panel Design, Analytical Sensitivity, and Clinical Application
• Digital Droplet PCR (ddPCR) for Absolute Quantification of Swine Enteric Coronaviruses in Fecal and Oral Fluid Samples: Analytical Sensitivity and Clinical Utility

References and Further Reading

• Laboratory Quality Management System Handbook. World Health Organization.
• Laboratory Biosafety Manual. World Health Organization.
• Assay Guidance Manual. National Center for Advancing Translational Sciences.
• Bioanalytical Method Validation Guidance. U.S. Food and Drug Administration.
• NCBI Literature Resources. National Center for Biotechnology Information.
• PubMed. National Library of Medicine.
• Final database lock results of the phase 2 cohort of lenvatinib + pembrolizumab for progressive disease after a PD-1/PD-L1-containing therapy in metastatic clear cell renal cell carcinoma. 2023.
• What’s new on the UniProt website: features and improvements for 2026. 2026.
• Graph informed biomarker discovery framework using transcriptomic machine learning for glioblastoma prognosis.. 2026.
• Reducing blood culture contamination in the ED: impact of Kurin Lock® implementation.. 2026.
• Standalone RFID Access Control System with Data-Integrity Verification Capabilities.. 2026.
• Research on instance-level data cleaning technology. Proceedings 2021 International Conference on Artificial Intelligence Big Data and Algorithms Caibda 2021, 2021.
• UniClean: A Multi-Signal Fusion Pipeline for Optimizing Data Cleaning Workflow. Proceedings International Conference on Data Engineering, 2025.
• Application of Hadoop based distributed data cleaning technology in periodical meta data integration. Proceedings 2017 10th International Symposium on Computational Intelligence and Design Iscid 2017, 2018.
• An RFID Data cleaning strategy based on maximum entropy feature selection. Journal of Digital Information Management, 2015.
• Data quality analysis and cleaning strategy for wireless sensor networks. Eurasip Journal on Wireless Communications and Networking, 2018.

This article is educational and does not replace institutional policy, professional advice, or applicable safety and regulatory requirements.