[Conflict of Interest Management Practice] Case Study Summary - Background of Conflict of Interest, Governance, and Loss of Trust as Seen Through Cases
In this series, we have considered conflicts of interest not as abstract concepts, but concretely through actual incidents and research studies.
This page organizes the cases covered in each installment by the issues that should be verified in practice.
The problem of conflict of interest cannot be fully understood by looking only at the specificities of individual incidents.
It is important to confirm what interests and relationships influenced decision-making, whether it should be viewed as an individual problem or a problem that the organization should grasp and manage, and where the issues lay in terms of disclosure, supervision, information integration, and initial response.
Please use this as a reference page when considering "which issues are similar in my organization" and "what should be managed to prevent similar problems."
Detailed explanations of each case are provided in the relevant articles for each installment. This page will be updated as the series progresses.
■ 1 Cases Related to Individual Decision-Making Bias
● Research study on the relationship between pharmaceutical companies and physicians
・What happened:
Analysis in the United States showed that even the provision of small meals averaging around $20 resulted in a statistically significant increase in the prescription rate of specific drugs.
・Essence of the problem:
Conflicts of interest do not arise only from large monetary exchanges; even small benefits can influence an expert's judgment.
・Implications:
Conflict of interest management is not a matter of amount, but must be grasped through the relationship of "who is involved in which decision, and which interest that judgment is linked to."
→ Part 10 Explained in "1. Even small benefits can influence decision-making"
● Diederik Stapel case
・What happened:
A large-scale data fabrication case by a Dutch social psychologist.
・Essence of the problem:
Non-financial interests, such as fame, achievements, publication in high-impact journals, and evaluation within the academic community, other than financial interests, can distort judgment and behavior.non-financial interests can distort judgment and behavior.
・Implications:
In conflict of interest management, it is necessary to include not only financial interests but also fame and career incentives as factors that can distort decision-making.
→ Part 11 Explained in "2. Bias created by fame and career"
● Wakefield case
・What happened:
A case where a paper suggesting a link between the MMR vaccine and autism was later retracted, leading to a finding of misconduct.
・Essence of the problem:
The point lies in the fact that a strong attachment to a research hypothesis and ties to social or political activities brought bias to research judgments.
・Implications:
Conflicts of interest are not limited to "money" issues; integration with beliefs, positions, and claims can also be factors that distort judgment.
→ Part 11Explained in "3. Bias Created by Beliefs and Policy Positions"
● National Universities/Public Hospitals - Nihon Kohden Case, MDM Case, etc.
・What happened:
Cases where funding provided under "nominally appropriate titles" such as scholarship donations, manuscript fees, or fictitious expenses was deemed to be a substantial inducement for transactions (bribery), leading to arrests or severe warnings.
・Essence of the problem:
The conflict of interest appeared not as blatant entertainment, but in the form of normal institutional operations such as "donations" or "outsourcing," and the neutrality of professional judgment (procurement or prescriptions) was substantially compromised.
・Implication:
Even if the "title" is legal, if it is linked to specific decision-making, it becomes a significant risk that leads directly to criminal penalties and the loss of social trust.
→ Part 25Explained in "1. (1) What kind of cases exist"
■ 2 Cases Regarding Conflict of Commitment, Misuse of Resources, External Activities, and Research Security
● Charles Lieber Case
・What happened:
A case where a Harvard University researcher provided false explanations regarding their relationship with China's "Thousand Talents Program," etc.
・Essence of the problem:
The relationship with their primary duties was not managed appropriately due to excessive commitment to external organizations and the lack of transparency in those relationships.
・Implication:
The issue of conflict of commitment is not about "whether one is busy," but about the disclosure of external relationships, the organization of relationships with primary duties, and the existence of mechanisms for the university to grasp and supervise them.
→ Part 9Explained in "1. Conflict of commitment is a 'time and effort issue,' but it surfaces in other forms"
● Feng “Franklin” Tao Case
・What happened:
A case at the University of Kansas where the non-disclosure of employment at an external institution became an issue. Note that the evaluation of the criminal proceedings was later overturned, and it is important to note that the evaluation by the prosecuting authorities and the final legal conclusion did not match.
・Essence of the problem:
The point is that external employment or external relationships were not sufficiently disclosed to the university, and were not appropriately managed as a conflict of commitment.
・Implication:
In managing conflict of commitment, it is necessary to confirm the relationships between side jobs, external employment, external funding, and research activities based on actual conditions, rather than stopping at formal declarations.
→ Part 9Explained in "2. (1) Feng “Franklin” Tao Case"
● Cleveland Clinic Settlement Case
・What happened:
A case where the organization's failure to manage was questioned because the researcher's other funding sources were not sufficiently disclosed in an NIH grant application, leading to the payment of a large settlement.
・Essence of the problem:
The issue was not limited to individual non-disclosure, but that the organization could not integrate the necessary information and explain it consistently.
・Implication:
Conflict of interest management, external funding management, and research security responses are not separate tasks, but must be grasped as information connected as an organization.
→ Part 9Explained in "2. (2) Cleveland Clinic Case"・Part 15"1. The risk of 'inexplicability' caused by information silos"
● Van Andel Research Institute Settlement Case
・What happened:
A case settled due to the non-disclosure of research support from overseas.
・The essence of the problem:
External support and relationships were not sufficiently grasped by the organization, leading to a failure in oversight and accountability.
・Implications:
External relationships regarding research activities should not be left to the discretion of individual researchers; organizational confirmation and continuous monitoring mechanisms are necessary.
→ 9th SessionExplained in "2. (3) Van Andel Research Institute Case"
● Dongping “Daniel” Tao Case
・What happened:
A resource misappropriation case where university research funds, equipment, and supplies were allegedly diverted for personal consulting work.
・The essence of the problem:
The boundary between primary duties and private activities collapsed, and university resources were used for private purposes without appropriate control.
・Implications:
In managing conflicts of commitment, it is necessary to clarify the boundary between public and private, including not only permission for side jobs but also the use of equipment, research funds, personnel, and time.
→ 9th SessionExplained in "3. Resource Misappropriation Type - The Breakdown of Separation Between Primary Duties and External Activities"
● Unauthorized use of university names, logos, and research results in advertisements
What happened:
In advertisements for health foods and supplements, the fact of joint research with a university is exaggerated, or logos are posted without permission, giving consumers the misconception of "quality assurance" by the university.
The essence of the problem:
This is a state where the intangible brand resources held by the university are being "misappropriated" for corporate profit-making activities, posing a reputational risk that damages the university's social credibility.
Implications:
Resource misappropriation is not limited to equipment and money. A system is needed to continuously inspect how the university's "credibility" is being used in advertisements and to stop any deviations through cooperation between legal and public relations departments.
→ 21st SessionExplained in "1. (3) Need to perceive it as misappropriation of university credibility and research results"
■ 3 Cases regarding organizational conflict of interest / Institutional COI
● Jesse Gelsinger Case
・What happened:
A case where a subject died in clinical research on gene therapy.
・The essence of the problem:
The issue was not only the financial relationship of the individual principal investigator, but also the fact that the University of Pennsylvania itself held financial interests through stocks and royalties, meaning the organization itself was conducting research while having a vested interest.
・Implications:
Conflict of interest is not just an individual issue; when a university or research institution itself has a financial interest as a party, separate management as an Institutional COI is required. Also, even if there is an alert, if there is strong "pressure to proceed" such as the continuation of research or expectations for results, it becomes impossible to make an organizational decision to stop.
→ 7th SessionExplained in "2. (2) Evaluation of stocks held by universities"
17th SessionExplained in "2. (1) Case where the decision to stop could not be made"
● Lack of transparency in independence in joint research courses and endowed chairs
What happened:
In courses receiving continuous funding from companies, there is a risk that companies may gain de facto control over research themes and the publication of results because annual reviews during operation are insufficient, beyond just the screening at the time of establishment.
The essence of the problem:
The boundary between "support through donations" and "individual work for companies" becomes blurred, and the university's judgment (research leadership and freedom of publication) becomes dependent on corporate intentions at the organizational level.
Implications:
It is not enough to just establish them; to ensure "accountability" during operation, a mechanism to continuously update relationships through annual reviews, like at Stanford University, is essential.
→ 26th SessionExplained in "5. How to apply this to joint research courses and endowed chairs in Japan"
■ 4 Cases regarding information integration and organizational grasp
● Cleveland Clinic Settlement Case [Reprinted]
・What happened:
A case where the organization failed to sufficiently integrate information that should have been disclosed, leading to a lack of consistency with grant applications and other areas, which called the adequacy of its management into question.
・Essence of the problem:
Even if the information itself exists, if it is scattered across multiple departments such as human resources, research funding, conflict of interest declarations, external activities, and research security, the organization cannot make appropriate judgments or provide explanations.
・Implication:
Without a mechanism to connect the "dots" of information into a "line," even if individual departments can see them, the organization as a whole remains blind. Information integration is not about management efficiency, but a foundation that supports the organization's accountability.
→ Part 15 Explained in "1. The Risk of 'Inexplicability' Caused by Information Silos"
● Mid Staffordshire Case
What happened:
A case where, despite the existence of multiple alerts such as feedback from the field, patient complaints, high mortality rates, and low external evaluations, the board of directors failed to integrate them and recognize them as a serious risk, thereby exacerbating the damage.
Essence of the problem:
It was not that the information was missing, but that individual signals were processed within each department (silo) and were not converted into "meaningful information" for the organization as a whole. Furthermore, a major factor in the continued inaction was that the board of directors and supervisors did not listen sufficiently to the voices of patients and staff, and failed to properly accept the concerns raised as "important signals for the organization" (a lack of listening skills).
Implication:
Without a mechanism to connect fragmented information such as side jobs, human resources, external funding, and complaint reports into a "line" and grasp the overall picture, serious governance failure will occur.
Supporting an "organization that can speak up" is not just about the courage of those raising their voices. Only when the staff and managers who first encounter the information have the "etiquette" to accept it without dismissing their discomfort, do the dots of information begin to connect into a line.
→ Part 17 "2. (3) Cases where alerts were not integrated"
Part 18 "2. (2) Enabling cross-sectional understanding of information within the organization"
Part 19 Explained in "1. Why do problems not surface even if there is a system?"
■ 5 Cases Regarding Reputation Risk and Crisis Response
● Toshiba Accounting Fraud Case
・What happened:
A case where inappropriate accounting continued for a long period, damaging the trust of the entire organization.
・Essence of the problem:
Governance failure occurred due to a combination of pressure from excessive goal setting, an organizational culture where it was difficult to speak up, and weak supervisory functions.
Also, despite the existence of many management and supervisory functions, responsibility was diffused, and there was no entity within the organization to apply an effective brake.
Although an internal reporting system existed and there were dozens of reports per year, there were zero reports regarding accounting fraud. Under the involvement of top management and the pressure of "challenge," the function to properly accept the discomfort of the field was paralyzed, and the fact that information was not converted into organizational judgment was fatal.
・Implication:
What leads to a loss of trust is not just the fraud itself, but the organizational structure that cannot deter, grasp, or correct the fraud.
Also, an "organization that can speak up" cannot be realized just by setting up a system. Governance only becomes effective when there is a human-centered response where those in a position to hear information first accept the concerns raised with empathy and act toward correction.
→ Part 13 "3. Case 1: Toshiba Accounting Fraud Case"
Part 17 "2. (2) Cases where it was ambiguous who would stop it"
Part 19 "1. Why do problems not surface even if there is a system?" Explained in
● Snow Brand Milk Products Food Poisoning Case
・What happened:
A case where the initial response, explanation, and delay in recall during a food poisoning accident intensified social criticism and caused a significant loss of trust in the company.
・Essence of the problem:
In addition to the occurrence of the accident itself, the subsequent delay in response and insufficient explanation exacerbated the damage and distrust.
・Implication:
In a crisis, grasping the facts, responding quickly, and providing sincere and transparent explanations are essential, and failure in the initial response makes the loss of trust decisive.
→ Part 13 Explained in "4. Case 2: Snow Brand Milk Products Food Poisoning Incident"
● Tylenol Incident (Johnson & Johnson)
・What happened:
In the face of a serious crisis involving product tampering, this case involved a swift nationwide recall and highly transparent communication, prioritizing consumer protection above all else.
・Essence of the problem:
It demonstrates that even if a crisis itself cannot be completely prevented, the subsequent decisions and responses significantly determine the fate of social trust.
・Implications:
In crisis management, a stance that prioritizes transparency, speed, and the protection of victims and users is the key to restoring trust.
→ Part 13 Explained in "5. Case 3: Tylenol Incident"
■ 6 Cases Regarding Industry-Academia Collaboration and AI Development for Medical Data and Specimens
● MSK–Paige.AI / Tempus Case
・What happened:
This is a case involving Memorial Sloan Kettering Cancer Center (MSK) and Paige.AI.
It showed the trend of digitizing pathology slides and using them for AI-based pathology diagnostic support technology and the development of large-scale pathology AI models.
Subsequently, in the acquisition of Paige by Tempus, large quantities of digital pathology slides, clinical data, and specimen analysis data were mentioned in a way that linked them to corporate value.
・Essence of the problem:
Pathology slides are not just medical records or research materials; they can lead to AI models, diagnostic support software, foundation models, and corporate value.
The question is what medical institutions provide, what kind of deliverables it turns into, and how it can be explained to patients.
・Implications:
When providing medical data and specimens, it is necessary to confirm the purpose of use, ownership of deliverables, reuse, and conflict of interest management, looking ahead not only to the research purpose at the time of provision but also to developments in AI models, derivative products, and corporate acquisitions or licensing.
→ Part 30 Explained in "2. (1) Case 1: Pathology slides lead to AI models and corporate value: MSK–Paige.AI / Tempus"
● I-MED / Harrison.ai / Annalise.ai Case
・What happened:
This is a case where Australia's I-MED Radiology Network provided patient data to Annalise.ai for the development and training of diagnostic imaging support AI models.
The Office of the Australian Information Commissioner (OAIC) concluded its preliminary investigation without taking regulatory action, as the provided data was deemed sufficiently de-identified.
・Essence of the problem:
Even if data is de-identified and judged not to fall under personal information under the law, the issue remains as to how much patients anticipated the use of their image data for AI development, how medical institutions explained it, and how the company's reuse and handling of derivative products were contracted.
・Implications:
When using medical images for AI development, one should not be satisfied with de-identification alone; it is necessary to confirm the purpose of use, duration of use, additional training, use for other products, and the handling of models or derivative data after the contract ends before providing the data.
→ Part 30 Explained in "2. (2) Case 2: Using radiological images for AI training: I-MED / Harrison.ai / Annalise.ai"
● UK NHS Foresight AI Case
・What happened:
This is a case involving Foresight AI, which develops large-scale medical AI models using UK NHS medical records and other data.
Data covering approximately 57 million people in England and over 10 billion medical events was targeted, but concerns were raised from the clinical front lines regarding the use of GP data, and the project was temporarily suspended.
・Essence of the problem:
Even if names and other identifiers are removed and data is handled in a secure environment, using it for AI development beyond the initial purpose or the expectations of patients and medical staff can damage social trust.
Legal and technical safety alone does not constitute full accountability.
・Implication:
When using large-scale routine clinical data for AI development, it is necessary to be in a state where you can explain to patients and clinical settings whose data is being used, for what purpose, by which institution, and based on what procedures.
→ 30th EditionExplained in "2. (3) Case 3: Using Routine Clinical Data for Large-Scale AI: UK NHS Foresight AI"
● Roche v. Foresight Diagnostics / Stanford Case
・What happened:
This is a case where Roche sued Stanford University and the US company Foresight Diagnostics, alleging misappropriation of trade secrets regarding technology that detects cancer by analyzing DNA in the blood. The defendants have denied the allegations, and the facts have not yet been determined.
・Essence of the issue:
When university researchers are involved as corporate consultants while also being involved in the establishment of university-originated startups or technology development in the same field, the question arises as to which position they obtained the information from and which research or business they are using it for. If confidential information obtained through corporate contracts, university research results, inventions, and licensed technologies cannot be distinguished later, it leads to disputes over trade secrets and intellectual property.
・Implication:
Regarding analysis methods, testing technologies, algorithms, and know-how born from medical data and samples, it is necessary to confirm them as a series of relationships surrounding the same technology, rather than processing side jobs, joint research, non-disclosure agreements, invention disclosures, licenses, and university-originated startup support separately.
→ 30th EditionExplained in "2. (4) Case 4: Blood-Derived Testing Technology and University-Originated Startups: Roche v. Foresight Diagnostics / Stanford"
● UK Biobank Consumer Site Posting Case
・What happened:
This is a case where de-identified participant data from the UK Biobank was posted on an external consumer-facing site for sales purposes.
While the UK Biobank explained that no personally identifiable information was included and that no sales took place, it suspended the access rights of the involved academic institutions and individuals as a breach of contract.
・Essence of the issue:
Even with de-identified data, if post-provision usage management is insufficient, problems such as posting on external sites, resale, or inappropriate removal can occur.
Contracts at the time of provision alone are not enough to continuously guarantee the safe use of data.
・Implication:
In the provision of biobank data, it is necessary to incorporate post-provision management, including confirmation of the accessing entity, confirmation of the affiliated institution and country, download restrictions, restrictions on removal outside the platform, log monitoring, and access suspension in case of violations.
→ 30th EditionExplained in "2. (5) Case 5: Biobank Data Management Continues After Provision: UK Biobank"
■ 7 How to Read This Page: Perspectives for Cross-Sectional Viewing
While the above cases are separate events, they can also be read through the following common points of discussion.
● Issues of Disclosure
In the cases of Lieber, Feng Tao, Cleveland Clinic, and Van Andel, the central issue is the failure to appropriately disclose external relationships, funding sources, and support relationships.
In conflict of interest management, it is important not only what is required to be declared, but also how far that information is verified and which departments are coordinated with for management.
● Issues of Non-Financial Interests
The cases of Stapel and Wakefield show that non-financial interests such as fame, research attachment, and social standing can influence judgment.
If conflict of interest management is limited to "money issues," important risks will be overlooked.
● Issues of Organizational Responsibility
The Jesse Gelsinger case and the Cleveland Clinic case demonstrate that the question is not just about individual actions, but how much a university or research institution can grasp, manage, and explain.
Without organizational-level mechanisms, relying on individual integrity or ethics is insufficient.
● Issues of Initial Response and Accountability
The cases of Toshiba, Snow Brand, and Tylenol show that the response after a problem occurs significantly impacts social trust.
Not only system design during normal times, but also the ability to explain and respond during crises is part of governance in a broad sense.
● Issues of Organizational Inaction
The greatest governance failure is caused by structural weaknesses that prevent the organization from converting signs of trouble into a decision or action to "stop it here and now." It is necessary to focus on the organization's mechanisms, system design, and decision-making structure—asking why the organization could not stop—rather than attributing it to individual problems.
● Issues of Where Responsible Judgment Lies
It is not enough for committees to exist on top of systems and organizational charts. The question is whether it is clear who has the authority and responsibility to ultimately apply the brakes (the end point of escalation) and whether it is operated as an effective flow. Having a formal mechanism and the organization actually being able to stop are different issues.
● Issues of the Organization's Initial Reception
The cases of Toshiba and Mid Staffordshire show that even if there are misgivings or alerts on the front lines, governance will not be activated if the person who first receives the information fails in their "way of receiving it." An "organization that can speak up" is not created solely by the courage of those raising their voices. Systems only become effective when those in a position to listen first possess the manners (being a good listener) of "not dismissing lightly" and "organizing and connecting appropriately." This is an essential prerequisite for breathing life into the "vessel" of governance and supporting the organization's self-cleansing function.
● Ensuring Transparency Based on Global Standards (ICMJE)
These are international standards as preventive measures for cases like the Wakefield incident. I would add that disclosing AI usage, data access, and sponsor involvement as a "commitment to transparency" is a "manner" for protecting the credibility of scientific processes.
● Due Diligence of Counterparties (Confirmation of Appropriateness)
To prevent research security risks like the Charles Lieber case in advance, this is a method of organizationally investigating the background of partners (relationships with governments/militaries and control structures) from public information. It is crucial not to just "finish after checking," but to link it to contract terms and access restrictions based on the risk.
● Issues of Post-Provision Management of Medical Data and Samples
The cases of MSK–Paige.AI, I-MED, NHS Foresight AI, Roche v. Foresight Diagnostics/Stanford, and UK Biobank show that medical data and samples can be developed into AI models, diagnostic technologies, testing know-how, licenses, and corporate value. Medical institutions and universities need to continue to confirm not only through pre-provision screening and contracts, but also what is created after provision, who uses it, and how far it is reused. It should also be clarified who has the authority to stop usage in the event of a breach of contract. It is important to treat explanations to patients/research participants, conflict of interest management for researchers/organizations, information management, and confirmation of intellectual property/trade secrets as connected practices rather than ending them as separate procedures.
──
▶︎ Click here for the series list
▶︎ Click here for the reverse lookup guide