Who Pays for Expertise?

A conflict of interest is not the same as corruption

NIH describes a conflict of interest as a situation in which a financial, professional, personal, or other secondary interest may influence professional judgment.[1] ICMJE uses a similar logic:

a conflict exists when a secondary interest could influence judgment regarding a primary interest, such as the validity of research or the welfare of patients.[2]

The important word is:

could.

A conflict of interest is not proof of:

• fraud;
• fabrication;
• deliberate lying.

It is a risk of bias that should be:

• disclosed;
• managed;
• sometimes eliminated.

Why do we need disclosure at all?

Because otherwise the reader cannot evaluate the context. Imagine two studies of the same drug. The first says:

funded by an independent public grant.

The second says:

funded by the drug manufacturer; the sponsor participated in study design, analysis, and preparation of the manuscript.

Both may contain perfectly accurate data. But the second calls for closer attention to:

• design;
• access to data;
• selection of outcomes;
• manner of interpretation.

ICMJE therefore requires disclosure of:

• financial and other relevant relationships;
• the source of support;
• the funder's role in design, data collection, analysis, and interpretation;
• its role in writing;
• any restrictions on publication.[2]

“Industry funded” is not a scientific refutation

One of the most common mistakes is:

“The company paid for the study, therefore the study is wrong.”

That is an argument about motive, not an analysis of the research. A better formulation is:

“Because there is a direct financial interest, we should examine more closely the points where that interest could have affected the research process.”

Then we look at:

• the protocol;
• randomization;
• the comparator group;
• preregistered outcomes;
• participant attrition;
• statistics;
• researchers' access to raw data;
• consistency between results and conclusions.

But funding is not irrelevant

The opposite extreme is:

“Only the method matters. Who pays is irrelevant.”

The empirical evidence does not support that position. A Cochrane systematic review of 75 studies on sponsorship of drugs and medical devices found that industry-sponsored studies more often reported efficacy results favorable to the sponsor's product and more often reached favorable conclusions.[3]

In the pooled analysis, industry-sponsored studies were:

• about 27% more likely to report favorable efficacy results;
• about 34% more likely to report favorable conclusions,

compared with studies funded from other sources.[3] This is an association, not proof that every individual industry-funded study is biased.

An interesting problem: standard risk-of-bias measures did not explain everything

Cochrane found something especially important. Differences between industry-funded and non-industry-funded studies were not fully explained by classic elements such as:

• randomization;
• allocation concealment;
• participant attrition;
• standard selective reporting.

The authors therefore concluded that there appears to be industry bias that ordinary methodological risk-of-bias tools do not capture completely.[3] That means:

a well-executed study can still be designed around a question that is advantageous to the sponsor.

Bias can begin before the first patient

Imagine a new medicine. A sponsor may legitimately decide:

• which disease to study;
• which population;
• which comparator drug;
• at what dose;
• which outcome will be primary;
• how long follow-up will last.

Each choice may be scientifically defensible. But the combination of choices can increase the probability of a favorable result. That is why it is useful to distinguish:

fabrication bias

from

design bias.

A biased result does not necessarily require fabricating a single number.

The research question is already a selection

We often imagine science as:

data → conclusion.

In reality, another step comes first:

which question will be funded at all?

If industry funds research on:

how to reduce the risk of its product,

it may fund less research on:

whether the product is needed in the first place.

If the state strongly funds one field, that field develops more:

• laboratories;
• careers;
• datasets;
• publications.

That is not abuse by itself. It shows that funding also influences the agenda of knowledge.

What is not researched cannot enter the consensus in the same way

This resembles agenda-setting from earlier articles. If a hypothesis is:

• expensive;
• commercially unattractive;
• politically unpopular;
• institutionally undesirable,

it may receive less research funding. The result is not necessarily:

bad science.

It may be:

an asymmetric amount of knowledge.

So when examining a broad consensus, we should also ask:

Which questions have been systematically studied, and which have received much less attention?

Publication bias: the scientific literature is not a complete archive of all studies

Even if every published study is methodologically correct, the literature as a whole can still create a distorted picture. Why? Because positive or statistically interesting results are easier to publish than:

• null;
• negative;
• unexciting results.

This is publication bias. If ten studies find no effect, one finds an effect, and mainly that one gets published, the reader sees:

“there is a positive result”

instead of:

“one of eleven studies produced a positive result.”

Selective reporting within a published study

A study may measure:

• pain;
• quality of life;
• hospitalization;
• laboratory values;
• several time points.

If the publication emphasizes mainly the outcomes that are:

• statistically significant;
• favorable;
• consistent with the preferred story,

we get selective outcome reporting. Cochrane warns that selective inclusion of outcomes and analyses based on their results can also mislead systematic reviews.[4] That is one reason preregistration matters so much.

Preregistration is a time machine against changing the story after the fact

If a researcher publicly records before analysis:

• the main hypothesis;
• the primary outcome;
• the statistical plan;

we can later compare:

what they intended to test

with

what they eventually published.

Preregistration does not prevent every form of bias. But it makes it harder to retroactively turn:

an accidental interesting finding

into

the original main hypothesis.

A sponsor can influence the process after data collection too

ICMJE specifically requires disclosure of:

• whether authors had access to the data;
• what kind of access they had;
• whether the sponsor imposed publication restrictions.[2]

Why does this matter? Because there is a major difference between:

an academic author having the full dataset and conducting an independent analysis

and

an academic author receiving the sponsor's analysis and writing from that.

Both situations may appear under the same label:

“industry-funded trial.”

An academic author on the title page does not necessarily mean academic control over the study

A BMJ cross-sectional study of academic-industry collaboration in clinical trials found that sponsors often retained or owned data, while publications did not always make clear what access academic authors had.[5] The study also found evidence of ghost authorship in some of the trials examined.[5] So the question is not only:

Who is listed as an author?

It is:

Who designed the protocol, who had the data, who analyzed them, and who wrote the paper?

Ghostwriting

Ghostwriting in scientific publishing means that an important contributor to writing or preparing the research is not properly identified. That is a problem for two reasons.

First

The reader cannot see the real origin of the text.

Second

Responsibility becomes unclear. If an industry statistician prepared key analyses, an academic name on the title page does not by itself reveal the full process. BMJ and related research have documented ghost authorship in some industry-funded clinical trials.[5][6]

“Doubt is our product”

One of the most famous documents in the history of industrial influence on science is an internal Brown & Williamson document from 1969. It contains the line:

“Doubt is our product ...”

The document goes on to describe doubt as a means of competing with the “body of fact” in the public mind and of establishing controversy.[7] This is a primary document. It matters because it is not merely an accusation made by critics of the tobacco industry.

It is an internal record of a communications strategy.

The tobacco industry: a documented case of managing scientific doubt

Later research into internal tobacco-industry documents revealed a broader pattern. In work on environmental tobacco smoke, Drope and Chapman documented:

• networks of industry-linked scientists;
• a significant role for lawyers in directing research activity;
• funding of organizations that appeared more independent from industry;
• use of symposia and other channels to spread preferred conclusions;
• examples in which unfavorable research did not reach the public.[8]

This is much stronger evidence than:

“the company funded the study.”

Here we have a documented strategy for influencing the production and communication of knowledge.

So not every conflict is the same

Compare:

Example A

A pharmaceutical company funds a registered randomized trial; the protocol is public; independent authors have the full data; and the sponsor has no right to block publication.

Example B

Industry defines the question, selects the data, controls the analysis, conceals the funder's role, and prevents unfavorable results from being published. Both studies are:

“industry funded.”

Epistemically, they are not equivalent. So we need degrees of conflict and control.

Sugar and coronary heart disease: a useful historical case that requires caution

Kearns, Schmidt, and Glantz analyzed internal Sugar Research Foundation documents from the 1960s in 2016. According to their historical review, the SRF:

• funded a literature review on coronary heart disease;
• set the project's objective;
• supplied literature;
• received drafts;
• did not have its financial and substantive role disclosed in the published review.[9]

The authors concluded that industry helped steer the debate away from the potential role of sucrose and toward fat and cholesterol.[9]

The history of sugar does not prove that “fat was innocent”

This is an important safeguard against creating a new simplification. A document showing industry influence does not automatically prove:

the opposite nutritional hypothesis is entirely correct.

A historical analysis of funding primarily speaks to:

• agenda;
• disclosure;
• sponsor influence.

The actual causes of coronary heart disease require the full body of later biomedical evidence. The broader principle is:

evidence that a debate was manipulated is not automatically evidence that the opposing scientific thesis is correct.

Industry is not the only source of interests

Financial conflict is easiest to see:

a company wants profit.

But ICMJE also points to:

• professional relationships;
• academic competition;
• personal rivalries;
• intellectual beliefs.[2]

A researcher may invest twenty years of a career in one theory. Even without a single euro of industry money, they have an interest:

in the theory surviving.

So the scientific system does not rest on the idea:

“scientists have no interests.”

It rests on:

methods and institutions that allow other people to check their work.

Public funding is not perspective-free either

The state can set:

• research priorities;
• strategic areas;
• defense programs;
• health goals.

That does not mean:

publicly funded research is propaganda.

It means:

public funding also contains an agenda about which questions are worth studying.

So “publicly funded” is not synonymous with:

completely interest-free.

Its advantage is often a different accountability structure and less direct commercial interest in a specific outcome.

Philanthropy also shapes the research agenda

Large foundations can finance:

• diseases;
• technologies;
• social programs;
• research centers.

Their support can be extremely valuable. At the same time, every major funder influences, through its priorities:

which questions receive institutions, teams, and career paths.

Funding transparency should therefore be a universal standard, not a requirement imposed only on companies.

A think tank is not a university, and that is not necessarily a problem

A think tank often exists precisely to:

• analyze public policy;
• propose solutions;
• influence public debate.

That is a legitimate function. But the reader should know whether a document is being read as:

• an academic study;
• policy analysis;
• an advocacy paper;
• an industry-funded report.

The main problem appears when a normative or interest-driven report is presented as neutral expertise without visible context.

Who funds the think tank?

For a politically sensitive policy document, it is therefore useful to check:

• who funds the organization;
• whether donors are public;
• whether a funder has a direct interest in the subject of the report;
• whether the funder had editorial control;
• whether the methodology is described adequately;
• whether the report uses primary data;
• whether it can be replicated.

The identity of the funder is context. It is not an automatic answer to the question of truth.

“Follow the money” is a good first rule and a poor final one

Following the money can reveal:

• motive;
• conflict;
• networks;
• links between institutions.

But if we end the investigation with:

“This person received money from X, therefore their result is false,”

we have not examined the research. A better rule is:

follow the money → then follow the method.

Critics of industry can have interests too

If we demand sponsor transparency, the standard should apply symmetrically. A researcher studying industry influence may receive support from:

• a public agency;
• a foundation;
• an advocacy organization.

That does not discredit the work. It means:

disclosure should remain visible even when we like the researcher's conclusion.

In the article on the history of sugar-industry funding, JAMA Internal Medicine later published a correction because one source of travel support had been omitted from the acknowledgments.[10] Even that small example illustrates the principle:

disclosure rules should apply to all sides.

WHO and the commercial determinants of health

WHO uses the concept commercial determinants of health for private-sector activities and the broader political-economic conditions through which commercial actors can positively or negatively influence health.[11] Among problematic practices, the organization lists:

• lobbying;
• donations;
• misinformation;
• influence on public policy.

At the same time, WHO explicitly recognizes that the private sector is an important partner in:

• development of medicines;
• vaccines;
• infrastructure;
• health services.[11]

This is a useful model for our series:

do not demonize the private sector, but clearly analyze conflicting incentives.

A conflict of interest is a reason for safeguards, not automatic disqualification

NIH lists possible ways of managing conflicts such as:

• public disclosure;
• independent monitoring;
• modifying the research plan;
• changing personnel responsibilities;
• reducing or eliminating the financial interest;
• ending the conflicting relationship.[1]

This is more sophisticated than:

“A conflict exists → the researcher is compromised.”

The question is:

Is the conflict manageable, or serious enough to require exclusion?

Consensus is not a vote by funders

Scientific consensus is not:

how many researchers support a claim.

At its best, it means:

different methods, groups, and datasets converge on a similar conclusion.

That is why the funding of one study becomes less decisive if we have:

• multiple independent teams;
• different countries;
• different methods;
• replications;
• meta-analyses;
• public data.

This is the power of convergence of evidence.

But a meta-analysis can inherit the bias of its ingredients

If the literature is:

• selectively published;
• funded homogeneously;
• oriented toward similar questions,

then even a statistically sophisticated meta-analysis is limited by the information that exists. Modern evidence-synthesis methodology therefore emphasizes analyzing:

• what was published;
• what may be missing;
• how questions were framed;
• whether commercial conflicts are present.[12]

Statistics cannot recover data that were never published.

Open data reduces dependence on reputation

If researchers can obtain:

• the protocol;
• anonymized data;
• analysis code;
• documentation;

the question shifts from:

“Do we trust the author?”

toward:

“Can other people verify the result?”

That is one of the strongest ways to reduce dependence on personal reputation or funding source.

Replication is better than moral speculation

If there is concern that a sponsor's interest influenced the result, the strongest response is often:

independent replication.

If another group:

• does not have the same sponsor;
• uses a similar method;
• obtains the same result,

confidence increases. Not because the first conflict disappeared. But because the conclusion is no longer dependent on one interest structure.

Expert consensus and dissent

Expert agreement has epistemic weight because no individual can check everything personally. But dissent has a role too. A good system needs:

• a dominant model;
• the possibility of criticism;
• publication of negative results;
• methods that allow a minority argument to win if its evidence is better.

A minority position is not automatically courageous and correct. A majority position is not automatically purchased. The deciding question is:

Which model survives scrutiny better?

“All experts are bought” is a self-sealing theory

If someone says:

“Every expert who disagrees with me is funded by the system,”

it becomes almost impossible to present counterevidence. If the expert has no visible conflict:

the conflict is hidden.

If the expert does have one:

proof.

If multiple experts reach the same conclusion:

proof of coordination.

This is the same epistemological problem we discussed in relation to conspiracy theories. Conflict of interest must not become a universal device for discrediting an unwanted result.

“No conflicts of interest” is not a certificate of truth either

Conversely:

the author reports no financial conflicts

does not mean:

the article is correct.

There can still be:

• methodological errors;
• bad statistics;
• confirmation bias;
• poor data;
• intellectual commitments;
• chance.

Disclosure is one layer. It does not replace peer review, replication, or methodology.

Who chose the comparator?

In clinical studies, a very useful question is:

What was the new product compared with?

With:

• placebo?
• the best existing treatment?
• an inferior dose of a competitor?

A study can be technically impeccable and still answer the question that is commercially most favorable. That is why comparator choice belongs in an assessment of sponsor influence.

Who chose the primary outcome?

If a sponsor wants to demonstrate success, it matters:

• which outcome counts as primary;
• how long patients are followed;
• whether the measure is a symptom or a clinical event;
• whether a surrogate endpoint is used.

Again:

this is not automatic proof of manipulation.

It is a point where the protocol and public registration should match.

Who had the right to say “publish”?

This is one of the most important questions. If a sponsor can:

• delay publication;
• prohibit publication;
• demand a change in the conclusion,

research independence is substantially weaker. ICMJE therefore requires disclosure of restrictions on manuscript submission and of the sponsor's role in writing.[2]

Who had the raw data?

If an academic author cannot independently inspect the raw data, their ability to take responsibility for the conclusion is limited. So in major disputed studies, THY-REALITY should check:

DATA ACCESS.

It is not enough to say:

“Professor X is a coauthor.”

EXPERTISE / FUNDING CHAIN

For future articles, I would introduce a standard model:

FUNDER → RESEARCH QUESTION → PROTOCOL → DATA COLLECTION → DATA OWNERSHIP → ANALYSIS → AUTHORS → PEER REVIEW → PUBLICATION → POLICY USE

At each link we ask:

• who had influence;
• what was disclosed;
• what was preregistered;
• who had the right to stop or change the process.

Five degrees of sponsor independence

In practice, studies can be described—not morally ranked—along five levels:

Level 1 – funding without an operational role

The sponsor provides money; the research team has an independent protocol, data, analysis, and publication rights.

Level 2 – participation in design

The sponsor contributes to the protocol, but data and publication remain independently controlled.

Level 3 – the sponsor has a data/analysis role

Stronger independent oversight is needed.

Level 4 – the sponsor controls the manuscript or publication

High risk of interpretive and publication bias.

Level 5 – concealed sponsor role

Funding, writing, or strategic involvement is not disclosed. This is a substantially more serious transparency problem.

What to check when an “expert” appears in the media

For a television or online expert, ask:

1. What is their actual field of specialization?
2. Are they speaking within that field?
3. Where do they work?
4. Who funds their relevant research or organization?
5. Do they have consulting contracts?
6. Are the data they cite public?
7. Are they describing a research finding or expressing a political/moral opinion?
8. Are there other independent experts reaching a similar conclusion?

This is more useful than:

“Is this expert ours or theirs?”

What to check in a think-tank report

1. Is the funder disclosed?
2. Does the funder have a direct interest in the issue?
3. Is the report peer reviewed or editorially reviewed?
4. Is the method reproducible?
5. Are the data accessible?
6. Are alternatives treated fairly?
7. Does the title go beyond the actual findings?
8. Is the document research or advocacy?

A think tank can produce excellent work. Its institutional label does not automatically give it academic weight.

What to check in a scientific study

For THY-REALITY:

FUNDING ≠ VERDICT

In addition to the funder, check:

• design;
• sample;
• comparator;
• primary outcomes;
• preregistration;
• statistical power;
• missing data;
• multiplicity;
• data access;
• authorship;
• publication history;
• independent replication.

The strongest evidence of influence is not the money—it is a documented intervention

If we discover:

“the company paid for the research,”

we have a conflict-of-interest risk. If we discover:

“the company selected the outcome, blocked publication of another result, and concealed its role,”

we have much stronger evidence of influence. That distinction must remain visible.

Conflict-of-interest policy can itself become ritual

An institution may have a form:

“Disclose your conflicts.”

Everyone fills it in. But nobody acts. Disclosure without management can become:

transparency theatre.

NIH therefore does not require only disclosure; it may also require a management plan involving independent monitoring, changes to the research design, or changes in personnel responsibilities.[1]

A good system does not require pure people

This is one of the most important lessons. We do not need researchers without:

• careers;
• ambition;
• funders;
• beliefs.

Such people do not exist. We need a system in which:

• conflicts are visible;
• methods are public;
• data can be checked;
• findings can be replicated;
• criticism is permitted;
• the funder cannot invisibly control the entire process.

THY-REALITY must not use the funder as a shortcut

This is an important editorial rule for the project. We will not write:

“This research was funded by X, therefore we cannot trust it.”

The correct form is:

“The research was funded by X. That creates a relevant conflict/interest context. We then examine what role the funder had and whether independent sources confirm the result.”

The same applies when the funding supports a conclusion we like.

Conclusion: money can shape knowledge, but money is not proof of truth or falsehood

Research needs funding. So completely “unfunded science” is not a realistic ideal. A better goal is:

transparently funded, methodologically verifiable, and independently corrigible science.

Money can influence:

• the question;
• methodology;
• publication;
• interpretation;
• communication.

History contains well-documented examples in which that influence was strategic and hidden. It also contains a vast amount of industry-funded research without which we would not have:

• many medicines;
• medical devices;
• new technologies.

So we do not need the formula:

industry = falsehood.

We need the harder question:

Who paid, what could they influence, who had the data, and can the result survive verification outside the interest structure that financed it?

If it can, the source of money becomes less important. If we cannot even see how money shaped the process, it becomes more important. The best scientific system therefore does not demand:

“Trust the expert.”

It demands:

“Show the expertise, disclose the interests, and let others verify the result.”

Methodological note

This article does not claim that:

• industry-funded science is automatically wrong;
• publicly or philanthropically funded science is automatically independent;
• a conflict of interest proves corruption;
• an association between sponsorship and favorable results in a particular literature proves bias in every individual study;
• evidence of industry influence on a historical debate by itself proves the opposing scientific thesis correct.

Funding is treated as one layer in evaluating evidence, not as a substitute for methodological review.