Research Waste: The $170 Billion Problem Decided Before the First Patient Enrolls

The most expensive fact in health research is also the least comfortable: an estimated 85% of research investment is avoidably wasted — close to $170 billion a year (Chalmers & Glasziou, 2009; Glasziou & Chalmers, 2016). Not because science is uncertain. Because of problems that are correctable.

That estimate comes from Iain Chalmers and Paul Glasziou, two of the most cited methodologists in medicine. Their arithmetic is plain: health research spends on the order of $200 billion a year globally, and the losses accumulate across four successive stages — questions that did not need answering, designs that could not answer the question being asked, results that were never published, and reports too unusable to change clinical practice. Because every study must pass through all four stages, the waste is cumulative: roughly half lost at each of the later stages compounds to more than 85% (Chalmers & Glasziou, 2009).

If you lead a research team, fund studies, or make policy from evidence, this is not an academic footnote. It is the largest avoidable line item in biomedical spending — and most of it is decided before the first patient enrolls.

Five faces of waste

A 2024 scoping review of the entire research-waste literature classified the phenomenon into five aspects: methodological (flaws in study design, conduct, or analysis), invisible (nonpublication, discontinuation, lack of data sharing), negligible (redundant or low-value research), underreported (biased or incomplete reporting), and structural (systemic features such as perverse incentives) (Rosengaard et al., 2024).

The striking finding for anyone planning a study: more than half of everything ever written about research waste concerns the methodological kind — flaws in design, conduct, and analysis (Rosengaard et al., 2024). And design comes first. A question chosen badly, a plan left unwritten, a dataset that cannot answer the causal question: each of these decides the fate of the study long before a single participant is recruited.

A bill you can count: the $535,000 amendment

Waste is not an abstract aggregate. It shows up in your spreadsheet as protocol amendments — changes to a finalized study protocol that require internal approval, then ethics or regulatory review.

In a Tufts Center for the Study of Drug Development analysis of more than 800 protocols, 57% of protocols had at least one substantial global amendment, and nearly half of those amendments were deemed avoidable. The median direct cost of implementing one: $141,000 for a Phase II protocol and $535,000 for a Phase III protocol, plus an average of three additional months of study conduct (Getz et al., 2016). The share of protocols requiring at least one substantial amendment has since climbed — to 76% in the 2024 update (Getz et al., 2024).

Read that number the right way and it is a diagnosis: every avoidable amendment is a design flaw that was not caught when it cost nothing to fix — it was caught after enrollment began, when fixing it costs half a million dollars and three months.

The cost that does not appear on a budget

The dollar figures understate the damage, because research waste has human as well as economic consequences (Chalmers & Glasziou, 2009).

Patients are enrolled in trials that cannot answer the question they were designed to answer — exposure to experimental risk with no possibility of knowledge in return. Funders commit resources to studies that duplicate work already done. Policy and treatment decisions are shaped by evidence that later fails to replicate. And each uninformative or unpublished study compounds the problem, because the next team cannot build on results that were never produced or never reported. The cheapest moment to prevent all of it is the same moment in every case: before the study begins.

Where the waste is decided: the design stage

The patterns that produce waste are not a single villain. They are pervasive expressions of the broader validity crisis, and each one is checkable:

• Questions not built on existing evidence. New research should not be done unless the question cannot already be answered by systematic reviews of what exists. In a sobering audit, only 11 of 24 trial authors were even aware of the relevant systematic reviews when they designed their new trials (Chalmers & Glasziou, 2009).
• Plans that do not exist yet. A protocol and analysis plan that are improvised after the data are gathered invite results that reflect the analyst's choices rather than the world — the flexibility problem known as HARKing (hypothesizing after results are known) (Ioannidis et al., 2014).
• Data that cannot answer the question. Observational data — electronic health records, claims, registries — is collected for care and administration, not for research. Using it to support causal claims requires methods that account for what it cannot see; treating association as causation is among the most studied sources of invalid findings (Rothman, Greenland, & Lash, 2008).
• Samples that manufacture their own answer. Who is missing from your data decides what your data says — conditioning on a collider (for example, admission to hospital) forces correlations that do not exist in the wider population (Rothman, Greenland, & Lash, 2008).

None of these is exotic. All of them are detectable at the design stage, where the cost of correction is a day of review rather than a $535,000 amendment.

What prevents it

The remedies are known, published, and cheap:

1. Start from what is already known. Require that new studies build on systematic reviews of existing evidence (Chalmers & Glasziou, 2009).
2. Lock the plan before the data. Register the protocol at inception and pre-specify the analysis; reporting guidelines such as CONSORT and STARD make results usable (Chalmers & Glasziou, 2009).
3. Bring methodologists in at design time. The median cost of one Phase III amendment would fund extensive methodological review of dozens of protocols — review that costs a fraction of the amendment it prevents (Getz et al., 2016).
4. For causal questions from real-world data, emulate the trial you cannot run. When randomization is unethical or impractical, target trial emulation designs the analysis of existing data as a careful imitation of the trial you would have run — removing the easy ways to be wrong (see our guide on causal inference).
5. Make the protocol checkable, not just readable. A protocol written as structured logic can be checked by deterministic rules the way an IDE checks code — which is the design philosophy behind Studio by Outcome Project: the Scientific Validity Linter reviews your design as you write it, before a single patient enrolls. The AI assists; the deterministic checks decide. Validation you can trace, not a black box you must trust — and you stay the scientist.

Rigor is a product, not a virtue

At Outcome Project, we believe research waste is eliminated at the point of design. Our work separates research that merely analyzes data from research that can stand in front of a decision:

• For researchers: build and stress-test your protocol inside Studio — try it free at studio.outcomeproject.com, with deterministic checks as you write.
• For institutions: see the environment applied to your own teams with a demo for your organization.
• For public-health authorities and teams designing studies now: the Outcome Squad runs specialized epidemiological services — design validation, causal inference from real-world data, and surveillance — built on exactly these principles: outcomeproject.com/services/outcome-squad.

The $170 billion is an estimate. That it is preventable is not. Waste is decided at the design stage — and so is its opposite: research worth every dollar invested.

References

• Chalmers I, Glasziou P. Avoidable waste in the production and reporting of research evidence. The Lancet. 2009;374(9683):86–89.
• Glasziou P, Chalmers I. Is 85% of health research really "wasted"? The BMJ (blog). 2016. https://blogs.bmj.com/bmj/2016/01/14/paul-glasziou-and-iain-chalmers-is-85-of-health-research-really-wasted/
• Getz KA, Stergiopoulos S, Short M, et al. The impact of protocol amendments on clinical trial performance and cost. Therapeutic Innovation & Regulatory Science. 2016;50(4):436–441. (Tufts Center for the Study of Drug Development)
• Getz K, Smith Z, Botto E, Murphy E, Dauchy A. New benchmarks on protocol amendment practices, trends and their impact on clinical trial performance. Therapeutic Innovation & Regulatory Science. 2024;58(3):539–548.
• Ioannidis JPA, Greenland S, Hlatky MA, et al. Increasing value and reducing waste in research design, conduct, and analysis. The Lancet. 2014;383(9912):166–175.
• Rosengaard LO, Andersen MZ, Rosenberg J, et al. Five aspects of research waste in biomedicine: a scoping review. Journal of Evidence-Based Medicine. 2024;17(2):351–359.
• Rothman KJ, Greenland S, Lash TL. Modern Epidemiology. 3rd ed. Philadelphia: Lippincott Williams & Wilkins; 2008.