In April 1965, an English statistician named Austin Bradford Hill stood up in front of the Section of Occupational Medicine at the Royal Society of Medicine and gave his presidential address.¹ Hill had spent his career trying to figure out what causes disease in populations, and he had just helped prove that smoking causes lung cancer. The question he wanted to talk about that afternoon was a simple one: when you see an association between two things, how do you decide whether one of them is causing the other?
Hill offered what he called nine viewpoints, but he stressed they were not the only criteria and ironically not a checklist. No single viewpoint could prove causation by itself, and Hill did not require all nine to be satisfied for a claim to hold. He made it clear that they were nothing more than the normal questions an honest person should ask before deciding that A causes B. The nine are association, consistency across studies, specificity, temporality, biological gradient, plausibility, coherence with existing knowledge, experimental evidence, and analogy.
Sixty years later, Hill's nine viewpoints are still used by epidemiologists to evaluate causal claims about population health. The framework can be applied to cancer clusters, environmental exposures, vaccine safety signals, or even drug adverse events. They are the closest thing public health has to a rulebook for deciding what we know and how we know it.
On April 28, 2026, RealClearInvestigations published a 6,000-word article by Paul Thacker entitled, "COVID Cover-Up: Hiding Star Researcher Ralph Baric's Ties to Global Pandemic." The piece argues that virologist Ralph Baric of the University of North Carolina played a central role in creating SARS-CoV-2 through gain-of-function research conducted with the Wuhan Institute of Virology, and that he and his colleagues then orchestrated a years-long cover-up to hide that role. The article opens by noting that nearly two-thirds of Americans now believe the virus came from a Chinese lab.²
I am not going to argue that the lab-leak hypothesis is false. The origin of SARS-CoV-2 is a legitimate scientific question, and reasonable people can reasonably disagree about how it should be investigated. My point is to ask what evidence is put forward to support its causal claim, and whether that evidence meets the standard Hill laid out for us in 1965.
The claim itself needs to be identified cleanly, because the article bounces around between different versions of itself. The strongest version, which is crystal clear in the closing paragraphs, holds that the COVID virus was created in Baric's lab. A senior HHS official is quoted saying, "Baric designed the gun. But the Chinese built it, and then they pulled the trigger." Former CDC director Robert Redfield is quoted saying there is "a real possibility that the virus's birthplace was Chapel Hill." Thacker has asserted a causal claim about a population-level event that killed an estimated 20 million people. Fortunately, we have Hill's framework, which he created to evaluate exactly this type of claim.
Strength of association is the first viewpoint, and Hill considered it the most useful. A weak association can still be causal, but a strong one is harder to explain away through bias or confounding. The smoking-and-lung-cancer association came in around 9-to-1; smokers were nine times more likely to develop lung cancer than non-smokers. Thacker doesn't put forward any quantitative association at all. He can't, because the underlying claim is about a single virus and a single research program, not a population-level exposure with measurable cases. What he substitutes for strength of association is a series of suggestive details: Baric pioneered gain-of-function research; the 2018 DEFUSE proposal described inserting a furin cleavage site into bat coronaviruses; and SARS-CoV-2 has a furin cleavage site. The pandemic started in the same city where the proposed work would have been done. Each of these is a temporal or geographic coincidence that may or may not mean anything. None of them is a measured association, because there is nothing here to measure.
Consistency across independent studies is Hill's second viewpoint. Causal claims gain credibility when multiple investigators using different methods arrive at the same finding. Thacker doesn't cite any convergent body of work. What he does reference is a small group of scientists and journalists who have come to the same conclusion through the same line of documentary evidence (leaked emails, the DEFUSE proposal, congressional depositions). The sources are not independent investigators, they are participants in a single investigative thread, citing each other and reaching the same conclusion. Hill's notion of consistency requires more than a chorus. What Thacker describes is what happens when one source provides material to several outlets and each one repeats it.
Specificity asks whether the proposed cause produces the observed effect specifically, or whether it would predict many possible outcomes. Thacker's argument is weakest here. The DEFUSE proposal, even if it had been funded and the experiments conducted, described a research program that could have produced any number of viral constructs. He asserts that the furin cleavage site in SARS-CoV-2 matches what DEFUSE proposed to engineer, but furin cleavage sites occur naturally in many coronaviruses, and the specific sequence in SARS-CoV-2 (PRRAR) is not what an engineer trying to optimize transmissibility would design. Andersen and colleagues argued in their 2020 Nature Medicine analysis that the sequence is suboptimal for that purpose.³ If you accept the article's logic, the same proposal predicts dozens of possible engineered viruses, only one of which would be SARS-CoV-2. Specificity, in Hill's framework, means that the cause must be tied to the effect in a way that excludes alternatives. Thacker doesn't do that. He assumes them away.
Temporality is the one Hill viewpoint Thacker clearly satisfies. The DEFUSE proposal was submitted in 2018, and the pandemic began in late 2019. The cause precedes the effect. Hill, however, was clear that temporality is necessary for causation but not nearly sufficient. Many things happened in 2018 that preceded the pandemic. The wildlife trade in southern China continued. Bat coronavirus surveillance generated thousands of sequences. Multiple labs around the world conducted coronavirus research. Citing the DEFUSE proposal as the relevant antecedent requires assuming the conclusion the temporality is supposed to support.
Biological gradient (the dose-response relationship) is unaddressed in the article for good reason given the nature of the claim. Plausibility is the one viewpoint where the article has a reasonable case to make. Lab origin is biologically plausible, and gain-of-function research does carry risks that have been understood for years. Hill warned explicitly, though, that plausibility is the weakest of the viewpoints because it depends on what current science considers possible. Almost any causal claim can be made plausible if you describe it in enough detail.
Coherence asks whether the proposed causal explanation fits with everything else known about the disease. Thacker runs into trouble here too, and doesn't even acknowledge it. The early epidemiology of SARS-CoV-2, the patterns of viral diversity in the Huanan Seafood Market, the absence of any pre-pandemic detection of the virus in lab surveillance records, and the genomic features that point toward a recent zoonotic origin are all unaddressed. He treats coherence as if it ran in only one direction.
Experimental evidence is where the Hill's framework eviscerates the article. Hill considered experimental or quasi-experimental evidence the strongest support for a causal claim. The article puts forward no experimental evidence whatsoever. No recovered viral sequences from a Baric or Wuhan lab match SARS-CoV-2. No laboratory records confirm the proposed constructs were ever made. Thacker cites no biosafety incident reports, no documented exposures, no early infections among lab personnel that would build a chain of transmission from lab to population. The absence is described as a cover-up, which may or may not be true, but an absence of evidence is not evidence of cover-up.
Analogy, the ninth viewpoint, asks whether similar exposures have produced similar effects. Yes, lab-acquired infections do happen. After all, the SARS-1 virus escaped from research laboratories on at least four documented occasions between 2003 and 2004.⁴ It's a reasonable analogy to support the plausibility of the lab origin idea in a general sense, though it do not support the specific claim that Baric's work caused this pandemic.
Grade the article honestly against Hill, and you get one viewpoint partially satisfied (temporality), one with a defensible analogy (analogy), one with a weak but reasonable claim (plausibility), and six viewpoints either unaddressed, weak, or contradicted. Hill himself would have laughed and refused to call this a causal argument. He spent the second half of his 1965 paper warning against Thacker's brand of reasoning, where a series of suggestive observations get arranged into a narrative that feels like enough proof to satisfy Columbo.
The lab-leak hypothesis itself may yet turn out to be correct. The honest answer about where the pandemic started is... honestly, we do not know yet, and we may never know with the level of certainty an epidemiologist would want. Sure, politics and the like will always have their answer. What I am saying is that the RealClearInvestigations article is doing work different from investigating origins. Thacker holds up a causality he molded from evidence that does not meet Hill's standards. He then asks us to accept it because the documents are damning, the personnel actions are suggestive, and the timing is striking. Those are journalistic instincts, and they are not unreasonable to use as jumping points. None of them, on their own or together, makes an argument that holds up to epidemiological standards for causation.
The 1998 Lancet paper by Andrew Wakefield proposed that the MMR vaccine causes autism in children.⁵ Wakefield's paper was retracted in 2010, and the British General Medical Council struck him from the medical register for fraud, but the claim has had a longer half-life than the paper.⁶
In the Wakefield case, the strength of association relied on only twelve children with no comparison group. Subsequent population studies covering millions of children in Denmark, Finland, Sweden, the United Kingdom, and the United States found no association at all.⁷ Consistency across investigators came back at zero, because multiple independent research groups across multiple countries failed to reproduce the finding. The proposed mechanism (chronic measles infection in the gut producing neurological injury) would have predicted effects beyond autism that nobody ever saw, which sinks its specificity. Temporality was there but artifactual, because MMR is administered around the age when autism is typically diagnosed, which makes it look like a temporal association. No dose-response relationship between vaccination and autism risk has ever been found, eliminating the biological gradient. Plausibility ran low against what is known about autism's prenatal origins, and coherence was contradicted by the genetic and prenatal-imaging evidence on autism etiology. When investigators went looking for experimental evidence, they found no measles virus RNA in the bowel tissue of children with autism that matched the controls.⁸ No other vaccine has ever been linked to autism despite extensive search, leaving analogy in the dust as well.
Wakefield's claim fails Hill's framework on every viewpoint that counts. The RealClearInvestigations article mostly fails it too. The two have a structure in common. Each relies on temporal coincidence, suggestive but unmeasured association, and a narrative arranged from fragments into something that looks like a picture of causation. An epidemiologist trying to write a causal-inference paper would toss both out the nearest airlock. The only thing either piece has done is persuade large factions of the public.
When the reasoning behind Wakefield and Thacker gets treated as authoritative, readers learn that loose causal inference is what investigative rigor looks like. The same readers, coming across whatever RFK Jr. is going to come up with next, or a new viral report about VAERS, or the next anti-vaccine pamphlet making its rounds in their kid's school, will evaluate each by the same flimsy standards. The damage adds up, and it comes with a body count.
On October 2, 2025, the South Carolina Department of Public Health confirmed a measles outbreak in the Upstate region.⁹ Eight cases at the start, centered in Spartanburg County, where MMR coverage in schools was 88.9% against the 95% threshold needed for herd immunity.¹⁰ One elementary and middle school in the wider Upstate region had only 17% of students with required immunizations.¹⁰ Six months later, on April 27, 2026, state officials declared the outbreak over after 42 days without a new case. By then 997 people had been infected, 953 of them unvaccinated, and 21 had been hospitalized.¹¹ The Spartanburg cluster was the largest measles outbreak in the United States in more than 35 years.¹²
The Spartanburg outbreak followed a smaller but deadlier one in West Texas earlier in 2025, in Gaines County, where kindergarten MMR coverage was 77% and two unvaccinated children died.¹³ Nationally, the United States recorded 2,288 confirmed measles cases in 2025, the highest total since the disease was declared eliminated in 2000.¹⁴ National MMR coverage among kindergartners fell from 95.2% in the 2019-2020 school year to 92.5% in 2024-2025, leaving about 286,000 kindergartners without protection.¹⁵ Non-medical exemptions hit 3.4%, an all-time high.¹⁵
I am not specifically blaming Paul Thacker, who has done legitimate work on conflicts of interest in biomedical research over many years. His article belongs to a journalistic tradition that has produced findings of consequence. The trouble is that he applies that tradition to a question (population-level disease causation) that needs a different evidentiary standard than the one journalism alone can satisfy. Hill's nine viewpoints are not a stylistic preference; they are the minimum bar for claiming that A caused B in a population, and he just does not clear it.
The Spartanburg outbreak ended the way Hill would have wanted it to. South Carolina health workers gave 81,096 MMR doses statewide during the outbreak, a 31.3% increase over the previous year, and doses in Spartanburg County itself nearly doubled.¹¹ Parents who had previously claimed exemptions watched their neighbors' children get sick and changed their minds. The framework worked because enough people stopped treating doubt as evidence and started treating exposure as a reason to act. All it cost was 997 cases of a preventable disease in one corner of one state, plus two dead children in another, plus whatever the next outbreak claims before its community decides to come to its senses.
References
- Hill AB. The environment and disease: association or causation? Proc R Soc Med. 1965;58(5):295-300.
- Thacker PD. COVID Cover-Up: Hiding Star Researcher Ralph Baric's Ties to Global Pandemic. RealClearInvestigations. April 28, 2026. https://www.realclearinvestigations.com/articles/2026/04/28/covid_cover-up_campaign_to_hide_star_researcher_ralph_barics_ties_to_global_pandemic_1179562.html
- Andersen KG, Rambaut A, Lipkin WI, Holmes EC, Garry RF. The proximal origin of SARS-CoV-2. Nat Med. 2020;26(4):450-452.
- Lim PL, Kurup A, Gopalakrishna G, et al. Laboratory-acquired severe acute respiratory syndrome. N Engl J Med. 2004;350(17):1740-1745.
- Wakefield AJ, Murch SH, Anthony A, et al. Ileal-lymphoid-nodular hyperplasia, non-specific colitis, and pervasive developmental disorder in children. Lancet. 1998;351(9103):637-641. [Retracted]
- Editors of The Lancet. Retraction of an interpretation. Lancet. 2010;375(9713):445.
- Madsen KM, Hviid A, Vestergaard M, et al. A population-based study of measles, mumps, and rubella vaccination and autism. N Engl J Med. 2002;347(19):1477-1482.
- Hornig M, Briese T, Buie T, et al. Lack of association between measles virus vaccine and autism with enteropathy: a case-control study. PLoS One. 2008;3(9):e3140.
- South Carolina Department of Public Health. DPH confirms measles outbreak in Upstate region. October 2, 2025. https://dph.sc.gov/news/dph-confirms-measles-outbreak-upstate-region
- South Carolina Department of Public Health. 2025 measles outbreak. https://dph.sc.gov/diseases-conditions/infectious-diseases/measles-rubeola/2025-measles-outbreak. Accessed April 30, 2026.
- Soucheray S. South Carolina measles outbreak ends as US cases near 1,800. CIDRAP News. April 28, 2026. https://www.cidrap.umn.edu/measles/south-carolina-measles-outbreak-ends-us-cases-near-1800
- WBTV Web Staff. South Carolina says Upstate measles outbreak is over after weeks without new cases. WBTV. April 27, 2026. https://www.wbtv.com/2026/04/28/south-carolina-says-upstate-measles-outbreak-is-over-after-weeks-without-new-cases/
- Centers for Disease Control and Prevention. Measles update, United States, January 1 to April 17, 2025. MMWR Morb Mortal Wkly Rep. 2025;74(14).
- Centers for Disease Control and Prevention. Measles cases and outbreaks. https://www.cdc.gov/measles/data-research/index.html. Accessed April 30, 2026.
- Seither R, Yusuf OB, Calhoun K, et al. Vaccination coverage with selected vaccines and exemption rates among children in kindergarten, United States, 2024-25 school year. MMWR Morb Mortal Wkly Rep. 2025. Cited in International Vaccine Access Center summary, Johns Hopkins Bloomberg School of Public Health.