On January 18, 1996, the National Cancer Institute and Fred Hutchinson Cancer Research Center held a press conference to announce that they had stopped a clinical trial twenty-one months early. The trial in question was called CARET — the Beta-Carotene and Retinol Efficacy Trial — and it had been running since 1985. It had enrolled 18,314 men and women at high risk of lung cancer: heavy smokers, former smokers, and workers with occupational asbestos exposure. The participants had been randomized to receive either a daily combination of 30 milligrams of beta-carotene and 25,000 IU of retinyl palmitate or a matching placebo. The expectation, grounded in roughly two decades of observational evidence, was that the supplemented group would develop fewer lung cancers and live longer than the placebo group.

The Data and Safety Monitoring Committee had reviewed the interim data and reached the opposite conclusion. The supplement group was developing more lung cancers, not fewer. They were also dying at higher rates from all causes. The point estimates were not subtle: a 28% relative increase in lung cancer incidence and a 17% relative increase in total mortality in the active-treatment arm. The committee concluded that continuing the trial would be unethical, because participants were being harmed by the intervention. They recommended early termination. The investigators agreed, the supplements were discontinued, and the participants were informed.

Two years earlier, in 1994, a parallel trial in Finland called the Alpha-Tocopherol Beta-Carotene Cancer Prevention Study — ATBC for short — had already published a similar result. ATBC had randomized 29,133 Finnish male smokers to alpha-tocopherol (vitamin E), beta-carotene, both, or placebo. The beta-carotene arm in ATBC showed an 18% increase in lung cancer incidence and an 8% increase in total mortality compared to placebo. The Finnish investigators, who had launched their trial in 1985 with the same observational-evidence hypothesis as CARET, had written cautious conclusions. They suggested the result might be a chance finding given the unexpected direction. CARET was the trial that would, if confirmatory, settle the question.

CARET settled it. Two large, well-conducted, independently-run randomized trials in different populations had both demonstrated that beta-carotene supplementation, far from preventing lung cancer in high-risk individuals, appeared to cause it. The observational literature — which had been so consistent, so biologically plausible, so apparently bulletproof — had reversed completely under direct experimental test.

This article walks through the observational evidence that motivated CARET and ATBC, the design of the trials, the early-stopping decision and what it revealed, the parallel ATBC findings, the long follow-up that confirmed persistent harm even after supplementation stopped, the mechanism that retrospectively explained the reversal, and what working strategists should take from the story when evaluating any “supplement X prevents disease Y” claim. The story is one of the cleanest examples in modern medicine of a consensus collapsing under randomized testing, and the structural lessons travel well beyond nutrition.

The 1980s Consensus On Carrots And Cancer

To understand why CARET and ATBC were launched, you have to understand how confident the field was that beta-carotene prevented cancer. The confidence was not casual. It was built on roughly two decades of converging evidence from multiple independent observational sources, and it had moved beyond academic epidemiology into mainstream public health advice.

The first signal came from large prospective cohort studies in the 1970s. Investigators following tens of thousands of adults observed that participants with higher dietary intake of carotenoid-rich foods — carrots, sweet potatoes, leafy greens, winter squash, cantaloupe — had lower rates of cancer at multiple sites, with lung cancer showing the strongest and most consistent association. The pattern survived adjustment for smoking, age, sex, and standard demographic confounders. People who ate more orange and dark-green vegetables got less cancer.

The signal was reinforced by serum biomarker studies. Investigators stored blood samples from cohort participants at baseline, followed them for cancer outcomes over years or decades, and then thawed the stored samples to measure serum beta-carotene concentrations in cancer cases versus matched controls. The pattern repeated. People who later developed cancer had, on average, lower serum beta-carotene at baseline than people who did not. The biomarker evidence was particularly compelling because it was harder to dismiss as a memory or reporting artifact: serum measurements were objective, the samples had been collected before disease onset, and the temporal ordering was clear.

A 1981 review by Richard Peto, Richard Doll, Jillian Buckley, and Michael Sporn, published in Nature, synthesized the evidence and proposed that beta-carotene specifically — as the principal dietary precursor of vitamin A and as a potent antioxidant — was a plausible cancer-preventive nutrient that warranted direct testing. The review article was titled “Can dietary beta-carotene materially reduce human cancer rates?” It answered the title with a qualified yes and called for randomized trials.

By the mid-1980s, the evidence base had broadened further. Case-control studies in the United States, Europe, and Asia consistently showed lower carotenoid intake in cancer cases versus controls. Ecological studies showed that populations with higher per-capita consumption of carotenoid-rich produce had lower cancer mortality. Mechanistic studies in cell lines and animal models showed that beta-carotene could quench reactive oxygen species, inhibit certain markers of tumor promotion, and modulate immune responses. The biological story was coherent. The epidemiological story was consistent. The public-health translation was already happening: dietary guidelines in multiple countries were emphasizing colorful fruits and vegetables, supplement manufacturers were marketing beta-carotene capsules to health-conscious consumers, and individual physicians were recommending the supplements to patients at elevated cancer risk.

The case for randomized trials was straightforward. The observational evidence was strong but not experimental. People who ate more vegetables differed from people who ate fewer in many ways beyond their carotenoid intake — they exercised more, smoked less, had higher education and income, ate less processed food, and had hundreds of other lifestyle and metabolic correlates. The standard epidemiological adjustments could handle the obvious confounders but could not rule out residual confounding by all the unmeasured factors that travel with vegetable consumption. The only way to isolate the effect of beta-carotene specifically was to randomize people to take it or not and follow them.

The National Cancer Institute funded two such trials in the United States. CARET was the larger of the two and recruited participants at elevated lung cancer risk specifically — heavy smokers and asbestos workers. The Physicians’ Health Study tested beta-carotene in a lower-risk population of male physicians. In Finland, the National Public Health Institute launched ATBC, which recruited male smokers aged 50 to 69 and tested both beta-carotene and vitamin E in a factorial design. The three trials together would constitute the definitive test.

The CARET Design

CARET was conceived as a large, simple, prevention-focused randomized trial in a population with a high enough baseline cancer incidence that a meaningful effect could be detected within a reasonable follow-up period. The principal investigators, Gilbert Omenn at the University of Washington and Gary Goodman at Fred Hutchinson, designed the trial to address the practical question that mattered for clinical and public-health translation: if you give beta-carotene plus retinyl palmitate (vitamin A) to people at high risk of lung cancer, do they get less lung cancer?

The intervention was a daily capsule containing 30 milligrams of beta-carotene and 25,000 IU of retinyl palmitate. The dose of beta-carotene was approximately ten times higher than typical dietary intake. The combination with retinyl palmitate reflected the biological rationale: beta-carotene is a vitamin A precursor and may exert some of its hypothesized chemopreventive effects through conversion to retinol, so adding the active vitamin removed any rate-limiting step.

Enrollment ran from 1985 through 1994. The trial recruited two populations. The asbestos cohort included 4,060 men with documented occupational asbestos exposure, recruited primarily through union locals and industrial-medicine clinics. The smoker cohort included 14,254 men and women aged 50 to 69 who were currently smoking at least 20 pack-years or had quit within the previous six years. Participants were randomized one-to-one to the active combination or matching placebo, with stratification by recruitment site and risk category.

Follow-up was active. Participants returned annually for symptom checks, supplement-bottle returns, and adherence counseling. Lung cancers and other cancer diagnoses were ascertained through linkage to state cancer registries, supplemented by self-report and confirmed by medical-record review. Deaths were ascertained through linkage to the National Death Index. The primary outcome was lung cancer incidence; secondary outcomes included other cancers, cardiovascular disease incidence, and total mortality.

The trial was designed for completion in 1998, with approximately 4 years of average follow-up after the end of enrollment. The original power calculations assumed a hazard ratio for lung cancer of 0.77 — that is, a 23% reduction in lung cancer in the active arm — based on the observational literature. The actual hazard ratio observed at the interim analysis was 1.28 — a 28% increase in the active arm. The trial was stopped early because the result was clear and the direction was harmful.

The Early-Stopping Decision

Data and Safety Monitoring Boards exist for a specific purpose. They review accumulating outcome data from ongoing trials at predetermined intervals, before the investigators are unblinded, and they have the authority to recommend early termination if the data show conclusive evidence of benefit, conclusive evidence of harm, or futility. Their decisions are governed by pre-specified stopping rules — typically based on group-sequential boundaries — that protect against making strong claims on the basis of random fluctuation.

The CARET Data and Safety Monitoring Committee met in late 1995. The data they reviewed showed 388 lung cancers among the 18,314 participants — 254 in the active-treatment group and 134 in the placebo group. The hazard ratio was 1.28, with a 95% confidence interval that excluded 1.0. The total-mortality data showed a 17% relative increase in the active group, again with confidence intervals that excluded 1.0. The committee crossed its pre-specified stopping boundary for harm.

The recommendation went to the trial steering committee, then to the National Cancer Institute, then to the investigators. There was, by accounts published in subsequent retrospectives, no serious internal debate about whether to stop. The data were clear. Continuing to randomize a known-harmful intervention into 18,000 participants — even ones who had volunteered for a research protocol — was not defensible. The supplements were discontinued in early 1996. The participants were informed by letter. The press conference was held on January 18.

The findings were published in the New England Journal of Medicine on May 2, 1996. The paper, by Gilbert Omenn and colleagues, reported the headline result and the breakdown by subgroup. The lung cancer excess was most pronounced in current smokers and in those with the heaviest exposure (in the asbestos cohort, asbestosis on chest X-ray, and in the smoker cohort, current rather than former smoking). The cardiovascular mortality was elevated as well, with the active group showing a 26% relative increase in death from cardiovascular disease. The paper concluded with what was, for an NEJM publication, an unusually direct policy statement: “The combination of beta carotene and vitamin A had no benefit and may have had an adverse effect on the incidence of lung cancer and on the risk of death from lung cancer, cardiovascular disease, and any cause in smokers and workers exposed to asbestos.”

The conclusion was strengthened by the parallel ATBC result. The Finnish trial had been published two years earlier, in April 1994, also in NEJM. The ATBC investigators had reported an 18% relative increase in lung cancer incidence in the beta-carotene arm versus the no-beta-carotene arm, with confidence intervals that excluded 1.0. Total mortality was 8% higher in the beta-carotene arm, also statistically significant. The ATBC paper had been cautious in interpretation — the result was unexpected, the chance of a false-positive finding could not be excluded, and the magnitude was smaller than CARET would later report. But the direction was the same, the population was similar (male smokers), and the two trials together produced a meta-analytic estimate that was both consistent across studies and unambiguously harmful.

The Physicians’ Health Study, which had tested beta-carotene in a lower-risk population of nonsmokers and former smokers, was completed in 1995 and published in 1996. It showed no benefit and no harm, with a hazard ratio of 1.00. The pattern across the three trials was coherent: in low-risk populations, beta-carotene supplementation did nothing; in high-risk smokers and asbestos workers, it caused harm.

Long Follow-Up

The CARET investigators continued to follow participants after the supplements were discontinued, with the question of whether the harm would persist, attenuate, or reverse once exposure ended. A 2004 paper in the Journal of the National Cancer Institute by Gary Goodman and colleagues reported the six-year post-intervention follow-up. The lung cancer incidence and cardiovascular mortality remained elevated in the formerly-supplemented group, although the magnitude of the excess attenuated over time. By the end of the post-intervention period, the cumulative incidence curves were beginning to converge, suggesting that the harm was at least partially reversible.

The most important finding of the long follow-up, beyond the persistence of harm, was that it ruled out the possibility that the early-stopping decision had been premature or that the original finding was an artifact of imbalances at randomization. Six additional years of follow-up — during which participants were no longer taking the supplements — produced cumulative outcome data that confirmed the original interpretation. The supplementation had increased lung cancer and cardiovascular mortality in a high-risk population, and the increase persisted for years after exposure stopped.

A 2010 systematic review and meta-analysis by Nathalie Druesne-Pecollo and colleagues in the International Journal of Cancer pooled the available randomized trial evidence on beta-carotene supplementation and cancer outcomes. The pooled analysis included data from CARET, ATBC, the Physicians’ Health Study, the Women’s Health Study, and several smaller trials, with a combined sample size well above 100,000 participants. The headline result was that beta-carotene supplementation was associated with a small but statistically significant overall increase in cancer incidence and mortality, with the effect concentrated in smokers and asbestos-exposed workers. In the non-smoking, non-exposed populations, the effect was null. The meta-analysis confirmed what CARET and ATBC had shown individually: beta-carotene supplements do not prevent cancer, and in vulnerable populations they cause it.

The Mechanism Story

A negative randomized trial is a falsification of a hypothesis. It does not automatically deliver the mechanism by which the falsified hypothesis failed. But for the beta-carotene reversal, a coherent mechanistic story emerged in the years after the trials were published, drawing on basic-science work that had been done both before and after the clinical results. The story has three components, and together they explain why the observational evidence had been so misleading.

The first component is the difference between nutrients in food and nutrients in supplements. Beta-carotene in carrots, sweet potatoes, and dark-green leafy vegetables does not occur in isolation. It is embedded in a matrix that includes hundreds of other carotenoids (alpha-carotene, lycopene, lutein, zeaxanthin, beta-cryptoxanthin, and many others), polyphenols, fiber, water, micronutrients, and a structural cellular matrix that affects bioavailability. The body absorbs beta-carotene from this matrix slowly, in modest quantities, in proportion with the absorption of many other co-occurring compounds. The pharmacokinetic profile of dietary beta-carotene is a slow, sustained, low-concentration delivery that has co-evolved with the human absorption system over millions of years.

A 30-milligram daily supplement is a different intervention. It delivers an isolated compound, in a concentrated bolus, without the co-occurring matrix, at concentrations that produce peak serum beta-carotene levels far above what dietary intake can achieve. The pharmacokinetic profile is that of a drug, not a food. And drugs at high doses can do things that the same molecules at dietary doses cannot. The conflation of “beta-carotene the molecule in a vegetable” with “beta-carotene the active ingredient in a 30-mg capsule” was, in retrospect, the central conceptual error. The observational evidence was about the former. CARET tested the latter.

The second component is the specific biology of beta-carotene in oxidative-stress environments. Beta-carotene is an antioxidant at typical dietary concentrations, but at very high concentrations and in the presence of oxygen tension and reactive species characteristic of smoke-exposed lung tissue, it can switch behavior and become a pro-oxidant. Basic-science work after the CARET results, including a series of papers by Xiang-Dong Wang and colleagues at Tufts using ferret models exposed to cigarette smoke, demonstrated that high-dose beta-carotene in smoke-exposed tissue produces oxidation products that can activate proliferation pathways and downregulate retinoic acid signaling. The same molecule that quenches free radicals at low concentration generates damaging oxidation byproducts at high concentration in inflamed tissue. The smokers were precisely the population in which this biology would matter, and they were the population in which the harm appeared.

The third component is the difference between what observational epidemiology can identify and what causes disease. The observational studies had shown that people with higher dietary carotenoid intake had less cancer. The most natural interpretation was that beta-carotene was the causal agent. But beta-carotene was a marker for a complex constellation of healthful dietary and lifestyle patterns: vegetable consumption in general, lower processed-food consumption, lower obesity, lower alcohol intake, higher socioeconomic status, more exercise, less smoking. The marker was correlated with the underlying causes, and adjustment for known confounders captured only the named ones. The randomized trial isolated the molecule from all of those confounders and revealed that the molecule itself, in isolation, did not have the effect that the marker had appeared to predict.

This pattern — observational marker that captures a constellation of healthful behaviors, randomized trial of the isolated molecule that fails to reproduce the benefit — is not unique to beta-carotene. It is the recurring lesson of nutritional epidemiology versus nutritional randomized trials. Antioxidant vitamin E supplementation failed similarly. Vitamin C supplementation showed minimal effects when randomized. Selenium for cancer prevention in well-nourished populations showed no benefit. The list is long enough that the field has, gradually and with reluctance, come to accept that single-nutrient supplementation is generally a poor strategy for chronic-disease prevention in adequately-fed populations, regardless of how clean the observational signal looked.

What Strategists Should Take From This

If you are evaluating any claim of the form “Supplement X prevents Disease Y, based on observational evidence,” CARET and ATBC give you a specific framework for discounting the claim. The framework has four steps.

Step one is to ask whether the claim is grounded in observational epidemiology or in randomized trials. Observational nutritional epidemiology is a methodologically valuable enterprise for hypothesis generation, but it is systematically vulnerable to confounding by the healthful-lifestyle constellation. People who eat more of any single healthful food tend to eat more healthful foods in general, exercise more, smoke less, and have higher socioeconomic status. Standard adjustments capture the obvious confounders but cannot rule out residual confounding by the unmeasured covariates of healthful living. The signal in an observational nutrition study should be treated as a hypothesis to be tested, not as a conclusion to be acted on.

Step two is to ask whether the proposed intervention is “the food” or “the molecule.” Eating more vegetables is a behavior that is supported by a large body of converging evidence and has a plausible biological story that does not depend on any single nutrient. Taking a 30-mg supplement of an isolated nutrient is a pharmacological intervention that may or may not reproduce the apparent benefit of the dietary pattern. The relationship between dietary intake of a nutrient and supplementation with that nutrient is not linear and is sometimes inverted. The default assumption should be that they are different interventions until a randomized trial shows otherwise.

Step three is to ask whether the population in which the intervention will be used resembles the population in which it was tested. CARET tested beta-carotene in smokers and asbestos workers and found harm. The Physicians’ Health Study tested it in low-risk physicians and found null. A claim that beta-carotene “is safe at typical doses” requires specifying for whom it is safe. Vulnerable subpopulations — defined by underlying disease, baseline exposures, genetic susceptibility, age, or co-medications — can experience harms from interventions that are neutral in the average population. The harms can be invisible in trials that did not specifically enroll the vulnerable group.

Step four is to ask whether the evidence base includes adequately-powered randomized trials in relevant populations or whether it consists primarily of observational data and mechanism. If the evidence is primarily observational and mechanistic, your prior on a positive RCT result should be modest, not high. The reversal rate of plausible-and-observational hypotheses under randomized testing in nutrition is high enough that the base rate must inform your individual judgment. A well-known empirical analysis by John Ioannidis and others has shown that a substantial fraction of highly-cited observational nutrition findings fail to replicate when tested in randomized trials. CARET and ATBC are not anomalies; they are representative cases.

The general lesson is one that the broader replication crisis literature has been making across many fields. The conditions that produce a strong, consistent, biologically-plausible observational signal — many independent studies, large sample sizes, multiple lines of converging evidence, a coherent mechanism — are not the same as the conditions that establish causation. Causation in heterogeneous human populations, with all the confounding that human behavior produces, requires randomization. When randomization is absent, the literature can reach high apparent consistency on a hypothesis that turns out to be wrong. CARET is the cleanest available example of how complete that reversal can be.

For a strategist evaluating any product, intervention, or policy that is supported primarily by observational evidence, the practical question to ask is: what would the randomized trial look like, and has it been done? If it has not, the claim is provisional. If it has been done and the result is positive, the claim is strengthened. If it has been done and the result is negative or reversed, the observational consensus is wrong, and your decision should follow the trial, not the consensus.

Sources And Verification

  • Omenn, G. S., Goodman, G. E., Thornquist, M. D., Balmes, J., Cullen, M. R., Glass, A., et al. (1996). Effects of a combination of beta carotene and vitamin A on lung cancer and cardiovascular disease. New England Journal of Medicine, 334(18), 1150–1155. DOI: 10.1056/NEJM199605023341802
  • The Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study Group. (1994). The effect of vitamin E and beta carotene on the incidence of lung cancer and other cancers in male smokers. New England Journal of Medicine, 330(15), 1029–1035. DOI: 10.1056/NEJM199404143301501
  • Goodman, G. E., Thornquist, M. D., Balmes, J., Cullen, M. R., Meyskens, F. L. Jr., Omenn, G. S., et al. (2004). The Beta-Carotene and Retinol Efficacy Trial: Incidence of lung cancer and cardiovascular disease mortality during 6-year follow-up after stopping beta-carotene and retinol supplements. Journal of the National Cancer Institute, 96(23), 1743–1750. DOI: 10.1093/jnci/djh320
  • Druesne-Pecollo, N., Latino-Martel, P., Norat, T., Barrandon, E., Bertrais, S., Galan, P., & Hercberg, S. (2010). Beta-carotene supplementation and cancer risk: A systematic review and metaanalysis of randomized controlled trials. International Journal of Cancer, 127(1), 172–184. DOI: 10.1002/ijc.25008
  • Peto, R., Doll, R., Buckley, J. D., & Sporn, M. B. (1981). Can dietary beta-carotene materially reduce human cancer rates? Nature, 290(5803), 201–208.

FAQ

Why did the CARET investigators not anticipate the harmful direction of the result?

The observational evidence was uniformly positive. Multiple cohort studies, case-control studies, and biomarker studies had shown lower cancer rates and lower mortality in people with higher beta-carotene intake. The biological story — antioxidant activity, vitamin A precursor function, immune modulation — was coherent. The pre-CARET expectation in the field was that beta-carotene would reduce lung cancer incidence by roughly 20–25%, which was the basis for the original power calculation. The possibility that the supplement would cause harm in smokers was discussed in the early planning literature but treated as a low-probability scenario that the trial would rule out. The trial did not rule it out; it confirmed it.

Does the CARET result mean people should eat fewer vegetables containing beta-carotene?

No. The CARET and ATBC results apply to high-dose isolated beta-carotene supplementation, not to dietary intake of carotenoid-rich vegetables. The observational evidence that vegetable consumption is associated with lower cancer rates is robust, and the dietary intake of beta-carotene from vegetables is far below the doses that produced harm in the trials. Current evidence supports continued recommendations to eat carotenoid-rich vegetables and does not support the use of beta-carotene supplements for cancer prevention.

Is the harm specific to smokers, or does beta-carotene supplementation harm everyone?

The randomized trial evidence shows clear harm in heavy smokers and asbestos workers. In low-risk populations such as the Physicians’ Health Study, beta-carotene supplementation showed no significant benefit or harm. The 2010 Druesne-Pecollo meta-analysis estimated a small overall increase in cancer mortality across the pooled randomized trial population, with the effect concentrated in smokers. The most defensible interpretation is that high-dose beta-carotene supplementation provides no cancer-prevention benefit in any tested population, and produces measurable harm in smokers and asbestos workers specifically.

Was the early-stopping decision controversial?

Early termination of a large prevention trial is always consequential and is rarely undertaken lightly. The CARET Data and Safety Monitoring Committee acted on pre-specified stopping rules and the data crossed the boundary for harm. There was little internal disagreement about whether to stop. There has been subsequent academic discussion about whether early termination of trials in general produces inflated effect estimates (the “early-stopping bias”), but in CARET this concern was addressed by the long post-intervention follow-up, which confirmed the original findings using additional outcome data that accumulated after the supplements were discontinued.

Did the observational evidence for beta-carotene have methodological problems that should have been flagged earlier?

In retrospect, yes. The principal vulnerability was confounding by the healthful-lifestyle constellation: people who ate more vegetables differed from people who ate fewer in many ways beyond their carotenoid intake, and the standard adjustments captured only the named confounders. The biomarker evidence added an apparent objectivity that masked the same underlying confounding (serum beta-carotene was itself a marker for the same dietary and lifestyle patterns). The field-wide expectation that the trials would simply confirm the observational signal reflects a general overconfidence in observational evidence that the broader replication-crisis literature has since documented across many fields.

Is the broader lesson that observational nutrition epidemiology is unreliable?

The narrower and more defensible lesson is that observational nutrition epidemiology is reliable for identifying dietary patterns that are correlated with health outcomes, but is unreliable for identifying single nutrients that, when isolated and supplemented, will reproduce the apparent benefit. The CARET–ATBC reversal, the Women’s Health Initiative HRT reversal, the vitamin E supplementation null trials, the high-dose vitamin C null trials, and several other randomized failures of observational nutritional hypotheses are consistent with this narrower lesson. The general advice “eat a varied diet rich in vegetables, fruits, whole grains, and legumes” is robustly supported. The general advice “take a supplement of nutrient X to prevent disease Y” is, in most cases, not supported.

How should this story affect my evaluation of current supplement-prevention claims?

The structural lesson is to discount any “supplement prevents disease” claim that is grounded primarily in observational evidence and mechanism. The base rate of such claims surviving randomized trials in nutrition is low. The claims that survive are typically ones with strong mechanistic grounding and replication in multiple randomized trials in the relevant population. If a supplement is being marketed on the basis of observational and mechanistic evidence alone, the appropriate prior is skepticism. If randomized trials in the relevant population are positive and replicated, the claim is strengthened. If the trials are negative or absent, the absence of evidence is itself informative.

replication-crisis beta-carotene caret-trial nutritional-epidemiology evidence-evaluation
Share this article
LinkedIn (opens in new tab) X / Twitter (opens in new tab)
Atticus Li

Experimentation and growth leader. CXL-certified CRO practitioner, Mindworx-certified behavioral economist (1 of ~1,000 worldwide). 200+ A/B tests across energy, SaaS, fintech, e-commerce, and marketplace verticals.

About LinkedIn Newsletter