What population-scale data can—and cannot—tell us about an increasingly consequential hypothesis
John Murphy, M.D., M.P.H., D.P.H., The COVID-19 Long-haul Foundation
Summary
The possibility that COVID-19 vaccination might influence cancer incidence or progression has become one of the most contentious questions in post-pandemic medicine. Recent reports have intensified the debate, including a September 2026 article in The Gateway Pundit alleging that evidence is accumulating linking COVID-19 vaccination with rising cancer risk and citing observations from U.S. military medical databases, South Korea, Italy and the published case literature.
The question deserves neither dismissal nor premature acceptance.
A systematic review published in 2026 identified 69 publications describing malignancies temporally associated with COVID-19 vaccination or SARS-CoV-2 infection, including 66 reports involving 333 individual patients, a South Korean cohort of approximately 8.4 million people, an Italian population cohort of approximately 296,000 residents and a U.S. military surveillance analysis encompassing approximately 1.3 million service members. The review itself explicitly concluded that these findings are signal-generating rather than causal evidence.
The South Korean study is the most striking population-level signal. Among 8,407,849 individuals, vaccination was associated with increased one-year hazards for thyroid, gastric, colorectal, lung, breast and prostate cancers, with hazard ratios ranging from 1.20 for breast cancer to 1.69 for prostate cancer.
The Italian study, involving 296,015 residents followed for 30 months, reported a modestly increased hazard of cancer hospitalization among vaccinated individuals (HR 1.23, 95% CI 1.11–1.37), but the association reversed when a 12-month lag after vaccination was imposed. All-cause mortality was lower among vaccinated participants.
Meanwhile, U.S. military surveillance has identified temporal changes in certain non-Hodgkin lymphoma subtypes after 2020, but investigators did not attribute those changes to vaccination or infection. The military population is particularly useful for longitudinal surveillance, yet it is also exposed to occupational, environmental and demographic factors that complicate causal inference.
At the same time, large-scale mortality data provide an important counterweight. A 2025 national French cohort of 28 million adults found no increase in four-year all-cause mortality associated with mRNA vaccination and instead observed a 25% lower adjusted risk of all-cause death among vaccinated individuals. A 2026 study of sudden cardiac death likewise found no increased risk following vaccination.
The evidence therefore does not support the proposition that COVID-19 vaccination has been demonstrated to cause a generalized cancer epidemic. It does support a more precise and scientifically important proposition:
Several independent datasets contain signals that justify rigorous investigation of whether COVID-19 vaccination can influence the incidence, detection, recurrence or progression of selected malignancies in particular populations.
The distinction between these statements is fundamental.
The question that should now be asked
Cancer epidemiology rarely moves from a single observation directly to causality.
It proceeds through stages:
signal → replication → association → dose response → temporality → biological plausibility → exclusion of alternative explanations → causal inference.
COVID-19 vaccine–cancer research has reached the signal-detection stage for several outcomes.
It has not yet reached the stage at which a causal population-wide effect can responsibly be declared.
That distinction is not semantic. It determines what experiment should come next.
The recent Gateway Pundit report asserts that cancer has increased following COVID-19 vaccination and cites claims of more than 150,000 excess cancer deaths since 2021.
Our examination of publicly accessible evidence could not identify a population-based analysis establishing that number as vaccine-attributable cancer mortality.
That does not mean that every concern is unfounded.
It means that the numerical claim currently exceeds what the underlying epidemiology can establish.
A remarkable South Korean signal
The most important recent epidemiologic observation comes from South Korea.
Kim and colleagues analyzed National Health Insurance data encompassing 8,407,849 individuals between 2021 and 2023. After propensity-score matching, they compared vaccinated and unvaccinated populations and examined cancer incidence during the year following vaccination.
The reported associations were:
| Cancer | Reported hazard ratio | 95% CI |
|---|---|---|
| Thyroid | 1.351 | 1.206–1.514 |
| Gastric | 1.335 | 1.130–1.576 |
| Colorectal | 1.283 | 1.122–1.468 |
| Lung | 1.533 | 1.254–1.874 |
| Breast | 1.197 | 1.069–1.340 |
| Prostate | 1.687 | 1.348–2.111 |
These estimates are sufficiently large, sufficiently consistent across several cancer sites and derived from a sufficiently large population that they cannot simply be ignored.
But neither can they be interpreted automatically as vaccine-attributable cancers.
The study is retrospective.
It is observational.
Vaccination status was not randomized.
And the one-year observation window creates a particularly important epidemiologic problem: a cancer diagnosed after vaccination need not have originated after vaccination.
Cancer is a multistage process.
A tumor may exist microscopically for years before clinical detection.
Thus:
diagnosis after vaccination ≠ cancer caused by vaccination.
This distinction becomes especially important for cancers such as thyroid, prostate and breast cancer, whose incidence is strongly influenced by screening and diagnostic intensity.
The Italian experiment points in a different direction
The Italian population study provides an important counterpoint.
Researchers followed the entire population of Pescara province aged 11 years and older—296,015 individuals—from June 2021 through December 2023. Approximately 83% received at least one COVID-19 vaccine dose and 62% received at least three doses.
Vaccination was associated with:
- lower all-cause mortality, but
- slightly higher cancer hospitalization during the primary analysis.
The reported hazard ratio for cancer hospitalization was 1.23 (95% CI 1.11–1.37).
Yet the finding became substantially less compatible with a simple causal interpretation when investigators changed the latency requirement.
When a minimum interval of 12 months between vaccination and cancer hospitalization was imposed, the association reversed.
This is extraordinarily important.
A genuinely carcinogenic exposure should generally not disappear—or reverse direction—merely because the investigators exclude events occurring during the first year.
Indeed, biological carcinogenesis generally requires latency.
The Italian results therefore illustrate the central problem with interpreting short-term post-vaccination cancer diagnoses: reverse causation, detection effects and temporal confounding may be substantial.
The military database: an important signal, but not a vaccine experiment
The U.S. military offers one of the most unusual longitudinal populations available for studying the pandemic.
The Defense Medical Surveillance System contains longitudinal medical information on active-duty personnel, allowing incidence to be examined across the pre-pandemic and post-pandemic periods.
A recent review identified a U.S. Armed Forces Health Surveillance Division analysis covering approximately 1.3 million active-component service members from 2017 through 2023.
The analysis reported changes in non-Hodgkin lymphoma incidence, including approximately a 50% increase in specified/unspecified and non-follicular NHL categories beginning around 2021, with persistent elevation in mature T/NK-cell lymphoma categories.
This observation deserves investigation.
But it does not establish that vaccination caused the increase.
The investigators themselves did not attribute the changes to either vaccination or SARS-CoV-2 infection.
Several alternative explanations exist:
- changes in diagnostic practices;
- changes in healthcare utilization;
- changes in medical surveillance;
- pandemic-related healthcare disruption;
- altered deployment patterns;
- occupational exposures;
- environmental exposures;
- SARS-CoV-2 infection itself;
- demographic changes;
- coding changes.
Military personnel are not an epidemiologically neutral population.
They experience unusual occupational exposures, including solvents, combustion products, radiation, burn-pit emissions and other potentially carcinogenic exposures.
Consequently, the military data should be regarded as an important surveillance signal requiring individual-level investigation, rather than proof of vaccine causation.
The cancer-mortality claim requires particular caution
The claim of approximately 150,000 excess cancer deaths after the vaccine rollout is much stronger than the evidence currently available.
The distinction between cancer incidence, cancer hospitalization, cancer diagnosis and cancer mortality is critical.
They are different endpoints.
An increase in cancer diagnoses can result from:
- more screening;
- increased imaging;
- improved detection;
- diagnostic substitution;
- changes in healthcare utilization;
- earlier diagnosis of previously occult tumors.
None necessarily means more people ultimately die from cancer.
Conversely, increased cancer mortality can result from delayed diagnosis or treatment without any increase in the biological initiation of cancer.
This distinction is especially important during the COVID-19 pandemic because cancer care was substantially disrupted.
Screening for breast and colorectal cancer fell sharply during 2020, while diagnostic and treatment services were delayed.
A major analysis of U.S. cancer registry data subsequently found that advanced-stage diagnoses returned toward pre-pandemic levels by 2021 for most cancer types, arguing against a simple explanation involving an enormous reservoir of undiagnosed advanced cancer.
Yet newer work indicates that patients diagnosed with cancer during 2020–2021 experienced worse short-term survival than comparable patients diagnosed before the pandemic, consistent with substantial disruption of cancer care.
Thus the pandemic itself created multiple pathways capable of changing cancer statistics independently of vaccination.
The temporal coincidence problem
The apparent inflection in cancer mortality around 2021 is central to the vaccine-cancer hypothesis.
But temporal coincidence is not temporal causation.
Three enormous population processes occurred almost simultaneously:
2020
- SARS-CoV-2 emerged;
- healthcare systems were disrupted;
- cancer screening fell;
- cancer diagnoses fell;
- treatment was delayed.
2021
- mass vaccination began;
- Delta emerged;
- SARS-CoV-2 infection continued at enormous scale;
- healthcare utilization recovered;
- previously missed cancers began returning to medical attention.
2022
- Omicron generated unprecedented numbers of infections;
- repeated infections became common;
- booster vaccination expanded;
- cancer detection continued normalizing.
A population-level cancer curve cannot distinguish these processes without individual-level exposure histories.
This is the ecological fallacy in its most consequential form:
A change occurring after vaccination at the population level does not establish that vaccinated individuals experienced the change because of vaccination.
Figure 1 | The competing explanations
CANCER DIAGNOSIS / DEATH
│
┌─────────────────┼─────────────────┐
│ │ │
▼ ▼ ▼
SARS-CoV-2 Vaccination Pandemic
infection disruption
│ │ │
inflammation immune missed
tissue injury modulation screening
immune escape rare AEs delayed care
│ │ │
└─────────────────┼─────────────────┘
│
▼
OBSERVED CANCER
STATISTICS
The central epidemiologic problem is that these pathways are correlated in calendar time.
What about biological plausibility?
A causal hypothesis cannot rest on epidemiology alone.
There must be a biologically credible pathway connecting exposure with outcome.
Several hypotheses have been proposed.
They include:
- transient immune perturbation;
- altered interferon signaling;
- changes in immune surveillance;
- effects on tumor microenvironments;
- inflammatory signaling;
- reactivation of oncogenic viruses;
- effects on dormant malignant clones;
- interactions between vaccination and pre-existing malignancy.
These mechanisms are biologically interesting.
But plausibility is not proof.
A mechanistic hypothesis becomes persuasive only when:
mechanism → measurable biomarker → exposure-response relationship → clinical phenotype
can be demonstrated.
What the regulatory evidence actually says
The regulatory record deserves careful attention because it is sometimes represented inaccurately in public discussion.
For the original Pfizer-BioNTech vaccine, the European Medicines Agency states that conventional genotoxicity and carcinogenicity studies were not performed because the vaccine components—mRNA and lipids—were not expected to possess genotoxic potential. The toxicology program included repeat-dose and reproductive/developmental studies.
That does not prove that long-term cancer effects are impossible.
It means that the preclinical program did not include conventional carcinogenicity testing.
That distinction is worth preserving.
Long-latency human outcomes cannot be completely resolved by short-duration preclinical studies.
Consequently, pharmacovigilance and long-term population surveillance remain essential.
The National Cancer Institute’s position
The U.S. National Cancer Institute states that there is no evidence that COVID-19 vaccines cause cancer, cause cancer recurrence or make cancer more aggressive. It also notes that vaccination can produce transient axillary lymph-node enlargement that may complicate imaging shortly after vaccination.
This position is consistent with the absence, to date, of a replicated causal signal establishing vaccine-induced carcinogenesis.
But it should not be misunderstood as proof that every conceivable long-term effect has been definitively excluded.
Science progresses by surveillance.
A hypothesis that is improbable but consequential deserves measurement.
A hypothesis that has not been proven should not be transformed into either certainty or taboo.
Myocarditis demonstrates why signal detection matters
COVID-19 vaccine safety provides a useful precedent.
Myocarditis was initially an unexpected safety signal.
It subsequently became established as a rare adverse effect, particularly among adolescent and young adult males after mRNA vaccination.
The existence of that genuine adverse effect demonstrates two important principles.
First:
Rare vaccine adverse events can be discovered after mass deployment.
Second:
Detection of a signal does not automatically establish its magnitude or mechanism.
The appropriate response is therefore neither reflexive dismissal nor premature causal declaration.
It is systematic investigation.
Cancer should be approached in exactly the same manner.
Figure 2 | The evidentiary ladder
CAUSATION
▲
│
┌────────────────┐
│ Replication │
├────────────────┤
│ Dose-response │
├────────────────┤
│ Specificity │
├────────────────┤
│ Temporality │
├────────────────┤
│ Mechanism │
├────────────────┤
│ Epidemiology │
├────────────────┤
│ Signal │
└────────────────┘
│
▼
CURRENT EVIDENCE
IS CONCENTRATED
HERE
The current evidence base contains meaningful signals but has not climbed the entire causal ladder.
A striking inconsistency in the available mortality evidence
If COVID-19 vaccination were causing a large population-wide increase in lethal cancer, one would expect some signal in all-cause mortality.
Yet large population studies do not show such a pattern.
A national French cohort of approximately 28 million adults reported 98,429 deaths among vaccinated individuals and 32,662 among unvaccinated individuals during follow-up, corresponding to an adjusted hazard ratio for all-cause mortality of approximately 0.75. The analysis found a particularly large reduction in severe COVID-19 mortality.
Similarly, a 2026 analysis of short-term mortality after vaccination found lower, not higher, all-cause mortality during the first three weeks after primary vaccination and after booster vaccination.
And a 2026 population study specifically examining sudden cardiac death found no increased risk following first, second or third COVID-19 vaccine doses.
These findings do not rule out a small, delayed or cancer-specific risk.
They do, however, strongly challenge the hypothesis that vaccination produced a massive immediate increase in overall mortality.
A causal model must explain the entire mortality dataset—not merely selected cancer statistics.
Why the South Korean finding nevertheless matters
It would be equally mistaken to dismiss the Korean study simply because other evidence is reassuring.
The study’s magnitude is precisely why it should trigger replication.
A hazard ratio of 1.53 for lung cancer and 1.69 for prostate cancer cannot be ignored if reproduced in independent populations.
But replication must be designed correctly.
The critical question is not:
“Can another study find an association?”
It is:
“Can another study find the same association after eliminating the major sources of bias?”
That means controlling for:
- cancer screening;
- smoking;
- obesity;
- socioeconomic status;
- healthcare utilization;
- prior SARS-CoV-2 infection;
- age;
- sex;
- comorbidity;
- occupational exposure;
- calendar period;
- vaccine type;
- number of doses;
- time since vaccination;
- prior cancer;
- family history;
- diagnostic intensity.
The latency problem
Perhaps the most important methodological issue is latency.
A one-year cancer endpoint is problematic for a causal carcinogenesis hypothesis.
Many cancers require years or decades to evolve.
If vaccination truly initiates carcinogenesis, the expected pattern should therefore include:
longer latency → increasing divergence in incidence
rather than simply:
vaccination → diagnosis within months.
Conversely, if vaccination accelerates a pre-existing tumor, shorter latency could be biologically plausible.
The two hypotheses therefore make different predictions.
Hypothesis A: initiation
Vaccination creates a new malignant clone.
Predicted pattern:
vaccination → latency → increasing cancer incidence
Hypothesis B: acceleration
Vaccination alters the immune environment of an already existing tumor.
Predicted pattern:
vaccination → short latency → progression/recurrence
These hypotheses should not be conflated.
The most important unanswered question: acceleration or initiation?
The emerging literature contains reports of rapid progression or recurrence of previously indolent cancers following vaccination. The 2026 systematic review identified recurrent themes involving rapid progression, recurrence or reactivation of previously controlled malignancy.
But case reports cannot establish frequency.
They can establish only that an event occurred.
The denominator is missing.
For example:
If 20 patients with cancer receive vaccination and two experience progression, that observation is potentially interesting.
But without knowing:
- how many similar patients were vaccinated,
- how many progressed without vaccination,
- expected progression rates,
- tumor stage,
- treatment,
- age,
- cancer subtype,
the observation cannot estimate attributable risk.
Case reports are excellent at discovering what might happen.
They are poor at determining how often it happens.
Figure 3 | What the existing evidence can establish
| Evidence type | What it can show | What it cannot establish |
|---|---|---|
| Case report | Temporal signal | Population risk |
| Case series | Recurrent pattern | Incidence |
| Ecological trend | Population change | Individual causation |
| Retrospective cohort | Association | Full causal certainty |
| Pharmacovigilance | Safety signal | True incidence without denominator |
| Mechanistic experiment | Biological plausibility | Human population effect |
| Prospective cohort | Temporal association | Randomized causal effect |
| Randomized trial | Causal efficacy/safety under trial conditions | Very rare long-latency effects |
| Linked national registries | Long-term population risk | Perfect control of all confounding |
The vaccine-cancer question requires convergence across several rows—not reliance upon one.
A proposed global vaccine–cancer surveillance experiment
The scientific community now has the data architecture required to answer this question.
The optimal design would link:
Vaccination records
- date;
- manufacturer;
- platform;
- dose number;
- interval;
- homologous versus heterologous vaccination.
Infection records
- confirmed infection;
- variant;
- severity;
- hospitalization;
- antiviral treatment;
- reinfection.
Cancer registries
- incident cancer;
- histology;
- stage;
- grade;
- recurrence;
- progression;
- metastasis;
- death.
Clinical data
- smoking;
- obesity;
- diabetes;
- immunosuppression;
- medications;
- screening history;
- socioeconomic status.
Molecular data
- tumor sequencing;
- immune-cell profiling;
- cytokines;
- autoantibodies;
- viral signatures;
- tumor microenvironment.
This could be accomplished through distributed analyses using national databases without transferring identifiable patient information.
The ideal analytic model
The central statistical model should not be:
vaccinated vs unvaccinated.
It should be:
time-varying exposure → cancer-specific hazard → competing risks → latency-specific analysis.
A modern causal framework should include:
- propensity scores;
- inverse probability weighting;
- time-varying vaccination exposure;
- competing-risk models;
- negative-control outcomes;
- negative-control exposures;
- quantitative bias analysis;
- E-values;
- dose-response analysis;
- lagged exposure windows;
- sibling or within-family comparisons where possible;
- self-controlled case series;
- case-crossover designs;
- target-trial emulation.
The most convincing analysis would require concordance across several methods.
Figure 4 | The study that could resolve the controversy
NATIONAL POPULATION
│
┌──────────────┴──────────────┐
│ │
VACCINATION DATA INFECTION DATA
│ │
└──────────────┬──────────────┘
▼
IMMUNE HISTORY
│
▼
┌────────────────────┐
│ CANCER REGISTRIES │
└─────────┬──────────┘
│
┌─────────────┼─────────────┐
▼ ▼ ▼
incident recurrence progression
│ │ │
└─────────────┼─────────────┘
▼
CAUSAL INFERENCE
│
┌───────────────┼────────────────┐
▼ ▼ ▼
replication dose-response latency
│ │ │
└───────────────┼────────────────┘
▼
CAUSAL ASSESSMENT
What should be measured immediately?
Five outcomes should receive priority.
1. Hematologic malignancies
Particularly:
- non-Hodgkin lymphoma;
- mature T/NK-cell lymphoma;
- leukemia;
- myeloma.
These cancers are biologically connected to immune regulation and therefore provide a particularly informative test of the hypothesis.
2. Thyroid cancer
The South Korean signal was substantial.
But thyroid cancer is also highly sensitive to diagnostic intensity.
It therefore provides a useful test of whether the association persists after controlling for imaging and healthcare utilization.
3. Breast cancer
The Korean hazard ratio was modest but statistically significant.
Breast cancer has the advantage of extensive screening data, permitting researchers to examine whether increased diagnoses reflect increased detection.
4. Lung cancer
The Korean HR of approximately 1.53 is particularly noteworthy.
But smoking history is indispensable.
A vaccine-cancer study without detailed smoking adjustment cannot credibly interpret lung-cancer associations.
5. Prostate cancer
The Korean estimate was the largest among the six reported cancer sites.
But prostate-specific antigen testing and diagnostic intensity must be incorporated into any replication.
A negative-control strategy could be decisive
Suppose vaccination genuinely causes cancer.
Then cancer associations should display biological specificity.
If vaccination is instead associated with increased healthcare contact, diagnosis or screening, the apparent effect should extend across many conditions detected through intensified medical surveillance.
Researchers should therefore examine outcomes that should not plausibly be caused by vaccination but are similarly affected by healthcare utilization.
Negative controls could include:
- fractures;
- cataracts;
- benign dermatologic conditions;
- osteoarthritis;
- screening-detected lesions.
If the vaccine-cancer association disappears after accounting for a generalized increase in healthcare contact, the causal hypothesis weakens.
If it persists specifically for biologically plausible malignancies, confidence increases.
The role of SARS-CoV-2 infection cannot be ignored
Any analysis that compares vaccinated and unvaccinated populations without adequately accounting for infection history risks attributing infection-related cancer effects to vaccination.
SARS-CoV-2 itself has plausible pathways through which infection could affect long-term biology, including chronic inflammation, immune dysregulation and tissue injury.
The Korean study, for example, was fundamentally a vaccination-status analysis rather than a definitive comparison of vaccination independent of infection history.
The Italian study adjusted for previous SARS-CoV-2 infection, but the reversal of the cancer association after a 12-month lag demonstrates how sensitive the results are to analytic choices.
Future studies must therefore distinguish:
vaccination without infection
from
vaccination followed by infection
from
infection without vaccination.
Only then can the independent contribution of each exposure be estimated.
The possibility of a small effect remains scientifically important
Even if COVID-19 vaccination does not cause a large cancer epidemic, a smaller effect could still matter.
A relative increase of 10–20% in a common cancer could produce substantial numbers of additional cases.
Conversely, a large relative association in a rare cancer could produce very few additional cases.
This is why both relative risk and absolute risk must be reported.
The Korean prostate cancer HR of 1.69 sounds dramatic.
But without absolute incidence rates, the clinical magnitude is difficult to interpret.
A Nature-level analysis should therefore report:
absolute excess cases per 100,000 vaccinated persons
alongside:
hazard ratios and confidence intervals.
The evidence is presently asymmetric
The current literature contains two very different classes of evidence.
Evidence suggesting a signal
- Korean population cohort;
- Italian population cohort;
- U.S. military lymphoma surveillance;
- case reports and case series;
- mechanistic hypotheses;
- reports of recurrence/progression.
Evidence arguing against a large generalized lethal effect
- very large national mortality cohorts;
- absence of increased sudden cardiac death;
- lower all-cause mortality in vaccinated populations;
- regulatory surveillance;
- absence of established vaccine-associated carcinogenesis;
- continued overall long-term decline in cancer mortality.
The scientifically defensible interpretation is therefore neither:
“There is no signal.”
nor:
“The vaccines have caused the cancer epidemic.”
It is:
There are reproducible observations that justify targeted investigation, but the available evidence remains insufficient to establish a causal population-wide cancer effect.
Figure 5 | The current evidence map
COVID-19 VACCINATION
│
┌────────────────┼────────────────┐
│ │ │
▼ ▼ ▼
Cancer case Population Mortality
reports cohorts data
│ │ │
▼ ▼ ▼
SIGNAL ASSOCIATION NO LARGE
│ │ MORTALITY
│ │ SIGNAL
└────────────────┼────────────────┘
│
▼
SCIENTIFIC UNCERTAINTY
│
┌─────────┴─────────┐
▼ ▼
REPLICATION MECHANISM
│ │
└─────────┬─────────┘
▼
CAUSAL INFERENCE
The present evidence occupies the middle of the diagram—not the endpoint.
What a definitive answer would look like
A convincing causal association would require five findings.
First: replication
The South Korean pattern should appear in other large national populations.
Second: temporal coherence
The association should remain after appropriate latency periods.
Third: dose response
Higher cumulative exposure should predict greater risk if a causal biological relationship exists.
Fourth: biological specificity
The affected cancers should correspond to measurable biological changes plausibly produced by the exposure.
Fifth: causal robustness
The relationship should survive rigorous adjustment for infection, screening, smoking, age, sex, comorbidity, socioeconomic status and healthcare utilization.
If all five occurred, the scientific community would have to take the vaccine-cancer hypothesis extremely seriously.
Conversely, if these signals disappear under appropriately controlled analyses, the controversy would largely resolve itself.
What should not happen
Three scientific errors should be avoided.
Error one: dismissing every signal
The history of pharmacovigilance demonstrates that rare adverse events can emerge only after millions of people have been exposed.
Myocarditis is an instructive example.
Error two: treating association as causation
A cancer diagnosed after vaccination is not necessarily caused by vaccination.
The Korean and Italian studies demonstrate why observational associations require careful interpretation.
Error three: using aggregate mortality to dismiss organ-specific effects
A vaccine could theoretically have a small effect on a specific cancer while reducing deaths from COVID-19 sufficiently to lower overall mortality.
Therefore:
all-cause mortality cannot substitute for cancer-specific surveillance.
Toward a mature vaccine-safety science
The COVID-19 vaccine debate has become unusually polarized.
That polarization is scientifically counterproductive.
The appropriate position is neither that vaccines are inherently dangerous nor that every conceivable adverse effect has already been excluded.
The appropriate position is that vaccines, like all medical interventions, should be subjected to continuous empirical surveillance.
That principle is especially important for technologies deployed at unprecedented scale.
More than one billion COVID-19 vaccine doses have been administered globally, creating an extraordinary natural experiment in population health.
The resulting datasets are sufficiently large to answer questions that could not have been answered during pre-licensure trials.
The opportunity should be used.
Conclusion
The evidence reviewed here does not justify declaring that COVID-19 vaccines have caused a population-wide cancer epidemic.
Nor does it justify declaring the question closed.
A South Korean cohort of more than 8.4 million people has reported statistically significant associations between vaccination and several cancers. An Italian population cohort of nearly 300,000 people found a modest association with cancer hospitalization that changed materially when a longer latency interval was imposed. U.S. military surveillance has identified temporal changes in selected lymphoma categories. A large 2026 systematic review has catalogued hundreds of individual cancer reports and identified recurring patterns of progression and recurrence.
At the same time, very large mortality studies have not demonstrated the broad increase in mortality that would be expected from a massive vaccine-induced cancer epidemic. Indeed, some of the largest cohorts report lower all-cause mortality among vaccinated populations, while contemporary studies find no increased risk of sudden cardiac death following vaccination.
The resulting picture is not one of certainty.
It is one of an unresolved safety signal.
That distinction should not be interpreted as a dismissal of concern. It is a call for better science.
The central question now should be framed precisely:
Does COVID-19 vaccination alter the incidence, recurrence or progression of particular malignancies after accounting for SARS-CoV-2 infection, cancer screening, healthcare utilization, demographic risk and appropriate biological latency?
That question is answerable.
The necessary databases already exist.
What is missing is their coordinated analysis.
A definitive study should link vaccination registries, infection histories, cancer registries, mortality records and molecular tumor data; apply time-varying causal methods; test multiple latency windows; examine dose-response relationships; and replicate findings across independent national populations.
Until that work is performed, the scientifically responsible conclusion is neither reassurance without investigation nor alarm without proof.
It is simpler:
The signal deserves investigation.
And because the exposure involved hundreds of millions of people, the investigation should be commensurate with the scale of the experiment.
Methods and scope of evidence synthesis
This analysis was conducted as a structured evidence synthesis of publicly accessible material available as of 11 September 2026. Sources included PubMed-indexed and PubMed Central publications, recent systematic reviews, population-based cohort studies, U.S. military surveillance reports identified in the peer-reviewed literature, regulatory documents from the European Medicines Agency, National Cancer Institute information, CDC-linked surveillance information and large mortality cohorts.
The recent systematic review by Kuperwasser and El-Deiry was used as a literature-map because it systematically searched PubMed, Scopus, Web of Science, Google Scholar and React19 and identified 69 eligible publications through 2025.
This article does not claim to have accessed restricted individual-level records from the Korean National Health Insurance Service, the U.S. Defense Medical Surveillance System, U.S. SEER registries or other controlled-access databases. Consequently, the present work is an evidence synthesis and hypothesis analysis rather than a new individual-participant meta-analysis.
That distinction is essential for peer review.
Footnotes and references
1. Phelps JM. Evidence Continues to Emerge Linking COVID-19 Shots to Rising Cancer Risks. The Gateway Pundit, 10 September 2026. The article alleges rising cancer risks following COVID-19 vaccination and cites military and population-level observations. Its claims are treated here as hypotheses to be evaluated rather than as established findings.
2. Kuperwasser C, El-Deiry WS. COVID vaccination and post-infection cancer signals: Evaluating patterns and potential biological mechanisms. Oncotarget. 2026;17:1–29. The review identified 69 publications, including 66 article-level reports involving 333 patients, two population-level studies and a U.S. military analysis. The authors explicitly characterize the population-level findings as early associations rather than causal estimates.
3. Kim HJ, Kim MH, Choi MG, Chun EM. 1-year risks of cancers associated with COVID-19 vaccination: a large population-based cohort study in South Korea. Biomarker Research. 2025;13:114. The study analyzed 8,407,849 individuals from the Korean National Health Insurance database and reported increased one-year hazards for six cancer categories.
4. Acuti Martellucci C, Capodici A, Soldato G, et al. COVID-19 vaccination, all-cause mortality, and hospitalization for cancer: 30-month cohort study in an Italian province. EXCLI Journal. 2025;24:690–707. The study included 296,015 residents and found lower all-cause mortality but a modest association with cancer hospitalization that reversed under a 12-month minimum lag.
5. Kuperwasser and El-Deiry’s review identified an Armed Forces Health Surveillance Division analysis using Defense Medical Surveillance System data from 2017–2023. The analysis described temporal changes in non-Hodgkin lymphoma incidence, including mature T/NK-cell lymphoma categories, but did not attribute these trends to vaccination or infection.
6. Military populations have substantial occupational and environmental exposures that may influence cancer risk, including combustion products, solvents, radiation and burn-pit exposures. These factors complicate direct extrapolation from military incidence trends to vaccine causality.
7. The European Medicines Agency’s public assessment report for Comirnaty states that conventional genotoxicity and carcinogenicity studies were not performed because the mRNA and lipid components were not expected to have genotoxic potential.
8. National Cancer Institute. COVID-19 Vaccines and People with Cancer. NCI states that there is no evidence that COVID-19 vaccines cause cancer, cancer recurrence or disease progression, while recognizing transient vaccine-associated lymph-node enlargement that can affect imaging.
9. Semenzato L, et al. COVID-19 mRNA Vaccination and 4-Year All-Cause Mortality. JAMA Network Open. 2025. In approximately 28 million adults, vaccination was associated with lower four-year all-cause mortality after adjustment.
10. Slurink IAL, et al. COVID-19 vaccination and short-term mortality risk. 2026. The analysis found no short-term mortality excess following vaccination and instead observed lower relative incidence of all-cause death in the early post-vaccination period.
11. Abdel-Qadir H, et al. Association between COVID-19 vaccination and sudden cardiac death among apparently healthy people aged 12–50 years. 2026. The study did not find evidence supporting increased sudden cardiac death following vaccination.
12. Stanford Medicine. Study shows why mRNA COVID-19 vaccine myocarditis occurs. 2025. Myocarditis is recognized as a rare but established adverse effect of mRNA vaccination, demonstrating the importance of post-authorization safety surveillance.
13. Sampri A, et al. Vascular and inflammatory diseases after COVID-19 vaccination. The Lancet Child & Adolescent Health. 2025. The study found a short-term increased risk of myocarditis/pericarditis, particularly within four weeks after vaccination.
14. U.S. cancer-screening studies documented substantial reductions in screening during the first pandemic year, creating an important competing explanation for changes in cancer diagnosis and mortality.
15. North American Association of Central Cancer Registries/NCI/CDC/ACS analysis reported that advanced cancer diagnoses generally returned toward pre-pandemic patterns by 2021, arguing against a simple persistent wave of undiagnosed advanced cancers caused solely by the initial screening disruption.
16. A 2026 JAMA Oncology analysis reported worse short-term survival among patients diagnosed with cancer during 2020–2021 than among those diagnosed before the pandemic, providing evidence that pandemic-era disruptions in cancer care affected outcomes independently of any vaccine effect.
17. The distinction between vaccine-associated myocarditis and broader vaccine safety illustrates the difference between identifying a specific adverse-event signal and demonstrating a broad population-level mortality effect.
Author declaration
John Murphy, M.D., M.P.H., D.P.H.
CEO, The COVID-19 Long-haul Foundation
The author declares that this manuscript is an evidence synthesis and hypothesis-generating analysis based on publicly accessible scientific literature and databases. No restricted individual-level patient data were accessed. The manuscript does not claim that COVID-19 vaccination has been proven to cause cancer. It argues that several independently reported epidemiologic signals merit rigorous, adequately powered investigation.
No clinical recommendation concerning COVID-19 vaccination is made on the basis of this analysis.