The Vaccine Confidence Project

Introduction

Vaccination ranks among humanity's greatest medical achievements. Smallpox has been eradicated, polio pushed to the edge of extinction, and measles, mumps, rubella, tetanus, diphtheria, influenza, HPV-related cancers, COVID-19, and many other diseases prevented, controlled, or mitigated through programs that have saved millions of lives. Yet despite this record, vaccine confidence has grown increasingly fragile.

People decline vaccination for many reasons. Some are misled by misinformation; others act on genuine uncertainty; many simply feel that modern medicine asks them to accept an intervention without fully understanding how it might affect them personally. Science has responded by demonstrating, repeatedly and at scale, that vaccines are overwhelmingly safe and effective. But one question remains comparatively unexplored: can we identify, before vaccination, the small number of individuals who may face an elevated risk of a serious adverse reaction — and, just as importantly, identify those who stand to benefit most?

The answer may lie at the intersection of immunology, genetics, artificial intelligence, systems biology, and precision medicine. This proposal outlines a research initiative built on that possibility — the development of pre-vaccination risk-assessment tools intended to improve both vaccine safety and public confidence.

The Current State of Vaccine Medicine

Modern medicine increasingly recognizes that people respond differently to the same treatment. Cancer therapies are tailored to genetic mutations, many medications are prescribed in light of pharmacogenomic testing, and organ-transplant recipients undergo extensive compatibility screening. Vaccination, by contrast, remains largely universal: a vaccine shown to be safe and effective across large populations is generally given in the same way to nearly everyone.

That strategy has been extraordinarily successful, and it should remain the default. The principle here is deliberately broad, spanning both viral vaccines (COVID-19, influenza, measles, HPV, varicella, polio) and bacterial ones (tetanus, diphtheria, pertussis, pneumococcal and meningococcal disease, Haemophilus influenzae type b). Nothing in the approach depends on whether the pathogen is a virus or a bacterium; what matters is that individual risk and benefit vary. But rare adverse events do occur. Documented examples include vaccine-induced immune thrombotic thrombocytopenia (VITT, also called thrombosis with thrombocytopenia syndrome, or TTS) associated with certain adenovirus-vector COVID-19 vaccines; myocarditis and pericarditis following mRNA COVID-19 vaccination, concentrated in specific groups such as adolescent and young adult males, typically after the second dose; Guillain-Barré syndrome following a small number of vaccines in very small numbers of recipients; and severe allergic reactions, including anaphylaxis, in individuals with particular sensitivities. Some rarer neurological and autoimmune reactions remain incompletely understood.

These events are uncommon, but they are real. The central question is therefore not whether vaccines are safe for most people — they are — but whether science can identify the few individuals for whom a different vaccine, a modified schedule, enhanced monitoring, or additional medical evaluation might be appropriate.

The Cost of Uncertainty

When a serious vaccine reaction occurs, its effects reach far beyond the affected person. Public trust erodes, families grow fearful, and social media amplifies individual stories faster than institutions can respond. Public-health agencies can become defensive and scientists frustrated, and the result is too often polarization rather than understanding.

That polarization carries measurable costs. Declining vaccination rates have contributed to outbreaks of measles and other preventable diseases in several countries. At the same time, people who experienced genuine adverse reactions frequently feel dismissed. Both groups lose confidence, and the public conversation hardens into a forced choice between "vaccines are perfectly safe" and "vaccines are dangerous."

A better approach may be possible. Rather than accept that framing, we might ask a more productive question: can we develop tools that identify individual risk while preserving the enormous population-level benefits of vaccination?

A New Research Direction

The Vaccine Confidence Project proposes a multi-layered, deliberately multi-modal pre-vaccination assessment platform. It would not rely on blood tests alone. Candidate inputs span several kinds of data: genomic testing (including HLA typing, Fc-receptor and other immune-gene variants, and polygenic risk scores where validated); molecular and laboratory markers (blood chemistry, immune and inflammatory profiles, platelet-activation markers); clinical history (comorbidities, including respiratory and pulmonary history; current medications; prior infections; and allergy or autoimmune history); and family history together with, where available, physiologic or functional-immune data. Genetic information deserves particular emphasis, because — as the literature below shows — host genotype shapes both how a person responds to a given vaccine and how severely they would suffer the disease that vaccine prevents. Artificial intelligence and modern statistical models could then be trained on these inputs to detect patterns associated with rare adverse events and, equally, to identify who stands to benefit most from vaccination.

The aim would not be to deny vaccination but to optimize it. For any given person, such a system might recommend proceeding normally, proceeding with enhanced monitoring, selecting an alternative vaccine, delaying vaccination pending further evaluation, or seeking specialist consultation. The overwhelming majority of people would almost certainly be classed as low risk — and that reassurance, delivered individually, could itself improve confidence and uptake.

Both Sides of the Ledger: Who Should Be Vaccinated

A tool that only flags who might be harmed by a vaccine is half a tool — and the more dangerous half to build in isolation, because on its own it can be misread as a search for reasons to refuse. The same precision logic carries a corollary that is at least as important: identifying who is most endangered by the disease itself, and for whom the benefit of vaccination is therefore greatest. Precision vaccination is properly understood not as risk-minimization but as net-benefit maximization — weighing the usually tiny risk of an adverse event against the often substantial risk of the disease, for each individual.

COVID-19 made this concrete. Its burden fell disproportionately on older people and those with pre-existing conditions. The OpenSAFELY study, which examined the records of more than 17 million adults in England, quantified sharply higher risks of COVID-19 death with older age, male sex, diabetes, severe asthma, obesity, and a range of cardiac, kidney, and other conditions, with over 90% of UK deaths occurring in people over 60. In that sense the pandemic was as much a comorbidity crisis as a viral one. The implication for a precision platform is not that such people should be screened away from vaccination but the opposite: they are precisely the people for whom the benefit was greatest, and for whom the recommendation should read "yes — and first."

Age is the spine of this calculus. The risk-benefit balance genuinely inverts across the lifespan: very high benefit at the upper end, where the disease is most lethal; a different and more finely balanced calculus in healthy younger adults, where disease risk is lower (though never zero); and a distinct calculus again in infants and children, whose risk from many diseases — and from the vaccines that prevent them — differs from adults'. A precision approach makes these differences explicit instead of papering over them with a single universal recommendation.

Respiratory history deserves particular attention on this benefit side, for a simple reason: many of the proposal's priority vaccines — against COVID-19, influenza, RSV, pneumococcal disease, and pertussis — target pathogens of the lung, so a person's pulmonary reserve directly shapes how dangerous an infection will be for them. Conditions that erode that reserve raise the benefit of vaccination accordingly: chronic obstructive pulmonary disease, severe asthma (identified as a risk factor in the OpenSAFELY analysis), cystic fibrosis, and interstitial lung disease or pulmonary fibrosis; thoracic malignancies such as lung cancer and pleural mesothelioma, which both reduce lung function and are often treated with immunosuppressive therapy; and a history of severe or recurrent lower-respiratory infection such as pneumonia — a softer signal, but one that can flag diminished reserve or an underlying susceptibility worth weighing. For someone in any of these groups, a respiratory infection is far more likely to progress to hospitalisation or death, so the platform should capture respiratory and pulmonary history as a distinct input rather than folding it into a generic comorbidity count.

This is not speculative. Unlike adverse-event prediction, disease-risk stratification is already operational. The QCOVID algorithm, derived and validated on millions of NHS records, estimated an individual's risk of COVID-19 hospitalization and death from routine clinical and demographic data; in practice its scores added roughly 1.5 million people to England's shielding list and prioritized around 800,000 of them for vaccination. A precision-vaccination platform can therefore stand on a proven half — established benefit-side risk models — while it researches the hard half, the adverse-event side. Host genetics tightens the link between the two: the GenOMICC study identified genetic variants that predispose to critical COVID-19, so genetic testing can inform both sides of the ledger at once — susceptibility to severe disease and, aspirationally, susceptibility to rare adverse events.

Framed this way, the platform is far more often an engine for saying "yes, especially you" than "no." That is the honest center of gravity of precision vaccination, and it should be stated plainly to funders and the public alike: the goal is to get the right vaccine to the right person at the right time — including, and especially, the person whose comorbidities or genotype make the disease most dangerous to them.

Natural, Vaccine-Induced, and Hybrid Immunity

A precision platform must also reckon honestly with how protection is acquired, because that bears directly on an individual's marginal benefit from another dose. Immunity comes in three broad forms, and the evidence distinguishes them clearly. Infection-acquired ("natural") immunity is real and can be robust, but it is purchased at the price of having the disease — with all the risk of severe illness, hospitalization, lasting effects, and death that entails. Vaccine-induced immunity is strong and far safer, though protection against infection can wane and new variants can erode it. Hybrid immunity — the combination of infection and vaccination, in either order — is, on current evidence, the strongest and most durable of the three.

The most authoritative synthesis is the systematic review and meta-regression by Bobrovitz and colleagues (Lancet Infectious Diseases, 2023), which found that hybrid immunity offered the highest and most durable protection — on the order of 97% against hospitalization or severe disease at twelve months — while protection against reinfection waned more quickly. Two points from that work matter enormously for this proposal, and the document states both plainly. First, the order does not change the headline: infection-then-vaccination and vaccination-then-infection both confer strong protection. Second, the same authors are emphatic that this is not an argument for seeking infection — the risks of acquiring immunity the "natural" way are real, and the finding reinforces rather than undercuts the case for vaccinating people who have already been infected. A precision platform can legitimately weigh documented prior infection when estimating an individual's marginal benefit from a further dose; it cannot treat deliberate infection as a strategy.

What a Blood Test Can and Cannot Reveal

A recurring and reasonable question is whether a blood test can simply read off a person's history — how many times they were vaccinated, how many times infected. The honest answer is partly yes, mostly no, and the distinction is worth making precisely because it is so often misunderstood.

Serology can broadly separate infection from vaccination by reading the pattern of two antibodies. Because most vaccines encode only the spike protein, anti-spike (anti-S) antibodies arise after either infection or vaccination, whereas anti-nucleocapsid (anti-N) antibodies arise essentially only after natural infection. A positive anti-N therefore points to prior infection, and the combination of the two markers can distinguish someone vaccinated-only from someone infected from someone with hybrid immunity. The major caveat is that anti-N wanes — often falling below detection within six to nine months — so it underestimates more distant infection and is an unreliable marker of whether someone was ever infected. What serology cannot do is count: a single sample cannot reliably tell you how many doses a person received or how many separate infections they had, because antibody levels are shaped by age, disease severity, timing, comorbidities, and the assay itself. Nor does it measure cellular (T-cell) immunity or guarantee protection. The useful framing is that serology is a powerful but partial tool: it reads broad exposure history, not a precise ledger of events.

Initial Research Targets

A phased strategy should begin with adverse events that are clinically significant, biologically plausible, measurable, and supported by existing data.

The first phase would focus on conditions with known or partially understood biological pathways, where existing laboratory markers could serve as starting points: VITT/TTS associated with adenovirus-vector COVID-19 vaccines, myocarditis associated with mRNA COVID-19 vaccines, Guillain-Barré syndrome associated with selected vaccines, and severe allergic reactions and anaphylaxis.

A second phase would extend the work to vaccines where public confidence has itself become a significant obstacle — among them MMR, HPV, influenza, polio, and varicella — and, on the bacterial side, to long-established vaccines such as those against pertussis and pneumococcal disease. Here the goal would not be to validate unsupported claims but to test, transparently, whether individualized screening can improve trust and understanding.

A third phase would aim to generalize these findings into a precision-vaccination framework adaptable to future vaccine technologies and to both viral and bacterial targets.

Why This Matters: The Economics and the Human Stakes

The coming decades of medicine will almost certainly be personalized. We no longer expect every cancer patient to receive identical treatment, nor every drug to act identically in every body. Vaccination may eventually follow the same path — not because vaccines have failed, but because they have succeeded so thoroughly that the next question is how to make them safer still.

A successful pre-vaccination assessment system could reduce rare adverse events, improve public confidence, increase uptake, lower long-term healthcare costs, speed acceptance of future vaccines, and deepen scientific understanding of human immune variability. And by weighing benefit as carefully as risk, it would more often identify the people who most need vaccination than those who should avoid it. Most importantly, it could turn a polarized debate into a shared scientific effort.

Literature Review

From population medicine to precision medicine

The intellectual ground for this proposal was laid when precision medicine moved from aspiration to national policy. Collins and Varmus set out the case in 2015 for matching prevention and treatment to individual variability rather than to population averages, and the All of Us Research Program then demonstrated that a cohort of roughly a million participants, contributing electronic health records, biospecimens, and survey data under a single governance structure, can be assembled and sustained. Two lessons from that program matter here: first, that very large, consented, richly phenotyped cohorts are operationally achievable; and second, that their scientific value depends on deliberate recruitment of populations historically underrepresented in genomics, without which any resulting risk model will perform unevenly across groups.

Vaccinomics and adversomics

The specific application of these ideas to vaccines has an established literature and its own vocabulary. Gregory Poland and colleagues at the Mayo Clinic Vaccine Research Group developed the concept of vaccinomics — the study of how genetic and other individual factors shape the heterogeneity of vaccine-induced immune responses — and, in parallel, adversomics, the study of the individual determinants of vaccine side effects. Their 2013 synthesis is the conceptual anchor for the present proposal. Crucially, the literature is candid about a gap the proposal must confront: adversomics is far less developed than vaccinomics. Omersel and Karas Kuželički's 2020 review surveyed what is known about the genes — particularly those encoding human leukocyte antigens, cytokines, and their receptors — associated with the efficacy and safety of hepatitis B, MMR, and HPV vaccines, and concluded that precision approaches to vaccination, while proposed, remain far from clinical implementation. The field supplies a credible rationale and candidate biomarkers, not yet a validated predictive test.

The mechanistic basis of the Phase One targets — and an honest limitation

The Phase One adverse events were chosen because they have the best-characterized biology, and the 2021 work on VITT/TTS illustrates both the promise and the central challenge. Greinacher, Schultz, Scully, and their co-authors independently established that VITT is mediated by antibodies against platelet factor 4 (PF4), a syndrome resembling heparin-induced thrombocytopenia — exactly the kind of measurable, mechanistically grounded event the proposal targets. The limitation must be stated plainly: those anti-PF4 antibodies develop after vaccination and are largely transient, so they are a diagnostic marker, not a pre-vaccination predictive one. The same caution applies to mRNA-associated myocarditis and to Guillain-Barré syndrome — at present there is no validated germline or baseline biomarker that identifies, before injection, who will experience these events. Phase One is therefore a discovery effort to determine whether pre-existing host factors can stratify risk, not a deployment of tests already known to work.

Why the approach is nonetheless credible: the abacavir precedent

The proposal's plausibility rests on a real, decisive precedent. Abacavir causes a hypersensitivity reaction in a few percent of patients, strongly linked to a single germline allele, HLA-B*57:01. The PREDICT-1 trial — a double-blind, prospective, randomized study of 1,956 patients across 19 countries — showed that screening for the allele before prescribing eliminated immunologically confirmed hypersensitivity (0% screened versus 2.7% control) and roughly halved clinically diagnosed cases, with a negative predictive value near 100%; cost-effectiveness analysis supported routine use, and screening entered standard care in 2008. PREDICT-1 establishes that a host genetic marker can predict a serious immune-mediated reaction, that a prospective trial can prove a screening strategy prevents harm, and that such a strategy can be adoptable. It also sets the bar: the abacavir predictor is a germline allele present before exposure, with near-perfect negative predictive value. The open question for vaccines is whether comparable pre-exposure predictors exist — until answered, abacavir is the model of success, not proof the success is achieved.

The other half of the ledger: disease-risk stratification is already real

If the adverse-event side is a research frontier, the benefit side is mature. The OpenSAFELY study (Williamson and colleagues, 2020) analysed more than 17 million adult NHS records and quantified how age, sex, and comorbidities such as diabetes, obesity, and chronic respiratory, cardiac, and kidney disease raise the risk of COVID-19 death — confirming that mortality was concentrated in identifiable high-risk groups. The QCOVID algorithm (Clift and colleagues, 2020) turned that evidence into a validated, deployed tool that England used to extend shielding and prioritise vaccination. Host-genetic studies — the GenOMICC genome-wide association study (Pairo-Castineira and colleagues, 2021) and the international COVID-19 Host Genetics Initiative — have begun to map variants predisposing to critical illness. And the immunity and serology literatures (Bobrovitz and colleagues, 2023, on hybrid immunity; the anti-N/anti-S distinction in serological surveillance) show how exposure history can be measured and weighed. Together these establish that individualised disease-risk prediction is feasible, validated, and already in use — which is precisely why a precision-vaccination platform should incorporate the benefit side from the outset rather than treating safety screening as the whole task.

Governance

The project should be governed through a structure that separates scientific direction, financial stewardship, and independent oversight, because the credibility of a vaccine-safety initiative depends on its being visibly free of conflicts of interest. A proposed structure has four layers: a Steering Committee responsible for strategy, milestones, and budget; a Scientific Advisory Board of immunologists, geneticists, biostatisticians, vaccinologists, and public-health practitioners that reviews the science and approves each phase; an independent Ethics and Data Oversight Board — including bioethicists, a patient or community representative, a privacy expert, and at least one member with no institutional tie to the project — that reviews consent materials, data-access decisions, and any return-of-results policy, and can pause the work; and a standing Data Access Committee that adjudicates every request to use the cohort's data against pre-published criteria.

Three commitments are worth stating explicitly in any funding document. First, all financial relationships with vaccine manufacturers, diagnostics companies, and other commercial parties are declared and managed under a written conflict-of-interest policy, and the project does not accept funding structured to favor a predetermined finding. Second, the project works with national immunization authorities and pharmacovigilance systems rather than around them. Third, governance documents, protocols, and aggregate results are made public, so that a sceptical reader can audit the project rather than being asked to trust it.

Ethics

The ethical framework follows the long-established principles codified in the Declaration of Helsinki and the Belmont Report — respect for persons, beneficence, and justice — and adds two considerations specific to this project.

The first is the duty not to harm vaccine confidence itself. A study of rare vaccine reactions can, if communicated carelessly, be weaponized to discourage vaccination. The project's framing must be consistent at every level: the goal is to optimize and protect vaccination by reassuring the low-risk majority and appropriately managing the rare high-risk individual, never to supply ammunition for blanket refusal — and the project commits in advance to publishing null and reassuring findings as prominently as positive ones. The second is equity. Because HLA and other immunogenetic markers vary in frequency across ancestral populations, a model trained on a narrow population will misclassify others, and a screening test available only to some would widen disparities. Justice here means diverse recruitment, validation across populations before use, a plan for equitable access, non-discrimination protections (in the United States, the Genetic Information Nondiscrimination Act), a clear policy on returning individual findings, and genuine equipoise.

Informed Consent

Consent for a long-lived biobank-and-data project cannot be a single signature at enrolment, because future uses of samples and data cannot all be specified in advance. The literature offers two workable models, and the project should combine them. Broad consent, used by large biobanks, asks participants to agree to a defined framework of future research subject to independent ethical review of each new study. Dynamic consent, developed by Kaye and colleagues, adds a digital interface through which participants stay informed and can grant, refine, or withdraw permissions over time. Pairing the two gives participants a workable enrolment process and continuing control.

Whatever the mechanism, the process must meet substantive standards, not merely documentary ones: genuine plain-language comprehension of what genomic and immunological data reveal; withdrawal that is real and easy, while honest about what cannot be undone once data are aggregated; explicit treatment of secondary use, commercial involvement, and data sharing; a stated policy on whether clinically actionable findings are returned and through what clinical pathway; and acknowledgment that genetic findings can carry implications for biological relatives who never consented. Consent for any minors would require separate, stricter safeguards and is not contemplated in the early phases.

Data Privacy

The project's central privacy fact, which should be stated to funders rather than buried, is that genomic data cannot be reliably anonymized. Gymrek and colleagues demonstrated that individuals could be re-identified from supposedly de-identified genetic data by combining Y-chromosome short-tandem-repeat profiles with public genealogy databases. Removing names and dates of birth, the standard sufficient for ordinary health records under HIPAA, does not make a genome anonymous. Any honest data-protection plan therefore relies on governed access rather than a false promise of anonymity.

The protection model should combine several layers: storage of identifiable data separately from research data under coded linkage; a controlled-access regime in which external researchers obtain data only through the Data Access Committee under binding agreements; preference for federated, "bring-the-analysis-to-the-data" approaches over distributing raw genomes; encryption in transit and at rest; audit logging; and a defined breach-response plan. Compliance must be explicit with the relevant regimes — HIPAA and GINA in the United States, the GDPR in the European Union — and the project should commit to the strictest standard across the jurisdictions in which it operates. Participants should be told candidly both what protections exist and what residual re-identification risk cannot be eliminated.

Study Design

The scientific design follows the phased logic above and inherits its central caveat: for most target events, no validated pre-vaccination predictor yet exists, so the early work is discovery and the later work is validation. The two must not be conflated.

Discovery phase. The most efficient and ethical starting point is to mine adverse events that have already occurred rather than expose new participants to risk: nested case-control studies drawn from large vaccinated populations, using linkage to established pharmacovigilance and vaccine-safety surveillance to identify confirmed cases of VITT, myocarditis, GBS, and anaphylaxis, matched to vaccinated controls. Cases and controls are then compared across HLA type, Fc-receptor and other immune-gene polymorphisms, baseline inflammatory and platelet biology, prior infection and allergy history, and, where samples allow, multi-omic profiles. Because these events are rare — VITT, for instance, on the order of single-digit to low-double-digit cases per million first doses — adequate power requires very large source cohorts or international case-pooling, and the power calculation should be presented honestly up front.

Model development. Candidate predictors are combined using statistical and machine-learning models, with three non-negotiable safeguards: strict separation of training and test data, external validation in an independent cohort, and explicit assessment of calibration and of performance across ancestral groups. The deliverable is not an algorithm but a quantified claim — the positive and negative predictive values a screening test would achieve in a defined population, with the abacavir standard of a very high negative predictive value.

Prospective validation. Only a marker that survives the previous phases justifies a prospective trial, with the PREDICT-1 design as the template: a prospective, ideally randomized comparison of screening-guided vaccination against standard practice, powered on a clinical endpoint and paired with a health-economic analysis.

The benefit side. Running in parallel, and drawing on the same data platform, the design should incorporate disease-risk stratification — who is most endangered by the target disease and therefore benefits most. This half is methodologically easier and already proven: validated tools such as QCOVID show that routine clinical and demographic data, augmented by host genetics, can guide real prioritisation decisions. Building it in from the start keeps the platform honest about its purpose — maximising net benefit — and ensures that for most people, and especially those with high-risk comorbidities, the output is a clear recommendation to vaccinate.

Epidemiology and real-time data capture. A central lesson of the COVID-19 pandemic is that the value of outbreak data depends on capturing it systematically while a disease is still fresh, and that this opportunity was only partly seized — data were fragmented across siloed health systems, slowed by privacy and consent constraints, by the sheer urgency of the moment, and in places by politics. A precision-vaccination platform is, among other things, a standing epidemiological instrument: a consented, linked, continuously updated cohort capable of capturing exposure, outcome, and host-factor data in real time. Designing for that from the outset — interoperable standards, pre-negotiated data-use frameworks, and surveillance linkage — is how the project turns the next emerging pathogen from a missed opportunity into a rapidly characterised one.

Throughout, the design should pre-register hypotheses and analysis plans, define the clinical action attached to each risk category before any participant is screened, and treat a reassuring or null result — finding that no useful pre-vaccination predictor exists for a given event — as a legitimate and publishable scientific outcome.

Objections, Autonomy, and the Limits of Science

A credible proposal must engage fairly with the people who decline vaccination, and doing so honestly requires separating two very different kinds of objection. After a careful review of the scientific literature, the empirical case against vaccination does not hold up: there is no credible body of peer-reviewed, double-blind evidence showing vaccines to be net-harmful, the foundational anti-vaccine paper (Wakefield's 1998 MMR–autism study) was retracted as fraudulent, and the highest-quality double-blind, placebo-controlled trials — the large COVID-19 vaccine trials among them — support vaccination. Once the empirical claims are set aside, what remains of a principled refusal is not a scientific dispute but a values claim. It is most fairly described not as "anti-vaccine" but as health naturalism, or a preference for natural immunity: the view, resting on what philosophers call the appeal to nature, that what is natural is inherently safer or better than what is manufactured.

That worldview comes in two distinct versions, and they deserve different responses. The individual version holds that natural infection confers more robust immunity than a vaccine. For some diseases this contains a kernel of truth about immune breadth, but for others it is simply dangerous — deliberate "measles parties," for instance, are condemned by every pediatric body because measles can cause pneumonia, encephalitis, and the delayed and fatal complication SSPE; the practice is historically continuous with pre-Jenner variolation, and the immunity literature is clear that hybrid immunity, where it arises, is best obtained without seeking infection. The population or evolutionary version holds that herd immunity is better "earned" through natural infection, or that allowing disease to run its course strengthens the species. This is where the gravest ethical objection lies, and the proposal presents it in the voice of its critics rather than branding adherents: disability advocates and bioethicists argued throughout the pandemic that a "let it spread" or mass-infection logic quietly treats the deaths of the elderly and disabled as an acceptable price — a social-Darwinist premise that most find unacceptable once it is made explicit. The free-rider variant of the argument — that the healthiest course is to remain unvaccinated while everyone else is immune — is individually rational and collectively self-defeating: it works only so long as almost no one else reasons the same way.

These objections also drive the most contested policy question of the pandemic era: mandates. Here the proposal takes no partisan side but frames the genuine tension — individual autonomy and proportionality against collective protection and the load on health systems — and offers precision vaccination as a middle path. Rather than compel everyone uniformly or abandon policy altogether, a risk-benefit approach can concentrate effort on those who clearly benefit, which is both more effective and less coercive. The relevant legal backdrop runs from the U.S. Supreme Court's 1905 decision in Jacobson v. Massachusetts, which upheld compulsory vaccination on the principle that individual liberty does not extend to exposing others to serious harm, to the present patchwork of exemptions and the bodies, such as the American Medical Association, that oppose broad ones.

Because so many objections are filed as religious, the proposal addresses that framing directly and accurately. No major world religion doctrinally opposes vaccination; Catholic, Jewish, Islamic, and Buddhist authorities have all supported it, and as Pelčić and colleagues document, most invocations of "religious exemption" are better understood as personal, philosophical, or political beliefs expressed in religious language. American law largely sidesteps doctrine and turns instead on the sincerity of the individual's belief, which is why courts are repeatedly drawn into the genuinely hard problem of deciding what counts as religious at all. The point for this proposal is clean: precision vaccination resolves the empirical question with evidence — who is genuinely at risk, who genuinely benefits — and by doing so it isolates the genuinely values-based residue, which is properly handled through the rights-and-exemption framework rather than litigated as though it were a scientific disagreement. That is how one respects a belief-based objection without pretending the science is in doubt.

Tailoring the Material to Each Audience

The same evidence base supports four different documents, and the framing should shift with the reader. For foundations, lead with vision, equity, and public trust, foregrounding the diverse-recruitment and equitable-access commitments that mission-driven funders weigh heavily. For nonprofits and patient-advocacy partners, foreground the people on both sides of the current divide and the community-advisory and dynamic-consent structures that give participants a real voice. For grant proposals, convert the formal sections into specific aims with testable hypotheses, the nested case-control and validation designs with explicit power calculations and milestones, and the abacavir/PREDICT-1 precedent as the feasibility argument. For public-awareness campaigns, translate everything into plain language and hold to the project's framing: precision screening exists to support vaccination by protecting the few and reassuring the many — never implying that screening is a reason to delay a vaccine recommended now — and campaign material should be reviewed by the ethics board precisely because well-intentioned safety messaging is the easiest thing for bad-faith actors to misquote.

Conclusion

Science advances when it refuses false choices. We need not choose between vaccine safety and vaccine confidence, between protecting public health and listening to individuals, or between innovation and caution. The future may instead lie in precision vaccination: understanding not only whether a vaccine works across populations, but how it interacts with each individual — and recognizing that for most people, and especially for the most vulnerable, that understanding points toward vaccination, not away from it.

The Vaccine Confidence Project is an invitation to explore that future — not because vaccines have failed, but because they have worked well enough that we can now ask how to make them better.

References

All entries were located and checked against the published literature. Items flagged "confirm" need a final detail verified before formal submission; policy and legal citations should be styled to the target funder's or journal's house format.

1. Collins FS, Varmus H. A new initiative on precision medicine. New England Journal of Medicine. 2015;372(9):793–795. doi:10.1056/NEJMp1500523.
2. All of Us Research Program Investigators. The "All of Us" Research Program. New England Journal of Medicine. 2019;381(7):668–676. doi:10.1056/NEJMsr1809937.
3. Poland GA, Ovsyannikova IG, Kennedy RB, Haralambieva IH, Jacobson RM. Vaccinomics, adversomics, and the immune response network theory: individualized vaccinology in the 21st century. Seminars in Immunology. 2013;25(2):89–103. PMID:23755893. (Confirm pagination.)
4. Omersel J, Karas Kuželički N. Vaccinomics and adversomics in the era of precision medicine: a review based on HBV, MMR, HPV, and COVID-19 vaccines. Journal of Clinical Medicine. 2020;9(11):3561. doi:10.3390/jcm9113561.
5. Mallal S, Phillips E, Carosi G, et al; PREDICT-1 Study Team. HLA-B*5701 screening for hypersensitivity to abacavir. New England Journal of Medicine. 2008;358(6):568–579. doi:10.1056/NEJMoa0706135.
6. Hughes S, Hughes A, Brothers C, Spreen W, Thorborn D. PREDICT-1 (CNA106030): the first powered, prospective trial of pharmacogenetic screening to reduce drug adverse events. Pharmaceutical Statistics. 2008;7(2):121–129. PMID:17534855.
7. Hughes DA, Vilar FJ, Ward CC, Alfirevic A, Park BK, Pirmohamed M. Cost-effectiveness analysis of HLA B*5701 genotyping in preventing abacavir hypersensitivity. Pharmacogenetics. 2004;14(6):335–342. PMID:15247625.
8. Greinacher A, Thiele T, Warkentin TE, Weisser K, Kyrle PA, Eichinger S. Thrombotic thrombocytopenia after ChAdOx1 nCov-19 vaccination. New England Journal of Medicine. 2021;384(22):2092–2101.
9. Schultz NH, Sørvoll IH, Michelsen AE, et al. Thrombosis and thrombocytopenia after ChAdOx1 nCoV-19 vaccination. New England Journal of Medicine. 2021;384(22):2124–2130.
10. Scully M, Singh D, Lown R, et al. Pathologic antibodies to platelet factor 4 after ChAdOx1 nCoV-19 vaccination. New England Journal of Medicine. 2021;384(23):2202–2211.
11. Kaye J, Whitley EA, Lund D, Morrison M, Teare H, Melham K. Dynamic consent: a patient interface for twenty-first century research networks. European Journal of Human Genetics. 2015;23(2):141–146. doi:10.1038/ejhg.2014.71.
12. Budin-Ljøsne I, Teare HJA, Kaye J, et al. Dynamic consent: a potential solution to some of the challenges of modern biomedical research. BMC Medical Ethics. 2017;18:4. doi:10.1186/s12910-016-0162-9.
13. World Medical Association. WMA Declaration of Helsinki: ethical principles for medical research involving human subjects (2013 revision). JAMA. 2013;310(20):2191–2194.
14. Gymrek M, McGuire AL, Golan D, Halperin E, Erlich Y. Identifying personal genomes by surname inference. Science. 2013;339(6117):321–324. doi:10.1126/science.1229566.
15. Regulation (EU) 2016/679 of the European Parliament and of the Council (General Data Protection Regulation). Official Journal of the European Union, 2016.
16. Health Insurance Portability and Accountability Act of 1996, Pub. L. No. 104-191; and the U.S. Genetic Information Nondiscrimination Act of 2008 (GINA), Pub. L. No. 110-233.
17. Williamson EJ, Walker AJ, Bhaskaran K, et al. (OpenSAFELY Collaborative). Factors associated with COVID-19-related death using OpenSAFELY. Nature. 2020;584(7821):430–436. doi:10.1038/s41586-020-2521-4.
18. Clift AK, Coupland CAC, Keogh RH, et al. Living risk prediction algorithm (QCOVID) for risk of hospital admission and mortality from coronavirus 19 in adults: national derivation and validation cohort study. BMJ. 2020;371:m3731. doi:10.1136/bmj.m3731.
19. Pairo-Castineira E, Clohisey S, Klaric L, et al. Genetic mechanisms of critical illness in COVID-19. Nature. 2021;591(7848):92–98. doi:10.1038/s41586-020-03065-y.
20. COVID-19 Host Genetics Initiative. Mapping the human genetic architecture of COVID-19. Nature. 2021;600(7889):472–477. (DOI to confirm: 10.1038/s41586-021-03767-x.)
21. Bobrovitz N, Ware H, Ma X, et al. Protective effectiveness of previous SARS-CoV-2 infection and hybrid immunity against the omicron variant and severe disease: a systematic review and meta-regression. Lancet Infectious Diseases. 2023;23(5):556–567. doi:10.1016/S1473-3099(22)00801-5.
22. Pelčić G, Karačić S, Mikirtichan GL, Kubar OI, Leavitt FJ, Cheng-Tek Tai M, Morishita N, Vuletić S, Tomašević L. Religious exception for vaccination or religious excuses for avoiding vaccination. Croatian Medical Journal. 2016;57(5):516–521. doi:10.3325/cmj.2016.57.516.
23. Jacobson v. Massachusetts, 197 U.S. 11 (1905).
24. Distinguishing SARS-CoV-2 infection and vaccine responses up to 18 months post-infection using nucleocapsid protein and receptor-binding domain antibodies. Microbiology Spectrum. 2023. doi:10.1128/spectrum.01796-23. (Confirm author list.)

Appendix A — A Fair Reading of the Skeptic Literature

There is no peer-reviewed, double-blind body of evidence supporting an anti-vaccine conclusion, so this is not a bibliography of such studies — assembling one would misrepresent the record. It is instead the credible literature that gives careful skepticism its legitimate footing, together with the discredited landmark cited explicitly as discredited so a reader sees the full terrain. Exact details should be confirmed before submission.

• Wakefield AJ, et al. Ileal-lymphoid-nodular hyperplasia, non-specific colitis, and pervasive developmental disorder in children. Lancet. 1998;351:637–641. RETRACTED (Lancet. 2010;375:445).
• Deer B. How the case against the MMR vaccine was fixed. BMJ. 2011;342:c5347.
• Institute of Medicine. Adverse Effects of Vaccines: Evidence and Causality. Washington, DC: National Academies Press; 2011. (A balanced review cataloguing which vaccine adverse events are causally established and which are not.)
• MacDonald NE; SAGE Working Group on Vaccine Hesitancy. Vaccine hesitancy: definition, scope and determinants. Vaccine. 2015;33(34):4161–4164. (Confirm details.)
• Pelčić G, et al. (2016) — see reference 22 — on religious exemptions and the distinction between religious objection and personal belief.

Appendix B — Supplementary Figures

Appendix C — Infographic Binding Table

All twenty figures are present at full resolution in the build folder. The HTML build embeds them as base64 (compressed for size, full width for legibility).