Vaccines

How they work and how we know

Andreas Handel

First-Year Odyssey, University of Georgia

November 2026

Question

In the WHO reading, a vaccine has 80% efficacy against disease. This means:

  1. A vaccinated person has an 80% chance of not getting the disease.

  2. A vaccinated group has an 80% lower disease risk compared to an unvaccinated group.

  3. The vaccine reduces symptoms by 80%.

  4. All of A)–C).

  5. None of A)–C).

About me

  • UGA Foundation Professor in Public Health, Department of Epidemiology and Biostatistics.
  • I model and analyze infectious diseases, especially influenza and norovirus.
  • I study how infections spread and how we can control them.
  • I teach infectious diseases, data analysis, and epidemiology.

Current headshot of Andreas Handel, from his personal website.

More about me

  • Originally from Germany, I moved to the US for graduate school.
  • I trained as a physicist. I have never taken an epidemiology or biostatistics class.
  • Away from work, I grow food, cook, eat, and mountain bike.

Andreas Handel mountain biking.

What would you like to know about vaccines?

XKCD comic: asked what they will do after vaccination, a character describes making spike proteins and presenting antigens to T cells.

Comic: Randall Munroe, xkcd: First Thing, CC BY-NC 2.5.

How vaccines work

How natural infection builds protection

Infection can create immune memory, but the person faces the disease itself.

Natural infection can produce antibodies and memory cells that help during later exposure, but the initial infection also carries a risk of serious disease.

Concepts: WHO, How do vaccines work?. Original AI-generated illustration. Illness severity and later protection vary.

How vaccination builds protection

Vaccination prepares immune defenses before natural exposure.

Vaccination leads to antibodies and immune memory and a faster response on later exposure. A separate downward arrow from vaccination shows the rare risk of serious side effects.

Concepts: WHO, How do vaccines work?; WHO, Vaccine safety. Original AI-generated illustration; branch sizes do not represent probabilities.

Different ways to make vaccines

Three vaccine approaches: a weakened pathogen, selected protein components, and RNA instructions that enable a cell to make an antigen.

Weakened pathogen

Replicates, but is much less able to cause disease.

Example: measles, polio

Selected component

A particular part of the pathogen.

Example: HPV, influenza

Genetic instructions

Cells make an antigen using the instructions.

Example: mRNA COVID-19, influenza

Selected approaches, not a complete list. Original AI-generated illustration.

Vaccine development

Original vaccine safety evaluation figure: preclinical work, phases 1–3, FDA review, and phase 4 monitoring. The illustrative stages progress to larger populations.

Knipe et al. (2020), Ensuring vaccine safety, Science. Original lecture figure; sample sizes are illustrative.

Measuring vaccine impact

The thing we can never measure

Suppose you get vaccinated and stay well. Did the vaccine prevent an illness?

One student has two possible paths at the same decision: vaccination followed by an observed healthy outcome, or no vaccination with an unobserved outcome marked by a question mark.

Your counterfactual is what would have happened to you, over the same period, if you had not been vaccinated.

We observe only one path. You might have stayed well either way, or vaccination might have prevented illness.

Concept: Hernan & Robins, Causal Inference: What If, Chapter 1. Same person, same decision, same follow-up period. This is a thought experiment, not a prediction. Original AI-generated illustration.

The thing we can measure

We can compare groups of people that do and do not get the vaccine.

Volunteers are assigned by chance to vaccine or comparison groups, followed over the same period, and their outcomes compared.

Assign by chance, then compare the same outcomes over the same period.

FDA, Vaccine Development 101. Original AI-generated illustration. The comparison can be placebo or another appropriate treatment.

What outcome do we measure?

Four illustrated outcomes: infection, illness, severe illness, and transmission. These are distinct risks that vaccination may reduce.

A vaccine can protect more strongly against one outcome than another.

Concepts: WHO, Vaccine efficacy, effectiveness and protection. Original AI-generated illustration. Effects vary across vaccines and pathogens.

How to quantify vaccine impact

Hypothetical trial, one-year follow-up

Group Participants Developed disease Disease risk
Comparison 1,000 100 10%
Vaccine 1,000 20 2%

Relative reduction: (10% − 2%) / 10% = 80%

Hypothetical teaching example. Both groups use the same disease definition and follow-up period. The estimate concerns this disease outcome and population.

A real trial: COVID-19 in 2020

Cumulative percentage of trial participants with COVID-19 over days after the first dose. The placebo curve rises substantially more than the vaccine curve. An inset enlarges the early period.

Polack et al. (2020), NEJM, Figure 3. Historical Pfizer–BioNTech trial. This is not a current effectiveness estimate.

Evidence after a trial

Efficacy: protection in a controlled trial.
Effectiveness: protection in everyday use.

Clinicians record vaccination and health outcomes during routine care. Researchers compare those records while accounting for differences in age, health, access, and exposure.

Researchers observe vaccination and outcomes without assigning vaccination. Comparisons must account for differences between the groups.

WHO, Vaccine efficacy, effectiveness and protection; FDA, Vaccine Development 101. Original AI-generated illustration.

Protection can extend to other people

With easier spread, infection can travel through susceptible people to an infant. When vaccinated people interrupt transmission, the infant can receive indirect protection.

Vaccination can interrupt chains of infection, helping protect people who remain susceptible.

Concept: WHO, How do vaccines work?. Original AI-generated illustration. Simplified example when vaccination reduces transmission. Protection varies and is not guaranteed.

Measles before and after vaccination

UK measles reports, 1950–2016: large recurring epidemics before vaccination began in 1968 and much lower reported counts afterward. The original measles panel is shown with its axes and introduction marker.

Historical UK data. Measles panel from Figure 1, Pollard & Bijker (2021), adapted there from Public Health England.

Recent U.S. measles cases

U.S. confirmed measles cases: 59 in 2023, 285 in 2024, 2,289 in 2025, and 4,852 through October 29, 2026. The partial-year bar is hatched.

CDC, Measles Cases and Outbreaks, checked November 2, 2026. Includes international visitors. 2025–2026 preliminary; 2026 is a partial year, not a projection.

Vaccine safety

Vaccine safety and monitoring

Illustrations of common mild vaccine reactions, rare serious reactions, and researchers comparing health records during ongoing safety monitoring.

Researchers compare event rates and check whether other studies find the same pattern. Monitoring continues after approval.

WHO, Vaccine safety; CDC, Safety monitoring. An event after vaccination does not, by itself, establish causation. Original AI-generated illustration.

Safety monitoring in action

Vaccine Rare problem found after approval Response
RotaShield rotavirus (1998) Bowel blockage (intussusception), about 1 in 10,000 infants Withdrawn in 1999. Later rotavirus vaccines tested in 60,000+ infants.
Johnson & Johnson COVID-19 (2021) Rare blood clots with low platelets, mostly in women under 50 Paused, then restricted. No longer used in the U.S.
Dengvaxia dengue (2015) Higher risk of severe dengue in people never infected before Now given only to people with a confirmed past infection.

Problems that are very rare, or limited to one group, can be missed in trials of thousands. Monitoring continues after approval.

Murphy et al. (2001), NEJM; Ruiz-Palacios et al. (2006) and Vesikari et al. (2006), NEJM; MacNeil et al. (2021), MMWR; Sridhar et al. (2018), NEJM. Anyone can submit a VAERS report; reports are signals to investigate, not proof of causation.

Comparing health benefits and harms

What benefits does the vaccine provide?

Apply the trial’s 10% versus 2% disease risks to 100,000 people over one year.

Strategy Disease cases Deaths from disease
No vaccination 10,000 100
Vaccination 2,000 20

8,000 cases and 80 deaths prevented.

Fictional teaching example. Assume 1 in 100 disease cases is fatal in either group. These are illustrative assumptions, not estimates for any real vaccine.

What harms does the vaccine cause?

Now include reactions caused by vaccinating those 100,000 people.

Vaccine-caused reaction Number affected
Temporary mild reaction, such as a sore arm 20,000
Serious nonfatal reaction 10
Fatal reaction 1

A brief sore arm and a death have very different consequences.

Fictional teaching example. These are assumed excess harms caused by vaccination, not all health events reported afterward. The reaction rates are invented for illustration.

Putting benefits and harms together

Deaths in this hypothetical population Number
Without vaccination: disease deaths 100
With vaccination: 20 disease deaths + 1 vaccine-caused death 21
Net deaths prevented 79

Vaccination also prevents 8,000 disease cases, while causing 20,000 temporary mild reactions and 10 serious nonfatal reactions.

Fictional comparison of 100,000 people over one year. Outcome counts should be interpreted by severity, not added together as equivalent events.

Making a personal decision under uncertainty

Our example estimates 79 net deaths prevented in a population, but does not identify whose death was prevented.

Researchers study groups, focus on people with similar characteristics, and use that evidence to help one person discuss possible benefits and harms with a clinician.

Evidence about people with similar age, health, and exposure helps estimate your chances of benefit and harm.

Evidence can guide your choice without guaranteeing your outcome.

Concept: Hernan & Robins, Causal Inference: What If, Chapter 1. The 79-death estimate is from our fictional example. Individual reactions can still be investigated and sometimes strongly attributed to a vaccine. Original AI-generated illustration.

Vaccine challenges

Basic immunological challenges

  • If you recover from measles, how long are you immune for?

  • If you get a measles vaccine, for how long are you protected?

  • If you recover from chlamydia, how long are you immune for?

  • If you get a chlamydia vaccine, for how long are you protected?

Dosing and timing challenges

Three vaccine schedule decisions: choosing a starting age, determining dose spacing, and deciding whether later doses are needed.

Starting age: When is protection needed, and when can the vaccine work well?

Initial series: What dose amount, how many doses, and how far apart?

Later doses: Are boosters or updated vaccines needed, and how often?

Concepts: CDC, Timing and spacing of immunobiologics; WHO, Vaccine protection and boosters. Schedules differ by vaccine and recipient. Original AI-generated illustration.

Vaccination challenges

Three considerations for vaccine uptake: learning about a vaccine, discussing benefits and concerns with a clinician, and reaching an accessible vaccination clinic.

Awareness: Knowing the vaccine exists and who it is for.

Acceptance: Willingness to receive it, shaped by confidence and concerns.

Access: Being able to get it: cost, transport, clinic hours, and supply.

Adapted from the 5As in Thomson et al. (2016), Vaccine; here Access includes affordability, and activation (e.g., reminders) is omitted. Related: WHO, Behavioural and social drivers of vaccination (2022). Original AI-generated illustration.

Vaccines and autism: timing and cause

Autism signs and routine vaccination can occur during the same age period.

A vaccination visit and a developmental observation both occur during early childhood. The two scenes share an age period, with no causal arrow connecting them.

We can’t directly measure the counterfactual but we can use the comparison groups approach again.

Children’s Hospital of Philadelphia, Vaccines and autism; WHO GACVS (2025). Original AI-generated illustration.

Testing the concern: a Danish study

657,461 children in Denmark, followed using health records (2019 study).

Researchers compare existing health records for children with and without MMR vaccination, then compare autism diagnoses during follow-up. Vaccination was not assigned by the researchers.

MMR-vaccinated children had no higher autism risk.

Researchers also considered differences such as age and family history.

WHO’s 2025 review: vaccines do not cause autism.

Hviid et al. (2019); WHO GACVS (2025). MMR = measles, mumps, rubella. Observational study, not a randomized trial. Original AI-generated illustration.

Questions?

XKCD comic: after reading vaccine studies, a character concludes that vaccines are pretty good and hopes more people will receive them.

Slides: https://www.andreashandel.com/presentations/

Reading 1: WHO: How do vaccines work?

Reading 2: WHO: Vaccine efficacy, effectiveness and protection

Comic: Randall Munroe, xkcd: Vaccine Research, CC BY-NC 2.5.