When Pfizer announced in November 2020 that its COVID-19 vaccine had 95% efficacy, the number seemed straightforward. Most people interpreted it as: "If 100 vaccinated people are exposed to the virus, only 5 will get sick." That interpretation is wrong, and the misunderstanding has practical consequences for how people assess their risk during outbreaks.
What efficacy actually measures
Vaccine efficacy is a relative risk reduction, not an absolute probability. It's calculated by comparing infection rates between a vaccinated group and an unvaccinated control group in a clinical trial, using this formula: (1 - risk in vaccinated group / risk in unvaccinated group) x 100.
Here is what the Pfizer trial data actually showed. The phase 3 trial enrolled 43,548 participants, split roughly evenly between vaccine and placebo groups. During the observation period, 8 out of 18,198 people in the vaccine group developed symptomatic COVID-19, compared to 162 out of 18,325 in the placebo group. Plugging those numbers in: 1 - (8/18,198) / (162/18,325) = 0.9504, or approximately 95%.
The 95% means that among people who would have gotten infected without the vaccine, 95% of them were protected. It does not mean that 5% of all vaccinated people will get infected. The baseline infection rate during the trial period was low - only about 0.88% of the placebo group got sick. So the absolute risk dropped from 0.88% to 0.04%, a difference of 0.84 percentage points.
Absolute vs. relative risk reduction
This distinction matters and is routinely overlooked. Relative risk reduction sounds dramatic: 95%. Absolute risk reduction sounds modest: 0.84 percentage points. Both numbers are correct. They describe the same data from different angles.
Absolute risk reduction depends on how common the disease is in the first place. During a massive outbreak wave where 10% of unvaccinated people get infected over a month, a 95% efficacy vaccine would reduce that to 0.5% - an absolute risk reduction of 9.5 percentage points. During a lull when only 0.1% of unvaccinated people would be infected, the same vaccine reduces risk to 0.005% - an absolute reduction of just 0.095 percentage points.
The vaccine didn't change. The context did. This is why vaccine benefits are most visible during high-transmission periods and harder to perceive during quiet periods. It's also why people sometimes feel that a vaccine "stopped working" when transmission declines - the absolute benefit becomes small even if relative protection remains strong.
Efficacy vs. effectiveness
Efficacy is measured under controlled clinical trial conditions with carefully selected participants, strict dosing protocols, proper cold chain storage, and defined follow-up periods. Effectiveness is what happens in the real world.
Real-world effectiveness is typically lower than trial efficacy for several reasons. Trial participants are often healthier than the general population (exclusion criteria screen out people with certain medical conditions). Storage and handling may be imperfect in mass vaccination campaigns. Timing between doses may vary. And the population mix includes people with compromised immune systems who may respond less robustly to vaccination.
The gap between efficacy and effectiveness varies by vaccine. Measles vaccine has a trial efficacy of about 95% for a single dose and 97% for two doses - and real-world effectiveness is nearly identical because the immune response to measles vaccination is strong and durable. Influenza vaccines show a bigger gap: typical efficacy in trials runs 40-60%, but real-world effectiveness in some seasons drops to 20-30% because the circulating strains may differ from those targeted by that season's formulation.
For COVID-19 vaccines, initial real-world effectiveness closely matched trial efficacy. Israeli data from early 2021, covering over 1.2 million Pfizer-vaccinated individuals, showed 94% effectiveness against symptomatic infection and 97% against hospitalization. Those numbers held for approximately 4-5 months before declining.
Why efficacy drops over time
Waning immunity is not a vaccine defect. It's immunology. Most vaccines trigger strong initial antibody responses that decline over months as short-lived plasma cells die off. Longer-lived memory B cells and T cells persist, providing durable protection against severe disease, but circulating antibody levels drop enough that breakthrough infections become possible.
A study published in The Lancet in October 2021 analyzed data from 3.4 million adults in the US and found that Pfizer vaccine effectiveness against infection declined from 88% in the first month after full vaccination to 47% after five months. Effectiveness against hospitalization declined more slowly: from 93% to 88% over the same period.
This pattern - faster waning of protection against infection than against severe disease - is common across many vaccines. Your immune system may not prevent the pathogen from gaining an initial foothold (infection), but memory immune cells mobilize quickly enough to prevent it from causing serious damage (hospitalization or death). This is why booster doses focus on restoring sterilizing immunity rather than building it from scratch.
Different endpoints tell different stories
A single efficacy number can be misleading because vaccines often protect against different outcomes at different rates. The most commonly measured endpoints are:
Infection (testing positive regardless of symptoms). This is the hardest endpoint for a vaccine to achieve because it requires preventing the pathogen from replicating at all.
Symptomatic disease (testing positive with symptoms). This is the standard endpoint in most vaccine trials, including the original COVID-19 trials.
Hospitalization (severe enough to require hospital care). Vaccines almost always perform better here than against infection alone.
Death. Vaccines typically show their highest efficacy against mortality.
During the Omicron wave, COVID-19 vaccine effectiveness against infection dropped significantly - in some studies to 30-40% for two doses. But effectiveness against hospitalization remained 70-80%, and against death, 85-90% with a booster. A vaccine that is 40% effective against infection and 90% effective against death is still an extraordinarily useful medical intervention, but the headline number of 40% sounds underwhelming to anyone who remembers 95%.
When you evaluate vaccines during an outbreak, always ask: efficacy or effectiveness against what? The endpoint matters more than the number.
How variants affect efficacy
RNA viruses mutate constantly. When mutations occur in the regions targeted by vaccine-induced antibodies - particularly the spike protein for COVID-19 vaccines - the antibodies may bind less effectively to the new variant. This is antigenic drift, and it's the same reason influenza vaccines need annual reformulation.
The original COVID-19 vaccines targeted the Wuhan strain spike protein. Against Alpha, efficacy held well (roughly 90%+ for mRNA vaccines). Against Delta, it dipped to approximately 80-88% for symptomatic disease. Against Omicron BA.1, it dropped to approximately 30-40% for infection after two doses, though booster doses partially restored protection.
This dynamic connects to disease severity scoring and threat assessment. A new variant that substantially evades vaccine-derived immunity is a different threat level than one that doesn't, even if its inherent severity is similar. Monitoring variant-specific vaccine effectiveness data is one of the inputs PandemicAlarm uses when evaluating outbreak severity.
Vaccines and the R0 connection
Vaccination's population-level effect ties directly back to the basic reproduction number. If a pathogen has an R0 of 5, and a vaccine is 90% effective at preventing transmission, then vaccinating at least 89% of the population (the herd immunity threshold of 1 - 1/R0 = 80%, adjusted for imperfect vaccine efficacy) should push R-effective below 1 and halt transmission.
But this calculation depends on the vaccine preventing transmission, not just disease. Some vaccines do both - measles vaccination both prevents disease and blocks onward spread. Other vaccines primarily prevent severe disease without fully stopping transmission. Early COVID-19 vaccines reduced transmission by an estimated 40-60% - meaningful, but not enough to achieve herd immunity through vaccination alone given the Omicron variant's R0 of 8-15.
Understanding this distinction prevents both overconfidence ("everyone's vaccinated, we're safe") and dismissal ("the vaccine doesn't even stop spread, why bother"). Vaccines that reduce but don't eliminate transmission still lower R-effective, slow outbreak growth, and save lives. They just can't end a pandemic on their own for highly transmissible pathogens. Other measures - improved ventilation, mask use, testing and isolation - remain part of the toolkit.
The numbers on a vaccine data sheet are not simple. But the core message is: read the endpoint, watch for waning, and distinguish relative from absolute risk. That's enough to evaluate outbreak-related vaccine claims without being misled.