Which vaccines are safe?

The short version: no vaccine is "safe" in the abstract. Safety is a property of the match between a specific person, a specific product, and the evidence that was actually run.

The question "which vaccines are safe" cannot be answered as a list. Safety is not a stamp a regulator gives. It is a question of whether the evidence for a specific vaccine in a specific population is strong enough to justify the risk. The burden of proof is on whoever asks you to accept the injection, and that burden has not been met equally for every vaccine.

For the COVID-19 mRNA vaccines, the evidence on all-cause mortality is deeply contradictory. A Norwegian cohort study of 4.6 million adults found that people with three or more doses had about 58% lower adjusted all-cause mortality than the unvaccinated [2]. But a UK analysis of ONS data found that the relative risk of death for vaccinated people started lower and then rose over time, crossing the unvaccinated line between September 2022 and January 2025 depending on age [3]. A US county-level study of over 3,000 counties found that higher vaccination uptake was associated with more excess mortality in 2022 and 2023, not less [5]. And a re-analysis of Australian data concluded that the claim that vaccination reduced all-cause mortality by 51% in older adults is mathematically implausible, because it would require an excess death rate of 121% in the unvaccinated, which has never been seen in any OECD country [1]. These studies cannot all be true. The evidence base is not settled. It is a mess.

For influenza vaccination in older adults and people with diabetes, the evidence is more consistent but still observational. A meta-analysis of six studies in people with diabetes found that flu vaccination was associated with about 46% lower all-cause mortality [12]. A Dutch cohort of 223,580 adults found a 10% reduction [6]. But these are observational studies, not randomized trials. The healthy vaccinee effect, where healthier people are more likely to get vaccinated, can easily produce these numbers. The studies try to adjust for it, but adjustment is never perfect. The real effect is probably smaller than the headline number.

For the childhood vaccines, the evidence is older and thinner than most people realize. The 1982 review by Fenichel cataloged neurological complications after routine childhood immunizations: encephalopathy after DPT at about 1 in 110,000 doses, paralytic polio from the live oral polio vaccine at about 1 in 3 million doses, and encephalopathy after measles vaccine at about 1.16 per million doses [43]. Rotavirus vaccines carry a known risk of intussusception, with a relative incidence of about 7 to 10 in the first week after the first dose [44]. These are real risks, and they are not zero.

My call: no vaccine is universally safe. The evidence for influenza vaccine in older adults and people with diabetes is the most consistent, but it is observational and the real benefit is probably modest. The evidence for COVID-19 mRNA vaccines is contradictory and unsettled. The childhood vaccines have documented risks that are small but real, and the long-term safety studies that would settle the question were never run.

Confidence: low for any blanket statement about vaccine safety. The evidence is too thin, too conflicted, and too dependent on who funded the study and who was left out of the comparison group.