The debate around COVID vaccination did not end when clinics closed, and mandates faded. It shifted. Early questions focused on short-term reactions, but later concerns turned toward the future. People wanted to know what vaccination meant years down the line. Online discussions grew louder, yet they often relied on fragments, personal stories, and selective screenshots. However, claims about long-term harm demand long-term evidence. Short trials cannot answer questions about mortality across several years. Therefore, large population studies become essential when uncertainty persists.
In December 2025, a French research team published a nationwide analysis in JAMA Network Open. Epidemiologist Laura Semenzato, PhD, and colleagues from EPI-PHARE led the work in collaboration with the French National Agency for Medicines and Health Products Safety. Their question was direct but consequential: do mRNA COVID vaccines increase long-term all-cause mortality in adults aged 18 to 59? The study’s Key Points state there was “no increased risk of all-cause mortality” following vaccination. The sections that follow explain how the researchers tested this question, what the data showed, and where careful limits still apply.
How the study was built
Semenzato and colleagues used the French National Health Data System, which links vaccination dates, healthcare use, and deaths from the civil registry. They included adults aged 18 to 59 who were alive on November 1, 2021. Exposure meant receiving a first mRNA dose between May 1 and October 31, 2021. Adults who stayed unvaccinated by November 1, 2021, entered the comparison group. The team assigned those adults a random index date based on vaccination dates. They censored follow-up at death, vaccination for previously unexposed people, or study end on March 31, 2025. They also ran complementary analyses using causes of death available through the end of 2023.
The team began follow-up 6 months after each person’s index date. This choice helped address immortal time bias, because everyone had to be alive to enter follow up. They then used weighted Cox models to balance sociodemographic factors and 41 comorbidities. In the Key Points section, the authors summarize the result in one sentence: “vaccinated individuals had a 74% lower risk of death from severe COVID-19 and no increased risk of all-cause mortality.” They also examined short-term mortality in a separate design, because early risks need different methods. The authors framed the question as long-term, not just weeks after dosing. That focus helps answer the claim of delayed fatal harm.
Who was included
This study did not rely on a small volunteer sample. It followed 22,767,546 vaccinated adults and 5,932,443 unvaccinated adults. Those counts cover a large share of France’s working-age population. A dataset this large can detect modest shifts in mortality risk. It also lets researchers test results across sex, age bands, and regions. The authors report that vaccinated adults were slightly older on average and more often women. They also had more cardiometabolic comorbidities at baseline, which would usually increase mortality risk. Because the data cover the whole country, the analysis avoids single-hospital bias. It also reflects real rollout rules and real access barriers. That makes the findings easier to apply to daily life.
The authors also made a restriction that strengthens fairness. They only included people with at least 1 healthcare reimbursement in 2020. That step aimed to keep both groups connected to the healthcare system. It reduces a bias where one group contains many people who rarely appear in records. Missing care records can hide diagnoses and delay documentation of events. By focusing on people with recent healthcare contact, the comparison becomes more like apples to apples. The paper also reports baseline differences, like cardiometabolic conditions, that could affect mortality if ignored. Weighting can only work when you measure these differences first. The authors used that measured detail to stabilize the comparison.
Mortality results
Over the median follow-up of 45 months, the study recorded 98,429 deaths in vaccinated adults and 32,662 deaths in unvaccinated adults. Raw totals can mislead, because the vaccinated group was much larger. The paper, therefore, reports event proportions. It found 0.4% mortality in vaccinated adults and 0.6% mortality in unvaccinated adults. Those percentages look small, yet they represent many thousands of deaths. They also provide a clear test for claims of delayed lethal harm. The authors also report that vaccinated adults had more cardiometabolic comorbidities at baseline, which would usually increase mortality risk. That makes an excess death signal even less likely to hide.
After weighting, the estimated hazard ratio for all-cause mortality was 0.75. The authors also found a marked reduction in severe COVID deaths. They report a weighted hazard ratio of 0.26 for death from severe COVID. Importantly, they saw “a similar association” for all-cause mortality even after excluding severe COVID deaths. That point is significant because it tests a simple explanation. If the entire difference came only from prevented COVID deaths, the overall association should shrink sharply. The Results section also states, “Sensitivity analysis revealed that vaccinated individuals consistently had a lower risk of death, regardless of the cause.” The paper also reports that mortality was 29% lower within 6 months following vaccination in a separate analysis. The authors describe this as a relative incidence of 0.71. These early findings match the long-term direction, even if bias plays a role.
Bias and limits
Observational studies can produce wrong conclusions when groups differ in unmeasured ways. Semenzato and colleagues took several steps to reduce this risk. They balanced groups using weighting across social factors and 41 comorbidities. They also delayed the start of the follow-up by 6 months. These choices reduce two major biases that often inflate vaccine benefits. They also make it harder for critics to argue that the results came from basic design errors. The paper also stratified results by age, sex, region, and deprivation measures, and it found consistent associations. Consistency does not erase bias, but it reduces the chance of a single subgroup driving the result.
The authors state this limitation, writing, “unmeasured confounders or healthy vaccinee bias may also contribute to this observed protective effect.” In everyday terms, vaccinated adults may seek care earlier and manage illness more consistently. They may also differ in occupational risks and household exposure. Those differences could lower mortality even without any direct non-COVID benefit. However, that does not change the central safety signal. A claim of hidden lethal harm predicts an upward mortality shift over time. This study did not show that shift. The authors also mention that comparisons require “specific methodological precautions.” They chose those precautions and reported them transparently. Readers can then judge the approach instead of guessing.
Cause-specific signals
Some long-term fears focus on specific causes of death, such as cancer or heart disease. The JAMA study explored this question using cause of death information available through December 31, 2023. It grouped causes into major ICD-10 categories, such as tumors, circulatory diseases, external causes, and COVID-19. The article also presents relative incidence estimates in defined windows after each vaccine dose. In those tables, values for several major categories stayed below 1.0 after vaccination. In practice, this approach asks whether deaths shift into particular buckets after vaccination. If a new fatal mechanism existed, it should leave a footprint in categories like circulatory disease. The study does not show that footprint. The study also reports cause-specific estimates across multiple time windows after doses.
The authors also provide clinical context about serious adverse events. They note that myocarditis, anaphylaxis, and transverse myelitis were among the main serious events reported after mRNA vaccination. Then they add a critical qualifier: these events “remained rare relative to the number of individuals vaccinated and were predominantly nonfatal.” That line matters when people jump from “event reported” to “long-term deaths must rise.” Rare nonfatal events can still require care, but they do not automatically translate into population-level mortality increases. The paper also notes that prior studies found no association with myocardial infarction, pulmonary embolism, or stroke after mRNA vaccination. Those outcomes connect to the most common alarm claims. This study extends that reassurance into a longer follow-up.
What surveillance adds
A long follow-up cohort can answer a mortality question, but it does not replace safety surveillance. Surveillance looks for rare events, especially those that occur soon after vaccination. In the United States, the CDC describes a layered system that mixes passive reporting with active monitoring. On its vaccine safety page, CDC states, “COVID-19 vaccines underwent the most intensive safety analysis in U.S. history.” It also says COVID vaccines “continue to be monitored for safety, even after FDA approval.” This monitoring matters because the background rate of illness stays high in large populations. A passive report alone cannot prove causation. Active systems compare vaccinated and unvaccinated rates and adjust for time.
CDC also explains what monitoring has detected as credible signals. It states that systems have identified “anaphylaxis and myocarditis or pericarditis” as serious types of adverse events after COVID vaccination. It also notes myocarditis has been most frequent in adolescent and young adult males within 7 days after a second mRNA dose. This detail helps interpret the French mortality findings. A rare risk can exist and still fail to shift overall death rates. CDC also explains why death reports can look alarming without implying a vaccine cause. It notes that reporting rules can require submissions even when the cause is unclear. This context reduces confusion when social media posts cite raw report counts.
How regulators respond
Vaccine oversight does not end when people start lining up for doses. Regulators require evidence before authorization, but they also require continued reporting afterward. WHO published a joint statement with the International Coalition of Medicines Regulatory Authorities to explain this oversight to clinicians. The statement says, “Regulators rigorously evaluate scientific and clinical evidence provided by vaccine manufacturers.” It also explains that manufacturers must follow defined standards for data and operations. Regulators also review manufacturing consistency, because quality issues can change risk. They inspect facilities and require documentation across batches. This work sits outside public view, yet it shapes safety outcomes.
The same WHO and ICMRA statement highlights what happens after approval. It explains how vaccine safety is “closely and continually monitored after approval.” Post approval monitoring looks for rare events and changing risk as circumstances shift. New variants appear, boosters roll out, and vaccine formulas update. Regulators respond by reviewing new data and issuing guidance. The statement also stresses confidence in the evaluation process before approval. It says the public can have confidence in the rigour of that process. This supports the idea that safety is an ongoing obligation, not a one-time decision.
Read More: FDA To Put Top-Level Warning for COVID-19 Vaccines?
What this can and cannot prove
This research provides strong evidence about one outcome in one age band. It covers adults aged 18 to 59, so it does not directly answer long-term mortality questions for older adults. The main cohort follow-up also begins 6 months after the index date. That choice helps address immortal time bias, yet it means the primary analysis does not cover the earliest weeks after vaccination. The authors assessed short-term mortality in a separate study using a different method. Another limit is the changing pandemic context across 2021 to 2025, including variants and treatment improvements. These changes affect baseline risk, yet they affect both groups living in the same society.
Residual confounding remains the biggest limitation. The database cannot fully capture occupation risks, exercise habits, diet, or housing conditions. The authors therefore warn that “unmeasured confounders or healthy vaccinee bias” may contribute to the protective association. Still, the main safety question is straightforward. If vaccination caused a meaningful long-term rise in deaths, it should appear in a cohort this large. Instead, the Conclusions and Relevance section states, “the results found no increased risk of 4-year all-cause mortality.” The study also cannot separate vaccine effects from broader behaviour differences that follow vaccination status. Still, its central finding remains: “no increased risk of 4-year all-cause mortality.” That statement addresses the core fear directly.
Conclusion
In this national French cohort, Semenzato and colleagues tracked more than 28 million adults for years. They found no increase in all-cause mortality after mRNA COVID vaccination. They also found far fewer deaths from severe COVID among vaccinated adults. The authors describe their work as the first national study to examine mortality differences at 4 years. The scale and the endpoint make the findings hard to dismiss. The Key Points section also states the study question plainly, and the findings answer it just as plainly.
People can still ask smart questions about rare side effects and updated boosters. Continuous monitoring exists for that reason, and CDC and WHO describe it plainly. Yet long-term all-cause mortality is a high-stakes outcome. This study did not show a rise in that outcome for vaccinated adults aged 18 to 59. For readers who want a grounded takeaway, this is it. The best available long follow-up evidence supports the long-term safety of mRNA COVID vaccination in this age group. When combined with ongoing monitoring described by CDC and WHO, the evidence supports continued confidence in COVID vaccination.
A.I. Disclaimer: This article was created with AI assistance and edited by a human for accuracy and clarity.
Read More: University Researchers Report Progress Toward Universal Cancer Vaccine