A new U.S. analysis involving 90,041 adults is examining whether meditation practice is associated with long-term mortality. The study, posted as a medRxiv preprint on September 20, 2026, combined four cohorts from the National Health Interview Survey with mortality follow-up through 2019.
Researchers from Pace University, Columbia University and New York University analyzed whether people who reported practicing meditation during the previous 12 months had different rates of death from any cause or cardiovascular disease compared with people who did not report recent meditation practice.
The headline finding is more complicated than a simple claim that meditation may extend life. After extensive statistical adjustment, researchers found no overall association between meditation and all-cause or cardiovascular mortality. The adjusted hazard ratio was 1.02 for all-cause mortality and 0.96 for cardiovascular mortality, with confidence intervals that included no association.
At the same time, the analysis identified potentially important differences by income. Among adults with family incomes below 200% of the federal poverty level, the estimates moved in a protective direction, although they did not reach conventional statistical significance. The researchers describe these findings as hypothesis-generating, rather than evidence that meditation prevents death.
That distinction matters because the study is based on observational survey data rather than a randomized trial.
The Analysis Included 90,041 U.S. Adults
The research team pooled data from the 2002, 2007, 2012 and 2017 National Health Interview Survey cohorts. These survey records were linked to mortality information, allowing researchers to follow participants after their original interviews.
The final analytic sample contained 90,041 adults, of whom 8,402, or 9.3%, reported meditation practice during the previous 12 months. The researchers accumulated more than 818,248 person-years of follow-up, with a median follow-up period of 7.75 years. During that period, there were 10,027 deaths from all causes and 3,040 cardiovascular deaths.
The National Health Interview Survey is a major source of information about health among the U.S. civilian, noninstitutionalized population. Its linkage with the National Death Index allows researchers to examine relationships between survey-reported health characteristics and subsequent mortality. The CDC National Death Index program explains how these linked datasets are constructed and used for mortality research.
The scale of the dataset is one of the study’s major strengths. A sample of more than 90,000 adults provides considerably more statistical information than the small clinical trials that have traditionally examined meditation and health outcomes.
However, the size of the dataset does not eliminate the central limitation of observational research: people who choose to meditate can differ from people who do not meditate in many ways.
Meditation Was Defined By Recent Self-Reported Practice
The researchers classified participants according to whether they reported any meditation practice during the previous 12 months.
This definition is important because it does not mean that all 8,402 meditation practitioners followed the same program, practiced for the same amount of time or meditated with the same frequency.
The NHIS questions changed somewhat across survey years. In earlier cohorts, meditation appeared within questions about relaxation techniques, while the 2017 questionnaire asked about spiritual, mindfulness or mantra meditation.
As a result, the analysis captures meditation as a broad category rather than as one standardized intervention.
Someone practicing mindfulness for several years could therefore be classified in the same broad exposure group as someone who had practiced meditation occasionally during the previous year.
The researchers recognized this problem and conducted additional analyses using the recency of meditation practice where comparable information was available. Those analyses produced a notable result: past meditation use was associated with lower all-cause mortality, while current past-12-month use was not.
That inconsistency is another reason the findings should not be interpreted as evidence that meditation itself changes survival.
For people interested in establishing a consistent practice, the broader principles behind meditation basics remain more relevant than the mortality association examined in this study.
The Overall Mortality Results Were Essentially Neutral
The central statistical results were not strongly protective.
After the researchers used their doubly robust statistical approach, the hazard ratio for all-cause mortality was 1.02, with a 95% confidence interval of 0.90 to 1.16.
For cardiovascular mortality, the hazard ratio was 0.96, with a 95% confidence interval of 0.79 to 1.16.
In practical terms, these estimates do not provide convincing evidence that meditation practitioners had lower long-term mortality overall.
The researchers also standardized estimated 15-year risks. For all-cause mortality, the standardized risk was approximately 14.2% under the meditation-practice scenario versus 14.7% under no practice, representing a difference of about 0.4 percentage points. The confidence interval ranged from a 1.8-percentage-point reduction to a 1.3-percentage-point increase.
For cardiovascular mortality, the corresponding standardized risks were approximately 5.5% versus 5.8%, a difference of about 0.2 percentage points.
These numbers illustrate why the headline should not be that meditation was associated with substantially longer survival. The adjusted overall estimates were close to the null.
Why The Initial Survival Difference Looked More Promising
Before statistical adjustment, the researchers observed better crude survival among meditation practitioners.
The difference was particularly noticeable among people with lower incomes. In adults below 200% of the federal poverty level, crude 15-year all-cause survival was 10.1 percentage points higher among meditation practitioners. Among those at or above that income threshold, the corresponding difference was 3.5 percentage points.
At first glance, those numbers could appear to support a substantial relationship between meditation and survival.
But the researchers then adjusted for characteristics that differed between groups.
Those crude survival differences largely disappeared after adjustment. This is a classic issue in observational research: people who meditate may also differ in education, income, health behaviors, health status or access to care.
If those characteristics are associated with both meditation and mortality, a simple comparison can make meditation appear more protective than it actually is.
The study’s statistical design was specifically intended to address this problem.
The Researchers Used A Doubly Robust Statistical Approach
Rather than relying on a single conventional regression model, the researchers used doubly robust survey-weighted Cox models.
The approach combined stabilized inverse-probability weighting with covariate adjustment. The researchers also standardized 15-year risks using g-computation and conducted probabilistic bias analyses to examine how exposure misclassification could affect the findings.
This level of statistical analysis is important because the question involves an exposure that was not randomly assigned.
A randomized clinical trial would theoretically balance measured and unmeasured characteristics between meditation and control groups. In this study, people decided for themselves whether to meditate, so the researchers had to use statistical methods to make the comparison more comparable.
The analysis found that weighting alone continued to suggest lower mortality, but adding the outcome model moved the estimates toward the null. The researchers interpreted this pattern as evidence that residual confounding was influencing the simpler analysis.
That is one of the most informative aspects of the paper: the statistical process itself demonstrated why an apparently large survival advantage should not be accepted at face value.
Lower-Income Adults Produced A Different Pattern
The study became more interesting when researchers separated participants according to family income.
Among adults with family income below 200% of the federal poverty level, the doubly robust hazard ratio was 0.85 for all-cause mortality, with a 95% confidence interval of 0.71 to 1.03.
For cardiovascular mortality, the corresponding hazard ratio was 0.80, with a 95% confidence interval of 0.57 to 1.10.
Both estimates point toward a possible protective association, but neither confidence interval excludes the possibility of no association.
Among participants at or above 200% of the federal poverty level, the estimates moved in the opposite direction. The all-cause mortality hazard ratio was 1.10, while the cardiovascular mortality hazard ratio was 1.02.
However, statistical tests of the interaction between income and meditation were not significant. The interaction p-value was 0.19 for all-cause mortality and 0.62 for cardiovascular mortality.
Therefore, the researchers cannot claim that meditation works differently according to income based on this analysis alone.
The income pattern is better viewed as a signal that deserves testing in future research.
A Smaller Subgroup Connected The Findings To Earlier Trials
The study included a second analytical component involving 2,786 non-Hispanic Black adults aged 55 or older who reported having hypertension.
This subgroup was selected because it more closely resembled the population examined in previous randomized trials of Transcendental Meditation and mortality.
Within the subgroup, 198 participants reported meditation practice. The observational estimates were imprecise because of the smaller sample.
For all-cause mortality, the hazard ratio was 0.97, with a 95% confidence interval of 0.63 to 1.49.
For cardiovascular mortality, it was 0.70, with a 95% confidence interval of 0.36 to 1.38.
The researchers then used a Bayesian analysis to combine this observational evidence with previous randomized-trial evidence.
Using a trial prior whose precision was discounted by 50%, the resulting posterior hazard ratio was 0.84 for all-cause mortality and 0.70 for cardiovascular mortality.
The probability that the hazard ratio was below 1 was estimated at 89% for all-cause mortality and 96% for cardiovascular mortality under that discounted prior.
Those numbers may sound compelling, but they are not equivalent to results from a new randomized trial. The Bayesian estimates incorporate information from earlier trials by design, so they should be interpreted as a synthesis of existing evidence rather than as an independent demonstration of benefit.
Earlier Cardiovascular Research Provides Context
The mortality question is not appearing in isolation.
The American Heart Association previously reviewed meditation as a possible component of cardiovascular risk reduction and concluded that the available evidence was limited. Its scientific statement noted possible benefits but emphasized methodological weaknesses and the need for stronger randomized research.
The new analysis extends that conversation by using a very large national dataset and long-term mortality follow-up.
At the same time, it does not resolve the underlying causal question.
The American Heart Association’s meditation evidence continues to provide useful context because cardiovascular mortality is one of the two major outcomes examined by the new study.
The broader evidence base also contains research examining meditation in relation to blood pressure, anxiety, depression, pain and sleep. The National Center for Complementary and Integrative Health notes that results vary considerably according to the specific meditation approach and health outcome being studied.
That distinction becomes particularly important when moving from short-term physiological changes to an outcome as complex as mortality.
The Study Cannot Show That Meditation Extends Life
The most important limitation is the observational nature of the research.
Participants were not randomly assigned to meditate or avoid meditation. Their practice was reported through survey questions, and the researchers did not control which meditation technique they used, how often they practiced or how long individual sessions lasted.
That leaves several potential sources of confounding.
People who meditate may have different health behaviors, social circumstances, education, healthcare access or baseline health conditions than people who do not. Even sophisticated statistical adjustment cannot guarantee that all relevant differences have been measured.
There is also the problem of exposure misclassification.
A broad question about meditation during the previous year cannot capture the difference between occasional practice and a sustained daily routine. The researchers therefore performed probabilistic bias analyses to estimate how measurement error could influence their findings.
The researchers also acknowledge that the meditation questions changed across survey years and that different respondents may have understood the term differently.
These limitations mean the paper can identify associations and patterns, but it cannot demonstrate that meditation caused any change in mortality.
The Findings Are More Useful For The Next Research Question
The most interesting contribution of this analysis may therefore be the question it raises rather than a definitive answer.
Overall, meditation practice was not associated with lower mortality after doubly robust adjustment. Yet the estimates among lower-income adults moved in a potentially protective direction, while the trial-matched subgroup produced Bayesian estimates that favored cardiovascular benefit.
Those signals are enough to justify more targeted research, but not enough to recommend meditation as a mortality-prevention treatment.
A stronger next step would be a large randomized study with a standardized meditation intervention, objective measures of adherence, long-term follow-up and clinically adjudicated cardiovascular outcomes.
Such a study could also examine whether socioeconomic conditions modify the effect of meditation.
That question is particularly relevant because the new analysis suggests that the relationship between meditation and mortality may not be uniform across populations.
The researchers themselves describe their findings as hypothesis-generating and call for additional work using more detailed measurements of meditation exposure.
What The 90,041-Person Study Actually Shows
This new analysis provides one of the broadest U.S. examinations to date of meditation practice and long-term mortality, using 90,041 adults, 818,248 person-years of follow-up, 10,027 all-cause deaths and 3,040 cardiovascular deaths.
Its central result is cautious: there was no overall association between meditation practice and mortality after extensive adjustment.
The study did identify potentially important patterns. Adults below 200% of the federal poverty level had mortality estimates that leaned toward protection, and a trial-informed Bayesian analysis of older non-Hispanic Black adults with hypertension produced a high probability of benefit for cardiovascular mortality. But neither finding establishes causation.
The timing of the research also matters. The paper was posted on September 20, 2026 as a medRxiv preprint, meaning it had not yet gone through conventional peer review at the time of publication.
For now, the evidence supports a narrower interpretation: meditation remains a promising subject for cardiovascular and longevity research, but this large U.S. dataset does not show that people who meditate live longer.
The next generation of studies will need to determine whether the patterns observed here reflect a genuine effect of meditation, differences between people who choose to meditate and those who do not, or some combination of both.