CBT-I science

Can treating insomnia slow biological aging? What one new trial found

A 2026 randomized trial measured DNA methylation before and after CBT-I. The result is genuinely interesting, and weaker than the headline it will get.

October 7, 20267 min read

Last reviewed October 7, 2026

Two lines tracked over time - one holding level, the other climbing away from it.

One trial, published in August 2026, found that older adults who received CBT-I aged more slowly on one measure of biological aging than older adults who received sleep education instead. That is a real result from a randomized trial, and it is considerably weaker than the headline it is going to get.

Here is what it actually showed.

What the study did

Researchers at UCLA had already run a randomized trial of CBT-I in adults aged 60 and over who met diagnostic criteria for insomnia disorder. Participants were assigned either to CBT-I or to sleep education therapy - not a waiting list, but an active comparison condition covering sleep hygiene, sleep biology, and stress biology. Both ran over eight weeks in group sessions.

Blood was drawn before treatment and again at a follow-up visit roughly two years later. This analysis took those samples and measured DNA methylation, then ran three established epigenetic clocks on the results.

Ninety-two people had usable paired samples: 47 in the CBT-I group, 45 in sleep education. Mean age was 69.5. The average gap between the two blood draws was about 24 and a half months.

The authors describe this as, to their knowledge, the first study to examine whether treating insomnia affects the pace of biological aging. The parent trial and this analysis were funded by the National Institute on Aging, and the authors declare no competing interests.

What it found

Three clocks were used, and they did not agree.

DunedinPACE, which estimates how fast a person is currently aging rather than how old they look, showed a difference between the groups: a group-by-time coefficient of −0.02 (95% CI −0.04 to −0.01), significant after correction for multiple testing. Read that as 0.02 fewer years of biological aging per chronological year in the CBT-I group.

GrimAge showed no significant difference between the groups (−0.33; 95% CI −0.68 to 0.03).

PCPhenoAge showed no significant difference either (−0.49; 95% CI −1.34 to 0.36).

Consistent with the parent trial, the CBT-I group was more likely to reach full remission of their insomnia: 34% (16 of 47) against 13% (6 of 45).

The finding underneath the finding

This is the part worth slowing down for, because almost every summary of this paper will skip it.

Within the CBT-I group, the pace of aging did not significantly change from baseline (mean change −0.01; 95% CI −0.03 to 0.01). Within the sleep education group, it increased (mean change 0.03; 95% CI 0.01 to 0.05), and that increase was significant.

So the difference between the groups was driven substantially by the comparison group getting worse, not by the CBT-I group getting better. The honest description is that CBT-I was associated with holding steady while the group without effective treatment accelerated.

An exploratory analysis pushes in the same direction: among people in the sleep education group whose insomnia did not remit, the pace of aging rose most of all (mean change 0.04; 95% CI 0.01 to 0.06).

That framing is less dramatic than "therapy reverses aging" and it is more interesting, because it points at untreated insomnia as the thing doing damage.

One clock did move within the CBT-I group on its own terms: GrimAge acceleration fell by 0.84 years (95% CI −1.35 to −0.34). But the comparison between groups on that same clock was not statistically significant, which is the test that matters for attributing anything to the treatment.

What an epigenetic clock is, and is not

DNA methylation is a set of chemical marks on DNA that change in patterned ways as people age. Epigenetic clocks read those patterns and produce an estimate - either of biological age relative to chronological age, or of the current rate of aging.

These measures predict disease and mortality risk across large populations, which is why researchers use them. They are not a clinical test, they are not diagnostic, and a number from one is not a fact about how long a particular person will live. In this study they are a surrogate measure: no participant's health outcomes were tracked. Nobody in this trial was shown to have avoided a disease.

Take the free assessment to see:

✓Your Insomnia Severity Index (ISI) score
✓A personalized 6-week CBT-I plan, built by a sleep physician
✓How your sleep improves, night by night

Start free assessment

Why this is preliminary, in the authors' own terms

The paper is unusually candid about its limits, and they matter.

It is a secondary analysis. Biological aging was not the parent trial's primary outcome. The trial was not designed to answer this question.

It used a fraction of the randomized sample. 431 people were assessed, 291 were randomized, and 92 ended up in this analysis - selected on the basis of which stored blood samples were available, for reasons of cost. The authors write that this "might have resulted in a loss of randomization" and raise selection bias and imprecision directly.

It was underpowered. A post-hoc calculation put the study at 73% power to detect an effect in at least one clock after correction. Below the conventional 80%.

The sample was narrow. Adults 60 and over, one university medical center in southern California, 84% white, and healthy enough to exclude major depression and serious medical conditions. The authors note findings might differ in populations with chronic inflammatory conditions, and that generalizing beyond this group is not warranted yet.

There were only two measurements per person. Two points describe a line; they cannot describe a trajectory.

The groups differed by sex. 45% of the CBT-I group were women against 73% of the comparison group. It was adjusted for, and the authors still say sex-specific effects cannot be ruled out.

Their conclusion is that replication in larger, adequately powered studies "is required." That is the correct reading of this paper, and it is theirs, not ours.

What this does not say

It does not say CBT-I is an anti-aging treatment. It does not say a specific number of years were added to anyone's life. It says nothing about adults under 60, because none were studied.

And it says nothing about self-guided programs or apps, including this one. The intervention here was eight weeks of therapist-led group sessions in a clinic. Whether a self-guided format produces the same biological signal is untested. The evidence comparing self-guided and therapist-led CBT-I is about sleep outcomes, not methylation.

Anyone citing this study as a benefit of a product is going considerably further than the data.

Why it is still worth knowing about

Because it changes the shape of the question.

Chronic insomnia has been linked to cardiovascular disease, cognitive decline, frailty and mortality in observational research, where the obvious objection is that sick people sleep badly. A randomized trial that measures a biological marker before and after treatment is a different kind of evidence. It is the first attempt at it, it is small, and it points in the direction the observational work already pointed.

The practical implication is unchanged: chronic insomnia is worth treating, and the treatment with the strongest evidence behind it is CBT-I. This trial adds a reason to think the stakes of leaving it untreated may be higher than a bad night, and it does not change what you would do about it.

If you want the evidence on sleep itself rather than on aging, that is in what CBT-I does and does not improve, and the rest of this section covers the method.

For how CBT-I is adapted later in life, see CBT-I for older adults.

Somnera is a self-guided education program built on CBT-I, founded and written by Dr. Camilo Ruiz, DO, FACOI, FAASM. It is not a diagnosis and does not replace care from your own clinician. The assessment is free and takes about two minutes.

Citations

Carroll JE, Kusters CD, Taha HB, Olmstead R, Breen EC, Irwin MR. Cognitive behavioural therapy for insomnia and epigenetic ageing: secondary analysis from a randomised controlled trial. Lancet Healthy Longev. 2026;7:100861. doi:10.1016/j.lanhl.2026.100861
Belsky DW, Caspi A, Corcoran DL, et al. DunedinPACE, a DNA methylation biomarker of the pace of aging. eLife. 2022;11:e73420
Lu AT, Quach A, Wilson JG, et al. DNA methylation GrimAge strongly predicts lifespan and healthspan. Aging (Albany NY). 2019;11:303–327
Irwin MR, Carrillo C, Sadeghi N, Bjurstrom MF, Breen EC, Olmstead R. Prevention of incident and recurrent major depression in older adults with insomnia: a randomized clinical trial. JAMA Psychiatry. 2022;79:33–41
Edinger JD, et al. Behavioral and psychological treatments for chronic insomnia disorder in adults: an American Academy of Sleep Medicine clinical practice guideline. J Clin Sleep Med. 2021;17(2):255–262. doi:10.5664/jcsm.8986

Frequently asked questions

One randomized trial published in 2026 found that adults over 60 who received CBT-I had a slower pace of biological aging on one epigenetic measure than adults who received sleep education instead. It was a secondary analysis of 92 people, two of the three measures showed no difference, and the authors call for replication. It is a promising early finding, not an established effect.

An estimate of biological age built from DNA methylation patterns - chemical marks on DNA that change with age. Some clocks estimate how old you look biologically; DunedinPACE, used in this trial, estimates how fast you are currently aging. They are research measures that predict disease risk across populations, not a diagnostic test.

On DunedinPACE, the difference between groups over roughly two years was 0.02 years of biological aging per chronological year, with a confidence interval of 0.01 to 0.04. That is a small difference, and it is reported as a group average rather than something an individual would notice.

Not on the main measure. The CBT-I group's pace of aging did not change significantly from baseline. What changed was the comparison group, whose pace of aging increased. The between-group difference was driven substantially by that.

The trial tested CBT-I delivered over eight weeks in group sessions to adults aged 60 and over at a university clinic. It did not test a self-guided program, an app, or younger adults. Nothing in it can be transferred to those without being tested.