Acupuncture for Insomnia: What PSG Sleep Data Really Shows

Acupuncture cleared a better test for insomnia—but not the final one

Acupuncture for primary insomnia has evidence that goes beyond people reporting that they slept better. A 2021 meta-analysis restricted to polysomnography (PSG)—the overnight EEG-based test used to measure sleep—found improvements in total sleep time, sleep efficiency, wake time after sleep onset, and awakenings. That matters because insomnia often includes a mismatch between how much people feel they slept and what a sleep recording shows.

The result is encouraging, not conclusive. The PSG-only evidence base included 11 randomized controlled trials and 775 participants, all in Chinese clinical populations. The review found substantial methodological bias, heterogeneity among treatment protocols, and publication bias. Acupuncture looks like a reasonable non-drug option for some people with insomnia, particularly when access to cognitive behavioral therapy for insomnia (CBT-I) is limited. It has not shown that it beats CBT-I, sleeping pills, or every other non-drug treatment.

Why objective sleep measurement changes the conversation

Most acupuncture-insomnia research has leaned heavily on questionnaires such as the Pittsburgh Sleep Quality Index and Insomnia Severity Index. Those measures capture the experience of insomnia, which is clinically important, but they cannot by themselves show that sleep architecture changed.

That distinction is unusually relevant in insomnia. Yoon et al. (2022) found that sleep-state misperception can run in both directions: some patients underestimate objectively measured sleep, while others have a different pattern of mismatch between perceived and recorded sleep. Cini et al. (2025) described sleep-state misperception as related to sleep instability and emotional dysregulation, rather than mere questionnaire noise. A treatment can improve how a person experiences sleep without changing EEG-defined sleep to the same degree, or vice versa.

PSG does not settle every clinical question. It records sleep under laboratory conditions and may not capture every night at home. But it makes it harder to confuse a shift in perception with a shift in measured sleep continuity.

What the PSG-only meta-analysis found

Zhao et al. (2021) pooled 11 randomized trials of acupuncture for primary insomnia that reported PSG outcomes. Trials compared real acupuncture with sham needling, placebo, or a waitlist. Across the pooled analysis, acupuncture increased total sleep time by 55.29 minutes (95% confidence interval [CI], 29.16 to 81.42; p<0.01) and sleep efficiency by 8.96 percentage points (95% CI, 3.97 to 13.95; p<0.01).

The findings also point to better sleep maintenance. Wake after sleep onset fell by 49.54 minutes (95% CI, −82.98 to −16.09; p<0.01), and the number of awakenings fell by 6.29 events per night (95% CI, −10.75 to −1.82; p<0.01). These are measured sleep-continuity outcomes, not simply better questionnaire scores.

The confidence intervals excluded no difference for those four pooled outcomes, so the results were statistically significant. Clinical importance is harder to judge. The trials differed in their acupuncture protocols and control conditions, and the evidence available here does not provide a direct PSG benchmark against CBT-I or approved hypnotics. A 55-minute pooled gain in total sleep time sounds large; it should not be treated as a personalized forecast for the next person who books an appointment.

The effect appears specific to sleep maintenance

The same meta-analysis did not find statistically significant pooled effects on sleep-onset latency or the percentage of N3 slow-wave sleep. Sleep-onset latency matters most for people whose central problem is falling asleep. N3 is commonly called deep sleep and is a major part of restorative sleep physiology. The available PSG evidence therefore supports an effect on fragmentation and sleep consolidation more than a demonstrated effect on falling asleep faster or increasing deep sleep.

That is a useful boundary. Someone who wakes repeatedly or wakes early may fit the measured pattern better than someone who lies awake for an hour at bedtime despite otherwise consolidated sleep.

Session count may matter, although the dose rule is not settled

Zhao et al. found measurable pooled effects only in protocols with at least 12 acupuncture sessions. Protocols below that threshold did not reach statistical significance in the pooled analyses. The evidence supports evaluating acupuncture as a course of treatment, not a one-off rescue maneuver after a bad week of sleep.

There is an important exception—or at least an unresolved contradiction. Wang et al. (2024) ran a sham-controlled randomized clinical trial of a specific pairing, HT7 (Shenmen) and KI7 (Fuliu), over 10 sessions. The real-acupuncture group improved on both subjective sleep measures and objective PSG measures compared with sham, and the trial also reported improvements in anxiety and mood symptoms.

Because 10 sessions fall below the meta-analysis threshold, neither finding cancels the other. The pooled threshold may describe an average across varied protocols rather than a universal biological minimum. Alternatively, a small positive trial may not reproduce at larger scale. The current evidence cannot determine whether this acupoint pair needs fewer sessions or whether the dose-response pattern depends on treatment details.

The main reasons to stay skeptical

The Zhao meta-analysis is more rigorous about sleep measurement than a questionnaire-only review, but objective endpoints do not erase trial quality problems. The authors reported high risk of bias across the included studies, including frequent concerns about randomization, allocation concealment, and outcome reporting. Egger's test also detected publication bias, which means unpublished null trials may be missing from the estimate. The pooled effects may therefore be upper-bound estimates.

Blinding adds another problem. Participants may recognize real versus sham needling through sensation, practitioner behavior, or context. Expectations can affect arousal and sleep, including sleep measured by PSG. The meta-analysis did not provide a pooled assessment showing that blinding worked consistently.

Generalizability is also limited. Every included trial took place in Chinese clinical populations, where treatment expectations and the cultural role of acupuncture may differ from those in other settings. The studies measured sleep at the end of treatment or soon afterward; PSG-confirmed durability at 3, 6, or 12 months remains unknown.

A 2024 clinical review by Zhao et al. places acupuncture among options that may be useful when CBT-I is difficult to access, while retaining CBT-I as the recommended first-line treatment for chronic insomnia. That hierarchy is sensible because no primary-insomnia PSG trial in this evidence base directly compares acupuncture with CBT-I or with an approved hypnotic.

Other non-drug options belong in the decision

Acupuncture is not the only intervention with sleep data. Turmel et al. (2022) reported improved sleep quality, fatigue, anxiety, and depression after tailored yoga practice in chronic insomnia, although its design was a single-group pre-post study rather than a sham-controlled trial. Li et al. (2025) found that active high-definition transcranial direct current stimulation improved questionnaire sleep scores versus sham in chronic insomnia; exploratory PSG analyses found improvements in sleep-onset latency and sleep efficiency, without changes in sleep-stage ratios.

Exercise has PSG evidence in other sleep conditions, but population matters. Chen et al. (2024) assessed exercise programs in obstructive sleep apnea, not primary insomnia. Its results should not be converted into a head-to-head comparison with acupuncture for insomnia.

Post-stroke insomnia offers an interesting but separate signal. Zhang et al. (2025) randomized 93 people with post-ischemic stroke insomnia to Tiaowei Jiannao acupuncture, eszopiclone, or sham and reported improvements across several PSG measures, including N3 percentage. Stroke-related insomnia has a different cause and stroke severity can affect sleep outcomes, so those findings do not establish the same benefit in primary insomnia.

Verdict

Acupuncture has PSG-confirmed evidence for improving sleep continuity in primary insomnia: more total sleep, greater efficiency, less wake time after sleep onset, and fewer awakenings in a pooled analysis of 11 trials. The evidence supports considering a course of at least 12 sessions with a trained practitioner if acupuncture fits your preferences and access to CBT-I is limited.

It does not support treating acupuncture as a proven replacement for CBT-I, a shortcut to deeper N3 sleep, or a reliable fix for sleep-onset insomnia. The best current reading is narrower and more useful: acupuncture may help consolidate sleep, but the size, durability, and comparative value of that benefit still need better trials.