The study, led by Zihang Tong and colleagues at the First Teaching Hospital of Tianjin University of Traditional Chinese Medicine, pooled evidence from 96 randomized controlled trials encompassing 18,419 adult participants. The research team searched PubMed, the Cochrane Library, Embase, and Web of Science through October 2025, then used network meta-analysis — a statistical framework that combines both direct comparisons between treatments and indirect comparisons linked through a common control — to rank six digital interventions: digital cognitive behavioral therapy (dCBT), digital mindfulness-based therapy (dMBT), digital brief behavioral therapy (dBBT), circadian rhythm support (CRS), virtual reality (VR), and tele-neurofeedback (NFB). By folding every trial into a single evidence network anchored on a shared control node, the method allows treatments never tested against each other in the same trial to be compared with remarkable statistical efficiency.

The headline finding concerns the Insomnia Severity Index, or ISI, a seven-item questionnaire that scores insomnia symptoms from 0 to 28. Compared with control treatment, dCBT reduced ISI scores by an average of 4.24 points, with a 95 percent confidence interval of −4.83 to −3.65 — an estimate supported by moderate-certainty evidence under the CINeMA framework. That figure is not just statistically significant; it crosses the minimal clinically important difference, the threshold — set at 4 points for the ISI in this analysis — beyond which patients and clinicians should actually notice a change. In other words, dCBT is the only intervention in the network whose effect on the core outcome of insomnia severity was unambiguously meaningful for real patients.

The detailed picture across other outcomes is more nuanced. dCBT also significantly improved PSQI scores (a reduction of 2.28 points), shortened subjective sleep onset latency by about 12.6 minutes, raised subjective sleep efficiency by 7.27 percentage points, and cut subjective wake after sleep onset by roughly 17.2 minutes. Yet none of these secondary effects reached their respective minimal clinically important differences, which the authors defined as, for example, at least a 3-point PSQI change or a 20-minute reduction in latency. Digital mindfulness-based therapy produced a smaller but significant ISI reduction of 2.28 points and shortened objectively measured sleep onset latency by 7.96 minutes, the latter based on moderate-certainty evidence from polysomnography and actigraphy studies. Virtual reality, evaluated through an inconsistency model because of statistical heterogeneity, reduced objectively measured wake after sleep onset by 13.38 minutes — significant, but again below the clinical threshold.

Direct head-to-head comparisons within the network reinforced dCBT’s dominance. It outperformed dMBT on the ISI (by 1.96 points) and on subjective total sleep time (by about 10 minutes), beat VR on the PSQI, and surpassed circadian rhythm support on sleep latency and sleep efficiency measures. Using SUCRA statistics, which express the probability that a treatment ranks best, dCBT topped the charts for the ISI (93.3 percent), the PSQI (95.8 percent), subjective sleep onset latency (95.0 percent), and subjective sleep efficiency (93.0 percent). Three interventions — tele-neurofeedback, digital brief behavioral therapy, and circadian rhythm support — showed no statistically significant benefits over control for any outcome, effectively dropping out of the clinical conversation until better evidence emerges.

The authors went well beyond the headline averages. Subgroup analyses probed whether effects differed by control group type, therapist guidance, treatment duration, and population. Notably, dCBT performed as well as face-to-face CBT where such comparisons existed, held up against active controls rather than only passive waitlists, and worked whether delivered with a therapist in the loop or fully automated — the ISI effect was −4.48 points with guidance versus −4.10 points without, a gap of less than half a point. Treatment duration mattered little for the core outcome: the ISI effect was identical (−4.24 points) whether programs ran eight weeks or longer. Effects were somewhat larger in people with insomnia symptoms who lack a formal diagnosis than in diagnosed patients, suggesting those with less entrenched sleep pathology may respond more readily to structured digital programs.

Sensitivity analyses added important caveats about who benefits. dCBT’s effects proved stable across pregnant women, cancer patients, and people with depression or anxiety, indicating broad applicability. dMBT’s efficacy, by contrast, appeared inflated in pregnant women under certain control designs and weaker in cancer survivors, and its improvement on the PSQI was fragile among depressed patients. VR was the most volatile intervention of all: its estimated effects swung dramatically depending on which small studies were included, leading the authors to urge caution about deploying VR in oncology, acute medical settings, and populations with severe somatic disease.

The study is candid about its limitations, and readers should absorb them. Only two of the 96 trials were judged low risk of bias overall, with 94 showing some concerns — especially around outcome measurement, where participants inevitably know what treatment they are receiving, inflating expectancy effects on subjective questionnaires. Heterogeneity was high for the ISI and PSQI, objective sleep outcomes rested on few trials with small samples, roughly 40 trials had potential industry funding, and the analysis captured only immediate post-treatment results with no follow-up data on whether benefits persist. The treatment taxonomy also followed a Chinese expert consensus rather than international regulatory frameworks, which may limit direct comparability with other classification schemes.

Still, the practical implications are hard to escape. For clinicians building insomnia care pathways, the evidence reasonably supports dCBT as the primary digital option, given its clinically meaningful reduction in insomnia severity. dMBT emerges as a reasonable alternative or adjunct — particularly for patients who prefer mindfulness approaches or struggle to adhere to CBT — even though its benefits did not reach clinical significance thresholds. The absence of patient and public involvement in the analysis, the authors acknowledge, means real-world preferences around acceptability remain underexplored. What the study delivers is something prior reviews could not: a single network in which six different digital therapies, from smartphone-delivered CBT to immersive VR, compete on equal statistical footing to answer the question patients actually ask — which one is best.

The research team, funded by Tianjin science and health programs and registered prospectively on PROSPERO, concludes that future work should prioritize large-scale, high-quality randomized trials with standardized interventions, objective outcome frameworks, and long-term follow-up. Until then, this analysis stands as the most complete map of the digital sleep-therapy landscape — one that confirms the promise of treatment delivered by app, but also warns that statistical significance alone is not the same as a patient sleeping meaningfully better. For the 16.2 percent of adults worldwide reporting clinically significant insomnia, the difference between those two thresholds is exactly the kind of precision that evidence-based medicine is meant to provide.

Subject of Research: Comparative effectiveness of digital therapeutics for improving sleep quality in adults with insomnia

Article Title: The effectiveness of digital therapeutics in improving sleep quality: a systematic review and network meta-analysis

Article References: Tong, Z., Ye, G., Fan, S., Li, Y., Zhang, J., Zhang, H., Chen, Q., Wang, H., Li, H., & Wang, J. (2026). The effectiveness of digital therapeutics in improving sleep quality: a systematic review and network meta-analysis. BMC Medicine, 24(1), Article 526. https://doi.org/10.1186/s12916-026-05129-8

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05129-8

Keywords: digital therapeutics, insomnia, sleep quality, network meta-analysis, digital cognitive behavioral therapy, mindfulness, virtual reality, tele-neurofeedback, circadian rhythm support, randomized controlled trials, PSQI, Insomnia Severity Index