183 trials ranked the ways to feel better. Almost everything worked about the same
Mindfulness, yoga, exercise and positive psychology landed within a narrow band of each other. Only combining exercise with a psychological approach pulled clear, and nature-based programmes did not beat controls.
The market for feeling better is enormous and almost entirely unranked. Mindfulness apps, gratitude journals, yoga studios, forest bathing, running clubs and acceptance and commitment therapy all promise improved well-being, and until now there has been no synthesis putting them in the same statistical room.
A preregistered network meta-analysis published in Nature Human Behaviour on 2 January does that [s1]. Its central finding is not that one approach wins. It is that most of them arrive at roughly the same place.
What was pooled
The authors searched MEDLINE, PsycINFO, CENTRAL and Scopus to March 2023 for randomised controlled trials of well-being interventions in adults without diagnosed conditions [s1]. That last restriction matters: this is evidence about the general population, not about treating depression or anxiety.
They found 183 trials covering 22,811 participants [s1]. The interventions included mindfulness-based, compassion-based, acceptance and commitment therapy and positive psychology approaches, alongside exercise, yoga, educational programmes, nature-based programmes and combined exercise-psychological approaches [s1]. Risk of bias was assessed with RoB 2 and the data synthesised using random-effects network meta-analysis [s1].
The ranking, such as it is
Most interventions improved well-being compared with inactive controls [s1].
Combined exercise-psychological interventions produced the largest effect, a standardised mean difference of 0.73 (95% CI 0.27 to 1.20) [s1]. Everything else clustered: mindfulness, compassion, single positive psychology interventions, yoga and exercise all landed between 0.41 and 0.49, with no significant differences between them [s1].
Nature-based interventions were not significantly more effective than controls, though the authors note the evidence there was limited by conceptual and methodological heterogeneity — "nature-based" covers a wide and inconsistently defined range of things [s1].
A standardised mean difference around 0.45 is a moderate effect. The honest summary is that if you are going to do something structured for your well-being, the evidence does not favour one option over another within that cluster, and the one thing that did stand out combined movement with a psychological component rather than choosing between them.
The caveats are load-bearing
The authors are direct about the quality problem. Risk of bias was frequently moderate to high across the included trials [s1]. Funnel plot asymmetry suggested potential publication bias — the well-known pattern where trials that found nothing are less likely to be published [s1].
They ran multiple sensitivity analyses to test whether the conclusions survived: including grey literature, excluding studies at high risk of bias, and excluding small studies [s1]. The overall conclusions held [s1]. Using the CINeMA framework, 71% of comparisons were rated moderate in certainty of evidence [s1]. That leaves 29% below moderate.
The search also closed in March 2023, so trials published since are not in it. And no external funding was received for the work [s1].
Where the digital versions sit
A separate systematic review and meta-analysis published in PLOS Digital Health on 15 January looked at one branch the network analysis could not resolve: immersive virtual reality interventions, in this case for older adults [s2].
Ten randomised controlled trials covering 746 older adults were included, searched across six databases from 2010 to December 2024 [s2]. The interventions were mostly reminiscence (40%) and garden or forest therapy (40%), with cognitive stimulation and multi-sensory stimulation making up the rest [s2]. Pooled results showed reduced depressive symptoms (SMD −0.83, 95% CI −1.05 to −0.60, I² = 21%) and reduced anxiety (SMD −0.77, 95% CI −1.32 to −0.22, I² = 70%) [s2].
The anxiety estimate carries substantial heterogeneity, and for stress and affect the authors could not pool at all, reporting a synthesis without meta-analysis instead [s2]. Ten trials and 746 participants is a thin base for a conclusion about a technology, and the population — older adults — is narrower than the Nature Human Behaviour review's.
It is worth noticing the tension. Nature-based interventions did not beat controls in the larger review [s1]; virtual garden and forest therapy sits among the interventions that did show effects in the smaller one [s2]. That could reflect the different populations, the different comparators, or the small-study effects both papers warn about.
What this is not
None of this is advice about what any individual should do. These are population-average effects in people without diagnosed conditions, drawn from a literature the authors themselves rate as frequently at moderate to high risk of bias [s1]. A moderate average effect is compatible with a large benefit for some people and none for others, and neither review can identify which is which.
What to watch
Whether preregistration and larger trials narrow the confidence intervals in the next generation of well-being research, and whether "nature-based intervention" acquires a definition specific enough to be tested.
Sources
- [s1] A systematic review and network meta-analysis of randomized controlled trials of well-being-focused interventions. Nature Human Behaviour, 2 January 2026. https://doi.org/10.1038/s41562-025-02369-1
- [s2] Immersive virtual reality-based intervention for psychological wellbeing among older adults: A systematic review and meta-analysis. PLOS Digital Health, 15 January 2026. https://doi.org/10.1371/journal.pdig.0001110
Sources
- A systematic review and network meta-analysis of randomized controlled trials of well-being-focused interventions — Nature Human Behaviour , January 2, 2026
- Immersive virtual reality-based intervention for psychological wellbeing among older adults: A systematic review and meta-analysis — PLOS Digital Health , January 15, 2026
Air pollution and lower life satisfaction: a twin study asks whether it's really the air
Analysing 4,504 German twins, researchers found the pollution–wellbeing link ran mostly between families, not within twin pairs — pointing to shared circumstances as much as the air itself.
On the evidence so far, working from home has no clear effect on health, good or bad
A Swiss study tracking people who switched to remote work found no measurable change across 13 health outcomes. A review of pandemic homeworking found the relationship varies so much that no consensus emerged.
'Cycle syncing' your workouts has little evidence behind it for ordinary women
A meta-analysis of 78 studies found the menstrual cycle has at most a trivial effect on performance. A randomised trial found timing strength training to cycle phase built no more muscle than training consistently.
Greener streets, slightly happier people, and a bigger gain for the poorest
A UK survey of 26,322 urban adults finds a small but consistent life-satisfaction gain from neighbourhood greenery. A Stockholm study of 46,282 shows who ends up living without it.