Touch is crucial for mental and physical well-being, and its reduction during events like the COVID-19 pandemic highlighted the need for further research. While many studies show benefits of touch interventions (e.g., massage, kangaroo care), there's large variability in studied cohorts, touch types, outcomes measured, and the toucher's identity. Previous meta-analyses had limited scope, failing to identify key moderators of efficacy. This study aimed to address this gap using a large-scale systematic review and multivariate meta-analysis to quantitatively assess (1) the effect of touch interventions on physical and mental health and (2) the influence of moderators such as the toucher's identity (human vs. object/robot, familiar vs. unfamiliar), demographics (clinical status, age, sex), delivery method (type of touch, body part), and procedure (duration, number of sessions). The study focused on humans but included animal studies in the systematic review to explore potential mechanisms. The large number of studies included allowed for detailed subgroup analyses to investigate these moderators' impact across varied health outcomes in adults/children and newborns separately.

Prior meta-analyses on touch interventions focused on specific touch types, cohorts, or health outcomes, limiting their scope and preventing identification of significant moderators. This study aimed to address these limitations by using a comprehensive approach that encompasses a wider range of interventions, cohorts, and outcomes to determine the specific factors influencing efficacy. Although previous research indicated broad benefits of touch across various groups and outcomes, the need to identify crucial moderators was crucial in order to tailor interventions and focus research to maximize well-being.

This study employed a pre-registered (PROSPERO: CRD42022304281) systematic review and multilevel, multivariate meta-analysis. Databases (Google Scholar, PubMed, Web of Science) were searched until October 1, 2022. Studies included a touch vs. no-touch control intervention with diverse health outcomes. Risk of bias was assessed using small study effects, randomization, sequencing, performance, and attrition bias. Data from adults (n = 2841 touch, 2556 control, 85 studies, 103 cohorts) and newborns (n = 2134 touch, 2086 control, 52 studies, 63 cohorts) were analyzed separately using R's 'rma.mv' function from the 'metafor' package. Multilevel models accounted for the hierarchical structure (study, cohort, effect). The variance-covariance matrix addressed dependencies within cohorts and studies. Robust variance estimation with cluster-robust inference at the cohort level was employed. Power sensitivity analysis ensured reliable effect detection. Subgroup analyses investigated moderators, only proceeding if sufficient power (≥80%) was achieved. Effect sizes were calculated as Hedges' g. Heterogeneity was assessed using Cochran's Q. Small-study bias was assessed visually (funnel plots) and statistically. Omnibus F-tests determined moderator significance, while t-tests compared effect sizes between moderator levels. Animal studies (19 studies, n = 911), primarily rodents, were included in the systematic review but excluded from the meta-analysis due to insufficient data. Risk of bias was assessed considering randomization, sequence bias, performance bias (blinding impossible), and attrition bias.

Overall, touch interventions had a medium-sized effect (adults: Hedges' g = 0.52, 95% CI [0.42, 0.63]; newborns: Hedges' g = 0.56, 95% CI [0.41, 0.71]). Benefits were similar for mental and physical outcomes. For adults, benefits were highest for pain, anxiety, and depression, and lower for respiratory, sleep, and heart rate parameters. In newborns, positive effects were found for cortisol, liver enzymes, respiration, temperature regulation, and weight gain. Human-human touch showed medium-sized benefits overall; however, object/robot touch showed similar physical benefits but significantly lower mental health improvements than human touch. Sub-analysis suggested skin-to-skin contact might mediate the higher mental health effects in human-human interactions. No significant differences were found between different touch types (massage, kangaroo care) or between familiar and unfamiliar touchers in adults. However, parental touch proved significantly more beneficial for newborns. Increasing the number of touch sessions increased benefits in adults (especially for trait anxiety, depression, and pain), but increasing session duration did not yield significant improvement and showed indications of negative correlations with cortisol and heart rate reduction in adults. No significant effects were found relating to sex ratio or age as moderators. Head touch showed significantly higher benefits than arm or torso touch. Unidirectional touch demonstrated higher health benefits than bidirectional touch, particularly in mental health domains. Study location showed that South American cohorts showed significantly stronger effects than North American or European cohorts for adults while North American cohorts for newborns showed weaker effects than Asian or European cohorts. Finally, a systematic review of studies not included in the meta-analysis due to missing effect sizes is available via the OSF project.

This study confirms the widespread benefits of touch interventions across various health outcomes, ages, and clinical statuses. The medium effect size is robust across a vast dataset, surpassing limitations of previous meta-analyses by including multiple dependent variables and using robust statistical methods like multilevel and multivariate models. While the type of touch seems less crucial, session frequency positively impacts outcomes, especially for adults, indicating a need for repeated sessions. Conversely, session duration doesn't improve benefits, and longer durations may even have negative impacts on some outcomes (cortisol, blood pressure in adults). Human-human touch is more effective than object/robot touch, specifically for mental health, highlighting the importance of skin-to-skin contact. Parental touch significantly benefits newborns. These findings have major implications for tailoring interventions and future research, which should focus on optimizing session frequency, considering skin-to-skin contact, and exploring how individual differences (sex, culture, relationship) influence touch's effect.

Touch interventions demonstrate significant benefits across numerous physical and mental health outcomes. Session frequency, but not duration, strongly impacts effectiveness, particularly for anxiety and depression. Human touch surpasses object/robot touch, especially for mental health. Future research should investigate factors such as skin-to-skin contact, individual differences, and long-term effects to further refine touch-based interventions for optimal outcomes. The study's animal data suggests promising avenues for translational research to understand the underlying mechanisms and optimize touch therapy.

Small-study bias and the inability to blind participants and researchers to the experimental condition could influence the results. The predominantly Western-focused language selection might have excluded relevant studies. The reliance on immediately measured outcomes might not fully capture long-term effects. The large number of post-hoc tests increases the risk of Type I errors. The lack of data points for some moderator levels limited the scope of detailed analysis.