Eighteen RCTs, describing 21 interventions, were included. Sample sizes ranged from 40 to 924 participants, where reported. Ten trials were rated as good quality, seven were fair, and one was poor.
Across all outcomes, workplace programmes were associated with a small statistically significant improvement (ES 0.24, 95% CI 0.14 to 0.34). Similar pooled effect sizes were found for individual outcomes (details in the paper), but the effect for work ability was not statistically significant.
Effect sizes were higher for trials of poor or fair quality than for those of good quality. Correct and clear description of randomisation, blinding of participants, use of intention-to-treat analysis, control for confounders, and longer follow-up were associated with lower effect sizes. The features associated with larger effect sizes were: younger participants (≤40 years), cluster randomisation, no intervention in the control group, low participation rates, high frequency of contact, and an educational component to the intervention.