Summary Background Invasive community-onset staphylococcal disease has emerged worldwide associated with Panton-Valentine leucocidin (PVL) toxin. Whether PVL is pathogenic or an epidemiological marker is unclear. We investigate the role of PVL in disease, colonisation, and clinical outcome. Methods We searched Medline and Embase for original research reporting the prevalence of PVL genes among Staphylococcus aureus pneumonia, bacteraemia, musculoskeletal infection, skin and soft-tissue infection, or colonisation published before Oct 1, 2011. We calculated odds ratios (ORs) to compare patients with PVL-positive colonisation and each infection relative to the odds of PVL-positive skin and soft-tissue infection. We did meta-analyses to estimate odds of infection or colonisation with a PVL-positive strain with fixed-effects or random-effects models, depending on the results of tests for heterogeneity. Results Of 509 articles identified by our search strategy, 76 studies from 31 countries met our inclusion criteria. PVL strains are strongly associated with skin and soft-tissue infections, but are comparatively rare in pneumonia (OR 0·37, 95% CI 0·22–0·63), musculoskeletal infections (0·44, 0·19–0·99), bacteraemias (0·10, 0·06–0·18), and colonising strains (0·07, 0·01–0·31). PVL-positive skin and soft-tissue infections are more likely to be treated surgically than are PVL-negative infections, and children with PVL-positive musculoskeletal disease might have increased morbidity. For other forms of disease we identified no evidence that PVL affects outcome. Interpretation PVL genes are consistently associated with skin and soft-tissue infections and are comparatively rare in invasive disease. This finding challenges the view that PVL mainly causes invasive disease with poor prognosis. Population-based studies are needed to define the role of PVL in mild, moderate, and severe disease and to inform control strategies. Funding None.