Given the pressures on land to produce ever more food, doing it ‘sustainably’ is growing in importance. However, ‘sustainable agriculture’ is complex to define, not least because agriculture impacts in many different ways and it is not clear how different aspects of sustainability may be in synergy or trade off against each other. We conducted a meta-analysis to assess the relationships between multiple measures of sustainability using novel analytical methods, based around defining the efficiency frontier in the relationship between variables, as well as using correlation analysis. We define 20 grouped variables of agriculture's impact (e.g. on soil, greenhouse gas, water, biodiversity) and find evidence of both strong positive and negative correlations between them. Analysis based on the efficiency frontier suggests that trade-offs can be ‘softened’ by exploiting the natural between-study variation that arises from a combination of farming best practice and context. Nonetheless, the literature provides strong evidence of the relationship between yields and the negative externalities created by farming across a range of measures.