Disease-modifying drugs are changing the natural history of multiple sclerosis (MS). However, currently available clinical trial data are insufficient to develop accurate personalised treatment algorithms to assign the best possible treatment to each person with MS according to disease features, treatment history and comorbidities. Such accurate algorithms would require the presence of numerous head-to-head trials of long duration, which is virtually impossible, given the economic costs, required time and difficulties with attrition. Thus, efforts are being made to compare relative treatment efficacy through observational designs, using large multi-centre prospective cohorts or ‘big MS data’, and network meta-analyses. Although such studies can yield useful information, they are liable to biases and their results should be confirmed in other study populations, including smaller, single-centre cohorts, where some of these biases can be minimised. In this View paper we analyse the potential benefits and biases of all these strategies alternative to head-to-head trials in MS. Finally, we propose the combination of all these types of studies to obtain reliable head-to-head drug comparisons in the absence of randomised designs.