Magnetic resonance imaging as a potential surrogate for relapses in multiple sclerosis: a meta-analytic approach