Idiopathic intracranial hypertension raises pressure inside the skull, causes severe headache and swells the optic nerve, and weight loss is one of the few things established to help it. A drug that produces large weight loss should therefore work, and a meta-analysis of nine studies covering 13,257 participants reports that it does [1]. The eye is somewhere these drugs have produced mixed signals before, as the optic nerve evidence by indication shows.
The clinical outcomes go the expected way. Headache risk came in at a risk ratio of 0.73 (95% CI 0.61–0.86), papilledema at 0.38 (95% CI 0.25–0.56), visual worsening at 0.51 (95% CI 0.37–0.68), and needing surgical intervention at 0.76 (95% CI 0.65–0.89). Mechanistically this is about as plausible as findings in this field get.
The reason to dwell on an implausible number rather than ignore it is that it calibrates the others. The mortality estimate and the papilledema estimate come from the same studies, comparing the same people. Whatever made the treated group so much less likely to die — being younger, healthier, better engaged with care, well enough to be offered an injection — was also making them less likely to lose vision. So the direction of the vision findings is probably right and the size of them is almost certainly inflated.
Nine studies, mostly observational, is what this evidence base consists of. The meta-analysis did exclude studies of patients with prior IIH treatment, which removes one source of confusion, but nothing randomized anyone to anything. The single randomized trial in this area is covered on this site’s sister publication rather than restated here.
For someone with this diagnosis the practical position is reasonable and the numbers are not. Weight loss helps intracranial pressure, these drugs produce weight loss, and a neuro-ophthalmologist weighing one is on defensible ground. Reading 0.38 as the papilledema benefit to expect is not defensible, for the same reason that a mortality hazard ratio the authors disowned and the dialysis cohort need their baseline tables read first — and the same reason the overdose comparison gave five different answers.