Mannitol is currently used to reduce intracranial pressure (ICP), but the evidence supporting its usefulness has been questioned. We aim to meta-analyze the effectiveness of mannitol in reducing ICP in adult patients with cerebral injuries and its dependency on baseline ICP values, comparing findings from individual patient data (IPD) and aggregated data (AD) meta-analysis performed on the same studies. We searched the Medline database, with no time limitation, through March 1, 2019. We selected studies for which IPD were available, with a before-after design, concerning adult patients with traumatic cerebral hemorrhages, subarachnoid hemorrhages, or hemorrhagic and ischemic stroke, treated with mannitol for increased intracranial hypertension. We extracted ICP values at baseline and at different time-points, and mannitol doses. We used a multilevel approach to account for multiple measurements on the same patient and for center variability. The AD meta-analysis and meta-regression were conducted using random-effects models. Three studies published IPD, and four authors shared their datasets. Two authors did not own their datasets anymore. Eight authors were unreachable, while 14 did not answer to our request. Overall, 7 studies provided IPD for 98 patients. The linear mixed-effects model showed that ICP decreased significantly after mannitol administration from an average baseline value of 22.1 mmHg to 16.8, 12.8, and 9.7 mmHg at 60, 120, and 180 min after mannitol administration. ICP reduction was proportional to baseline values with a 0.64 mmHg decrease for each unitary increment of the initial ICP value. Dose did not influence ICP reduction. The AD meta-analysis, based on data collected between 30 and 60 min from mannitol administration not accounting for multiple time-point measurements, overestimated ICP reduction (10 mmHg), while meta-regression provided similar results (0.66 mmHg decrease for each unitary increase of initial ICP). Mannitol is effective in reducing pathological ICP, proportionally to the degree of intracranial hypertension. IPD meta-analysis provided a more precise quantification of ICP variation than the AD approach.