Mannitol Use in Acute Stroke
Case Fatality at 30 Days and 1 Year
Background and Purpose— Mannitol is used worldwide to treat acute stroke, although its efficacy and safety have not been proven by randomized trials.
Methods— In a tricenter, prospective study, we analyzed the 30-day and 1-year case fatality with respect to mannitol treatment status in 805 patients consecutively admitted within 72 hours of stroke onset. Confounding factors were compared between treated and nontreated patients.
Results— Two thirds of the patients received intravenous mannitol as part of their routine treatment (mean dose, 47±22 g/d; mean duration, 6±3 days). The case fatality was 25% versus 16% (P=0.006) at 30 days and 38% versus 25% (P<0.001) at 1 year in the-mannitol treated and nontreated groups, respectively. Mannitol treatment effect was adjusted for age, stroke severity, fever in the first 3 days, and aspirin treatment (for ischemic strokes) in logistic regression models. Depending on the factors entered into the model, either no effect or harm could be attributed to mannitol. When the analysis was restricted to those admitted within 24 hours (n=568), case fatality differed significantly only at 1 year (35% in treated and 26% in nontreated patients, P=0.044). Although the prognostic scores of the Scandinavian Neurological Stroke Scale were similar in treated and nontreated patients, both in ischemic and hemorrhagic strokes, the patient groups differed in several factors that might also have influenced survival.
Conclusions— Based on the results of this study, no recommendations can be made on the use of mannitol in acute stroke, and properly randomized, controlled trials should be performed to come to a final conclusion.
The role of mannitol therapy in acute stroke is controversial. Because of its osmotic effect, mannitol is assumed to decrease cerebral edema. Mannitol might improve cerebral perfusion by decreasing viscosity, and as a free-radical scavenger, it might act as a neuroprotectant. Among the possible adverse effects, fluid and electrolyte imbalances, cardiopulmonary edema, and rebound cerebral edema have been listed. Mannitol can activate the process of apoptotic cell death and has the potential to activate the inflammatory mediators that aggravate the neuronal injury due to ischemia.1 In acute stroke, mannitol preferentially shrinks the nonaffected parts of the brain.2 In patients with cerebral edema after large, hemispheric infarction, acute mannitol use did not alter midline tissue shifts.3 Although mannitol decreases elevated intracranial pressure after stroke4 and results in increased cerebral perfusion pressure and brain oxygenation in large, hemispheric strokes,5 its overall effect on stroke outcome is unclear. Clinical observations could not prove the beneficial effects of mannitol in ischemic or hemorrhagic strokes in humans,6,7 and a systematic review of randomized trials could not draw a conclusion on the effects of mannitol in acute stroke.8 Despite the lack of evidence of an obvious benefit, mannitol has been used to treat human stroke for >30 years, and both American and European guidelines9,10 mention its use in certain clinical conditions of acute stroke. In an initial analysis of almost 1000 consecutively admitted stroke patients, we found that mannitol treatment was associated with a higher hospital case fatality, regardless of the age of the patient and the level of consciousness on admission.11 We could not decide whether this higher case fatality was caused by the effect of mannitol or by differences in prognostic factors other than age and disturbance of consciousness between the treated and nontreated patients. Therefore, in the present article, by considering several additional prognostic and confounding factors, we analyzed the case fatality at 30 days and 1 year after stroke in the same cohort in a prospective, observational study to see whether mannitol treatment results in obvious benefit or harm when administered to patients admitted within 72 hours of stroke onset. We further analyzed those admitted within 24 hours, and within this group, we investigated those with ischemic and hemorrhagic strokes.
The database of the Mures-Uzhgorod-Debrecen study12 was analyzed. Data of all patients consecutively admitted with acute cerebrovascular disease between October 1, 1999 and September 30, 2000 to the 3 centers were prospectively entered into a database. The database includes information on risk factors; patient condition on admission, including prognostic and long-term items of the Scandinavian Neurological Stroke Scale13 (SNSS); treatment on the ward; and condition at discharge. For discharge condition, outcome according to the International Stroke Trial14 was used, with a minor addition. Follow-up was performed at 30 days and 1 year after stroke by personal visits, postal questionnaires, or telephone calls to the patients, their relatives, or their family practitioners. From the current analysis, we excluded those with transient ischemic attack, cases of subarachnoid hemorrhage, and those who were admitted after 72 hours of stroke onset. The study was approved by the ethics committee of the University of Debrecen.
The database was originally designed for epidemiological and audit purposes and not to test treatment effects in a controlled fashion; therefore, some information important for the present analysis had to be additionally obtained. Such information was extracted from the patients’ documents retrospectively for this analysis.
The following factors were considered in the analysis: age; prestroke dependency; time to admission from stroke onset; diabetes; previous stroke, malignancy, and peripheral arterial disease in the history; smoking status; serum glucose level on admission; disturbance of the level of consciousness (LOC) on admission; the prognostic and long-term scores according to the SNSS (smaller SNSS scores indicate more severe strokes); white cell count in the first 3 days after admission; fever in the first 3 days after admission, defined as axillary temperature >37°C; fever at any time during the hospital stay; chronic obstructive pulmonary disease; atrial fibrillation during the hospital stay; antibiotic use, aspirin treatment, and heparin treatment during hospitalization; respirator use other than during attempts of resuscitation; and nasogastric tube feeding.
Continuous variables were compared by ANOVA. Stroke scale scores were compared by the Mann-Whitney test. The Pearson χ2 test was used to compare frequencies. Logistic regression models were used to evaluate whether 30-day and 1-year case fatalities depended on mannitol treatment status. In the models, survival was the dependent variable, and those factors that were found to be different by univariate analyses between the treated and nontreated groups were entered as continuous predictors or categorical factors. We used Statistica for Windows, version 6.1 (StatSoft) and the Proc Logistic procedure of SAS, version 8.02 (SAS Institute).
Characteristics of Patients
Of the acute stroke patients registered in the database, 805 were admitted within 3 days of stroke onset. Of these patients, 666 had ischemic stroke. The mean±SD age was 66±12.5 years; there were 471 men and 334 women. Time from stroke to admission was <24 hours in 568 patients; of these, 457 had ischemic and 111 had hemorrhagic strokes. Thirty-day and 1-year survival data were available for 782 of 805 (97%) and for 768 of 805 (95.4%) patients, respectively. The overall case fatality was 22.1% at 30 days and 33.6% at 1 year. Case fatality in the mannitol-treated subgroup was significantly higher both at 30 days (25% versus 16%, P=0.006) and at 1 year (38% versus 25%, P<0.001).
Characteristics of Mannitol Treatment
Based on local traditions, in 2 of the centers (Debrecen and Targu Mures), mannitol was frequently administered, whereas at the third center (Uzhgorod), mannitol was rarely used. The rate of mannitol treatment significantly differed among the 3 centers but did not differ between sexes. Mannitol was given according to the discretion of the treating physician, and when given, the mannitol solution was administered intravenously for 3 to 10 days. The mean dose of mannitol was 47±22 g/d, and the mean duration of mannitol treatment was 6±3 days. Mannitol treatment was initiated on the day of admission in 97%, and only 3% of the patients received mannitol for worsening of their condition. Except for mannitol, no other osmotically active medications (glycerol, hypertonic saline, urea, etc) were used by the centers.
Comparison of Mannitol-Treated And Nontreated Patients
Characteristics of the mannitol-treated and nontreated groups are given in Table 1. Mannitol-treated patients were older, stayed longer in hospital, and had somewhat lower SNSS prognostic and long-term scores than did those who were not treated with mannitol. More of the mannitol-treated patients were dependent before their current stroke. Artificial ventilation and nasogastric tube feeding were more frequent in mannitol-treated patients. Of the 666 patients with ischemic strokes, aspirin was given to 72% of the mannitol-treated and to 50% of the nontreated patients (P<0.001).
Patients Admitted <24 Hours After Stroke
Mannitol use was >70% in those admitted within 24 hours, 56% among those admitted between 24 and 48 hours, and 31% among those admitted between 48 and 72 hours; ie, with a longer delay to admission, the application rate of mannitol significantly decreased (P<0.0001). Therefore, we performed a separate analysis in the subgroup of those 568 patients who were admitted within the first 24 hours after stroke onset, and within this group, 2 further analyses were performed for those with ischemic and hemorrhagic strokes.
Characteristics of the patients admitted within 24 hours of stroke onset are given in Table 2. Mannitol use was the same in ischemic and hemorrhagic strokes. The proportion of prestroke dependency was 3.6% in nontreated and 13.5% in mannitol-treated patients (P<0.001). Fewer patients in the mannitol group were smokers. Disturbance of consciousness on admission was more frequent in the nontreated group, whereas fever within the first 72 hours, respirator use, and nasogastric feeding was more common in the mannitol-treated group. Although SNSS prognostic scores were similar, there was a tendency in the mannitol-treated group for a higher 30-day case fatality (23% versus 17%, P=0.13). The marginally lower long-term SNSS score was associated with higher 1-year case fatality in the mannitol group.
Outcome in All Patients
Odds ratios and their 95% confidence intervals for survival regarding mannitol treatment status, after adjustment for other variables, are given in Table 3. Without considering other factors, mannitol treatment was associated with significantly decreased odds of 30-day and 1-year survival (P=0.005 and P=0.0002, respectively). When mannitol treatment status was adjusted for age and the presence of disturbance of consciousness on admission, mannitol still seemed to have an adverse effect (P<0.0028 for 30-day and P<0.0017 for 1-year survival). When the prognostic score of the SNSS (ie, the sum score of the LOC, eye movements, and arm and leg strength items; score range, 0 to 22) was used in the model instead of the presence or absence of disturbance of consciousness, mannitol treatment did not have a significant effect on survival (P=0.1931 and P=0.1241 for 30-day and 1-year survival, respectively).
Case fatality at 1 year was 38% in the mannitol group and 25% in the nontreated group (P=0.0002). Age, presence of a disturbance of consciousness on admission, and mannitol treatment were significantly associated with 1-year case fatality (P<0.002 for all). In an extended model, age (P<0.001), SNSS long-term score (P<0.001), and fever in the first 72 hours (P=0.011) were significantly associated with case fatality, whereas mannitol treatment had no effect (P=0.8).
Outcome in Patients Admitted Within 24 Hours of Stroke Onset
In those admitted in the first 24 hours of stroke, case fatality was significantly associated with mannitol treatment at 1 year but not at 30 days (P=0.0388 and P=0.12, respectively; Table 3). When the effect of mannitol was adjusted for age and the presence of disturbed LOC on admission, mannitol had a significant adverse effect on both 30-day and 1-year survival (P=0.024 and P=0.0277, respectively). When the SNSS prognostic score was used instead of the LOC, the effect of mannitol became nonsignificant.
Ischemic Stroke Patients Admitted Within 24 Hours
When the analysis was further restricted to those 457 patients who had ischemic stroke and were admitted within 24 hours, mannitol treatment status was not associated with 30-day case fatality (15% in treated and 12.7% in nontreated groups, P=0.51). SNSS prognostic score and case fatality at 30 days and 1 year are shown in Figure 1. The SNSS prognostic score was 16.3±5.6 and 16.8±5.4 in the treated (n=315) and nontreated (n=142) patients, respectively (P=0.21). There was no difference in the frequency of patients with disturbed LOC on admission (P=0.08). Respirator use was similar, whereas nasogastric tube feeding was more frequent in the mannitol group. SNSS long-term score was 26.8±14.7 and 29.9±15.3 in the treated and nontreated groups, respectively (P=0.01). Case fatality at 1 year was 27.7% in treated and 22.9% in nontreated patients (P=0.28). Although mannitol treatment had the tendency to increase the chances for survival, when treatment effect was adjusted for age, SNSS long-term score, fever in the first 3 days, and aspirin treatment in the acute phase (odds ratio, 1.869; Table 3), the confidence intervals were wide and included the possibility of harm.
Cerebral Hemorrhage Patients Admitted Within 24 Hours
Of the 111 patients with cerebral hemorrhages, 84 were treated and 27 were not treated with mannitol. Although treated patients were older than nontreated patients (65.0±12 and 56.3±12.3 years, respectively; P=0.002), they had similar scores on both the prognostic and long-term items of the SNSS (10.5±6.5 versus 10.1±6.4, P=0.82, and 15.0±14.1 versus 13.6±13.4, P=0.74, in treated and nontreated patients, respectively) and had similar white cell counts and glucose levels on admission than nontreated patients. Disturbance of the LOC on admission was more frequent in the nontreated group (P=0.02). Treated and nontreated patients did not differ significantly regarding the frequency of prestroke dependency, chronic obstructive pulmonary disease, malignancy, fever in first 72 hours, atrial fibrillation, antibiotic use, respirator use, and nasogastric tube feeding. Case fatality was not significantly higher in the treated group at 30 days and 1 year (52% versus 41%, P=0.31, and 62% versus 44%, P=0.12; Figure 2). Although the odds ratios for survival were <0.6 in all models (Table 3), suggesting an adverse effect of mannitol treatment, the 95% confidence intervals were wide and included the possibility of a beneficial effect.
Although the results of observational studies on treatment effects and on case fatality should be very cautiously interpreted because of the potential of large biases,15 the results were surprising to us. In contrast to the expected favorable effect of mannitol, we could not find any association between mannitol use and better prognosis at 30 days and 1 year after stroke. Depending on the factors included in the logistic regression models, mannitol either did not have a significant effect on case fatality or was associated with an adverse outcome. Of the neurologic signs on admission, in the analysis we first used only the LOC, because among the clinical signs, this was found to be the most significant prognostic factor in ischemic and hemorrhagic strokes.16,17 When disturbed LOC and age were used as confounding factors in the model, mannitol had a significant negative effect. However, when LOC was changed to a more complex, quasi-continuous index of neurologic damage, ie, the prognostic score of the SNSS including LOC, eye movements, and severity of paresis of the affected upper and lower extremities, the effect of mannitol became nonsignificant. When the analysis was restricted to those who were admitted within 24 hours, the findings were similar.
The fact that mannitol use was associated with higher short-term and long-term case fatality in the total group might be partly or totally explained by the differences between treated and nontreated patients in prognostic factors. The absolute difference in SNSS prognostic scores between treated and nontreated patients was small, though statistically significant, in the total group, but was not significant in the subgroup comparisons. Most prognostic and confounding factors did not differ significantly between treated and nontreated patients admitted within 24 hours with ischemic stroke, and several factors were even more favorable for treated patients in the hemorrhagic subgroup.
This analysis has several limitations. First, this is a prospective, observational study and not a randomized, controlled trial; therefore, selection bias could have affected the results. Second, computed tomography (CT) was performed in only 73% of the patients, and in 10% of the patients, no lesion was detected. In these patients, no repeated scans were performed. Therefore, it was not possible to perform an analysis by the volume of the lesions. It has been reported that although a visible infarct on the CT scan is associated with adverse prognosis, at least 30% of patients with ischemic strokes have normal CT scans.18 Third, there were missing data for some of the patients for several parameters; eg, white cell count was present for >95% of patients, but data for glucose level on admission were missing in close to 20% of patients. Fourth, mannitol was administered according to the discretion of the treating neurologist, and the dose of mannitol used and the duration of treatment varied.
We attempted to analyze the effect of mannitol on case fatality by considering as many confounding factors as possible. It was not possible to enter into the model all factors that possibly influence short- and long-term survival. Depending on the number and type of prognostic factors used in the multivariate analysis, mannitol had either a nonsignificant or an adverse effect. This observational study does not prove that mannitol is harmful if given for acute stroke, but it raises concerns and emphasizes the need for properly designed, randomized, clinical trials to decide whether the practice of routine mannitol use in patients with acute stroke is justified, should be restricted to subgroups, or should be stopped altogether.
The work was supported by grants No. ETT 490/2000, ETT 578/2000, and RO-8/2000. The authors thank János Steiber (Biostatistical Group, Gedeon Richter Ltd, Hungary) for calculating the odds ratios and their confidence intervals for Table 3.
- Received November 9, 2002.
- Revision received February 27, 2003.
- Accepted March 7, 2003.
Videen TO, Zazulia AR, Manno EM, Derdeyn CP, Adams RE, Diringer MN, Powers WJ. Mannitol bolus preferentially shrinks non-infarcted brain in patients with ischemic stroke. Neurology. 2001; 57: 2120–2122.
Mano EM, Adams RE, Derdeyn CP, Powers WJ, Diringer MN. The effects of mannitol on cerebral edema after large hemispheric cerebral infarct. Neurology. 1999; 52: 583–587.
Schwarz S, Schwab S, Bertram M, Aschoff A, Hacke W. Effects of hypertonic saline hydroxyethyl starch solution and mannitol in patients with increased intracranial pressure after stroke. Stroke. 1998; 29: 1550–1555.
Steiner T, Pilz J, Schellinger P, Wirtz R, Friederichs V, Aschoff A, Hacke W. Multimodal online monitoring in middle cerebral artery territory stroke. Stroke. 2001; 32: 2500–2506.
Candelise L, Colombo A, Spinnler H. Therapy against brain swelling in stroke patients: a retrospective clinical study on 227 patients. Stroke. 1975; 6: 353–356.
Santambrogio S, Martinotti R, Sardella F, Porro F, Randazzo A. Is there a real treatment for stroke? clinical and statistical comparison of different treatments in 300 patients. Stroke. 1978; 9: 130–132.
Bereczki D, Liu M, do Prado GF, Fekete I. Cochrane report: a systematic review of mannitol therapy for acute ischemic stroke and cerebral parenchymal hemorrhage. Stroke. 2000; 31: 2719–2722.
Adams HP, Brott TG, Crowell RM, et al. Guidelines for the management of patients with acute ischemic stroke: a statement for healthcare professionals from a special writing group of the Stroke Council, American Heart Association. Stroke. 1994; 25: 1901–1914.
Bereczki D, Mihálka L, Szatmári S, Di Cesar D, Fülesdi B, Csiba L, Fekete I. Mannitol use and outcome in acute stroke: results from the Mures-Uzhgorod-Debrecen study. Eur J Neurol. 2002; 9 (suppl 2): 173.Abstract.
Scandinavian Stroke Study Group. Multicenter trial of hemodilution in ischemic stroke. Stroke. 1985; 16: 885–890.
Howard G, Walker MD, Becker C, Coull B, Feibel J, McLeroy K, Toole JF, Yatsu F. Community hospital-based stroke programs: North Carolina, Oregon, and New York, III: factors influencing survival after stroke: proportional hazards analysis of 4219 patients. Stroke. 1986; 17: 294–299.
Lampl Y, Gilad R, Eshel Y, Sarova-Pinhas I. Neurological and functional outcome in patients with supratentorial hemorrhages: a prospective study. Stroke. 1995; 26: 2249–2253.
Wardlaw JM, Lewis SC, Dennis MS, Counsell C, McDowall M. Is visible infarction on computed tomography associated with an adverse prognosis in acute ischemic stroke? Stroke. 1998; 29: 1315–1319.