Association of body mass index with mortality and functional outcome after acute ischemic stroke

The relation between obesity and stroke outcome has been disputed. This study was aimed to determine the association of body mass index (BMI) with mortality and functional outcome in patients with acute ischemic stroke. Data were from a national, multi-centre, prospective, hospital-based register: the ChinaQUEST (Quality Evaluation of Stroke Care and Treatment) study. Of 4782 acute ischemic stroke patients, 282 were underweight (BMI < 18.5 kg/m2), 2306 were normal-weight (BMI 18.5 to < 24 kg/m2), 1677 were overweight (BMI 24 to <28 kg/m2) and 517 were obese (BMI ≥ 28 kg/m2). The risks of death at 12 months and death or high dependency at 3 and 12 months in overweight (HR: 0.97, 95% CI: 0.78–1.20; OR: 0.93, 95% CI: 0.80–1.09; OR: 0.95, 95% CI: 0.81–1.12) and obese patients (HR: 1.07, 95% CI: 0.78–1.48; OR: 0.96, 95% CI: 0.75–1.22; OR: 1.06, 95% CI: 0.83–1.35) did not differ from normal-weight patients significantly after adjusting for baseline characteristics. Underweight patients had significantly increased risks of these three outcomes. In ischemic stroke patients, being overweight or obese was not associated with decreased mortality or better functional recovery but being underweight predicted unfavourable outcomes.

and conduct of the ChinaQUEST has been described previously 21,22 . In brief, consecutive patients (aged ≥ 15 years) admitted to a hospital within 30 days from onset of acute stroke due to ischemic stroke or intracerebral hemorrhage were included during a 5-month period in 2006 22 . Patients with subarachnoid hemorrhage were excluded.
Patients were assessed 4 times during the course of the study, at admission, discharge, and approximately 3 and 12 months post-stroke 22 . Baseline information such as socio-demographic variables, medical history and clinical features of the index stroke, was obtained predominantly by face-to-face interviews. In-hospital details on diagnostic and treatment strategies were obtained through medical records (46%) and interviews with patients or their proxies (54%). Follow-up information about functional outcome as measured by the modified Rankin scale (mRS) was collected mainly by telephone interview (>80%) at 3 and 12 months. Death and cause of death were ascertained by family members (64.7%), a health professional (18.4%), records (e.g. medical, police [11.8%]), death certificate (3.5%) or other sources (1.5%).
Written informed consent was obtained from all patients or an appropriate family member (if the patient was unable to give it) to participate. The study was approved by the ethics committees of Peking University First Hospital (Beijing), Ruijin Hospital (Shanghai), Prince of Wales Hospital (Hong Kong), and The University of Sydney. Good Clinical Practice guidelines in accordance with the Declaration of Helsinki were used and the privacy of patients was strictly protected.
Variables of interest and outcomes. BMI was calculated from height and weight (kilograms per square metre) measured by nurses or obtained from the patient or the accompanying relatives on admission. The patient population was sub-grouped into the following categories according to the Chinese Guidelines on Prevention and Control of Obesity in Adults 23,24 : underweight (BMI < 18.5 kg/m 2 ), normal weight (BMI 18.5 to <24 kg/m 2 ), overweight (BMI 24 to <28 kg/m 2 ) and obese (BMI ≥ 28 kg/m 2 ).
The outcomes included all-cause mortality at 12 months, and a composite of death or high dependency at 3 and 12 months post-stroke. High dependency was defined as a mRS score of 3-5.
Covariates considered included socio-demographic variables, medical history and clinical features of the index stroke, including: age, sex, partnership (living alone or living with a partner), education level ('low level' , being illiterate or having only primary education; 'high level' , having secondary education or higher) and annual household income (<20000 Chinese Yuan [CNY], approximately $2857 US; ≥20000 CNY; or declined to respond/unknown); medical history of hypertension, diabetes, hyperlipidaemia, atrial fibrillation, prior stroke/ TIA, coronary artery disease (any history of heart attack and/or myocardial infarction, angina or coronary heart disease) either self-reported or diagnosed in hospital post-stroke; prior dependency, current cigarette smoking, and regular alcohol consumption within the 3 months before stroke onset; time of presentation at the hospital (<6 hour; ≥6 hour; or unknown), Glasgow coma scale (GCS) score on admission ('mild' 13-15; 'moderate' 8-12;  Statistical analysis. Baseline characteristics were compared among different BMI subgroups using the one-way analysis of variance (ANOVA) and Chi-square test for continuous and for categorical variables as appropriate. The Cox proportional hazard model was performed to assess the association between BMI and all-cause mortality at 12 months. Similarly, logistic regression models were performed to evaluate the association between BMI and the composite outcome of death or high dependency at 3 and 12 months post-stroke. The association of BMI and high dependency was also evaluated in the survivors at 3 and 12 months after stroke. The distribution of causes of death in those who had died during the follow-up period was compared with Chi-square test among the BMI subgroups. We performed four additional analyses to assess the robustness of our main findings. First, we performed a sensitivity analysis using the World Health Organization (WHO) BMI classification criteria 25 , which defined underweight (BMI < 18.5 kg/m 2 ), normal weight (BMI 18.5 to <25 kg/m 2 ), overweight (BMI 25 to <30 kg/m 2 ) and obese (BMI ≥ 30 kg/m 2 ). Second, we restricted the study population to those who had not received thrombolysis treatment. Third, another definition of high dependency as a mRS score of 4-5 was used to validate the association between BMI and functional outcome 15 . Fourth, we performed a subgroup analysis according to smoking status (current smoker and current non-smoker). All P values were two-sided, with P < 0.05 considered statistically significant. All statistical analyses were performed using SPSS version 13.0 (SPSS Inc., Chicago, IL, USA).

Results
Baseline characteristics. There were 6508 acute stroke patients enrolled in the ChinaQUEST study. After excluding 1572 patients with intracerebral hemorrhage and 154 with uncertain diagnosis and missing data, a total of 4782 patients with acute ischemic stroke were included in our analysis. The mean age of those with ischemic stroke was 64.6 years (range 15-102 years) and 37.9% were female. Their average BMI was 23.8 ± 3.5 kg/m 2 . Of them, 282 patients (5.9%) were underweight, 2306 (48.2%) had normal weight, 1677 (35.1%) were overweight, and 517 (10.8%) were obese.
Baseline characteristics between the BMI subgroups are shown in Table 1. Underweight patients were older and more likely to have had little education, a low income, atrial fibrillation, prior dependency and low level of consciousness at baseline. Compared with normal-weight patients, obese patients were younger and more often had a medical history of hypertension, diabetes, hyperlipidaemia, previous stroke/TIA, and coronary artery disease. Overweight group had the highest proportions of male, current smoker and regular alcohol user. Body mass index and stroke outcomes. There were 484 patients who had died within 12 months after stroke. High dependency was found in 1221 patients at 3 months and in 942 patients at 12 months. Unadjusted analysis showed obese patients did not differ from patients of normal BMI significantly in the risks for death and death or high dependency after stroke. Overweight patients had a slightly lower proportion of death or high dependency at 12 months than normal-weight patients. Those who were underweight had the highest risks of the three endpoints ( ; compared with normal-weight group) was associated with death at 12 months, and death or high dependency at 3 and 12 months significantly. Underweight patients remained at significantly highest risk of poor outcome (Table 2).
In the surviving patients, there was no significant difference in the risk of high dependency in overweight and obese patients compared with normal-weight group, either (Table 3). In those who had died during the follow-up period, the distribution of causes of death did not differ significantly among the BMI subgroups (p = 0.200; Table 4).

Sensitivity analyses.
The relation between BMI and the outcomes were consistent in sensitivity analyses ( Table 5). The multivariable analyses showed the risks for death and death or high dependency in overweight and obese patients did not differ from normal-weight patients significantly when using the WHO BMI classification criteria, or excluding those patients who had received thrombolysis treatment. Using another definition of high dependency as a mRS score of 4-5 did not change the association between BMI and functional disability. The stratified analysis by smoking status showed no significant differences in these outcomes among normal-weight, overweight and obese patients in both subgroups.

Discussion
The present study did not find any outcome advantage associated with being overweight or obese at the time of experiencing an acute ischemic stroke. The risks of death and functional disability in overweight and obese patients did not differ significantly from normal-weight patients. However, underweight patients had consistently the worst outcomes after ischemic stroke. Several studies reported a lower mortality in obese patients than normal-weight patients after stroke 9-15 . Doehner et al. also found obesity predicted a decreased risk of severe disability 15 . While in our study, the risks of death and death or high dependency after ischemic stroke in overweight and obese patients did not differ significantly from that of normal-weight patients. In the surviving patients, overweight and obesity was not associated with better functional recovery, either. The discrepancy of the results may be due to several reasons. First, our study included only subjects after ischemic stroke, whereas some previous studies included also subjects with hemorrhagic stroke 10,12,14,15 . Taking into account the significant differences in pathogenesis and prognostic pattern, combining ischemic and hemorrhagic stroke was not appropriate for the evaluation of the relationship between obesity and ischemic stroke outcome. Second, our analysis controlled for a wide range of covariates, including stroke severity and socioeconomic status, which were the important confounders of the association of obesity and stroke outcome but often absent in previous studies [10][11][12][13][14][16][17][18] . In a cohort of Korean patients with ischemic stroke, the favorable effects of overweight and obesity disappeared after adjustment of initial stroke severity 18 . Third, the outcomes were evaluated at different time point. Those studies which found a benefit of obesity often assessed the mortality and functional outcome after a relatively long follow-up period, ranging from 30 months to 10 years 9-12, 15, 26 . Considering that many other diseases might influence the survival and functional status of stroke patients within a longer follow-up period and limited information of these co-morbidities was collected and adjusted in previous analyses 9-12, 15, 16, 26 , their findings were more vulnerable to confounding bias. In contrast, our study selected the mortality within 12 months and the proportion of death or high dependency at 3 and 12 months after stroke as the primary outcomes, which were well-recognized measures of stroke outcome and used as study endpoints in many clinical trials on stroke 27,28 . So our result was a more reliable estimate of the independent impact of obesity on stroke prognosis. Another study using stroke-related mortality at 1 week and 1 month post-stroke as the outcomes also found no survival advantage in obese patients 16 , which was consistent with our findings. Additionally, the sensitivity analysis with a different definition of high dependency validated our result on functional outcome.
Although we found no significant differences of survival and functional disability in overweight and obese patients compared with normal-weight patients, our results still support the recommendation of current guidelines to strive for normal weight after stroke, given the association between weight loss and improvements in major cardiovascular risk factors, including dyslipidaemia, diabetes and hypertension 19,29,30 . Our study showed underweight patients had the highest all-cause mortality and the worst functional recovery after stroke, which was consistent with previous studies 8,9,11,12,26 . Although underweight individuals may more often have serious systemic diseases or cancer 16 , our analysis found the distribution of causes of death in underweight group did not differ from that of other BMI subgroups significantly, which indicated the increased mortality associated with underweight could not be attributed to other co-morbidities.
Given the BMI distribution in Asian population is left-shifted as compared with that of Western population 26, 31 , our study utilized the obesity criteria according to the Chinese Guidelines on Prevention and Control of Obesity in Adults 23,24 . Additionally, the current WHO obesity definition was used in the sensitivity analysis and did not change our results 25 .
Stroke severity is closely associated with the prognosis after stroke [16][17][18]32 . Although the National Institutes of Health Stroke Scale (NIHSS) score was not collected in the ChinaQUEST database, we included GCS score and OCSP classification as confounders in our analyses, which had been used as measures of stroke severity in previous studies and had high inter-observer reliability [33][34][35] . Additionally, GCS score and OCSP classification had been validated as the independent predictors of stroke outcome in the ChinaQUEST cohort 22 .
Some studies indicated smoking was an important confounder of the relation between BMI and mortality 3 . To avoid the potential influence of smoking, we performed a subgroup analysis to assess the association of BMI and stroke outcome in current smokers and current non-smokers, respectively. The analysis showed there was no advantage of survival or functional recovery associated with overweight and obese in both subgroups, which supported our main findings.
There are several limitations to the present study that should be addressed. First, our study is an observational investigation that suffers from the typical limitations of such studies. We adjusted the outcome analyses for a range of identified confounding variables; however, the findings may be influenced by unknown confounders. Second, we could not exclude the possibility of bias due to other treatment and intervention that might influence stroke outcome, although the sensitivity analysis showed that the intravenous thrombolysis treatment did not affect the results of our main analysis. Third, the definition of 'overweight' and 'obesity' was based on BMI in our  study. However, BMI does not reflect body fat distribution, which is also important to the evaluation of obesity 36 .
Further studies with measures of body fat distribution, such as waist circumference or waist-to-hip ratio, may provide a more accurate assessment of the association between obesity and stroke outcome.
In conclusion, the present study shows that there is no survival advantage of overweight or obesity after ischemic stroke. The risk of death and functional disability is not significantly different in overweight and obese patients as compared with normal-weight patients. Underweight is associated with the highest risk of death and high dependency after stroke. Our results support the rationality of weight control for overweight and obese patients recommended by current guidelines on secondary prevention of stroke. However, the observational nature of our study warrants further validation.