Low Bone Mineral Density Could Predict Cognitive Impairment during the Acute and Recovery Phases of Ischemic Stroke

Background Little is known about the effect of osteoporosis at the acute and recovery phase of stroke on cognitive function. We evaluated the effect of osteoporosis on cognitive function in patients with ischemic stroke aged >50 years. Methods We retrospectively examined consecutive patients with acute stroke hospitalized between 2016 and 2018. Osteoporosis was defined as a T score of <-2.5 for the femoral neck or lumbar spine bone mineral density. The primary outcome was cognitive impairment measured by the Korean Mini-Mental State Examination in the acute phase and recovery phase of ischemic stroke. The association between osteoporosis and the severity of cognitive impairment was investigated using a multivariate analysis. Results Of the 260 included subjects (107 men and 153 women), 70 (26.9%) had osteoporosis. Cognitive impairment was more severe in the osteoporosis group than in the non-osteoporosis group (30.5% versus 47.1%, p=0.001). After recovery phase of stroke, the proportion of patients with cognitive impairment remained higher in the osteoporosis group. The multivariate analysis revealed a correlation between a low femoral neck bone mineral density and severe cognitive impairment in the acute and recovery phases of stroke (adjusted odds ratio [OR] 3.66, 95% confidence interval [CI] 1.05-12.79 in the acute phase and adjusted OR 11.17, 95% CI 1.12-110.98 in the recovery phase), whereas lumbar spine osteoporosis was not associated with cognitive impairment. Conclusions Low bone mineral density is associated with poor cognitive function in patients with acute stroke. Early bone mineral density assessments during acute stroke, particularly at the femoral neck, may be a useful marker of cognitive function.


INTRODUCTION
Several studies have investigated the relationship between low bone mineral density (BMD) and ischemic stroke and have shown that low BMD represents a potential risk factor for stroke and affects the long-term outcome of ischemic stroke [1][2][3][4]. Nonetheless, researchers have not clearly determined whether osteoporosis is a marker for stroke. However, a potential complex causal link between stroke and osteoporosis has been reported. Since ischemic stroke can result in physical disability and cognitive impairment [5] and 40% of patients with stroke present with preexisting osteoporosis [6], a study designed to evaluate the effect of osteoporosis on cognitive impairment after ischemic stroke would be interesting. Relatively few studies of BMD in patients with cerebrovascular disease have been performed compared with the number of studies investigating BMD in patients with cardiovascular disease.
Recently, BMD was reported to be associated with microvascular diseases, such as cerebral white matter disease (WMD); a lower BMD indicates a more severe WMD [7,8]. Additionally, WMD is a well-known risk factor for dementia in several studies [9].
Hence, we postulated that a low BMD may be associated with cognitive impairment.
In this study, we investigated the relationship between BMD and cognitive function in patients with ischemic stroke during the acute and recovery phases patients using a single-center registry database in an attempt to reveal a new prognostic factor for cognitive impairment in patients with stroke.

Study population
We included consecutive patients with acute ischemic stroke who were hospitalized 5 at the Chuncheon Sacred Heart Hospital between January 2016 and June 2018. All patients over 50 years of age were subjected to routine bone densitometry within 7 days of stroke onset. We excluded patients who had a surgical or medical history affecting bone turnover (traumatic fracture; spinal surgery; liver, renal, or thyroid disease; and hormonal agent use) and a pre-stroke modified Rankin Scale (mRS) score of >3. All females enrolled in this study were menopausal. Unconscious patients whose cognitive function was unable to be assessed were also excluded.
We established 3 databases to investigate the effects of osteoporosis on cognitive function and WMD. Because patients with severe neurological deficits were unable to perform or refuse cognitive function tests, we established three separate databases: database 1 excluded subjects who were not able to complete the Korean Mini-Mental State Examination (K-MMSE) evaluating the effect of osteoporosis on cognitive function during hospitalization; database 2 excluded subjects who were not able to complete the K-MMSE at 1 year after stroke onset; and database 3 included all eligible subjects with WMD ( Fig. 1).

Data collection and outcome measures
The following data were obtained directly from the registry database: (1) demographic characteristics: age and sex; (2) stroke risk factors and medical history: previous stroke, hypertension, diabetes mellitus, hyperlipidemia, atrial fibrillation, current smoking status, and prior antithrombotic drug use; (3) clinical characteristics and acute stroke treatment: initial National Institute of Health Stroke Severity Scale (NIHSS) score, ischemic stroke mechanism according to the TOAST criteria with some modifications [10], educational level (years), body mass index and systolic blood pressure; (4) laboratory data: white blood cell count and 6 hemoglobin, creatinine, initial random glucose, low-density lipoprotein and glycated hemoglobin levels; and (5) functional status defined by the mRS score at discharge and at 3 months.
The primary outcome measure was the severity of cognitive impairment, and the secondary outcome measure was the extent of WMD stratified by BMD T scores of <-2.5 and ≥-2.5.

BMD measurements
The BMD at the lumbar spine and femoral neck was estimated using dual-energy Xray absorptiometry (Horizon W DXA system, HOLOGIC, USA). The coefficient of variation was approximately 1% for the BMD of the lumbar spine and femoral neck.
The lumbar spine BMD was measured as the mean BMD from the first to the fourth lumbar vertebra. The femoral neck BMD was measured on only the right side.
Osteoporosis was defined as a BMD at the lumbar spine or femoral neck that was at least 2.5 standard deviations (SDs) below the reference mean (T score <-2.5) [11].
The volumetric BMD (g/cm 2 ) was also measured at both the lumbar spine and femoral neck. The BMD was measured in all subjects during hospitalization.

Cognitive function test
The K-MMSE was used to assess the cognitive function of the subjects [12]. The K-MMSE is the most widely applied cognitive function screening tool used in clinical research in Korea. The initial K-MMSE was administered by an expert neuropsychologist within 7 days of stroke onset. At 1 year after ischemic stroke onset, which was considered the recovery phase of stroke, a follow-up K-MMSE was administered to each patient. We categorized cognitive impairment as follows: no 7 cognitive impairment (K-MMSE score of 24-30 points), mild cognitive impairment (K-MMSE score of 18-23 points) and severe cognitive impairment (K-MMSE score of 0-17 points) [13].

Assessment of WMD
All subjects were subjected to brain magnetic resonance imaging (MRI). The Fazekas scale was used to quantify the periventricular and deep subcortical white matter on the MRI fluid attenuation inversion recovery images and assess the presence of a WMD [14]. The Fazekas scale is recommended to assess the extent of WMD due to its simplicity and applicability in MRI. The severity of WMD was defined as follows: no punctate white matter foci (Fazekas score=0), multiple punctate white matter foci (Fazekas score=1, mild), incipient confluence or bridging of punctate foci (Fazekas score=2, moderate) and confluent white matter areas (Fazekas=3, severe). Two expert neurologists (S-H Lee and MU Jang) assessed the WMD severity.
The interobserver reliability of the WMD scale was acceptable (kappa coefficient=0.81).

Statistical analysis
The summary statistics are presented as the numbers of subjects (percentage) for categorical variables and as the means±SDs or medians for continuous variables.
Group comparisons were performed using Pearson's chi-square test or Fisher's exact test for categorical variables and Student's t-test or the Mann-Whitney U test for continuous variables, as appropriate.
In this study, we separately analyzed the effect of BMD of the femoral neck (Model 1) and BMD of the lumbar spine (Model 2) on the outcomes. Regarding the primary 8 and secondary outcome measures, the osteoporosis and non-osteoporosis groups were compared using Pearson's chi-square test, and the independent effects of osteoporosis on those outcome measures were analyzed using a multivariate regression analysis. The multivariate analysis was adjusted for variables selected based on p values <0.1 in the comparisons of groups stratified according to osteoporosis severity, and clinically plausible associations with each outcome variable were determined. Moreover, an interaction analysis was performed to whether WMD modified the association between BMD and cognitive impairment.
We performed a multivariate analysis to evaluate the impact of BMD on cognitive function during the acute and recovery phases after stroke using databases 1 and 2, and the impact of BMD on WMD using database 3.
As a sensitivity analysis, we evaluated the impact of osteoporosis on cognitive function as defined by the Montreal Cognitive Assessment (MoCA) using the van Steenovan MoCA-MMSE conversion scale. We divided the MoCA scores into three groups (MoCA score of 0~10, 11~17 and 18~30 points) [15][16][17] Crude and adjusted odds ratios (ORs) and 95% confidence intervals (CIs) were estimated. The sensitivity of the association between the volumetric BMD and the K-MMSE score was investigated using a linear regression analysis. All data analyses were performed with IBM SPSS version 21.0 software (IBM Corporation, Armonk, NY, USA).

RESULTS
Over the study period, 302 patients (125 males and 177 females, aged 73.3±10.5 years) with acute ischemic stroke underwent bone densitometry (database 3, Additional file 1: Table S1). Of the 302 patients, 260 patients were assessed using the K-MMSE (database 1, 107 males and 153 females, aged 72.9±10.5 years). Of the 260 patients, 239 patients were assessed with the follow-up K-MMSE in the recovery phase of stroke (database 2, 98 males and 141 females, aged 73.2±10.5). Seventy patients (26.9%) in database 1 were diagnosed with osteoporosis (BMD T score <-2.5), 38 (14.6%) had a femoral neck BMD T score of <-2.5, 51 (19.6%) had a lumbar spine BMD T score of <-2.5, and 19 had both a femoral neck and lumbar spine BMD T score of <-2.5. The baseline clinical characteristics stratified by the BMD at the femoral neck and lumbar spine are shown in Table 1, Tables S2 and S3 (Additional file 1). Patients in the osteoporosis group tended to be older and were more likely to be female, have higher creatinine levels and not be current smokers.
However, the stroke severity measured using the NIHSS and the stroke subtypes were not different between the osteoporosis and non-osteoporosis groups. In addition, the functional status (mRS) at hospitalization and at 3 months did not differ between the osteoporosis and non-osteoporosis groups.
Cognitive impairment measured using the K-MMSE was more severe in the osteoporosis group than in the non-osteoporosis group. (47.1% versus 30.5%, p=0.001). Additionally, the cognitive impairment of the group with femoral neck osteoporosis was significantly more severe than the group without femoral neck osteoporosis (45.1% versus 32.5%, p=0.002), whereas the cognitive impairment of the group with lumbar spine osteoporosis was not different from the group without lumbar spine osteoporosis (Fig. 2a). During the recovery phase of ischemic stroke, the proportion of patients with a severe cognitive impairment in the osteoporosis group remained higher than the non-osteoporosis group (44.3% versus 29.8%, p=0.002). The proportion of patients with a severe cognitive impairment in the group with femoral neck osteoporosis during the recovery phase of stroke remained higher than the group without femoral neck osteoporosis (51.5% versus 30.6%, p=0.01, Fig. 2b). The extent of WMD was more likely to be severe in the osteoporosis group than in the non-osteoporosis group (p=0.02). Furthermore, the group with femoral neck osteoporosis had significantly more severe WMD than the group without femoral neck osteoporosis, but a similar difference was not observed for patients with lumbar spine osteoporosis (Fig. 3). With respect to the results according to the MoCA scores, the proportion of patients with lower MoCA scores (0~10 points) was relatively higher in the group with femoral neck osteoporosis in both the acute and recovery phases of stroke (31.1% versus 57.9%, p=0.001 in the acute phase and 47.1% versus 78.8%, p=0.002 in the recovery phase. Additional file 1: figure S1).
The multivariate analysis revealed a correlation between osteoporosis and mild cognitive impairment in the acute phase, but not with severity of cognitive impairment during the recovery phase of ischemic stroke ( Table 2, Additional file   1:Tables S4 and S5). Osteoporosis of the femoral neck was associated with a severe cognitive impairment (adjusted OR 3.66, 95% CI 1.05-12.79 in the acute phase; adjusted OR 11.17, 95% CI 1.12-110.98 in the recovery phase), whereas osteoporosis of the lumbar spine was not associated with cognitive impairment. The multivariate analysis did not reveal a correlation between osteoporosis assessed by measuring the total BMD and the extent of WMD. However, osteoporosis of the femoral neck increased the risk of moderate and severe WMD. Osteoporosis of the lumbar spine did not increase the extent of WMD (Table 3 and Additional file 1: Table S6). The mediator effect observed in the interaction analysis showed that the association between BMD and cognitive impairment was significantly modified by WMD during the recovery phase of ischemic stroke (  Table S7).
The linear regression sensitivity analysis revealed a positive correlation between the volumetric BMD at the femoral neck and increasing K-MMSE scores during both the acute and recovery phases of ischemic stroke (β=0.23, R 2 =0.26, p<0.001 in the acute phase; β=0.19, R 2 =0.36, p=0.01 in the recovery phase). However, the volumetric BMD at the lumbar spine was not significantly correlated with the K-MMSE score in either phase of ischemic stroke (Additional file 1: Table S8).

DISCUSSION
We are the first group to show that osteoporosis of the femoral neck is an independent predictor of cognitive impairment in the acute phase and recovery phase of ischemic stroke. Additionally, osteoporosis of the femoral neck increases the risk of either moderate or severe WMD. These results have significant clinical implications. Despite recommendations to measure BMD in patients with acute stroke presenting risk factors for osteoporosis, BMD is not evaluated routinely in most stroke centers [18,19]. Approximately 40% of patients with hip fracture have a cognitive impairment [20]. Because stroke could lead to an increased risk of cognitive impairment, the early detection and treatment of osteoporosis might promote improved cognitive function in patients with stroke.
In the present study, the percentages of patients presenting a cognitive impairment based on a K-MMSE score <24 score were 63.8% (28.8% with a mild cognitive impairment and 35.0% with a severe cognitive impairment) during the acute phase and 63.6% (30.1% with a mild cognitive impairment and 33.5% with a severe cognitive impairment) during the recovery phase of ischemic stroke. Although the pre-stroke cognitive status was not available in the present study, the proportion of patients presenting with a cognitive impairment during the recovery phase of stroke is comparable to previous studies reporting a prevalence of cognitive impairment after stroke ranging from 24% to 70% [21]. Regarding the prevalence of post-stroke cognitive impairment, findings from this study support emerging evidence of the importance of improving strategies designed for the secondary prevention of poststroke cognitive impairment. In addition, the associations between osteoporosis and MoCA scores in the sensitivity analysis support our main findings. Based on our results, we cautiously recommend that clinicians use the BMD as a new marker of cognitive impairment after stroke.
Our study revealed the association between osteoporosis and cognitive impairment but did not demonstrate causality. Osteoporosis is known to affect microvascular diseases, such as cerebral WMD. The same mechanisms of dysregulated proteolytic processing and ion and electron transport have been described in genetic and laboratory studies of both cerebral WMD and bone osteoporosis [22,23]. In addition, WMD is a strong predictor of vascular cognitive impairment and dementia [24][25][26][27].
Notably, the mediator effect analysis showed that WMD was potentially attributed to the BMD and cognitive impairment during the recovery phase of acute ischemic stroke. Based on these laboratory and statistically results, we carefully postulated that osteoporosis affects cognitive function by inducing microvascular damage during WMD progression. However, we should be cautious when generalizing this finding, since we did not consider unmeasured confounding factors.
Interestingly, femoral neck BMD, but not lumbar spine BMD, was associated with 13 severity of cognitive impairment and WMD. In previous studies of cardiovascular disease, femoral neck BMD was significantly associated with severe coronary atherosclerosis and an increased risk of coronary artery disease [28,29]. This association was explained by the greater likelihood that spinal BMD will be affected by drugs, medical conditions and degenerative arthritis in elderly people than femoral neck BMD [30]. The correlation between the volumetric femoral neck and lumbar spine BMDs and the MMSE score and the association between the femoral neck BMD and MoCA score in this study strongly support this hypothesis. Therefore, femoral neck BMD potentially represents a more useful marker for predicting cognition and WMD than lumbar spine BMD. However, since both femoral neck and lumbar spine BMD were associated with a poor outcome in a previous analysis of cerebrovascular disorders [2], and other studies measured only the femur neck BMD to investigate stroke risk [1, 3], we should use caution when generalizing these results. Nonetheless, these results could prompt researchers to identify a reasonable BMD site for predicting cognition in patients with stroke. Further studies are needed to clarify this issue.
Osteoporosis is associated with atherosclerosis [31-34], but this association remains controversial [35]. Most of these studies have focused on coronary artery disease.
In the study, large artery atherosclerosis tended to be an uncommon ischemic stroke mechanism, whereas the incidence of stroke caused by small vessel occlusion tended to be high in patients with osteoporosis. Although osteoporosis is known to share risk factors with atherosclerosis, studies examining the relationship between atherosclerosis and osteoporosis in patients with ischemic stroke are rare.
Regarding the ischemic stroke mechanism, the proportion of patients presenting with atherosclerosis who are diagnosed with ischemic stroke is relatively low (ranging from 9.3% to 20.9% in patients with ischemic stroke) than the proportion of patients presenting with atherosclerosis who are diagnosed with coronary artery disease [10]. Thus, osteoporosis may exhibit a stronger correlation with microvascular disease than with atherosclerosis in subjects with a brain infarction.
However, this conclusion should be interpreted with caution due to the small sample size of our study.
Although this study is the first to show an association between a low BMD and cognition in patients with stroke, it has several limitations. First, this study was a single-center study with a relatively small number of subjects, despite the use of a prospective database. Further multicenter studies assessing cognitive function should be established to confirm our results. Second, since this study employed a cross-sectional design, we were unable to determine causality between cognition and BMD. The levels of vitamin D and osteocalcin, potential mediators of the observed associations, were not available for all subjects. However, we proposed that WMD had an important role in the effect of BMD on cognitive decline. These findings might allow clinicians to establish a clinical strategy that predicts cognitive impairment after stroke. Third, in the present study, the mean age of all subjects was high (72.9±10.5 years), and the osteoporosis group tended to be older than the non-osteoporosis group, which may have affected the outcome. Post-stroke depression also could affect the cognition of the subjects. After reviewing the Geriatric Depression Scale (GDS), which was administered at the same time as the MMSE, the GDS scores were not different between the two groups (data were not shown). Since we adjusted for age in all multivariate models and found no difference in GDS scores between the osteoporosis group and the non-osteoporosis group, we assume that the impacts of age and depression on the outcomes have 15 been attenuated in this study. Fourth, although we controlled for known measurable confounders in the multivariate analysis, unmeasured confounding factors (physical activity and gait speed prior to stroke, acute stressors such as emotional stress and depression due to hospitalization or stroke symptoms, etc.) could remain and may threaten the generalizability of our results. Although scores for physical activity and gait speed prior to stroke were not available in this study, we excluded subjects with independent functional statuses (mRS >3), and the 3-month functional status scores (mRS >2) were not different between the two groups. In addition, acute stressful situations can affect the cognitive status of a patient in the acute phase of stroke. However, we aimed to evaluate the impact of osteoporosis on cognition 1 year after stroke onset, which should therefore be free from confounding by the effects of acute stressful situations. We additionally found that the adjusted ORs for the effect of osteoporosis on cognition tended to increase in the recovery phase (Table 2). Fifth, the effects of the stroke itself (lesion locations and infarct volume, etc.) on delayed cognitive impairment were not considered in the present study.
However, the locations of stroke lesions were not significantly different between the osteoporosis group and non-osteoporosis group. Although the infarct volume was not available in our study, we cautiously suggest that a lack of difference in the proportions of lesion locations could increase the validity of our results. Last, since the cognitive function test was not administered prior to stroke, the patients' prestroke cognitive statuses were unavailable. However, the aim of our study was to determine the trend in cognitive decline during the acute and recovery phases in both the osteoporosis and non-osteoporosis groups and to compare the impact of osteoporosis on cognitive function in those two phases of stroke. Since we compared the effect of BMD on cognitive impairment in both the acute and recovery 16 phases of stroke, our results could have important clinical implications.

CONCLUSIONS
In conclusion, a low BMD is associated with poor cognitive function. Although we did not determine causality between cognition and BMD, this study might allow clinicians to consider the association between osteoporosis and cognitive decline in patients with ischemic stroke. Also, we cautiously suggest that the extent of WMD

AVAILABILITY OF DATA AND MATERIALS
The datasets in the current study are available from the corresponding author on reasonable request.    Supplemental_20191122.pdf