INTRODUCTION
Post-stroke osteoporosis is clinically important as stroke survivors face increased risks of skeletal fragility, impaired balance, and falls. Studies show that they can experience significant bone loss, particularly in the paretic extremities [1-3]. Impaired locomotor function in these patients often reduces their ability to break a fall toward the paretic side [4,5]. These factors raise fracture incidence, especially in the paretic hip, resulting in further disability and reduced survival [4-9].
To diagnose osteoporosis, the International Society for Clinical Densitometry (ISCD) recommends measuring bone mineral density (BMD) at the lumbar spine (LS) and hip. Clinical practice usually involves dual-energy X-ray absorptiometry (DXA) scans of the LS and one hip, given the high correlation between the BMD of bilateral hips in the general population [10,11]. Unilateral hip measurements reduce scan time and radiation exposure, which remains low compared to other X-ray procedures [12,13].
Although unilateral hip DXA may be acceptable, studies suggest that bilateral hip DXA scans offer superior sensitivity, resulting in a more severe diagnosis classification in 5.35%–11% of cases compared to unilateral scans [12-20]. Bilateral BMD measurements are especially crucial for those with unilateral hip conditions. This has clinical implications, particularly in settings where osteoporosis treatment requires a World Health Organization (WHO)-classified T-score of ≤-2.5 [21]. Despite this, bilateral hip DXA is often overlooked in stroke rehabilitation, likely due to ISCD guidelines and logistical constraints.
Disagreements in BMD categorization based on the WHO criteria are classified into major and minor discordances based on the definition by Woodson [22]. Major discordance refers to a situation in which the BMD measurement results indicate osteoporosis in one measurement site while indicating normal bone density in another site. Minor discordance, in comparison to major discordance, refers to a smaller discrepancy in BMD measurement results between two measurement sites, wherein one site is diagnosed with osteoporosis and the other site is diagnosed with osteopenia, or one site is diagnosed with osteopenia while the other site is diagnosed as normal. Studies report that major T-score discordance between the spine and hip occurs in 2.7%–4.3% of cases, while minor discordance is more common, seen in 35%–51% of cases [23-25]. Factors influencing spine-hip BMD discordance include age, obesity, menopause, multiple pregnancies, and metabolic factors [26]. However, research on the prevalence of major and minor T-score discordance between the LS, paretic, and non-paretic hips, as well as related factors in stroke patients, remains limited.
To address the aforementioned knowledge gaps in screening of osteoporosis after stroke, this study investigated the prevalence of spine-hip and affected-unaffected hip T-score discordance and associated factors in hemiplegic stroke survivors using the DXA BMD of the LS and bilateral hips.
METHODS
Study design and patient selection
This was a multicenter, retrospective, cross-sectional study of hemiplegic stroke patients who visited the Department of Rehabilitation Medicine at two teaching hospitals from October 2014 to February 2022. Subjects were categorized based on the phase of stroke at the time of admission: acute patients (within one month), subacute patients (between one and six months), and chronic patients (beyond six months post-stroke). All subjects participated in individualized weight-bearing activities. Non-ambulatory patients received training using devices like tilt tables and standing frames or through robot-assisted and partial body weight-supported treadmill training, while ambulatory patients engaged in manual or robotic gait training. Activity intensity and duration were tailored to each patient’s functional status. All inpatients (n=534) received physiotherapy five days a week for 30 to 90 minutes, involving tilt tables, standing frames, or gait training. Of the 6 outpatients, 2 required assistance to walk, while 4 were independent ambulators. Hemiplegic stroke patients who had undergone LS and bilateral hip DXA and who had a first diagnosis of stroke using brain magnetic resonance imaging or computed tomography were included. Then, male≥50 years and postmenopausal female were analyzed. The exclusion criteria were as follows: age younger than 20 years old (n=32), ataxia (n=2), previous strokes affecting the sensorimotor system (n=8), cases with inadequate DXA data for analysis, such as patients who had undergone LS but only unilateral hip DXA (n=162); comorbidities affecting calcium homeostasis (i.e., thyroid and parathyroid disorders, chronic kidney disease, rheumatoid arthritis, and malignancies, such as Paget’s disease and osteomalacia) (n=5); comorbidities of neurodegenerative disorders, such as Parkinson’s disease (n=1); unilateral or bilateral implantation state or history of fracture at any site (n=4) and multiple BMD assessments across stroke phases (n=7). No patients were included multiple times across these categories (Fig. 1).
BMD measurement
BMD was measured at the first to fourth LS, total proximal femur, and femoral neck of both lower extremities using GE Lunar Prodigy Advance (GE Healthcare) DXA systems. Ward’s triangle was excluded from the analysis. BMD test results included areal BMD (aBMD) and T-score. Normative data derived from the general South Korean population were used as the reference standards for T-scores in the LS and hip areas. BMD was assessed according to the ISCD measurement criteria. Spine within the region of interest exhibiting artifacts that could potentially alter bone density or showing degenerative changes that could falsely elevate bone density, as well as spines demonstrating a T-score difference of 1 standard deviation (SD) or more between adjacent regions, were excluded from the interpretation.
Following the WHO criteria, participants were classified as having osteoporosis, if their BMD T-scores were ≥2.5 SD below the mean BMD of young adults, and osteopenia, if their lowest BMD T-scores were 1.0–2.5 SD below the mean at either the LS, femoral neck, or total proximal femur [27]. Patients with BMD T-scores >-1 in the LS and both hip areas were classified as normal. To determine the number of participants with osteoporosis, osteopenia, or normal BMD, three approaches to BMD measurement were separately utilized: measuring the LS and only the paretic hip, measuring the LS and only the non-paretic hip, and measuring the LS and both the paretic and contralateral hips.
Then, the patients were categorized into one of three groups based on their BMD measurements: concordance (consistent BMD classification of osteoporosis, osteopenia, or normal at the LS and bilateral hips), minor discordance (discordant BMD classification between the LS and the hip site, or between two hip sites), and major discordance (BMD classification of osteoporosis at one site and normal at all other sites) according to the definition by Woodson [22].
The term “low LS discordance” was applied to classify cases where the LS T-score WHO diagnosis indicated advanced bone loss compared to both hips. “Low hip discordance” categorized cases where either hip T-score WHO diagnosis showed advanced bone loss compared to the LS. “Low paretic hip discordance” and “low non-paretic hip discordance” reflected discrepancies between the paretic/non-paretic hip and contralateral hip or LS. We used “low bilateral hips discordance” for cases where bilateral hip T-score WHO classifications were identical and showed discordance, signifying more advanced bone loss than the LS. Each site-specific 'low discordance' category encompasses both minor and major discordance.
Clinical data and assessment
Data on age, sex, body mass index (BMI, kg/m2), total serum 25-hydroxy vitamin D and calcium levels, stroke type, and stroke severity were obtained. Stroke severity was assessed using the National Institutes of Health Stroke Scale. We further collected data on clinical parameters related to post-stroke functional ability, such as lower-extremity muscle strength and spasticity on the affected side, balance function, cognitive function, and activities of daily living (ADL). The assessment of muscular strength was conducted using the sum of manual muscle testing scores for the paralyzed lower limb muscles, which included the hip flexor, hip extensor, hip abductor, hip adductor, knee flexor, knee extensor, ankle dorsiflexor, and ankle plantarflexor. Additionally, the modified Ashworth scale and the Berg Balance Scale were used to assess spasticity in the affected lower limb and balance function. Cognitive function and physical disability related to ADL were assessed using the Korean Mini-Mental State Examination (K-MMSE) and the Korean version of the modified Barthel Index, respectively.
Ethics statement
Ethical approval was obtained from the Institutional Review Board of National Traffic Injury Rehabilitation Hospital (No. NTRH-22001), and the study conformed to the principles of the Declaration of Helsinki. Patient’s permission was not required as the data were de-identified using existing records. The article does not include any figure or video of a recognizable patient.
Statistical analysis
All statistical analyses were performed using a statistical program (IBM SPSS Statistics 27; IBM SPSS Inc.). Results are presented as means±SDs, medians and interquartile ranges, or counts and percentages. We used Kruskal–Wallis test followed by the Mann–Whitney U-test, and ANOVA followed by Student’s t-test for group comparisons, as appropriate. The paired t-test was applied to compare BMD measurements between the bilateral hips. Potential risk factors for T-score discordance were subjected to a multivariate logistic regression analysis, and the resulting odds ratios with 95% confidence intervals were reported. p-values<0.05 indicated statistical significance.
RESULTS
This analysis included 540 stroke patients, including 265 (49.1%) male and 275 (50.9%) female individuals. An analysis of patients based on the timing of BMD testing revealed that there were 197 acute patients, 278 subacute patients, and 65 chronic patients. The average age of the patients was 74.4±9.9 (range, 50–101) years, and the average BMI was 23.3±3.3 kg/m2. Of all patients enrolled in the study, 211 (39.07%) were diagnosed with osteoporosis. Among them, 200 patients were diagnosed with osteoporosis for the first time at the rehabilitation medicine department, 8 patients had already been diagnosed and were receiving osteoporosis treatment, and the remaining 3 patients had no records of prior medication history. Of the 8 patients receiving osteoporosis treatment, 7 patients were on oral bisphosphonates, and 1 patient was receiving teriparatide injections. Additionally, 3 of these patients were taking calcium and vitamin D supplements. Patients categorized as concordant, low LS discordance, and low hip discordance constituted 48.2%, 12.2%, and 39.6% of the total study population, respectively (Table 1).
The distribution of the T-score classifications at the LS and hips, based on the WHO criteria, is summarized in Table 2. The DXA scan at the LS and paretic hip revealed osteoporosis in 26.5% (143 patients) of the total cases. Likewise, the DXA scan at the LS and non-paretic hip detected osteoporosis in 25.0% (135 patients) of the total cases. Notably, when both hips were included in the DXA scan, osteoporosis was identified in 30.2% (163 patients) of all cases, indicating a higher detection rate of osteoporosis compared to single hip measurements with the LS. As the WHO classification of T-scores progressed from normal to osteopenia and then to osteoporosis, concordance decreased from 100% to 48.3% and 39.5%, respectively (Fig. 2).
Of the total 540 cases, major discordance was observed in 25 (4.6%) and 24 (4.5%) participants based on the LS and only paretic hip DXA or the LS and only non-paretic hip DXA, respectively (Fig. 3, Table 2). The detection rate of major discordance was 0.7%–0.9% higher when BMD was measured in LS and both hips compared to when it was measured in the LS and a single hip (29 participants, 5.4%). Minor discordance between osteopenia and osteoporosis that may influence the initiation of osteoporosis treatment was observed in 8.3% and 17.8% of the total subjects between bilateral hips and between LS and bilateral hips, respectively (Fig. 3).
The duration of stroke morbidity in the low hip discordance patient group did not show a significant difference compared to the concordance group. However, it was significantly longer than in the low LS discordance patient group (p=0.014, Table 1). Patients with low LS discordance accounted for 13.7%, 13.3%, and 3.1% of cases in the acute, subacute, and chronic phases, respectively, while low hip discordance was more prevalent, at 26.3%, 27.4%, and 41.5%. Low paretic hip discordance appeared in 3.1%, 6.8%, and 10.8% of cases across these phases (Fig. 4). These findings highlight the importance of screening the affected hip T-score in post-osteoporosis care for patients in the chronic phase of stroke.
The mean aBMD (g/cm2) and T-score for the LS were 1.04±0.23 and -1.67±1.73, respectively. In this study, the total hip values on the paretic side were significantly lower for T-scores (p<0.001), compared to those for the non-paretic hip. Subgroup analysis based on the phases of stroke was conducted to examine the differences in T-scores between both hips. The results showed that, while no significant differences were found in the acute phase, statistically significant differences were observed in the subacute and chronic phases (Table 3).
We conducted multivariate logistic regression analyses to identify the risk factors associated with low discordance between the LS, paretic hip, and non-paretic hip, compared to concordance (Table 4). The results indicated a significant association between younger age and low LS discordance (p=0.025). Additionally, a lower K-MMSE score was significantly correlated with low paretic hip discordance (p=0.024).
DISCUSSION
To the best of our knowledge, this is the first study to investigate T-score discordance characteristics between the spine and hip, as well as between affected and unaffected hips, along with associated factors, using the DXA BMD of the LS and bilateral hips in a sizable cohort of hemiplegic stroke patients. In the typical clinical assessment of BMD, DXA measurements of the LS and a single hip BMD are utilized. However, given the elevated risk of fractures, especially in the paralyzed limb, this approach may potentially delay the diagnosis of osteoporosis and increase the risk of suboptimal treatment among stroke patients. According to the findings of this study, approximately 20% of the population exhibited minor discordance, with one site showing osteoporosis and another osteopenia, or major discordance. In both cases, these discrepancies directly influence the decision to initiate osteoporosis treatment. Notably, the detection rate of major discordance was 0.7%–0.9% higher when BMD was measured in the LS and both hips compared to when it was measured in the LS and a single hip. Investigating and recognizing the prevalence of major spine-hip discordance can play a valuable role in clinical decision-making, as stroke patients are at an increased risk of hip fractures.
This study differs from previous research, which has primarily focused on accelerated trabecular versus cortical bone loss after menopause in relation to T-score discordance. Post-stroke osteoporosis exhibits unique pathophysiological differences from post-menopausal, senile, or secondary osteoporosis. Risk factors for post-stroke osteoporosis include brain-bone interaction, reduced physical activity, altered mechanical loading, and poor nutrition, compounded by other medical conditions and aging [28]. Bone loss after stroke is unevenly distributed throughout the skeleton.
This study’s findings of a significant decline in aBMD and T-scores on the paretic side during the subacute and chronic phases align with previous research showing a 12%–17% drop in BMD on the hemiplegic side within 1 year after a stroke [29]. Post-stroke bone loss begins shortly after a cerebrovascular accident and continues for approximately 3–4 months, followed by a slower progression until the end of the first year. Beyond this period, a persistent state of permanent bone loss is observed [8]. This decline is likely due to loss of mobility of hemiplegic limbs as well as reduced stress on the bones, resulting in bone loss on the affected side. These findings highlight the importance of measuring both bilateral hip BMD and LS BMD, as well as the need for early intervention to prevent bone loss.
In addition, younger age was significantly linked to a higher risk of low LS discordance. This may be due to the impact of severe lumbar disc degeneration and degenerative arthritis, which increase LS BMD in older individuals [30,31]. Additionally, LS BMD may decrease in early stroke patients as they remain bed-bound. This may explain the gradual increase in the ratio of low paretic hip discordance compared to low LS discordance, as the patient group in this study has difficulty walking but can sit as they progress from the acute to the chronic stage.
Moreover, we found a significant correlation between low paretic hip discordance and the severity of cognitive impairment, but no such correlation in the non-paretic hip or LS. Building on previous findings, we hypothesize that cognitive function influences the relationship between muscles and bones. Bone strength adapts to mechanical demands, with muscle contractions creating physiological loads necessary for osteogenesis [32,33]. To support bone adaptation and prevent bone loss, stroke rehabilitation programs should include weight-bearing exercises, resistance training, and aerobic activity. These activities promote bone formation by inhibiting resorption and stimulating nitric oxide release from osteocytes, which generates piezoelectric charges, fostering bone growth [34-37]. However, these programs may not be fully understood or consistently applied in stroke patients with cognitive impairments [7,38].
This study possesses several strengths. Firstly, the large patient sample size facilitated a comprehensive analysis of spine-hip and bilateral hip discordance, categorized as major and minor discordances, enabling a more intuitive interpretation of BMD data. Secondly, the study examined the prevalence and patterns of discordance in relation to the temporal profile after stroke. Lastly, the retrospective investigation of hemiplegic patients offers a useful model for exploring T-score discordance, as it controls for genetic and environmental factors when comparing LS, paretic hip, and non-paretic hip BMD values in each patient.
Like other retrospective observational studies, this study has several limitations. Although previous research has shown greater BMD loss in the paretic upper limb compared to the paretic lower limb, with more pronounced side-to-side differences, this study did not include BMD data for the upper limbs due to domestic regulations. Health insurance reimbursement is only approved for follow-up tests using DXA of the spine and hip. Another limitation is the relatively small number of patients in the chronic phase of stroke, which may limit the generalizability of the findings. As a result, there were constraints in analyzing risk factors for BMD discordance across different stroke phases. More comprehensive, population-based studies are needed to expand upon the current findings. Also, BMD was not continuously measured over time in the same patients; instead, concordance rates were analyzed by dividing the study population by different periods. Further longitudinal studies are necessary to validate and extend these results. Additionally, this study included stroke patients admitted to the rehabilitation unit, most of whom were moderately to severely disabled. Stroke patients with better functional abilities were likely discharged home. Therefore, more comprehensive studies involving patients across all levels of severity are needed. Moreover, this study focused on clinical variables related to stroke severity and functional outcomes. However, given that osteoporosis is influenced by various factors affecting bone metabolism, future research should explore the effects of pharmacological treatments, endocrine and metabolic conditions, and dietary patterns.
In conclusion, our study found that a lower aBMD and T-score in the paretic hip compared to the non-paretic hip and low hip discordance became more prevalent as the phase progressed from acute to chronic. This emphasizes the need for focused T-score assessments of the affected hip in post-osteoporosis care. DXA scans that combined measurements of the LS and bilateral hips increased osteoporosis detection rates, identifying more cases of major discordance than scans that measured the LS and a single hip alone. Multivariate logistic regression pinpointed a lower K-MMSE score as being significantly correlated with low discordance in the paretic hip. This research underscores the critical importance of comprehensive BMD assessments, which include the LS and bilateral hips, in post-stroke osteoporosis care to enhance diagnostic accuracy and inform treatment strategies.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Full text via PMC
Download Citation
