Bone structure and remodelling in stroke patients: early effects of zoledronate.

INTRODUCTION: We have reported that after an acute stroke, intravenous zoledronate prevented bone loss in the hemiplegic hip. Participants from the trial also volunteered for trans-iliac bone biopsy, to assess the early effects of stroke and zoledronate on iliac bone remodelling.
METHODS: Patients with acute stroke were randomly assigned to a single intravenous dose of zoledronate 4 mg or placebo within 5 weeks of stroke. Biopsies from 14 patients (3 female, 11 male, mean age 71+/-11) were suitable for analysis. These were taken at mean 10 weeks (+/-2) post-stroke, and included 5 patients who had received zoledronate. Histomorphometry was performed on undecalcified sections using light and fluorescence microscopy. Static and dynamic indices of remodelling were compared to a local reference range from healthy controls. Osteoclasts and their precursors were identified on frozen sections using tartrate resistant acid phosphatase (TRAP) staining. Dual-energy x-ray absorptiometry (DXA) of the proximal femora was performed at baseline and 6 months later.
RESULTS: The eroded surface in cancellous bone (ES/BS) was significantly higher in stroke patients than controls (5.7% vs. ref 1.6%, p<0.0001). Although ES/BS did not differ between zoledronate and placebo-treated groups, there were significantly fewer osteoclasts and their precursors in zoledronate-treated individuals (p=0.023). Bone formation indices (osteoid surface, OS/BS and mineralising surface, MS/BS) were significantly lower in stroke patients than controls and although OS/BS was higher in the zoledronate group than the placebo group (p=0.033), MS/BS was not different (p=0.924). There were no differences between hemiplegic and unaffected sides for any histomorphometric parameter despite asymmetric reductions in hip bone mineral density (p=0.013).
CONCLUSION: Stroke patients had higher resorption indices and lower bone forming surfaces than controls, consistent with uncoupling of bone remodelling. These findings are preliminary and a larger study is required to evaluate the contributions of gender, age and hemiplegic status to the remodelling imbalance. Zoledronate therapy was associated with a reduction in osteoclastic cell numbers consistent with its known mode of action in bone.

RCT Entities:   Population Interventions Outcomes

Introduction

We previously demonstrated that a single intravenous infusion of zoledronate given within 5 weeks of acute stroke protected against the deleterious effects of hemiplegia on hip bone mineral density [1]. After a stroke, there is a reduction in bone mineral density in the hemiplegic hip as assessed by DXA. Despite the magnitude and rapidity of these effects, the underlying mechanisms at the bone tissue level in stroke patients are unexplained. Biochemical bone marker studies have suggested an increase in bone resorption and reduced bone formation in the first week after stroke compared to controls [2], which persisted throughout the first year [3,4]. Although bone density reduction following stroke is site specific (with the greatest losses at sites such as the affected hip [5]), more generalised bone loss as a result of bed-rest and alterations in calciotropic hormone regulation is also described [6]. Year-long treatment with daily oral risedronate after acute stroke was associated with suppression of bone resorption markers, but also an unexpected increase in bone formation markers [4]. Bone formation might be suppressed after stroke due to a combination of reduced PTH and reduced active vitamin D [7], the result of raised ionised calcium inhibiting the parathyroid gland [8]. The aims of this study were three-fold: to assess indices of cancellous, endocortical and cortical bone turnover in the iliac bone 10 weeks following stroke; to evaluate the early effects of zoledronate on bone remodelling; and to compare histomorphometric parameters in patients with stroke to a UK healthy reference range. Secondary aims were to investigate the relative contributions of the side of biopsy (hemiplegic or contra-lateral), baseline 25 hydroxyvitamin D (25OHD), stroke severity and functional status to changes in bone remodelling in stroke patients.

Methods

Approval for the study was obtained from the Cambridge Local Research Ethics Committee (Cambs LREC C, 2001). All the study participants were admitted to an acute stroke unit with first ever stroke and were taking part in the randomised placebo controlled trial of 4 mg intravenous zoledronate vs. placebo to prevent bone loss in hemiplegia. As part of that study protocol, all patients who were randomised to either zoledronate or placebo were invited to volunteer for a single trans-iliac bone biopsy. Patients received calcium (1 g) and vitamin D (800 IU) daily during the trial. Full inclusion and exclusion criteria for this trial have been published [1]. Additional exclusion criteria for the biopsy study were disorders of blood coagulation, warfarin therapy, respiratory disease and obesity. Patients had been independent and ambulant before admission with a stroke and all had hemiplegia affecting the lower limb. They were unable to walk 1 week following stroke. All patients underwent measurement of bone mineral density of both hips by DXA at study entry, repeated at 6 months. Two scales were used to evaluate patients' activity and motor power at study entry and immediately prior to biopsy. These were the Barthel Index (/100) [9] and the long term score (/48) of the Scandinavian Stroke Scale [10].

Fifteen of the 31 trial subjects volunteered for a single trans-iliac bone biopsy that was timed to occur approximately 10 weeks after the stroke (mean 10.4 +/− 1.9 SD). In one subject, the biopsy specimen was incomplete leaving 14 samples for analysis. The reasons for not having a biopsy were as follows: 9 patients declined, 5 were taking oral warfarin therapy, and 2 had discontinued the study. The trans-iliac bone biopsies were taken between January 2002 and October 2003 from 4 women (1 unsuitable for analysis) and 11 men, aged between 56 and 89 (mean 77.1 +/− 11) years. In order to explore the side-specific effects of hemiplegia on bone turnover, it was decided at the start of the study to sample alternately from the affected and unaffected side, so 7 patients had a biopsy from the hemiplegic side and 7 from the contra-lateral side. Upon unblinding, five patients had received zoledronate and 9 had received placebo infusions. Biopsies were obtained using a 7.5 mm internal diameter modified Bordier trephine using local anaesthetic infiltration and intravenous sedation. In 11 patients it was possible to double demeclocycline labelling before the biopsy (300 mg demeclocycline twice daily for 2 days, then a ten day gap, followed by 300 mg twice daily for 2 days, followed by a biopsy 4 days after the last dose) [11]. Biopsies were cut in half longitudinally, coded by the laboratory technician and all histomorphometric measurements on blinded biopsies were made by the same observer (KESP) with the exception of ES/BS (done by SV).

One half of the biopsy was embedded in methylmethacrylate for histomorphometry (British Drug House Chemicals Ltd, Poole, Dorset), the other was briefly immersed in polyvinylalcohol (PVA) and chilled in hexane for 10 min at − 70 °C before mounting in PVA on a brass chuck for TRAP staining and immunohistochemistry on cryosections [12]. Undecalcified cryosections (8 μm) were cut onto sectioning tape using a Bright cryostat microtome (Huntingdon, UK) and reacted for tartrate resistant acid phosphatase activity [13]. After air drying, cryotape sections were placed in a freshly prepared buffered solution of 18.75 mg of naphthol AS B1 phosphate, 93.75 mg of sodium tartrate and 37 ml of 0.1 M tri sodium citrate (buffered to pH 4.5 with 2.1% citric acid). They were then washed rapidly in a solution of 0.25 g of sodium fluoride in 250 ml distilled water before placing in a reagent consisting of 18.75 mg of fast garnet in 35 ml of buffer (250 ml distilled water and 3.4 g sodium acetate buffered to pH 6.2 with acetic acid). Tape sections were then washed in distilled water twice and cover slipped with an aqueous mounting media.

After methylmethacrylate embedding, 8 μm undecalcified sections were cut with a Jung Polycut microtome (Leica, Milton Keynes, UK) and stained with Von Kossa/Van Gieson stain for osteoid measurement, trabecular surface and structural parameters and toluidine blue stain for eroded surface and wall thickness (using polarised light) before mounting on slides with DPX (Fisher Chemicals). Additional 15 μm sections were examined for fluorescent labels using ultraviolet light microscopy and a 365-nm filter. For all cancellous indices at least two sections from levels separated by more than 150 μm (a reduced thickness due to the limitations of using a half core biopsy) were analysed from each biopsy to avoid replicate sampling of a single surface event. Our detailed histomorphometric methods have been published previously [14,15] but additions are described here. For the present study, cancellous bone was defined as the area bounded by, parallel to and separated from the endocortical surface by 250 μm. The remaining surface including the entire endosteum was analysed separately. Osteoid thickness (O.Th) was measured at × 200 magnification, bone surface (BS) at × 80, wall thickness (W.Th) using polarised light at × 125 and eroded surface (ES) at × 200. For O.Th, a minimum number of 20 osteoid seams was analysed from each biopsy. All osteoid widths greater than 3 μm were included in the analysis [16]. For W.Th a minimum of 25 bone remodelling units was measured for each biopsy. Mineralising surface (MS/BS%) was defined as the surface extent of labelled (double-labelled + half single-labelled surface, dL + 1/2 sL) surface to cancellous bone surface. Mineral apposition rate (distance between labels) in placebo patients included values of 0.3 μm/d for 2/7 biopsies that had single cortical labels only [17]. Bone formation rate (BFR/BS um3/um2/d) was defined as the amount of new bone mineralised per day per unit of cancellous bone surface. Based on the geometric probability density function, measures of mean apparent widths were transformed into 3D mean apparent thicknesses by multiplying them by π/4 [18]. All histomorphometric measurements are described using nomenclature approved by the American Society for Bone and Mineral Research (ASBMR [19]), with the exception of ‘cortical osteoid area’, Ct. OAr/BAr.

Endocortical histomorphometry

The endocortical surfaces were defined for the purposes of this study as both inner bone surfaces and cancellous surfaces up to 250 μm into the medullary cavity measuring parallel to the plane of the inner surfaces. Osteoid surface (Ec. OS/BS) was measured on these surfaces in the same manner as the cancellous OS/BS.

Cortical histomorphometry

Mean cortical thickness (Ct. Th) was measured automatically after drawing periosteal and endocortical surfaces from three sections and using correction for section obliquity. Cortical osteoid area (Ct. OAr/BAr) and cortical porosity (%) were measured as follows. The entire Von Kossa stained cortex was captured at 100× magnification (2 cortices per biopsy) using automated montaging software and a digital camera. Canals were defined on the cortex image using a digital drawing pen and osteoid was filled in. The osteoid area was measured directly using ImageJ software (ImageJ 1.32e, NIH, USA available to download at http://rsb.info.nih.gov/ij/) and osteoid area as a percentage of cortical bone area was calculated. Assumptions of isotropy used to convert 2D to 3D values in cancellous bone are not met in cortical bone [19,20], so mean cortical areas in 2D cannot be extrapolated to mean volumes in 3D. Therefore, Ct. OAr/BAr is reported, acknowledging that the correct ASBMR nomenclature is to use either 2D or 3D referents only in a single publication.

Comparison with healthy reference data

Mean histomorphometric parameters from the stroke patients were compared with mean values from the UK healthy adult reference range of Vedi et al [14].

Statistical analysis

Data were analysed using the JMP statistical package JMP (v 4.0, SAS institute, Cary, NC, USA). Significance was determined by Student's t-test and either a 2-sample t-test or Dunnett's test when comparing stroke patients with reference data. For non-normally distributed data, significance was determined by use of the Wilcoxon/Kruskals–Wallis 2 sample test. To calculate the individual short term precision of the measurements, five repeated measurements were made on 1 day. Each measurement was taken from the same section. Intra-observation (precision) was as follows: BV/TV 1.9%, OS/BS 4.2%, O.Th 2.4%, and W.Th 3.1%. Since the reference histomorphometry data were generated locally, coefficients of inter-observer variation were calculated after both observers (KESP and SV) recorded three measurements for each parameter after examining 2–3 sections from the same test biopsy. The inter-observer coefficients of variation were BV/TV 9.1%, OS/BS 11.3%, O.Th 8.8%, and W.Th 8.2%.

Results

Patient demographics are shown in Table 1. Static histomorphometry could be performed in 5 zoledronate patients and 9 placebo patients. The groups were comparable in terms of age and severity of stroke (as assessed by the Scandinavian Stroke Scale, SSS and Barthel Index, BI) at the time of biopsy. Qualitative assessment of the stroke biopsies showed no evidence of woven bone, marrow fibrosis or signs of cellular toxicity. The sample sizes for comparison were very small (n = 7 per group), but there were no differences between the iliac biopsies from the hemiplegic side and unaffected side for any parameter (0.90 > p(all) > 0.13). However, in the placebo group there was a markedly greater loss of total hip BMD on the hemiplegic side compared to the unaffected side in the 6 months following stroke (Table 1). Leg power impairment at 3 months after stroke (< 5 on the SSS) was not associated with statistically significant differences in any parameter (0.8 > p(all) > 0.13).

Table 1

Demographics in stroke patients [mean (95% CI)]

Measurement	Zoledronate 4 mg	Placebo	pValuea
Demographics
Age	67.2 (52.8, 81.6)	72.7 (64.3, 81.0)
n (female, male) (affected, unaffected)	5 (1, 4) (2, 3)	9 (2, 7) (5, 4)
Interval (weeks) from stroke to biopsy	9.3 (7.5, 11.1)	10.8 (9.4, 12.1)
Interval (weeks) from zol to biopsy	5.6 (4.2, 7.1)
SSS/48	29 (21, 37)	28 (22, 39)
SSS Leg Power/5	4.2 (2.6, 5.8)	3.6 (2.2, 5.0)
Barthel Index/100	58 (40, 76)	57 (34, 80)
25OHD (mmol/l)	41.4 (20.5, 62.7)	42.9 (29.5, 56.7)
DXA results
Baseline total hip BMD (g/cm2)	1.08 (0.93, 1.22)	0.98 (0.87, 1.09)
ΔTotal hip BMDb hemiplegic side (%)	+ 1.4 (− 0.2, + 3.0)	−3.3 (− 6.4, − 0.2)	0.013
Δ Total hip BMDb unaffected side (%)	+ 1.9 (− 2.0, + 5.7)	− 0.5 (− 2.7, + 1.7)	0.160

Student's t-test.

Change in BMD over 6 months.

The ES/BS was not significantly different between groups, but the zoledronate treated group had fewer TRAP+ cells per mm2 than placebo. The OS/BS was significantly higher in zoledronate-treated patients than placebo-treated patients (Table 2). The zoledronate-treated stroke patients also had significantly higher endocortical OS/BS (but not cortical OAr/BAr) than placebo-treated patients, with no difference in cortical porosity or thickness.

Table 2

Basic static, endocortical and dynamic histomorphometry in stroke patients [median (IQR)]

Measurement	n	Zoledronate 4 mg	n	Placebo	p Valuea
Static parameters
BV/TV (%)	5	21.4 (14.0, 25.6)	9	18.2 (16.6, 21.9)	0.894
OS/BS (%)	5	8.06 (4.73, 1.53)	9	2.26 (1.59, 4.91)	0.033
OV/BV (%)	5	0.89 (0.48, 1.46)	9	0.23 (0.18, 0.45)	0.046
O.Th (um)	5	6.35 (6.01, 7.68)	9	6.83 (5.00, 7.35)	0.689
ES/BS (%)	5	4.98 (3.67, 7.46)	9	5.76 (4.89, 6.54)	0.594
W.Th (um)	5	32.5 (28.4, 35.4)	9	28.3 (24.4, 29.1)	0.046
Frozen sections
TRAP+ cells (/mm2)	5	0.05 (0.01, 0.17)	9	0.35 (0.11, 0.52)	0.023
Endocortical parameters
Ec. OS/BS (%)	5	14.0 (9.0, 18.8)	9	3.8 (2.7, 7.5)	0.011
Ct. OAr/BAr (%)	5	0.43 (0.10, 0.58)	9	0.11 (0.06, 020)	0.110
Ct. Porosity (%)	5	5.84 (3.60, 7.38)	9	5.07 (2.93, 8.16)	0.790
Ct. Th (um)	5	708 (523, 875)	9	777 (559, 1166)	0..351
Dynamic parameters
MS/BS (%)	4	1.80 (0.29, 2.62)b	7	1.25 (0.00, 2.86)c	0.924
MAR (um/d)	4	0.51 (0.11, 0.91)b	7	0.55 (0.30, 0.86)d	0.925
BFR/BS (um3/um2/d)	4	0.01 (0.001, 0.03)b	7	0.01 (0.00, 0.03)c	0.924

Wilcoxon Ranks Sums test.

Includes zero value from 1 zoledronate patient with no labels present.

Includes zero values from 2 placebo patients with only single cortical labels.

Assumed MAR of 0.3 um/d in 2 placebo patients with only single cortical labels (Foldes method).

Of the 11 demeclocycline labelled patients, 4 were from the zoledronate group and 7 from the placebo group. Among the 7 placebo group biopsies, 2 samples displayed single cortical labelling only (with no double labels in cancellous or cortical bone). In those 2 samples, the MAR was assumed to be 0.3 μm/d [17]. Double labelling was seen in three biopsies from zoledronate treated patients, but in one biopsy there were no labels in any compartment (Table 2). There was no significant difference in MS/BS or BFR/BS between zoledronate-treated and placebo-treated stroke patients. However, values for these indices in both groups were substantially lower than in the healthy controls (Table 3). MAR was also similar in the two patient groups and similar to that seen in healthy controls.

Table 3

Histomorphometric parameters in stroke patients vs. healthy controls [mean (SD)]

Measurement	n		All Stroke Pts		n	Controls	p Value
Age range			56–89			51–80
Sex			3f, 11m			16f, 12m
BV/TV (%)	14		19.6 (4.6)		28	22.3 (4.3)	0.065
O.Th (um)	14		6.5 (1.1)		28	6.1 (2.1)	0.537
ES/BS (%)	14		5.7 (1.7)		28	1.6 (0.7)	< 0.0001a
MS/BS (%)	11		1.6 (1.3)c		28	11.1 (5.5)	< 0.0001a
MAR (um/d)	11		0.6 (0.3)d		28	0.6 (0.1)	0.454
			Zol	Pla
OS/BS (%)	5	9	7.7 (3.9)	3.0 (1.7)	28	27.8 (11.8)	< 0.05b
OV/BV (%)	5	9	1.0 (0.6)	0.3 (0.2)	28	4.0 (2.2)	< 0.05b
W.Th (um)	5	9	32.0 (4.0)	28.3 (3.4)	28	34.9 (3.0)	< 0.05b

Significantly different from control mean by 2 sided t-test.

Significantly different from control mean by Dunnett's test.

Includes zero values from 1 zoledronate patient (no labels present) and from 2 placebo patients (single cortical labels).

Includes assumed MAR of 0.3 um/d in 2 placebo patients with only single cortical labels (Foldes method).

Discussion

In iliac bone biopsies obtained within 3 months after acute stroke, bone formation at the tissue level was reduced when compared to healthy controls, irrespective of whether patients were receiving zoledronate therapy or placebo. However, the number of osteoclastic cells was lower in patients treated with zoledronate than with placebo, consistent with the known anti-resorptive effects of zoledronate. The increase in eroded surface in both zoledronate and placebo treated patients when compared to controls may thus reflect reduced bone formation rather than increased resorption, resulting in the failure of bone formation to occur within previously formed resorption cavities. The reason for the higher osteoid surface in zoledronate-treated patients than in those receiving placebo is unclear since the mineralising surface was similar in the two groups. The apparent difference in osteoid surface may reflect the large variance in the small number of samples measured or multiple testing of significance. The lack of demeclocyline label in one zoledronate-treated patient was noted; potential explanations for this include failure to absorb or take demeclocycline (for any reason), suppression of bone remodelling by zoledronate and absent labelling as occasionally observed in normal subjects.

The finding of reduced tissue level bone formation early after stroke is in agreement with studies demonstrating that biochemical markers of bone formation were reduced (when compared to controls) in the early phase of stroke recovery [2] and throughout the first year [3] in addition to a marked early increase in resorption markers. We did not measure turnover markers in the present study and bone histomorphometric data have not been previously reported in the stroke population. Because the biopsies were obtained within 3 months of stroke, it is assumed that bone formation measures reported here predominately reflect the activity within cancellous Basic Multicellular Units (BMUs) that were initiated prior to stroke, with fewer measurements from BMUs initiated after stroke. Since pre-stroke biopsies were not available, indices of bone formation prior to stroke are unknown. Nevertheless, the markedly lower mineralising surfaces in the stroke patients, irrespective of treatment, when compared to healthy controls support the contention that tissue level formation is significantly suppressed early after stroke, whether in new or previously initiated BMUs. Our results suggest that one explanation for bone loss in stroke may be an increase in the remodelling space through failure of the coupled formation that would normally accompany bone resorption. There may be local or systemic factors that are responsible for this altered dynamic; for instance, reversal may be prolonged or osteoblast recruitment and/or activity may be impaired.

Histomorphometric assessment of biopsies obtained in the pivotal Phase III zoledronate study (HORIZON) demonstrated the expected reduction in remodelling rate but also suggested that osteoblastic activity, as assessed by mineral apposition rate, was increased by zoledronate relative to placebo [21]. HORIZON also showed preservation of cancellous bone microarchitecture in zoledronate-treated patients, consistent with an anti-resorptive effect, although no increase in the degree of matrix mineralisation was demonstrated. In the present study no significant difference in mineral apposition rate was seen between zoledronate and placebo-treated patients, perhaps due to the small sample available here [21].

In the present study, there were no differences in indices of bone remodelling between the hemiplegic and unaffected sides (although the sample size was very small for this comparison), in contrast to the marked DXA BMD changes at the hemiplegic and unaffected hips observed in the patients (Table 1). Using DXA, hemiplegia has been shown to have both generalised and localised effects on the skeleton [22-24]. Iliac bone remodelling may therefore be representative of the generalised (or indirect) skeletal effects of stroke, rather than those directly due to altered mechanical loading or muscle forces. It has been suggested that such generalised bone loss after stroke occurs because of changes in calciotropic hormone secretion such as raised ionised calcium inhibiting parathyroid hormone [25].

The main limitation of the present study is the modest sample size, particularly for comparison of demeclocycline-derived indices. The large measurement and sampling variance associated with bone histomorphometry means that with the numbers included in our study, only large differences between zoledronate and placebo-treated patients could reliably be demonstrated. The lack of a contemporaneous control group for comparison is a shortcoming common to most histomorphometry studies, but this is mitigated by the similarities between our findings and those of Recker et al [21]. The control data in this study were derived from healthy subjects and were obtained by ourselves using similar methods. Finally, since the biopsies were obtained relatively early after stroke, the long-term steady state effects of stroke and of zoledronate could not be determined in this study.

In conclusion, our results indicate reduced tissue level bone formation in iliac bone early after stroke with a reduction in osteoclastic cells only in zoledronate-treated patients. Further studies are required to elucidate further the effects of acute stroke on bone remodelling and the mechanisms by which bisphosphonates influence both bone resorption and formation in stroke patients.