Differences in Geometric Strength at the Contralateral Hip between Men with Hip Fracture and Non-Fractured Comparators

Abstract

Older men sustain excess bone mineral density (BMD) declines after hip fracture; however, BMD provides no information on mechanical structure and strength. The aim was to assess whether changes in hip bone geometry in older men after hip fracture differ than that expected with aging. Two cohorts were used: Baltimore Hip Studies 7th cohort (BHS-7) and Baltimore Men’s Osteoporosis Study (MOST). The sample (N=170) included older Caucasian men with hip fracture that were propensity score matched (1:1) to community-dwelling non-fractured comparators. Hip Structural Analysis (HSA) calculated aerial BMD and metrics of bone structural strength: cross-sectional bone area (CSA), cortical outer diameter (OD), section modulus (SM), and centroid position (CP). Mixed-effect models estimated changes in HSA parameters and adjusted robust regression models evaluated between-cohort differences in annual percent change at the narrow neck (NN), intertrochanteric (IT), and femoral shaft (FS). Hip fracture was associated with statistically greater declines in NN CSA (β = −2.818; 95% CI: −3.300%, −2.336%), SM (β = −1.896%; 95% CI: −2.711%, −1.080%) and CP (β = −0.884%; 95% CI: −0.889%, −0.880%) and significantly larger increases in NN OD (β = 0.187%; 95% CI: 0.185%, 0.190%). Differences in IT HSA parameters were like the NN but larger in magnitude, while there were favorable changes in FS geometry where fragility fractures are rare. Findings indicate there are declines in bone structure and strength at the NN and IT regions of the proximal femur in older men during hip fracture recovery that far exceed what occurs during normal aging.

1.0. Introduction

Osteoporosis is a silent disease that affects as many as 2 million older American men [1]. The condition is often under-recognized in men, and tends to be under-treated, leading to preventable deteriorations in bone structure and strength and resulting fragility fractures [2, 3]. Older men with osteoporosis are at an increased risk of fragility fractures; more than 70,000 hip fractures occur in this group each year in the United States, the most significant consequence of osteoporosis [2, 4]. Among older men who experience hip fracture, one in three die within the subsequent year, and another one-third will fracture again [2]. Bone mineral density (BMD) has traditionally been used as a proxy metric of bone strength because it is statistically associated with hip fracture risk [2]. However, strength is a function of bone tissue’s ability to resist loading forces (material strength) as well as the amount of bone tissue and where it is placed to resist those forces (structural geometry) [5]. Despite misconceptions to the contrary, BMD provides no data about material strength, but it does correlate with geometry albeit with some ambiguity. By contrast, parameters that measure spatial distribution, resistance to bending, axial compression, and cortical stability provide information on how changes in bone geometry relate to osteoporosis and osteoporotic fractures [6]. Unfortunately, there is a paucity of longitudinal investigations on bone geometry in older men, and results from studies conducted in women may not be generalizable to men.

Cross-sectional studies indicate that older age in men is associated with lower BMD, but there are minimal age-related differences in bending strength, a consequence of increased bone width that seems to offset reduced bone tissue during aging [7, 8]. Despite compensatory changes to overcome less bone tissue that results in maintenance of structural strength, men largely retain stability within thinner cortical walls of the proximal femur [7, 8]. Equally important as pre-fracture changes in skeletal composition are post-fracture declines in bone structure that occur in the recovery period, during which time patients receive interventions to reduce disability and risk of new fractures [9]. Recent reports indicate that men experience significant declines in total hip and femoral neck BMD that are three to five times greater than decreases that occur in similar non-fractured comparators [9, 10]. Current evidence also implies that the structural advantage of males (i.e., maintenance of bending resistance and cortical stability in the presence of bone loss) is reversed after the hip fracture [11]. However, changes in bone geometry in older men that are associated with hip fracture remain unexplored. Thus, the study objective was to compare changes in geometric properties of bone structural geometry between older men after hip fracture and a similar group of community-dwelling men who did not have a hip fracture.

2.0. Methods

2.1. Hip Fracture Sample

Older men who fractured a hip were identified from the Baltimore Hip Studies 7^th (BHS-7) cohort, a prospective, observational study designed to examine sex differences in the sequelae of hip fracture [9–13]. The study protocol was approved by the University of Maryland Baltimore Institutional Review Board and review boards of participating hospitals; methodologic details have been published previously [14]. Briefly, older adults (N=362; 180 males and 182 females) hospitalized for hip fracture were recruited from 2006 to 2011 and enrolled from eight participating BHS network hospitals in the Baltimore metropolitan area. Men were continuously enrolled into the study, while recruitment of women was frequency matched with men on fracture timing within each hospital. Thus, the recruitment strategy ensured that an equal number of men and women were enrolled throughout the study and minimized the effect of secular changes in care and hospital practice differences. Participants were adults aged 65 years or older at the time of hospital admission for hip fracture (ICD-9 codes 820.00–820.9) who consented to enroll or had a proxy provide informed consent within 15 days of being admitted.

Exclusion criteria included pathologic fracture, not community-dwelling at the time of fracture, non-English speaker, being bedbound for 6 months before fracture, residence > 70 miles from the hospital, weight > 300 pounds, no surgery, and hardware in the contralateral (i.e., non-fractured) hip. Five participants did not provide data at the baseline or 2 month follow-up visit and another 18 participants were removed as a result of an IRB-requested post procedure audit (6 participants were subsequently found to be ineligible because they did not meet study inclusion criteria and 12 participants were determined to be ineligible due to secondary failures of the informed consent process), leaving a sample of 339 participants. Patients or their proxies consented to enroll in the study, and data collection occurred within 15 days of hospitalization and at 2, 6, and 12 months follow-up. The BHS-7 sample (Figure 1) was restricted to older Caucasian men (n=85) with complete baseline covariate data and ≥ 2 dual-energy x-ray absorptiometry (DXA) scans suitable for Hip Structural Analysis (HSA).

Figure 1.

Study sample flow diagram

2.2. Non-Fractured Comparators

Community-dwelling older Caucasian men who served as non-fractured comparators were selected from the Baltimore Men’s Osteoporosis Study (MOST). Between-group comparisons for other racial and ethnic groups were not possible due to the small number of nonwhite men (n=17) with hip fracture enrolled into BHS-7 that comprised individuals who were black, Asian, and from other backgrounds (e.g., American Indian/Native American). MOST was a prospective, observational cohort study, which evaluated racial differences in BMD in older men; recruitment methods and inclusion and exclusion criteria are available elsewhere [15, 16].

To summarize, participants provided written informed consent and were recruited from population-based listings of age-eligible men male drivers in the Baltimore metropolitan area and surrounding counties. Men with bilateral total hip replacements, weight over 300 pounds, or who were unable to give informed consent were excluded from the study. Participants (N=694; 503 Caucasian men) aged 65 years or older were enrolled from July 2000 through July 2001 and completed a baseline examination that included a clinical evaluation, self-administered and interviewer-administered questionnaires, as well as other assessments [15]. Among surviving participants (N=542), a second follow-up assessment was conducted between 10 and 31 months (mean=18 months) after baseline [16]. MOST participants used as comparators included individuals (n=249) who attended the baseline and second follow-up visit and had complete baseline covariate data and 2 DXA scans suitable for HSA.

2.3. DXA Measurement

Proximal femur DXA scans were performed in BHS-7 on the non-fractured hip with a Hologic (Waltham, MA, USA) or Lunar Prodigy (Madison, WI, USA) machine at baseline (≤ 15 days of hospitalization) and at 2, 6, and 12 months follow-up [9]. Among BHS-7 participants, all available DXA data during the post-fracture recovery period in those with at least two scans were used to assess changes in hip bone structural geometry. Seven facilities conducted DXA scans, and three and four sites used Hologic and Lunar prodigy machines, respectively. Participants’ DXA measurement site may have changed, but every individual had scans performed on a machine of the same manufacturer at each study visit. Seven participants in the BHS-7 cohort changed DXA measurement sites during follow-up, and only two of the 85 subjects included in the current study did not have all their DXA scans conducted on the same machine at the same location. Quality control, certification of DXA operators, and scanning procedures followed standardized methods to guarantee reproducibility of the results [9]. Across all DXA sites for BHS-7, only one to two technicians conducted the scans and placement of leg rotation was cross-trained for the study protocol. The coefficient of variation on Hologic and Lunar Prodigy machines ranged from 0.18–0.23% and 0.17–0.19%, respectively [10]. DXA machines were calibrated daily and reproducibility was assessed separately at each clinical site to ensure there was no significant measurement drift over time.

Two DXA scans of the proximal femur were conducted in MOST at baseline (Visit 1) and at the follow-up visit (Visit 2). Baseline DXA scans were measured using a QDR-2000 (Hologic, Waltham, MA, USA), but due the release of updated models during the study period, a QDR-4500 (Hologic) was used for all DXA scans at the follow-up visit [16]. In the current study, DXA data at both the baseline and follow-up visit among participants who completed these assessments were used to evaluate changes in hip bone structural geometry. Standard methods were used for quality control, certification of DXA operators, and scanning procedures to guarantee measurement reproducibility. The DXA systems were calibrated daily to provide accurate measurements in vivo when using an anthropometric phantom, and precision error rates for the QDR-2000 and QDR-4500 were estimated to be 1% or less [15, 16]. All DXA measurements were conducted by Hologic-trained and -certified technicians.

2.4. Structural Parameters

Hip Structural Analysis (HSA) was used to measure geometric properties of bone tissue. HSA software was developed at The Johns Hopkins University to analyze standard DXA images converted into digital bone mass data using proprietary file format provided from the DXA manufacturers using non-disclosure agreements [6]. DXA images are converted into bone mass images where only pixels containing bone mineral (g/cm²) are non-zero, and soft tissues and marrow cavity contents are subtracted from the image. Since DXA images are two dimensional, bending geometry and outer dimensions measured by the HSA algorithm are only relevant to forces within the plane of measurement [6]. DXA scans determined to be of insufficient quality (e.g., poor image, inadequate anatomical coverage, etc.) were excluded.

The validated HSA algorithm derives three areas for cross-sectional analysis on proximal femur DXA images: narrow neck (NN), the slimmest point of the femoral neck; intertrochanteric (IT), on the bisector of the neck-shaft angle; and femoral shaft (FS), a distance of 1.5 times the narrowest neck diameter distal to the neck-shaft axis intersection [5]. HSA software generates five parallel mineral mass profiles (lines of pixel values) traversing across the local femur axis at each analytic cross-section [5]. Geometric parameters that can be evaluated in the image are measured from each of the five profiles and then averaged at each cross-sectional region [5]. The basic principles and mathematical formulae are illustrated in Figure 2. Aerial BMD estimated from HSA generally differs when compared to values reported by manufacturers because the thin cross-sectional locations typically exclude most of the pixels in the manufacturer’s larger regions [6]. Nonetheless, the underlying mathematics (i.e., averaging of pixel values in a region) is otherwise identical. Geometric parameters used in this study include: cross-sectional area (CSA; a measure of bone surface area in cm²); outer diameter (OD: a measure of outer cortical width in cm); section modulus (SM; a measure of bending resistance in cm³); and centroid position (CP; dimensionless) as the distance from the profile center of mass to the outer (thicker) cortical surface on the medial side divided by the OD [6]. Research has shown CP to have lower values in fracture cases and is indicative of greater asymmetry in the distribution of bone mass in the cross-section [17]. Areal BMD measures derived using HSA were considered potential confounders, while other geometric parameters were assessed as outcome variables, but buckling ratio was excluded for technical reasons (see Discussion).

Figure 2.

Illustration of how HSA algorithm derives geometric properties from a projected profile of bone cross-section into DXA image. Cross-section through femur shaft is shown for clarity. Note that the bone mass image eliminates marrow cavities and trabecular voids, so that summations and dimensions intrinsically exclude them. CSA= cross-sectional area, CSMI = cross-sectional moment of inertia, SM = Section Modulus, OD = Outer Diameter, CP = centroid position, a_i = incremental subunit of area, d_i = incremental subunit of a_i distance from center of mass, d_max = maximum distance of center of mass from medial or lateral cortical surface, d_med = distance of center of mass from medial cortical surface. Note that CSMI, SM, OD and CP are referenced to the plane of the image only.

In prior research, the accuracy and precision of HSA geometry has been evaluated on Hologic Discovery and GE Prodigy DXA scanner models using a specially constructed hip geometry phantom [18]. Their phantom contained 6 different sized bone-simulating neck segments incorporating a trabecular core and an outer cortical shell, and a Hologic Discovery scanner is technically equivalent to the QDR 4500 used in the present study. Pearson correlation coefficients between DXA and phantom geometry on both scanners were 1.0 for all parameters used in the present study [18]. Furthermore, precision error estimates (i.e., % coefficient of variation) ranged from between 0.3% to 3.9 % over 10 repetitions, and accuracy tended to show positive errors for the narrowest neck segment and negative errors for the widest [18]. Measurement errors over all six segments (max, min) were between 0.06 and −.03 g/cm² for aerial BMD, 0.20 and −0.23 cm² for CSA, 0.04, −0.14 cm for OD, and 0.11 and −3.75 cm³ for SM [18].

2.5. Confounding Variables

Potential confounders were selected a priori based on review of the research literature to identify factors associated with BMD changes in older men and have been used in prior cross-cohort comparisons [10]. Variables measured at study baseline were assessed from medical record abstractions and interviews by BHS-7 research staff and a modified self-administered questionnaire and interview by MOST study personnel [9, 15]. Baseline covariates included age (years), height (meters), weight (kilograms), smoking (ever smoked 100 cigarettes), alcohol consumption (any in the last 12 months), comorbidities, concomitant medications, and areal BMD (NN, IT, and FS). Comorbidity was measured in both cohorts using a continuous count of prior conditions with variables mutually shared by the BHS-7 and MOST studies; the conditions included chronic obstructive pulmonary disease, depression, diabetes, kidney disease, liver disease, osteoporosis, Parkinson’s disease, rheumatoid arthritis, and peptic ulcer disease. Concomitant medications were measured as use within the 6 months preceding BHS-7 enrollment and concurrent to MOST study entry and included the following: bone-active drugs (BHS-7: etidronate, alendronate, risedronate, ibandronate, teriparatide, calcitonin, zoledronic acid, pamidronate; MOST: etidronate, alendronate, risedronate, calcitonin), glucocorticoids (prednisone), hormone therapy (testosterone), and calcium supplements [9, 15].

2.6. Matching

Propensity score matching was used to account for differences in baseline characteristics between the BHS-7 and MOST cohorts and to create a single homogeneous sample of community-dwelling older men [19]. BHS-7 participants with hip fracture (n=85) and non-fractured MOST participants (n=249) who had complete baseline covariate data and ≥ 2 DXA scans suitable for HSA were matched (1:1) using a propensity score (estimated probability of cohort membership conditional on potential confounders). Logistic regression was used to calculate a propensity score for each subject, and model variables included all baseline covariates selected a priori in addition to aerial BMD measured at the NN, IT, and FS. Nearest neighbor matching without replacement was conducted using the propensity score as a distance measure. Matched pairs were selected one at a time, and at each matching step, a non-fractured comparator who was not yet matched but was closest to a given hip fracture participant on the distance measure was chosen. Absolute standardized mean differences were examined in the original and matched samples to evaluate covariate balance and bias reduction [19]. Differences in baseline variables between cohorts of ≥ 0.2 standard deviations were considered evidence for covariate imbalance [19].

2.7. Statistical Analysis

Differences in geometric parameters of hip bone structure and strength between older men with hip fracture and non-fractured comparators were evaluated in two stages. At the first stage, linear mixed-effects models were used to estimate subject-specific rates of change in CSA, OD, SM, and CP at the NN, IT, and FS. Exploratory data analyses examined functional forms of geometric parameters over time, and there was no evidence for departures from linearity. Each geometric parameter was regressed on time (years) separately by cohort, and random intercepts and slopes accounted for within-subject clustering of observations. Individual rates of change were derived from random slopes for time; interpreted as one-year changes in geometric parameters.

During the second stage, robust linear regression models were used to assess differences in annual percent change ([slope/baseline value] × 100), an approach that has been used in prior HSA studies and cross-cohort comparisons [10, 11]. Robust regression models used an M-estimator (“maximum likelihood type”) and were implemented to account for potential outlier values in response variables [20]. Annual percent changes were modeled as a function of cohort membership (BHS-7 vs. MOST) to estimate between-group differences. Robust regression models adjusted for potential confounders were used to control for any residual confounding in calculating cohort-specific adjusted mean annual percent changes in geometric parameters and their corresponding 95% confidence intervals (95% CI). Statistical tests were two-sided, and all analyses were conducted using R statistical software (version 3.4.1).

3.0. Results

3.1. Sample Characteristics

The potential sample for analysis comprised 85 BHS-7 participants with 265 observations and 249 MOST participants with 498 observations. Older men who experienced hip fracture were older, taller, and weighed less than community-dwelling non-fractured comparators (Table 1). Also, BHS-7 participants were less likely to consume alcohol and use calcium supplements and had more comorbid conditions than MOST participants. Areal BMD was lower at the NN and IT in those who experienced hip fracture but higher at the FS compared to their community-dwelling non-fractured counterparts. In addition, proximal femur geometry (Table 2) at the NN and IT regions in participants with hip fracture was associated with more resistance to bending but wider bones and greater asymmetry in the distribution of bone within a cross-section. Similar baseline differences were observed at the FS region, but BHS-7 participants also had significantly greater CSA at this measurement site.

Table 1.

Baseline characteristics by cohort in the original and matched study samples.

Original Sample
Variable (n (%) or mean (SD)	MOST (n=249)		BHS-7 (n=85)		P	D
Age (years)	73.84	(5.12)	81.18	(7.21)	<0.001	1.02
Height (cm)	172.96	(6.47)	176.63	(7.66)	<0.001	0.48
Weight (kg)	83.02	(12.93)	79.18	(12.40)	0.017	0.31
Smoking	134	(53.8)	51	(60.0)	0.388	0.13
Alcohol consumption	213	(85.5)	57	(67.1)	<0.001	0.39
Comorbidity	0.57	(0.78)	1.21	(1.01)	<0.001	0.64
Calcium supplements	80	(32.1)	15	(17.6)	0.016	0.38
Glucocorticoids	25	(10.0)	5	(5.9)	0.348	0.18
Hormone therapy	11	(4.4)	1	(1.2)	0.294	0.30
Bone-active drugs	12	(4.8)	7	(8.2)	0.367	0.12
NN BMD (g/cm2)	0.73	(0.13)	0.67	(0.15)	0.001	0.39
IT BMD (g/cm2)	0.76	(0.13)	0.7	(0.17)	0.003	0.32
FS BMD (g/cm2)	1.26	(0.18)	1.33	(0.28)	0.007	0.26
Matched Sample
Age (years)	76.15	(5.24)	81.18	(7.21)	<0.001	0.70
Height (cm)	174.05	(6.82)	176.63	(7.66)	0.021	0.34
Weight (kg)	81.3	(9.56)	79.18	(12.40)	0.214	0.17
Smoking	48	(56.5)	51	(60.0)	0.756	0.07
Alcohol consumption	69	(81.2)	57	(67.1)	0.054	0.30
Comorbidity	0.84	(0.94)	1.21	(1.01)	0.013	0.37
Calcium supplements	13	(15.3)	15	(17.6)	0.836	0.06
Glucocorticoids	5	(5.9)	5	(5.9)	1.000	0.00
Hormone therapy	0	(0.0)	1	(1.2)	1.000	0.11
Bone-active drugs	5	(5.9)	7	(8.2)	0.765	0.09
NN BMD (g/cm2)	0.7	(0.13)	0.67	(0.15)	0.164	0.20
IT BMD (g/cm2)	0.74	(0.13)	0.7	(0.17)	0.149	0.20
FS BMD (g/cm2)	1.29	(0.19)	1.33	(0.28)	0.298	0.14

BHS-7: Baltimore Hip Studies 7th cohort; CM: Centimeters; D: Absolute standardized mean covariate difference; FS: Femoral shaft; G/CM²: Grams per centimeter squared; IT: Intertrochanteric; KG: Kilograms; MOST; Baltimore Men’s Osteoporosis Study; NN: Narrow neck; P: P value.

Table 2.

Baseline geometry characteristics by cohort in the original study samples.

Variable (n (%) or mean (SD)	MOST (n=249)		BHS-7 (n=85)		P
NN CSA (cm2)	2.39	(0.43)	2.32	(0.55)	0.269
NN OD (cm)	3.47	(0.22)	3.67	(0.26)	<0.001
NN SM (cm3)	1.25	(0.25)	1.35	(0.37)	0.009
NN CP (dimensionless)	0.46	(0.02)	0.45	(0.02)	<0.001
IT CSA (cm2)	4.20	(0.77)	4.2	(1.05)	0.996
IT OD (cm)	5.82	(0.36)	6.26	(0.41)	<0.001
IT SM (cm3)	4.08	(0.82)	4.47	(1.22)	0.001
IT CP (dimensionless)	0.45	(0.02)	0.43	(0.03)	<0.001
FS CSA (cm2)	3.71	(0.56)	4.17	(0.89)	<0.001
FS OD (cm)	3.10	(0.19)	3.30	(0.24)	<0.001
FS SM (cm3)	2.08	(0.36)	2.61	(0.57)	<0.001
FS CP (dimensionless)	0.49	(0.01)	0.49	(0.01)	0.565

BHS-7: Baltimore Hip Studies 7th cohort; CM: Centimeters; CM²: Centimeters squared; CM³: Centimeters cubed; FS: Femoral shaft; IT: Intertrochanteric; KG: Kilograms; MOST; Baltimore Men’s Osteoporosis Study; NN: Narrow neck; P: P value.

All BHS-7 participants were retained after 1:1 propensity matching to non-fractured comparators from MOST, and matching resulted in a decrease in absolute standardized mean differences for all baseline covariates. However, variables that still showed evidence of covariate imbalance included age, height, alcohol consumption, comorbidity, and NN and IT areal BMD. Notably, use of bone-active drugs was similar during follow-up (n=7 at 2 months, n=9 at 6 months, n=8 at 12 months) compared to baseline (n=7) in BHS-7 participants, while their use of calcium supplementation increased from 15 to 40 participants over this time period. Utilization of calcium supplements and bone-active drugs was comparable between the baseline and follow-up visit in matched MOST participants (data not shown). When compared to the included hip fracture participants, older Caucasian men in BHS-7 excluded from the potential sample for analysis were more likely to consume alcohol, and absolute standardized mean differences suggested there were additional discrepancies in terms of them being heavier, taking more glucocorticoids, and having higher aerial BMD (Supplemental Table 1). Excluded Caucasian men from MOST were older and more likely to be smokers and taking glucocorticoids than those selected to be potential comparators (Supplemental Table 2).

3.2. NN Geometry

At the NN, men in BHS-7 experienced significant mean declines in CSA and SM in the year after hip fracture (Table 3). The NN CP shifted toward the medial (thicker) cortex while the OD expanded; both changes were statistically significant. By contrast, non-fractured comparators from MOST had almost no change in NN OD and slight average increases in NN CSA, SM, and CP that generally did not reach statistical significance. Estimates of between-cohort differences in change in geometric parameters indicated that compared to MOST participants, older men enrolled in BHS-7 had significantly greater mean declines in NN CSA (β = −2.818; 95% CI: −3.300%, −2.336%), SM (β = −1.896%; 95% CI: −2.711%, −1.080%), and CP (β = −0.884%; 95% CI: −0.889%, −0.880%) that were accompanied by significantly larger increases in NN OD (β = 0.187%; 95% CI: 0.185%, 0.190%).

Table 3.

Adjusted cohort-specific percent changes and between-group differences in hip structural analysis narrow neck geometric parameters.

Parameter	BHS-7a		MOSTa		Differenceb
	% Change	95% CI	% Change	95% CI	% Change	95% CI
CSA	−2.647	−2.831, −2.464	0.171	−0.013, 0.354	−2.818	−3.300, −2.336
OD	0.188	0.186, 0.189	0.000	−0.001, 0.002	0.187	0.185, 0.190
SM	−1.466	−1.839, −1.092	0.430	0.056, 0.804	−1.896	−2.711, −1.080
CP	−0.886	−0.887, −0.884	−0.002	−0.003, 0.000	−0.884	−0.889, −0.880

BHS-7: Baltimore Hip Studies 7th cohort; CP: Centroid position; CSA: Cross sectional area; OD: Outer diameter; MOST: Baltimore Men’s Osteoporosis Study; SM: Section modulus; 95% CI: 95% confidence interval.

^aCohort-specific annual change holding model covariates at the combined sample mean.

^bAdjusted for age, height, weight, smoking status, alcohol consumption, glucocorticoids, hormone therapy, bone-active drugs, calcium supplements, comorbidity count, and baseline areal BMD (NN, IT, and FS).

3.3. IT Geometry

Older men with hip fracture had a statistically significant decrease in IT CSA and an increase in IT OD (Table 4). In the year following hip fracture, men in BHS-7 also experienced decreases in IT SM and in IT CP. Community-dwelling older men in MOST had minimal changes in IT CSA, OD, and SM but did sustain a significant decrease in IT CP toward the medial cortex. The difference in changes in geometric parameters between cohorts indicated that hip fracture in older men was associated with significantly greater declines in IT CSA and SM: −0.909% (95%: −1.026%, −0.793%) and −2.548% (95% CI: −2.821%, −2.274%), respectively. BHS-7 participants also had significantly larger increases in IT OD (β = 0.941%; 95% CI: 0.929%, 0.952%) and decreases in CP (β = −1.1388%; 95% CI: −1.171%, −1.105%) compared to those in MOST.

Table 4.

Adjusted cohort-specific percent changes and between-group differences in hip structural analysis intertrochanteric geometric parameters.

Parameter	BHS-7a		MOSTa		Differenceb
	% Change	95% CI	% Change	95% CI	% Change	95% CI
CSA	−0.902	−0.938, −0.867	0.007	−0.028, 0.042	−0.909	−1.026, −0.793
OD	0.942	0.936, 0.948	0.001	−0.004, 0.007	0.941	0.929, 0.952
SM	−2.492	−2.594, −2.391	0.055	−0.046, 0.157	−2.548	−2.821, −2.274
CP	−1.113	−1.127, −1.098	0.025	0.011, 0.040	−1.138	−1.171, −1.105

BHS-7: Baltimore Hip Studies 7th cohort; CP: Centroid position; CSA: Cross sectional area; OD: Outer diameter; MOST: Baltimore Men’s Osteoporosis Study; SM: Section modulus; 95% CI: 95% confidence interval.

^aCohort-specific annual change holding model covariates at the combined sample mean.

^bAdjusted for age, height, weight, smoking status, alcohol consumption, glucocorticoids, hormone therapy, bone-active drugs, calcium supplements, comorbidity count, and baseline areal BMD (NN, IT, and FS).

3.4. FS Geometry

At the FS cross-section, older men enrolled in BHS-7 experienced significant positive increases in CSA, OD, and SM following hip fracture (Table 5). There were no differences in cortical asymmetry between cohorts as measured by CP. Like in other measurement sites, non-fractured comparators from MOST had changes in FS CSA, OD, and SM that were much smaller in magnitude than changes among older men in BHS-7. Correspondingly, hip fracture was associated with significantly greater changes in FS geometric parameters compared to non-fracture comparators that were 2.063% (95% CI: 1.930%, 2.196%) for CSA, 0.310% (95% CI: 0.307%, 0.313%) for OD, and 2.742% (95% CI: 2.435%, 3.049%) for SM.

Table 5.

Adjusted cohort-specific percent changes and between-group differences in hip structural analysis femoral shaft geometric parameters.

Parameter	BHS-7a		MOSTa		Differenceb
	% Change	95% CI	% Change	95% CI	% Change	95% CI
CSA	2.039	1.994, 2.084	−0.024	−0.069, 0.021	2.063	1.930, 2.196
OD	0.313	0.311, 0.314	0.003	0.001, 0.004	0.310	0.307, 0.313
SM	2.576	2.458, 2.694	−0.166	−0.284, −0.048	2.742	2.435, 3.049
CP	0.010	0.003, 0.017	−0.006	−0.013, 0.000	0.017	−0.005, 0.038

BHS-7: Baltimore Hip Studies 7th cohort; CP: Centroid position; CSA: Cross sectional area; OD: Outer diameter; MOST: Baltimore Men’s Osteoporosis Study; SM: Section modulus; 95% CI: 95% confidence interval.

^aCohort-specific annual change holding model covariates at the combined sample mean.

^bAdjusted for age, height, weight, smoking status, alcohol consumption, glucocorticoids, hormone therapy, bone-active drugs, calcium supplements, comorbidity count, and baseline areal BMD (NN, IT, and FS).

4.0. Discussion

Interpreting the implications of results presents a challenging task, and fundamentally, determining how these data contribute to knowledge on structural strength of the proximal femur and the ways in which changes it over time in older men. Without a single strength estimate, the available geometric parameters inform how changes in dimensions affect the proximal femur during aging. Our study findings highlight the magnitude and direction of different patterns of changes in proximal femur geometry between older men after hip fracture and community-dwelling non-fractured comparators. More specifically, our HSA results demonstrate between-cohort differences in annual percent change with statistically greater declines at the NN and IT regions for CSA, SM, and CP, as well as significantly larger increases in NN and IT OD in the hip fracture men than non-fractured men.

If structural geometry at the fractured hip is bilaterally symmetric with the measured contralateral hip, BHS-7 participants are already at risk of secondary fractures at the NN and IT regions, both likely sites of fragility fracture [21]. Hip fracture at baseline was associated with differences in geometry that included significantly lower strength in axial compression and bending, wider bone diameters, and more medially shifted centroids. Also, older men enrolled in BHS-7 showed degradation in CSA and SM following hip fracture as well as expansion of diameter and more medial shifting of the centroid. The relationship between lower BMD and hip fracture incidence is well recognized as is the observation that the magnitude of the association for the bone mineral content (BMC) component is weaker compared to BMD alone [17]. BMD is a superior index of hip fracture than BMC because it contains both BMC and region area (a measure of OD), and although measurement units and derivation differ, CSA can be interpreted as BMC of the cross-section (BMD = BMC/region area or CSA/OD) [5, 17]. In the current study, hip fracture cases had expansions in OD and decreases in CSA, which are consistent with prior BHS research and findings from the Study of Osteoporotic Fractures, where women with hip fracture had not only lower CSA than comparators but also significantly wider femurs and more medially shifted centroids [17, 22].

The most common modality for hip fracture in older adults is a fall to the side affecting the greater trochanter, during which femur cross-sections are subject to a combination of bending and axial compression stresses, and bending stresses are generally dominant. Axial compression stresses are distributed over the total bone surface of the cross-section, and reductions in CSA will increase this stress component, but bending forces are dependent on both the amount of surface and how far the surface is from the bending axis (center of mass). Bending stresses are greatest on points in the cross-section furthest from the bending axis and are always largest in magnitude on the cortical surface. Maximum bending stress is inversely related to the SM, and at the NN and IT cross-sections, the lateral cortical surface is farther from the center of mass because the relatively thicker medial cortex shifts the center of mass medially (i.e., CP <0.5). Accordingly, cortices are more uniform in thickness at the FS region because the center of mass is typically near the central axis (CP ~0.5). As demonstrated by Fox and Keaveny, a shift in the centroid will increase bending stress on the further cortex while reducing it in the nearer cortex [23]. The limited dimensions from HSA indicate BHS-7 participants have less bone and wider diameters at baseline compared to non-fractured comparators, both of which contribute to lower BMD, and subsequent bone loss and bone gain evidenced by significantly greater declines in CSA and increases in OD at the NN and IT regions. However, it is not possible to ascertain whether bone gain is confined to the medial outer cortex (strains are greatest), or if bone loss tends toward the opposite cortical surface (strains are least), but the medial shift in CP at the NN and IT regions intimates this possibility.

In context, findings suggest that BMD declines in older men at the proximal femur during the post-fracture recovery period are (in part) related to loss of bone tissue (CSA) and to expansions in diameter at the NN and IT regions. These data are congruent with previous research demonstrating that decreases in BMD after hip fracture in women occurs in conjunction with wider bones and reductions in bone tissue [11, 22]. Specifically, results imply that both men and women who suffer fragility fractures of the hip may have a larger proportion BMD decline after hip fracture due to expanded diameter rather than to actual bone loss. Whether this observation can lead to improved identification of those at higher fracture risk by conventional DXA is a subject that will require further study. Nonetheless, it is worth noting that older men enrolled in BHS-7 would have benefited from pharmacological treatment with bisphosphonates. Given a study by Antonelli and colleagues suggests that only 8% of men initiate oral bisphosphates during hip fracture recovery, structural declines at the NN and IT regions may represent a critical gap in clinical care [24]. More broadly, our results illustrate the need for protocols in older adults who experience a hip fracture that utilize a more comprehensive strategy for prevention of secondary factures.

Results from the present study compared to prior research differ with respect to structural geometry at the FS region following hip fracture that were not evident in the MOST comparison group or in prior studies in women [11, 22]. Structural trends at the FS were reversed compared to the NN and IT regions, where unlike their non-fractured comparators, BHS-7 participants had increases in CSA, OD, and SM. Interestingly, the wider FS OD and higher CSA among BHS-7 participants versus to MOST participants was evident at the baseline comparison of structural geometry. Since method accuracy is greater at the FS region, it is unlikely that this finding is due to a methodological error, although it cannot be ruled out. Given these unexpected FS findings among hip fracture cases were present at baseline and not evident at the nearby IT and NN regions, the potential mechanisms to explain such structural trends are not clear but may have biomechanical origins. In any case, the clinical utility of these results is unlikely to have a role in secondary fracture prevention because fractures initiate from points where the femur is weaker (NN and IT) and are highly unlikely at sites where structural strength is strongest (FS). Consequently, these changes may not represent improvements in bone strength since the FS region is not a site where fragility fractures are likely to occur.

Study results need to be interpreted in terms of the design limitations. Older men from BHS-7 experienced a hip fracture, a group that is intrinsically less healthy than non-fractured comparators from MOST, and both samples may lack generalizability to more diverse populations. Notwithstanding, propensity score matching and additional multivariable adjustment for potential confounders ensured that estimated changes in geometric parameters were adjusted for measurable between-cohort differences in participant characteristics. The HSA method has several limitations that are a consequence of the two-dimensional nature of DXA images [5]. Femurs are three-dimensional objects, but dimensions and distribution of mineral mass in a DXA scan strongly depend on how the femur is positioned when it is scanned. These positioning uncertainties require larger samples and longer follow-up periods to enable detection. In prior studies, buckling ratio (BR) has been used as an index of susceptibility to failure by local buckling. While local buckling is a likely failure mechanism at the proximal femur, as estimated by HSA the BR is relatively crude, and it requires an assumption involving the distribution of mass between cortical and trabecular bone that cannot always be justified. Finally, as in any observational study, there is potential for confounding by factors that were not measured or could not be harmonized; however, there would need to be a strong unmeasured confounder or many weak unmeasured confounders that are independent of the measured confounders to bias results [25]. The risk of large biases is mitigated by the fact that BHS-7 and MOST were cohorts specifically designed to study hip fracture and bone health in older men, and therefore, collected information on the most prominent confounders. Study findings represent a reasonable estimate of the changes in bone structure and strength at the proximal femur that occur after hip fracture in older men.

5.0. Conclusions

In conclusion, older men who sustain a hip fracture experience bone loss and expansions in bone width, particularly at the NN and IT regions, resulting in centroid shifts and large declines in bending resistance at the proximal femur that are several times greater than geometric changes among non-fractured comparators. When considered in context with the changes that occur at the FS region during the post-fracture recovery period, declines in bone structure and strength at the NN and IT portions of the proximal femur highlight an area where more attention is needed in the care of older men after hip fracture. Future research should strive to identify the mechanisms that explain excess declines in bone structure and strength associated with hip fracture in older men and to develop rehabilitation strategies that integrate bone restoring medications and physical therapy regimens in order to optimize both functional recovery and prevention of secondary fractures.

Highlights.

• After hip fracture older men experience accelerated bone loss and expansions in bone width at the narrow neck and intertrochanteric

• Resultantly there are centroid shifts and large declines in bending resistance at the proximal femur exceeding changes associated with aging

• Post-fracture declines in hip bone structure and strength highlight that more attention is needed in osteoporosis care for older men

• Rehabilitation strategies that integrate bone restoring medications and physical therapy regimens are needed to optimize hip fracture recovery

Abbreviations:

BMD

bone mineral density

HSA

hip structural analysis

CSA

cross-sectional area

OD

outer diameter

SM

section modulus

CP

centroid position

NN

narrow neck

IT

intertrochanteric

FS

femoral shaft

BHS

Baltimore Hip Studies

MOST

Baltimore Men’s Osteoporosis Study

DXA

dual-energy x-ray absorptiometry

BMC

bone mineral content