Measurement of cortical porosity of the proximal femur improves identification of women with nonvertebral fragility fractures

These hypotheses were tested in the Tromsø Study, a single-center population-based health study in Norway, which conducted six surveys in 1974, 1979–1980, 1986–1987, 1994–1995, 2001–2002, and 2007–2008, respectively [21]. During the Tromsø 4 survey in 1994–1995, all 37,558 eligible inhabitants in Tromsø over 24 years were invited and 27,158 subjects (72 %) participated. All their nonvertebral fractures were registered from the x-ray archives of the University Hospital of North Norway, Tromsø between Tromsø 4 (1994–1995) and 1 January 2010 [22, 23].

In a nested case-control design, we identified 1250 women participating in Tromsø 4, who suffered a fracture at the hip, wrist, or proximal humerus after the age of 50 years, during the 15-year registry of fractures (1994–1995 to 2010). We invited all 760 who still were alive and living in Tromsø. We recruited 264 fracture cases after excluding premenopausal women, women with bisphosphonate exposure for osteoporosis, pathological fractures or women with hip prostheses, or metal screws in the hip region, because metal on one side can make noise in the CT images of both sides, and therefore many women with hip fractures could not be included. Age-matched fracture-free women were randomly selected among the Tromsø 4 participants, 1186 were invited, and after applying the same exclusion criteria, 260 controls were recruited. Of these 524 participants, we excluded 15 currently receiving hormone replacement therapy and 66 with movement artifacts during CT scanning. This left 443 women in the final analyses: 232 controls and 211 fracture cases (4 hip, 181 wrist, and 26 proximal humeral). The median time since their index fracture was 6.6 years. All measurements were performed from November 2011 through January 2013. All participants gave written informed consent. The study was approved by the Regional Committee of Research Ethics and the Norwegian Data Inspectorate and was conducted in accordance with the World Medical Association Declaration of Helsinki.

A self-administered questionnaire included information concerning all fractures after the age of 50 years, diseases, use of medication, and lifestyle. Height and weight were measured in light clothing without shoes. Total hip and FN aBMD were measured at the nondominant side using DXA (GE Lunar Prodigy, Lunar Corporation, Madison, WI, USA), and at the opposite side in those with hip fracture at the nondominant side, and the coefficients of variation (CV) were 1.2 and 1.7 %, respectively [24]. The women were categorized into those with normal FN aBMD, osteopenia, and osteoporosis using the WHO classification, and the T-score was based on the National Health and Nutrition Examination Survey (NHANES III) reference [24].

The 10-year probability of a major osteoporotic fracture was calculated using the country-specific WHO FRAX algorithm for Norwegian women, including age, height, weight, parental hip fracture, rheumatoid arthritis, smoking, alcohol intake, glucocorticoid therapy, and FN aBMD [6‐8]. The nonvertebral fractures used as inclusion criteria of this study were not included as “previous” fracture in the calculation of the FRAX score, as the aim was to assess 10-year probability of fracture before the event, not the probability of fracture after this event. We did include other fractures that occurred prior to the index fracture in the FRAX score, and these fractures were identified from information in questionnaires at the Tromsø 4 and the current study. FRAX is designed to be used with and without FN aBMD.

CT scans (Siemens Somatom Sensation 16, Erlangen, Germany) of the nondominant hip were performed at the Department of Radiology, University Hospital of North Norway. The CT machine had an in-plane resolution of 0.74 mm, the slice thickness was 0.6 mm, the hip was scanned from just above the femoral head to 2 cm below the lesser trochanter (Fig. 1), and the exposure dose of radiation was ~1.5 mSv. Images were analyzed in Melbourne, Australia, using StrAx1.0 software [25]. As cortices are thin at the proximal femur (femoral head, neck, and trochanter), analyses were confined to a region of interest (ROI) where the cortices are thicker. This 3.7-mm subtrochanteric region starts at tip of the lesser trochanter.

The subtrochanteric region within the ROI in low resolution CT images was segmented into the compact-appearing cortex, transitional zones, and trabecular compartment using StrAx1.0, a non-thresholding method which automatically selects attenuation profile curves as reported using HR-pQCT images [25]. Porosity within each cortical compartment was also quantified automatically throughout the ROI, and similarly in CT images as reported in HR-pQCT images [25]. Local bone edges were identified as the beginning and the end of the rising and falling S-shaped portions of the profile curve enabling the delineation of the compartments [25]. Compartments were segmented by analyzing ~3600 consecutive overlapping profiles around the perimeter of each cross-sectional slice. The density profile curve produced had two plateaus: one corresponding to the compact-appearing cortex and one corresponding to the trabecular compartment. Between these plateaus was a descending S-shaped curve or transition between the two plateaus. This was the transitional zone. The density profile curve was expressing the mineralized bone area as the percentage of total area within each column.

Porosity presented in this study is the average void volume fraction summed using all voxels within the total cortex (compact-appearing cortex, outer and inner transitional zones (Fig. 1a, b)). The porosity quantified by this algorithm is the proportion of emptiness within each voxel, or the fraction of the bone volume occupied by void (porosity). The size and number of pores were not determined by using this software. However, it is common practice to be able to accurately quantify structures or objects invisible to the naked eye. To measure porosity at the subvoxel level, below visibility, two referent attenuation values are required P: the background (muscle, water etc.) and B: the fully mineralized bone matrix (1200 mg HA/cm³). The proportion of the voxel volume occupied by mineralized bone matrix volume is its level of fullness (LOF). As previously reported, LOF of each voxel is estimated as (LOF) % = (A_i-P)/(B-P), where A_i is the attenuation of voxel i [25]. From LOF, the void volume of each voxel or level of emptiness (porosity) = 100–LOF (%) (Fig. 2). Accuracy of porosity measurements using CT, voxel size 740 microns, was validated ex vivo by testing the agreement with HR-pQCT measurements, voxel size 82 microns, of the same ROI at the femoral subtrochanteric region in 5 cadaveric specimens. The correlation between porosity of the compact appearing cortex by these methods (R ²) was 0.96 (Fig. 3a). The corresponding R ² were 0.87, 0.87, and 0.94 for porosity of the outer and inner transitional zones, and the total cortex, respectively (Fig. 3b–d). As shown in the Bland-Altman plots (right panels), the error (difference between measurements by CT and HR-pQCT scanning) ranged from 0 to 10 % depending on the compartment, and agreement between both measurements exceeded 90 %. Validation of the StrAx 1.0 software analyses of the femoral subtrochanter cortical porosity, as well as all standard CT parameters in this study, was performed by repositioning and rescanning a human hip phantom 10 times, and CVs were between 0.3 and 2.3 %. This human hip phantom was delivered with the CT scanner (Siemens Somatom Sensation 16, Erlangen, Germany).