The influence of bone mineral density and cortical index on the complexity of fractures of the proximal humerus

Bone mineral density and cortical index

BMD was measured at the hip, FN and LS with DXA (Hologic, Inc., Marlborough, Massachusetts). For each of these regions, T-scores were obtained, representing normal BMD (T-scores ⩾ -1), osteopaenia (T-scores -1 to -2.5) and osteoporosis (T-scores ⩽ -2.5), according to the WHO criteria.¹² When distinguishing between normal BMD, osteopaenia or osteoporosis, the lowest T-score was the decisive factor. All patients had at least one evaluable DXA measurement of the three anatomical regions (hip, FN or LS).

The CI indicates the relative thickness of the cortex in relation to the total shaft diameter. The CI was measured using a method adapted from a technique described in previous studies (Fig. 1).¹⁰ Plain anteroposterior (AP) radiographs of the proximal humerus were used to obtain the CI. The best-quality radiograph taken within four weeks of the trauma was used. A digital radiographic image processing software with a digital measuring device (IMPAX 6.6.1; Agfa HealthCare N.V, Mortsel, Belgium) was used for measuring the CI. The interobserver variability for measuring the CI and cortical thickness described by Tingart et al,⁹ was good, hence the analysis of the CI was performed by a single investigator (BP).

Fig. 1

Example for region of interest (ROI) for analysis of cortical index. * denotes the area of marrow cavity (MA). + denotes the area of the cortex. The area of the shaft (TA) = the sum of * and both +. The cortical index (CI) is calculated by using the following formula: CI = (TA - MA)/TA or [(* and + medial and + lateral) - (*)]/(*and + medial and + lateral).

Complexity of fracture

Plain AP radiographs of the proximal humerus (the same as used for obtaining the CI) were classified according to the AO/OTA classification.¹³ Next, fractures were divided into “type A” or “simple” fractures, and “type B/C” or “complex” fractures (Fig.2). This assessment was independently carried out by two investigators (JT and BP). In the event of a disagreement between the two investigators, a third independent senior investigator (JM) decided on the definitive classification.

Fig. 2

Proximal humeral fractures divided into simple and complex fractures using the AO/OTA criteria.

Statistical analysis

Data analysis was performed using SPSS statistics software version 21.0 (IBM Corp, Armonk, NY). Patients were characterised by age, gender, mean T-scores, BMD (normal/osteopaenia/osteoporosis), mean CI and type of fracture (simple/complex). Continuous outcome variables were checked for a normal distribution. To study the differences in mean T-scores and CI between the simple and complex fracture groups, independent samples t-tests were used. This relationship was also studied by testing whether the frequencies of osteoporosis and osteopaenia were equally distributed among simple and complex fracture groups, using a chi-squared test. In addition, we tested whether there was a correlation between T-scores and CI using Pearson’s correlation coefficients.

Consequently, we studied the effect of gender, age and BMI on our primary outcome measures. The independent samples t-test was used to investigate differences in age between the simple and complex fracture group, and to examine differences in T-scores and CI between men and women.

Finally, we investigated the relationship between T-scores and fracture complexity, and between CI and fracture complexity, adjusted for age, gender and BMI, using a logistic regression analysis.

Patient characteristics

In this retrospective chart review study, 175 patients with a proximal humeral fracture visited the FLS. Seven patients were excluded from this study because of unusual AP radiographs of the proximal humerus (n = 2), a previous humeral fracture (n = 1), incomplete fracture lines or disruption at the location of the CI measurements (n = 3), and a high-energy trauma and additional injuries to the humerus (n = 1). A total of 168 patients (24 men and 144 women) with a mean age of 67.2 years (sd 9.4, range 51.0 to 88.7) were included for analysis. All patient characteristics are listed in Table I. The humeral fractures in all included patients resulted from low-energy trauma. A direct fall on the shoulder from standing height was the most frequent trauma mechanism. According to the AO/OTA classification, 46 fractures were classified as “simple” and 122 fractures were classified as “complex”. The median time between the radiograph used for the measurements (CI and fracture classification) and the FLS visit was 68 days (interquartile range 39 to 101 days).

BMD and CI in simple and complex fractures

Based on the T-scores, 23 (14%) patients had a normal BMD, 73 (43%) patients had osteopaenia, and 72 (43%) osteoporosis. No significant differences were found between simple and complex fractures of the proximal humerus in the mean T-score of the hip (respectively, 1.53 and -1.38, p = 0.38), FN (respectively, -1.90 and -1.78, p = 0.47) or LS (respectively, -2.00 and -1.76, p = 0.31). Accordingly, the proportions of simple and complex fractures were not significantly different between patients with normal BMD, osteopaenia or osteoporosis. In addition, the mean CI of the simple and complex fracture groups (0.221 versus 0.235, p = 0.14) were also not significantly different (Table I).

The CI showed a moderate correlation with T-scores of the hip (r = 0.38, p < 0.01) and the FN (r = 0.41, p < 0.01), and only a weak correlation with the T-score of the LS (r = 0.18, p = 0.03).

Gender, age and BMI

The mean age and the gender distribution of the patients in the simple fracture group was similar to that of the patients in the complex fracture group (age: 68.5 versus 66.8 respectively, p = 0.29; male gender: 11% versus 16% respectively, p = 0.44). Mean T-scores and cortical index were not significantly different between men and women (Table II). The mean BMI in the complex fracture group was significantly higher than that in the simple fracture group (26.9 versus 25.2; p = 0.05). When adjusted for age, gender and the BMD using logistic regression analysis, the effect of BMI on fracture complexity remained marginally significant, depending on the T-scores used to determine the BMD (odds ratio = 1.09 to 1.10; p = 0.05 to 0.08).

Finally, age, gender and BMI did not qualitatively influence the relationship between the BMD and fracture complexity (p > 0.6 for all T-scores), or between the CI and fracture complexity (p = 0.23).

Study goal

This is the largest study to date looking at BMD and the CI of the proximal humerus, in relation to the occurrence of complex fractures of the proximal humerus. In contrast to the authors’ hypothesis, osteoporosis does not result in more complex fractures of the proximal humerus. However, the BMI was significantly higher in patients with complex fractures.

Relation between osteoporosis and fracture complexity

The current benchmark for diagnosing osteoporosis or osteopaenia is the DXA scan. Despite DXA being regarded as the best possible option for assessing the BMD in different anatomical sites,^6,7,14 this diagnostic tool is not without its failings. Richmond et al¹⁵ have showed that the accuracy of this test depends on the placement and sizing of the region of interest (ROI), overlying material in the ROI (fat or soft-tissue calcifications) and the absence of normal structures (e.g. after laminectomy). As cadaveric studies show, osteoporosis plays an important role in the failure of osteosynthesis in fracture treatment.^16-18 The quality of bone post-trauma may play a more important role in the future to optimise treatment of fractures of the proximal humerus.¹⁹ In contrast to a previous study in which a clear correlation was found between the BMD (of the hip) and the severity of distal radial fractures, we did not find the same association between the BMD (of the hip, FN or LS) and fractures of the proximal humerus.³

The absolute cortical thickness and the CI have proven to be reliable ways of assessing osteoporosis from conventional radiographs in previous studies.^8,9 The advantage of the CI over the cortical thickness is that the CI does not depend on magnification differences between images. In contrast to the DXA scan, the CI can easily be measured directly after trauma and it is determined using conventional radiographs. Therefore, it is a measurement for assessing bone quality and can be used to select the best treatment option following a fracture of the proximal humerus. Osterhoff et al¹⁰ demonstrated in a small population that the CI was not significantly different between two-part fractures of the proximal humerus and complex fractures. In the present study, with a larger cohort of patients, we found a similar result and therefore can confirm their findings.

There are weak to moderate correlation between the T-scores (hip, FN and LS) and the CI of the proximal humerus. As other studies showed, the BMD varies for each anatomical area and therefore the hip, FN or LS BMD (as commonly measured) might not correspond to the BMD of the proximal humerus.^6,7 Tingart et al,⁹ however, found significant good to excellent correlations between T-scores of different sites within the proximal humerus and the absolute cortical thickness of the proximal humerus (r = 0.64 to 0.84). However, in daily practice, only the BMD of the hip, FN and LS are commonly measured to evaluate a patient’s bone quality. Although these parameters might give an idea about the osteoporotic status in general, they seem inadequate to give detailed information about the proximal humerus in particular. The CI appears to be an appropriate alternative for evaluating local bone quality.

Effects of gender, age and BMI

In contrast to a previous study that showed that older women experienced a greater decline in BMD than men of the same age,²⁰ our results do not show an association between BMD or CI and gender. As for other studies, our results show a decrease in BMD and CI in relation to increasing age.^21-23 Nevertheless, age does not seem to be a discriminating factor for a simple or complex proximal humeral fracture.

In this study, we found a significantly higher BMI in the complex fracture group than in the simple fracture group. Previous studies have already shown that populations with a higher BMI are more at risk of fracturing a proximal humerus than people with normal or low BMI.^24,25 In a large population-based study, Prieto-Alhambra et al²⁴ found that post-menopausal women with obesity are protected against hip and pelvic fractures but have an almost 30% increased risk of fracturing the proximal humerus in comparison with normal/underweight women. They indicated that the reason might be related to the trauma mechanism and the padding provided by adipose tissue. It is suggested that people with higher body weight experience an increased traumatic force during a fall. According to our data, a higher BMI results in increased BMD of the femur (T-scores of the hip and FN). Beck et al²⁶ suggest that as the hip is a weight bearing joint, the greater force of soft-tissue mass in heavier individuals stimulates the bone to increase the BMD in this region.

Study limitations

The retrospective design is the most important limitation in this study. Due to this design, we could not investigate the influence of other variables on the complexity of fractures of the proximal humerus, such as the specific trauma mechanism. Osterhoff et al¹⁰ also studied the relationship between the CI and the complexity of fractures of the proximal humerus after low-energy trauma. They observed that more falls from a certain height were seen in the complex fracture group than in the simple fracture group. This might suggest that trauma mechanism plays a more important role in the complexity of fractures of the proximal humerus than bone parameters such as BMD or CI, although this has not yet been investigated. Another limitation is the minimal intra- and interobserver variability of the AO/OTA classification.^27,28 Nevertheless, in our opinion it is the most suitable method for classifying simple and complex fractures of the proximal humerus. To optimise reliability, the classification was independently carried out by two investigators (JT) and (BP) and, in the event of disagreement, a third independent investigator (JM) would decide on the definitive classification. Finally, the DXA scans were not performed immediately after the trauma, which could have resulted in changes to the bone parameters over time. However, the median time between the radiographs performed, used for the (CI and fracture classification) measurements and the DXA was only two months. This period was similar for both the simple and complex fracture groups. Therefore, we deemed the influence on our results of possible bone changes in this period to be minimal.

In conclusion, this study shows no difference in the BMD of the hip, FN or LS, and no difference in the CI of the proximal humerus, in simple compared with complex fractures of the proximal humerus after a low-energy trauma. However, BMI in the complex fracture group is significantly higher than in the simple fracture group.

Consequently, a simple proximal humeral fracture does not necessarily imply a better bone quality than would be found in a complex fracture. These results suggest that factors other than the BMD and the CI, such as the BMI, play a more important role in the complexity of fractures of the proximal humerus. Further studies should focus on the relationship between the BMD of the proximal humerus and the complexity of a fracture of the proximal humerus, and also the role of trauma mechanisms on the complexity of these fractures.