Sie können Operatoren mit Ihrer Suchanfrage kombinieren, um diese noch präziser einzugrenzen. Klicken Sie auf den Suchoperator, um eine Erklärung seiner Funktionsweise anzuzeigen.
Findet Dokumente, in denen beide Begriffe in beliebiger Reihenfolge innerhalb von maximal n Worten zueinander stehen. Empfehlung: Wählen Sie zwischen 15 und 30 als maximale Wortanzahl (z.B. NEAR(hybrid, antrieb, 20)).
Findet Dokumente, in denen der Begriff in Wortvarianten vorkommt, wobei diese VOR, HINTER oder VOR und HINTER dem Suchbegriff anschließen können (z.B., leichtbau*, *leichtbau, *leichtbau*).
Comparison of the time-dependent discriminatory accuracy of femoral strength and bone mineral density for predicting future hip and major osteoporotic fractures: a 16-year follow-up of the AGES-Reykjavik cohort
The discriminative accuracy of femoral strength was significantly higher than that of aBMD over 16 years of follow-up for classifying hip fractures and major osteoporotic fractures. The use of accurate thresholds, whether for aBMD or other imaging-based biomarkers, is crucial to improve sensitivity and identify high-risk older adults.
Background
Areal bone mineral density (aBMD) is a surrogate for bone strength but has limited prognostic value. Finite element (FE)–derived femoral strength offers a biomechanical alternative to aBMD for fracture risk assessment, but its long-term predictive value remains unclear. This study compared the discriminatory accuracy of aBMD and femoral strength for hip (HFs) and major osteoporotic fractures (MOFs) over 16 years, accounting for mortality risk.
Methods
In the prospective Age Gene/Environment Susceptibility‐Reykjavik (AGES‐Reykjavik) Study, elderly participants underwent CT scans at entry and automated algorithms were used to compute aBMD and femoral strength. Time-dependent area under the receiver operating characteristic curves (AUC) was used to compare the predictive abilities of aBMD and femoral strength. Optimal cutoffs at the Youden’s index were compared with the World Health Organization (WHO)–defined aBMD cutoffs at various time points.
Results
The cohort comprised 4621 older adults (mean age 76 ± 5 years). Femoral strength had a significantly higher AUC than aBMD in identifying HFs (p < 0.05) from the 6th year in males and females, while their AUCs in predicting MOFs were similar. WHO-defined aBMD showed low sensitivity (17–52%) but high specificity (78–94%) for both HFs and MOFs. The sensitivity of optimal femoral strength was significantly higher than that of aBMD at comparable specificity by 5–19% for HFs and 2–10% for MOFs (p < 0.05).
Conclusions
Both image-based markers predict long-term fracture risk and enable opportunistic screening with existing CT scans. However, femoral strength demonstrates better discriminatory accuracy than aBMD. The low sensitivity of the WHO-defined aBMD demonstrates the necessity to revise current risk assessment criteria.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Introduction
Osteoporosis, marked by loss of bone mass and deterioration associated with aging and chronic conditions, progresses silently but results in osteoporotic fractures, which carry increased mortality and lifelong fracture risk [1]. The socioeconomic burden is high, with fracture-related costs in Europe reaching $37.5 billion in 2017 and expected to rise by 27% by 2030 [2]. Despite being underestimated, osteoporosis-related disability is comparable to or greater than that caused by other chronic conditions [3]. The World Health Organization (WHO) currently advocates for the use of areal bone mineral density (aBMD) measured from dual-energy X-ray absorptiometry (DXA) to detect osteoporosis as a precursor for elevated fracture risk. In addition, the Fracture Risk Assessment Tool (FRAX) is widely used to estimate fracture risk by combining clinical risk factors with aBMD. While the aBMD classifies individuals as normal, osteopenic, or osteoporotic, the FRAX estimates the 10-year probability of hip fractures (HFs) or major osteoporotic fractures (MOFs). Yet, aBMD and FRAX both have limited sensitivity and specificity in fracture risk prediction [4, 5].
Finite-element (FE) analysis offers an alternative to aBMD by incorporating the three-dimensional (3D) mechanical structure of bone to improve fracture prediction [6]. However, studies comparing FE-derived strength and aBMD show mixed results [7‐16], with only one study reporting improved predictive power [10]. Variability in sample sizes, study designs, and follow-up periods, even within studies on the same cohorts, contributes to these inconsistencies [10, 13, 15, 16]. Existing studies of the discriminatory power of aBMD and FE-derived strength are based on a single follow-up time point, limiting the evaluation of long-term efficacy in managing osteoporotic fracture risk. Finally, while aBMD studies typically use the universal cutoff established by WHO, predictive accuracy can be affected by the choice of cutoff values.
Anzeige
To address these gaps, this study will compare FE-derived femoral strength with DXA-equivalent aBMD, computed from CT scans, to assess their biomechanical relevance in long-term fracture risk prediction, independent of clinical risk factors. The findings could support the potential use of CT-based opportunistic screening, enabling clinicians to assess fracture risk from scans obtained for other medical indications. The study will investigate the time-dependent measures of area under the curve (AUC) for predicting osteoporotic fractures, while considering cutoff values and the competing risk of mortality. The aim is to compare the discriminatory accuracy of aBMD and FE-derived femoral strength in classifying future HFs and MOFs using the full Age Gene/Environment Susceptibility‐Reykjavik (AGES‐Reykjavik) Study cohort [17].
Subjects and methods
Participants
The full AGES‐Reykjavik Study cohort [17] comprises 5764 community-dwelling elderly Icelandic males and females. For this analysis, participants were included only if they had CT scan data available from study entry (n = 4831). Subjects were excluded if they did not provide informed consent (n = 48). The study’s last follow-up period for this investigation was 16.6 years, which is rounded down to 16 years henceforth for ease of reference. Throughout the study, participants were not lost to follow-up, except in cases of death or migration, though migration-related losses were minimal (< 1%). Demographic information used in this study was collected at baseline including age, sex, height, and weight.
Fracture and mortality outcomes
Incident fractures that occurred after baseline CT acquisition (study entry) were tracked until the end of follow-up. MOFs were defined as osteoporotic fractures occurring in the hip, spine, upper arm, or forearm. Fractures were confirmed using electronic health records, accessed via unique personal identifiers, primarily from hospital records within Iceland’s medical system. Fractures were categorized using International Classification of Diseases (ICD-9 and ICD-10) codes for low-trauma fractures. HFs were identified using ICD-9 codes 820–820.2, 820.8, and 821–821.2, and ICD-10 codes S72.0–4 and S72.7–9. In addition to these codes, MOFs were further defined using ICD-9 codes 805.0, 805.2, 805.4, 805.6, 805.8, 806, 806.0, 806.2, 806.4, 806.6, 808.0, 808.2, and 808.4, and ICD-10 codes S12.0–2, S22.0–1, S22.8, S32.0–5, and S32.7–8 for spine fractures; ICD-9 codes 812, 812.0, and 812.2–4, and ICD-10 codes S42.2–4 and S42.8–9 for upper arm fractures; and ICD-9 codes 813, 813.0, 813.2, and 813.4–5, and ICD-10 codes S52.0–6 and S52.8–9 for forearm fractures. Mortality data for study participants were obtained and validated through the Icelandic registry.
CT imaging
CT scans were acquired using a Siemens Somatom Sensation 4 multi-detector CT scanner (Siemens Medical Solutions, Erlangen, Germany). These CT scans were performed with a standard protocol (slice thickness 1 mm; tube voltage 140 kV; voxel size of 0.977 × 0.977 × 1 mm) and incorporated a hydroxyapatite calibration phantom (Image Analysis, Columbia, KY, USA) [17].
Anzeige
Calculation of DXA-equivalent aBMD
DXA scans were not acquired for the cohort. Instead, subjects were scanned in a position representing a DXA scan, with toes internally rotated using a spacer placed between the legs. Volumetric BMD (vBMD) was obtained from calibrated CT scans and then projected onto the coronal plane to calculate CT-derived total hip aBMD [17]. Keyak et al. [15] demonstrated a strong correlation (r = 0.94) between CT-derived and DXA-derived total hip aBMD in a subsample of 132 subjects from the AGES-Reykjavik cohort who underwent both scans (DXA: Lunar Prodigy, GE Medical Systems, Milwaukee, WI). The relationship was described by the equation: DXA-derived total hip aBMD (g/cm2) = 0.924 × CT-derived total hip aBMD (g/cm2) + 0.137, which was used to estimate DXA-equivalent total hip aBMD in this study [15].
FE analysis
Both femurs were segmented from the CT images using an automatic segmentation method [18]. The femur volumes were discretized into 10-noded tetrahedral elements, with a target length of 3 mm, by using commercially available software (Ansa 20.1.4; Beta CAE Systems, Root, Switzerland). Material properties were mapped to the FE meshes based on the density-calibrated gray levels of the CT images using the method described in a previous study [15]. The femurs were positioned in a sideways fall configuration with a − 5° adduction and 0° internal rotation, but our previous work indicates a limited influence of adduction and internal rotation angles on the discriminatory power of the FE models [10]. A hinge joint was used to constrain the distal end while rigid bodies with frictionless contact supported the femoral head and greater trochanter. A prescribed displacement rate of 1 m/s was applied to the greater trochanter based on experimental validation studies (see Fig. A1A; Online Resource 1) [6, 19]. The models were analyzed with a commercially available explicit finite element solver (LS-Dyna R12, LS-Dyna, Livermore, CA, USA). The strength of the femur was recorded as the peak force from the force–time response recorded at the femoral head (see Fig. A1B; Online Resource 1). The whole FE analysis process, from pre-processing to post-processing, was wrapped into a fully automatic software pipeline written in Python (version 3.8). Only the results for the left femurs were used in the present study, as aBMD was only available for the left femur.
Statistical analysis
Baseline demographic and clinical characteristics were reported as means and standard deviations for continuous variables and as absolute numbers and percentages for categorical variables. Descriptive statistics were compiled, and Student’s t-tests compared differences between subjects with fractures and those without, as well as between deceased subjects and those still alive, using the non-fracture and alive group as a reference. aBMD and femoral strength were adjusted for age using univariate linear regression before assessing statistical differences. Time-dependent receiver operator characteristic (ROC) curves and the area under the curve (AUC) values were evaluated with the “timeROC” R package to ensure robust analysis of biomarker performance, as continuous variables, in the context of competing risks [20, 21]. Stratified by sex, Cox regression was used to adjust aBMD and femoral strength for age, and the predicted residuals from the models were used to compute the time-dependent AUC values for predicting HFs and MOFs [22]. The AUC values and their 95% CIs, along with sensitivity, specificity, and optimal cutoff points, were obtained at each time point. Statistical significance of differences between ROC curves was tested using the “compare” function in the “timeROC” R package, and the adjusted p-values were used to account for the multiplicity of tests. The ROC curves at 5, 10, and 15 years were also plotted to visually assess the performance of the markers. The optimal cutoff values for aBMD and femoral strength at 5, 10, and 15 years, determined using Youden’s index, were compared with the WHO-defined osteoporotic thresholds. The Youden’s index (sensitivity + specificity − 1) was used, assuming equal weighting of false positives and false negatives. These WHO-defined thresholds are set at aBMD < 0.64 g/cm2 for females and < 0.68 g/cm2 for males, based on DXA scans (Hologic QDR 1000; Waltham, MA, USA) from the third National Health and Nutrition Examination Survey (NHANES III) Caucasian cohort [23]. The NHANES III total femur conversion equations for GE Lunar DXA systems were applied to account for differences in DXA systems [24]. Consequently, the thresholds were corrected to aBMD < 0.70 g/cm2 for females and < 0.74 g/cm2 for males to match the DXA system used in the AGES-Reykjavik cohort. Sensitivity and specificity for both biomarkers at Youden’s Index and for aBMD at the WHO-defined thresholds were evaluated in predicting HFs and MOFs. The sensitivities of optimal femoral strength (at Youden’s Index) at 5, 10, and 15 years were compared with aBMD at equivalent specificities (referred to as comparator aBMD) and vice versa. Statistical differences between the sensitivities were assessed with McNemar’s test for categorical data. Age adjustments were applied only in the initial Cox regression analysis to compute the time-dependent AUC values for predicting HFs and MOFs. No adjustments were made after the thresholds were applied in subsequent analyses. A p-value of 0.05 was considered to be significant. All statistical analyses were performed using Python (version 3.8) and R (version 3.4.5 for Windows) programming environments.
Results
In this study cohort, subjects with FE model generation errors (n = 162) were excluded. Finally, there were 4621 subjects (2592 females; 2029 males), with a mean age of 76.29 ± 5.47 years, for which aBMD and femoral strength were available for comparison. The median age was 76 years (range 66–96). The last follow-up was at 16.6 years. The median follow-up time was 8.9 years (range less than 1.0–16.4) for deceased subjects and 14.8 years (range 13.2–16.6) for patients who were alive. There were 698 subjects with HFs (485 females (19%); 213 males (10%)), and 1484 subjects (940 females (36%); 544 males (27%)) who did not sustain HFs at the end of follow-up (Table 1). The majority of subjects (1167 females (45%); 1272 males (63%)) experienced competing mortality events and did not sustain HFs (Table 1). As for MOFs, there were 1363 subjects with MOFs (1003 females (39%); 360 males (18%)), 1198 subjects (700 females (27%); 498 males (25%)) alive without MOFs and 2060 subjects (889 females (34%); 1171 males (58%)) experienced competing mortality events and did not sustain MOFs (Table 1). The cumulative incidence of HFs and MOFs increased steadily over time, with females having a higher incidence of HFs and MOFs compared to males (see Fig. A2; Online Resource 1). After adjusting for age, both females and males with HFs, as well as those with MOFs, showed significant differences in aBMD and femoral strength compared to subjects alive without fractures (Table 1). This significance did not persist in deceased males without HFs or MOFs.
Table 1
Baseline characteristics of subjects at study entry stratified by HF and MOF status at the end of 16 years of follow-up
Alive without HF
With HF
Deceased without HF
Alive without HF
With HF
Deceased without HF
Females (2592)
Males (2029)
Sex (%)
940 (36)
485 (19)
1167 (45)
544 (27)
213 (10)
1272 (63)
Age (years)
73.1 ± 4.1
77.7 ± 5.2*
77.9 ± 5.6*
73.2 ± 3.7
78.0 ± 5.3*
77.6 ± 5.4*
Total hip aBMD (g/cm2)^
0.83 ± 0.04
0.78 ± 0.05*
0.78 ± 0.05*
0.96 ± 0.02
0.94 ± 0.03*
0.94 ± 0.03*
Femoral strength (kN)^
5.38 ± 0.40
4.92 ± 0.52*
4.90 ± 0.55*
8.23 ± 0.30
7.83 ± 0.44*
7.86 ± 0.45*
Follow-up (years)
14.9 ± 0.9
11.7 ± 3.5*
8.7 ± 4.0*
14.8 ± 0.9
10.6 ± 4.1*
8.1 ± 4.0*
Alive without MOF
With MOF
Deceased without MOF
Alive without MOF
With MOF
Deceased without MOF
Females (2592)
Males (2029)
Sex (%)
700 (27)
1003 (39)
889 (34)
498 (25)
360 (18)
1171 (58)
Age (years)
72.9 ± 4.0
76.8 ± 5.3*
77.9 ± 5.8*
73.1 ± 3.6
77.5 ± 5.2*
77.7 ± 5.5*
Total hip aBMD (g/cm2)^
0.83 ± 0.04
0.79 ± 0.05*
0.78 ± 0.05*
0.96 ± 0.02
0.94 ± 0.03*
0.94 ± 0.03*
Femoral strength (kN)^
5.40 ± 0.39
5.00 ± 0.52*
4.90 ± 0.57*
8.24 ± 0.30
7.88 ± 0.44*
7.86 ± 0.46*
Follow-up (years)
14.9 ± 0.9
12.1 ± 3.5*
8.2 ± 4.0*
14.8 ± 0.9
10.9 ± 3.9*
7.9 ± 4.0*
^Adjusted for age at entry into study
*Significantly different from female or male subjects alive without HF/MOF (p < 0.05)
Figure 1 shows AUC estimates for age-adjusted aBMD (denoted in gray solid lines) and FE-derived femoral strength (denoted in blue solid lines) in predicting HFs over 16 years. The light blue dashed lines represent the cumulative incidence of HFs or MOFs over the follow-up period. Femoral strength consistently outperformed aBMD, with significant differences occurring from 6 to 15 years for females (Fig. 1A) and from 6 to 14 years in males (Fig. 1B) (p < 0.05). At 15 years, femoral strength had an AUC of 0.70 (95% CI 0.69–0.73) in females and 0.75 (95% CI 0.72–0.78) in males, surpassing the AUC estimate of aBMD of 0.67 (95% CI 0.65–0.70) in females and 0.72 (95% CI 0.69–0.75) in males by 3 points. For MOFs, AUC estimates were generally lower than for HFs (Fig. 1D). At 15 years, the AUC for femoral strength was 0.64 (95% CI 0.62–0.66) in females and 0.69 (95% CI 0.67–0.71) in males, compared to the AUC of aBMD of 0.63 (95% CI 0.62–0.65) in females and 0.68 (95% CI 0.66–0.70) in males (Fig. 1C, D). ROC curves show a clear distinction between femoral strength and aBMD for HFs (Fig. 2A, B), whereas they were similar for MOFs (Fig. 2C, D).
Fig. 1
AUC estimates as a function of follow-up time for age-adjusted FE-derived femoral strength (blue solid lines) and DXA-equivalent aBMD (gray solid lines), as continuous variables, in predicting HF for females (A), HF for males (B), MOF for females (C) and MOF for males (D). Light blue dashed lines represent the cumulative incidence of HFs or MOFs over the follow-up period. Statistical significance between the two metrics is denoted in bold
Time‐dependent ROC curves for the prediction of HFs in females (A), males (B), MOFs in females (C), and MOFs in males (D) using age-adjusted DXA-equivalent aBMD and femoral strength at 5, 10, and 15 years. The gray dashed line represents an AUC of 0.50
The WHO-defined cutoff for aBMD was compared with cutoff values for aBMD and femoral strength, determined using the Youden’s index at 5, 10, and 15 years (Table 2). The sensitivities of optimal femoral strength (using Youden’s Index) at 5, 10, and 15 years were compared to those of aBMD at corresponding specificities (denoted as comparator aBMD) (Tables 3 and 4). Similarly, the sensitivities of optimal aBMD (using Youden’s Index) were compared to those of femoral strength at corresponding specificities (denoted as comparator femoral strength) (Tables 3 and 4). This analysis aims to identify how many additional individuals at high risk of fractures could be detected using femoral strength compared to optimal aBMD, and how many by aBMD compared to the optimal femoral strength threshold.
Table 2
WHO-defined cutoff for aBMD, optimal cutoffs for aBMD and femoral strength based on the Youden’s index, and cutoffs for comparator aBMD and femoral strength at 5, 10, and 15 years in predicting HFs or MOFs
Year
WHO-defined aBMD (g/cm2)
Optimal aBMD (g/cm2)
Comparator femoral strength (kN)
Comparator aBMD (g/cm2)
Optimal femoral strength (kN)
Females
HF
5
0.70
0.74
4.33
0.76
4.54
10
0.70
0.75
4.47
0.74
4.36
15
0.70
0.77
4.63
0.79
4.99
MOF
5
0.70
0.75
4.47
0.80
5.01
10
0.70
0.75
4.46
0.78
4.82
15
0.70
0.78
4.85
0.80
5.06
Males
HF
5
0.74
0.89
7.15
0.85
6.60
10
0.74
0.89
7.20
0.85
6.60
15
0.74
0.91
7.53
0.90
7.33
MOF
5
0.74
0.89
7.15
0.85
6.62
10
0.74
0.90
7.33
0.85
6.60
15
0.74
0.91
7.52
0.90
7.26
Table 3
Proportion of HFs stratified according to the WHO-defined aBMD, optimal aBMD, and femoral strength based on Youden’s index, as well as comparator aBMD and femoral strength for females and males
Year
WHO-defined aBMD
Optimal aBMD
Comparator femoral strength
Comparator aBMD
Optimal femoral strength
Females
High-risk subjects among HF cases (sensitivity) (%)
5
52
67
71*
71*
79
10
46
66
69*
61*
66
15
39
63
67*
70*
78
Low-risk subjects among non-HF cases (specificity) (%)
5
78
68
68
61
61
10
80
66
66
70
70
15
80
62
62
53
53
Males
High-risk subjects among HF cases (sensitivity) (%)
5
33
84
82
56*
75
10
30
78
82
58*
74
15
23
75
78*
69*
76
Low-risk subjects among non-HF cases (specificity) (%)
5
93
64
64
74
74
10
93
64
64
76
76
15
94
59
59
63
63
*Significantly different from respective optimal values
Table 4
Proportion of MOFs stratified according to the WHO-defined aBMD, optimal aBMD, and femoral strength based on Youden’s index, and comparator aBMD and femoral strength for females and males
Year
WHO-defined aBMD
Optimal aBMD
Comparator femoral strength
Comparator aBMD
Optimal femoral strength
Females
High-risk subjects among MOF cases (sensitivity) (%)
5
40
60
61
73*
77
10
37
55
56
63*
68
15
32
60
64*
64*
69
Low-risk subjects among non-MOF cases (specificity) (%)
5
79
65
65
51
51
10
82
69
69
59
59
15
82
60
60
55
55
Males
High-risk subjects among MOF cases (sensitivity) (%)
5
23
65
65
49*
59
10
22
68
67
50*
56
15
17
66
66
60*
62
Low-risk subjects among non-MOF cases (specificity) (%)
5
93
64
64
74
74
10
94
63
63
77
77
15
94
60
60
65
65
*Significantly different from respective optimal values
For HF cases, the WHO-defined aBMD cutoff had the lowest sensitivity and highest specificity at all time points compared to optimal and comparator values (Table 3). At 5, 10, and 15 years, the sensitivity of comparator femoral strength was 3–4% higher in females compared to optimal aBMD (Table 3), with all three time points reaching statistical significance. Similarly, in males, the sensitivity of comparator femoral strength was 2–4% higher than aBMD, but statistical significance was only attained at 15 years (p < 0.05) (Table 3). Optimal femoral strength showed 8%, 5%, and 8%, and 19%, 16%, and 7% significantly higher sensitivity in females and males, respectively, compared to comparator aBMD (p < 0.05) (Table 3). For MOFs, the WHO-defined aBMD again had the lowest sensitivity and highest specificity (Table 4). There were no significant differences between comparator femoral strength and optimal aBMD, except at 15 years, where comparator femoral strength was significantly more sensitive by 4% in females (p < 0.05) (Table 4). Similar to HFs, optimal femoral strength was significantly more sensitive than comparator aBMD in females (4%, 5%, 5%) and in males (10%, 6%, 2%) at 5, 10, and 15 years respectively (p < 0.05) in predicting MOFs (Table 4).
Discussion
In this prospective study of elderly participants with osteoporotic fractures and long-term follow-up, the first key finding was that FE-predicted femoral strength had higher discriminatory accuracy than aBMD in predicting HFs and similar accuracy in predicting MOFs over the 16 years of follow-up. To our knowledge, the longitudinal discriminative accuracy of aBMD and/or femoral strength has not been previously evaluated. The second key finding of the study is that both measures exhibit a decline in predictive value with longer follow-up times. This raises the question of whether identifying future fractures remains valuable despite the decline in long-term prediction. In addition to the time-dependent measures of AUC, we assessed the sensitivity and specificity of both markers and their changes over time. The last key finding was that optimal femoral strength (at the Youden’s Index) showed significantly higher sensitivity than the comparator aBMD (at equivalent specificity). For MOFs, though the AUCs for both aBMD and femoral strength were not statistically different, optimal femoral strength (at the Youden’s Index) still demonstrated significantly higher sensitivity than the comparator aBMD (at equivalent specificity) in both males and females.
Anzeige
It remains to be investigated whether utilizing CT imaging data from over a decade ago is still effective for assessing fracture risk. For instance, assuming that the AUC of femoral strength at 14 years is adequate for clinical prognosis of HFs, aBMD would only be considered equally effective for up to 10 years in females and 11 years in males. This suggests that femoral strength can be utilized for a longer duration with reliable discriminatory accuracy compared to aBMD in predicting future HFs. For MOFs, the ability of femoral strength to perform comparably to aBMD in predicting these fractures highlights its potential as a prognostic biomarker in clinical settings, especially for opportunistic screening with existing CT scans [25].
AUC was generally found to be higher for males than for females for both HFs and MOFs. These differences might not solely reflect male–female disparities but could be influenced by the higher fracture incidence among females, which is nearly double that of males for both HFs and MOFs. The higher incidence in females highlights the need for more sensitive predictive biomarkers to accurately identify fractures, leading to poorer AUC estimates compared to males. In addition, there seems to be unstable predictive performances, especially in the early years when fracture incidences are low. This volatility demonstrates how the number of cases in a sample population can significantly impact AUC values. To address these challenges, statistical methods such as oversampling of minority cases could be employed to enhance model stability and predictive accuracy [26].
However, previous literature has not conclusively supported the superiority of femoral strength derived from CT imaging compared to aBMD. Among eight studies that have reported AUC, sensitivity, or specificity values of femoral strength [7‐14], only five studies have compared that to aBMD [7‐10, 12] (see Table A1; Online Resource 1). Three of those studies used a case–control study design using post-fracture imaging [8, 9, 12]. Although they indicate that FE-derived bone strength is superior to aBMD, these studies do not reflect the real clinical challenge of identifying future fracture risk [7‐10, 12]. Adams et al. [7] conducted a rigorous study with nearly 2800 older adults over 5 years and a 47% HF incidence, reporting AUC values of 0.72 to 0.75 for femoral strength and aBMD (see Table A1; Online Resource 1). In our study of almost 4600 older adults over 5 years with a HF incidence of 5% in females and 3% in males, we found higher AUC estimates for femoral strength compared to aBMD [7]. These differences could be attributed to either population differences or the use of case–control study designs that do not reflect true population cohort characteristics [7].
Currently, there is only one universally established cutoff for aBMD specified by the WHO. This study demonstrated the limitations of the WHO-defined criteria for assessing fracture risk. These findings align with the study by Stone et al., which reported similar sensitivities across various fracture sites [27]. The corresponding femoral strength cutoff values at the WHO-defined aBMD threshold were 4.0 kN in females and 5.2 kN in males for predicting either HFs or MOFs, lower than the optimal femoral strength cutoff values obtained in our study at 15 years, highlighting the need for revising the WHO-defined criteria for accurate risk assessment. Also, the corresponding femoral strength cutoff values at the WHO-defined criteria obtained in our study are higher than the thresholds established by Keaveny and colleagues [28], which are 3.0 kN for fragile bone strength in females and 3.5 kN in males. This could be attributed to the differences in the FE methodology applied but may also indicate a need for population-specific analyses to determine the thresholds.
Anzeige
It should be noted that selecting a threshold based on maximizing Youden’s index, which treats false results as clinically and economically equivalent, is not necessarily optimal, as the incremental cost and health impact of a false negative (missed fracture) is not equal to a false positive (unnecessarily treating to prevent fracture). Osteoporotic fractures are associated with high costs, disability, and mortality risk. With this in mind, high sensitivity might be preferred, even if this comes at the expense of some specificity [29]. Therefore, prioritizing sensitivity in predictive biomarkers can be a strategic choice to reduce the overall burden of osteoporotic fractures, provided that the cost of preventive treatment remains relatively low compared to the cost of fracture management. Our comparison of aBMD and femoral strength at equal specificity showed that femoral strength had superior sensitivity. At 15 years, with the same specificity, femoral strength identified 8% more high-risk females with HFs and 5% more high-risk females with MOFs, while the increases were 7% and 2% more in males, respectively. To put these percentages into context, 1052 HFs occurred in Iceland between 2008 and 2012 among adults aged 67 and older [30]. An 8% increase in sensitivity for HF prediction in females and 7% in males translates to identifying 61 additional high-risk females and 21 additional high-risk males, compared to aBMD. Given the severe morbidity, mortality, and healthcare burden associated with hip fractures, even modest improvements in risk stratification could have important implications for early intervention and targeted treatment. For MOFs, femoral strength and aBMD performed similarly. While 5% and 2% improvements in MOF detection for females and males may have a limited impact on clinical decision-making, they highlight the potential utility of femoral strength, especially in settings where aBMD screening is unavailable. FE-derived femoral strength could enable opportunistic screening using existing CT scans from other medical evaluations, complementing DXA-based assessments and providing additional biomechanical insights for fracture risk prediction.
This study has several limitations. First, the cohort only included CT imaging without DXA imaging, meaning subjects did not have a clinical diagnosis of osteoporosis. Comparing the results with DXA-derived aBMD would have provided a direct comparison to the current clinical standard. However, the CT scans were taken mimicking the leg position of DXA scans, and a relationship between CT-derived vs. DXA-derived aBMD has been established to support the use of the CT data for simulating the current clinical standard [15]. Second, the findings are specific to the Icelandic population; thus, caution is needed when applying them to other populations. Third, the use of dichotomous risk thresholds for both metrics may not be optimal. In cardiovascular risk assessment, 5-year risk is categorized into four risk levels [31], which helps identify high-risk individuals more precisely and ensures a more nuanced distribution of risk. A limitation of this study is that FE-derived strength predictions were not adjusted for clinical risk factors, such as those in the FRAX tool, despite fracture risk assessment often relying on a combination of aBMD and clinical risk factors. Future research should assess whether integrating FE-derived strength with FRAX or other clinical risk factors provides incremental predictive value. While this study focuses on biomechanical markers for fracture prediction, it is important to acknowledge that extra-skeletal factors such as fall risk and comorbidities also play a significant role. Since these factors are not fully accounted for in our models, there is an inherent ceiling effect on the AUC values that can be achieved. Nevertheless, the longest follow-up for femoral strength estimation using FE analysis has been 5 years, and no studies have examined whether femoral strength can predict MOFs. Our study addresses this gap by assessing the long-term predictive value of FEA strength compared to aBMD.
In conclusion, time-dependent measures of AUC, sensitivity, and specificity have demonstrated that femoral strength can be effectively used over longer follow-up periods compared to aBMD, providing reliable discriminatory accuracy in predicting future HFs and MOFs. Future work should focus on health economic assessments to determine whether the statistical superiority of femoral strength observed in this study translates into meaningful clinical benefits. The limitations of the WHO-defined aBMD thresholds demonstrate the need for population-specific and time-dependent adjustments to improve predictive accuracy. Further research with diverse cohorts and long-term follow-ups is needed to validate these findings. Additionally, existing CT scans from over a decade ago could be used for opportunistic risk assessment, potentially refining screening strategies for osteoporotic fracture risk management.
Acknowledgements
The research was conducted at the Future Health Technologies at the Singapore-ETH Centre, which was established collaboratively between ETH Zürich and the National Research Foundation Singapore.
Anzeige
Declarations
Conflict of interest
None.
Statement of human and animal rights
The manuscript does not contain clinical studies or patient data.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License, which permits any non-commercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc/4.0/.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Comparison of the time-dependent discriminatory accuracy of femoral strength and bone mineral density for predicting future hip and major osteoporotic fractures: a 16-year follow-up of the AGES-Reykjavik cohort
Verfasst von
Anitha D. Praveen
Dheeraj Jha
Alexander Baker
Ingmar Fleps
Páll Björnsson
Lotta María Ellingsen
Thor Aspelund
Sigurdur Sigurdsson
Vilmundur Gudnason
Halldór Pálsson
David Matchar
Fjola Johannesdottir
Stephen J. Ferguson
Benedikt Helgason
Small RE (2005) Uses and limitations of bone mineral density measurements in the management of osteoporosis. MedGenMed 7:3PubMedPubMedCentral
5.
Beaudoin C, Moore L, Gagné M et al (2019) Performance of predictive tools to identify individuals at risk of non-traumatic fracture: a systematic review, meta-analysis, and meta-regression. Osteoporos Int 30:721–740. https://doi.org/10.1007/s00198-019-04919-6CrossRefPubMed
Adams AL, Fischer H, Kopperdahl DL et al (2018) Osteoporosis and hip fracture risk from routine computed tomography scans: the fracture, osteoporosis, and CT utilization study (FOCUS). J Bone Miner Res 33:1291–1301. https://doi.org/10.1002/jbmr.3423CrossRefPubMed
8.
Bhattacharya P, Altai Z, Qasim M, Viceconti M (2019) A multiscale model to predict current absolute risk of femoral fracture in a postmenopausal population. Biomech Model Mechanobiol 18:301–318. https://doi.org/10.1007/s10237-018-1081-0CrossRefPubMed
9.
Falcinelli C, Schileo E, Balistreri L et al (2014) Multiple loading conditions analysis can improve the association between finite element bone strength estimates and proximal femur fractures: a preliminary study in elderly women. Bone 67:71–80. https://doi.org/10.1016/j.bone.2014.06.038CrossRefPubMed
10.
Fleps I, Pálsson H, Baker A et al (2022) Finite element derived femoral strength is a better predictor of hip fracture risk than aBMD in the AGES Reykjavik study cohort. Bone 154:116219. https://doi.org/10.1016/j.bone.2021.116219CrossRefPubMed
Qasim M, Farinella G, Zhang J et al (2016) Patient-specific finite element estimated femur strength as a predictor of the risk of hip fracture: the effect of methodological determinants. Osteoporos Int 27:2815–2822. https://doi.org/10.1007/s00198-016-3597-4CrossRefPubMedPubMedCentral
13.
Kopperdahl DL, Aspelund T, Hoffmann PF et al (2014) Assessment of incident spine and hip fractures in women and men using finite element analysis of CT scans. J Bone Miner Res 29:570–580. https://doi.org/10.1002/jbmr.2069CrossRefPubMed
14.
Yosibash Z, Trabelsi N, Buchnik I et al (2023) Hip fracture risk assessment in elderly and diabetic patients: combining autonomous finite element analysis and machine learning. J of Bone Mineral Res 38:876–886. https://doi.org/10.1002/jbmr.4805CrossRef
15.
Keyak JH, Sigurdsson S, Karlsdottir G et al (2011) Male-female differences in the association between incident hip fracture and proximal femoral strength: a finite element analysis study. Bone 48:1239–1245. https://doi.org/10.1016/j.bone.2011.03.682CrossRefPubMedPubMedCentral
16.
Enns-Bray WS, Bahaloo H, Fleps I et al (2019) Biofidelic finite element models for accurately classifying hip fracture in a retrospective clinical study of elderly women from the AGES Reykjavik cohort. Bone 120:25–37. https://doi.org/10.1016/j.bone.2018.09.014CrossRefPubMed
17.
Harris TB, Launer LJ, Eiriksdottir G et al (2007) Age, Gene/environment susceptibility-reykjavik study: multidisciplinary applied phenomics. Am J Epidemiol 165:1076–1087. https://doi.org/10.1093/aje/kwk115CrossRefPubMed
18.
Bjornsson PA, Baker A, Fleps I et al (2023) Fast and robust femur segmentation from computed tomography images for patient-specific hip fracture risk screening. Comput Methods Biomech Biomed Eng: Imaging Vis 11:253–265. https://doi.org/10.1080/21681163.2022.2068160CrossRef
19.
Enns-Bray WS, Bahaloo H, Fleps I et al (2018) Material mapping strategy to improve the predicted response of the proximal femur to a sideways fall impact. J Mech Behav Biomed Mater 78:196–205. https://doi.org/10.1016/j.jmbbm.2017.10.033CrossRefPubMed
20.
Blanche P, Dartigues J, Jacqmin-Gadda H (2013) Estimating and comparing time-dependent areas under receiver operating characteristic curves for censored event times with competing risks. Stat Med 32:5381–5397. https://doi.org/10.1002/sim.5958CrossRefPubMed
21.
Li X, Yin Z, Li L (2016) tdROC: nonparametric estimation of time-dependent ROC, Brier score, and survival difference from right censored time-to-event data with or without competing risks. https://doi.org/10.32614/CRAN.package.tdROC
Binkley N, Kiebzak GM, Lewiecki EM et al (2005) Recalculation of the NHANES database SD improves T-score agreement and reduces osteoporosis prevalence. J Bone Miner Res 20:195–201. https://doi.org/10.1359/JBMR.041115CrossRefPubMed
Bach M, Werner A, Żywiec J, Pluskiewicz W (2017) The study of under- and over-sampling methods’ utility in analysis of highly imbalanced data on osteoporosis. Inf Sci 384:174–190. https://doi.org/10.1016/j.ins.2016.09.038CrossRef
27.
Stone KL, Seeley DG, Lui L-Y et al (2003) BMD at multiple sites and risk of fracture of multiple types: long-term results from the study of osteoporotic fractures. J Bone Miner Res 18:1947–1954. https://doi.org/10.1359/jbmr.2003.18.11.1947CrossRefPubMed
Turner DA, Khioe RFS, Shepstone L et al (2018) The cost-effectiveness of screening in the community to reduce osteoporotic fractures in older women in the UK: economic evaluation of the SCOOP study. J Bone Miner Res 33:845–851. https://doi.org/10.1002/jbmr.3381CrossRefPubMed
Grundlagenwissen der Arthroskopie und Gelenkchirurgie erweitert durch Fallbeispiele, Videos und Abbildungen. Zur Fortbildung und Wissenserweiterung, verfasst und geprüft von Expertinnen und Experten der Gesellschaft für Arthroskopie und Gelenkchirurgie (AGA).
Lange Nüchternzeiten vor einer OP können unerwünschte gesundheitliche Konsequenzen haben. Ein multidisziplinäres internationales Team hat deswegen in einem Delphi-Prozess acht Empfehlungen formuliert, die die perioperative Versorgung verbessern sollen.
Es ist jetzt auch schon wieder rund 600 Millionen Jahre her, dass sich mit dem Erscheinen der Zweiseitentiere die Lateralität ausgebildet hat. In chirurgischen Fächern wie der Orthopädie hat das Folgen bis auf den heutigen Tag. Linkshänder wissen, wovon die Rede ist.
Glukokortikoide, die zu Beginn einer krankheitsmodifizierenden Rheumatherapie begleitend eingesetzt werden, lassen sich in den allermeisten Fällen planmäßig ausschleichen und absetzen: Das Ergebnis der ARCTIC-Studie spricht diesbezüglich für die europäische und gegen die US-amerikanische Leitlinie.
Laut einer Querschnittstudie leiden rund 7% der in Deutschland lebenden über 16-Jährigen unter chronischen Schmerzen, die ihren Alltag stark beeinträchtigen. Außer biologischen scheinen auch psychische und soziale Faktoren mit sogenanntem High-Impact Chronic Pain assoziiert zu sein.
