Vertebral fracture assessment (VFA) for osteoporosis screening in US postmenopausal women: is it cost-effective?

Cost-Effectiveness of Artificial Intelligence-Based Opportunistic Compression Fracture Screening of Existing Radiographs

PURPOSE

Osteoporotic vertebral compression fractures (OVCFs) are a highly prevalent source of morbidity and mortality, and preventive treatment has been demonstrated to be both effective and cost effective. To take advantage of the information available on existing chest and abdominal radiographs, the authors' study group has developed software to access these radiographs for OVCFs with high sensitivity and specificity using an established artificial intelligence deep learning algorithm. The aim of this analysis was to assess the potential cost-effectiveness of implementing this software.

METHODS

A deterministic expected-value cost-utility model was created, combining a tree model and a Markov model, to compare the strategies of opportunistic screening for OVCFs against usual care. Total costs and total quality-adjusted life-years were calculated for each strategy. Screening and treatment costs were considered from a limited societal perspective, at 2022 prices.

RESULTS

In the base case, assuming a cost of software implantation of $10 per patient screened, the screening strategy dominated the nonscreening strategy: it resulted in lower cost and increased quality-adjusted life-years. The lower cost was due primarily to the decreased costs associated with fracture treatment and decreased probability of requiring long-term care in patients who received preventive treatment. The screening strategy was dominant up to a cost of $46 per patient screened.

CONCLUSIONS

Artificial intelligence-based opportunistic screening for OVCFs on existing radiographs can be cost effective from a societal perspective.

Analysis of Long-Term Medical Expenses in Vertebral Fracture Patients

Background

The objective of this study was to analyze the direct medical expenses of a vertebral fracture cohort (VC) and a matched cohort (MC) over 5 years preceding and following the fracture, analyze the duration of the rise in medical expenses due to the fracture, and examine whether the expenses vary with age group, utilizing a national claims database.

Methods

Subjects with vertebral fractures and matched subjects were chosen from the National Health Insurance Service Sample cohort (NHIS-Sample) of South Korea. Patients with vertebral fractures were either primarily admitted to acute care hospitals (index admissions) or those who received kyphoplasty or vertebroplasty during the follow-up period (2002-2015). A risk-set matching was performed using 1 : 5 random sampling to simulate a real-world situation. Individual-level direct medical expenses per quarter were calculated for 5 years prior and subsequent to the vertebral fracture. In this analysis using a comparative interrupted time series design, we examined the direct medical expenses of a VC and an MC.

Results

A total of 3,923 incident vertebral fracture patients and 19,615 matched subjects were included in this study. The mean age was 75.5 ± 7.4 years, and 69.5% were women. The mean difference in medical expenses between the two groups increased steadily before the fracture. The medical expenses of the VC peaked in the first quarter following the fracture. The cost changes were 1.82 times higher for the VC than for the MC (95% confidence interval, 1.62-2.04; p < 0.001) in the first year. Subsequently, there were no differential changes in medical expenses between the two groups (p > 0.05). In the < 70-year subgroup, there were no differential changes in medical expenses between the two groups (p > 0.05). However, in the ≥ 80-year subgroup, the cost changes for the VC were higher than those for the MC up to 5 years after time zero.

Conclusions

Based on our study results, we suggest that health and medical policies for vertebral fractures should be designed to last up to approximately 1 year after the fracture. Health policies should be differentiated according to age group.

Adjusting FRAX Estimates of Fracture Probability Based on a Positive Vertebral Fracture Assessment

Importance

FRAX is the most widely used and validated fracture risk prediction tool worldwide. Vertebral fractures, which are an indicator of subsequent osteoporotic fractures, can be identified using dual-energy x-ray absorptiometry (DXA) vertebral fracture assessment (VFA).

Objective

To assess the calibration of FRAX and develop a simple method for improving FRAX-predicted fracture probability in the presence of VFA-identified fracture.

Design, Setting, and Participants

This prognostic study analyzed the DXA and VFA results of all individuals who underwent a VFA between March 31, 2010, and March 31, 2018, who were included in the Manitoba Bone Mineral Density Registry. These individuals were randomly assigned to either the development cohort or validation cohort. A modified algorithm-based qualitative approach was used by expert readers to code VFAs as positive (≥1 vertebral fractures detected) or negative (0 vertebral fracture detected). Statistical analysis was conducted from August 7, 2022, to May 22, 2023.

Exposures

FRAX scores for major osteoporotic fracture (MOF) and hip fracture were calculated with or without VFA results.

Main Outcomes and Measures

Incident fractures and death were ascertained using linked population-based health care provincial data. Cumulative incidence curves for MOF and hip fracture were constructed, including competing mortality, to predict the 10-year observed risk of fracture. The observed probability was compared with FRAX-predicted fracture probability with and without VFA results and recalibrated FRAX from derived multipliers.

Results

The full cohort of 11 766 individuals was randomly allocated to the development cohort (n = 7854; 7349 females [93.6%]; mean [SD] age, 75.7 [6.8] years) or the validation cohort (n = 3912; 3713 females [94.9%]; mean [SD] age, 75.5 [6.9] years). Over a mean (SD) observation time of 3.8 (2.3) years, with the longest observation at 7.5 years, FRAX was well calibrated in subgroups with negative VFA results. For individuals without a prior clinical fracture but with a positive VFA result, the 10-year FRAX-predicted MOF probability was 16.3% (95% CI, 15.7%-16.8%) without VFA information and 23.4% (95% CI, 22.7%-24.1%) with VFA information. The observed 10-year probabilities were 26.9% (95% CI, 26.0%-27.8%) and 11.2% (95% CI, 10.3%-12.1%), respectively, resulting in recalibration multipliers of 1.15 (95% CI, 0.87-1.43) for MOF and 1.31 (95% CI, 0.75-1.87) for hip fracture. For individuals with a prior clinical fracture and a positive VFA result, the 10-year FRAX-predicted probabilities were 25.0% (95% CI, 24.2%-25.7%) for MOF and 9.3% (95% CI, 8.7%-10.0%) for hip fracture. The observed 10-year probabilities were 38.1% (95% CI, 37.0%-39.1%) for MOF and 16.4% (95% CI, 15.4%-17.4%) for hip fracture, resulting in a recalibration multiplier of 1.53 (95% CI, 1.10-1.96) for MOF and 1.76 (95% CI, 1.17-2.35) for hip fracture. Good calibration (>0.90) was confirmed using the derived multipliers in the validation cohort.

Conclusions and Relevance

Results of this prognostic study suggest that FRAX underestimated fracture risk in patients with VFA-identified fractures. Simple multipliers could recover FRAX calibration in individuals with VFA-identified fractures.

Estimating the future clinical and economic benefits of improving osteoporosis diagnosis and treatment among postmenopausal women across eight European countries

A population-level, cross-sectional model was developed to estimate the clinical and economic burden of osteoporosis among women (≥ 70 years) across eight European countries. Results demonstrated that interventions aimed at improving fracture risk assessment and adherence would save 15.2% of annual costs in 2040.

PURPOSE

Osteoporosis is associated with significant clinical and economic burden, expected to further increase with an ageing population. This modelling analysis assessed clinical and economic outcomes under different hypothetical disease management interventions to reduce this burden.

METHODS

A population-level, cross-sectional cohort model was developed to estimate numbers of incident fractures and direct costs of care among women (≥ 70 years) in eight European countries under different hypothetical interventions: (1) an improvement in the risk assessment rate, (2) an improvement in the treatment adherence rate and (3) a combination of interventions 1 and 2. A 50% improvement from the status quo, based on existing disease management patterns, was evaluated in the main analysis; scenario analyses evaluated improvement of either 10 or 100%.

RESULTS

Based on existing disease management patterns, a 44% increase in the annual number of fractures and costs was predicted from 2020 to 2040: from 1.2 million fractures and €12.8 billion in 2020 to 1.8 million fractures and €18.4 billion in 2040. Intervention 3 provided the greatest fracture reduction and cost savings (a decrease of 17.9% and 15.2% in fractures and cost, respectively) in 2040 compared with intervention 1 (decreases of 8.7% and 7.0% in fractures and cost, respectively) and intervention 2 (10.0% and 8.8% reductions in fracture and cost, respectively). Scenario analyses showed similar patterns.

CONCLUSION

These analyses suggest that interventions which improve fracture risk assessment and adherence to treatments would relieve the burden of osteoporosis, and that a combination strategy would achieve greatest benefits.

Routine Use of Lateral Vertebral Assessment With DXA Scan for Detection of Silent But Debilitating Vertebral Fractures

PURPOSE OF THE REPORT

Reduced bone mineral density is a major public health dilemma with high prevalence. Vertebral fracture (VF) is an independent risk factor for fragility fracture. Lateral vertebral assessment (LVA) in dual-energy x-ray absorptiometry is a reliable, low-radiation, accurate, and cost-effective method for VF assessment.

PATIENTS AND METHODS

Five hundred seventy-five scans of oncologic and nononcologic patients were retrospectively reviewed irrespective of age or sex. Patients' symptoms, bone mineral density, and risk factors were also evaluated. Scans in which LVA was not acquired or had previously known VFs were excluded.

RESULTS

The mean age of patients was 66 ± 11.5 years. Eleven percent of patients had VFs on LVA, of which 7 were excluded due to known VFs. Ten percent had new VFs, most of whom were women (n = 42). The most common risk factor was secondary osteoporosis in women and rheumatoid arthritis in men. Sixty-eight percent of the patients had solitary fractures, whereas 32% had multiple fractures. Most of these patients had underlying osteopenia (n = 19). FRAX was calculated twice: once with the history of personal fracture marked and the other time unmarked as these would not have been discovered if LVA was not acquired. Statistically significant mean percent difference of 5.4% was found in probability of major osteoporotic fracture and 2.1% in the mean risk of hip fracture.

CONCLUSIONS

In our population, 10% patients had unsuspected VFs on LVA in dual-energy x-ray absorptiometry scan. Most of these were nononcologic patients with associated risk factors. Based on the FRAX tool, there is a significant difference in the 10-year risk of fracture when unsuspected fractures discovered on LVA are marked.

Romosozumab for the treatment of postmenopausal women at high risk of fracture

INTRODUCTION

Romosozumab is a monoclonal antibody that binds to sclerostin (an inhibitor of the Wingless-related integration site (Wnt) signaling pathway). It is a new osteoanabolic drug that simultaneously increases bone formation and decreases bone resorption. It has recently been approved by the US and EU authorities in postmenopausal women with at high risk of fractures.

AREAS COVERED

The literature on romosozumab in preclinical and in phase II and III clinical studies has been reviewed about the effect on bone, bone markers, and fracture reduction and its safety.

EXPERT OPINION

Compared to antiresorptive agents, its unique mechanism of action results in a quicker and greater increase in bone mineral density, it repairs and restores trabecular and cortical bone microarchitecture, and reduces fracture risk more rapidly and more effectively than alendronate, with persisting effects for at least two years after transition to antiresorptive agents. This finding has introduced the concept that, in patients at very high risk of fractures, the optimal sequence of treatment is to start with an osteoanabolic agent, followed by a potent AR drug. Recent national and international guidelines recommend the use of romosozumab as an initial treatment in patients at very high fracture risk without a history of stroke or myocardial infarction.

Recommendations on the post-acute management of the osteoporotic fracture - Patients with "very-high" Re-fracture risk

Osteoporosis is a systemic skeletal disease where there is low bone mass and deterioration of bone microarchitecture, leading to an increased risk of a fragility fracture. The aim of this clinical guideline from Fragility Fracture Network Hong Kong SAR, is to provide evidence-based recommendations on the post-acute treatment of the osteoporotic fracture patient that presents for clinical care at the Fracture Liaison Service (FLS). It is now well established that the incidence of a second fracture is especially high after the first 2 years of the initial osteoporotic fracture. Therefore, the recent osteoporotic fracture should be categorized as “very-high” re-fracture risk. Due to the significant number of silent vertebral fractures in the elderly population, it is also recommended that vertebral fracture assessment (VFA) should be incorporated into FLS. This would have diagnostic and treatment implications for the osteoporotic fracture patient. The use of a potent anti-osteoporotic agent, and preferably an anabolic followed by an anti-resorptive agent should be considered, as larger improvements in BMD is strongly associated with a reduction in fractures. Managing other risk factors including falls and sarcopenia are imperative during rehabilitation and prevention of another fracture. Although of low incidence, one should remain vigilant of the atypical femoral fracture. The aging population is increasing worldwide, and it is expected that the treatment of osteoporotic fractures will be routine. The recommendations are anticipated to aid in the daily clinical practice for clinicians.

The Translational potential of this article

Fragility fractures have become a common encounter in clinical practise in the hospital setting. This article provides recommendations on the post-acute management of fragility fracture patients at the FLS.

Development of a manufacturer-independent convolutional neural network for the automated identification of vertebral compression fractures in vertebral fracture assessment images using active learning

BACKGROUND

Convolutional neural networks (CNNs) can identify vertebral compression fractures in GE vertebral fracture assessment (VFA) images with high balanced accuracy, but performance against Hologic VFAs is unknown. To obtain good classification performance, supervised machine learning requires balanced and labeled training data. Active learning is an iterative data annotation process with the ability to reduce the cost of labeling medical image data and reduce class imbalance.

PURPOSE

To train CNNs to identify vertebral fractures in Hologic VFAs using an active learning approach, and evaluate the ability of CNNs to generalize to both Hologic and GE VFA images.

METHODS

VFAs were obtained from the OsteoLaus Study (labeled Hologic Discovery A, n = 2726), the Manitoba Bone Mineral Density Program (labeled GE Prodigy and iDXA, n = 12,742), and the Canadian Longitudinal Study on Aging (CLSA, unlabeled Hologic Discovery A, n = 17,190). Unlabeled CLSA VFAs were split into five equal-sized partitions (n = 3438) and reviewed sequentially using active learning. Based on predicted fracture probability, 17.6% (n = 3032) of the unlabeled VFAs were selected for expert review using the modified algorithm-based qualitative (mABQ) method. CNNs were simultaneously trained on Hologic, GE dual-energy and GE single-energy VFAs. Two ensemble CNNs were constructed using the maximum and mean predicted probability from six separately trained CNNs that differed due to stochastic variation. CNNs were evaluated against the OsteoLaus validation set (n = 408) during the active learning process; ensemble performance was measured against the OsteoLaus test set (n = 819).

RESULTS

The baseline CNN, prior to active learning, achieved 55.0% sensitivity, 97.9% specificity, 57.9% positive predictive value (PPV), F1-score 56.4%. Through active learning, 2942 CLSA Hologic VFAs (492 fractures) were added to the training data-increasing the proportion of Hologic VFAs with fractures from 4.2% to 12.5%. With active learning, CNN performance improved to 80.0% sensitivity, 99.7% specificity, 94.1% PPV, F1-score 86.5%. The CNN maximum ensemble achieved 91.9% sensitivity (100% for grade 3 and 95.5% for grade 2 fractures), 99.0% specificity, 81.0% PPV, F1-score 86.1%.

CONCLUSION

Simultaneously training on a composite dataset consisting of both Hologic and GE VFAs allowed for the development of a single manufacturer-independent CNN that generalized to both scanner types with good classification performance. Active learning can reduce class imbalance and produce an effective medical image classifier while only labeling a subset of available unlabeled image data-thereby reducing the time and cost required to train a machine learning model.

The prevalence of vertebral fractures among Indian perimenopausal women and its association with ovarian biomarkers

INTRODUCTION

There is dearth of data on prevalent vertebral fractures in perimenopausal women in India and limited literature on the utility of FSH, AMH and estradiol in evaluating bone health them. The objective was to study the prevalence of vertebral fractures (VF) and to assess the utility of FSH, estradiol and AMH in predicting them in Indian perimenopausal women MATERIALS AND METHODS: It was a cross-sectional study. Perimenopausal women aged 40-49 years underwent assessment for prevalent vertebral fractures, bone mineral density (BMD) and trabecular bone score (TBS). Utility of serum FSH, estradiol and AMH in predicting prevalent vertebral fractures was also assessed.

RESULTS

A total of 300 perimenopausal women with mean (SD) age of 43.2 (2.8) years was recruited and 18% had moderate-severe VF. Mean (SD) serum AMH was lower in perimenopausal women with VF as compared to those without fractures [0.752 (0.594) vs 1.023 (0.704) P = 0.006]. AMH showed significant positive correlation with TBS (r = 0.3; P < 0.001) and BMD at the femoral neck (r = 0.2; P < 0.001) and lumbar spine (r = 0.3; P < 0.001).On ROC analysis, AMH demonstrated good performance in predicting prevalent VF with an AUC of 0.800 (95% CI 0.705-0.880) and a sensitivity of 85% and specificity of 60% at a cut-off of 1.12 ng/mL. On an exploratory multivariate logistic regression analysis, AMH significantly predicted prevalent fractures with an adjusted OR (OR) of 1.85 (95% CI: 1.03-3.00; P = 0.04). The performance of FSH and estradiol in predicting prevalent fractures was sub-optimal.

CONCLUSION

About one-fifth of the study subjects had prevalent vertebral fractures. AMH may be a menstrual cycle independent biomarker and may reflect bone loss in perimenopausal women. Further prospective studies are needed to validate these findings.

Inter-observer agreement of vertebral fracture assessment with dual-energy x-ray absorptiometry equipment

PURPOSE

To investigate the time and effort needed to perform vertebral morphometry, as well as inter-observer agreement for identification of vertebral fractures on vertebral fracture assessment (VFA) images.

METHODS

Ninety-six images were retrospectively selected, and three radiographers independently performed semi-automatic 6-point morphometry. Fractures were identified and graded using the Genant classification. Time needed to annotate each image was recorded, and reader fatigue was assessed using a modified Simulator Sickness Questionnaire (SSQ). Inter-observer agreement was assessed per-patient and per-vertebra for detecting fractures of all grades (grades 1-3) and for grade 2 and 3 fractures using the kappa statistic. Variability in measured vertebral height was evaluated using the intraclass correlation coefficient (ICC).

RESULTS

Per-patient agreement was 0.59 for grades 1-3 fracture detection, and 0.65 for grades 2-3 only. Agreement for per-vertebra fracture classification was 0.92. Vertebral height measurements had an ICC of 0.96. Time needed to annotate VFA images ranged between 91 and 540 s, with a mean annotation time of 259 s. Mean SSQ scores were significantly lower at the start of a reading session (1.29; 95% CI: 0.81-1.77) compared to the end of a session (3.25; 95% CI: 2.60-3.90; p < 0.001).

CONCLUSION

Agreement for detection of patients with vertebral fractures was only moderate, and vertebral morphometry requires substantial time investment. This indicates that there is a potential benefit for automating VFA, both in improving inter-observer agreement and in decreasing reading time and burden on readers.

The current situation in the approach to osteoporosis in older adults in Turkey: areas in need of improvement with a model for other populations

PURPOSE

The total number of older adults in Turkey is striking, amounting to around 8 million, and this translates into considerably higher numbers of cases of osteoporosis (OP) and fractures in older adults. In this article, we outlined the current situation of OP in older adults in Turkey and investigated the differences between Turkey and a representative developed European country (Belgium), in terms of the screening, diagnosis, and treatment of OP. Our intention in this regard was to identify areas in need of improvement and subsequently to make a clear call for action to address these issues.

METHODS

Herein, considering the steps related to the OP approach, we made a complete review of the studies conducted in Turkey and compared with the literature recommendations.

RESULTS

There is a need for a national osteoporotic fracture registry; measures should be taken to improve the screening and treatment of OP in older males, such as educational activities; technicians involved in dual-energy X-ray absorptiometry (DXA) scanning should undergo routine periodic training; all DXA centers should identify center-specific least significant change values; all older adults should be considered for routine lateral dorsolumbar X-ray imaging for the screening of vertebral fractures while ordering DXA scans; the inclusion of vertebral fracture assessment (VFA) software in DXA assessments should be considered; screening using a fracture risk assessment tool (FRAX) algorithm that is specific to Turkey should be integrated; the fortification of foods with vitamin D is required; the high fracture risk by country-specific FRAX algorithm and the presence of falls/high fall risk should be integrated in reimbursement terms; and finally, more "fracture liaison services" should be established.

CONCLUSION

We suggest that the practical consideration of our suggestions will provide considerable support to the efforts for combating with the adverse consequences of OP in society. This approach can be subsequently modeled for other populations to improve the management of OP globally.

Feasibility of a generalized convolutional neural network for automated identification of vertebral compression fractures: The Manitoba Bone Mineral Density Registry

BACKGROUND

Vertebral fracture assessment (VFA) images are acquired in dual-energy (DE) or single-energy (SE) scan modes. Automated identification of vertebral compression fractures, from VFA images acquired using GE Healthcare scanners in DE mode, has achieved high accuracy through the use of convolutional neural networks (CNNs). Due to differences between DE and SE images, it is uncertain whether CNNs trained on one scan mode will generalize to the other.

PURPOSE

To evaluate the ability of CNNs to generalize between GE DE and GE SE VFA scan modes.

METHODS

12,742 GE VFA images from the Manitoba Bone Mineral Density Program, obtained between 2010 and 2017, were exported in both DE and SE modes. VFAs were classified by imaging specialists as fracture present or absent using the modified algorithm-based qualitative (mABQ) method. VFA scans were randomly divided into independent training (60%), validation (10%), and test (30%) sets. Three CNN models were constructed by training separately on DE only, SE only, and a composite dataset comprised of both SE and DE VFAs. All three trained CNN models were separately evaluated against both SE and DE test datasets.

RESULTS

Good performance was seen for CNNs trained and evaluated on the same scan mode. DE scans used for both training and evaluation (DE/DE) achieved 87.9% sensitivity, 87.4% specificity, and an area under the receiver operating characteristic curve (AUC) of 0.94. SE scans used for both training and evaluation (SE/SE) achieved 78.6% sensitivity, 90.6% specificity, AUC = 0.92. Conversely, CNNs performed poorly when evaluated on scan modes that differed from their training sets (AUC = 0.58). However, a composite CNN trained simultaneously on both SE and DE VFAs gave performance comparable to DE/DE (82.4% sensitivity, 94.3% specificity, AUC = 0.95); and provided improved performance over SE/SE (82.2% sensitivity, 92.3% specificity, AUC = 0.94). Positive predictive value was higher with the composite CNN compared with models trained solely on DE (74.5% vs. 58.7%) or SE VFAs (68.6% vs. 62.9%).

CONCLUSION

CNNs for vertebral fracture identification are highly sensitive to scan mode. Training CNNs on a composite dataset, comprised of both GE DE and GE SE VFAs, allows CNNs to generalize to both scan modes and may facilitate the development of manufacturer-independent machine learning models for vertebral fracture detection.

Machine learning for automated abdominal aortic calcification scoring of DXA vertebral fracture assessment images: A pilot study

BACKGROUND

Abdominal aortic calcification (AAC) identified on dual-energy x-ray absorptiometry (DXA) vertebral fracture assessment (VFA) lateral spine images is predictive of cardiovascular outcomes, but is time-consuming to perform manually. Whether this procedure can be automated using convolutional neural networks (CNNs), a class of machine learning algorithms used for image processing, has not been widely investigated.

METHODS

Using the Province of Manitoba Bone Density Program DXA database, we selected a random sample of 1100 VFA images from individuals qualifying for VFA as part of their osteoporosis assessment. For each scan, AAC was manually scored using the 24-point semi-quantitative scale and categorized as low (score < 2), moderate (score 2 to <6), or high (score ≥ 6). An ensemble consisting of two CNNs was developed, by training and evaluating separately on single-energy and dual-energy images. AAC prediction was performed using the mean AAC score of the two models.

RESULTS

Mean (SD) age of the cohort was 75.5 (6.7) years, 95.5% were female. Training (N = 770, 70%), validation (N = 110, 10%) and test sets (N = 220, 20%) were well-balanced with respect to baseline characteristics and AAC scores. For the test set, the Pearson correlation between the CNN-predicted and human-labelled scores was 0.93 with intraclass correlation coefficient for absolute agreement 0.91 (95% CI 0.89-0.93). Kappa for AAC category agreement (prevalence- and bias-adjusted, ordinal scale) was 0.71 (95% CI 0.65-0.78). There was complete separation of the low and high categories, without any low AAC score scans predicted to be high and vice versa.

CONCLUSIONS

CNNs are capable of detecting AAC in VFA images, with high correlation between the human and predicted scores. These preliminary results suggest CNNs are a promising method for automatically detecting and quantifying AAC.