Abstract. Introduction. Precision health aims to develop personalised and proactive strategies for predicting, preventing, and treating complex diseases such as osteoarthritis (OA), a degenerative joint disease affecting over 300 million people worldwide. Due to OA heterogeneity, which makes developing effective treatments challenging, identifying patients at risk for accelerated disease progression is essential for efficient clinical trial design and new treatment target discovery and development. Objectives. This study aims to create a trustworthy and interpretable precision health tool that predicts rapid knee OA progression based on baseline patient characteristics using an advanced automated machine learning (autoML) framework, “Autoprognosis 2.0”. Methods. All available 2-year follow-up periods of 600 patients from the FNIH OA Biomarker Consortium were analysed using “Autoprognosis 2.0” in two separate approaches, with distinct definitions of clinical outcomes: multi-class predictions (categorising patients into non-progressors, pain-only progressors, radiographic-only progressors, and both pain and radiographic progressors) and binary predictions (categorising patients into non-progressors and progressors). Models were developed using a training set of 1352 instances and all available variables (including clinical, X-ray, MRI, and biochemical features), and validated through both stratified 10-fold cross-validation and hold-out validation on a testing set of 339 instances. Model performance was assessed using multiple evaluation metrics, such as AUC-ROC, AUC-PRC, F1-score, precision, and recall. Additionally, interpretability analyses were carried out to identify important predictors of rapid disease progression. Results. Our final models yielded high accuracy scores for both multi-class predictions (AUC-ROC: 0.858, 95% CI: 0.856–0.860; AUC-PRC: 0.675, 95% CI: 0.671–0.679; F1-score: 0.560, 95% CI: 0.554–0.566) and binary predictions (AUC-ROC: 0.717, 95% CI: 0.712–0.722; AUC-PRC: 0.620, 95% CI: 0.616–0.624; F1-score: 0.676, 95% CI: 0.673–0679). Important predictors of rapid disease progression included the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) scores and MRI features. Our models were further successfully validated using a hold-out dataset, which was previously omitted from model development and training (AUC-ROC: 0.877 for multi-class predictions; AUC-ROC: 0.746 for binary predictions). Additionally, accurate ML models were developed for predicting OA progression in a subgroup of patients aged 65 or younger (AUC-ROC: 0.862, 95% CI: 0.861–0.863 for multi-class predictions; AUC-ROC: 0.736, 95% CI: 0.734–0.738 for binary predictions). Conclusions. This study presents a reliable and interpretable precision health tool for predicting rapid knee OA progression using “Autoprognosis 2.0”. Our models provide accurate predictions and offer insights into important predictors of rapid disease progression. Furthermore, the transparency and interpretability of our methods may facilitate their acceptance by clinicians and patients, enabling effective utilisation in clinical practice. Future work should focus on refining these models by increasing the sample size, integrating additional features, and using independent datasets for external validation. Declaration of Interest. (b) declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported:I declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research project

Precision health aims to develop personalised and proactive strategies for predicting, preventing, and treating complex diseases such as osteoarthritis (OA). Due to OA heterogeneity, which makes developing effective treatments challenging, identifying patients at risk for accelerated disease progression is essential for efficient clinical trial design and new treatment target discovery and development. To create a reliable and interpretable precision health tool that predicts rapid knee OA progression over a 2-year period from baseline patient characteristics using an advanced automated machine learning (autoML) framework, “Autoprognosis 2.0”. All available 2-year follow-up periods of 600 patients from the FNIH OA Biomarker Consortium were analysed using “Autoprognosis 2.0” in two separate approaches, with distinct definitions of clinical outcomes: multi-class predictions (categorising disease progression into pain and/or radiographic progression) and binary predictions. Models were developed using a training set of 1352 instances and all available variables (including clinical, X-ray, MRI, and biochemical features), and validated through both stratified 10-fold cross-validation and hold-out validation on a testing set of 339 instances. Model performance was assessed using multiple evaluation metrics. Interpretability analyses were carried out to identify important predictors of progression. Our final models yielded higher accuracy scores for multi-class predictions (AUC-ROC: 0.858, 95% CI: 0.856-0.860) compared to binary predictions (AUC-ROC: 0.717, 95% CI: 0.712-0.722). Important predictors of rapid disease progression included WOMAC scores and MRI features. Additionally, accurate ML models were developed for predicting OA progression in a subgroup of patients aged 65 or younger. This study presents a reliable and interpretable precision health tool for predicting rapid knee OA progression. Our models provide accurate predictions and, importantly, allow specific predictors of rapid disease progression to be identified. Furthermore, the transparency and explainability of our methods may facilitate their acceptance by clinicians and patients, enabling effective translation to clinical practice

Total knee and hip arthroplasty (TKA and THA) are the most commonly performed surgical procedures, the costs of which constitute a significant healthcare burden. Improving access to care for THA/TKA requires better efficiency. It is hypothesized that this may be possible through a two-stage approach that utilizes prediction of surgical time to enable optimization of operating room (OR) schedules. Data from 499,432 elective unilateral arthroplasty procedures, including 302,490 TKAs, and 196,942 THAs, performed from 2014-2019 was extracted from the American College of Surgeons (ACS) National Surgical and Quality Improvement (NSQIP) database. A deep multilayer perceptron model was trained to predict duration of surgery (DOS) based on pre-operative clinical and biochemical patient factors. A two-stage approach, utilizing predicted DOS from a held out “test” dataset, was utilized to inform the daily OR schedule. The objective function of the optimization was the total OR utilization, with a penalty for overtime. The scheduling problem and constraints were simulated based on a high-volume elective arthroplasty centre in Canada. This approach was compared to current patient scheduling based on mean procedure DOS. Approaches were compared by performing 1000 simulated OR schedules. The predict then optimize approach achieved an 18% increase in OR utilization over the mean regressor. The two-stage approach reduced overtime by 25-minutes per OR day, however it created a 7-minute increase in underutilization. Better objective value was seen in 85.1% of the simulations. With deep learning prediction and mathematical optimization of patient scheduling it is possible to improve overall OR utilization compared to typical scheduling practices. Maximizing utilization of existing healthcare resources can, in limited resource environments, improve patient's access to arthritis care by increasing patient throughput, reducing surgical wait times and in the immediate future, help clear the backlog associated with the COVID-19 pandemic

Background. Although there are predictive equations that estimate the total fat mass obtained from multiple-site ultrasound (US) measurements, the predictive equation of total fat mass has not been investigated solely from abdominal subcutaneous fat thickness. Therefore, the aims of this study were; (1) to develop regression-based prediction equations based on abdominal subcutaneous fat thickness for predicting fat mass in young- and middle-aged adults, and (2) to investigate the validity of these equations to be developed. Methods. The study was approved by the Local Research Ethics Committee (Decision number: GO 19/788). Twenty-seven males (30.3 ± 8.7 years) and eighteen females (32.4 ± 9.5 years) were randomly divided into two groups as the model prediction group (19 males and 12 females) and the validation group (8 males and 6 females). Total body fat mass was determined by dual-energy X-ray absorptiometry (DXA). Abdominal subcutaneous fat thickness was measured by US. The predictive equations for total fat mass from US were determined as fat thickness (in mm) × standing height (in m). Statistical analyses were performed using R version 4.0.0. The association between the total fat mass and the abdominal subcutaneous fat thickness was interpreted using the Pearson test. The linear regression analysis was used to predict equations for total body fat mass from the abdominal subcutaneous fat thickness acquired by US. Then these predictive equations were applied to the validation group. The paired t-test was used to examine the difference between the measured and the predicted fat masses, and Lin's concordance correlation coefficient (CCC) was used as a further measure of agreement. Results. There was a significant positive moderate correlation between the total fat mass and the abdominal subcutaneous fat thickness × height in the model prediction group of males (r = 0.588, p = 0.008), whereas significant positive very strong correlation was observed in the model prediction group of females (r = 0.896, p < 0.001). Predictive equations for DXA-measured total body fat mass from abdominal subcutaneous fat thickness using US were as follows: for males “Fat mass-DXA = 0.276 × (Fat thickness-US × Height) + 17.221” (R. 2. = 0.35, SEE = 3.6, p = 0.008); for females “Fat mass-DXA = 0.694 x (Fat thickness-US × Height) + 7.085” (R. 2. = 0.80, SEE = 2.8, p < 0.001). When fat mass prediction equations were applied to the validation groups, measured- and estimated-total fat masses in males and females were found similar (p = 0.9, p = 0.5, respectively). A good level of agreement between measurements in males and females was attained (CCC = 0.84, CCC = 0.76, respectively). Conclusion. We developed and validated prediction equations that are convenient for determining total fat masses in young- and middle-aged adults using abdominal subcutaneous fat thickness obtained from the US. The abdominal subcutaneous fat thickness obtained from a single region by US might provide a noninvasive quick and easy evaluation not only in clinical settings but also on the field

Although 3D-printed porous dental implants may possess improved osseointegration potential, they must exhibit appropriate fatigue strength. Finite element analysis (FEA) has the potential to predict the fatigue life of implants and accelerate their development. This work aimed at developing and validating an FEA-based tool to predict the fatigue behavior of porous dental implants. Test samples mimicking dental implants were designed as 4.5 mm-diameter cylinders with a fully porous section around bone level. Three porosity levels (50%, 60% and 70%) and two unit cell types (Schwarz Primitive (SP) and Schwarz W (SW)) were combined to generate six designs that were split between calibration (60SP, 70SP, 60SW, 70SW) and validation (50SP, 50SW) sets. Twenty-eight samples per design were additively manufactured from titanium powder (Ti6Al4V). The samples were tested under bending compression loading (ISO 14801) monotonically (N=4/design) to determine ultimate load (F. ult. ) (Instron 5866) and cyclically at six load levels between 50% and 10% of F. ult. (N=4/design/load level) (DYNA5dent). Failure force results were fitted to F/F. ult. = a(N. f. ). b. (Eq1) with N. f. being the number of cycles to failure, to identify parameters a and b. The endurance limit (F. e. ) was evaluated at N. f. = 5M cycles. Finite element models were built to predict the yield load (F. yield. ) of each design. Combining a linear correlation between FEA-based F. yield. and experimental F. ult. with equation Eq1 enabled FEA-based prediction of F. e. . For all designs, F. e. was comprised between 10% (all four samples surviving) and 15% (at least one failure) of F. ult. The FEA-based tool predicted F. e. values of 11.7% and 12.0% of F. ult. for the validation sets of 50SP and 50SW, respectively. Thus, the developed FEA-based workflow could accurately predict endurance limit for different implant designs and therefore could be used in future to aid the development of novel porous implants. Acknowledgements: This study was funded by EU's Horizon 2020 grant No. 953128 (I-SMarD). We gratefully acknowledge the expert advice of Prof. Philippe Zysset

Background. There is a 20% dissatisfaction rate with knee replacements. Calls for tools that can pre-operatively identify patients at risk of being dissatisfied postoperatively have been widespread. However, it is unclear what sort of information patients would want from such a tool, how it would affect their decision making process, and at what part of the pathway such a tool should be used. Methods. Using focus groups involving 12 participants and in-depth interviews with 10 participants, we examined the effect outcome prediction has by providing fictitious predictions to patients at different stages of treatment. A qualitative analysis of themes, based on a constant comparative method, is used to analyse the data. This study was approved by the Dyfed Powys Research Ethics Committee (13/WA/0140). Results. Our results demonstrate several interesting findings. Firstly, patients who have received information from friends and family are unwilling to adjust their expectation of outcome down (i.e. to a worse outcome), but highly willing to adjust it up (to a better outcome). This is an example of the optimism bias, and suggests the effect on expectation of any poor outcome prediction would be blunted. Secondly, patients generally wanted a “bottom line” outcome, rather than lots of detail. Thirdly, patients who were earlier in their treatment for osteoarthritis were more likely to find the information useful, and for it to affect their decision, than patients later in their pathway. Conclusion. An outcome prediction tool would have most effect targeted towards people at the start of their treatment pathway, with a “bottom line” prediction of outcome. However, any effect on expectation and decision making of a poor outcome prediction is likely to be blunted by the optimism bias. Level of Evidence. 4

During OA the homeostasis of healthy articular chondrocytes is dysregulated, which leads to a phenotypical transition of the cells, further influenced by external stimuli. Chondrocytes sense those stimuli, integrate them at the intracellular level and respond by modifying their secretory and molecular state. This process is controlled by a complex interplay of intracellular factors. Each factor is influenced by a myriad of feedback mechanisms, making the prediction of what will happen in case of external perturbation challenging. Hampering the hypertrophic phenotype has emerged as a potential therapeutic strategy to help OA patients (Ripmeester et al. 2018). Therefore, we developed a computational model of the chondrocyte's underlying regulatory network (RN) to identify key regulators as potential drug targets. A mechanistic mathematical model of articular chondrocyte differentiation was implemented with a semi-quantitative formalism. It is composed of a protein RN and a gene RN(GRN) and developed by combining two strategies. First, we established a mechanistic network based on accumulation of decades of biological knowledge. Second, we combined that mechanistic network with data-driven modelling by inferring an OA-GRN using an ensemble of machine learning methods. This required a large gene expression dataset, provided by distinct public microarrays merged through an in-house pipeline for cross-platform integration. We successfully merged various micro-array experiments into one single dataset where the biological variance was predominant over the batch effect from the different technical platforms. The gain of information provided by this merge enabled us to reconstruct an OA-GRN which subsequently served to complete our mechanistic model. With this model, we studied the system's multi-stability, equating the model's stable states to chondrocyte phenotypes. The network structure explained the occurrence of two biologically relevant phenotypes: a hypertrophic-like and a healthy-like phenotype, recognized based on known cell state markers. Second, we tested several hypotheses that could trigger the onset of OA to validate the model with relevant biological phenomena. For instance, forced inflammation pushed the chondrocyte towards hypertrophy but this was partly rescued by higher levels of TGF-β. However, we could annihilate this rescue by concomitantly mimicking an increase in the ALK1/ALK5 balance. Finally, we performed a screening of in-silico (combinatorial) perturbations (inhibitions and/or over-activations) to identify key molecular factors involved in the stability of the chondrocyte state. More precisely, we looked for the most potent conditions for decreasing hypertrophy. Preliminary validation experiments have confirmed that PKA activation could decrease the hypertrophic phenotype in primary chondrocytes. Importantly the in-silico results highlighted that targeting two factors at the same time would greatly help reducing hypertrophic changes. A priori testing of conditions with in-silico models may cut time and cost of experiments via target prioritization and opens new routes for OA combinatorial therapies

Abstract

Quantitative ultrasound (QUS) is a promising tool to estimate bone structure characteristics and predict fragile fracture. The aim of this pilot cross-sectional study was to evaluate the performance of a multi-channel residual network (MResNet) based on ultrasonic radiofrequency (RF) signal to discriminate fragile fractures retrospectively in postmenopausal women.

Abstract

Abstract

Abstract

The development of an algorithm that provides accurate individualised estimates of revision risk could help patients make informed surgical treatment choices. This requires building a survival model based on fixed and modifiable risk factors that predict outcome at the individual level. Here we compare different survival models for predicting prosthesis survivorship after hip replacement for osteoarthritis using data from the National Joint Registry for England, Wales, Northern Ireland and the Isle of Man.

In this comparative study we implemented parametric and flexible parametric (FP) methods and random survival forests (RSF). The overall performance of the parametric models was compared using Akaike information criterion (AIC). The preferred parametric model and the RSF algorithm were further compared in terms of the Brier score, concordance index (C index) and calibration.

The dataset contains 327 238 hip replacements for osteoarthritis carried out in England and Wales between 2003 and 2015. The AIC value for the FP model was the lowest. The averages of survival probability estimates were in good agreement with the observed values for the FP model and the RSF algorithm. The integrated Brier score of the FP model and the RSF approach over 10 years were similar: 0.011 (95% confidence interval: 0.011–0.011). The C index of the FP model at 10 years was 59.4% (95% confidence interval: 59.4%–59.4%). This was 56.2% (56.1%–56.3%) for the RSF method.

The FP model outperformed other commonly used survival models across chosen validation criteria. However, it does not provide high discriminatory power at the individual level. Models with more comprehensive risk adjustment may provide additional insights for individual risk.

The management of displaced forearm diaphyseal fractures in adults is predominantly operative. Anatomical reduction is necessary to infer optimal motion and strength. The authors have observed an intraoperative technique where passive pronosupination is examined to assess quality of reduction as a surrogate marker for active movement.

We aimed to assess the value of this technique, but intentionally malreducing a simulated diaphyseal fracture of a radius in a cadaveric model, and measuring the effect on pronosupination.

A single cadaveric arm was prepared and pronation/supination was examined according to American Academy of Orthopaedic Surgeons guidance. A Henry approach was then performed and a transverse osteotomy achieved in the radial diaphysis. A volar locking plate was used to hold the radius in progressive amounts of translation and rotation, with pronosupaintion measured with a goniometer.

The radius could be grossly malreduced with no effect on pronation and supination until the extremes of deformity. The forearm showed more tolerance with rotational malreduction than translation. Passive pronation was more sensitive for malreduction than supination.

The use of passive pronosupination to assess quality of reduction is misleading.

Introduction

We have evaluated the circulation of the femoral head after multiple pinning for femoral neck fractures by bone SPECT.

Abstract. Objectives. Knee alignment affects both the development and surgical treatment of knee osteoarthritis. Automating femorotibial angle (FTA) and hip-knee-ankle angle (HKA) measurement from radiographs could improve reliability and save time. Further, if the gold-standard HKA from full-limb radiographs could be accurately predicted from knee-only radiographs then the need for more expensive equipment and radiation exposure could be reduced. The aim of this research is to assess if deep learning methods can predict FTA and HKA angle from posteroanterior (PA) knee radiographs. Methods. Convolutional neural networks with densely connected final layers were trained to analyse PA knee radiographs from the Osteoarthritis Initiative (OAI) database with corresponding angle measurements. The FTA dataset with 6149 radiographs and HKA dataset with 2351 radiographs were split into training, validation and test datasets in a 70:15:15 ratio. Separate models were learnt for the prediction of FTA and HKA, which were trained using mean squared error as a loss function. Heat maps were used to identify the anatomical features within each image that most contributed to the predicted angles. Results. FTA could be predicted with errors less than 3° for 99.8% of images, and less than 1° for 89.5%. HKA prediction was less accurate than FTA but still high: 95.7% within 3°, and 68.0 % within 1°. Heat maps for both models were generally concentrated on the knee anatomy and could prove a valuable tool for assessing prediction reliability in clinical application. Conclusions. Deep learning techniques could enable fast, reliable and accurate predictions of both FTA and HKA from plain knee radiographs. This could lead to cost savings for healthcare providers and reduced radiation exposure for patients

Artificial Intelligence (AI) is becoming more powerful but is barely used to counter the growth in health care burden. AI applications to increase efficiency in orthopedics are rare. We questioned if (1) we could train machine learning (ML) algorithms, based on answers from digitalized history taking questionnaires, to predict treatment of hip osteoartritis (either conservative or surgical); (2) such an algorithm could streamline clinical consultation. Multiple ML models were trained on 600 annotated (80% training, 20% test) digital history taking questionnaires, acquired before consultation. Best performing models, based on balanced accuracy and optimized automated hyperparameter tuning, were build into our daily clinical orthopedic practice. Fifty patients with hip complaints (>45 years) were prospectively predicted and planned (partly blinded, partly unblinded) for consultation with the physician assistant (conservative) or orthopedic surgeon (operative). Tailored patient information based on the prediction was automatically sent to a smartphone app. Level of evidence: IV. Random Forest and BernoulliNB were the most accurate ML models (0.75 balanced accuracy). Treatment prediction was correct in 45 out of 50 consultations (90%), p<0.0001 (sign and binomial test). Specialized consultations where conservatively predicted patients were seen by the physician assistant and surgical patients by the orthopedic surgeon were highly appreciated and effective. Treatment strategy of hip osteoartritis based on answers from digital history taking questionnaires was accurately predicted before patients entered the hospital. This can make outpatient consultation scheduling more efficient and tailor pre-consultation patient education

Currently implemented accuracy metrics in open-source libraries for segmentation by supervised machine learning are typically one-dimensional scores [1]. While extremely relevant to evaluate applicability in clinics, anatomical location of segmentation errors is often neglected. This study aims to include the three-dimensional (3D) spatial information in the development of a novel framework for segmentation accuracy evaluation and comparison between different methods. Predicted and ground truth (manually segmented) segmentation masks are meshed into 3D surfaces. A template mesh of the same anatomical structure is then registered to all ground truth 3D surfaces. This ensures all surface points on the ground truth meshes to be in the same anatomically homologous order. Next, point-wise surface deviations between the registered ground truth mesh and the meshed segmentation prediction are calculated and allow for color plotting of point-wise descriptive statistics. Statistical parametric mapping includes point-wise false discovery rate (FDR) adjusted p-values (also referred to as q-values). The framework reads volumetric image data containing the segmentation masks of both ground truth and segmentation prediction. 3D color plots containing descriptive statistics (mean absolute value, maximal value,…) on point-wise segmentation errors are rendered. As an example, we compared segmentation results of nnUNet [2], UNet++ [3] and UNETR [4] by visualizing the mean absolute error (surface deviation from ground truth) as a color plot on the 3D model of bone and cartilage of the mean distal femur. A novel framework to evaluate segmentation accuracy is presented. Output includes anatomical information on the segmentation errors, as well as point-wise comparative statistics on different segmentation algorithms. Clearly, this allows for a better informed decision-making process when selecting the best algorithm for a specific clinical application

The lateral wall thickness (LWT) in trochanteric femoral fractures is a known predictive factor for postoperative fracture stability. Currently, the AO/OTA classification uses a patient non-specific measure to assess the absolute LWT (aLWT) and distinguish stable A1.3 from unstable A2.1 fractures based on a threshold of 20.5 mm. This approach potentially results in interpatient deviations due to different bone morphologies and consequently variations in fracture stability. Therefore, the aim of this study was to explore whether a patient-specific measure for assessment of the relative LWT (rLWT) results in a more precise threshold for prediction of unstable fractures. Part 1 of the study evaluated 146 pelvic radiographs to assess left-right symmetry with regard to caput-collum-angle (CCD) and total trochanteric thickness (TTT), and used the results to establish the rLWT measurement technique. Part 2 reevaluated 202 patients from a previous study cohort to analyze their rLWT versus aLWT for optimization purposes. Findings in Part 1 demonstrated a bilateral symmetry of the femur regarding both CCD and TTT (p ≥ 0.827) allowing to mirror bone's morphology and geometry from the contralateral intact to the fractured femur. Outcomes in Part 2 resulted in an increased accuracy for the new determined rLWT threshold (50.5%) versus the standard 20.5 mm aLWT threshold, with sensitivity of 83.7% versus 82.7% and specificity 81.3% versus 77.8%, respectively. The novel patient-specific rLWT measure can be based on the contralateral femur anatomy and is a more accurate predictor of a secondary lateral wall fracture in comparison to the conventional aLWT. This study established the threshold of 50.5% rLWT as a reference value for prediction of fracture stability and selection of an appropriate implant for fixation of trochanteric femoral fractures

The primary aim was to assess the reliability of ultrasound in the assessment of humeral shaft fracture healing. The secondary aim was to estimate the accuracy of ultrasound assessment in predicting humeral shaft nonunion. Twelve patients (mean age 54yrs [20–81], 58% [n=7/12] female) with a non-operatively managed humeral diaphyseal fracture were prospectively recruited and underwent ultrasound scanning at six and 12wks post-injury. Scans were reviewed by seven blinded observers to evaluate the presence of sonographic callus. Intra- and inter-observer reliability were determined using the weighted kappa and intraclass correlation coefficient (ICC). Accuracy of ultrasound assessment in nonunion prediction was estimated by comparing scans for patients that united (n=10/12) with those that developed a nonunion (n=2/12). At both six and 12wks, sonographic callus was present in 11 patients (10 united, one developed a nonunion) and sonographic bridging callus (SBC) was present in seven patients (all united). Ultrasound assessment demonstrated substantial intra- (6wk kappa 0.75, 95% CI 0.47-1.03; 12wk kappa 0.75, 95% CI 0.46-1.04) and inter-observer reliability (6wk ICC 0.60, 95% CI 0.38-0.83; 12wk ICC 0.76, 95% CI 0.58-0.91). Absence of sonographic callus demonstrated a sensitivity of 50%, specificity 100%, positive predictive value (PPV) 100% and negative predictive value (NPV) 91% in nonunion prediction (accuracy 92%). Absence of SBC demonstrated a sensitivity of 100%, specificity 70%, PPV 40% and NPV 100% (accuracy 75%). Of three patients at risk of nonunion based on reduced radiographic callus formation (Radiographic Union Score for HUmeral fractures <8), one had SBC on 6wk ultrasound (and united) and the other two had non-bridging or absent sonographic callus (both developed a nonunion). Ultrasound assessment of humeral shaft fracture healing was reliable and predictive of nonunion, and may be a useful tool in defining the risk of nonunion among patients with reduced radiographic callus formation