Asia-Pacific consensus on long-term and sequential therapy for osteoporosis.

Objectives: This study aimed to present the Asia-Pacific consensus on long-term and sequential therapy for osteoporosis, offering evidence-based recommendations for the effective management of this chronic condition. The primary focus is on achieving optimal fracture prevention through a comprehensive, individualized approach. Methods: A panel of experts convened to develop consensus statements by synthesizing the current literature and leveraging clinical expertise. The review encompassed long-term anti-osteoporosis medication goals, first-line treatments for individuals at very high fracture risk, and the strategic integration of anabolic and antiresorptive agents in sequential therapy approaches. Results: The panelists reached a consensus on 12 statements. Key recommendations included advocating for anabolic agents as the first-line treatment for individuals at very high fracture risk and transitioning to antiresorptive agents following the completion of anabolic therapy. Anabolic therapy remains an option for individuals experiencing new fractures or persistent high fracture risk despite antiresorptive treatment. In cases of inadequate response, the consensus recommended considering a switch to more potent medications. The consensus also addressed the management of medication-related complications, proposing alternatives instead of discontinuation of treatment. Conclusions: This consensus provides a comprehensive, cost-effective strategy for fracture prevention with an emphasis on shared decision-making and the incorporation of country-specific case management systems, such as fracture liaison services. It serves as a valuable guide for healthcare professionals in the Asia-Pacific region, contributing to the ongoing evolution of osteoporosis management.

Duration of fracture prevention after zoledronate treatment in women with osteopenia: observational follow-up of a 6-year randomised controlled trial to 10 years.

BACKGROUND: We previously identified that zoledronate administered at 18-month intervals reduced fragility fractures by a third in a 6-year trial of women older than 65 years with osteopenia. This extension aims to identify the persistence of these effects. METHODS: Of the 2000 ambulant, community dwelling, postmenopausal women older than 65 years recruited in Auckland, New Zealand, with T-scores at the total hip or femoral neck in the range -1·0 to -2·5, we invited participants who received four doses of intravenous zoledronate, completed follow-up to year 6 of the core trial, did not have metabolic bone disease (other than osteoporosis), and were not using bone-active drugs into this 4-year observational study extension, during which further treatment was at the discretion of their own doctors. Participants were asked to notify study staff of any new fractures and were telephoned at 7·5 years and 9·0 years to update their health status. Participants were then invited to an onsite visit at 10·0 years. Fractures and other health events were documented at each contact and analysed in all women who entered the extension, and bone mineral density (BMD; analysed in participants without notable use of bone-active medications who attended an onsite visit at 10 years) and turnover markers (measured from fasting morning blood in a random subset of 50 participants) were measured at year 10. FINDINGS: Of the 1000 women randomly assigned to receive zoledronate in the core trial, 796 participants were eligible for the extension, of whom 762 (96%) entered the extension between Sept 24, 2015, and Dec 13, 2017. Mean follow-up duration was 4·24 years (SD 0·57, range 0·61-6·55; final follow-up on May 25, 2022). 727 (91%) of participants were assessed at 10 years. 25 women died during the extension, six withdrew for medical reasons, and four were lost to follow-up. 92 women suffered 114 non-vertebral fractures during the extension. Non-vertebral fracture rates increased from a nadir of 15 fractures per 1000 woman-years (95% CI 10-21) in the last 2 years of the core trial to 24 fractures (17-33) in years 6-8 and 42 fractures (32-53) in years 8-10, similar to that in the placebo group in the last 2 years of the core trial. Total hip BMD (relative risk per 0·1 g/cm2 0·73, 95% CI 0·57-0·93; p=0·011) and a previous history of non-vertebral fracture (1·74, 1·12-2·69; p=0·013) at year 6 predicted incident fractures but change in total hip BMD did not. Total hip BMD decreased from 4·2% above study baseline to 0·8% above baseline (p<0·0001) during the extension. Turnover markers were not useful for predicting BMD loss in individuals. Osteonecrosis of the jaw or atypical femoral fractures did not occur in any participants. INTERPRETATION: The reduced fracture rates following zoledronate in the core trial were substantially maintained for 1·5-3·5 years after the last zoledronate infusion, but not thereafter. FUNDING: Health Research Council of New Zealand.

Comparative analysis of functional diversity of rumen microbiome in bison and beef heifers.

IMPORTANCE: Ruminants play a key role in the conversion of cellulolytic plant material into high-quality meat and milk protein for humans. The rumen microbiome is the driver of this conversion, yet there is little information on how gene expression within the microbiome impacts the efficiency of this conversion process. The current study investigates gene expression in the rumen microbiome of beef heifers and bison and how transplantation of ruminal contents from bison to heifers alters gene expression. Understanding interactions between the host and the rumen microbiome is the key to developing informed approaches to rumen programming that will enhance production efficiency in ruminants.

No Effect of NSAID Use on Efficacy of Zoledronate: A Post Hoc Analysis of a Randomized Trial in Osteopenic Older Women.

Bisphosphonates are widely used for the prevention and treatment of osteoporosis. Nonsteroidal anti-inflammatory drugs (NSAIDs) are also widely used among the older population group at high risk of fractures. NSAIDs have been shown to impact on bone turnover, and a recent reanalysis of a clinical trial of clodronate found that NSAID use at baseline abrogated any effect of clodronate on either bone density (BMD) or fracture risk. To determine whether NSAIDs influence the efficacy of other bisphosphonates, we have reanalyzed our 6-year randomized controlled trial of zoledronate in 2000 osteopenic postmenopausal women. NSAID use was reported at baseline in 38% of the cohort and anytime use was reported by 65%. The evolution of the zoledronate effects on BMD were almost identical whether or not women were using NSAIDs at baseline and were significant in both subgroups at all BMD sites (p < 0.0001). The significant reduction in the risk of fracture in those allocated to zoledronate (p < 0.0001) showed no interaction with baseline use of NSAIDs (p = 0.33) nor with NSAID use at any time during the study (p = 0.28). The odds of fracture were significantly reduced in both NSAID users and nonusers. We conclude that the present analysis provides no support for the suggestion that NSAIDs interfere with the efficacy of potent bisphosphonates in terms of their effects on bone density or fracture. © 2023 The Authors. Journal of Bone and Mineral Research published by Wiley Periodicals LLC on behalf of American Society for Bone and Mineral Research (ASBMR).

How Trustworthy are Performance Evaluations for Basic Vision Tasks?

This article examines performance evaluation criteria for basic vision tasks involving sets of objects namely, object detection, instance-level segmentation and multi-object tracking. The rankings of algorithms by a criterion can fluctuate with different choices of parameters, e.g. Intersection over Union (IoU) threshold, making their evaluations unreliable. More importantly, there is no means to verify whether we can trust the evaluations of a criterion. This work suggests a notion of trustworthiness for performance criteria, which requires (i) robustness to parameters for reliability, (ii) contextual meaningfulness in sanity tests, and (iii) consistency with mathematical requirements such as the metric properties. We observe that these requirements were overlooked by many widely-used criteria, and explore alternative criteria using metrics for sets of shapes. We also assess all these criteria based on the suggested requirements for trustworthiness.

Effect of a Three-Day Course of Dexamethasone on Acute Phase Response Following Treatment With Zoledronate: A Randomized Controlled Trial.

Zoledronate is a potent intravenous bisphosphonate effective in the management of osteoporosis, Paget's disease and skeletal-related events in malignancy. Its most frequent adverse effect is the acute phase response (APR), an inflammatory reaction characterized by fever, musculoskeletal pain, headache, and nausea. This randomized, placebo-controlled, double-blind study investigated the efficacy of a three-day course of dexamethasone 4 mg daily in reducing incidence of APR. Participants (n = 60) were randomized to receive either 4 mg of oral dexamethasone 1.5 hours before zoledronate and once a day for the following 2 days, or placebo. Oral temperature was measured at baseline and three times a day for the following 3 days, and questionnaires assessing symptoms of the APR were completed at baseline and for 3 days following zoledronate. Use of anti-inflammatory medication in the 3 days following zoledronate was recorded. The primary outcome was the temperature change from baseline. There was a significant difference in the primary outcome between the dexamethasone and placebo groups (p < 0.0001), with a mean decrease in temperature of 0.10°C (95% confidence interval [CI], -0.34 to 0.14) in the dexamethasone group compared with a mean increase in temperature of 0.84°C (95% CI, 0.53-1.16) in the placebo group on the evening following zoledronate. There was also a difference in APR-related symptom score over time between the two groups (p = 0.0005), with a median change in symptom score in the dexamethasone group 1 day after zoledronate of 0 (95% CI, 0-1) compared with 3 (95% CI, 0-5) in the placebo group. An increase in temperature of ≥1°C to a temperature of >37.5°C occurred in two of 30 (6.7%) participants in the dexamethasone group compared with 14 of 30 participants (46.7%) in the placebo group (p = 0.0005). This study demonstrates that a 3-day course of dexamethasone substantially reduces the APR following zoledronate infusion. © 2023 American Society for Bone and Mineral Research (ASBMR).

Asia-pacific consensus on osteoporotic fracture prevention in postmenopausal women with low bone mass or osteoporosis but no fragility fractures.

Postmenopausal women are at significant risk for osteoporotic fractures due to their rapid bone loss. Half of all postmenopausal women will get an osteoporosis-related fracture over their lifetime, with 25% developing a spine deformity and 15% developing a hip fracture. By 2050, more than half of all osteoporotic fractures will occur in Asia, with postmenopausal women being the most susceptible. Early management can halt or even reverse the progression of osteoporosis. Consequently, on October 31, 2020, the Taiwanese Osteoporosis Association hosted the Asia-Pacific (AP) Postmenopausal Osteoporotic Fracture Prevention (POFP) consensus meeting, which was supported by the Asian Federation of Osteoporosis Societies (AFOS) and the Asia Pacific Osteoporosis Foundation (APOF). International and domestic experts developed ten applicable statements for the prevention of osteoporotic fractures in postmenopausal women with low bone mass or osteoporosis but no fragility fractures in the AP region. The experts advocated, for example, that postmenopausal women with a high fracture risk be reimbursed for pharmaceutical therapy to prevent osteoporotic fractures. More clinical experience and data are required to modify intervention tactics.

Looking Beyond Two Frames: End-to-End Multi-Object Tracking Using Spatial and Temporal Transformers.

Tracking a time-varying indefinite number of objects in a video sequence over time remains a challenge despite recent advances in the field. Most existing approaches are not able to properly handle multi-object tracking challenges such as occlusion, in part because they ignore long-term temporal information. To address these shortcomings, we present MO3TR: a truly end-to-end Transformer-based online multi-object tracking (MOT) framework that learns to handle occlusions, track initiation and termination without the need for an explicit data association module or any heuristics. MO3TR encodes object interactions into long-term temporal embeddings using a combination of spatial and temporal Transformers, and recursively uses the information jointly with the input data to estimate the states of all tracked objects over time. The spatial attention mechanism enables our framework to learn implicit representations between all the objects and the objects to the measurements, while the temporal attention mechanism focuses on specific parts of past information, allowing our approach to resolve occlusions over multiple frames. Our experiments demonstrate the potential of this new approach, achieving results on par with or better than the current state-of-the-art on multiple MOT metrics for several popular multi-object tracking benchmarks.

Effect of oral zoledronate administration on bone turnover in older women.

AIM: The aim of this work is to assess the safety and efficacy of two oral zoledronate preparations by determining their effects on bone resorption in healthy postmenopausal women. METHODS: The preparations studied were zoledronic acid in enteric-coated capsules or a microparticle preparation of zoledronic acid in these capsules. Bone resorption was measured as ß-C-telopeptideof type I collagen (CTX) in fasting serum. Separate cohorts, each of five women, were recruited and allocated in sequence to single doses of 20 mg, 40 mg, or 60 mg of oral zoledronate. RESULTS: Zoledronate 20 mg enteric capsules were well tolerated, reduced serum CTX by a median 51% at 1 week, but by only 17% at 1 month. Doses of 40 or 60 mg of this preparation produced APR and/or gastrointestinal symptoms in more than half of participants. With these doses, median CTX reduction at 1 week was >80%, ~70% at 1 month, but only ~30% at 6 months. Enteric capsules containing microparticles of zoledronate 20 mg reduced CTX by a median 53% at 1 week, with offset over 3 months. Two or three of these capsules dosed weekly reduced CTX by ~50% at 1 month, and by ~30% at 3 and 6 months. CONCLUSIONS: Oral zoledronate 20 mg circumvents the problem of APR symptoms but, even with multiple doses, the anti-resorptive effect is smaller and less sustained than with intravenous zoledronate. Probably a viable oral regimen of zoledronate dosing at intervals of weeks to months could be developed, but the advantage of infrequent dosing would be lost.

Zoledronate Reduces Height Loss Independently of Vertebral Fracture Occurrence in a Randomized Trial in Osteopenic Older Women.

Vertebral fractures are associated with height loss, reduced quality of life, and increased mortality and are an important endpoint for osteoporosis trials. However, height loss is associated with quality of life and mortality independent of associations with fracture. We have used data from a recent 6-year trial of zoledronate in 2000 osteopenic women aged >65 years to assess the impact of the semiquantitative and quantitative components of the definition of vertebral fracture on the outcome of that trial, to determine what factors impacted on height loss and to test whether height loss can be used as a surrogate for vertebral fracture incidence. In the trial protocol, an incident vertebral fracture was defined as a change in Genant grade plus both a 20% and 4 mm decrease in a vertebral height. The addition of the quantitative criteria reduced the number of fractures detected but did not change the size of the anti-fracture effect (odds ratios of 0.49 versus 0.45) nor the width of the confidence intervals for the odds ratios. Multivariate analysis of baseline predictors of height change showed that age accelerated height loss (p < 0.0001) and zoledronate reduced it (p = 0.0001). Incident vertebral fracture increased height loss (p = 0.0005) but accounted for only 0.7% of the variance in height change, so fracture could not be reliably inferred from height loss. In women without incident vertebral fractures, height loss was still reduced by zoledronate (height change: zoledronate, -1.23; placebo -1.51 mm/yr, p < 0.0001). This likely indicates that zoledronate prevents a subtle but widespread loss of vertebral body heights not detected by vertebral morphometry. Because height loss is associated with quality of life and mortality independent of associations with fracture, it is possible that zoledronate impacts on these endpoints via its effects on vertebral body integrity. © 2022 American Society for Bone and Mineral Research (ASBMR).

EXTENSIVE EXPERTISE IN ENDOCRINOLOGY: Osteoporosis management.

Fractures occur in about half of older White women, and almost a third of older White men. However, 80% of the older individuals who have fractures do not meet the bone density definition of osteoporosis, suggesting that this definition is not an appropriate threshold for offering treatment. Fracture risk can be estimated based on clinical risk factors with or without bone density. A combination of calculated risk, fracture history, and bone density is used in treatment decisions. Medications available for reducing fracture risk act either to inhibit bone resorption or to promote bone formation. Romosozumab is unique in that it has both activities. Bisphosphonates are the most widely used interventions because of their efficacy, safety, and low cost. Continuous use of oral bisphosphonates for >5 years increases the risk of atypical femoral fractures, so is usually punctuated with drug holidays of 6-24 months. Denosumab is a further potent anti-resorptive agent given as 6-monthly s.c. injections. It is comparable to the bisphosphonates in efficacy and safety but has a rapid offset of effect after discontinuation so must be followed by an alternative drug, usually a bisphosphonate. Teriparatide stimulates both bone formation and resorption, substantially increases spine density, and reduces vertebral and non-vertebral fracture rates, though data for hip fractures are scant. Treatment is usually limited to 18-24 months, followed by the transition to an anti-resorptive. Romosozumab is given as monthly s.c. injections for 1 year, followed by an anti-resorptive. This sequence prevents more fractures than anti-resorptive therapy alone. Because of cost, anabolic drugs are usually reserved for those at very high fracture risk. 25-hydroxyvitamin D levels should be maintained above 30 nmol/L, using supplements if sunlight exposure is limited. Calcium intake has little effect on bone density and fracture risk but should be maintained above 500 mg/day using dietary sources.

Prediction of Vitamin D Deficiency in Older Adults: The Role of Machine Learning Models.

CONTEXT: Conventional prediction models for vitamin D deficiency have limited accuracy. BACKGROUND: Using cross-sectional data, we developed models based on machine learning (ML) and compared their performance with those based on a conventional approach. METHODS: Participants were 5106 community-resident adults (50-84 years; 58% male). In the randomly sampled training set (65%), we constructed 5 ML models: lasso regression, elastic net regression, random forest, gradient boosted decision tree, and dense neural network. The reference model was a logistic regression model. Outcomes were deseasonalized serum 25-hydroxyvitamin D (25(OH)D) <50 nmol/L (yes/no) and <25 nmol/L (yes/no). In the test set (the remaining 35%), we evaluated predictive performance of each model, including area under the receiver operating characteristic curve (AUC) and net benefit (decision curves). RESULTS: Overall, 1270 (25%) and 91 (2%) had 25(OH)D <50 and <25 nmol/L, respectively. Compared with the reference model, the ML models predicted 25(OH)D <50 nmol/L with similar accuracy. However, for prediction of 25(OH)D <25 nmol/L, all ML models had higher AUC point estimates than the reference model by up to 0.14. AUC was highest for elastic net regression (0.93; 95% CI 0.90-0.96), compared with 0.81 (95% CI 0.71-0.91) for the reference model. In the decision curve analysis, ML models mostly achieved a greater net benefit across a range of thresholds. CONCLUSION: Compared with conventional models, ML models predicted 25(OH)D <50 nmol/L with similar accuracy but they predicted 25(OH)D <25 nmol/L with greater accuracy. The latter finding suggests a role for ML models in participant selection for vitamin D supplement trials.

Effects of elevated serum urate on cardiometabolic and kidney function markers in a randomised clinical trial of inosine supplementation.

In observational studies, serum urate positively associates with cardiometabolic and kidney diseases. We analyzed data from a randomised placebo-controlled trial to determine whether moderate hyperuricemia induced by inosine affects cardiometabolic and kidney function markers. One hundred and twenty post-menopausal women were recruited into a 6-month randomised, double-blind, placebo-controlled trial of inosine for bone health. Change from baseline in the following pre-specified endpoints was analyzed: body mass index; blood pressure; lipid profile; C-reactive protein; fasting glucose; insulin; HbA1c; serum creatinine; and estimated glomerular filtration rate (eGFR). Despite increases in serum urate levels (+ 0.17 mmol/L at week 6, P < 0.0001), no significant between-group differences were observed in cardiometabolic markers, with the exception of lower fasting glucose concentrations with inosine at week 19. In the inosine group, change in serum urate correlated with change in serum creatinine (r = 0.41, P = 0.0012). However, there was no between-group difference in serum creatinine values. Over the entire study period, there was no significant difference in eGFR (ANCOVA P = 0.13). Reduction in eGFR was greater in the inosine group at Week 13 (mean difference - 4.6 mL/min/1.73 m2, false detection rate P = 0.025), with no between-group difference in eGFR at other time points. These data indicate that increased serum urate does not negatively influence body mass index, blood pressure, lipid profile, or glycaemic control. Serum urate changes associated with inosine intake correlate with changes in serum creatinine, but this does not lead to clinically important reduction in kidney function over 6 months.Clinical trial registration number: Australia and New Zealand Clinical Trials Registry (ACTRN12617000940370), registered 30/06/2017.

Efficacy and Safety of Romosozumab Among Postmenopausal Women With Osteoporosis and Mild-to-Moderate Chronic Kidney Disease.

Patients with osteoporosis and chronic kidney disease (CKD) are at increased risk of fracture and associated negative outcomes, including increased mortality. The present post hoc analysis of two randomized, multicenter, phase 3 clinical trials-Fracture Study in Postmenopausal Women with Osteoporosis (FRAME) and Active-Controlled Fracture Study in Postmenopausal Women with Osteoporosis at High Risk (ARCH)-investigated the efficacy and safety of romosozumab in postmenopausal women with osteoporosis and mild-to-moderate CKD. The analysis included data from 7147 patients from FRAME and 4077 from ARCH. Eighty-one percent of patients from FRAME and 85% from ARCH had mild or moderate reduction in estimated glomerular filtration rate (eGFR) at baseline, and part of this reduction is likely age related. During the 1-year double-blind phases of the trials, patients received romosozumab 210 mg sc or placebo monthly in FRAME and romosozumab 210 mg sc monthly or alendronate 70 mg po weekly in ARCH. Bone mineral density (BMD) at the lumbar spine, total hip, and femoral neck and vertebral and nonvertebral fractures were assessed at baseline and month 12. In both trials, the least-square mean percent change from baseline BMD was significantly greater in the romosozumab groups versus controls across all kidney function categories at month 12. Romosozumab reduced the relative risk of new vertebral fractures at month 12 among patients with eGFR of 30-59, 60-89, and ≥90 mL/min by 72% (95% confidence interval [CI] 14-91; p = 0.017), 70% (40-85; p < 0.001), and 84% (30-96; p = 0.005), respectively, in FRAME versus placebo, and by 51% (5-75; p = 0.04), 19% (-28 to 49; p = 0.39), and 57% (1-81, p = 0.04), respectively, in ARCH versus alendronate. Incidences of adverse events, asymptomatic decreases in serum calcium, and evolution of kidney function during the studies were similar across all baseline kidney function groups. Romosozumab is an effective treatment option for postmenopausal women with osteoporosis and mild-to-moderate reduction in kidney function, with a similar safety profile across different levels of kidney function. © 2022 The Authors. Journal of Bone and Mineral Research published by Wiley Periodicals LLC on behalf of American Society for Bone and Mineral Research (ASBMR).

Drug therapy for osteoporosis in older adults.

The goal of osteoporosis management is to prevent fractures. Several pharmacological agents are available to lower fracture risk, either by reducing bone resorption or by stimulating bone formation. Bisphosphonates are the most widely used anti-resorptives, reducing bone turnover markers to low premenopausal concentrations and reducing fracture rates (vertebral by 50-70%, non-vertebral by 20-30%, and hip by ~40%). Bisphosphonates bind avidly to bone mineral and have an offset of effect measured in months to years. Long term, continuous use of oral bisphosphonates is usually interspersed with drug holidays of 1-2 years, to minimise the risk of atypical femoral fractures. Denosumab is a monoclonal antibody against RANKL that potently inhibits osteoclast development and activity. Denosumab is administered by subcutaneous injection every 6 months. Anti-fracture effects of denosumab are similar to those of the bisphosphonates, but there is a pronounced loss of anti-resorptive effect from 7 months after the last injection, which can result in clusters of rebound vertebral fractures. Two classes of anabolic drugs are now available to stimulate bone formation. Teriparatide and abaloparatide both target the parathyroid hormone-1 receptor, and are given by daily subcutaneous injection for up to 2 years. Romosozumab is an anti-sclerostin monoclonal antibody that stimulates bone formation and inhibits resorption. Romosozumab is given as monthly subcutaneous injections for 1 year. Head-to-head studies suggest that anabolic agents have greater anti-fracture efficacy and produce larger increases in bone density than anti-resorptive drugs. The effects of anabolic agents are transient, so transition to anti-resorptive drugs is required. The optimal strategy for cycling anabolics, anti-resorptives, and off-treatment periods remains to be determined.

Auto-Rectify Network for Unsupervised Indoor Depth Estimation.

Single-View depth estimation using the CNNs trained from unlabelled videos has shown significant promise. However, excellent results have mostly been obtained in street-scene driving scenarios, and such methods often fail in other settings, particularly indoor videos taken by handheld devices. In this work, we establish that the complex ego-motions exhibited in handheld settings are a critical obstacle for learning depth. Our fundamental analysis suggests that the rotation behaves as noise during training, as opposed to the translation (baseline) which provides supervision signals. To address the challenge, we propose a data pre-processing method that rectifies training images by removing their relative rotations for effective learning. The significantly improved performance validates our motivation. Towards end-to-end learning without requiring pre-processing, we propose an Auto-Rectify Network with novel loss functions, which can automatically learn to rectify images during training. Consequently, our results outperform the previous unsupervised SOTA method by a large margin on the challenging NYUv2 dataset. We also demonstrate the generalization of our trained model in ScanNet and Make3D, and the universality of our proposed learning method on 7-Scenes and KITTI datasets.

Learn to Predict Sets Using Feed-Forward Neural Networks.

This paper addresses the task of set prediction using deep feed-forward neural networks. A set is a collection of elements which is invariant under permutation and the size of a set is not fixed in advance. Many real-world problems, such as image tagging and object detection, have outputs that are naturally expressed as sets of entities. This creates a challenge for traditional deep neural networks which naturally deal with structured outputs such as vectors, matrices or tensors. We present a novel approach for learning to predict sets with unknown permutation and cardinality using deep neural networks. In our formulation we define a likelihood for a set distribution represented by a) two discrete distributions defining the set cardinally and permutation variables, and b) a joint distribution over set elements with a fixed cardinality. Depending on the problem under consideration, we define different training models for set prediction using deep neural networks. We demonstrate the validity of our set formulations on relevant vision problems such as: 1) multi-label image classification where we outperform the other competing methods on the PASCAL VOC and MS COCO datasets, 2) object detection, for which our formulation outperforms popular state-of-the-art detectors, and 3) a complex CAPTCHA test, where we observe that, surprisingly, our set-based network acquired the ability of mimicking arithmetics without any rules being coded.