Effect of preceding drug therapy on the renal and cardiovascular outcomes of combined sodium-glucose cotransporter-2 inhibitor and glucagon-like peptide-1 receptor agonist treatment in patients with type 2 diabetes and chronic kidney disease.

AIM: To conduct a post hoc subgroup analysis of patients with type 2 diabetes (T2D) from the RECAP study, who were treated with sodium-glucose cotransporter-2 (SGLT2) inhibitor and glucagon-like peptide 1 receptor agonist (GLP-1RA) combination therapy, focusing only on those patients who had chronic kidney disease (CKD), to examine whether the composite renal outcome differed between those who received SGLT2 inhibitor treatment first and those who received a GLP-1RA first. METHODS: We included 438 patients with CKD (GLP-1RA-first group, n = 223; SGLT2 inhibitor-first group, n = 215) from the 643 T2D patients in the RECAP study. The incidence of the composite renal outcome, defined as progression to macroalbuminuria and/or a ≥50% decrease in estimated glomerular filtration rate (eGFR), was analysed using a propensity score (PS)-matched model. Furthermore, we calculated the win ratio for these composite renal outcomes, which were weighted in the following order: (1) both a ≥50% decrease in eGFR and progression to macroalbuminuria; (2) a decrease in eGFR of ≥50% only; and (3) progression to macroalbuminuria only. RESULTS: Using the PS-matched model, 132 patients from each group were paired. The incidence of renal composite outcomes did not differ between the two groups (GLP-1RA-first group, 10%; SGLT2 inhibitor-first group, 17%; odds ratio 1.80; 95% confidence interval [CI] 0.85 to 4.26; p = 0.12). The win ratio of the GLP-1RA-first group versus the SGLT2 inhibitor-first group was 1.83 (95% CI 1.71 to 1.95; p < 0.001). CONCLUSION: Although the renal composite outcome did not differ between the two groups, the win ratio of the GLP-1RA-first group versus the SGLT2 inhibitor-first group was significant. These results suggest that, in GLP-1RA and SGLT2 inhibitor combination therapy, the addition of an SGLT2 inhibitor to baseline GLP-1RA treatment may lead to more favourable renal outcomes.

Influence of the combination of SGLT2 inhibitors and GLP-1 receptor agonists on eGFR decline in type 2 diabetes: post-hoc analysis of RECAP study.

Accumulating evidence has demonstrated that both SGLT2 inhibitors (SGLT2i) and GLP-1 receptor agonists (GLP1Ra) have protective effects in patients with diabetic kidney disease. Combination therapy with SGLT2i and GLP1Ra is commonly used in patients with type 2 diabetes (T2D). We previously reported that in combination therapy of SGLT2i and GLP1Ra, the effect on the renal composite outcome did not differ according to the preceding drug. However, it remains unclear how the initiation of combination therapy is associated with the renal function depending on the preceding drug. In this post hoc analysis, we analyzed a total of 643 T2D patients (GLP1Ra-preceding group, n = 331; SGLT2i-preceding group, n = 312) and investigated the differences in annual eGFR decline. Multiple imputation and propensity score matching were performed to compare the annual eGFR decline. The reduction in annual eGFR decline in the SGLT2i-preceding group (pre: -3.5 ± 9.4 mL/min/1.73 m2/year, post: -0.4 ± 6.3 mL/min/1.73 m2/year, p < 0.001), was significantly smaller after the initiation of GLP1Ra, whereas the GLP1Ra-preceding group tended to slow the eGFR decline but not to a statistically significant extent (pre: -2.0 ± 10.9 mL/min/1.73 m2/year, post: -1.8 ± 5.4 mL/min/1.73 m2/year, p = 0.83) after the initiation of SGLT2i. After the addition of GLP1Ra to SGLT2i-treated patients, slower annual eGFR decline was observed. Our data raise the possibility that the renal benefits-especially annual eGFR decline-of combination therapy with SGLT2i and GLP1Ra may be affected by the preceding drug.

Pretreatment body mass index affects achievement of target blood pressure with sodium-glucose cotransporter 2 inhibitors in patients with type 2 diabetes mellitus and chronic kidney disease.

Sodium-glucose cotransporter 2 inhibitor (SGLT2-I) shows excellent antihypertensive effects in addition to its hypoglycemic effects. However, whether body mass index (BMI) affects the antihypertensive effect of SGLT2-I remains unknown. We investigated the impact of baseline BMI on the achievement of target blood pressure (BP) with SGLT2-I treatment in Japanese patients with type 2 diabetes mellitus (T2DM) and chronic kidney disease (CKD). We retrospectively evaluated 447 Japanese patients with T2DM and CKD treated with SGLT2-I for at least 1 year. The primary outcome was achieving the target BP (<130/80 mmHg) after SGLT2-I treatment. Patients were divided into two groups according to a baseline BMI of 29.1 determined by receiver operating characteristic analysis and analyzed in a cohort model with propensity score matching. In each group, 130 patients were compared by propensity score matching. The target BP achievement rate was significantly higher in the BMI < 29.1 group than in the BMI ≥ 29.1 group (34% and 21%, respectively, p = 0.03). The odds ratio for achieving the target BP in the BMI ≥ 29.1 group was 0.50 (95% confidence interval, 0.28-0.90, p = 0.02). The BMI < 29.1 group had significantly lower systolic and diastolic BPs after SGLT2-I treatment than the BMI ≥ 29.1 group. Only the BMI < 29.1 group was showed a significant decrease in the logarithmic albumin-to-creatinine ratio from baseline after SGLT2-I treatment. In patients with T2DM and CKD, baseline BMI was associated with the antihypertensive effects of SGLT2-I. Patients in the lower baseline BMI group were more likely to achieve the target BP after SGLT2-I treatment. Pretreatment BMI affects the antihypertensice effect of SGLT2 inhibirors in patients with T2DM and CKD.

Renoprotective effects of combination treatment with sodium-glucose cotransporter inhibitors and GLP-1 receptor agonists in patients with type 2 diabetes mellitus according to preceding medication.

AIMS: Combination therapy with sodium-glucose cotransporter inhibitors (SGLT2is) and GLP-1 receptor agonists (GLP1Ras) is now of interest in clinical practice. The present study evaluated the effects of the preceding drug type on the renal outcome in clinical practice. METHODS: We retrospectively extracted type 2 diabetes mellitus patients who had received both SGLT2i and GLP1Ra treatment for at least 1 year. A total of 331 patients in the GLP1Ra-preceding group and 312 patients in the SGLT2i-preceding group were ultimately analyzed. Either progression of the albuminuria status and/or a ≥30% decrease in the eGFR was set as the primary renal composite outcome. The analysis using propensity score with inverse probability weighting was performed for the outcome. RESULTS: The incidences of the renal composite outcome in the SGLT2i- and GLP1Ra-preceding groups were 28% and 25%, respectively, with an odds ratio [95% confidence interval] of 1.14 [0.75, 1.73] (p = .54). A logistic regression analysis showed that the mean arterial pressure (MAP) at baseline, the logarithmic value of the urine albumin-to-creatinine ratio at baseline, and the change in MAP were independent factors influencing the renal composite outcome. CONCLUSION: With combination therapy of SGLT2i and GLP1Ra, the preceding drug did not affect the renal outcome.

The concomitant use of sodium-glucose co-transporter 2 inhibitors improved the renal outcome of Japanese patients with type 2 diabetes treated with glucagon-like peptide 1 receptor agonists.

Aims: This study aimed to clarify the renal influence of glucagon-like peptide 1 receptor agonists (GLP1Ras) with or without sodium-glucose co-transporter 2 inhibitors (SGLT2is) on Japanese patients with type 2 diabetes mellitus (T2DM). Methods: We retrospectively extracted 547 patients with T2DM who visited the clinics of members of Kanagawa Physicians Association. The progression of albuminuria status and/or a ≥ 15% decrease in the estimated glomerular filtration rate (eGFR) per year was set as the renal composite outcome. Propensity score matching was performed to compare GLP1Ra-treated patients with and without SGLT2i. Results: After matching, 186 patients in each group were compared. There was no significant difference of the incidence of the renal composite outcomes (17% vs. 20%, P = 0.50); however, the annual decrease in the eGFR was significantly smaller and the decrease in the urine albumin-to-creatinine ratio was larger in GLP1Ra-treated patients with the concomitant use of SGLT2is than in those without it (-1.1 ±â€…5.0 vs. -2.8 ±â€…5.1 mL/min/1.73 m2, P = 0.001; and -0.08 ±â€…0.61 vs. 0.05 ±â€…0.52, P = 0.03, respectively). Conclusion: The concomitant use of SGLT2i with GLP1Ra improved the annual decrease in the eGFR and the urine albumin-to-creatinine ratio in Japanese patients with T2DM.

Langevin Cooling for Unsupervised Domain Translation.

Domain translation is the task of finding correspondence between two domains. Several deep neural network (DNN) models, e.g., CycleGAN and cross-lingual language models, have shown remarkable successes on this task under the unsupervised setting-the mappings between the domains are learned from two independent sets of training data in both domains (without paired samples). However, those methods typically do not perform well on a significant proportion of test samples. In this article, we hypothesize that many of such unsuccessful samples lie at the fringe-relatively low-density areas-of data distribution, where the DNN was not trained very well, and propose to perform the Langevin dynamics to bring such fringe samples toward high-density areas. We demonstrate qualitatively and quantitatively that our strategy, called Langevin cooling (L-Cool), enhances state-of-the-art methods in image translation and language translation tasks.

Comparison of renal outcomes between sodium glucose co-transporter 2 inhibitors and glucagon-like peptide 1 receptor agonists.

AIMS: This study aimed to clarify the differences in how sodium glucose co-transporter 2 inhibitors (SGLT2i) and glucagon-like peptide 1 receptor agonists (GLP1Ra) influence kidney function in Japanese patients with type 2 diabetes mellitus (T2DM). METHODS: We retrospectively built two databases of patients with T2DM who visited the clinics of members of Kanagawa Physicians Association. We defined the renal composite outcome as either progression of albuminuria status and/or > 15% deterioration in estimated glomerular filtration rate (eGFR) per year. We used propensity score matching to compare patient outcomes after SGLT2i and GLP1Ra treatments. RESULTS: The incidence of renal composite outcomes was significantly lower in SGLT2i-treated patients than in GLP1Ra-treated patients (n = 15[11%] and n = 27[20%], respectively, P = 0.001). Annual eGFR changes (mL/min/1.73 m2/year) between the two groups differed significantly (-1.8 [95 %CI, -2.7, -0.9] in SGLT2i-treated patients and - 3.4 [95 %CI, -4.6, -2.2] in GLP1Ra-treated patients, P = 0.0049). The urine albumin-to-creatinine ratio changed owing to a significant interaction between the presence or absence of a decrease in systolic blood pressure and the difference in treatments (P < 0.04). CONCLUSION: Renal composite outcome incidence was lower in SGLT2i-treated patients than in GLP1Ra-treated patients.

Higher-Order Explanations of Graph Neural Networks via Relevant Walks.

Graph Neural Networks (GNNs) are a popular approach for predicting graph structured data. As GNNs tightly entangle the input graph into the neural network structure, common explainable AI approaches are not applicable. To a large extent, GNNs have remained black-boxes for the user so far. In this paper, we show that GNNs can in fact be naturally explained using higher-order expansions, i.e., by identifying groups of edges that jointly contribute to the prediction. Practically, we find that such explanations can be extracted using a nested attribution scheme, where existing techniques such as layer-wise relevance propagation (LRP) can be applied at each step. The output is a collection of walks into the input graph that are relevant for the prediction. Our novel explanation method, which we denote by GNN-LRP, is applicable to a broad range of graph neural networks and lets us extract practically relevant insights on sentiment analysis of text data, structure-property relationships in quantum chemistry, and image classification.

Prognosis of hemodialysis patients with progressive aortic stenosis: a prospective cohort study.

BACKGROUND: Whether progressive mild to moderate aortic stenosis in hemodialysis patients influences their prognosis has not been elucidated. This prospective cohort study explored whether progressive aortic stenosis predicted the rate of cardiac events and mortality in those patients. METHODS: A total of 283 consecutive hemodialysis patients (no aortic stenosis, 248; progressive aortic stenosis, 35) underwent echocardiography for assessment of aortic stenosis, with a median follow-up period of 4.1 years. Study endpoints were cardiac events, all-cause mortality, and cardiac death. Kaplan-Meier analysis and multivariate Cox proportional hazard analysis were performed to estimate cardiac events, all-cause mortality, and cardiac death. RESULTS: Cumulative cardiac event rate, all-cause mortality rate, and the rate of cardiac death at 3-year follow-up were 44.9%, 40.5%, and 26.4% in patients with progressive aortic stenosis and 22.1%, 19.0%, and 7.5% in those without aortic stenosis, respectively. Kaplan-Meier analysis demonstrated the cumulative rates of cardiac events and all-cause mortality. And cardiac death was significantly higher in patients with progressive aortic stenosis than in those without aortic stenosis. Multivariate Cox proportional hazard analysis revealed that progressive aortic stenosis was predictive of cardiac events (adjusted hazard ratio 2.47; 95% confidence interval 1.38-4.39) and cardiac death (adjusted hazard ratio 4.21; 95% confidence interval 2.10-8.46). Age, physical activity, C-reactive protein, and serum albumin levels-but not progressive aortic stenosis-predicted all-cause mortality. CONCLUSIONS: The rates of cardiac events and cardiac death were higher in hemodialysis patients with progressive aortic stenosis than in those without aortic stenosis. Furthermore, progressive aortic stenosis predicted cardiac events and cardiac death. Compared with those without aortic stenosis, patients with progressive aortic stenosis had higher all-cause mortality, which was related to their comorbidities.Trial registration This study was retrospectively registered with University Hospital Medical Information Network Clinical Trials Registry (registration number, UMIN 000024023) at September 12th, 2016.

Estimation of Thermodynamic Observables in Lattice Field Theories with Deep Generative Models.

In this Letter, we demonstrate that applying deep generative machine learning models for lattice field theory is a promising route for solving problems where Markov chain Monte Carlo (MCMC) methods are problematic. More specifically, we show that generative models can be used to estimate the absolute value of the free energy, which is in contrast to existing MCMC-based methods, which are limited to only estimate free energy differences. We demonstrate the effectiveness of the proposed method for two-dimensional ϕ^{4} theory and compare it to MCMC-based methods in detailed numerical experiments.

Robustifying models against adversarial attacks by Langevin dynamics.

Adversarial attacks on deep learning models have compromised their performance considerably. As remedies, a number of defense methods were proposed, which however, have been circumvented by newer and more sophisticated attacking strategies. In the midst of this ensuing arms race, the problem of robustness against adversarial attacks still remains a challenging task. This paper proposes a novel, simple yet effective defense strategy where off-manifold adversarial samples are driven towards high density regions of the data generating distribution of the (unknown) target class by the Metropolis-adjusted Langevin algorithm (MALA) with perceptual boundary taken into account. To achieve this task, we introduce a generative model of the conditional distribution of the inputs given labels that can be learned through a supervised Denoising Autoencoder (sDAE) in alignment with a discriminative classifier. Our algorithm, called MALA for DEfense (MALADE), is equipped with significant dispersion-projection is distributed broadly. This prevents white box attacks from accurately aligning the input to create an adversarial sample effectively. MALADE is applicable to any existing classifier, providing robust defense as well as off-manifold sample detection. In our experiments, MALADE exhibited state-of-the-art performance against various elaborate attacking strategies.

Asymptotically unbiased estimation of physical observables with neural samplers.

We propose a general framework for the estimation of observables with generative neural samplers focusing on modern deep generative neural networks that provide an exact sampling probability. In this framework, we present asymptotically unbiased estimators for generic observables, including those that explicitly depend on the partition function such as free energy or entropy, and derive corresponding variance estimators. We demonstrate their practical applicability by numerical experiments for the two-dimensional Ising model which highlight the superiority over existing methods. Our approach greatly enhances the applicability of generative neural samplers to real-world physical systems.

Optimizing for Measure of Performance in Max-Margin Parsing.

Many learning tasks in the field of natural language processing including sequence tagging, sequence segmentation, and syntactic parsing have been successfully approached by means of structured prediction methods. An appealing property of the corresponding training algorithms is their ability to integrate the loss function of interest into the optimization process improving the final results according to the chosen measure of performance. Here, we focus on the task of constituency parsing and show how to optimize the model for the F1 -score in the max-margin framework of a structural support vector machine (SVM). For reasons of computational efficiency, it is a common approach to binarize the corresponding grammar before training. Unfortunately, this introduces a bias during the training procedure as the corresponding loss function is evaluated on the binary representation, while the resulting performance is measured on the original unbinarized trees. Here, we address this problem by extending the inference procedure presented by Bauer et al. Specifically, we propose an algorithmic modification that allows evaluating the loss on the unbinarized trees. The new approach properly models the loss function of interest resulting in better prediction accuracy and still benefits from the computational efficiency due to binarized representation. The presented idea can be easily transferred to other structured loss functions.

Relation between Blood Pressure Management and Renal Effects of Sodium-Glucose Cotransporter 2 Inhibitors in Diabetic Patients with Chronic Kidney Disease.

AIM: The renoprotective effect of sodium-glucose cotransporter 2 inhibitors is thought to be due, at least in part, to a decrease in blood pressure. The aim of this study was to determine the renal effects of these inhibitors in low blood pressure patients and the dependence of such effect on blood pressure management status. METHODS: The subjects of this retrospective study were 740 patients with type 2 diabetes mellitus and chronic kidney disease who had been managed at the clinical facilities of the Kanagawa Physicians Association. Data on blood pressure management status and urinary albumin-creatinine ratio were analyzed before and after treatment. RESULTS: Changes in the logarithmic value of urinary albumin-creatinine ratio in 327 patients with blood pressure < 130/80 mmHg at the initiation of treatment and in 413 patients with BP above 130/80 mmHg were -0.13 ± 1.05 and -0.24 ± 0.97, respectively. However, there was no significant difference between the two groups by analysis of covariance models after adjustment of the logarithmic value of urinary albumin-creatinine ratio at initiation of treatment. Changes in the logarithmic value of urinary albumin-creatinine ratio in patients with mean blood pressure of <102 mmHg (n = 537) and those with ≥102 mmHg (n = 203) at the time of the survey were -0.25 ± 1.02 and -0.03 ± 0.97, respectively, and the difference was significant in analysis of covariance models even after adjustment for the logarithmic value of urinary albumin-creatinine ratio at initiation of treatment (p < 0.001). CONCLUSION: Our results confirmed that blood pressure management status after treatment with SGLT2 inhibitors influences the extent of change in urinary albumin-creatinine ratio. Stricter blood pressure management is needed to allow the renoprotective effects of sodium-glucose cotransporter 2 inhibitors.

Retrospective analysis of effects of sodium-glucose co-transporter 2 inhibitor in Japanese type 2 diabetes mellitus patients with chronic kidney disease.

AIM: The aim of this study was to assess the renal effects of the glucose-lowering SGLT2 inhibitors in Japanese type 2 diabetes mellitus patients with chronic kidney disease. METHODS: The Kanagawa Physicians Association maintains a registry of patients who visit their 31 clinics. Clinical data of type 2 diabetes mellitus patients with chronic kidney disease, who were prescribed SGLT2 inhibitors in addition to other treatments, were collected and analysed. RESULTS: SGLT2i was associated with a fall in HbA1c from 64.1 ± 16.7 mmol/mol (8.0 ± 1.5%) to 56.5 ± 12.9 mmol/mol (7.3 ± 1.2%) ( p < 0.01) in 869 analysed cases, a decrease in urine albumin-creatinine ratio from a median of 47.1 to 41.1 mg/gCr, and decrease in estimated glomerular filtration rate from 77.7 ± 23.9 to 75.0 ± 23.9 mL/min/1.73 m2 ( p < 0.01). The effect on albumin-creatinine ratio was independent of age or stage of estimated glomerular filtration; however, there was a significant negative correlation between albumin-creatinine ratio at the initiation of SGLT2 inhibitor and change in ACR. Multiple linear regression analysis identified use of empagliflozin, ß-blockers, and sulphonylureas, Δsystolic blood pressure at office, serum Cr and albumin-creatinine ratio value at initiation of SGLT2 inhibitor as independent and significant determinants of change in ACR. CONCLUSIONS: This study confirmed that the beneficial renal effects of SGLT2 inhibitor in Japanese type 2 diabetes mellitus patients with chronic kidney disease, similar to those reported in large-scale clinical trials conducted in Western countries.

Renal effects of sodium glucose co-transporter 2 inhibitors in Japanese type 2 diabetes mellitus patients with home blood pressure monitoring.

Decrease in blood pressure contributes to the reno-protective effects of sodium-glucose cotransporter 2 inhibitors; however, its relationship with home monitoring of blood pressure is unclear. We retrospectively analyzed 101 visiting members of the Kanagawa Physicians Association with type 2 diabetes mellitus and chronic kidney disease who were taking sodium-glucose cotransporter 2 inhibitors and who monitored blood pressure at home for a median treatment period of 14 months. At baseline, the mean value of HbA1c was 59.3 mmol/mol (7.6%) and the median value of albumin-creatinine ratio was 30.9 mg/gCr that was evaluated in 88 patients. The mean blood pressure both at office and home significantly decreased, and there was a significant positive correlation between the change in albumin-creatinine ratio and both blood pressures. Controlled hypertension, masked hypertension, white coat hypertension, and sustained hypertension were observed in 10.9%, 13.9%, 12.9%, and 62.4% of patients at the initiation of therapy, which changed to 10.9%, 16.8%, 17.8%, and 54.5% at the time of the survey, respectively. In conclusion, management of blood pressure both at office and home was found to be important for the reno-protective effects of sodium-glucose cotransporter 2 inhibitors along with strict blood pressure management.

Transductive Regression for Data With Latent Dependence Structure.

Analyzing data with latent spatial and/or temporal structure is a challenge for machine learning. In this paper, we propose a novel nonlinear model for studying data with latent dependence structure. It successfully combines the concepts of Markov random fields, transductive learning, and regression, making heavy use of the notion of joint feature maps. Our transductive conditional random field regression model is able to infer the latent states by combining limited labeled data of high precision with unlabeled data containing measurement uncertainty. In this manner, we can propagate accurate information and greatly reduce uncertainty. We demonstrate the usefulness of our novel framework on generated time series data with the known temporal structure and successfully validate it on synthetic as well as real-world offshore data with the spatial structure from the oil industry to predict rock porosities from acoustic impedance data.

Support Vector Data Descriptions and $k$ -Means Clustering: One Class?

We present ClusterSVDD, a methodology that unifies support vector data descriptions (SVDDs) and $k$ -means clustering into a single formulation. This allows both methods to benefit from one another, i.e., by adding flexibility using multiple spheres for SVDDs and increasing anomaly resistance and flexibility through kernels to $k$ -means. In particular, our approach leads to a new interpretation of $k$ -means as a regularized mode seeking algorithm. The unifying formulation further allows for deriving new algorithms by transferring knowledge from one-class learning settings to clustering settings and vice versa. As a showcase, we derive a clustering method for structured data based on a one-class learning scenario. Additionally, our formulation can be solved via a particularly simple optimization scheme. We evaluate our approach empirically to highlight some of the proposed benefits on artificially generated data, as well as on real-world problems, and provide a Python software package comprising various implementations of primal and dual SVDD as well as our proposed ClusterSVDD.

On robust parameter estimation in brain-computer interfacing.

OBJECTIVE: The reliable estimation of parameters such as mean or covariance matrix from noisy and high-dimensional observations is a prerequisite for successful application of signal processing and machine learning algorithms in brain-computer interfacing (BCI). This challenging task becomes significantly more difficult if the data set contains outliers, e.g. due to subject movements, eye blinks or loose electrodes, as they may heavily bias the estimation and the subsequent statistical analysis. Although various robust estimators have been developed to tackle the outlier problem, they ignore important structural information in the data and thus may not be optimal. Typical structural elements in BCI data are the trials consisting of a few hundred EEG samples and indicating the start and end of a task. APPROACH: This work discusses the parameter estimation problem in BCI and introduces a novel hierarchical view on robustness which naturally comprises different types of outlierness occurring in structured data. Furthermore, the class of minimum divergence estimators is reviewed and a robust mean and covariance estimator for structured data is derived and evaluated with simulations and on a benchmark data set. MAIN RESULTS: The results show that state-of-the-art BCI algorithms benefit from robustly estimated parameters. SIGNIFICANCE: Since parameter estimation is an integral part of various machine learning algorithms, the presented techniques are applicable to many problems beyond BCI.

Efficient Exact Inference With Loss Augmented Objective in Structured Learning.

Structural support vector machine (SVM) is an elegant approach for building complex and accurate models with structured outputs. However, its applicability relies on the availability of efficient inference algorithms--the state-of-the-art training algorithms repeatedly perform inference to compute a subgradient or to find the most violating configuration. In this paper, we propose an exact inference algorithm for maximizing nondecomposable objectives due to special type of a high-order potential having a decomposable internal structure. As an important application, our method covers the loss augmented inference, which enables the slack and margin scaling formulations of structural SVM with a variety of dissimilarity measures, e.g., Hamming loss, precision and recall, Fß-loss, intersection over union, and many other functions that can be efficiently computed from the contingency table. We demonstrate the advantages of our approach in natural language parsing and sequence segmentation applications.