Brain-derived neurotrophic factor (BDNF) is essential for maintaining energy and glucose balance within the central nervous system. Because the study of its metabolic actions has been limited to effects in neuronal cells, its role in other cell types within the brain remains poorly understood. Here we show that astrocytic BDNF signaling within the ventromedial hypothalamus (VMH) modulates neuronal activity in response to changes in energy status. This occurs via the truncated TrkB.T1 receptor. Accordingly, either fasting or central BDNF depletion enhances astrocytic synaptic glutamate clearance, thereby decreasing neuronal activity in mice. Notably, selective depletion of TrkB.T1 in VMH astrocytes blunts the effects of energy status on excitatory transmission, as well as on responses to leptin, glucose and lipids. These effects are driven by increased astrocytic invasion of excitatory synapses, enhanced glutamate reuptake and decreased neuronal activity. We thus identify BDNF/TrkB.T1 signaling in VMH astrocytes as an essential mechanism that participates in energy and glucose homeostasis.

The thrombospondin receptor alpha2delta-1 (α2δ-1) plays essential roles promoting the activity of SF1 neurons in the ventromedial hypothalamus (VMH) and mediating glucose and lipid metabolism in male mice. Its role in the VMH of female mice remains to be defined, especially considering that this hypothalamic region is sexually dimorphic. We found that α2δ-1 depletion in SF1 neurons differentially affects glucose and lipid balance control and sympathetic tone in females compared to males. Mutant females show a modest increase in relative body weight gain when fed a high-fat diet (HFD) and normal energy expenditure, indicating that α2δ-1 is not a critical regulator of energy balance in females, similar to males. However, diminished α2δ-1 function in the VMH leads to enhanced glycemic control in females fed a chow diet, in contrast to the glucose intolerance reported previously in mutant males. Interestingly, the effects of α2δ-1 on glucose balance in females are influenced by diet. Accordingly, females but not males lacking α2δ-1 exhibit diminished glycemic control as well as susceptibility to hepatic steatosis when fed a HFD. Increased hepatic sympathetic tone and CD36 mRNA expression and reduced adiponectin levels underlie these diet-induced metabolic alterations in mutant females. The results indicate that α2δ-1 in VMH SF1 neurons critically regulates metabolic function through sexually dimorphic mechanisms. These findings are clinically relevant since metabolic alterations have been reported as a side effect in human patients prescribed gabapentinoid drugs, known to inhibit α2δ-1 function, for the treatment of seizure disorders, neuropathic pain, and anxiety disorders.

BACKGROUND The uptake of digital health technology (DHT) has been surprisingly low in clinical practice. Despite showing great promise to improve patient outcomes and disease management, there is limited information on the factors that contribute to the limited adoption of DHT, particularly for hypertension management. OBJECTIVE This scoping review provides a comprehensive summary of barriers to and facilitators of DHT adoption for hypertension management reported in the published literature with a focus on provider- and patient-related barriers and facilitators. METHODS This review followed the methodological framework developed by Arskey and O’Malley. Systematic literature searches were conducted on PubMed or Medical Literature Analysis and Retrieval System Online, Cumulative Index to Nursing and Allied Health Literature, and Excerpta Medica database. Articles that reported on barriers to and/or facilitators of digital health adoption for hypertension management published in English between 2008 and 2017 were eligible. Studies not reporting on barriers or facilitators to DHT adoption for management of hypertension were excluded. A total of 2299 articles were identified based on the above criteria after removing duplicates, and they were assessed for eligibility. Of these, 2165 references did not meet the inclusion criteria. After assessing 134 studies in full text, 98 studies were excluded (full texts were either unavailable or studies did not fulfill the inclusion criteria), resulting in a final set of 32 articles. In addition, 4 handpicked articles were also included in the review, making it a total of 36 studies. RESULTS A total of 36 studies were selected for data extraction after abstract and full-text screening by 2 independent reviewers. All conflicts were resolved by a third reviewer. Thematic analysis was conducted to identify major themes pertaining to barriers and facilitators of DHT from both provider and patient perspectives. The key facilitators of DHT adoption by physicians that were identified include ease of integration with clinical workflow, improvement in patient outcomes, and technology usability and technical support. Technology usability and timely technical support improved self-management and patient experience, and positive impact on patient-provider communication were most frequently reported facilitators for patients. Barriers to use of DHTs reported by physicians include lack of integration with clinical workflow, lack of validation of technology, and lack of technology usability and technical support. Finally, lack of technology usability and technical support, interference with patient-provider relationship, and lack of validation of technology were the most commonly reported barriers by patients. CONCLUSIONS Findings suggest the settings and context in which DHTs are implemented and individuals involved in implementation influence adoption. Finally, to fully realize the potential of digitally enabled hypertension management, there is a greater need to validate these technologies to provide patients and providers with reliable and accurate information on both clinical outcomes and cost effectiveness.

Many health care organizations use value-based care strategies that include population health management programs and data analytics to stratify their population and identify high-risk and high-cost patients. Most of these programs target the top 5% most expensive patients. However, little is known about these patients prior to reaching the top 5% of cost, or how their characteristics change over time. To address these gaps, we analyzed the differences in characteristics of patients from 3 different cost segments over 5 years (2011-2015). To evaluate potentially avoidable hospitalizations and associated savings in the health care cost of older patients using Personal Emergency Response Service (PERS). We conducted a retrospective, longitudinal, multicenter study to evaluate potentially avoidable hospitalizations of 2643 older patients over 5 years (2011-2015). All patients had at least one inpatient and/or outpatient encounter, and at least one episode of home health care during the study period. Additionally, all patients used PERS at home anytime during the study period. We ranked patients by their annual health care cost and then grouped them into the following segments for each respective year: T-segment constitutes the top 5% most expensive patients; M-segment comprises the middle 45% of patients; B-segment includes the bottom 50% least expensive patients. We then evaluated differences in the characteristics of patients in the B-, M- and T-cost segments in each study year. Continuous variables were compared by t test (two-tailed) for normally distributed variables and Kruskal-Wallis Rank Sum test for skewed variables. The chi-square test was used for categorical variables. The three cost segments differed significantly each year (P<.05) with respect to: demographics (age, education), PERS utilization (Incidents, ER transport), health care utilization (hospitalizations, length of stay, 30-, 90-, and 180-day readmissions, outpatient encounters) and medical conditions (number of conditions, Charlson Comorbidity Index). Further, we analyzed the number of potentially avoidable hospitalizations (as defined by CMS) and associated cost savings in each segment. All hospitalizations occurred among patients in the T- and M-segments while the B-segment was hospitalization-free each year. The percentage of avoidable hospitalizations in the M-segment compared with the T-segment was 3 times greater (75% vs 25%, P<.001). While the potential cost saving from avoidable hospitalizations in the entire population increased from $3.0M to $8.2M (2011-2015), the majority of these cost savings were in the M-segment compared with the T-segment (60% vs 40%, P<.001). Although many health care organizations target intensive and costly interventions to their most expensive patients, this analysis suggests there is untapped potential to control costs and improve care beyond focusing on the highest cost patients. Namely, targeting patients in the middle cost segment may offer great opportunity for population management programs to maximize both potentially avoidable hospitalizations and cost savings.

Heart failure is one of the leading causes of hospitalization in the United States. Advances in big data solutions allow for storage, management, and mining of large volumes of structured and semi-structured data, such as complex healthcare data. Applying these advances to complex healthcare data has led to the development of risk prediction models to help identify patients who would benefit most from disease management programs in an effort to reduce readmissions and healthcare cost, but the results of these efforts have been varied. The primary aim of this study was to develop a 30-day readmission risk prediction model for heart failure patients discharged from a hospital admission.

Despite widespread adoption and demonstrated value in a range of industries, machine learning predictive algorithms are yet to be routinely used in frontline medical care. Significant health system and industry-based resources are allocated towards validating and refining predictive algorithms for a range of applications to ensure accuracy and reliability. For these algorithms to be useful and useable, further work is required to understand how and why they might fit into, and augment existing clinical workflows. This qualitative study assessed the value and usability of a novel machine learning technology to predict and explain the risk of 30-day hospital readmission in patients with heart failure (HF). It involved exploring opportunities for integration of the technology within existing clinical workflows, and investigating key roles that use current readmission risk scores and may use future scores. Semi-structured interviews (n=27) and targeted observations (n=3) were carried out with key stakeholders, including physicians, nurses, hospital administration, and non-clinical support staff. Participants were recruited from cardiology and general medicine units at an academic medical center within the Partners HealthCare system. Data was analyzed via inductive thematic and workflow analysis. Findings were validated via member checking across limited key roles (n=3). Results highlighted a number of factors that were deemed necessary by staff for successful integration of a risk prediction tool into existing clinical workflow. These included, but were not limited to the following. Staff clearly stated that any new tool must be easily accessible from within the electronic health record, which dictates the majority of existing clinical workflow. Staff emphasized that information should be consistently accurate and that any display must be digestible efficiently, intuitively and quickly (ie, within <5 seconds). Additionally, staff discussed that outputs of the risk prediction tool must match their clinical intuition, experience and interactions with the patient. To be truly valuable, the tool must also provide added value over and above these factors: some staff indicated that provision of role-specific and actionable next steps based on the system output would provide novel value to their daily work. Using these considerations, a number of role groups were identified as potentially able to derive value from the proposed risk prediction tool, including case managers, attending RNs, responding clinicians, hospital administration staff, nursing directors and attending physicians. Acceptability and value varied by role, specialization and clinical context. For example, cardiology-trained clinicians reported feeling well-versed in providing good clinical care and minimizing preventable readmissions, and thus saw less value in the tool. General medicine staff, however, indicated that a HF-specific tool may be impractical for their day-to-day work given the range of clinical presentations seen by them. Findings resonate with existing literature around successful implementation and adoption of technologies in health care. Frontline clinicians are incredibly discerning around proposed changes to their existing workflow. Many HF readmission risk tools and initiatives have been trialled with mixed success; frontline staff demonstrated fatigue around piloting new initiatives. However, given the right conditions, staff reported some perceived value in machine learning-based tools to improve their daily work.

BACKGROUND Heart failure (HF) patients have a high readmission rate with approximately 20% of patients being readmitted within 30-days after discharge. Hospital interventions to reduce HF readmissions are resource- and effort-intensive. Widespread availability of electronic medical record data has spurred interest in using machine learning-based techniques for risk stratification of heart failure patients. The predictive performance of machine learning-based predictive models is often evaluated solely using the Area Under the Receiver Operating Characteristic (AUROC) curve. However, the AUROC is independent of prevalence therefore predictive models with the same AUROC can have differential clinical utility. Furthermore, the AUROC does not provide any insight about the presence of overfitting or decay in predictive performance of a model over time, both of which can affect its real-world performance. OBJECTIVE Our primary objective is to assess real-world performance of a 30-day readmission risk prediction model for HF patients, which had an AUROC of 0.71 in the training dataset. METHODS Predictions for risk of 30-day readmissions in HF patients in the Partners Healthcare System were prospectively obtained from the model. We assessed the positive (PPV) and negative predictive value (NPV), in addition to sensitivity, specificity, accuracy, model calibration and Brier score. RESULTS Four hundred twenty index admissions that were not part of the training dataset were included in this prospective evaluation. Readmission rate was 24% (101 30-day readmissions). The AUROC of the predictive model was 0.57. At a discrimination threshold of 0.2 for flagging high-risk index admissions, the sensitivity and specificity of the model were 53.46% and 63.32%, respectively. The PPV and NPV were 31.57% and 81.12%, respectively. The Brier score was 0.19. CONCLUSIONS Our analysis offers important insights about the real-world performance of this predictive model. The NPV suggests that the model’s prediction about patients at low risk for readmission are reliable. This insight can be useful in optimizing resource allocation for patients with heart failure.

April 27, 2018April 10, 2018Free AccessDrivers of Epilepsy-related Cost of Care in an Integrated Healthcare System (P6.280)Sujay Kakarmath, Mahesh Agarwal, Sara Golas, Jennifer Felsted, Joseph Dye, Jesse Fishman, Marjory Levey, and Stephen AgboolaAuthors Info & AffiliationsApril 10, 2018 issue90 (15_supplement)https://doi.org/10.1212/WNL.90.15_supplement.P6.280 Letters to the Editor

Half of Medicare reimbursement goes toward caring for the top 5% of the most expensive patients. However, little is known about these patients prior to reaching the top or how their costs change annually. To address these gaps, we analyzed patient flow and associated health care cost trends over 5 years. To evaluate the cost of health care utilization in older patients by analyzing changes in their long-term expenditures. This was a retrospective, longitudinal, multicenter study to evaluate health care costs of 2643 older patients from 2011 to 2015. All patients had at least one episode of home health care during the study period and used a personal emergency response service (PERS) at home for any length of time during the observation period. We segmented all patients into top (5%), middle (6%-50%), and bottom (51%-100%) segments by their annual expenditures and built cost pyramids based thereon. The longitudinal health care expenditure trends of the complete study population and each segment were assessed by linear regression models. Patient flows throughout the segments of the cost acuity pyramids from year to year were modeled by Markov chains. Total health care costs of the study population nearly doubled from US $17.7M in 2011 to US $33.0M in 2015 with an expected annual cost increase of US $3.6M (P=.003). This growth was primarily driven by a significantly higher cost increases in the middle segment (US $2.3M, P=.003). The expected annual cost increases in the top and bottom segments were US $1.2M (P=.008) and US $0.1M (P=.004), respectively. Patient and cost flow analyses showed that 18% of patients moved up the cost acuity pyramid yearly, and their costs increased by 672%. This was in contrast to 22% of patients that moved down with a cost decrease of 86%. The remaining 60% of patients stayed in the same segment from year to year, though their costs also increased by 18%. Although many health care organizations target intensive and costly interventions to their most expensive patients, this analysis unveiled potential cost savings opportunities by managing the patients in the lower cost segments that are at risk of moving up the cost acuity pyramid. To achieve this, data analytics integrating longitudinal data from electronic health records and home monitoring devices may help health care organizations optimize resources by enabling clinicians to proactively manage patients in their home or community environments beyond institutional settings and 30- and 60-day telehealth services.

Big data solutions, particularly machine learning predictive algorithms, have demonstrated the ability to unlock value from data in real time in many settings outside of health care. Rapid growth in electronic medical record adoption and the shift from a volume-based to a value-based reimbursement structure in the US health care system has spurred investments in machine learning solutions. Machine learning methods can be used to build flexible, customized, and automated predictive models to optimize resource allocation and improve the efficiency and quality of health care. However, these models are prone to the problems of overfitting, confounding, and decay in predictive performance over time. It is, therefore, necessary to evaluate machine learning–based predictive models in an independent dataset before they can be adopted in the clinical practice. In this paper, we describe the protocol for independent, prospective validation of a machine learning–based model trained to predict the risk of 30-day re-admission in patients with heart failure. This study aims to prospectively validate a machine learning–based predictive model for inpatient admissions in patients with heart failure by comparing its predictions of risk for 30-day re-admissions against outcomes observed prospectively in an independent patient cohort. All adult patients with heart failure who are discharged alive from an inpatient admission will be prospectively monitored for 30-day re-admissions through reports generated by the electronic medical record system. Of these, patients who are part of the training dataset will be excluded to avoid information leakage to the algorithm. An expected sample size of 1228 index admissions will be required to observe a minimum of 100 30-day re-admission events. Deidentified structured and unstructured data will be fed to the algorithm, and its prediction will be recorded. The overall model performance will be assessed using the concordance statistic. Furthermore, multiple discrimination thresholds for screening high-risk patients will be evaluated according to the sensitivity, specificity, predictive values, and estimated cost savings to our health care system. The project received funding in April 2017 and data collection began in June 2017. Enrollment was completed in July 2017. Data analysis is currently underway, and the first results are expected to be submitted for publication in October 2018. To the best of our knowledge, this is one of the first studies to prospectively evaluate a predictive machine learning algorithm in a real-world setting. Findings from this study will help to measure the robustness of predictions made by machine learning algorithms and set a realistic benchmark for expectations of gains that can be made through its application to health care. RR1-10.2196/9466

Background: Telehealth programs have been successful in reducing 30-day readmissions and emergency department visits. However, such programs often focus on the costliest patients with multiple morbidities and last for only 30 to 60 days postdischarge. Inexpensive monitoring of elderly patients via a personal emergency response system (PERS) to identify those at high risk for emergency hospital transport could be used to target interventions and prevent avoidable use of costly readmissions and emergency department visits after 30 to 60 days of telehealth use. Objective: The objectives of this study were to (1) develop and validate a predictive model of 30-day emergency hospital transport based on PERS data; and (2) compare the model's predictions with clinical outcomes derived from the electronic health record (EHR). Methods: We used deidentified medical alert pattern data from 290,434 subscribers to a PERS service to build a gradient tree boosting-based predictive model of 30-day hospital transport, which included predictors derived from subscriber demographics, self-reported medical conditions, caregiver network information, and up to 2 years of retrospective PERS medical alert data. We evaluated the model's performance on an independent validation cohort (n=289,426). We linked EHR and PERS records for 1815 patients from a home health care program to compare PERS-based risk scores with rates of emergency encounters as recorded in the EHR. Results: In the validation cohort, 2.22% (6411/289,426) of patients had 1 or more emergency transports in 30 days. The performance of the predictive model of emergency hospital transport, as evaluated by the area under the receiver operating characteristic curve, was 0.779 (95% CI 0.774-0.785). Among the top 1% of predicted high-risk patients, 25.5% had 1 or more emergency hospital transports in the next 30 days. Comparison with clinical outcomes from the EHR showed 3.9 times more emergency encounters among predicted high-risk patients than low-risk patients in the year following the prediction date. Conclusions: Patient data collected remotely via PERS can be used to reliably predict 30-day emergency hospital transport. Clinical observations from the EHR showed that predicted high-risk patients had nearly four times higher rates of emergency encounters than did low-risk patients. Health care providers could benefit from our validated predictive model by targeting timely preventive interventions to high-risk patients. This could lead to overall improved patient experience, higher quality of care, and more efficient resource utilization.

Background: In the US healthcare system, half of overall Medicare and Medicaid Services reimbursement goes towards caring for the top 5% most expensive patients. However, little is known about how these patients' costs change annually prior to them reaching the top 5%. To address these gaps and investigate potential cost savings opportunities, we analyzed patient flow and associated healthcare cost trends over the period 2011-2015.

Background: Studies suggest that certain technologies may help regulate emotional states and reduce stress. Spire is a wearable device that measures breathing patterns to identify distinct emotional states—calm, tense or focus—to send feedback to the user. Muse is a meditation device that gives real time user audio-feedback based on the EEG wave patterns.

Background: With the worldwide increase in the elderly population, chronic diseases and associated healthcare utilization such as costly emergency department visits and subsequent hospitalizations are also on the rise. Predictive analytics can be used to identify patients at high risk for emergency utilization, using Electronic Health Record (EHR) data collected before or at hospital discharge. In addition, non-hospital data may be useful for prediction of changes in risk outside of hospital settings. Inexpensive monitoring of elderly via a Personal Emergency Response System (PERS) to identify patients at high risk for emergency hospital transport could be used to target interventions and prevent avoidable, costly long-term healthcare utilization.

The central mechanisms controlling glucose and lipid homeostasis are inadequately understood. We show that α2δ-1 is an essential regulator of glucose and lipid balance, acting in steroidogenic factor-1 (SF1) neurons of the ventromedial hypothalamus (VMH). These effects are body weight independent and involve regulation of SF1+ neuronal activity and sympathetic output to metabolic tissues. Accordingly, mice with α2δ-1 deletion in SF1 neurons exhibit glucose intolerance, altered lipolysis, and decreased cholesterol content in adipose tissue despite normal energy balance regulation. Profound reductions in the firing rate of SF1 neurons, decreased sympathetic output, and elevated circulating levels of serotonin are associated with these alterations. Normal calcium currents but reduced excitatory postsynaptic currents in mutant SF1 neurons implicate α2δ-1 in the promotion of excitatory synaptogenesis separate from its canonical role as a calcium channel subunit. Collectively, these findings identify an essential mechanism that regulates VMH neuronal activity and glycemic and lipid control and may be a target for tackling metabolic disease.

Brain‐derived neurotrophic factor (BDNF), a protein that supports neuronal survival and synaptic plasticity, is essential for energy balance regulation. Genetic polymorphisms that impede BDNF function pose risk factors for obesity in humans. Similarly, blunted BDNF signaling in the rodent brain is associated with overeating (OV), body weight (BW) gain and metabolic dysfunction. Here we identified alpha2delta‐1 (α2δ‐1), a high voltage‐gated calcium channel subunit and thrombospondin receptor, as a novel appetite regulator underlying BDNF function in the ventromedial hypothalamus (VMH). α2δ‐1 is highly expressed in the VMH of wild‐type (WT) mice. Moreover, mice lacking central Bdnf (CM) have reduced cell surface α2δ‐1 expression in VMH cells. Selective inhibition of VMH α2δ‐1 in WT mice through chronic delivery of gabapentin resulted in a significant increase in food intake and BW gain after 7 days. Rescuing the α2δ‐1 deficit in VMH cells of CM, through targeted delivery of an adeno‐associated virus encoding α2δ‐1, mitigated the OV and BW gain and normalized glucose homeostasis. These data inform mechanisms underlying effects of BDNF in the VMH and identify a critical role for α2δ‐1 in feeding control.

Brain-derived neurotrophic factor (BDNF) and its receptor, TrkB, are critical components of the neural circuitry controlling appetite and body weight. Diminished BDNF signaling in mice results in severe hyperphagia and obesity. In humans, BDNF haploinsufficiency and the functionalBdnfVal66Met polymorphism have been linked to elevated food intake and body weight. The mechanisms underlying this dysfunction are poorly defined. We demonstrate a chief role of α2δ-1, a calcium channel subunit and thrombospondin receptor, in triggering overeating in mice with central BDNF depletion. We show reduced α2δ-1 cell-surface expression in the BDNF mutant ventromedial hypothalamus (VMH), an energy balance-regulating center. This deficit contributes to the hyperphagia exhibited by BDNF mutant mice because selective inhibition of α2δ-1 by gabapentin infusion into wild-type VMH significantly increases feeding and body weight gain. Importantly, viral-mediated α2δ-1 rescue in BDNF mutant VMH significantly mitigates their hyperphagia, obesity, and liver steatosis and normalizes deficits in glucose homeostasis. Whole-cell recordings in BDNF mutant VMH neurons revealed normal calcium currents but reduced frequency of EPSCs. These results suggest calcium channel-independent effects of α2δ-1 on feeding and implicate α2δ-1–thrombospondin interactions known to facilitate excitatory synapse assembly. Our findings identify a central mechanism mediating the inhibitory effects of BDNF on feeding. They also demonstrate a novel and critical role for α2δ-1 in appetite control and suggest a mechanism underlying weight gain in humans treated with gabapentinoid drugs.

BACKGROUND:Smoking cessation is often followed by weight gain. Eating behaviors and weight change have been linked to the brain response to food, but it is unknown whether smoking influences this response.OBJECTIVE:We determined the influence of smoking status (smokers compared with nonsmokers) on the brain response to food in regions associated with weight changes in nonsmokers.DESIGN:In study 1, we used functional MRI (fMRI) to identify regions of the brain associated with weight change in nonsmokers. BMI and the brain response to a milk shake, which is a palatable and energy-dense food, were measured in a group of 27 nonsmokers (5 men). Sixteen subjects (3 men) returned 1 y later for BMI reassessment. The change in BMI was regressed against the brain response to isolate regions associated with weight change. In study 2, to determine whether smokers showed altered responses in regions associated with weight change, we assessed the brain response to a milk shake in 11 smokers. The brain response to a milk shake compared with a tasteless control solution was assessed in 11 smokers (5 men) in comparison with a group of age-, sex- and body weight-matched nonsmokers selected from the pool of nonsmokers who participated in study 1.RESULTS:The response in the midbrain, hypothalamus, thalamus, and ventral striatum was positively associated with weight change at the 1-y follow-up in 16 nonsmokers. Compared with nonsmokers, smokers had a greater response to milk shakes in the hypothalamus.CONCLUSION:Smokers display an altered brain response to food in the hypothalamus, which is an area associated with long-term weight change in nonsmokers.

Previous investigations consistently report a negative association between body mass index (BMI) and response in the caudate nucleus during the consumption of palatable and energy dense food. Since this response has also been linked to weight gain, we sought to replicate this finding and determine if the reduced response is associated with measures of impulsivity or food reward. Two studies were conducted in which fMRI was used to measure brain response to milkshake and a tasteless control solution. In Study 1 (n = 25) we also assessed self-reported impulsivity, willingness to work for food, and subjective experiences of the pleasantness of milkshake taste and aroma. Replicating prior work, we report a negative association between BMI and brain response to milkshake vs. tasteless in the caudate nucleus. The opposite pattern was observed in the ventral putamen, with greater response observed in the 13 overweight compared to the 12 healthy weight subjects. Regression of brain response against impulsivity and food reward measures revealed one significant association: in the overweight but not healthy weight group self-reported impulsivity was negatively associated with caudate response to milkshake. In Study 2 (n = 14), in addition to assessing brain response to milkshake and tasteless solutions subjects completed a go/no-go task outside the scanner. As predicted, we identified an inverse relationship between caudate response to milkshake vs. tasteless and failure to inhibit responses on the no-go trials. We conclude that the inverse correlation between BMI and caudate response to milkshake is associated with impulsivity but not food reward. These findings suggest that response to milkshake in the dorsal striatum may be related to weight gain by promoting impulsive eating behavior.