Robust and data-efficient generalization of self-supervised machine learning for diagnostic imaging.

Machine-learning models for medical tasks can match or surpass the performance of clinical experts. However, in settings differing from those of the training dataset, the performance of a model can deteriorate substantially. Here we report a representation-learning strategy for machine-learning models applied to medical-imaging tasks that mitigates such 'out of distribution' performance problem and that improves model robustness and training efficiency. The strategy, which we named REMEDIS (for 'Robust and Efficient Medical Imaging with Self-supervision'), combines large-scale supervised transfer learning on natural images and intermediate contrastive self-supervised learning on medical images and requires minimal task-specific customization. We show the utility of REMEDIS in a range of diagnostic-imaging tasks covering six imaging domains and 15 test datasets, and by simulating three realistic out-of-distribution scenarios. REMEDIS improved in-distribution diagnostic accuracies up to 11.5% with respect to strong supervised baseline models, and in out-of-distribution settings required only 1-33% of the data for retraining to match the performance of supervised models retrained using all available data. REMEDIS may accelerate the development lifecycle of machine-learning models for medical imaging.

Validation of clinical acceptability of deep-learning-based automated segmentation of organs-at-risk for head-and-neck radiotherapy treatment planning.

Introduction: Organ-at-risk segmentation for head and neck cancer radiation therapy is a complex and time-consuming process (requiring up to 42 individual structure, and may delay start of treatment or even limit access to function-preserving care. Feasibility of using a deep learning (DL) based autosegmentation model to reduce contouring time without compromising contour accuracy is assessed through a blinded randomized trial of radiation oncologists (ROs) using retrospective, de-identified patient data. Methods: Two head and neck expert ROs used dedicated time to create gold standard (GS) contours on computed tomography (CT) images. 445 CTs were used to train a custom 3D U-Net DL model covering 42 organs-at-risk, with an additional 20 CTs were held out for the randomized trial. For each held-out patient dataset, one of the eight participant ROs was randomly allocated to review and revise the contours produced by the DL model, while another reviewed contours produced by a medical dosimetry assistant (MDA), both blinded to their origin. Time required for MDAs and ROs to contour was recorded, and the unrevised DL contours, as well as the RO-revised contours by the MDAs and DL model were compared to the GS for that patient. Results: Mean time for initial MDA contouring was 2.3 hours (range 1.6-3.8 hours) and RO-revision took 1.1 hours (range, 0.4-4.4 hours), compared to 0.7 hours (range 0.1-2.0 hours) for the RO-revisions to DL contours. Total time reduced by 76% (95%-Confidence Interval: 65%-88%) and RO-revision time reduced by 35% (95%-CI,-39%-91%). All geometric and dosimetric metrics computed, agreement with GS was equivalent or significantly greater (p<0.05) for RO-revised DL contours compared to the RO-revised MDA contours, including volumetric Dice similarity coefficient (VDSC), surface DSC, added path length, and the 95%-Hausdorff distance. 32 OARs (76%) had mean VDSC greater than 0.8 for the RO-revised DL contours, compared to 20 (48%) for RO-revised MDA contours, and 34 (81%) for the unrevised DL OARs. Conclusion: DL autosegmentation demonstrated significant time-savings for organ-at-risk contouring while improving agreement with the institutional GS, indicating comparable accuracy of DL model. Integration into the clinical practice with a prospective evaluation is currently underway.

An NFIX-mediated regulatory network governs the balance of hematopoietic stem and progenitor cells during hematopoiesis.

The transcription factor (TF) nuclear factor I-X (NFIX) is a positive regulator of hematopoietic stem and progenitor cell (HSPC) transplantation. Nfix-deficient HSPCs exhibit a severe loss of repopulating activity, increased apoptosis, and a loss of colony-forming potential. However, the underlying mechanism remains elusive. Here, we performed cellular indexing of transcriptomes and epitopes by high-throughput sequencing (CITE-seq) on Nfix-deficient HSPCs and observed a loss of long-term hematopoietic stem cells and an accumulation of megakaryocyte and myelo-erythroid progenitors. The genome-wide binding profile of NFIX in primitive murine hematopoietic cells revealed its colocalization with other hematopoietic TFs, such as PU.1. We confirmed the physical interaction between NFIX and PU.1 and demonstrated that the 2 TFs co-occupy super-enhancers and regulate genes implicated in cellular respiration and hematopoietic differentiation. In addition, we provide evidence suggesting that the absence of NFIX negatively affects PU.1 binding at some genomic loci. Our data support a model in which NFIX collaborates with PU.1 at super-enhancers to promote the differentiation and homeostatic balance of hematopoietic progenitors.

Sclerosing Mesenteritis Complicated With Mesenteric Lymphoma Responsive to Ustekinumab.

A 45-year-old man with a 10-year history of biopsy-proven, steroid-dependent sclerosing mesenteritis failed/was intolerant to tamoxifen, azathioprine, colchicine, cyclophosphamide, and methotrexate. He developed osteoporosis, diabetes, and bilateral cataracts. He responded to infliximab but was diagnosed with mesenteric large B-cell lymphoma 6 months after treatment initiation. He achieved remission from lymphoma after chemotherapy, but the sclerosing mesenteritis remained poorly controlled. He was treated with ustekinumab (520 mg intravenously followed by 90 mg subcutaneously every 8 weeks), leading to complete steroid-free remission. He remains symptom and cancer-free 24 months after starting ustekinumab.

Selective brain entry of lipid nanoparticles in haemorrhagic stroke is linked to biphasic blood-brain barrier disruption.

Haemorrhagic stroke represents a significant public health burden, yet our knowledge and ability to treat this type of stroke are lacking. Previously we showed that we can target ischaemic-stroke lesions by selective translocation of lipid nanoparticles through the site of blood-brain barrier (BBB) disruption. The data we presented in this study provide compelling evidence that haemorrhagic stroke in mice induces BBB injury that mimics key features of the human pathology and, more importantly, provides a gate for entry of lipid nanoparticles-based therapeutics selectively to the bleeding site. Methods: Haemorrhagic stroke was induced in mice by intra-striatal collagenase injection. lipid nanoparticles were injected intravenously at 3 h, 24 h & 48 h post-haemorrhagic stroke and accumulation in the brain studied using in-vivo optical imaging and histology. BBB integrity, brain water content and iron accumulation were characterised using dynamic contrast-enhanced MRI, quantitative T1 mapping, and gradient echo MRI. Results: Using in-vivo SPECT/CT imaging and optical imaging revealed biphasic lipid nanoparticles entry into the bleeding site, with an early phase of increased uptake at 3-24 h post-haemorrhagic stroke, followed by a second phase at 48-72 h. Lipid nanoparticles entry into the brain post-haemorrhage showed an identical entry pattern to the trans-BBB leakage rate (Ktrans [min-1]) of Gd-DOTA, a biomarker for BBB disruption, measured using dynamic contrast-enhanced MRI. Discussion: Our findings suggest that selective accumulation of liposomes into the lesion site is linked to a biphasic pattern of BBB hyper-permeability. This approach provides a unique opportunity to selectively and efficiently deliver therapeutic molecules across the BBB, an approach that has not been utilised for haemorrhagic stroke therapy and is not achievable using free small drug molecules.

Yield and Risk Factors for Advanced Colorectal Neoplasia and Long-term Outcomes in Veterans With 3 or More Nonadvanced Adenomas.

BACKGROUND AND AIMS: Until recently, guidelines recommended a 3-year surveillance colonoscopy for persons with 3 to 10 nonadvanced adenomas (NAA). In this study, we quantify yield for metachronous advanced neoplasia (AN); attempt to identify risk factors for AN; and measure colorectal cancer (CRC) incidence and mortality. METHODS: We used natural language processing to screen an existing data set for Veterans with 3 to 10 NAA. We manually reviewed colonoscopy and pathology reports to verify baseline findings and determine results of subsequent colonoscopy (sCY). Baseline features were extracted from the electronic medical record (EMR) and a national data set, CRC incidence was obtained from the Veterans Affairs cancer registry, and CRC mortality from the National Death Index through September 30, 2017. CRC incidence and mortality were compared between Veterans who did versus did not have sCY. RESULTS: Natural language processing identified 3673 Veterans who potentially had 3 to 10 NAA, of which 1672 were excluded after EMR review. In the analytical cohort of 2001 subjects, 1178 (59%) had sCY at a mean (SD) follow-up of 4.3 (2.2) years. The sCY group was younger (mean age: 61 vs. 67 y; P<0.01) and were less likely to have diabetes (27% vs. 31%; P=0.02) and congestive heart failure (4% vs. 9%; P<0.01). sCY showed AN in 182 subjects (15.5%). Baseline features were no different between those with versus without metachronous AN. Subjects with sCY had a greater CRC incidence (n=7 vs. n=0; P=0.046), but there was no difference in CRC mortality (0 for both subgroups). CONCLUSIONS: Among patients with 3 to 10 NAA on index colonoscopy who underwent sCY, AN was present in 15.5% at mean follow-up of 4.3 years. No risk factors for AN were identified. CRC incidence, but not CRC mortality, was higher among those with sCY.

Four consecutive yearly point-prevalence studies in Wales indicate lack of improvement in sepsis care on the wards.

The 'Sepsis Six' bundle was promoted as a deliverable tool outside of the critical care settings, but there is very little data available on the progress and change of sepsis care outside the critical care environment in the UK. Our aim was to compare the yearly prevalence, outcome and the Sepsis Six bundle compliance in patients at risk of mortality from sepsis in non-intensive care environments. Patients with a National Early Warning Score (NEWS) of 3 or above and suspected or proven infection were enrolled into four yearly 24-h point prevalence studies, carried out in fourteen hospitals across Wales from 2016 to 2019. We followed up patients to 30 days between 2016-2019 and to 90 days between 2017 and 2019. Out of the 26,947 patients screened 1651 fulfilled inclusion criteria and were recruited. The full 'Sepsis Six' care bundle was completed on 223 (14.0%) occasions, with no significant difference between the years. On 190 (11.5%) occasions none of the bundle elements were completed. There was no significant correlation between bundle element compliance, NEWS or year of study. One hundred and seventy (10.7%) patients were seen by critical care outreach; the 'Sepsis Six' bundle was completed significantly more often in this group (54/170, 32.0%) than for patients who were not reviewed by critical care outreach (168/1385, 11.6%; p < 0.0001). Overall survival to 30 days was 81.7% (1349/1651), with a mean survival time of 26.5 days (95% CI 26.1-26.9) with no difference between each year of study. 90-day survival for years 2017-2019 was 74.7% (949/1271), with no difference between the years. In multivariate regression we identified older age, heart failure, recent chemotherapy, higher frailty score and do not attempt cardiopulmonary resuscitation orders as significantly associated with increased 30-day mortality. Our data suggests that despite efforts to increase sepsis awareness within the NHS, there is poor compliance with the sepsis care bundles and no change in the high mortality over the study period. Further research is needed to determine which time-sensitive ward-based interventions can reduce mortality in patients with sepsis and how can these results be embedded to routine clinical practice.Trial registration Defining Sepsis on the Wards ISRCTN 86502304 https://doi.org/10.1186/ISRCTN86502304 prospectively registered 09/05/2016.

Dissociating the effect of reward uncertainty and timing uncertainty on neural indices of reward prediction errors: A reward positivity (RewP) event-related potential (ERP) study.

Accurate reward predictions include forecasting both what a reward will be and when a reward will occur. We tested how variations in the certainty of reward outcome and certainty in timing of feedback presentation modulate neural indices of reward prediction errors using the reward positivity (RewP) component of the scalp-recorded brain event-related potential (ERP). In a within-subjects design, seventy-three healthy individuals completed two versions of a cued doors task; one cued the probability of a reward outcome while the other cued the probability of a delay before feedback. Replicating previous results, RewP amplitude was larger for uncertain feedback compared to certain feedback. Additionally, RewP amplitude was differentially associated with uncertainty of presence/absence of reward, but not uncertainty of feedback timing. Findings suggest a dissociation in that RewP amplitude is modulated by reward prediction certainty but is less affected by certainty surrounding timing of feedback.

Educational Intervention Improves Proton Pump Inhibitor Stewardship in Outpatient Gastroenterology Clinics.

BACKGROUND: Improper chronic proton pump inhibitor (PPI) use has risen significantly in the last few decades. In our gastroenterology trainees' clinics, we aimed to optimize PPI usage. METHODS: We collected baseline data on patients' PPI use for 8 weeks. Based on gastroenterology society guidelines, we determined conditions for appropriate PPI use. If the indication could not be determined, it was categorized as "unknown". Generated from the three most frequent causes for inappropriate PPI use, interventions were developed to correct each issue. Following a brief educational session, trainees implemented these interventions over a subsequent 8-week interval. RESULTS: During our pre-intervention period, trainees evaluated 263 patients who were prescribed a PPI. In 49% of the cases, the use of PPI was deemed inappropriate. The most common reasons were: gastroesophageal reflux disease (GERD) which was never titrated to the lowest effective dose, twice daily dosing for Barrett's esophagus (BE) chemoprevention and unknown indication. During our intervention period, trainees evaluated 145 patients prescribed a PPI for GERD with well-controlled symptoms in 101 cases. PPI had not been titrated to lowest effective dose in 37 cases prompting intervention which was successful in 23 cases. PPI indication was unknown in 17 cases prompting a message to the prescribing provider to review appropriateness. Two cases of BE chemoprevention with twice daily dosing were appropriately reduced to daily dosing. Ultimately, after intervention, PPI use was deemed appropriate after intervention in 172 (77%) cases. CONCLUSIONS: Improper chronic PPI use was significant. Focusing intervention efforts on PPI use for GERD, BE and unknown indications substantially increased appropriateness of PPI use.

Effective patient education techniques for preventing unintended pregnancies.

Ample resources exist detailing the variety of contraceptive options for patients who want to prevent an unintended pregnancy; however, few resources provide tools to guide clinicians in best practices for history taking and patient education about contraceptive use. This article attempts to fill this gap by reviewing current recommendations on timing of patient education, techniques to open a discussion, and common patient misunderstandings about pregnancy and contraception.

Real-world prevalence of the inclusion criteria for the LEADER trial: Data from a national general practice network.

AIMS: To explore the prevalence and describe the clinical characteristics of people with type 2 diabetes with a similar cardiovascular (CV) profile to that of the LEADER trial participants in a primary care setting in England. MATERIALS AND METHODS: In this cross-sectional analysis, using the Royal College of General Practitioners (RCGP) Research and Surveillance Centre (RSC) network database, we identified people with type 2 diabetes meeting the LEADER inclusion criteria. We identified people's CV risk factors using computerized medical records. Additionally, we assessed the prescription pattern of glucagon-like peptide-1 receptor agonists (GLP-1RAs) in this cohort. RESULTS: Of 1 275 461 adults, we identified 84 394 with type 2 diabetes, of whom 14 000 (16.6%) met the LEADER inclusion criteria for established or high-risk CV disease (RCGP RSC-CVD group). The LEADER cohort was younger than the RCGP RSC-CVD group (64.2 vs 73.2 years), had higher mean glycated haemoglobin (71.6 vs 67.1 mmol/mol) and blood pressure (BP) values (systolic BP: 135.9 vs 132.9 mmHg; diastolic BP: 77.2 vs 72.7 mmHg), and a higher mean body mass index (32.5 vs 30.9 kg/m2 ). In the RCGP RSC-CVD group, only 1215 people (8.7%) had ever been prescribed a GLP-1RA and 760 (5.4%) had ever received liraglutide. CONCLUSIONS: In a cohort of English general practice patients, one in six people with type 2 diabetes met the LEADER inclusion criteria, and less than one in 10 of these received liraglutide, a drug which has demonstrated CV benefits amongst others. There is scope to improve the outlook in people with type 2 diabetes and high CV risk through evidence-based use of specific GLP-1RAs.

Working memory load affects early affective responses to concrete and abstract words differently: Evidence from ERPs.

Early posterior negativity (EPN) is an early-occurring, event-related, potential that is elicited by pictures and words that have highly arousing characteristics. Whilst EPN has been found with words presented in isolation several times, different types of words have shown quite different effects across different types of tasks. One possible reason for this is that memory and attentional demands may affect the way semantic features of words are processed, and this may modulate EPN. This was investigated in a silent reading task using abstract and concrete words of negative and neutral valence and a dual phonological working memory task to manipulate memory load. The results showed that abstract but not concrete words elicited EPN, and this may have affected downstream processing. Further analyses examining alpha desynchronization showed that negative concrete words appeared to be significantly affected by the memory load manipulation, unlike negative abstract words. These results provide evidence that the processing of features in negative concrete words is more affected by working memory and attentional demands than the processing of features in negative abstract words, and this may be responsible for the failure of negative concrete words to elicit EPN in this study. Thus, the extent to which words elicit EPN appears to be dependent on both their semantic representations and competing cognitive processes. These results provide a potential explanation for some of the differences that have been reported in previous experiments as well as insight into how memory and attention can affect the processing of the semantic features of words.

Impact of a Resident-Centered Interprofessional Quality Improvement Intervention on Acute Care Length of Stay.

Competency in interprofessional quality improvement and performance measurement is required by the Accreditation Council for Graduate Medical Education. We implemented an interprofessional quality improvement project to support trainee involvement in systems-level improvement to reduce hospital length of stay and engage trainees in efforts to improve the validity and reliability of clinical documentation contributing to risk-adjusted performance measures. The intervention had three components: daily interprofessional disposition huddles to discuss discharge needs, medical documentation curriculum to improve clinical data accuracy, and scheduled coding huddles to provide real-time feedback on documentation. Outcome measures included an unadjusted and risk-adjusted measure of hospital length of stay. Case severity index (CSI) served as a process measure. Statistical process control charts were used to measure change over time. The mean unadjusted length of stay decreased from 5.84 to 4.98 days. Both the unadjusted and the risk-adjusted length of stay measures exceeded the lower control limit of the statistical control chart. The CSI increased and exceeded the upper control limit of the statistical control chart. Improvements were sustained in the year following implementation. The intervention offers a model for academic institutions to satisfy new Common Program Requirements by engaging trainees in performance measurement and interprofessional improvement efforts.

Nfix Promotes Survival of Immature Hematopoietic Cells via Regulation of c-Mpl.

Hematopoietic stem and progenitor cells (HSPCs) are necessary for life-long blood production and replenishment of the hematopoietic system during stress. We recently reported that nuclear factor I/X (Nfix) promotes HSPC survival post-transplant. Here, we report that ectopic expression of Nfix in primary mouse HSPCs extends their ex vivo culture from about 20 to 40 days. HSPCs overexpressing Nfix display hypersensitivity to supportive cytokines and reduced apoptosis when subjected to cytokine deprivation relative to controls. Ectopic Nfix resulted in elevated levels of c-Mpl transcripts and cell surface protein on primary murine HSPCs as well as increased phosphorylation of STAT5, which is known to be activated down-stream of c-MPL. Blocking c-MPL signaling by removal of thrombopoietin or addition of a c-MPL neutralizing antibody negated the antiapoptotic effect of Nfix overexpression on cultured HSPCs. Furthermore, NFIX was capable of binding to and transcriptionally activating a proximal c-Mpl promoter fragment. In sum, these data suggest that NFIX-mediated upregulation of c-Mpl transcription can protect primitive hematopoietic cells from stress ex vivo. Stem Cells 2018;36:943-950.

A Formant Range Profile for Singers.

Vowel selection is important in differentiating between singing styles. The timbre of the vocal instrument, which is related to its frequency spectrum, is governed by both the glottal sound source and the vowel choices made by singers. Consequently, the ability to modify the vowel space is a measure of how successfully a singer can maintain a desired timbre across a range of pitches. Formant range profiles were produced as a means of quantifying this ability. Seventy-seven subjects (including trained and untrained vocalists) participated, producing vowels with three intended mouth shapes: (1) neutral or speech-like, (2) megaphone-shaped (wide open mouth), and (3) inverted-megaphone-shaped (widened oropharynx with moderate mouth opening). The first and second formant frequencies (F1 and F2) were estimated with fry phonation for each shape and values were plotted in F1-F2 space. By taking four vowels of a quadrangle /i, æ, a, u/, the resulting area was quantified in kHz2 (kHz squared) as a measure of the subject's ability to modify their vocal tract for spectral differences.

Assessing early implementation of state autism insurance mandates.

In the United States, health insurance coverage for autism spectrum disorder treatments has been historically limited. In response, as of 2015, 40 states and Washington, DC, have passed state autism insurance mandates requiring many health plans in the private insurance market to cover autism diagnostic and treatment services. This study examined five states' experiences implementing autism insurance mandates. Semi-structured, key-informant interviews were conducted with 17 participants representing consumer advocacy organizations, provider organizations, and health insurance companies. Overall, participants thought that the mandates substantially affected the delivery of autism services. While access to autism treatment services has increased as a result of implementation of state mandates, states have struggled to keep up with the demand for services. Participants provided specific information about barriers and facilitators to meeting this demand. Understanding of key informants' perceptions about states' experiences implementing autism insurance mandates is useful for other states considering adopting or expanding mandates or other policies to expand access to autism treatment services.

Stop the Blame Game: Restructuring Morbidity and Mortality Conferences to Teach Patient Safety and Quality Improvement to Residents.

INTRODUCTION: Adverse events leading to patient harm are rarely the result of an individual error but are instead due to a series of errors resulting from system breakdowns. Thus, the Accreditation Council for Graduate Medical Education requires all residents to participate in quality improvement and patient safety programs. However, a major reported obstacle to sustainable quality improvement and patient safety curricula, as well as meaningful practice improvement, is the small number of faculty with expertise or training in these topics. METHODS: This workshop provides a simple framework for redesigning traditional morbidity and mortality conferences for faculty who have minimal quality improvement training. The materials associated with this publication include a standardized presentation template, sample teaching points, and a faculty facilitator's guide. RESULTS: Between August 2014 and February 2015, 135 trainees from one of our tertiary training sites attended seven of these redesigned conferences. The largest gains were made in teaching residents how to use a systems-based approach to analyze medical error and how to identify corresponding error-reduction strategies. Residents also perceived themselves as more likely to put their knowledge into action through filing an incident report after attending the conference. The one item that did not change was the residents' perception of safety culture at their institution, suggesting that attendance at a monthly conference is not sufficient to change culture. DISCUSSION: Similarly formatted M&Mconferences may help institutions address several aspectof the ACGME CLER program that provides programswith periodic feedback regarding trainee education onpatient safety and quality improvement as well.

Hip proprioceptive feedback influences the control of mediolateral stability during human walking.

Active control of the mediolateral location of the feet is an important component of a stable bipedal walking pattern, although the roles of sensory feedback in this process are unclear. In the present experiments, we tested whether hip abductor proprioception influenced the control of mediolateral gait motion. Participants performed a series of quiet standing and treadmill walking trials. In some trials, 80-Hz vibration was applied intermittently over the right gluteus medius (GM) to evoke artificial proprioceptive feedback. During walking, the GM was vibrated during either right leg stance (to elicit a perception that the pelvis was closer mediolaterally to the stance foot) or swing (to elicit a perception that the swing leg was more adducted). Vibration during quiet standing evoked leftward sway in most participants (13 of 16), as expected from its predicted perceptual effects. Across the 13 participants sensitive to vibration, stance phase vibration caused the contralateral leg to be placed significantly closer to the midline (by ∼2 mm) at the end of the ongoing step. In contrast, swing phase vibration caused the vibrated leg to be placed significantly farther mediolaterally from the midline (by ∼2 mm), whereas the pelvis was held closer to the stance foot (by ∼1 mm). The estimated mediolateral margin of stability was thus decreased by stance phase vibration but increased by swing phase vibration. Although the observed effects of vibration were small, they were consistent with humans monitoring hip proprioceptive feedback while walking to maintain stable mediolateral gait motion.

Sleep apnea syndrome represents a risk for glaucoma in a veterans' affairs population.

Purpose. To determine whether the diagnosis of sleep apnea syndrome (SAS) represents a risk-factor for glaucoma. Design. Retrospective records review. Methods. Records in an electronic database which exists at the Birmingham, Alabama Veterans' Affairs Medical Center (BVAMC) permit data retrieval and sorting based on diagnostic and procedural codes. Deidentified data of those having had an eye examination and a diagnostic code (ICD-9) for either sleep apnea or glaucoma were included. Statistical Analyses. SPSS version 19 was used to produce crosstabs and to conduct a bivariate logistic regression that examined the relationship between SAS and glaucoma. Results. A total of 70,960 unique records were included for analysis. Of the 2,725 patients with a diagnosis of sleep apnea, 228 (8.37%) also had a diagnosis of glaucoma. Diagnosis of glaucoma was present in 3,410 patients among 68,235 patients (5.00%) without sleep apnea. Bivariate logistic regression analysis yielded an odds ratio of 1.736 (P < 0.001) suggesting that individuals with SAS are more likely to have a coexisting diagnosis of glaucoma than individuals without SAS. Conclusions. Results of this investigation suggest that SAS may represent a significant risk factor for glaucoma and this should be considered when managing patients who report that diagnosis.