Search

Over the past years, research utilizing routine care data extracted from Electronic Medical Records (EMRs) has increased tremendously. Yet there are no straightforward, standardized strategies for pre-processing these data. We propose a dedicated medical pre-processing pipeline aimed at taking on many problems and opportunities contained within EMR data, such as their temporal, inaccurate and incomplete nature. The pipeline is demonstrated on a dataset of routinely recorded data in general practice EMRs of over 260,000 patients, in which the occurrence of colorectal cancer (CRC) is predicted using various machine learning techniques (i.e., CART, LR, RF) and subsets of the data. CRC is a common type of cancer, of which early detection has proven to be important yet challenging. The results are threefold. First, the predictive models generated using our pipeline reconfirmed known predictors and identified new, medically plausible, predictors derived from the cardiovascular and metabolic disease domain, validating the pipeline's effectiveness. Second, the difference between the best model generated by the data-driven subset (AUC 0.891) and the best model generated by the current state of the art hypothesis-driven subset (AUC 0.864) is statistically significant at the 95% confidence interval level. Third, the pipeline itself is highly generic and independent of the specific disease targeted and the EMR used. In conclusion, the application of established machine learning techniques in combination with the proposed pipeline on EMRs has great potential to enhance disease prediction, and hence early detection and intervention in medical practice., (Copyright © 2016 Elsevier Ltd. All rights reserved.)

Objective: Machine learning techniques can be used to extract predictive models for diseases from electronic medical records (EMRs). However, the nature of EMRs makes it difficult to apply off-the-shelf machine learning techniques while still exploiting the rich content of the EMRs. In this paper, we explore the usage of a range of natural language processing (NLP) techniques to extract valuable predictors from uncoded consultation notes and study whether they can help to improve predictive performance., Methods: We study a number of existing techniques for the extraction of predictors from the consultation notes, namely a bag of words based approach and topic modeling. In addition, we develop a dedicated technique to match the uncoded consultation notes with a medical ontology. We apply these techniques as an extension to an existing pipeline to extract predictors from EMRs. We evaluate them in the context of predictive modeling for colorectal cancer (CRC), a disease known to be difficult to diagnose before performing an endoscopy., Results: Our results show that we are able to extract useful information from the consultation notes. The predictive performance of the ontology-based extraction method moves significantly beyond the benchmark of age and gender alone (area under the receiver operating characteristic curve (AUC) of 0.870 versus 0.831). We also observe more accurate predictive models by adding features derived from processing the consultation notes compared to solely using coded data (AUC of 0.896 versus 0.882) although the difference is not significant. The extracted features from the notes are shown be equally predictive (i.e. there is no significant difference in performance) compared to the coded data of the consultations., Conclusion: It is possible to extract useful predictors from uncoded consultation notes that improve predictive performance. Techniques linking text to concepts in medical ontologies to derive these predictors are shown to perform best for predicting CRC in our EMR dataset., (Copyright © 2016 Elsevier B.V. All rights reserved.)

Background and Aims: Although optical diagnosis of small colorectal polyps can be achieved at expert centers, accurate prediction of histopathological outcomes has not been achieved in all practice settings. It is therefore important that resected polyps are retrieved for histology. The aims of this study were to evaluate the effect of education and competitive feedback on the overall polyp retrieval rate and to determine which polyp-related and procedure-related factors are associated with retrieval., Methods: We prospectively included consecutive colonoscopies performed at a single center between April 1, 2013 and April 1, 2014. Patients with inflammatory bowel disease or familial polyposis syndromes were excluded from analysis. Six months after the start of the study, all endoscopists were educated on the importance of polyp retrieval, and a competition was started by publicly providing feedback on the retrieval rate of all endoscopists and the monthly best 3 performers (or "golden retrievers") in particular. We compared overall retrieval rates in the 6 months before and after the start of the competition., Results: The overall polyp retrieval rate improved from 88% (525/594) to 93% (978/1047), comparing consecutive colonoscopies performed in the 6 months before and during the polyp retrieval competition (P < .01). The histopathological outcomes of retrieved polyps were not different before and during the competition. The retrieval rate of right-sided polyps increased from 85% to 95% during the competition (odds ratio [OR], 3.3; 95% confidence interval [CI], 2.0-5.4), whereas the left-sided retrieval rate remained 92%. On multivariable analysis, polyp size greater than 5 mm (OR, 4.1; 95% CI, 1.8-9.6) and in competition resection (OR, 1.8; 95% CI, 1.3-2.6) were significantly associated with polyp retrieval, respectively., Conclusion: Providing education and competitive feedback to endoscopists will improve polyp retrieval rates, especially for clinically relevant, right-sided polyps., (Copyright © 2016 American Society for Gastrointestinal Endoscopy. Published by Elsevier Inc. All rights reserved.)

Epidemiological studies indicate that vitamin D exerts a protective effect on the development of various solid cancers. However, concerns have been raised regarding the potential deleterious role of high vitamin D levels in the development of esophageal adenocarcinoma (EAC). This study investigated genetic variation in the vitamin D receptor (VDR) in relation to its expression and risk of Barrett esophagus (BE) and EAC. VDR gene regulation was investigated by immunohistochemistry, reverse transcriptase-polymerase chain reaction (RT-PCR) and gel shift assays. Fifteen haplotype tagging single-nucleotide polymorphisms (SNPs) of the VDR gene were analyzed in 858 patients with reflux esophagitis (RE), BE or EAC and 202 healthy controls. VDR mRNA expression was higher in BE compared with squamous epithelium. VDR protein was located in the nucleus in BE. An rs1989969T/rs2238135G haplotype was identified in the 5' regulatory region of the VDR gene. It was associated with an approximately two-fold reduced risk of RE, BE and EAC. Analysis of a replication cohort was done for BE that confirmed this. The rs1989969T allele causes a GATA-1 transcription factor binding site to appear. The signaling of GATA-1, which is regarded as a negative transcriptional regulator, could explain the findings for rs1989969. The rs2238135G allele was associated with a significantly reduced VDR expression in BE; for the rs1989969T allele, a trend in reduced VDR expression was observed. We identified a VDR haplotype associated with reduced esophageal VDR expression and a reduced incidence of RE, BE and EAC. This VDR haplotype could be useful in identifying individuals who benefit most from vitamin D chemoprevention.

Performing high-quality colonoscopy is one of the important goals of gastroenterology practices and requires achieving a high level of bowel cleansing, performing good and safe polypectomy, and detecting all polyps present in the colon. This article summarizes currently available techniques and technologies to maximize mucosal visualization. Several maneuvers can be applied during insertion and withdrawal of the colonoscope to optimize mucosal visualization and decrease the number of missed polyps. Newly developed technologies support the endoscopist in the detection of polyps. Each technique is reviewed, with emphasis on the impact on colorectal polyp detection., (Copyright © 2015 Elsevier Inc. All rights reserved.)

Background: Adequate bowel preparation is important for optimal colonoscopy. It is important to identify patients at risk for inadequate bowel preparation because this allows taking precautions in this specific group., Objective: To develop a prediction score to identify patients at risk for inadequate bowel preparation who may benefit from an intensified bowel cleansing regimen., Design: Patient and colonoscopy data were prospectively collected, whereas clinical data were retrospectively collected for a total of 1996 colonoscopies in participants who received split-dose bowel preparation. Multivariate logistic regression analyses were conducted in a random two-thirds of the cohort to develop a prediction model. Validation and evaluation of the discriminative power of the prediction model were performed within the remaining one-third of the cohort., Setting: Four centers, including one academic and three medium-to-large size nonacademic centers., Patients: Consecutive colonoscopies in November and December 2012. Mean age was 57.3 ± 15.9 years, 45.8% were male and indications for colonoscopy were screening and/or surveillance (27%), abdominal symptoms and/or blood loss and/or anemia (60%), inflammatory bowel disease (9%), and others (4%)., Interventions: Colonoscopy., Main Outcome Measurements: Inadequate bowel preparation defined as Boston Bowel Preparation Scale score <6., Results: A total of 1331 colonoscopies were included in the development cohort, of which 172 (12.9%) had an inadequate bowel preparation. Independent factors included in the prediction model were American Society of Anesthesiologists Physical Status Classification System score ≥3, use of tricyclic antidepressants, use of opioids, diabetes, chronic constipation, history of abdominal and/or pelvic surgery, history of inadequate bowel preparation, and current hospitalization. The discriminative ability of the scale was good, with an area under the curve of 0.77 in the validation cohort., Limitations: Study design partially retrospective, no data on patient compliance., Conclusion: We developed a validated, easy-to-use prediction scale that can be used to identify subjects with an increased risk of inadequate bowel preparation with good accuracy., (Copyright © 2015 American Society for Gastrointestinal Endoscopy. Published by Elsevier Inc. All rights reserved.)

Chronic gastrointestinal ischemia (CGI) is the result of decreased mucosal perfusion. Typical histological characteristics are lacking which hamper its early diagnosis. Hypoxia-inducible factor-1α (HIF-1α) is expressed under acute hypoxia. We investigated HIF-1α expression in chronic ischemic and inflammatory conditions of the human gastrointestinal (GI) tract. Immunohistochemical expression of HIF-1α was analyzed in 61 patients, including patients with CGI, Helicobacter pylori gastritis, ischemic colitis (IC), infectious colitis, and inflammatory bowel disease (IBD), and 22 controls. HIF-1α expression in >10 % of the cells was regarded as positive staining, and expression <10 % of the cells was considered as negative staining. In the upper GI tract, HIF-1α expression was found in 5/20 CGI patients, but not in controls (p = 0.08). The sensitivity and specificity of HIF-1α expression for diagnosing CGI were 25 and 84%, respectively. In the lower GI tract, HIF-1α was expressed in all patients with IC and infectious colitis and in a majority of IBD patients as well as in 7/12 controls. The sensitivity and specificity of HIF-1α for diagnosing IC were 100 and 51%, respectively. HIF-1α expression was more often (p = 0.02) observed in patients with histological signs of inflammation in the lower GI tract. HIF-1α is expressed in acute and chronic ischemic tissue, but also in normal colon tissue and inflammatory disorders.

Background & Aims: Chronic gastrointestinal ischemia (CGI) is more common than previously thought. Visible light spectroscopy (VLS) allows for noninvasive measurements of mucosal capillary hemoglobin oxygen saturation during endoscopy. We evaluated the response of patients with occlusive CGI to treatment after evaluation by radiologic imaging of the vasculature and VLS. We also identified factors associated with response to treatment in these patients., Methods: In a prospective study, we collected data from 212 patients referred for evaluation of suspected CGI from November 2008 through January 2011. Patients underwent an extensive evaluation that included visualization of gastrointestinal arteries and assessments of mucosal perfusion by means of VLS. Treatment response was evaluated in patients with occlusive CGI. Factors associated with response to therapy were assessed by using multivariate logistic regression analysis., Results: Occlusive CGI was diagnosed in 107 patients (50%); 96 were offered treatment (90%). After median follow-up period of 13 months, data on treatment response were available from 89 patients (93%); 62 patients had a sustained response (70%). Weight loss before treatment (odds ratio [OR], 1.93), presence of an abdominal bruit (OR, 2.36), and corpus mucosal saturation level <56% (OR, 4.84) were the strongest predictors of a positive response to treatment., Conclusions: Treatment of CGI, diagnosed by a multimodal approach, provides a substantial long-term rate of response (70% in 13 months). Weight loss, abdominal bruit, and low corpus mucosal saturation identify patients most likely to respond to treatment. Multiple techniques should therefore be used to assess patients with CGI, including VLS measurements, to detect mucosal hypoxia., (Copyright © 2015 AGA Institute. Published by Elsevier Inc. All rights reserved.)

Background and Study Aims: Local recurrence has been observed after endoscopic mucosal resection (EMR) of nonpedunculated colorectal lesions. The indications for follow-up colonoscopy and the optimal time interval are currently unclear. The aims of this systematic review were to assess the frequency of local recurrence after EMR, to identify risk factors for recurrence, and to provide follow-up recommendations., Methods: A literature search was performed in PubMed, EMBASE, and the Cochrane Library. EMR was defined as endoscopic snare resection after submucosal fluid injection for removal of nonpedunculated adenomas and early carcinomas. Local recurrence was subdivided into early recurrence (detected at the first follow-up colonoscopy) and late recurrence (detected after ≥ 1 previous normal colonoscopy). A random effects meta-analysis was performed to calculate the pooled estimate of risk of recurrence., Results: A total of 33 studies were included. The mean recurrence risk after EMR was 15 % (95 % confidence interval [CI] 12 % - 19 %). Recurrence risk was higher after piecemeal resection (20 %; 95 %CI 16 % - 25 %) than after en bloc resection (3 %; 95 %CI 2 % - 5 %; P < 0.0001). In 15 studies that differentiated between early and late recurrences, 152/173 recurrences (88 %) occurred early. In four studies with follow-up at 3, 6, and ≥ 12 months, 19/25 (76 %) recurrences were detected at 3 months, increasing to 24 (96 %) at 6 months. In multivariable analysis, only piecemeal resection was associated with recurrence (3 of 3 studies)., Conclusion: Local recurrence after EMR of nonpedunculated colorectal lesions occurs in 3 % of en bloc resections and 20 % of piecemeal resections. Piecemeal resection was the only independent risk factor for recurrence. As more than 90 % of recurrences are detected at 6 months after EMR, we propose that 6 months is the optimal initial follow-up interval., (© Georg Thieme Verlag KG Stuttgart · New York.)

Background: Although colonoscopy is the accepted standard for detection of colorectal adenomas and cancers, many adenomas and some cancers are missed. To avoid interval colorectal cancer, the adenoma miss rate of colonoscopy needs to be reduced by improvement of colonoscopy technique and imaging capability. We aimed to compare the adenoma miss rates of full-spectrum endoscopy colonoscopy with those of standard forward-viewing colonoscopy., Methods: We did an international, multicentre, randomised trial at three sites in Israel, one site in the Netherlands, and two sites in the USA between Feb 1, 2012, and March 31, 2013. Patients aged 18-70 years referred for colorectal cancer screening, polyp surveillance, or diagnostic assessment underwent same-day, back-to-back tandem colonoscopy with standard forward-viewing colonoscope and the full-spectrum endoscopy colonoscope. The patients were randomly assigned (1:1), via computer-generated randomisation with block size of 20, to which procedure was done first. The endoscopist was masked to group allocation until immediately before the start of colonoscopy examinations; patients were not masked. The primary endpoint was adenoma miss rates. We did per-protocol analyses. This trial is registered with ClinicalTrials.gov, number NCT01549535., Findings: 197 participants were enrolled. 185 participants were included in the per-protocol analyses: 88 (48%) were randomly assigned to receive standard forward-viewing colonoscopy first, and 97 (52%) to receive full-spectrum endoscopy colonoscopy first. By per-lesion analysis, the adenoma miss rate was significantly lower in patients in the full-spectrum endoscopy group than in those in the standard forward-viewing procedure group: five (7%) of 67 vs 20 (41%) of 49 adenomas were missed (p<0·0001). Standard forward-viewing colonoscopy missed 20 adenomas in 15 patients; of those, three (15%) were advanced adenomas. Full-spectrum endoscopy missed five adenomas in five patients in whom an adenoma had already been detected with first-pass standard forward-viewing colonoscopy; none of these missed adenomas were advanced. One patient was admitted to hospital for colitis detected at colonoscopy, whereas five minor adverse events were reported including vomiting, diarrhoea, cystitis, gastroenteritis, and bleeding., Interpretation: Full-spectrum endoscopy represents a technology advancement for colonoscopy and could improve the efficacy of colorectal cancer screening and surveillance., Funding: EndoChoice., (Copyright © 2014 Elsevier Ltd. All rights reserved.)

Background: Reflux esophagitis (RE) and Barrett's esophagus (BE) are predisposing factors for development of esophageal adenocarcinoma (EAC), the solid tumor with the fastest rising incidence in the Western world. This RE-BE-EAC cascade involves multiple host factors and consequently multiple genes. Polymorphisms in the 3' region of myosin IXB (Myo9B) are associated with chronic inflammatory gastrointestinal disorders like celiac disease and ulcerative colitis, assuming that variation in Myo9B influences the intestinal permeability., Aim: To determine esophageal expression and the genetic variation of the Myo9B gene in the RE-BE-EAC cascade., Methods: DNA from 886 Caucasian participants (198 non-reflux controls, 305 RE, 254 BE, 129 EAC) was collected for the determination of the Myo9B gene polymorphism (rs2305764). Esophageal Myo9B expression was determined on biopsies from normal, RE, BE and EAC epithelium., Results: Genotype G/G was more common in BE (p = 0.032) and EAC (p = 0.046), but not in RE (p = 0.126) compared with the control group. Cytoplasmic Myo9B expression was determined in RE, BE and EAC, but most prominent in epithelial cells of BE and EAC., Conclusions: Genetic variation of Myo9B may play a role in the etiology of BE and EAC by increasing the permeability of the epithelial barrier.

Objective: Esophageal cancer development is a sequence that starts with reflux esophagitis (RE), followed by Barrett's esophagitis (BE), dysplasia, and finally esophageal adenocarcinoma (EAC). Tumor necrosis factor (TNF) is a potent anti-neoplastic agent, hence DNA polymorphisms that reduce TNF levels potentially enhance the development of BE and EAC. The aim of the study was to determine the impact of TNF gene variation on the RE-BE-EAC cascade., Methods: DNA from 887 Caucasian participants (197 controls, 305 RE, 257 BE, 128 EAC) was tested for the gene polymorphism TNF-β NcoI, and TNF production was determined by TNF-α specific immunohistochemistry on esophageal biopsies from these BE (n = 31) and EAC (n = 4) patients., Results: As compared with healthy controls, the TNF-β NcoI A/A genotype was significantly more prevalent in BE (p = 0.04) and EAC patients (p = 0.02), but not in RE patients (p = 0.1). While TNF-α protein levels were invariably high in esophageal biopsies from EAC patients, most esophageal BE samples showed low to moderate TNF levels., Conclusions: Chronic inflammation, like in BE, markedly increase the risk of malignant transformation. In this study, the significantly higher frequency of the TNF-β NcoI A/A genotype and the local TNF expression indicate that the pro-inflammatory cytokine TNF plays a role in the development of BE and EAC.

Reflux esophagitis (RO) and Barrett's esophagus (BO) can cause esophageal adenocarcinoma (OAC). The esophageal mucosa in the RO-BO-OAC cascade is chronically exposed to gastro-esophageal reflux. Epidermal growth factor (EGF) has an important role in the protection and repair of mucosal damage, and non-physiologic levels are associated with gastrointestinal tumors. The aim is to determine the functional effect of EGF gene polymorphisms on RO, BO and OAC development. A cohort of 871 unrelated Dutch Caucasians consisted of 198 healthy controls, 298 RO patients, 246 BO patients and 129 OAC patients. The frequency of the EGF-production-associated 5'UTR A+61G polymorphism was determined in these four groups. EGF immunohistochemistry was performed on BO biopsies. EGF expression was significantly lower in the G/G genotype compared with the A/G (P=0.008) and A/A (P=0.002) group. The G/G genotype was significantly more prevalent in RO (odds ratios (OR)=2.6; 95% confidence intervals (95% CI): 1.3-5.2), BO (OR=3.0; 95% CI: 1.5-6.2) and OAC (OR=4.1; 95% CI: 1.8-9.7) than in controls. The G allele is associated with reduced EGF expression and increased risk for RO, BO and OAC development. This indicates that reduced mucosal protection resulting from genetically decreased EGF expression enhances esophageal tumor development.

Background & Aims: The diagnosis of chronic gastrointestinal ischemia (CGI) remains a clinical challenge. We aimed to assess the diagnostic value of clinical features, visualization of the gastrointestinal arteries, and evaluation of mucosal perfusion in patients clinically suspected of CGI., Methods: A total of 186 patients referred for suspicion of CGI were prospectively included and followed up. All patients had an extensive diagnostic work-up, including visualization of the gastrointestinal arteries with computed tomography, magnetic resonance, or conventional angiography, and mucosal perfusion with tonometry. The reference standard for CGI was persistent clinical response after adequate therapy. The diagnostic value of individual and combined tests was assessed with multivariable logistic regression analysis., Results: A total of 116 (62%) patients were diagnosed with CGI. In a multivariable model solely based on clinical features, the strongest predictors for CGI were the presence of postprandial pain, weight loss per month in kilograms, concomitant cardiovascular disease, and presence of an abdominal bruit. However, this model showed limited discriminative ability for the presence or absence of CGI (c-statistic, 0.62). Adding radiologic imaging to the prediction model improved the discriminative ability substantially (c-statistic, 0.81). Adding tonometry to the prediction model further improved the discriminative ability of the model (c-statistic, 0.90). The combination of clinical features and tonometry with a c-statistic of 0.88 approximated the discriminative ability of the latter model., Conclusions: Clinical features alone have a limited value to assess CGI correctly. Visualization of the gastrointestinal arteries and evaluation of mucosal perfusion substantially improve the diagnosis of CGI. The strongest diagnostic contribution comes from mucosal perfusion assessment., (Copyright © 2011 AGA Institute. Published by Elsevier Inc. All rights reserved.)

Barrett's oesophagus (BO) is thought to be an intermediate step in the progression from reflux oesophagitis (RO) to oesophageal adenocarcinoma. Premalignant conditions that develop in the presence of chronic inflammation are often associated with the development of a more pronounced humoral immune response during progression of the disease. The aim of this study was to determine whether BO is also associated with a more pronounced humoral immune response when compared to RO. Immunohistochemical studies were performed to quantify the mean numbers of Th2 effector cells (plasma cells and mast cells) and Th1 effector cells (macrophages and CD8(+) T cells) to detect the antibody classes produced by plasma cells (IgA, IgG, IgM or IgE) and to determine the presence of isolated lymph follicles [segregated B and T cell areas, follicular dendritic cells (CD23) and expression of CXCL13] in 124 oesophageal biopsies from 20 patients with BO and 20 patients with RO. The proportion of Th2 effector cells was higher in BO than in RO, mainly due to higher numbers of plasma cells and mast cells in BO (p < 0.001). Most plasma cells in BO and RO expressed IgG, but several IgE(+) plasma cells were detected in BO: these were rare in RO. In line with this, isolated lymph follicles were observed in 4/20 (20%) patients with BO, but not in RO. We therefore conclude that the inflammatory response is skewed towards a more pronounced humoral immune response when RO progresses to BO. It may be that this shift, which is similar to that found in other chronic inflammatory conditions, contributes to an increased cancer risk in BO., (Copyright (c) 2005 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.)