Geographic Atrophy Segmentation Using Multimodal Deep Learning.

Purpose: To examine deep learning (DL)-based methods for accurate segmentation of geographic atrophy (GA) lesions using fundus autofluorescence (FAF) and near-infrared (NIR) images. Methods: This retrospective analysis utilized imaging data from study eyes of patients enrolled in Proxima A and B (NCT02479386; NCT02399072) natural history studies of GA. Two multimodal DL networks (UNet and YNet) were used to automatically segment GA lesions on FAF; segmentation accuracy was compared with annotations by experienced graders. The training data set comprised 940 image pairs (FAF and NIR) from 183 patients in Proxima B; the test data set comprised 497 image pairs from 154 patients in Proxima A. Dice coefficient scores, Bland-Altman plots, and Pearson correlation coefficient (r) were used to assess performance. Results: On the test set, Dice scores for the DL network to grader comparison ranged from 0.89 to 0.92 for screening visit; Dice score between graders was 0.94. GA lesion area correlations (r) for YNet versus grader, UNet versus grader, and between graders were 0.981, 0.959, and 0.995, respectively. Longitudinal GA lesion area enlargement correlations (r) for screening to 12 months (n = 53) were lower (0.741, 0.622, and 0.890, respectively) compared with the cross-sectional results at screening. Longitudinal correlations (r) from screening to 6 months (n = 77) were even lower (0.294, 0.248, and 0.686, respectively). Conclusions: Multimodal DL networks to segment GA lesions can produce accurate results comparable with expert graders. Translational Relevance: DL-based tools may support efficient and individualized assessment of patients with GA in clinical research and practice.

Identifying clinically useful biomarkers in neurodegenerative disease through a collaborative approach: the NeuroToolKit.

BACKGROUND: Alzheimer's disease (AD) is a complex and heterogeneous disease, which requires reliable biomarkers for diagnosis and monitoring disease activity. Preanalytical protocol and technical variability associated with biomarker immunoassays makes comparability of biomarker data across multiple cohorts difficult. This study aimed to compare cerebrospinal fluid (CSF) biomarker results across independent cohorts, including participants spanning the AD continuum. METHODS: Measured on the NeuroToolKit (NTK) prototype panel of immunoassays, 12 CSF biomarkers were evaluated from three cohorts (ALFA+, Wisconsin, and Abby/Blaze). A correction factor was applied to biomarkers found to be affected by preanalytical procedures (amyloid-ß1-42, amyloid-ß1-40, and alpha-synuclein), and results between cohorts for each disease stage were compared. The relationship between CSF biomarker concentration and cognitive scores was evaluated. RESULTS: Biomarker distributions were comparable across cohorts following correction. Correlations of biomarker values were consistent across cohorts, regardless of disease stage. Disease stage differentiation was highest for neurofilament light (NfL), phosphorylated tau, and total tau, regardless of the cohort. Correlation between biomarker concentration and cognitive scores was comparable across cohorts, and strongest for NfL, chitinase-3-like protein-1 (YKL40), and glial fibrillary acidic protein. DISCUSSION: The precision of the NTK enables merging of biomarker datasets, after correction for preanalytical confounders. Assessment of multiple cohorts is crucial to increase power in future studies into AD pathogenesis.

Deep Learning to Predict Geographic Atrophy Area and Growth Rate from Multimodal Imaging.

OBJECTIVE: To develop deep learning models for annualized geographic atrophy (GA) growth rate prediction using fundus autofluorescence (FAF) images and spectral-domain OCT volumes from baseline visits, which can be used for prognostic covariate adjustment to increase power of clinical trials. DESIGN: This retrospective analysis estimated GA growth rate as the slope of a linear fit on all available measurements of lesion area over a 2-year period. Three multitask deep learning models-FAF-only, OCT-only, and multimodal (FAF and OCT)-were developed to predict concurrent GA area and annualized growth rate. PARTICIPANTS: Patients were from prospective and observational lampalizumab clinical trials. METHODS: The 3 models were trained on the development data set, tested on the holdout set, and further evaluated on the independent test sets. Baseline FAF images and OCT volumes from study eyes of patients with bilateral GA (NCT02247479; NCT02247531; and NCT02479386) were split into development (1279 patients/eyes) and holdout (443 patients/eyes) sets. Baseline FAF images from study eyes of NCT01229215 (106 patients/eyes) and NCT02399072 (169 patients/eyes) were used as independent test sets. MAIN OUTCOME MEASURES: Model performance was evaluated using squared Pearson correlation coefficient (r2) between observed and predicted lesion areas/growth rates. Confidence intervals were calculated by bootstrap resampling (B = 10 000). RESULTS: On the holdout data set, r2 (95% confidence interval) of the FAF-only, OCT-only, and multimodal models for GA lesion area prediction was 0.96 (0.95-0.97), 0.91 (0.87-0.95), and 0.94 (0.92-0.96), respectively, and for GA growth rate prediction was 0.48 (0.41-0.55), 0.36 (0.29-0.43), and 0.47 (0.40-0.54), respectively. On the 2 independent test sets, r2 of the FAF-only model for GA lesion area was 0.98 (0.97-0.99) and 0.95 (0.93-0.96), and for GA growth rate was 0.65 (0.52-0.75) and 0.47 (0.34-0.60). CONCLUSIONS: We show the feasibility of using baseline FAF images and OCT volumes to predict individual GA area and growth rates using a multitask deep learning approach. The deep learning-based growth rate predictions could be used for covariate adjustment to increase power of clinical trials. FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found after the references.

Clinical performance and robustness evaluation of plasma amyloid-ß42/40 prescreening.

INTRODUCTION: Further evidence is needed to support the use of plasma amyloid ß (Aß) biomarkers as Alzheimer's disease prescreening tools. This study evaluated the clinical performance and robustness of plasma Aß42 /Aß40 for amyloid positivity prescreening. METHODS: Data were collected from 333 BioFINDER and 121 Alzheimer's Disease Neuroimaging Initiative study participants. Risk and predictive values versus percentile of plasma Aß42 /Aß40 evaluated the actionability of plasma Aß42 /Aß40 , and simulations modeled the impact of potential uncertainties and biases. Amyloid PET was the brain amyloidosis reference standard. RESULTS: Elecsys plasma Aß42 /Aß40 could potentially rule out amyloid pathology in populations with low-to-moderate amyloid positivity prevalence. However, simulations showed small measurement or pre-analytical errors in Aß42 and/or Aß40 cause misclassifications, impacting sensitivity or specificity. The minor fold change between amyloid PET positive and negative cases explains the biomarkers low robustness. DISCUSSION: Implementing plasma Aß42 /Aß40 for routine clinical use may pose significant challenges, with misclassification risks. HIGHLIGHTS: Plasma Aß42 /Aß40 ruled out amyloid PET positivity in a setting of low amyloid-positive prevalence. Including (pre-) analytical errors or measurement biases caused misclassifications. Plasma Aß42 /Aß40 had a low inherent dynamic range, independent of analytical method. Other blood biomarkers may be easier to implement as robust prescreening tools.

Evaluating the Safety and Efficacy of Crenezumab vs Placebo in Adults With Early Alzheimer Disease: Two Phase 3 Randomized Placebo-Controlled Trials.

Importance: Alzheimer disease (AD), a neurodegenerative disease characterized by ß-amyloid plaques and τ tangles in the brain, represents an unmet medical need with no fully approved therapeutics to modify disease progression. Objective: To investigate the safety and efficacy of crenezumab, a humanized monoclonal immunoglobulin G4 antibody targeting ß-amyloid oligomers, in participants with prodromal to mild (early) AD. Design, Setting, and Participants: Two phase 3 multicenter randomized double-blind placebo-controlled parallel-group efficacy and safety studies of crenezumab in participants with early AD, CREAD and CREAD2, were initiated in 2016 and 2017, respectively, and were designed to evaluate the efficacy and safety of crenezumab in participants with early AD. CREAD (194 sites in 30 countries) and CREAD2 (209 sites in 27 countries) were global multicenter studies. A total of 3736 and 3664 participants were screened in CREAD and CREAD2, respectively. A total of 3736 and 3664 participants were screened in CREAD and CREAD2, respectively. Both trials enrolled individuals aged 50 to 85 years with early AD. Participants with some comorbidities and evidence of cerebral infarction or more than 4 microbleeds or areas of leptomeningeal hemosiderosis on magnetic resonance imaging were excluded. After 2923 and 2858 were excluded, respectively, 813 participants in CREAD and 806 in CREAD2 were randomly assigned in a 1:1 ratio to either placebo or crenezumab. In the final analysis, there were 409 participants in the placebo group and 404 in the crenezumab group in CREAD and 399 in the placebo group and 407 in the crenezumab group in CREAD2. Data were analyzed up until January 2019 and August 2019, respectively. Interventions: Participants received placebo or 60 mg/kg crenezumab intravenously every 4 weeks for up to 100 weeks. Main Outcomes and Measures: The primary outcome was change from baseline to week 105 in Clinical Dementia Rating-Sum of Boxes (CDR-SB) score. Results: There were 813 participants in CREAD (mean [SD] age, 70.7 [8.2] years; 483 female and 330 male) and 806 in CREAD2 (mean [SD] age, 70.9 [7.7] years; 456 female and 350 male). Baseline characteristics were balanced between both groups. The between-group difference in mean change from baseline in CDR-SB score (placebo minus crenezumab) was -0.17 (95% CI, -0.86 to 0.53; P = .63) at week 105 in the CREAD study (88 placebo; 86 crenezumab). Compared with previous trials, no new safety signals were identified, and amyloid-related imaging abnormalities with edema were rare, mild, and transient. No meaningful changes in AD biomarkers were observed. Both studies were discontinued following a preplanned interim analysis indicating that CREAD was unlikely to meet the primary end point. Conclusions and Relevance: Crenezumab was well tolerated but did not reduce clinical decline in participants with early AD. Trial Registration: ClinicalTrials.gov Identifiers: CREAD, NCT02670083; CREAD2, NCT03114657.

Target engagement in an alzheimer trial: Crenezumab lowers amyloid ß oligomers in cerebrospinal fluid.

OBJECTIVE: Oligomeric forms of amyloid ß protein (oAß) are believed to be principally responsible for neurotoxicity in Alzheimer disease (AD), but it is not known whether anti-Aß antibodies are capable of lowering oAß levels in humans. METHODS: We developed an ultrasensitive immunoassay and used it to measure oAß in cerebrospinal fluid (CSF) from 104 AD subjects participating in the ABBY and BLAZE phase 2 trials of the anti-Aß antibody crenezumab. Patients received subcutaneous (SC) crenezumab (300mg) or placebo every 2 weeks, or intravenous (IV) crenezumab (15mg/kg) or placebo every 4 weeks for 68 weeks. Ninety-eight of the 104 patients had measurable baseline oAß levels, and these were compared to levels at week 69 in placebo (n = 28), SC (n = 35), and IV (n = 35) treated patients. RESULTS: Among those receiving crenezumab, 89% of SC and 86% of IV patients had lower levels of oAß at week 69 versus baseline. The difference in the proportion of patients with decreasing levels was significant for both treatment arms: p = 0.0035 for SC and p = 0.01 for IV crenezumab versus placebo. The median percentage change was -48% in the SC arm and -43% in the IV arm. No systematic change was observed in the placebo group, with a median change of -13% and equivalent portions with negative and positive change. INTERPRETATION: Crenezumab lowered CSF oAß levels in the large majority of treated patients tested. These results support engagement of the principal pathobiological target in AD and identify CSF oAß as a novel pharmacodynamic biomarker for use in trials of anti-Aß agents. ANN NEUROL 2019;86:215-224.

Amyloid positron emission tomography and cerebrospinal fluid results from a crenezumab anti-amyloid-beta antibody double-blind, placebo-controlled, randomized phase II study in mild-to-moderate Alzheimer's disease (BLAZE).

BACKGROUND: We investigated the effect of crenezumab, a humanized anti-amyloid-beta (Aß) immunoglobulin (Ig)G4 monoclonal antibody, on biomarkers of amyloid pathology, neurodegeneration, and disease progression in patients with mild-to-moderate Alzheimer's disease (AD). METHODS: This double-blind, placebo-controlled, randomized phase II study enrolled patients with mild-to-moderate AD and a Mini-Mental State Examination (MMSE) score of 18-26. In part 1 of the study, patients were 2:1 randomized to receive low-dose subcutaneous (SC) 300 mg crenezumab every 2 weeks (q2w) or placebo for 68 weeks; in part 2, patients were 2:1 randomized to receive high-dose intravenous (IV) 15 mg/kg crenezumab every 4 weeks (q4w) or placebo for 68 weeks. The primary endpoint was change in amyloid burden from baseline to week 69 assessed by florbetapir positron emission tomography (PET) in the modified intent-to-treat population. Secondary endpoints were change from baseline to week 69 in cerebrospinal fluid (CSF) biomarkers and fluorodeoxyglucose PET, and change from baseline to week 73 in 12-point Alzheimer's Disease Assessment Scale cognitive subscale (ADAS-Cog12) and Clinical Dementia Rating Sum of Boxes (CDR-SB). Safety was assessed in patients who received at least one dose of study treatment. RESULTS: From August 2011 to September 2012, 91 patients were enrolled and randomized (low-dose SC cohort: crenezumab (n = 26) or placebo (n = 13); high-dose IV cohort: crenezumab (n = 36) or placebo (n = 16)). The primary endpoint was not met using a prespecified cerebellar reference region to calculate standard uptake value ratios (SUVRs) from florbetapir PET. Exploratory analyses using subcortical white matter reference regions showed nonsignificant trends toward slower accumulation of plaque amyloid in the high-dose IV cohort. In both cohorts, a significant mean increase from baseline in CSF Aß(1-42) levels versus placebo was observed. Nonsignificant trends toward ADAS-Cog12 and CDR-SB benefits were identified in a mild (MMSE 20-26) subset of the high-dose IV cohort. No amyloid-related imaging abnormalities due to edema/effusion were observed. CONCLUSION: The primary endpoint was not met. Exploratory findings suggest potential Aß target engagement with crenezumab and possible slower accumulation of plaque amyloid. Studies investigating the effects of higher doses of crenezumab on amyloid load and disease progression are ongoing. TRIAL REGISTRATION: ClinicalTrials.gov, NCT01397578 . Registered on 18 July 2011.

ABBY: A phase 2 randomized trial of crenezumab in mild to moderate Alzheimer disease.

OBJECTIVE: To evaluate the safety and efficacy of crenezumab in patients with mild to moderate Alzheimer disease (AD). METHODS: In this phase 2 trial, 431 patients with mild to moderate AD 50 to 80 years of age were randomized 2:1 (crenezumab:placebo). Patients received low-dose subcutaneous crenezumab 300 mg or placebo every 2 weeks (n = 184) or high-dose intravenous crenezumab 15 mg/kg or placebo every 4 weeks (n = 247) for 68 weeks. Primary outcome measures were change in Alzheimer's Disease Assessment Scale-Cognitive Subscale (ADAS-Cog12) and Clinical Dementia Rating-Sum of Boxes scores from baseline to week 73. RESULTS: The primary and secondary endpoints were not met. In an exploratory post hoc analysis, a reduction in decline on the ADAS-Cog12 was observed in the high-dose group. Separation from the placebo group on the ADAS-Cog12 was greatest in the milder subsets of AD patients and reached statistical significance in the group with Mini-Mental State Examination scores of 22 to 26. In both groups, there was a significant increase in CSF ß-amyloid1-42 levels that correlated with crenezumab CSF levels. The overall rate of adverse events was balanced between groups. One case of amyloid-related imaging abnormalities indicative of vasogenic edema or effusions was reported. CONCLUSIONS: Although prespecified criteria for testing treatment effects were not met, these data suggest a potential treatment effect in patients with mild AD treated with high-dose crenezumab. Together with the safety profile for crenezumab, these data support the exploration of crenezumab treatment at even higher doses in patients with early AD. CLINICALTRIALSGOV IDENTIFIER: NCT 01343966. CLASSIFICATION OF EVIDENCE: This study provides Class II evidence that, for people with AD, crenezumab does not significantly improve cognition or function at 18 months. The study is rated Class II because <80% of enrolled patients completed the study.

CSF biomarkers of Alzheimer's disease concord with amyloid-ß PET and predict clinical progression: A study of fully automated immunoassays in BioFINDER and ADNI cohorts.

INTRODUCTION: We studied whether fully automated Elecsys cerebrospinal fluid (CSF) immunoassay results were concordant with positron emission tomography (PET) and predicted clinical progression, even with cutoffs established in an independent cohort. METHODS: Cutoffs for Elecsys amyloid-ß1-42 (Aß), total tau/Aß(1-42), and phosphorylated tau/Aß(1-42) were defined against [18F]flutemetamol PET in Swedish BioFINDER (n = 277) and validated against [18F]florbetapir PET in Alzheimer's Disease Neuroimaging Initiative (n = 646). Clinical progression in patients with mild cognitive impairment (n = 619) was studied. RESULTS: CSF total tau/Aß(1-42) and phosphorylated tau/Aß(1-42) ratios were highly concordant with PET classification in BioFINDER (overall percent agreement: 90%; area under the curve: 94%). The CSF biomarker statuses established by predefined cutoffs were highly concordant with PET classification in Alzheimer's Disease Neuroimaging Initiative (overall percent agreement: 89%-90%; area under the curves: 96%) and predicted greater 2-year clinical decline in patients with mild cognitive impairment. Strikingly, tau/Aß ratios were as accurate as semiquantitative PET image assessment in predicting visual read-based outcomes. DISCUSSION: Elecsys CSF biomarker assays may provide reliable alternatives to PET in Alzheimer's disease diagnosis.

Technical performance of a novel, fully automated electrochemiluminescence immunoassay for the quantitation of ß-amyloid (1-42) in human cerebrospinal fluid.

INTRODUCTION: Available assays for quantitation of the Alzheimer's disease (AD) biomarker amyloid-beta 1-42 (Aß [1-42]) in cerebrospinal fluid demonstrate significant variability and lack of standardization to reference measurement procedures (RMPs). We report analytical performance data for the novel Elecsys ß-amyloid (1-42) assay (Roche Diagnostics). METHODS: Lot-to-lot comparability was tested using method comparison. Performance parameters were measured according to Clinical & Laboratory Standards Institute (CLSI) guidelines. The assay was standardized to a Joint Committee for Traceability in Laboratory Medicine (JCTLM) approved RMP. RESULTS: Limit of quantitation was <11.28 pg/mL, and the assay was linear throughout the measuring range (200-1700 pg/mL). Excellent lot-to-lot comparability was observed (correlation coefficients [Pearson's r] >0.995; bias in medical decision area <2%). Repeatability coefficients of variation (CVs) were 1.0%-1.6%, intermediate CVs were 1.9%-4.0%, and intermodule CVs were 1.1%-3.9%. Estimated total reproducibility was 2.0%-5.1%. Correlation with the RMP was good (Pearson's r, 0.93). DISCUSSION: The Elecsys ß-amyloid (1-42) assay has high analytical performance that may improve biomarker-based AD diagnosis.

Blood-sampling collection prior to surgery may have a significant influence upon biomarker concentrations measured.

BACKGROUND: Biomarkers can be subtle tools to aid the diagnosis, prognosis and monitoring of therapy and disease progression. The validation of biomarkers is a cumbersome process involving many steps. Serum samples from lung cancer patients were collected in the framework of a larger study for evaluation of biomarkers for early detection of lung cancer. The analysis of biomarker levels measured revealed a noticeable difference in certain biomarker values that exhibited a dependence of the time point and setting of the sampling. Biomarker concentrations differed significantly if taken before or after the induction of anesthesia and if sampled via venipuncture or arterial catheter. METHODS: To investigate this observation, blood samples from 13 patients were drawn 1-2 days prior to surgery (T1), on the same day by venipuncture (T2) and after induction of anesthesia via arterial catheter (T3). The biomarkers Squamous Cell Carcinoma antigen (CanAG SCC EIA, Fujirebio Diagnostics, Malvern, USA), Carcinoembrionic Antigen (CEA), and CYFRA 21-1 (Roche Diagnostics GmbH, Mannheim, Germany) were analyzed. RESULTS: SCC showed a very strong effect in relation to the sampling time and procedure. While the first two points in time (T1; T2) were highly comparable (median fold-change: 0.84; p = 0.7354; correlation ρ = 0.883), patients showed a significant increase (median fold-change: 4.96; p = 0.0017; correlation ρ = -0.036) in concentration when comparing T1 with the sample time subsequent to anesthesia induction (T3). A much weaker increase was found for CYFRA 21-1 at T3 (median fold-change: 1.40; p = 0.0479). The concentration of CEA showed a very small, but systematic decrease (median fold-change: 0.72; p = 0.0039). CONCLUSIONS: In this study we show the unexpectedly marked influence of blood withdrawal timing (before vs. after anesthesia) and procedure (venous versus arterial vessel puncture) has on the concentration of the protein biomarker SCC and to a less extent upon CYFRA21-1. The potential causes for these effects remain to be elucidated in subsequent studies, however these findings highlight the importance of a standardized, controlled blood collection protocol for biomarker detection.

Upregulation of Periostin and Reactive Stroma Is Associated with Primary Chemoresistance and Predicts Clinical Outcomes in Epithelial Ovarian Cancer.

PURPOSE: Up to one third of ovarian cancer patients are intrinsically resistant to platinum-based treatment. However, predictive and therapeutic strategies are lacking due to a poor understanding of the underlying molecular mechanisms. This study aimed to identify key molecular characteristics that are associated with primary chemoresistance in epithelial ovarian cancers. EXPERIMENTAL DESIGN: Gene expression profiling was performed on a discovery set of 85 ovarian tumors with clinically well-defined response to chemotherapies as well as on an independent validation dataset containing 138 ovarian patients from the chemotreatment arm of the ICON7 trial. RESULTS: We identified a distinct "reactive stroma" gene signature that is specifically associated with primary chemoresistant tumors and was further upregulated in posttreatment recurrent tumors. Immunohistochemistry (IHC) and RNA in situ hybridization (RNA ISH) analyses on three of the highest-ranked signature genes (POSTN, LOX, and FAP) confirmed that modulation of the reactive stroma signature genes within the peritumoral stromal compartments was specifically associated with the clinical chemoresistance. Consistent with these findings, chemosensitive ovarian cells grown in the presence of recombinant POSTN promoted resistance to carboplatin and paclitaxel treatment in vitro. Finally, we validated the reactive stroma signature in an independent dataset and demonstrated that a high POSTN expression level predicts shorter progression-free survival following first-line chemotherapy. CONCLUSIONS: Our findings highlight the important interplay between cancer and the tumor microenvironment in ovarian cancer biology and treatment. The identified reactive stromal components in this study provide a molecular basis to the further development of novel diagnostic and therapeutic strategies for overcoming chemoresistance in ovarian cancer.

IL-6 pathway-driven investigation of response to IL-6 receptor inhibition in rheumatoid arthritis.

OBJECTIVES: To determine whether heterogeneity in interleukin-6 (IL-6), IL-6 receptor and other components of the IL-6 signalling pathway/network, at the gene, transcript and protein levels, correlate with disease activity in patients with rheumatoid arthritis (RA) and with clinical response to tocilizumab. DESIGN: Biomarker samples and clinical data for five phase 3 trials of tocilizumab were analysed using serum (3751 samples), genotype (927 samples) and transcript (217 samples) analyses. Linear regression was then used to assess the association between these markers and either baseline disease activity or treatment response. RESULTS: Higher baseline serum IL-6 levels were significantly associated (p<0.0001) with higher baseline DAS28, erythrocyte sedimentation rate, C reactive protein and Health Assessment Questionnaire in patients whose responses to disease-modifying antirheumatic drugs (DMARD-IR) and to antitumour necrosis factor (aTNF-IR) were inadequate and patients who were naive/responders to methotrexate (MTX). Higher baseline serum IL-6 levels were also significantly associated with better clinical response to tocilizumab (versus placebo) measured by cDAS28 in the pooled DMARD-IR (p<0.0001) and MTX-naive populations (p=0.04). However, the association with treatment response was weak. A threefold difference in baseline IL-6 level corresponded to only a 0.17-unit difference in DAS28 at week 16. IL-6 pathway single nucleotide polymorphisms and RNA levels also were not strongly associated with treatment response. CONCLUSIONS: Our analyses illustrate that the biological activity of a disease-associated molecular pathway may impact the benefit of a therapy targeting that pathway. However, the variation in pathway activity, as measured in blood, may not be a strong predictor. These data suggest that the major contribution to variability in clinical responsiveness to therapeutics in RA remains unknown.

A PAUC-based estimation technique for disease classification and biomarker selection.

The partial area under the receiver operating characteristic curve (PAUC) is a well-established performance measure to evaluate biomarker combinations for disease classification. Because the PAUC is defined as the area under the ROC curve within a restricted interval of false positive rates, it enables practitioners to quantify sensitivity rates within pre-specified specificity ranges. This issue is of considerable importance for the development of medical screening tests. Although many authors have highlighted the importance of PAUC, there exist only few methods that use the PAUC as an objective function for finding optimal combinations of biomarkers. In this paper, we introduce a boosting method for deriving marker combinations that is explicitly based on the PAUC criterion. The proposed method can be applied in high-dimensional settings where the number of biomarkers exceeds the number of observations. Additionally, the proposed method incorporates a recently proposed variable selection technique (stability selection) that results in sparse prediction rules incorporating only those biomarkers that make relevant contributions to predicting the outcome of interest. Using both simulated data and real data, we demonstrate that our method performs well with respect to both variable selection and prediction accuracy. Specifically, if the focus is on a limited range of specificity values, the new method results in better predictions than other established techniques for disease classification.

Predictive factors of mortality after PEG insertion: guidance for clinical practice.

BACKGROUND: Percutaneous endoscopic gastrostomy (PEG) is considered the preferred route for long-term enteral feeding. The aim of this study was to determine predictors of an increased mortality risk after PEG insertion. METHODS: A retrospective study was conducted during a 13-year period in the gastroenterology department of Erlangen University Hospital. The authors completed a questionnaire with details of demographic data, diagnosis, indication for PEG, type of tube, and cause of death. Patients were contacted regularly at scheduled appointments. RESULTS: In total, 787 patients (574 male [72.9%]) underwent PEG placement by the pull technique. The main underlying disease was malignant (75.6%). By the end of the study period, 614 patients had died. The average survival time was 720 days. The 30-, 60-, 90-day and 1-, 3-, and 5-year mortality rates amounted to 6.5%, 9.8%, 13%, 32.1%, 59.3%, and 69.8%, respectively. Predictive factors of increased 30-day mortality were higher age, lower body mass index (BMI), and the presence of diabetes mellitus. The presence of all 3 variables served as an indicator to detect high-risk patients, with a sensitivity of 0.80 and a specificity of 0.64. CONCLUSION: Mortality predictors for patients after PEG insertion are higher age, lower BMI, and the presence of diabetes mellitus. To avoid unnecessary and dangerous examinations in high-risk patients, the above-mentioned predictive factors of mortality should be checked before PEG placement.

Alterations of taste perception in Crohn's disease and their dependency on disease activity and nutritional behavior.

BACKGROUND: The pathogenesis of Crohn's disease (CD) is unknown. Besides immunoregulatory, genetic and environmental aspects, a nutritional impact is also encountered. Whether taste perception exerts any influence on an increased consumption of carbohydrates is unknown. OBJECTIVE: To evaluate nutritional habits and taste perception in CD patients, either in active or inactive disease stages. DESIGN: A prospective study was performed with 31 active and 27 inactive CD patients, and 30 age and sex-matched healthy subjects. Nutritional behavior was determined using an extensive optical nutrition questionnaire and taste perception was assessed by a 3-drop method with exceeding dilution tests. RESULTS: Active and inactive CD patients exhibited a significant increased taste threshold for the detection of all solutions (bitter: P=0.0012; salty: P=0.0198; sour: P=0.0021; and sweet: P=0.0208). For recognition, the determination of bitter solution (P=0.0014) was significantly reduced in CD patients compared with healthy subjects. No impact of clinical or objective parameters of inflammation on taste perception could be established. The consumption of refined sugar in CD patients was higher than in healthy subjects, though not significant. CONCLUSIONS: An increased taste threshold for the detection of all 4 taste qualities in active and inactive CD patients suggests a systemic pathogenesis, such as an inflammation of the oral cavity, as a manifestation of CD. In this study, changes in taste threshold were not associated with altered sugar consumption.

Risk factors associated with adverse drug reactions following hospital admission: a prospective analysis of 907 patients in two German university hospitals.

BACKGROUND: Since the 1970s, studies have examined potential risk factors associated with adverse drug reactions (ADRs) in a variety of settings. However, no pharmacoepidemiological study exists that incorporates clinical and laboratory parameters in a multiple regression model in order to consider predictors for ADRs. OBJECTIVES: To characterize risk factors associated with ADRs in patients admitted to university hospital departments of internal medicine. DESIGN AND SETTING: Intensive pharmacovigilance was carried out in departments of internal medicine of two university hospitals. All admissions were followed prospectively for the occurrence of ADRs by members of a pharmacoepidemiological team consisting of physicians, pharmacologists and pharmacists. To identify patients at high risk for experiencing ADRs, patient histories and several clinical and laboratory data, determined at the time of admission, were taken into consideration. In addition to the drug prescribed, 40 parameters defined vital status at admission. These included temperature, heart rate, blood pressure (systolic-diastolic), body mass index, nicotine and alcohol use, and first laboratory test results after admission on nutrition status, inflammation, liver, kidney, pancreas or thyroid status, electrolytes, blood count and coagulation. RESULTS: 907 patients were observed during the study period. The mean age of the study population was 60 +/- 16 years. The median number of different drugs administered per patient during hospitalization was 9.6 +/- 7.7. In 345 patients, 592 ADRs were evaluated: 33.4% possible, 61.5% probable and 4.7% highly probable. Two ADR-related deaths were observed during the study period. Analysing ADR predictors, 17 of 40 parameters reached significance in univariate analysis, but only five in a multivariate binary regression model: raised temperature (odds ratio [OR] 1.609; 95% CI 1.133, 2.285), low erythrocyte levels (OR 0.386; 95% CI 0.194, 0.768), low thrombocyte levels (OR 0.788, 95% CI 0.627, 0.989), high number of drugs (OR 1.117; 95% CI 1.076, 1.159) and female sex (OR 1.562; 95% CI 0.785, 2.013) were independent predictors for ADRs. CONCLUSION: For the patients investigated, of the large number of clinical data available only five independent factors predict ADR occurrence. Taking these results into account, physicians will be able to focus early on patients at risk for ADRs. To minimize ADR occurrence, ADR predictors should be integrated into the clinical pathway.

Potential of ultrasound diagnosis for parotid tumors: analysis of qualitative and quantitative parameters.

OBJECTIVE: Histology of parotid tumors determines the extent of surgery. The aim was to test ultrasound (US) contrast enhancer-kinetics to identify histologic entities, possibly being superior to qualitative morphological parameters. STUDY DESIGN: In a cross-sectional assessment of ultrasound diagnosis, the subjective US-classification was compared with contrast analysis with histology as gold standard. SUBJECTS AND METHODS: A total of 64 male and 61 female patients with a mean age of 54 years were included, with 13 malignant tumors. These were classified with US morphology, then time-dependent contrast medium analysis. RESULTS: A total of 92.8% of tumors were classified correctly as malignant or benign. The sensitivity, specificity, positive- and negative-predictive values were 66.7%, 86.3%, 60.6%, and 89.1% for differentiating Warthin tumors, but only 46.2%, 98.2%, 75%, and 94% for malignant lesions. Contrast parameters yielded significant parameters for benign tumors, not for malignant entities. CONCLUSION: Although contrast medium analysis provided statistical criteria, these, however, do not possess the ability to improve the diagnostic prediction of tumor histology. Neither the morphologic classification nor contrast medium analysis was able to identify a malignant lesion sufficiently.

The value of dynamic MRI studies in parotid tumors.

RATIONALE AND OBJECTIVES: To evaluate the ability of dynamic contrast-enhanced magnetic resonance imaging (MRI) to differentiate several tumor entities of the parotid gland in a prospective clinical trial. MATERIALS AND METHODS: A total of 112 patients with parotid tumors were examined with dynamic contrast-enhanced 1.5 T MRI. Precontrast axial T1-weighted imaging was used to select five slices for the dynamic study. Subsequently, a T1-weighted FLASH sequence was used for the dynamic contrast study (0.2 ml Gd/kg x body weight). Contrast agent application and the FLASH sequence were started simultaneously. Ten acquisitions of 10 seconds' scan time each were performed (total acquisition time 1:40 minutes). Signal intensity versus time (SIvT) curves was obtained for all tumors. After correlation of the categorized SIvT curves, these were compared with histopathology. Finally, all MRIs together with the tumor specific SIvT curves were re-read and correlated with histopathologic diagnosis. All reading sessions were done by three experienced radiologists. RESULTS: Four characteristic intensity-time curves were observed: pleomorphic adenoma showed a gradual increase in signal intensity, followed by a plateau phase on a low intensity level. Cysts showed a vacillating course at a low signal intensity level. Adenolymphomas as well as carcinomas showed a rapid increase in signal intensity followed by a plateau phase. Statistic significance was found for the time-to-peak values for adenolymphomas and pleomorphic adenomas and for the maximum peak signal intensity values for carcinomas. Together with other morphologic MRI criteria (contrast enhancement, border characteristics) and clinical features, a differentiation between adenolymphoma and carcinoma was possible. CONCLUSIONS: With additional dynamic contrast-enhanced MRI, a more reliable differentiation between common parotid tumors is possible before surgery.

Generalized approaches to partitioning the attributable risk of interacting risk factors can remedy existing pitfalls.

OBJECTIVE: To clarify the properties of different approaches to estimate the contribution of single-risk factors to the disease load in a population. STUDY DESIGN AND SETTING: Three methods of partitioning attributable risks are reviewed and two additional procedures as modifications of the existing algorithms are introduced. Basis properties of the approaches are outlined in the simplest setting with two exposure variables. The extension to more complex settings is illustrated by an example involving three risk factors. RESULTS: The quantification of the impact of single-risk factors can vary considerably according to the method used. Different orderings of the risk factors with respect to their impact can occur. Approaches can be classified according to two features: (i) inclusion or exclusion of partial interactions between risk factors, (ii) equal or proportional distribution of the surplus resulting from the combined action of risk factors. Practical applications have to carefully consider intrinsic limitations of all partitioning approaches. CONCLUSION: The decision on which concept to use when partitioning attributable risks on the population level should be based on the desired properties the solution ought to have. Arguments from game-theoretical reasoning can help to guide further research in this area, especially in exploring the methods using proportional division rules that are not yet fully understood.