The use of classification and regression trees to predict the likelihood of seasonal influenza.

Background Individual signs and symptoms are of limited value for the diagnosis of influenza. Objective To develop a decision tree for the diagnosis of influenza based on a classification and regression tree (CART) analysis. Methods Data from two previous similar cohort studies were assembled into a single dataset. The data were randomly divided into a development set (70%) and a validation set (30%). We used CART analysis to develop three models that maximize the number of patients who do not require diagnostic testing prior to treatment decisions. The validation set was used to evaluate overfitting of the model to the training set. Results Model 1 has seven terminal nodes based on temperature, the onset of symptoms and the presence of chills, cough and myalgia. Model 2 was a simpler tree with only two splits based on temperature and the presence of chills. Model 3 was developed with temperature as a dichotomous variable (≥38°C) and had only two splits based on the presence of fever and myalgia. The area under the receiver operating characteristic curves (AUROCC) for the development and validation sets, respectively, were 0.82 and 0.80 for Model 1, 0.75 and 0.76 for Model 2 and 0.76 and 0.77 for Model 3. Model 2 classified 67% of patients in the validation group into a high- or low-risk group compared with only 38% for Model 1 and 54% for Model 3. Conclusions A simple decision tree (Model 2) classified two-thirds of patients as low or high risk and had an AUROCC of 0.76. After further validation in an independent population, this CART model could support clinical decision making regarding influenza, with low-risk patients requiring no further evaluation for influenza and high-risk patients being candidates for empiric symptomatic or drug therapy.

Radiotherapy and temozolomide for newly diagnosed glioblastoma: recursive partitioning analysis of the EORTC 26981/22981-NCIC CE3 phase III randomized trial.

PURPOSE: The European Organisation for Research and Treatment of Cancer and National Cancer Institute of Canada trial on temozolomide (TMZ) and radiotherapy (RT) in glioblastoma (GBM) has demonstrated that the combination of TMZ and RT conferred a significant and meaningful survival advantage compared with RT alone. We evaluated in this trial whether the recursive partitioning analysis (RPA) retains its overall prognostic value and what the benefit of the combined modality is in each RPA class. PATIENTS AND METHODS: Five hundred seventy-three patients with newly diagnosed GBM were randomly assigned to standard postoperative RT or to the same RT with concomitant TMZ followed by adjuvant TMZ. The primary end point was overall survival. The European Organisation for Research and Treatment of Cancer RPA used accounts for age, WHO performance status, extent of surgery, and the Mini-Mental Status Examination. RESULTS: Overall survival was statistically different among RPA classes III, IV, and V, with median survival times of 17, 15, and 10 months, respectively, and 2-year survival rates of 32%, 19%, and 11%, respectively (P < .0001). Survival with combined TMZ/RT was higher in RPA class III, with 21 months median survival time and a 43% 2-year survival rate, versus 15 months and 20% for RT alone (P = .006). In RPA class IV, the survival advantage remained significant, with median survival times of 16 v 13 months, respectively, and 2-year survival rates of 28% v 11%, respectively (P = .0001). In RPA class V, however, the survival advantage of RT/TMZ was of borderline significance (P = .054). CONCLUSION: RPA retains its prognostic significance overall as well as in patients receiving RT with or without TMZ for newly diagnosed GBM, particularly in classes III and IV.

Improved predictive mapping of indoor radon concentrations using ensemble regression trees based on automatic clustering of geological units.

PURPOSE: According to estimations around 230 people die as a result of radon exposure in Switzerland. This public health concern makes reliable indoor radon prediction and mapping methods necessary in order to improve risk communication to the public. The aim of this study was to develop an automated method to classify lithological units according to their radon characteristics and to develop mapping and predictive tools in order to improve local radon prediction. METHOD: About 240 000 indoor radon concentration (IRC) measurements in about 150 000 buildings were available for our analysis. The automated classification of lithological units was based on k-medoids clustering via pair-wise Kolmogorov distances between IRC distributions of lithological units. For IRC mapping and prediction we used random forests and Bayesian additive regression trees (BART). RESULTS: The automated classification groups lithological units well in terms of their IRC characteristics. Especially the IRC differences in metamorphic rocks like gneiss are well revealed by this method. The maps produced by random forests soundly represent the regional difference of IRCs in Switzerland and improve the spatial detail compared to existing approaches. We could explain 33% of the variations in IRC data with random forests. Additionally, the influence of a variable evaluated by random forests shows that building characteristics are less important predictors for IRCs than spatial/geological influences. BART could explain 29% of IRC variability and produced maps that indicate the prediction uncertainty. CONCLUSION: Ensemble regression trees are a powerful tool to model and understand the multidimensional influences on IRCs. Automatic clustering of lithological units complements this method by facilitating the interpretation of radon properties of rock types. This study provides an important element for radon risk communication. Future approaches should consider taking into account further variables like soil gas radon measurements as well as more detailed geological information.

The serine repeat antigen (SERA) gene family phylogeny in Plasmodium: the impact of GC content and reconciliation of gene and species trees.

Plasmodium falciparum is the parasite responsible for the most acute form of malaria in humans. Recently, the serine repeat antigen (SERA) in P. falciparum has attracted attention as a potential vaccine and drug target, and it has been shown to be a member of a large gene family. To clarify the relationships among the numerous P. falciparum SERAs and to identify orthologs to SERA5 and SERA6 in Plasmodium species affecting rodents, gene trees were inferred from nucleotide and amino acid sequence data for 33 putative SERA homologs in seven different species. (A distance method for nucleotide sequences that is specifically designed to accommodate differing GC content yielded results that were largely compatible with the amino acid tree. Standard-distance and maximum-likelihood methods for nucleotide sequences, on the other hand, yielded gene trees that differed in important respects.) To infer the pattern of duplication, speciation, and gene loss events in the SERA gene family history, the resulting gene trees were then "reconciled" with two competing Plasmodium species tree topologies that have been identified by previous phylogenetic studies. Parsimony of reconciliation was used as a criterion for selecting a gene tree/species tree pair and provided (1) support for one of the two species trees and for the core topology of the amino acid-derived gene tree, (2) a basis for critiquing fine detail in a poorly resolved region of the gene tree, (3) a set of predicted "missing genes" in some species, (4) clarification of the relationship among the P. falciparum SERA, and (5) some information about SERA5 and SERA6 orthologs in the rodent malaria parasites. Parsimony of reconciliation and a second criterion--implied mutational pattern at two key active sites in the SERA proteins-were also seen to be useful supplements to standard "bootstrap" analysis for inferred topologies.

In vivo targeting of an anti-tumor antibody coupled to antigenic MHC class I complexes induces specific growth inhibition and regression of established syngeneic tumor grafts.

The concept of antibody-mediated targeting of antigenic MHC/peptide complexes on tumor cells in order to sensitize them to T-lymphocyte cytotoxicity represents an attractive new immunotherapy strategy. In vitro experiments have shown that an antibody chemically conjugated or fused to monomeric MHC/peptide can be oligomerized on the surface of tumor cells, rendering them susceptible to efficient lysis by MHC-peptide restricted specific T-cell clones. However, this strategy has not yet been tested entirely in vivo in immunocompetent animals. To this aim, we took advantage of OT-1 mice which have a transgenic T-cell receptor specific for the ovalbumin (ova) immunodominant peptide (257-264) expressed in the context of the MHC class I H-2K(b). We prepared and characterized conjugates between the Fab' fragment from a high-affinity monoclonal antibody to carcinoembryonic antigen (CEA) and the H-2K(b) /ova peptide complex. First, we showed in OT-1 mice that the grafting and growth of a syngeneic colon carcinoma line transfected with CEA could be specifically inhibited by systemic injections of the conjugate. Next, using CEA transgenic C57BL/6 mice adoptively transferred with OT-1 spleen cells and immunized with ovalbumin, we demonstrated that systemic injections of the anti-CEA-H-2K(b) /ova conjugate could induce specific growth inhibition and regression of well-established, palpable subcutaneous grafts from the syngeneic CEA-transfected colon carcinoma line. These results, obtained in a well-characterized syngeneic carcinoma model, demonstrate that the antibody-MHC/peptide strategy can function in vivo. Further preclinical experimental studies, using an anti-viral T-cell response, will be performed before this new form of immunotherapy can be considered for clinical use.

An experimental study of discharge partitioning and flow structure at symmetrical bifurcations

Recent research has examined the factors controlling the geometrical configuration of bifurcations, determined the range of stability conditions for a number of bifurcation types and assessed the impact of perturbations on bifurcation evolution. However, the flow division process and the parameters that influence flow and sediment partitioning are still poorly characterized. To identify and isolate these parameters, three-dimensional velocities were measured at 11 cross-sections in a fixed-walled experimental bifurcation. Water surface gradients were controlled, and systematically varied, using a weir in each distributary. As may be expected, the steepest distributary conveyed the most discharge ( was dominant) while the mildest distributary conveyed the least discharge ( was subordinate). A zone of water surface super-elevation was co-located with the bifurcation in symmetric cases or displaced into the subordinate branch in asymmetric cases. Downstream of a relatively acute-angled bifurcation, primary velocity cores were near to the water surface and against the inner banks, with near-bed zones of lower primary velocity at the outer banks. Downstream of an obtuse-angled bifurcation, velocity cores were initially at the outer banks, with near-bed zones of lower velocities at the inner banks, but patterns soon reverted to match the acute-angled case. A single secondary flow cell was generated in each distributary, with water flowing inwards at the water surface and outwards at the bed. Circulation was relatively enhanced within the subordinate branch, which may help explain why subordinate distributaries remain open, may play a role in determining the size of commonly-observed topographic features, and may thus exert some control on the stability of asymmetric bifurcations. Further, because larger values of circulation result from larger gradient disadvantages, the length of confluence-diffluence units in braided rivers or between diffluences within delta distributary networks may vary depending upon flow structures inherited from upstream and whether, and how, they are fed by dominant or subordinate distributaries. Copyright (C) 2011 John Wiley & Sons, Ltd.

Random measurement error and regression dilution bias.

Summary points: - The bias introduced by random measurement error will be different depending on whether the error is in an exposure variable (risk factor) or outcome variable (disease) - Random measurement error in an exposure variable will bias the estimates of regression slope coefficients towards the null - Random measurement error in an outcome variable will instead increase the standard error of the estimates and widen the corresponding confidence intervals, making results less likely to be statistically significant - Increasing sample size will help minimise the impact of measurement error in an outcome variable but will only make estimates more precisely wrong when the error is in an exposure variable

Trophic resource partitioning and competition between the two sibling bat species Myotis myotis and Myotis blythii

1. Niche theory predicts that the stable coexistence of species within a guild should be associated, if resources are limited, with a mechanism of resource partitioning. Using extensive data on diets, the present study attempts: (i) to test the hypothesis that, in sympatry, the interspecific overlap between the trophic niches of the sibling bat species Myotis myotis and M. blythii-which coexist intimately in their roosts-is effectively lower than the two intraspecific overlaps; (ii) to assess the role played by interspecific competition in resource partitioning through the study of trophic niche displacement between several sympatric and allopatric populations. 2. Diets were determined by the analysis of faecal samples collected in the field from individual bats captured in various geographical areas. Trophic niche overlaps were calculated monthly for all possible intraspecific and interspecific pairs of individuals from sympatric populations. Niche breadth was estimated from: (i) every faecal sample; (ii) all the faecal samples collected per month in a given population (geographical area). 3. In every population, the bulk of the diets of M. myotis and M. blythii consisted of, respectively, terrestrial (e.g. carabid beetles) and grass-dwelling (mostly bush crickets) prey. All intraspecific trophic niche overlaps were significantly greater than the interspecific one, except in Switzerland in May when both species exploited mass concentrations of cockchafers, a non-limiting food source. This clearcut partitioning of resources may allow the stable, intimate coexistence observed under sympatric conditions. 4. Relative proportions of ground-and grass-dwelling prey, as well as niche breadths (either individual or population), did not differ significantly between sympatry and allopatry, showing that, under allopatric conditions, niche expansion does not take place. This suggests that active interspecific competition is not the underlying mechanism responsible for the niche partitioning which is currently observed between M. myotis and M. blythii.

Systematic analysis of factors associated with progression and regression of ulcerative colitis in 918 patients.

BACKGROUND: Studies that systematically assess change in ulcerative colitis (UC) extent over time in adult patients are scarce. AIM: To assess changes in disease extent over time and to evaluate clinical parameters associated with this change. METHODS: Data from the Swiss IBD cohort study were analysed. We used logistic regression modelling to identify factors associated with a change in disease extent. RESULTS: A total of 918 UC patients (45.3% females) were included. At diagnosis, UC patients presented with the following disease extent: proctitis [199 patients (21.7%)], left-sided colitis [338 patients (36.8%)] and extensive colitis/pancolitis [381 (41.5%)]. During a median disease duration of 9 [4-16] years, progression and regression was documented in 145 patients (15.8%) and 149 patients (16.2%) respectively. In addition, 624 patients (68.0%) had a stable disease extent. The following factors were identified to be associated with disease progression: treatment with systemic glucocorticoids [odds ratio (OR) 1.704, P = 0.025] and calcineurin inhibitors (OR: 2.716, P = 0.005). No specific factors were found to be associated with disease regression. CONCLUSIONS: Over a median disease duration of 9 [4-16] years, about two-thirds of UC patients maintained the initial disease extent; the remaining one-third had experienced either progression or regression of the disease extent.

Integrated Analysis of Molecular and Clinical Prognostic Factors in Stage II/III Colon Cancer.

Background The prognostic potential of individual clinical and molecular parameters in stage II/III colon cancer has been investigated, but a thorough multivariable assessment of their relative impact is missing. Methods Tumors from patients (N = 1404) in the PETACC3 adjuvant chemotherapy trial were examined for BRAF and KRAS mutations, microsatellite instability (MSI), chromosome 18q loss of heterozygosity (18qLOH), and SMAD4 expression. Their importance in predicting relapse-free survival (RFS) and overall survival (OS) was assessed by Kaplan-Meier analyses, Cox regression models, and recursive partitioning trees. All statistical tests were two-sided. Results MSI-high status and SMAD4 focal loss of expression were identified as independent prognostic factors with better RFS (hazard ratio [HR] of recurrence = 0.54, 95% CI = 0.37 to 0.81, P = .003) and OS (HR of death = 0.43, 95% CI = 0.27 to 0.70, P = .001) for MSI-high status and worse RFS (HR = 1.47, 95% CI = 1.19 to 1.81, P < .001) and OS (HR = 1.58, 95% CI = 1.23 to 2.01, P < .001) for SMAD4 loss. 18qLOH did not have any prognostic value in RFS or OS. Recursive partitioning identified refinements of TNM into new clinically interesting prognostic subgroups. Notably, T3N1 tumors with MSI-high status and retained SMAD4 expression had outcomes similar to stage II disease. Conclusions Concomitant assessment of molecular and clinical markers in multivariable analysis is essential to confirm or refute their independent prognostic value. Including molecular markers with independent prognostic value might allow more accurate prediction of prognosis than TNM staging alone.

Binary classification of dyslipidemia from the waist-to-hip ratio and body mass index: a comparison of linear, logistic, and CART models.

BACKGROUND: We sought to improve upon previously published statistical modeling strategies for binary classification of dyslipidemia for general population screening purposes based on the waist-to-hip circumference ratio and body mass index anthropometric measurements. METHODS: Study subjects were participants in WHO-MONICA population-based surveys conducted in two Swiss regions. Outcome variables were based on the total serum cholesterol to high density lipoprotein cholesterol ratio. The other potential predictor variables were gender, age, current cigarette smoking, and hypertension. The models investigated were: (i) linear regression; (ii) logistic classification; (iii) regression trees; (iv) classification trees (iii and iv are collectively known as "CART"). Binary classification performance of the region-specific models was externally validated by classifying the subjects from the other region. RESULTS: Waist-to-hip circumference ratio and body mass index remained modest predictors of dyslipidemia. Correct classification rates for all models were 60-80%, with marked gender differences. Gender-specific models provided only small gains in classification. The external validations provided assurance about the stability of the models. CONCLUSIONS: There were no striking differences between either the algebraic (i, ii) vs. non-algebraic (iii, iv), or the regression (i, iii) vs. classification (ii, iv) modeling approaches. Anticipated advantages of the CART vs. simple additive linear and logistic models were less than expected in this particular application with a relatively small set of predictor variables. CART models may be more useful when considering main effects and interactions between larger sets of predictor variables.

Generalized linear and generalized additive models in studies of species distributions: setting the scene

An important statistical development of the last 30 years has been the advance in regression analysis provided by generalized linear models (GLMs) and generalized additive models (GAMs). Here we introduce a series of papers prepared within the framework of an international workshop entitled: Advances in GLMs/GAMs modeling: from species distribution to environmental management, held in Riederalp, Switzerland, 6-11 August 2001.We first discuss some general uses of statistical models in ecology, as well as provide a short review of several key examples of the use of GLMs and GAMs in ecological modeling efforts. We next present an overview of GLMs and GAMs, and discuss some of their related statistics used for predictor selection, model diagnostics, and evaluation. Included is a discussion of several new approaches applicable to GLMs and GAMs, such as ridge regression, an alternative to stepwise selection of predictors, and methods for the identification of interactions by a combined use of regression trees and several other approaches. We close with an overview of the papers and how we feel they advance our understanding of their application to ecological modeling.

Predictive analysis and mapping of indoor radon concentrations in Switzerland

It is estimated that around 230 people die each year due to radon (222Rn) exposure in Switzerland. 222Rn occurs mainly in closed environments like buildings and originates primarily from the subjacent ground. Therefore it depends strongly on geology and shows substantial regional variations. Correct identification of these regional variations would lead to substantial reduction of 222Rn exposure of the population based on appropriate construction of new and mitigation of already existing buildings. Prediction of indoor 222Rn concentrations (IRC) and identification of 222Rn prone areas is however difficult since IRC depend on a variety of different variables like building characteristics, meteorology, geology and anthropogenic factors. The present work aims at the development of predictive models and the understanding of IRC in Switzerland, taking into account a maximum of information in order to minimize the prediction uncertainty. The predictive maps will be used as a decision-support tool for 222Rn risk management. The construction of these models is based on different data-driven statistical methods, in combination with geographical information systems (GIS). In a first phase we performed univariate analysis of IRC for different variables, namely the detector type, building category, foundation, year of construction, the average outdoor temperature during measurement, altitude and lithology. All variables showed significant associations to IRC. Buildings constructed after 1900 showed significantly lower IRC compared to earlier constructions. We observed a further drop of IRC after 1970. In addition to that, we found an association of IRC with altitude. With regard to lithology, we observed the lowest IRC in sedimentary rocks (excluding carbonates) and sediments and the highest IRC in the Jura carbonates and igneous rock. The IRC data was systematically analyzed for potential bias due to spatially unbalanced sampling of measurements. In order to facilitate the modeling and the interpretation of the influence of geology on IRC, we developed an algorithm based on k-medoids clustering which permits to define coherent geological classes in terms of IRC. We performed a soil gas 222Rn concentration (SRC) measurement campaign in order to determine the predictive power of SRC with respect to IRC. We found that the use of SRC is limited for IRC prediction. The second part of the project was dedicated to predictive mapping of IRC using models which take into account the multidimensionality of the process of 222Rn entry into buildings. We used kernel regression and ensemble regression tree for this purpose. We could explain up to 33% of the variance of the log transformed IRC all over Switzerland. This is a good performance compared to former attempts of IRC modeling in Switzerland. As predictor variables we considered geographical coordinates, altitude, outdoor temperature, building type, foundation, year of construction and detector type. Ensemble regression trees like random forests allow to determine the role of each IRC predictor in a multidimensional setting. We found spatial information like geology, altitude and coordinates to have stronger influences on IRC than building related variables like foundation type, building type and year of construction. Based on kernel estimation we developed an approach to determine the local probability of IRC to exceed 300 Bq/m3. In addition to that we developed a confidence index in order to provide an estimate of uncertainty of the map. All methods allow an easy creation of tailor-made maps for different building characteristics. Our work is an essential step towards a 222Rn risk assessment which accounts at the same time for different architectural situations as well as geological and geographical conditions. For the communication of 222Rn hazard to the population we recommend to make use of the probability map based on kernel estimation. The communication of 222Rn hazard could for example be implemented via a web interface where the users specify the characteristics and coordinates of their home in order to obtain the probability to be above a given IRC with a corresponding index of confidence. Taking into account the health effects of 222Rn, our results have the potential to substantially improve the estimation of the effective dose from 222Rn delivered to the Swiss population.