Modeling time-varying recruitment rates in multicenter clinical trials.

Multicenter phase II/III clinical trials are large-scale operations that often include hundreds of recruiting centers in several countries. Therefore, the operational aspects of a trial must be thoroughly planned and closely monitored to ensure better oversight and study conduct. Predicting patient recruitment plays a pivotal role in trial monitoring as it informs how many people are expected to be recruited on a given day. As such, study teams may rely on predictions to assess progress and detect any deviations from the original plan that might put the study and potentially, patients at risk. Recruitment predictions are often based on a Poisson-Gamma model that assumes that centers have a constant recruitment rate throughout the trial. The model has useful features and has extensively been used, yet its main assumption is often unrealistic. A nonhomogeneous Poisson process has been recently proposed that can incorporate time-varying recruitment rates. In this work, we predict patient recruitment using both approaches and assess the impact of said assumption in a real-world setting where studies may not necessarily have constant center-specific recruitment rates. The paper showcases experience from modeling recruitment in trials sponsored by AstraZeneca between 2005 and 2018. In these data, there is evidence of time-varying recruitment rates. The predictive performance of models that account for both constant and time-varying recruitment effects is presented. Following a descriptive analysis of data, we assess model performance and investigate the impact of model misspecification.

A review of spline function procedures in R.

BACKGROUND: With progress on both the theoretical and the computational fronts the use of spline modelling has become an established tool in statistical regression analysis. An important issue in spline modelling is the availability of user friendly, well documented software packages. Following the idea of the STRengthening Analytical Thinking for Observational Studies initiative to provide users with guidance documents on the application of statistical methods in observational research, the aim of this article is to provide an overview of the most widely used spline-based techniques and their implementation in R. METHODS: In this work, we focus on the R Language for Statistical Computing which has become a hugely popular statistics software. We identified a set of packages that include functions for spline modelling within a regression framework. Using simulated and real data we provide an introduction to spline modelling and an overview of the most popular spline functions. RESULTS: We present a series of simple scenarios of univariate data, where different basis functions are used to identify the correct functional form of an independent variable. Even in simple data, using routines from different packages would lead to different results. CONCLUSIONS: This work illustrate challenges that an analyst faces when working with data. Most differences can be attributed to the choice of hyper-parameters rather than the basis used. In fact an experienced user will know how to obtain a reasonable outcome, regardless of the type of spline used. However, many analysts do not have sufficient knowledge to use these powerful tools adequately and will need more guidance.

A special case of reduced rank models for identification and modelling of time varying effects in survival analysis.

Flexible survival models are in need when modelling data from long term follow-up studies. In many cases, the assumption of proportionality imposed by a Cox model will not be valid. Instead, a model that can identify time varying effects of fixed covariates can be used. Although there are several approaches that deal with this problem, it is not always straightforward how to choose which covariates should be modelled having time varying effects and which not. At the same time, it is up to the researcher to define appropriate time functions that describe the dynamic pattern of the effects. In this work, we suggest a model that can deal with both fixed and time varying effects and uses simple hypotheses tests to distinguish which covariates do have dynamic effects. The model is an extension of the parsimonious reduced rank model of rank 1. As such, the number of parameters is kept low, and thus, a flexible set of time functions, such as b-splines, can be used. The basic theory is illustrated along with an efficient fitting algorithm. The proposed method is applied to a dataset of breast cancer patients and compared with a multivariate fractional polynomials approach for modelling time-varying effects. Copyright © 2016 John Wiley & Sons, Ltd.

A feature selection method for classification within functional genomics experiments based on the proportional overlapping score.

BACKGROUND: Microarray technology, as well as other functional genomics experiments, allow simultaneous measurements of thousands of genes within each sample. Both the prediction accuracy and interpretability of a classifier could be enhanced by performing the classification based only on selected discriminative genes. We propose a statistical method for selecting genes based on overlapping analysis of expression data across classes. This method results in a novel measure, called proportional overlapping score (POS), of a feature's relevance to a classification task. RESULTS: We apply POS, along-with four widely used gene selection methods, to several benchmark gene expression datasets. The experimental results of classification error rates computed using the Random Forest, k Nearest Neighbor and Support Vector Machine classifiers show that POS achieves a better performance. CONCLUSIONS: A novel gene selection method, POS, is proposed. POS analyzes the expressions overlap across classes taking into account the proportions of overlapping samples. It robustly defines a mask for each gene that allows it to minimize the effect of expression outliers. The constructed masks along-with a novel gene score are exploited to produce the selected subset of genes.

Cox models with dynamic ridge penalties on time-varying effects of the covariates.

Analysis of long-term follow-up survival studies require more sophisticated approaches than the proportional hazards model. To account for the dynamic behaviour of fixed covariates, penalized Cox models can be employed in models with interactions of the covariates and known time functions. In this work, I discuss some of the suggested methods and emphasize on the use of a ridge penalty in survival models. I review different strategies for choosing an optimal penalty weight and argue for the use of the computationally efficient restricted maximum likelihood (REML)-type method. A ridge penalty term can be subtracted from the likelihood when modelling time-varying effects in order to control the behaviour of the time functions. I suggest using flexible time functions such as B-splines and constrain the behaviour of these by adding proper penalties. I present the basic methods and illustrate different penalty weights in two different datasets.

Skin colonisation at the catheter exit site is strongly associated with catheter colonisation and catheter-related sepsis.

AIM: The commonest mode of catheter colonisation is via the extraluminal route with skin bacteria. Catheter-related sepsis causes significant mortality and morbidity in neonates. Our aim was to study the relationships between culture-positive catheter exit site skin swabs, percutaneous central venous catheter segments and blood to determine the magnitude of associations between exit site skin colonisation, catheter colonisation and catheter-related sepsis. METHODS: In a prospective study, an exit site skin swab and three formerly in vivo catheter segments (proximal, middle and tip) were taken for culture at catheter removal. In those neonates who were clinically unwell at catheter removal, a peripheral blood culture was also collected. Univariate and multivariate analyses were used to study associations. RESULTS: Skin swabs were culture positive in 39 (21%) of 187 catheter removals. With a culture-positive skin swab, the risk of associated catheter colonisation was nearly eight times higher (OR: 7.84, 95% CI: 3.59-17.15) and the risk of definite catheter-related sepsis with the same organism was nearly 10 times higher (OR 9.86, 95% CI: 3.13-31.00). CONCLUSION: Culture-positive skin swabs from the catheter exit site were strongly associated with catheter colonisation and with definite catheter-related sepsis with the same organism. These data provide further evidence supporting catheter colonisation via the extraluminal route and highlight the importance of optimising skin disinfection before catheter insertion.

Reduced rank hazard regression with fixed and time-varying effects of the covariates.

Modelling survival data from long-term follow-up studies presents challenges. The commonly used proportional hazards model should be extended to account for dynamic behaviour of the effects of fixed covariates. This work illustrates the use of reduced rank models in survival data, where some of the covariate effects are allowed to behave dynamically in time and some as fixed. Time-varying effects of the covariates can be fitted by using interactions of the fixed covariates with flexible transformations of time based on b-splines. To avoid overfitting, a reduced rank model will restrict the number of parameters, resulting in a more sensible fit to the data. This work presents the basic theory and the algorithm to fit such models. An application to breast cancer data is used for illustration of the suggested methods.

Comparison of Olympic-Style Weightlifting Performances of Elite Athletes: Scaling Models Account for Body Mass.

PURPOSE: We developed a scale for comparison of performances by weightlifters of different body mass and compare this scaling formula to current systems. METHODS: Data from Olympics and World and Continental Championships from 2017 to 2021 were obtained; results from athletes with doping violations were excluded, resulting in performances from 1900 athletes from 150 countries for use in analysis. Functional relationships between performance and body mass were explored by testing various transformations of body mass in the form of fractional polynomials that include a wide range of nonlinear relationships. These transformations were evaluated in quantile regression models to determine the best fit, examine sex differences, and distinguish fits for different performance levels (90th, 75th, and 50th percentiles). RESULTS: The resulting model used a transformation of body mass with powers -2 and 2 for males and females and was used to specify a scaling formula. The small percentage deviations between modeled and actual performances confirm the high accuracy of the model. In the subset of medalists, scaled performances were comparable across different body masses, whereas both Sinclair and Robi scalings, currently used in competitions, were more variable. The curves had similar shapes for the 90th and 75th percentile levels but were less steep for the 50th percentile. CONCLUSIONS: The scaling formula we derived to compare weightlifting performances across a range of body mass can easily be implemented in the competition software to determine the overall best lifters. This is an improvement over current methods that do not accurately account for differences in body mass and result in bias or yield large variations even with small differences in body mass despite identical performances.

Atrial fibrillation in embolic stroke of undetermined source: role of advanced imaging of left atrial function.

AIMS: Atrial fibrillation (AF) is detected in over 30% of patients following an embolic stroke of undetermined source (ESUS) when monitored with an implantable loop recorder (ILR). Identifying AF in ESUS survivors has significant therapeutic implications, and AF risk is essential to guide screening with long-term monitoring. The present study aimed to establish the role of left atrial (LA) function in subsequent AF identification and develop a risk model for AF in ESUS. METHODS AND RESULTS: We conducted a single-centre retrospective case-control study including all patients with ESUS referred to our institution for ILR implantation from December 2009 to September 2019. We recorded clinical variables at baseline and analysed transthoracic echocardiograms in sinus rhythm. Univariate and multivariable analyses were performed to inform variables associated with AF. Lasso regression analysis was used to develop a risk prediction model for AF. The risk model was internally validated using bootstrapping. Three hundred and twenty-three patients with ESUS underwent ILR implantation. In the ESUS population, 293 had a stroke, whereas 30 had suffered a transient ischaemic attack as adjudicated by a senior stroke physician. Atrial fibrillation of any duration was detected in 47.1%. The mean follow-up was 710 days. Following lasso regression with backwards elimination, we combined increasing lateral PA (the time interval from the beginning of the P wave on the surface electrocardiogram to the beginning of the A' wave on pulsed wave tissue Doppler of the lateral mitral annulus) [odds ratio (OR) 1.011], increasing Age (OR 1.035), higher Diastolic blood pressure (OR 1.027), and abnormal LA reservoir Strain (OR 0.973) into a new PADS score. The probability of identifying AF can be estimated using the formula. Model discrimination was good [area under the curve (AUC) 0.72]. The PADS score was internally validated using bootstrapping with 1000 samples of 150 patients showing consistent results with an AUC of 0.73. CONCLUSION: The novel PADS score can identify the risk of AF on prolonged monitoring with ILR following ESUS and should be considered a dedicated risk stratification tool for decision-making regarding the screening strategy for AF in stroke.

One-third of patients with a type of stroke called embolic stroke of undetermined source (ESUS) also have a heart condition called atrial fibrillation (AF), which increases their risk of having another stroke. However, we do not know why some patients with ESUS develop AF. To figure this out, we studied 323 patients with ESUS and used a special device to monitor their heart rhythm continuously for up to 3 years, an implantable loop recorder. We also looked at their medical history, performed a heart ultrasound, and identified some factors that increase the risk of identifying AF in the future. Factors associated with future AF include older age, higher diastolic blood pressure, and problems with the co-ordination and function of the upper left chamber of the heart called the left atrium.Based on these factors, we created a new scoring system that can identify patients who are at higher risk of developing AF better than the current scoring systems, the PADS score. This can potentially help doctors provide more targeted and effective treatment to these patients, ultimately aiming to reduce their risk of having another stroke.

Assessment of Paclitaxel Drug-Coated Balloon Only Angioplasty in STEMI.

BACKGROUND: Primary percutaneous coronary intervention (pPCI) with drug-eluting stents (DES) has emerged as the standard of care, but stent-related events have persisted. Drug-coated balloon (DCB)-only angioplasty is an emerging technology, although it is not fully evaluated compared with DES in the context of pPCI. OBJECTIVES: The aim of this study was to investigate the safety of DCB-only angioplasty compared with second-generation DES in pPCI. METHODS: All-cause mortality and net adverse cardiac events (cardiovascular mortality, acute coronary syndrome, ischemic stroke or transient ischemic attack, major bleeding, and unplanned target lesion revascularization [TLR]) were compared among all patients treated with DCBs only or with second-generation DES only for first presentation of ST-segment elevation myocardial infarction (STEMI) due to de novo disease between January 1, 2016, and November 15, 2019. Patients treated with both DCBs and DES were excluded. Data were analyzed using Cox regression models, Kaplan-Meier estimator plots and propensity score matching. RESULTS: Among 1,139 patients with STEMI due to de novo disease, 452 were treated with DCBs and 687 with DES. After a median follow-up period of >3 years, all-cause mortality was 49 of 452 and 62 of 687 in the DCB and DES groups, respectively (P = 0.18). On multivariable Cox regression analysis, there was no difference in mortality between DCBs and DES in the full and propensity score-matched cohorts. Age, frailty risk, history of heart failure, and family history of ischemic heart disease remained significant independent predictors of mortality. There was no difference in any of the secondary endpoints, including unplanned TLR. CONCLUSIONS: DCB-only angioplasty appears safe compared with DES for STEMI in terms of all-cause mortality and all net adverse cardiac events, including unplanned TLR. DCB may be an efficacious and safe alternative to DES in selected patient groups. (Drug Coated Balloon Only vs Drug Eluting Stent Angioplasty; NCT04482972).

Post-COVID-19 syndrome risk factors and further use of health services in East England.

Post-COVID syndrome, defined as symptoms persisting for more than twelve weeks after the diagnosis of COVID-19, has been recognised as a new clinical entity in the context of SARS-CoV-2 infection. This study was conducted to characterise the burden and predictors for post-COVID-19 syndrome in the local population. It was a community-based web-survey study conducted in Norfolk, East England, UK. We sent the survey to patients with confirmed COVID-19 infection by real-time polymerase chain reaction by December 6th, 2020. Questions related to the pre-COVID and post-COVID level of symptoms and further healthcare use. Baseline characteristics were collected from the primary care records. Logistic regression analysis was conducted to establish predictors for post-COVID-19 syndrome and further healthcare utilisation. Of 6,318 patients, survey responses were obtained from 1,487 participants (23.5%). Post-COVID-19 syndrome symptoms were experienced by 774 (52.1%) respondents. Male sex compared to female sex was a factor protective of post-COVID symptoms; relative risk (RR) 0.748, 95% confidence interval (CI), 0.605-0.924. Body mass index was associated with a greater risk of developing post-COVID-19 symptoms (RR 1.031, 95% CI, 1.016-1.047, for 1 kg/m2). A total of 378 (25.4%) people used further health services after their index COVID-19 infection, of whom 277 (73.2%) had post-COVID symptoms. Male sex was negatively associated with the use of further health services (RR 0.618, 95% CI, 0.464-0.818) whereas BMI was positively associated (RR 1.027, 95% CI, 1.009-1.046). Overall, post-COVID-19 symptoms increased the probability of using health services with RR 3.280, 95% CI, 2.540-4.262. This survey of a large number of people previously diagnosed with COVID-19 across East England shows a high prevalence of self-reported post-COVID-19 syndrome. Female sex and BMI were associated with an increased risk of post-COVID-19 syndrome and further utilisation of healthcare.

Long-term safety of paclitaxel drug-coated balloon-only angioplasty for de novo coronary artery disease: the SPARTAN DCB study.

OBJECTIVES: We aimed to investigate long-term survival of paclitaxel DCB for percutaneous coronary intervention (PCI). BACKGROUND: Safety concerns have been raised over the use of paclitaxel devices for peripheral artery disease recently, following a meta-analysis suggesting increased late mortality. With regard to drug-coated balloon (DCB) angioplasty for coronary artery intervention however, there is limited data to date regarding possible late mortality relating to paclitaxel. METHODS: We compared all-cause mortality of patients treated with paclitaxel DCB to those with non-paclitaxel second-generation drug-eluting stents (DES) for stable, de novo coronary artery disease from 1st January 2011 till 31st December 2018. To have homogenous groups allowing data on safety to be interpreted accurately, we excluded patients with previous PCI and patients treated with a combination of both DCB and DES in subsequent PCIs. Data were analysed with Kaplan-Meier curves and Cox regression statistical models. RESULTS: We present 1517 patients; 429 treated with paclitaxel DCB and 1088 treated with DES. On univariate analysis, age, hypercholesterolaemia, hypertension, peripheral vascular disease, prior myocardial infarction, heart failure, smoking, atrial fibrillation, decreasing estimated glomerular filtration rate (eGFR) [and renal failure (eGFR < 45)] were associated with worse survival. DCB intervention showed a non-significant trend towards better prognosis compared to DES (p = 0.08). On multivariable analysis age, decreasing eGFR and smoking associated with worse prognosis. CONCLUSION: We found no evidence of late mortality associated with DCB angioplasty compared with non-paclitaxel second-generation DES in up to 5 years follow-up. DCB is a safe option for the treatment of de novo coronary artery disease.

Sex differences and impact of body mass on performance from childhood to senior athletes in Olympic weightlifting.

BACKGROUND: Olympic weightlifting requires technical skills, explosive power, strength, and coordination. Weightlifters can be competitive within a range of morphological characteristics due to competition body weight classes. To date no studies have examined when sex differences arise in weightlifting and the impact of body mass on performances at different ages. OBJECTIVES: To examine when sex-related differences emerge, to quantify the influence of body mass on performances at different ages, and to estimate the age at peak performance. METHODS: Competitions results from USA Weightlifting National Championships, Youth, Junior, and Senior from 2014 to 2019 were collected for weightlifters aged 6 to 30. RESULTS: At age 10 the median total weight lifted was 51kg and 54kg, respectively, for girls and boys. From age 10 to 12 a gender gap emerges with a sex difference of 11.7% at age 14 at 55kg body mass. At age 25 the sex-related performance difference is smaller for lighter athletes (23.6% at 69kg body mass) and larger for heavier athletes (29.9% at 81kg body mass). The median peak age for men is 26.5 years (95% CI: 25.7, 27.3) and for women 25.9 years (95% CI: 24.7, 27.3). CONCLUSION: We quantified the impact of body weight and age and sex differences for youth and young adults, ages 6 to 30 years old, participating in national level Olympic weightlifting competitions in the United States. Body weight at younger ages has less impact on performance compared to older ages, and boys and girls perform similarly. When reaching the ages typically associated with the onset of puberty, boys' performances rapidly increase and the gap between genders widens. Women achieve peak performance at a similar age than men. Such results may help to establish progression trajectories for talented athletes and inform coaches, athletes and national governing bodies.

State of the art in selection of variables and functional forms in multivariable analysis-outstanding issues.

BACKGROUND: How to select variables and identify functional forms for continuous variables is a key concern when creating a multivariable model. Ad hoc 'traditional' approaches to variable selection have been in use for at least 50 years. Similarly, methods for determining functional forms for continuous variables were first suggested many years ago. More recently, many alternative approaches to address these two challenges have been proposed, but knowledge of their properties and meaningful comparisons between them are scarce. To define a state of the art and to provide evidence-supported guidance to researchers who have only a basic level of statistical knowledge, many outstanding issues in multivariable modelling remain. Our main aims are to identify and illustrate such gaps in the literature and present them at a moderate technical level to the wide community of practitioners, researchers and students of statistics. METHODS: We briefly discuss general issues in building descriptive regression models, strategies for variable selection, different ways of choosing functional forms for continuous variables and methods for combining the selection of variables and functions. We discuss two examples, taken from the medical literature, to illustrate problems in the practice of modelling. RESULTS: Our overview revealed that there is not yet enough evidence on which to base recommendations for the selection of variables and functional forms in multivariable analysis. Such evidence may come from comparisons between alternative methods. In particular, we highlight seven important topics that require further investigation and make suggestions for the direction of further research. CONCLUSIONS: Selection of variables and of functional forms are important topics in multivariable analysis. To define a state of the art and to provide evidence-supported guidance to researchers who have only a basic level of statistical knowledge, further comparative research is required.

Usefulness of the NULL-PLEASE Score to Predict Survival in Out-of-Hospital Cardiac Arrest.

PURPOSE: Out-of-hospital cardiac arrest (OHCA) carries a very high mortality rate even after successful cardiopulmonary resuscitation. Currently, information given to relatives about prognosis following resuscitation is often emotive and subjective, and varies with clinician experience. We aimed to validate the NULL-PLEASE score to predict survival following OHCA. METHODS: A multicenter cohort study was conducted, with retrospective and prospective validation in consecutive unselected patients presenting with OHCA. The NULL-PLEASE score was calculated by attributing points to the following variables: Nonshockable initial rhythm, Unwitnessed arrest, Long low-flow period, Long no-flow period, pH <7.2, Lactate >7.0 mmol/L, End-stage renal failure, Age ≥85 years, Still resuscitation, and Extracardiac cause. The primary outcome was in-hospital death. RESULTS: We assessed 700 patients admitted with OHCA, of whom 47% survived to discharge. In 300 patients we performed a retrospective validation, followed by prospective validation in 400 patients. The NULL-PLEASE score was lower in patients who survived compared with those who died (0 [interquartile range 0-1] vs 4 [interquartile range 2-4], P < .0005) and strongly predictive of in-hospital death (C-statistic 0.874; 95% confidence interval, 0.848-0.899). Patients with a score ≥3 had a 24-fold increased risk of death (odds ratio 23.6; 95% confidence interval, 14.840-37.5; P < .0005) compared with those with lower scores. A score ≥3 has a 91% positive predictive value for in-hospital death, while a score <3 predicts a 71% chance of survival. CONCLUSION: The easy-to-use NULL-PLEASE score predicts in-hospital mortality with high specificity and can help clinicians explain the prognosis to relatives in an easy-to-understand, objective fashion, to realistically prepare them for the future.

Performance Development From Youth to Senior and Age of Peak Performance in Olympic Weightlifting.

A total of 3,782 performance results for male and female weightlifters, ages 14-30 from 123 countries, from Youth, Junior, and Senior World Championships and Olympic Games 2013-2017 were used to estimate the age at peak performance in Olympic weightlifting and quantify performance development from adolescence to adulthood. The age at peak performance was estimated for men and women globally and for different geographic regions. Overall, male and female weightlifters achieve their peak performance in weightlifting at similar ages. The median peak age is 26.0 years (95% CI: 24.9, 27.1) for men and 25.0 years (95% CI: 23.9, 27.4) for women, at the 90th percentile of performances. The median peak age was 26.3 years for men (95% CI: 24.5, 29.6) and 26.4 years for women (95% CI: 24.5, 29.6), at the 50th percentile. It is a novel finding that the age at peak performance varies for male and female athletes from different geographic regions (Western Europe, Eastern Europe, Middle East, Far East, North- and South America). For some regions men reach peak performance at a younger age than women, while this relationship is reversed for other regions. A possible explanation could be that socio-economic factors influence the pool of available athletes and thus may under- or overestimate the true peak age. Unlike in track and field where the discipline might determine specific body types, weightlifters at all ages compete in body weight classes, enabling us to compare performance levels and annual rate of change for athletes of different body mass. We quantified increases in performance in Olympic weightlifting for male and female adolescents. Sex-specific differences arise during puberty, boys outperform girls, and there is a rapid increase in their performance levels before the further growth slows down. The largest annual rate of increase in the total weight lifted was achieved between 16 and 17 years of age for both sexes with lower body mass and between 21 and 22 years with higher body mass. Such new information may help to establish progression trajectories for young athletes.

Age-associated Performance Decline and Sex Differences in Olympic Weightlifting.

INTRODUCTION: The rate of decline in physical performance for women is thought to be faster than that for men at any age in the Masters age classes in weightlifting and other sports. We quantified the age-associated decline in Olympic weightlifting separately for women and men and investigated possible impact of perimenopausal years on the performance decline. METHODS: Results from Masters Weightlifting competitions from 1993 to 2018 were compiled from original score sheets and meet results made available by International Weightlifting Federation. Quantile curves were estimated for the age-related performance decline, and confidence intervals (CI) for the fractional performance with reference age 35 yr were calculated. Age-related decline curves were estimated for different periods to examine changes in performance levels. RESULTS: A total of 10,225 performance results for male and female weightlifters age 35 to 90 yr from 71 countries were included in the analysis. At age 40 yr compared with age 35 yr, the fractional performance is 0.947 (95% CI, 0.926-0.975), for men and 0.952 (95% CI. 0.898-0.986) for women while this is reduced to 0.723 (95% CI: 0.651-0.800) at age 60 yr for men and 0.604 (95% CI. 0.543-0.706) for women. Female performance levels before 2000 were worse; however, they have stabilized since 2013. CONCLUSIONS: The performances of women weightlifters have improved over the last 25 yr. Thus, previous publications do not reflect current physical capabilities of women. The age-associated performance decline for female weightlifters mirrors the decline for men except for an accelerated decline during a 10-yr period across the age range from late 40s to late 50s thus coinciding with a transition into menopause.

Ensemble of a subset of kNN classifiers.

Combining multiple classifiers, known as ensemble methods, can give substantial improvement in prediction performance of learning algorithms especially in the presence of non-informative features in the data sets. We propose an ensemble of subset of kNN classifiers, ESkNN, for classification task in two steps. Firstly, we choose classifiers based upon their individual performance using the out-of-sample accuracy. The selected classifiers are then combined sequentially starting from the best model and assessed for collective performance on a validation data set. We use bench mark data sets with their original and some added non-informative features for the evaluation of our method. The results are compared with usual kNN, bagged kNN, random kNN, multiple feature subset method, random forest and support vector machines. Our experimental comparisons on benchmark classification problems and simulated data sets reveal that the proposed ensemble gives better classification performance than the usual kNN and its ensembles, and performs comparable to random forest and support vector machines.

A fast routine for fitting Cox models with time varying effects of the covariates.

The S-plus and R statistical packages have implemented a counting process setup to estimate Cox models with time varying effects of the covariates. The data set has to be re-arranged in a repeated measurement setting: the time is divided into small time intervals where a single event occurs and for each time interval, the covariate values and outcome in the interval for each subject still under observation are stacked to a large data set. This is the known (Tstart,Tstop] algorithm implemented in Therneau's Survival library (S-plus), which has been ported into an R package by Thomas Lumley. However, the expansion of a data set leads to a larger set, which can be hard to handle even with fast modern computers. We propose the use of a fast and efficient algorithm, written in R, which works on the original data without the use of an expansion. The computations are done on the original data set, with significant less memory resources used. This improves the computational time by orders of magnitude. The algorithm can also fit reduced rank Cox models with time varying effects. We illustrate the method on a large data set of 2433 breast cancer patients, a smaller study of 358 ovarian cancer patients, and compare the computational times on simulated data of up to 10,000 cases with SAS proc phreg and survival package in R. For larger data sets our algorithm was several times faster, and was able to handle larger data sets then SAS and R.

Relationship of cell proliferation (Ki-67) to 99mTc-(V)DMSA uptake in breast cancer.

INTRODUCTION: The aim of the present study was to identify the relationships between the uptake of radiotracers - namely pentavalent dimercaptosuccinic acid [(V)DMSA] and sestamibi (MIBI) - and the following parameters in primary breast cancer: steroid receptor concentrations (i.e. estrogen receptor [ER] and progesterone receptor [PR]), Ki-67 expression, tumor size, tumor grade, age, and levels of expression of p53 and c-erbB-2. In addition, by multivariate regression analysis, we further isolated those factors with independent associations with (V)DMSA and/or MIBI uptake in primary breast cancer. METHODS: Thirty-four patients with histologically confirmed breast carcinoma underwent preoperative scintimammography with technetium-99m (99mTc)-(V)DMSA and/or 99mTc-MIBI in consecutive sessions 10 and 60 min after administration of 925-1110 MBq of each radiotracer. The tumor-to-background ratio was calculated and correlated with the presence of ER, PR, Ki-67, tumor size, tumor grade, p53, and c-erbB-2. ER, PR, p53, and c-erbB-2 were determined immunohistochemically. The analysis included tumor-to-background ratio of (V)DMSA and MIBI uptake as dependent and all of the other parameters as independent variables. RESULTS: Correlation was positive between Ki-67 and (V)DMSA (r = 0.37 at 10 min, P = 0.038; r = 0.42 at 60 min, P = 0.018) and inverse between PR and (V)DMSA uptake (r = -0.46 at 10 min, P = 0.010; r = -0.51 at 60 min, P = 0.003). Multivariate regression analysis demonstrated a positive correlation between Ki-67 and (V)DMSA at 60 min (P = 0.045). Ki-67 was not significantly correlated with MIBI uptake, whereas tumor size was positively correlated with MIBI uptake at 60 min both in univariate (r = 0.45, P = 0.027) and multivariate analysis (P = 0.024). Negative correlations were observed between (V)DMSA uptake and ER, as well as between ER/PR and MIBI uptake, but these were not significant. CONCLUSION: Ki-67 appears to represent the major independent factor affecting (V)DMSA uptake in breast cancer. Tumor size was the only independent parameter influencing MIBI uptake in breast cancer. (V)DMSA appears to have an advantage over MIBI in that it can be used to visualize tumors with intense proliferative activity, and thus it can identify those tumors that are more aggressive.