Search

Background. N-of-1 designs gain popularity in nutritional research because of the improving technological possibilities, practical applicability and promise of increased accuracy and sensitivity, especially in the field of personalized nutrition. This move asks for a search of applicable statistical methods. Objective. To demonstrate the differences of three popular statistical methods in analyzing treatment effects of data obtained in N-of-1 designs. Method. We compare Individual-participant data meta-analysis, frequentist and Bayesian linear mixed effect models using a simulation experiment. Furthermore, we demonstrate the merits of the Bayesian model including prior information by analyzing data of an empirical study on weight loss. Results. The linear mixed effect models are to be preferred over the meta-analysis method, since the individual effects are estimated more accurately as evidenced by the lower errors, especially with lower sample sizes. Differences between Bayesian and frequentist mixed models were found to be small, indicating that they will lead to the same results without including an informative prior. Conclusion. For empirical data, the Bayesian mixed model allows the inclusion of prior knowledge and gives potential for population based and personalized inference.

Tiered or stepwise approaches to assess occupational exposure to nano-objects, and their agglomerates and aggregates have been proposed, which require decision rules (DRs) to move to a next tier, or terminate the assessment. In a desk study the performance of a number of DRs based on the evaluation of results from direct reading instruments was investigated by both statistical simulations and the application of the DRs to real workplace data sets. A statistical model that accounts for autocorrelation patterns in time-series, i.e. autoregressive integrated moving average (ARIMA), was used as 'gold' standard. The simulations showed that none of the proposed DRs covered the entire range of simulated scenarios with respect to the ARIMA model parameters, however, a combined DR showed a slightly better agreement. Application of the DRs to real workplace datasets (n = 117) revealed sensitivity up to 0.72, whereas the lowest observed specificity was 0.95. The selection of the most appropriate DR is very much dependent on the consequences of the decision, i.e. ruling in or ruling out of scenarios for further evaluation. Since a basic assessment may also comprise of other type of measurements and information, an evaluation logic was proposed which embeds the DRs, but furthermore supports decision making in view of a tiered-approach exposure assessment. © 2016 The Author.

Background and aims: The spread of smartphone applications and fitness monitors provides less expensive Methods for tracking time-location- activity data, which is a critical source of information for modelling personal exposure. The present study examines the potential use of smart consumer products data for predicting location status. Methods: A trial campaign of instrument reliability took place examining a series of monitors, such as the FitBit Flex and Moves app, for tracking people’s location and activities. Four participants in the city of Thessaloniki wore these devices along with a wearable temperature sensor, an Actigraph and a GPS sensor for a week. Since location data alone does not reliably determine whether a person is indoors outdoors or in transit, the predictive value of the aforementioned devices data was explored using an ANN, resulting to a time-activity model based solely on sensor records. The independent variables that fed the ANN input layer were consisted of a) personal temperature, Temp, b) the change of personal temperature with time, dTemp/dt, c) personal speed, Speed, e) the observed temperature, based on a central weather station measurements, Tempout, d) and the ratio of personal temperature to the observed one, Temp/Tempout. Moreover, day light information was transformed into a categorical element (day or night) which was also included as an input variable. The initial database was divided into training and validation set (85% and 15% of the total record entries, respectively) and the models developed from the training set were tested using the validation set. Results: The ANN predicted results were compared to real data based on time-activity log records, filled out on paper by participants. The accuracy of the ANN predictions is close to 85%. Conclusions: While the model is being refined, it is already clear that this kind of investigation provides useful information on the utility of commercial devices as modular add-ons to exposure studies.

Introduction: The determination of presence in micro environments including indoor vs outdoor is critical for modelling personal exposure based on time-location-activity data. The aim of this study was to investigate the potential use of smart consumer products in combination with other (sensor) data for predicting the presence in indoor and outdoor micro-environments . Methods: As part of the HEALS project time-location-activity data were collected from 28 office workers for 7 days with the MOVES app on a personal smartphone and the Fitbit Flex. In addition, real time personal air temperature (Elitech RC) was measured for all participants and real time personal UV level (Extech Luxmeter with Semrock 300/80 nm filter) was measured at 4 participants, both devices were attached to the outer clothing. Paper logs were kept by each participants for logging time-activity and indoor and outdoor locations. Results: The MOVES classification (place(=cluster), walk, cycle, transport) and the paper log correlated well. The predictive value of personal temperature, personal UV level, historical weather data (mean local temperate, rainy day) and day/time indicators (day of the week and time of the day) for further classification of indoor cluster versus outdoor cluster was explored using random forest models. Preliminary results indicate a moderate to high accuracy (65-99%) for the different study subjects. Discussion: The preliminary results indicate that when using MOVES to assess personal time-location-activity information additional (sensor) data may be used to further classify the places into indoor and outdoor places. Ongoing analyses focus on optimizing of the models for predicting indoor versus outdoor places and on generalizability of these models.

Chronic obstructive pulmonary disease (COPD) is a chronic lung disease that is thought to affect over one million people in Great Britain. The main factor contributing to the development of COPD is tobacco smoke. This paper presents a microsimulation model for the development of COPD, incorporating population dynamics and trends in smoking. The model simulates a population longitudinally throughout their lifetimes, providing projections of future COPD prevalence and evaluation of the effects of changes in risk factor prevalence such as smoking. Sensitivity analysis provides information on the most influential model parameters. The model-predicted prevalence of COPD in 2040 was 17% in males over the age of 35 years (13% amongst non-smokers and 22% amongst smokers), and a modest decline over the next 25 years due to recent trends in smoking rates. The simulation model provides us with valuable information on current and future trends in COPD in Great Britain. It was developed primarily to enable easy extension to evaluate the effects of occupational and environmental exposures on lung function and the prevalence of COPD and to allow evaluation of interventions, such as introducing health surveillance or policy changes. As longitudinal studies for investigating COPD are difficult due to the lengthy follow-up time required and the potentially large number of drop-outs, we anticipate that the model will provide a valuable tool for health impact assessment. An extended model for occupational exposures is under development and will be presented in a subsequent paper. © 2015 Published by Elsevier Ltd.

Background and aims: The determination of presence in micro-environments, including indoor vs outdoor spaces, is critical for modelling personal exposure based on time-location-activity data. The aim of this study was to investigate the potential use of smart consumer products in combination with other (sensor) data for predicting the presence of the wearer in indoor and outdoor micro-environments. Methods: As part of the HEALS project time- location-activity data were collected from 28 office workers for 7 days with the MOVES app on a personal smartphone and the Fitbit Flex. In addition, real time personal air temperature (Elitech RC) and global positioning system (GPS) coordinates (Qstarz) were measured for all participants and real time personal UV level (Extech Luxmeter with Semrock 300/80 nm filter) was measured for 4 participants, both devices were attached to the outer clothing. Paper logs were kept by each participant for logging time- activity and indoor and outdoor locations. Results: The MOVES classification (place, walk, cycle, transport) and the paper log correlated well. The predictive value of personal temperature, GPS, personal UV level, historical weather data (mean local temperate, rainy day) and day/time indicators (day of the week and time of the day) for further classification of indoor cluster versus outdoor cluster was explored using random forest models. Preliminary results indicate a moderate to high accuracy (65-99%) for the different study subjects. Conclusions: The preliminary results indicate that when using MOVES to assess personal timelocation-activity information additional (sensor) data may be used to more reliably classify the places visited into indoor and outdoor spaces. Ongoing analyses focuses on optimizing of the models for predicting indoor versus outdoor locations and on assessing the generalizability of these models.

Background: In many therapeutic areas, individual patient markers have been identified that are associated with differential treatment response. These markers include both baseline characteristics, as well as short-term changes following treatment. Using such predictive markers to select subjects for inclusion in randomized clinical trials could potentially result in more targeted studies and reduce the number of subjects to recruit. Methods: This study compared three trial designs on the sample size needed to establish treatment efficacy across a range of realistic scenarios. A conventional parallel group design served as the point of reference, while the alternative designs selected subjects on either a baseline characteristic or an early improvement after a short active run-in phase. Data were generated using a model that characterized the effect of treatment on survival as a combination of a primary effect, an interaction with a baseline marker and/or an early marker improvement. A representative scenario derived from empirical data was also evaluated. Results: Simulations showed that an active run-in design could substantially reduce the number of subjects to recruit when improvement during active run-in was a reliable predictor of differential treatment response. In this case, the baseline selection design was also more efficient than the parallel group design, but less efficient than the active run-in design with an equally restricted population. For most scenarios, however, the advantage of the baseline selection design was limited. Conclusions: An active run-in design could substantially reduce the number of subjects to recruit in a randomized clinical trial. However, just as with the baseline selection design, generalizability of results may be limited and implementation could be difficult.