Artificial neural networks are applied in many situations. neuralnet is built to train multi-layer perceptrons in the context of regression analyses, i.e. to approximate functional relationships between covariates and response variables. Thus, neural networks are used as extensions of generalized linear models. neuralnet is a very flexible package. The backpropagation algorithm and three versions of resilient backpropagation are implemented and it provides a custom-choice of activation and error function. An arbitrary number of covariates and response variables as well as of hidden layers can theoretically be included. The paper gives a brief introduction to multilayer perceptrons and resilient backpropagation and demonstrates the application of neuralnet using the data set infert, which is contained in the R distribution.

BackgroundThe survival time of patients with head and neck squamous cell carcinoma (HNSCC) is related to health behavior, such as tobacco smoking and alcohol consumption. Poor oral health (OH), dental care (DC) and the frequent use of mouthwash have been shown to represent independent risk factors for head and neck cancerogenesis, but their impact on the survival of HNSCC patients has not been systematically investigated.MethodsTwo hundred seventy-six incident HNSCC cases recruited for the ARCAGE study were followed through a period of 6–10 years. Interview-based information on wearing of dentures, gum bleeding, teeth brushing, use of floss and dentist visits were grouped into weighted composite scores, i.e. oral health (OH) and dental care (DH). Use of mouthwash was assessed as frequency per day. Also obtained were other types of health behavior, such as smoking, alcohol drinking and diet, appreciated as both confounding and study variables. Endpoints were progression-free survival, overall survival and tumor-specific survival. Prognostic values were estimated using Kaplan-Meier analysis and Cox proportional hazards regression models.ResultsA good dental care score, summarizing annual dental visits, daily teeth cleaning and use of floss was associated with longer overall survival time (p = .001). The results of the Cox regression models similarly suggested a higher risk of tumor progression and shortened overall survival in patients with poor dental care, but the results lost their statistical significance after other types of health behavior had been controlled for. Frequent use of mouthwash (≥ 2 times/day) significantly increased the risk of tumor-specific death (HR = 2.26; CI = 1.19–4.32). Alcohol consumption and tobacco smoking were dose-dependently associated with tumor progression and shorter overall survival.ConclusionFrequent mouthwash use of ≥ 2 times/day seems to elevate the risk of tumor-specific death in HNSCC patients. Good dental care scores are associated with longer overall survival.Electronic supplementary materialThe online version of this article (doi:10.1186/s12903-016-0185-0) contains supplementary material, which is available to authorized users.

Children carrying the protective FTO genotype TT seem to be more protected by a favourable social environment regarding the development of obesity than children carrying the AT or AA genotype.

BackgroundDuring the last decades, sex and gender biases have been identified in various areas of biomedical and public health research, leading to compromised validity of research findings. As a response, methodological requirements were developed but these are rarely translated into research practice. The aim of this study is to provide good practice examples of sex/gender sensitive health research.MethodsWe conducted a systematic search of research articles published in JECH between 2006 and 2014. An instrument was constructed to evaluate sex/gender sensitivity in four stages of the research process (background, study design, statistical analysis, discussion).ResultsIn total, 37 articles covering diverse topics were included. Thereof, 22 were evaluated as good practice example in at least one stage; two articles achieved highest ratings across all stages. Good examples of the background referred to available knowledge on sex/gender differences and sex/gender informed theoretical frameworks. Related to the study design, good examples calculated sample sizes to be able to detect sex/gender differences, selected sex/gender sensitive outcome/exposure indicators, or chose different cut-off values for male and female participants. Good examples of statistical analyses used interaction terms with sex/gender or different shapes of the estimated relationship for men and women. Examples of good discussions interpreted their findings related to social and biological explanatory models or questioned the statistical methods used to detect sex/gender differences.ConclusionsThe identified good practice examples may inspire researchers to critically reflect on the relevance of sex/gender issues of their studies and help them to translate methodological recommendations of sex/gender sensitivity into research practice.Electronic supplementary materialThe online version of this article (doi:10.1186/s12961-017-0174-z) contains supplementary material, which is available to authorized users.

BackgroundOur aim is to investigate the ability of neural networks to model different two-locus disease models. We conduct a simulation study to compare neural networks with two standard methods, namely logistic regression models and multifactor dimensionality reduction. One hundred data sets are generated for each of six two-locus disease models, which are considered in a low and in a high risk scenario. Two models represent independence, one is a multiplicative model, and three models are epistatic. For each data set, six neural networks (with up to five hidden neurons) and five logistic regression models (the null model, three main effect models, and the full model) with two different codings for the genotype information are fitted. Additionally, the multifactor dimensionality reduction approach is applied.ResultsThe results show that neural networks are more successful in modeling the structure of the underlying disease model than logistic regression models in most of the investigated situations. In our simulation study, neither logistic regression nor multifactor dimensionality reduction are able to correctly identify biological interaction.ConclusionsNeural networks are a promising tool to handle complex data situations. However, further research is necessary concerning the interpretation of their parameters.