Resolució XIV Concurs Student

Dels treballs presentats al XIV Concurs Student d'Estadística Aplicada:

  • Chronic kidney disease risk-assessment model, based upon genetic polymorphisms and classical risk factors. A case-control analysis in the NEFRONA study
  • Inferencia para datos funcionales: Estudio sobre la contaminación del aire en Barcelona
  • Algoritmo de diagnosis para el Alzheimer
  • Análisis de modelos estadísticos para estudios de asociación del genoma completo
  • Comprensión de la información clínica. Adherencia al tratamiento en pacientes mayores polimedicados
  • Celebridades vs Youtubers: Influencia en la decisión de compra de los millennials españoles
  • Análisis criminológicos con R: Datos espaciales y localización de puntos calientes
  • Generalized decline in tree resilience to drought in three dominant pine species in ne Iberian peninsula
  • Assessing different bivariate discrete distributions to detect disease outbreaks in reproduction reports of sow farrowings
  • Efficient clinical trial design for the selection of a primary binary endpoint

 

Els membres del Jurat del XIV Concurs STUDENT d'Estadística Aplicada, han resolt atorgar el premi Idescat al millor treball de màster a:

 

Marta Bofill Roig

 
  Estudiant del Màster Universitari en Estadística i Investigació Operativa
Universitat Politècnica de Catalunya
 

Pel treball:

 

Efficient clinical trial design for the selection of a primary binary endpoint

 
    The choice of the primary endpoint is an important issue when designing a clinical trial. Gomez and Lagakos [5] proposed the ARE (Asymptotic Relative Eciency) method to discern between using a composite endpoint or one of its component as the primary endpoint to compare two treatment in the context of survival analysis.
Aiming to provide statistical guidelines that would indicate when it is more efficient to use the composite endpoint over one of its components as the primary endpoint of the trial, we expand the ARE method to binary endpoints.
We show that the ARE method for binary composite endpoint depends on six parameters including the degree of association between the components of the composite endpoint on each group, the event proportion, and the e ffect of therapy given in each component.
Finally, we describe the method and illustrate its use in a real case study for the prevention of venous thromboembolism.
   

 

Han resolt atorgar el premi Almirall al millor treball de bioestadística a: 

 

Carles Pérez Guallar

 
  Unitat de Bioestadística i Epidemiologia, Institut de Recerca Biomèdica de Lleida
Estudiant del Grau d'Estadística Aplicada
Universitat Autònoma de Barcelona
 

Pel treball:

 

Chronic kidney disease risk-assessment model, based upon genetic polymorphisms and classical risk factors. A case-control analysis in the NEFRONA study

 
    Chronic kidney disease (CKD) is a major risk factor for end-stage renal disease, cardiovascular disease and premature death, with a high prevalence (10.4% and 11.8% respectively in men and women aged 20 and older), which requires enormous medical costs for its treatment. CKD is considered a major public health problem and international efforts are still needed for prevention, detection and treatment. Classical risk factors for CKD have been identified, being diabetes and hypertension the leading causes of CKD. However, few studies have assessed genetic risk factors and their combined effect with the classical ones. Here, we assessed the genetic association with CKD of 79 single nucleotide polymorphisms (SNPs) previously related with CKD risk factors, using for this purpose the baseline data provided by the multicenter prospective NEFRONA study, including 2445 CKD patients (cases) and 559 controls. Logistic regression models, considering different genetic inheritance models, were used to assess the association of the SNPs and haplotypes with CKD. Twelve polymorphisms and three haplotypes were found significantly associated with CKD, even after adjusting by sex, age, and race. From these, multivariate analysis revealed a combination of 8 SNPs associated with CKD. Furthermore, when additionally considering diabetes and hypertension in the multivariate model, the 8-SNP combination was found statistically significant (p<0.00001, adjusted p<0.00001). A final model containing a 5-SNP combination, diabetes and hypertension and the interaction of hypertension with one of the SNPs as found as the most accurate one, showing a ROC curve with an AUC of 0.82 and a sensitivity and specificity of 88.1% (86.0-90.0) and 66.9% (62.8-70.9). In conclusion, we successfully identified a set of SNPs and haplotypes associated with CKD, which could be used together with diabetes and hypertension as risk factors for CKD in clinical practice, prevention, treatment and genetic counselling.    

 

Han resolt atorgar el premi SEA al millor treball de grau a:

 

Nicolás Fernández Martínez

 
  Unitat de Bioestadística i Epidemiologia, Institut de Recerca Biomèdica de Lleida
Estudiant del Grau d'Estadística Aplicada
Universitat Autònoma de Barcelona
 

Pel treball:

 

Assessing different bivariate discrete distributions to detect disease outbreaks in reproduction records of sow farrowings

 
    Porcine reproductive and respiratory syndrome (PRRS) is a viral disease with negative impacts on reproduction of sows, being associated to a decrease in the number of born alive piglets (NBA) and an increase in the number of lost piglets (NLP) in the farrowings. At practice, it is of interest to assess whether a farm is under an (PRRS) outbreak situation. The aim is to develop a statistical method to distinguish healthy and disease phases, using farming data and considering different discrete bivariate distributions. Data from a farm in Lleida was used containing records on both NBA and NLP, date of the farrowing for each sow and parity. Different discrete bivariate probability distributions were considered to model NBA and NLP, including double Poisson, bivariate Poisson, diagonal-inflated bivariate Poisson and a conditional Poisson model, using different R packages. Since a relevant negative correlation between NBA and NLP was observed the conditional Poisson model was the one that reasonably handled best with the data, conditioning by NBA. The analysis of the data showed a clear outbreak period revealing a displacement of the mean NBA and NLP values (lower and higher respectively). For outbreak detection, the expected values of NBA and NLP under a nonoutbreak scenario were estimated using a ML procedure. For each farrowing a p-value was computed defined as the probability of jointly observing a lower NBA and higher NLP. A combined p-value using the last 100 p-values was computed using the ?²-inverse method and, alternatively, a Wald test using a single ML estimation considering the 100 last farrowing under a nonoutbreak scenario was also performed. A displacement of the hypothesized mean NBA and NLP values was considered in the p-value computation in order to be more specific. The presented method can be used to efficiently detect two periods with an outbreak situation. The combination of p-values and the direct use of a Wald test lead to similar results.    

 

 

 

El Servei d'Estadística de la UAB i l'Institut d'Estadística de Catalunya, volen expressar la seva enhorabona més efusiva als guanyadors, així com agraïr a la resta de participants l'interès i l'esforç mostrat.

L'acte d'entrega de premis va tenir lloc el passat 7 d'octubre de 2016, en motiu del XVI Dia de l'Estadística a Catalunya organitzat per l'Institut de Salut Global (ISGlobal), on es va fer públic el nom dels guanyadors i s'entregaren els premis i diplomes.

 

Campus d'excel·lència internacional U A B