Multicollinearity and misleading statistical results / 대한마취과학회지

Multicollinearity represents a high degree of linear intercorrelation between explanatory variables in a multiple regression model and leads to incorrect results of regression analyses. Diagnostic tools of multicollinearity include the variance inflation factor (VIF), condition index and condition number, and variance decomposition proportion (VDP). The multicollinearity can be expressed by the coefficient of determination (R(h)²) of a multiple regression model with one explanatory variable (X(h)) as the model’s response variable and the others (X(i) [i≠h] as its explanatory variables. The variance (σ(h)²) of the regression coefficients constituting the final regression model are proportional to the VIF(1/1−R(h)²). Hence, an increase in R(h)² (strong multicollinearity) increases σ(h)². The larger σ(h)² produces unreliable probability values and confidence intervals of the regression coefficients. The square root of the ratio of the maximum eigenvalue to each eigenvalue from the correlation matrix of standardized explanatory variables is referred to as the condition index. The condition number is the maximum condition index. Multicollinearity is present when the VIF is higher than 5 to 10 or the condition indices are higher than 10 to 30. However, they cannot indicate multicollinear explanatory variables. VDPs obtained from the eigenvectors can identify the multicollinear variables by showing the extent of the inflation of σ(h)² according to each condition index. When two or more VDPs, which correspond to a common condition index higher than 10 to 30, are higher than 0.8 to 0.9, their associated explanatory variables are multicollinear. Excluding multicollinear explanatory variables leads to statistically stable multiple regression models.

Ronda clínica y epidemiológica. Introducción al análisis multivariable (parte II) / Clinical and epidemiological round. Introduction to multivariable analysis (part II) / Ronda clínica e epidemiológica. Introdução à análise multi-variáveis (parte II

En la primera parte de este tema presentamos su definición, principales usos y los tres métodos de análisis multivariable más utilizados en la literatura científica. En esta segunda parte profundizaremos en los criterios para la incorporación de variables independientes al análisis, las herramientas para evaluar qué tan adecuado es el modelo seleccionado y la interpretación de los resultados y de los coeficientes en cada tipo de regresión.

In the first part of this topic, we explained the definition of multivariable analysis, its main uses, and the three most commonly used methods in the scientific literature. In this second part, we will delve into the inclusion criteria of independent variables to the analysis, the tools to assess how the model fits the data and the interpretation of the results and coefficients in each regression model.

Na primeira parte deste tema apresentamos sua definição, principais usos e os três métodos de análises multi-variável mais utilizados na literatura científica. Nesta segunda parte aprofundaremos nos critérios para a incorporação de variáveis independentes à análise, as ferramentas para avaliar que tão adequado é o modelo selecionado e a interpretação dos resultados e dos coeficientes em cada tipo de regressão.

Correlation study of ankle-brachial index and proteinuria in patients with acute cerebral infarction / 医学研究生学报

Objective Proteinuria is in close relation with arteriosclerosis,but little investigation was made on the correlation between ankle-brachial index( ABI) and proteinuria in patients with acute cerebral infarction. Methods 271 consecutive patients with acute cerebral infarction were enrolled from January 2013 and December 2013, which were divided into 2 groups ( patients with proteinuria and patients without proteinuria) according to urine analysis.ABI of all patients was measured and stratified into 4 groups:ABI≤0.9 (low);0.9

Risk factors for nosocomial infection in trauma patients during intensive care unit stay / 中华急诊医学杂志

Objective To determine risk factors in nosocomial infection of trauma patients during intensive care unit stay. Methods A retrospective study was carried out. A total of 1103 trauma patients admitted to the intensive care unit of five tertiary hospitals in Zhejiang Province in 2009 were reviewed. Demographic data, injury severity score and other variables related to the trauma services were collected. Univariate and multivariate analysis were processed to identify the independent risk factors of nosocomial infection in trauma patients during stay in intensive care unit. Results Overall, 171 patients( 15.5% )developed nosocomial infection during ICU stay. Of 1103 patients, 157 patients (14.2% ) died, and the 59 fatal patients were from infection group. The mortality rate in infection group was 34.7% , which was significantly higher than that in non - infection group (10.5% ). The independent risk factors of nosocomial infection in all the patients determined by using multivariate analysis included central venous monitoring, mechanical ventilation, age ≥65, the length of ICU stay > 14 days and injury severity score ≥ 16. For the severe trauma patients, central venous monitoring, mechanical ventilation, the length of ICU stay > 14 days were independent risk factors of nonsocomial infection. Conclusions The severity of injury, age, the length of ICU stay and invasive procedures were related to the nosocomial infection. To standardize the invasive procedures and to reduce the length of ICU stay may decrease the infection rate of trauma patients.

Molecular biological markers for prognosis of early stage adenocarcinoma of lung / 中国综合临床

Objective To analyze the expression of molecular biological markers in patients with early stage adenocarcinoma of lung and to determine the correlation between combined expression and prognosis. Methods Pathologic specimens were collected from 68 patients of early stage adenocarcinoma of lung in our hospital between 2000 to 2002. A panel of molecular markers, including Ki-67, Her-2, P16, P53, VEGF, MMP-9 and CD44v6, were chosen for immunohistochemical analysis of the tumor. Cox proportional hazards regression analysis was used to ana-lyze the relationship between expression of these factors and prognosis. Results Muhivariable analysis demonstrated significantly elevated risk for the following molecular markers: P53 (RR=3.228,95.0% CI 1.331-7.828, P=0.010),MMP-9(RR=2.071,95.0% CI 1.062-4.036,P=0.033),VEGF (RR=2.577,95.0% CI 1.124-5.908, P=0.025). The survival period in the patients with combined expression was remarkably shortened. Conclusions Combined evaluation of P53, VEGF and MMP-9 has clinical significance to the prognosis of early stage adenocarcino-ma of lung.

Análise de Sobrevida em Pacientes com Câncer de Pulmão

O câncer de pulmão é uma das maiores causas de morte no mundo, sendo a primeira causa em morte por câncer entre homens no Brasil. A história natural da doença inclui elevada letalidade e evolução agressiva, quase sempre com o paciente chegando ao médico quando a doença já se encontra em fase avançada. Nos últimos 40 anos a taxa de sobrevida em neoplasias de pulmão melhorou em 8 %, nos Estados Unidos, apesar do avanço nas tecnologias de detecção precoce e no refinamento das técnicas de tratamento. Um desafio contínuo é descrever o papel das variáveis clínicas, assistenciais e demográficas dos pacientes com câncer de pulmão. Neste sentido, este trabalho avaliou uma amostra da população de pacientes acompanhados pelo Programa de Oncopneumologia, do Instituto de Doenças do Tórax e do Hospital Universitário Clementino Fraga Filho (UFRJ). Neste estudo de coorte não concorrente foi estimado a importância das variáveis clínicas, demográficas e assistenciais, na sobrevida de pacientes com câncer de pulmão. Um total de 585 pacientes foram incluídos e suas principais características são: 77,8 % de homens, 85,6 % com tipo histológico não-pequenas células, 67,7 % em estágios IIIB/IV pelo TNM, e 84,4 % com história de tabagismo. A sobrevida mediana foi de 7 meses, com 5,4 % dos pacientes permanecendo vivos em 5 anos. Tabagismo, tratamento, níveis de PS e estagiamento foram determinantes prognósticos independentes para a sobrevida desta população. Apesar das limitações do desenho de estudo retrospectivo utilizando registros médicos, estudos como este são importantes para a compreensão do padrão da doença na população.

Lung cancer is a leading cause of death in the world, and the first one in cancer-related death, in Brazil. The natural history of disease includes high lethality and aggressive clinical course, with the patient usually presenting advanced disease at first medical evaluation. Although improvement of medical technologies, early detection and newest therapies, the mortality rate was reduced only 8% in the last 40 years, in U.S.A. The role of patient’s clinical, social and health assistance characteristics, in lung cancer survival, is a continuous challenge. In this way, we evaluated a sample of patients treated at Rio de Janeiro Federal University, by Onco-Pneumology Program, a multidisciplinary group of Thorax Diseases Institute and Clementino Fraga Filho Universitary Hospital. In this non concurrent cohort study was estimated the influence of clinical, demographic and health assistance related variables in survival of lung cancer patients. A total of 585 patients were reviewed and their main characteristics are: 77,8 % males, 85,6 % with non-small-cell histology, 67,7 % IIIB/IV TNM stages, and 84,4 % with positive smoke history. The median survival time was 7 months, with 5,4 % of the patients alive in 5 years. Smoking, treatment, PS levels and TNM stage were independent prognostic determinants of survival in this population. Despite of the limitations of a retrospective design using medical records, studies like this one are important for understanding the pattern of the diseases in a population.