Researchers are conducting a cross-sectional study to demonstrate the relationship between renal frame count (RFC), considered the dependent variable, and LDL, considered an independent variable. RFC is an objective quantitative angiographic method of measuring macrovascular blood flow in the main renal artery and its segmental branches. Two hundred patients were categorized into 2 groups according to serum LDL levels—LDL <130 mg/dL (n = 90) and LDL ≥130 mg/dL (n = 110). Other parameters included BMI, platelet count, and creatinine clearance. Which of the following is the best statistical approach for simultaneously evaluating the association between RFC and LDL cholesterol while adjusting for BMI, platelet count, and creatinine clearance?
Dependent variable | |||
Qualitative (categorical) | Quantitative | ||
Independent variable | Qualitative (categorical) | Chi-square, logistic regression* | t test, ANOVA, linear regression |
Quantitative | Logistic regression* | Correlation, linear regression | |
*Dependent variable must be dichotomous. ANOVA = analysis of variance. | |||
Identifying the best statistical approach for any research question depends on the type and number of variables studied. Variables can be broadly classified as follows:
This study estimates the relationship between renal frame count (RFC) (quantitative dependent variable) and LDL (independent variable); the analysis adjusts for BMI, platelet count (PC), and creatinine clearance (CC) (independent variables considered for adjustment). Although LDL is a quantitative variable, in this study it has been categorized (LDL <130 mg/dL versus LDL ≥130 mg/dL). Therefore, the best statistical approach must simultaneously accommodate the following variables:
| Parameter | Type | |
| Dependent variable | RFC | Quantitative |
| Independent variables | LDL | Qualitative (categorical) |
| BMI* | Quantitative | |
| PC* | Quantitative | |
| CC* | Quantitative | |
| CC = creatinine clearance; PC = platelet count; RFC = renal frame count. *Adjustment variables | ||
Multiple linear regression (MLR) is used to evaluate associations between 1 quantitative dependent variable (primary outcome) and ≥2 independent variables that can be either quantitative or qualitative. The effect of the main independent variable of interest (eg, LDL) on the primary outcome (eg, RFC) is adjusted to account for the effects of other independent variables in the model (eg, BMI, PC, CC). Controlling for these variables reduces the potential confounding effects they may have on the relationship under study.
(Choice B) Multiple logistic regression is used to estimate the association between ≥2 independent variables and 1 dichotomous dependent variable (ie, with 2 outcomes). For example, a logistic regression model would evaluate the association between presence or absence of type 2 diabetes mellitus (dichotomous dependent variable) and obesity while adjusting for age and smoking status (3 independent variables). In this study of RFC and LDL, the dependent variable, RFC, is quantitative (not dichotomous); therefore, logistic regression would not be used (even though LDL, one of the independent variables, is categorized as dichotomous).
(Choice C) The 1-way analysis of variance (ANOVA) is used to compare the means of a quantitative dependent variable among ≥2 group means. However, it cannot simultaneously adjust for other variables.
(Choice D) Correlation analysis measures the strength and direction of a linear relationship between 2 quantitative variables. For example, a study may report a correlation coefficient describing the relationship between RFC and BMI. However, this analysis does not adjust for the effect of other variables.
Educational objective:
Multiple linear regression evaluates the association between a quantitative dependent variable and independent variables of interest while controlling for the effects of other factors (adjustment variables).