Skip to main content

Table 3 Proposed Methods of Analysis

From: Evaluation for inherited and acquired prothrombotic defects predisposing to symptomatic thromboembolism in children with acute lymphoblastic leukemia: a protocol for a prospective, observational, cohort study

Analysis Hypothesis Independent variable Outcome variable Method of analysis
Name Variable type
Primary Presence of one or more prothrombotic defect/s increase the risk of TE Presence of one or more prothrombotic defect Symptomatic TE Binary Fisher’s Exact Test Analysis performed for overall prevalence of at least one prothrombotic defect and for individual defect.
Secondary
Aim 1. 1.a Older age increases the risk of TE, especially in presence of one or more prothrombotic defect Age of the patient (age < 10 years, age ≥ 10 years) Symptomatic TE Binary Logistic regression
1.b HR/VHR ALL increases the risk of TE, especially in presence of one or more prothrombotic defect Risk categorization of ALL (HR/VHR ALL, SR ALL)
1.c PEG ASP therapy increases the risk of TE, especially in presence of one or more prothrombotic defect Type of ASP (E. coli ASP, PEG ASP)
Aim 2 A mathematical model can be used to determine the risk of symptomatic TE Clinical Variables: age of the patient, risk categorization of ALL presence or absence of prothrombotic defect, Therapy variable: type of ASP used Symptomatic TE Binary Regression analysis will be used to predict the risk of TE
Aim 3 The overall (presence of at least one tested) prevalence of one or more prothrombotic defect is at least 20%   Prevalence of one or more prothrombotic defect/s Categorical Prevalence of individual and overall prothrombotic defect will be expressed as percentage of affected patients with 95% CI
  1. Logistic regression will be used to analyze the data for both primary and secondary outcomes. Analysis results of regression modeling will be expressed as coefficient, corresponding standard error, 95% CI and associated p-values. Variance inflation factors will be used to assess multi-collinearity among predictors. Missing values will be handled using multiple imputation or last observation carried forward. Model assumptions will be assessed through residual analysis. Goodness-of-fit will be evaluated using qq plots for normality and coefficient of determination and R2 for regression models