Quality of life outcomes in a cancer trial: dealing with missing data

A phase 3 cancer trial with EORTC QLQ-C30

The EORTC QLQ-C30 is a 30-item questionnaire that has been designed for use in a wide range of cancer patient populations and is a reliable and valid measure of the quality of life in cancer patients. It includes a number of different scales, but this challenge is focussed on the global health and quality of life scale (QL). Further details of EORTC QLQ-C30 can be found in the Sept 2022 challenge.

Details of the simulated study are:

Two arms (Experimental Treatment vs. Standard of Care);
100 participants in the Experimental Arm and 100 participants in the Standard of Care Arm;
Participants are followed for 48 weeks;
The PRO scores are collected at baseline and every 3 weeks

Dataset

The dataset contains the following variables:

USUBJID
ARM (“Experimental Treatment”, “Standard of Care”)
The longitudinal evaluation of QL from Week 0 (WEEK00, numeric) to Week 48 (WEEK48, numeric)
LASTVAL (last observed value before dropout, numeric)
AVGVAL (QL average over available assessments, numeric)
LASTVIS (last available assessment before dropout, numeric)
AGE (numeric)
AGEGR (“(35,45]”, “(45,55]”, “(55,65]”, “(65,75]”)
SEX (“F”, “M”)
BMI (numeric)
BMIGR (“Normal”, “Obese”, “Overweight”, “Underweight”)

The dataset is simulated with three patterns of missing data for QL:

a MCAR pattern where the probability of missingness does not depend on observed and unobserved characteristics
a MAR pattern where the probability of missingness depends on one covariate
a MNAR pattern, where the probability of dropout depends on the QL worsening

Visualization task

To produce data visualisations that provide insight on the missingness pattern of the data:

Summarise the overall pattern: how many patients withdrew, when did patients withdraw?
Use visualisation to show any covariates associated with an increased probability of missing data?
Use visualisation to sumarise whether a worsening of QL leads to early withdrawal?