Functional impairment, symptom severity, and overall quality of life in patients with advanced lung or colorectal cancer in six European countries: baseline findings from the ACTION study

Abstract

Background

People with advanced cancer often suffer from various symptoms, which can arise from the cancer itself and its treatment, the illness experience, and/or co-morbid conditions. Important patient-reported outcomes such as functional status, symptom severity, and quality of life (QoL) might differ between countries, as countries vary with regard to contextual factors such as their healthcare system.

Purpose

To assess self-reported emotional functioning, physical functioning, symptoms, and overall QoL in patients with advanced lung or colorectal cancer from six European countries, particularly in relation to their country of residence.

Methods

We used baseline patient data from the ACTION trial, including socio-demographic and clinical data as well as patient-reported data regarding functioning, symptoms, and overall QoL (EORTC QLQ-C15-PAL).

Results

Data from 1117 patients (55% lung cancer stage III/IV, 45% colorectal cancer stage IV) were used. The highest (worst) average symptom score was found for fatigue. We found similarities but also important differences in the outcomes across countries. The best scores (the highest for emotional functioning and QoL, the lowest for symptoms) were reported by Dutch and Danish patients. Belgian patients reported relatively low emotional functioning.

Conclusion

The optimization of functioning, symptom relief, and overall QoL should be important objectives of healthcare professionals who take care of patients with advanced cancer. There are similarities, but also substantial differences across countries in functional status, symptoms, and overall QoL. Policymakers should take these differences into account and invest in offering health care catered to the needs of their population.

Appendices

Appendix 1 Statistical methods

Statistical analyses were performed using R version 3.5.1 (2018-07-02). Linear mixed models were estimated in order to explore how each outcome (emotional functioning, physical functioning, symptoms, and overall quality of life) could be understood in terms of the patients’ country of residence, taking into account other socio-demographic and clinical patient characteristics. A few candidate predictor variables were left out of the analyses, because there was little variation in the observed values (living conditions), a considerable conceptual overlap between two predictor variables (marital status and living with a partner; we chose to use the latter) or a relatively large amount of missing information mainly corresponding to “prefer not to specify” responses (religion). Given the present study’s primary focus, country of residence was always considered part of the model as a fixed effect, along with the center (hospital), which was always considered a random effect. These variables were included in a stepwise selection procedure along with the other potential predictors: cancer treatments, primary diagnosis, gender, age, years of education, living with a partner, having children, years since diagnosis, and years since diagnosis of the current cancer stage—in this order.

For each separate outcome, we started with a base model containing only centers (hospitals) and the key predictor of interest (country of residence) and then added the above-mentioned additional candidate variables one at a time. In each forward step, the candidate predictor under consideration was either retained or deleted from the model depending on its significance. Variables with doubtful results were initially removed and retried at the end of the modeling process. Variable selection decisions were based on Akaike’s Information Criterion (AIC) to compare models and verified by applying likelihood ratio (LR) tests. The analysis continued until all potential predictors were evaluated resulting in a model for each of the outcome variables. The Netherlands (the country with the highest overall quality of life) was chosen as a reference category. No theoretical standard exists to determine which reference population is most “appropriate,” and the interpretation of the results can vary depending on which comparison group is chosen. To counter this concern, we implemented a second approach to analyzing the data to see whether the two parallel methods would lead to similar results. We conducted contrasts comparing each individual country’s mean score to the average mean of the remaining countries, for all outcomes separately. We corrected for multiple testing across contrasts (i.e., controlled the experiment-wise error rate) by calculating adjusted p values according to Shaffer’s method.

Appendix 2

Table 4 Appendix Table. Global overview of mean scores for functional impairment, symptom severity, and overall quality of life per country (N = 1117)