OBJECTIVE: When undergoing cancer therapy, disease- and treatment-related symptoms can affect HRQL. In some instances, analyses show that relationships between symptoms and HRQL outcomes are trivial or not significant. Using two measures of the EORTC as an example, the present study demonstrates that relationships among domains of patient-reported outcomes measures may be more complex than can be seen using conventional regression methods. Rather than estimating only direct effects of symptoms on distal HRQL outcomes, indirect effects can be modeled to better account for patient experiences.
METHODS: Data came from phase 2 (n=125) and phase 3 (n=131) trials of gemcitabine for treatment of locally advanced or metastatic NSCLC. All items from the EORTC QLQ-C30 (except financial difficulties) and LC-13 lung cancer module were included in the theory-based path analytic structural equation model. Cycle-3 data (following two doses of study medication) were used. Symptoms were traced directly, and through intermediate outcomes, to distal HRQL outcomes. Standardized direct and indirect effects on distal outcomes were estimated and compared with total effects. Models and effects were compared between the two time periods. Estimation was conducted using Mplus v7.4.
RESULTS: Results showed logical, causal ordering among QLQ-C30 and LC-13 outcomes. For example, the effects of many symptoms, such as pain, did not significantly directly affect distal outcomes, such as physical functioning, whereas the total effects of pain were substantial and significant: direct effect=0.0, ns; indirect effects=-0.27, p<0.001; and total effects=-0.27, p<0.001, with all of the effect of pain affecting physical functioning through fatigue (i.e., pain affects fatigue which affects physical functioning).
CONCLUSIONS: The direct effect of pain, for example, on distal outcomes (e.g., physical functioning or quality of life) may be neither substantial nor significant. Estimates of total treatment effects that considers indirect effects, however, can recover meaningful results useful in decision making.
