To the Editor
The widely used Functional Assessment of Cancer Therapy - Prostate Cancer (FACT-P) consists of the Functional Assessment of Cancer Therapy – General (FACT-G) and a prostate cancer subscale. In its second version, the FACT-G consists of five subscales measuring physical, functional, social/family, and emotional well-being in addition to satisfaction with the doctor-patient relationship (the current version 4 of the FACT-G no longer includes a relationship with doctor subscale). The prostate cancer subscale of the FACT-P includes 12 items specifically designed to measure important symptoms and concerns specific to men with prostate cancer. In an article published in 2007, Wu et al. [
1] described an algorithm for mapping FACT-P scores to EuroQol five-dimensional questionnaire scores.
Recently, we used this algorithm to map the FACT-P data to the EuroQol five-dimensional questionnaire by using data from a clinical trial. A discrepancy, however, was discovered in the published conversion equation. Specifically, our mapping effort was producing out-of-range (i.e., greater than 1) mean EuroQol five-dimensional questionnaire values.
We tried to reproduce the average utility value in the Wu et al. publication by inserting the identical input data that Wu et al. used (
Fig. 1 shows FACT-P variables;
Fig. 2 shows baseline patient characteristics) into the mapping equation presented (
Fig. 3; “Excluding EORTC” column). In doing this, we found a second discrepancy in the Wu et al. publication. In
Figure 2, the excerpt from the Wu et al. publication cites the mean body mass index (BMI) at 72.4. It is clear that this is a graphical error (the mean age value is used for mean BMI as well). To confirm this, we consulted the source publication for the input data that Wu et al. used, a summary of mCRPC observational study quality-of-life findings by Sullivan et al [
2]. In the Sullivan et al. article, we found that the mean BMI was 27, and confirmed the mean age to be 72.4 years. When applying input data from Table 1 and using average characteristics of the study population published by Wu et al. (including the corrected value for BMI) to the mapping algorithm, we arrived at an average mapped utility of 1.11 compared with the value of 0.62 reported by Wu et al.