We externally validated the ARC in CRPC patients who were administered DTX in 2 or more cycles and showed that there were statistically significant differences in OS among the ARC risk groups. The median OS and 95% CI for each ARC risk group were similar between the CRPC patients used for validation in this study (the validation group) and the CRPC patients used to develop the ARC  (the development group). The c-index for OS was 0.60, indicating that the ARC had a modest discriminatory ability in the validation group. A multivariate analysis revealed that the ARC was an independent prognostic factor. Thus, the ability of the ARC to classify and approximately predict the OS of CRPC patients with certain reproducibility was confirmed, suggesting that the ARC is useful when predicting prognosis in DTX-treated CRPC patients. This means that CRPC patients who are classified into good- and intermediate-risk groups are recommended for aggressive DTX administration, because these patients would be expected to experience prolonged OS in response to DTX. However, CRPC patients who are classified as poor-risk should be recommended for clinical trial participation or other treatments because given the poor outcomes, these patients would be expected to experience a limited prognosis despite the use of DTX.
ARC was also reported to be able to classify PSA response in CRPC patients, although this system was principally aimed at classifying OS . We also externally validated the usefulness of ARC in classifying the PSA response of CRPC patients. However, there were no statistically significant differences in PSA response between the ARC risk groups. This result was considered to be caused by the reason that most of CRPC patients in the validation group had a history of EMP use; the frequency of a history of EMP use was high in good-risk group (92%) compared to intermediate-risk (82%) and poor-risk (80%) groups. The CRPC patients with a history of EMP use showed significantly lower PSA response during DTX treatment than those without a history of EMP use . Thus, considering that PSA response during DTX treatment are likely to be low in the CRPC patients with a history of EMP use, our cohort might not be suitable for validating the usefulness of ARC in classifying the PSA response of CRPC patients.
This validation study exhibited the following characteristics with respect to the ARC: the development group presented with median PSA levels of 110 ng/mL at DTX initiation. From this group, symptoms and imaging test items that are often observed after some disease progression, including bone scan progression , significant pain , and visceral metastases [8, 21], were identified as risk factors. However, as the efficacy of DTX was established in CRPC patients and DTX initiation in patients with low PSA levels was found to confer better prognosis [9, 15, 17, 21], the likelihood of initiating DTX at lower PSA levels has increased to a level higher than those reported in the TAX327 and SWOG9916 trials. The validation group presented median PSA levels of 20 ng/mL at DTX initiation, a lower value than that of the development group, and possessed few ARC risk factors; this led to a disproportionate distribution in which 65% of CRPC patients in the validation group were classified as good-risk. Thus, the ARC tends to classify many CRPC patients with low PSA levels as good-risk. At the comparison of the prognoses and/or treatment responses of CRPC patients, the ARC would ensure more accurate outcomes while considering the above-mentioned ARC characteristics.
This study had the following limitations: a retrospective study design; a small sample size; different patient backgrounds in the validation and development groups in terms of EMP exposure, lower PSA levels, lack of visceral spread, and lower numbers of DTX cycles; and a different definition of significant pain as a risk factor. Although these major limitations could have possibly deteriorated the quality of this external validation study of ARC, it was noteworthy that the ARC indicated good discriminatory ability for OS even in this validation group.