To the best of our knowledge, this is the first real-world study evaluating PMS of patients with HR-LPC/LAPC. Importantly, PMS was unrelated to TTM or primary treatment (age-adjusted), suggesting that PMS may be a fixed time interval, independent of the factors promoting, or time it takes to develop, metastatic disease. Our finding that PMS is independent of TTM is consistent with the findings of the ICECaP working group, who report strong surrogacy for MFS and OS among patients with LPC [9, 10]. Overall, these results suggest that the most meaningful way to improve survival of patients with HR-LPC/LAPC is to delay or prevent the development of metastases.

The PMS of the post-metastasis cohort included in our study is similar to the post-metastasis prostate cancer-specific survival reported by a previous real-world evidence study by Pascale et al. [24]. Using a database of 1364 patients with prostate cancer, Pascale et al. examined prostate cancer-specific survival after metastasis, defined as the time interval from the date of the first radiographic metastasis to the date of prostate cancer-related death or the last follow-up. Among 913 patients who presented with LPC and underwent curative treatment, 136 developed metastatic hormone-sensitive disease and had a median prostate cancer-specific survival of 50.4 (95% CI 39.1–78.5) months [24]. The similar PMS across study populations suggests that PMS is a relative constant time interval.

As expected, the risk of death increased significantly after metastasis. However, our findings provide new insights into the magnitude of change in risk of death from the pre- to post-metastasis disease states: in patients who were treated with RP or RT with or without ADT, the relative risk of death increased by eightfold and fivefold after metastasis, respectively. Notably, despite patients with HR-LPC/LAPC being at increased risk of metastases and death, the overall SMR values among the pre-metastasis cohort were lower than those in the general US male population. A similar observation was reported for a consecutive series of patients with LPC treated with RP, who had an SMR below 1 compared with an age-matched background population [25]. Røder et al. surmised that this result might reflect selection bias, including low comorbidity, long life expectancy, and higher socioeconomic status of patients who are eligible for primary treatment [25]. Among patients who developed metastases after primary treatment for HR-LPC/LAPC, RP was associated with improved PMS compared with RT only and RT + ADT. Although after accounting for age at metastasis diagnosis the results suggest independence between primary treatment and PMS, other important baseline characteristics, such as ECOG PS and CCI, could not be accounted for due to the high level of missing data. Thus, caution is required when interpreting these results.

Our findings have important implications for optimizing the OS of patients with HR-LPC/LAPC. There is some evidence that treatment intensification with androgen receptor pathway inhibitors may improve outcomes in patients with non-metastatic prostate cancer. In an exploratory pooled clinical trial analysis that included 72 patients with intermediate- and high-risk disease, neoadjuvant therapy with abiraterone acetate or enzalutamide before RP was shown to potentially have a positive impact on time to biochemical recurrence rates [26]. In the EMBARK phase 3 trial [27], 1068 patients with prostate cancer with biochemical recurrence after local therapy were randomized to receive enzalutamide + ADT, ADT alone, or enzalutamide alone, and followed for a median of 60.7 months. Results showed that patients treated with enzalutamide + ADT or enzalutamide monotherapy had improved MFS compared with those who were treated with ADT alone. The STAMPEDE trial demonstrated that treatment with abiraterone acetate + prednisolone + ADT was associated with improved OS compared with ADT alone in patients with HR-LAPC or metastatic prostate cancer [28]. Apalutamide is currently being evaluated for the treatment of HR-LPC/LAPC before RP and concurrent with RT in the PROTEUS (ClinicalTrials.gov number, NCT03767244) and ATLAS (NCT02531516) phase 3 trials, respectively. Similarly, enzalutamide is being evaluated with concurrent RT in the ENZARAD trial (NCT02446444).

Interventions that focus on diet and lifestyle may also potentially delay the development of metastasis. In patients with prostate cancer with biochemical recurrence after local treatment, a low-carbohydrate diet may be associated with a longer PSA doubling time [29]. There is also evidence of a link between obesity and prostate cancer-specific death [30], suggesting that obesity may be associated with a more aggressive form of prostate cancer [31], giving rise to the possibility that weight-loss interventions may also delay metastatic spread, although this requires formal testing.

Limitations of the current study reflect those that are inherent to the use of administrative databases for epidemiological research, including the dependency of the accuracy of ICD codes and algorithms to identify medical conditions, missing or imprecise event dates, and the retrospective design with nonrandom assignment of patients to primary treatment groups. Although the duration of ADT may impact the time to metastasis, ADT treatment duration could not be estimated owing to missingness of the ADT end date. The use of androgen receptor pathway inhibitors among patients with metastatic disease is likely to impact PMS; however, sufficient data on the use of these agents were not available in this study. Additionally, the patient cohorts included in this study were from the USA, so the findings may not be generalizable to patients globally. For example, there are differences between the USA and other countries in both policies and healthcare practices that impact prostate cancer screening, diagnosis, and treatment. Moreover, as noted above, all patients in this study were seen by medical oncology providers; thus, the data were heavily skewed toward higher-risk patients and those who needed systemic therapy, which likely explains the very high (> 50%) rate of metastases.

In addition, relevant patient prognostic factors, such as ECOG PS, were not available from the database used for this study, which hindered a full multivariate analysis, and data were also not available to estimate the proportion of patients in the post-metastasis cohort who had mCSPC or mCRPC. Other potential confounders, such as the distribution of risk factors at diagnosis and subsequent therapies received pre-metastasis, were explored in a related dataset as shown in the Supplementary Materials; however, these seem to suggest minimal impact on the key research findings. Despite this, these analyses should be interpreted with caution as the related dataset differs slightly from the cohorts described herein due to the inclusion of additional patients. As such, future research should explore data sources with higher urology and radiation oncology practice representation and detailed information on prognostic factors, including ECOG PS, CCI, disease volume, whether the disease after metastasis is mCSPC or mCRPC, number of risk factors present at baseline, and the subsequent treatments and therapies after metastasis.