The role of the prognostic nutritional index at baseline and follow-up in patients with metastatic prostate cancer.

BACKGROUND/

OBJECTIVES

Patients with metastatic prostate cancer (mPC) represent a population with a higher proportion of elderly patients compared to other patient groups. Androgen receptor pathway inhibitors (ARPISs) are frequently used in current treatment regimens. Although the Prognostic Nutritional Index (PNI) is widely utilized, its stand alone capacity to accurately predict overall survival (OS) and progression-free survival (PFS) remains limited. This study aimed to evaluate the prognostic value of PNI in mPC patients treated with ARPIs and to optimize survival predictions by developing integrated multivariable logistic and Cox regression models.

METHODS

PNI was calculated at baseline and third month after treatment initiation. To assess cross-sectional predictive capability, multivariable logistic regression models were developed for OS and PFS events, and their diagnostic performance was evaluated using Receiver Operating Characteristic (ROC) curve analysis. Subsequently, time-to-event dynamics and hazard ratios were analyzed using univariable and multivariable Cox proportional hazards regression.

RESULTS

The multivariable logit model demonstrated robust predictive performance for survival outcomes. Utilizing the model-derived optimal probability cutoff of 0.343, patients were successfully stratified into distinct prognostic cohorts. The high-probability risk group (≥ 0.343, n = 86) exhibited a significantly worse prognosis, with accelerated curve decline and earlier, more frequent mortality. In contrast, the low-probability group (< 0.343, n = 51) showed a favorable prognosis, maintaining an approximate 30% survival rate up to the 75th month (Log-rank p = 0.009).

CONCLUSION

Evaluating nutrition alone is not enough to predict how mPC patients will progress. By tracking PNI changes over time and combining them into multivariable logistic model, we can much better predict overall survival and progression-free survival. In practice, using our model's 0.343 cutoff score allows us to easily catch high-risk patients early on, helping physician step in with personalized nutritional and medical support.