Background: Prostate cancer (PCa) is the most common cancer among men, and sepsis is a leading cause of death in PCa, yet their association has not been quantified. This study aims to quantify the association between sepsis severity (identified as sepsis, severe sepsis, and septic shock) and prostate complications among PCa subgroups.

Methods: We used the SEER-Medicare data to identify patients diagnosed with PCa in 2010 -2015. PCa patients were classified as: localized or regional (L/R) stage and distant (D) stage and further stratified as survivors and non-survivors. Patients with no history of PCa composed the control group. Our network-based modeling approach used sepsis and prostate complications (identified using ICD codes) as network nodes. The association between complication pairs, represented as edges, were weighted by the probability of complication pairs occurring concurrently. The network densities for each subgroup were computed at different thresholds based on the weights of the edges.

Results: We identified 27,166 patients with PCa and 22,223 as controls. Increase in threshold was associated with highest decline in network density among controls followed by L/R survivors (Table). Additionally, prostate and sepsis complications co-occur more frequently among stage D than stage L/R patients. Also, stage D patients experience prostate complications with more severe sepsis conditions (severe sepsis and septic shock) than other subpopulations. The variances and medians of edges' weights among all subgroups were statistically different (p<0.001).

Conclusions: This is the first population-based study using network-based modeling to quantify the association between varying degrees of sepsis severity and prostate complications among PCa subgroups. The differentiating complication pairs identified among subgroups can be used as features in machine-learning algorithms to predict the severity of patient outcomes and to improve patient prognosis.

Table: Network densities of sepsis and PCa complication pairs at different weight thresholds for each PCa subgroups

Network densities at various threshold weightsQuantiles of Weights in Network (Median (Q1,Q3))
Threshold Levels 0.000 0.010 0.020 0.030 0.040  
Control 0.193 0.005 0.003 0.001 0.001 0.005 (0.002, 0.016) 
Stage L\R survivors 0.241 0.029 0.011 0.007 0.005 0.007 (0.003, 0.021) 
Stage L\R non-survivors 0.158 0.038 0.014 0.007 0.004 0.012 (0.005, 0.029) 
Stage D survivors 0.084 0.027 0.006 0.002 0.001 0.013 (0.006, 0.032) 
Stage D non-survivors 0.130 0.036 0.017 0.009 0.004 0.012 (0.005, 0.029) 
Network densities at various threshold weightsQuantiles of Weights in Network (Median (Q1,Q3))
Threshold Levels 0.000 0.010 0.020 0.030 0.040  
Control 0.193 0.005 0.003 0.001 0.001 0.005 (0.002, 0.016) 
Stage L\R survivors 0.241 0.029 0.011 0.007 0.005 0.007 (0.003, 0.021) 
Stage L\R non-survivors 0.158 0.038 0.014 0.007 0.004 0.012 (0.005, 0.029) 
Stage D survivors 0.084 0.027 0.006 0.002 0.001 0.013 (0.006, 0.032) 
Stage D non-survivors 0.130 0.036 0.017 0.009 0.004 0.012 (0.005, 0.029) 

Citation Format: Ali Jazayeri, Niusha Jafari, Christopher C. Yang, Nikita Nikita, Grace Lu Yao. Risk of sepsis among patients with prostate cancer: A network-based modeling approach [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 234.