Estimating indirEct and dirEct EffEcts of a cancEr of Unknown Primary (cUP) diagnosis on sUrvival for a 6 month-PEriod aftEr diagnosis

Objective: Cancer for Unknown Primary is cancer with unidentified metastases, and it is believed that CUP diagnosis affects survival in two different pathways; CUP may innately differ from other cancers with identified primaries, so the diagnosis itself can influence how long a person can survive. Another way is that CUP diagnosis affects how much people receive treatment, and the treatment, the mediator, is the factor to influence the outcome. In this project, using the dataset of 10575 cancer patients, I assessed direct and indirect effects of CUP diagnosis onto survival. Method: I used Inverse Propensity Weight by calculating marginal probabilities, estimated from logistic models. For indirect effect, I calculated P(treatment|confounders) as marginal probabilities to estimate IPW, while using P(pcup|confounders,treatment) for direct effect. All marginal probabilities are controlled within 0.05 and 0.95 to prevent extreme weight. Result: IPW for direct effect is 7.376, and that for indirect effect is 1.545. Conclusion: Direct effect of CUP diagnosis onto survival is much greater than indirect effect. CUP may innately from other cancers with identified metastases, and thus diagnosis itself can influence the patient’s survival regardless of treatment.