Studies support the existence of psychosomatic phenomena that enable critically ill patients to postpone death until a specific event. We assessed for this effect in cancer by examining variability in deaths at the month and weekend levels using the National Center for Health Statistics database. We found that deaths from cancer were not uniformly distributed temporally. There was a relative 3.3% difference death rate between the peak on Saturday and nadir on Monday, and relative 10.2% difference in rate of death between the peak of deaths in January and nadir in February. The “weekend effect” could be present in 1 in 200 cancer deaths and the “holiday effect” in 1 in 100 cancer deaths. Temporal variation may reflect a small portion of patients are able to “hold on” for a limited amount of time. This uneven distribution of cancer deaths highlights the importance of improving communication and facilitating end-of-life discussions.