I'm not sure I agree with that, but if I had to support that assertion, I would say that while patients with COPD, CHF and neurodegenerative diseases may die of hypoxia (no fever there), patients with cancer more often die of opportunistic infections.
In patients being treated with chemotherapy (and for some time afterwards as well), myelosuppression - that is suppression of white blood cells vital to fighting infection - is a major risk factor, and is the major dose-limiting factor of systemic chemotherapy.
However, as I stated, I don't agree that fever is a predictor of end of life in cancer patients.
In one large study of terminally ill cancer inpatients, infection was the cause of death in only 15% of patients. More common causes included congestive heart failure (24%), renal failure (30%), liver disease (30%), cerebrovascular disease (30%), pulmonary embolism (25%) - very common in cancer patients - and lung disease. (Overlap exists because of comorbidity.)
Another study of hospice patients found the major problems in terminal cancer patients were the presence of cognitive failure, weight loss, dysphagia, anorexia and dyspnea to be independent survival predictors in this population (which somewhat supports the above study.)
The National Cancer Institute's Last Days of Life–Health Professional Version does not even list fever as a predictor of end of life (EOL).
Among a few others, the late signs found highly specific for impending death in 3 days involved instead signs of cardiac, cerebrovascular, respiratory, and renal failure.
Similarly, a study in the journal Supportive Care in Cancer found that according to palliative care experts (78 professionals, including physicians, nurses, and others):
changes in the following categories (each consisting of up to three phenomena) were considered highly relevant to clinicians in identifying and predicting a patient's last hours/days of life: "breathing", "general deterioration", "consciousness/cognition", "skin", "intake of fluid, food, others", "emotional state" and "non-observations/expressed opinions/other".
Coley's toxins have to do with treatment of cancer. I'm not sure that they have anything whatsoever to do with death from cancer.
Mortality, morbidity, and cost associated with febrile neutropenia in adult cancer patients
Invasive fungal infections in patients with hematologic malignancies in a tertiary care cancer center: an autopsy study over a 15-year period (1989-2003)
Last Days of Life–Health Professional Version
International palliative care experts' view on phenomena indicating the last hours and days of life.