During the pre-exam pathology review session at my medical school, one of the students asked about the relative incidence of metastases to the CNS versus primary CNS neoplasms. I answered that metastases are ten times more common than primary tumors. After the presentation, a colleague in the audience pointed out to me that the current issue of Robbins and Cotran (p. 1330) says: "about half to three quarters are primary tumors, and the rest are metastatic." I said, "No way!" and produced another textbook (the current edition of "Greenfield's Neuropathology"), which states the following on page 2116: "Metastatic tumors to the brain are approximately 10 times more common than primary intracranial neoplasms."
As we investigated the issue further, it became clear that the two textbooks were starting with a completely different denominator in arriving at their proportions. In Robbins and Cotran, the authors were looking at incidence rates of metastases in patients presenting with brain tumors. In Greenfield's Neuropathology, the authors appear to be extrapolating from autopsy series which included patients who never had a pre-mortem brain biopsy because metastasis was presumed. You might say that none of this matters too much. And, in a way, you would be right to say that. The bottom line is that a significant proportion of brain tumors are metastatic lesions. But, this discussion does matter in that it is a nice example of how statistical estimates of the prevalence of disease can vary widely depending on what denominator the author chooses to use. It is incumbent upon the author to be crystal clear about the denominator; but, unfortunately, that is not always the case – in which case it is incumbent upon the reader to beware.