We read with interest the article titled "Occurrence, Causes, and Outcome of Delirium in Patients With Advanced Cancer"1 in the March issue of the ARCHIVES. The statistical methodology is unclear to us, and we have noted some apparent errors. We are also concerned about the clinical research and ethical implications of the study as we understand it.
Lawlor et al1 used χ2 and t tests to make comparisons on a per-episode as opposed to a per-patient basis. This assumes delirium episodes are independent. It is questionable whether repeated episodes of delirium are truly independent events rather than an evolving terminal state (as supported by the 50% mortality rate in the palliative care unit). The data in Tables 3 and 4 of the article seem to indicate that the metabolic, nonrespiratory, infectious, and hematological factors have a 1:1 correspondence, but the numbers in the 2 tables do not match. The psychoactive drug factors noted in Table 4 seem to be a combination of the opioids and nonopioids listed in Table 3. Thus there should be at least 38 instances of psychoactive drug involvement in the 40 reversible episodes; 16 in the 31 nonreversible episodes. There are, however, only 15 instances of psychoactive drug involvement in the nonreversible episodes. If this is incorrect, then the univariant and multivariant Cox results are also in error.
Davis MP, Walsh D. Clinical and Ethical Questions Concerning Delirium Study on Patients With Advanced Cancer. Arch Intern Med. 2001;161(2):296-297. doi: