Cum hoc ergo propter hoc (With this, therefore because of this) — this logical fallacy often results in erroneous deductions. For example, the increase of ice cream consumption during summer time and the concomitant increase of sunburns does not necessarily mean that ice cream causes sunburn. Admittedly, our mind loves such correlations, and as a consequence we often believe that mathematical correlation is equivalent to causation.
This is also very true for many studies about transfusion medicine. There are virtually no studies that fails to demonstrate a significant correlation between blood transfusion and postoperative morbidity and mortality, suggesting deleterious effects of blood transfusion.
Chalfin and coworkers tried to overcome this phenomenon by adjusting for potential confounders in bladder cancer patients receiving neoadjuvant chemotherapy and radical cystectomy. The cohort evaluated in their study is of particular interest as neoadjuvant chemotherapy is known to be associated with haematologic side effects that may lead to relative immunosuppression, and as a consequence transfusion-related immunomodulation might be more pronounced in these patients.
Using logistic regression models controlling for confounders, the authors identified factors that are independently associated with perioperative morbidity and mortality. In multivariate analysis, perioperative blood transfusion was associated with increased surgical morbidity and length of stay, and intraoperative (but not perioperative) blood transfusion was associated with decreased overall survival and bladder cancer-specific survival. However, the number of patients included was fairly low, and thus only a small number of confounders could be controlled for. Importantly, this study warns us not to misinterpret correlation as caution.
– Jens Meier