We consider that your discussion on the possible overuse of blood transfusions simplifies a complex issue (Nature 520, 24–26; 2015).

Readers might infer, for example, that a standardized transfusion protocol is safer than individualized blood-management care, or that restricted transfusion in response to a particular haemoglobin concentration is at least as safe and effective as transfusion titrations determined by a range of haemoglobin levels. In fact, neither hypothesis has been tested. As you indicate, physicians need to take into account individuals' primary disease as well as any complications and accompanying disorders. Such factors can improve patient care in the longer term, beyond any simple numerical guidelines for restrictive-transfusion practices.

We have shown that a restrictive-transfusion threshold at 7 grams per decilitre of haemoglobin can be problematic for people with stable blood pressure and cardiovascular status (H. G. Klein and C. Natanson Ann. Intern. Med. 157, 753–754; 2012). We argue elsewhere that the pivotal trials you cite would have been more robust had they included a range of transfusion triggers, instead of just two arbitrarily selected haemoglobin levels, and a standard-of-care control arm to incorporate all clinical and lab observations for each patient (K. J. Deans et al. Vox Sang. 99, 16–23; 2010). Restrictive practices save blood, but they do not necessarily save lives.