We enjoyed reading the recent article by Borthakur et al (2007), entitled ‘Immune anaemias in patients with chronic lymphocytic leukaemia treated with fludarabine, cyclophosphamide and rituximab – incidence and predictors’, but have to disagree with their statements regarding the diagnosis of autoimmune haemolytic anaemia (AIHA) in patients with a negative direct antiglobulin test (DAT). In 17 patients with haemolytic anaemia following treatment with fludarabine, cyclophosphamide and rituximab, only three had positive DATs, thus fitting the generally accepted criteria for AIHA. Fourteen of the 17 patients (82·4%) had haemolytic anaemia (HA) with a negative DAT. If the patients had at least two indicators of haemolysis (increased indirect bilirubin, absolute reticulocyte count, lactase dehydroxinase or low haptoglobin), the authors classified them as having AIHA. These criteria are inadequate for a diagnosis of AIHA. DAT-negative AIHA is a well-described subgroup of the AIHAs and is difficult to diagnose, but there are suggested approaches (Petz & Garratty, 2004; Garratty, 2005). If a haematologist has a patient with HA and suspects an immune aetiology, the DAT is the first step (the positive predictive value of a positive DAT in this setting is 83%, and the negative PV is 99% for a negative DAT) (Kaplan & Garratty, 1985). If the DAT is negative, then extensive testing has to be performed to exclude the multiple other causes for haemolytic anaemia. Following this, a workup for DAT-negative AIHA can be performed [e.g. detection of small amounts of red blood cell (RBC)-bound IgG; detection of RBC-bound IgA, IgM; detection of low affinity IgG autoantibodies; studies on patient’s serum and eluates from the patient’s RBCs] (Petz & Garratty, 2004; and Garratty, 2005). If these tests are non-productive and there are haematological suggestions of immune HA (e.g. spherocytes on peripheral blood smear), then a possible diagnosis of DAT-negative AIHA can be entertained. None of these investigations were initiated by Borthakur et al (2007); their 14 patients were classified as AIHA on the basis of criteria that only suggested the presence of a haemolytic anaemia, not AIHA (i.e. haemolytic anaemia because of RBC autoantibodies).