To the best of our knowledge, this study is one of the largest to analyze transfusion practices in a PC setting.
Almost half of the patients analyzed had a gastrointestinal malignancy, as bleeding events are common in these tumors. Although, hematologic malignancies are between the most reported advanced cancer diagnosis leading to RBC transfusion , we report only a few cases (4%) of oncohematologic patients. This is in accordance with the literature where oncohematologic patients can be nearly 7% of all oncologic patients [13,14].
There is evidence supporting that these patients behave as a special group in PC when compared with other tumors: there is a smaller number of referrals, with more advanced and symptomatic disease status, and less time between the last treatment date and the referral date or death .
We report that at least 56% of patients received RBC transfusion in the previous 6 months. Thus, most of the patients were already receiving RBC before being admitted by the PC team, which could have affected the decision to repeat the procedure. Only 43% were under other therapies to treat or prevent anemia, reflecting an inadequate investigation and treatment before transfusion.
More than half of the patients were under drugs with impact on bleeding events, which may have affected the decision to transfuse. Anti-inflammatory drugs were the most frequent. These drugs should be used carefully or suspended in patient at risk for bleeding.
We found a mean pretransfusion Hb level of 6.85 g/dL, lower than Neoh et al  (7.5 g/dL) and Timothy et al retrospective  (7.8 g/dL) and prospective  (7.2 g/dL) studies but higher than Sirianni et al  (6,5 g/dL). Neoh et al  report that 70% of patients had an Hb above the trigger threshold of 7.0 g/dL, while we found an inferior frequency of 42%. Also, we found a median of RBC units given per TE (1.0) similar to Sirianni et al  and lower than Neoh et al  (2.0) and Timothy et al retrospective  (2.3) and prospective  (2.1) studies. These results may reflect doctors’ effort to adopt a restrictive blood transfusion strategy in a PC setting.
Impact on symptoms
Symptoms were recorded in 67% of TE. It is likely that a higher number of patients had symptoms, but they were not registered in patient files. Similarly, to other studies [3,10,15] the most reported symptom was asthenia/fatigue.
Considering the overall number of TE in which it was possible to assess the 15-day symptomatic benefit, we observed a symptomatic benefit in more than half of the cases. However, symptomatic benefit was not assessed in nearly one third of the TE and no standardized tool was used to demonstrate or refute benefit, since no specific validated tools to assess the utility of transfusions exist. Additionally, symptoms as fatigue are multifactorial and very difficult to measure, and a single intervention targeting one of these factors is unlikely to have a widespread benefit. Pre-transfusion ECOG-PS association with symptomatic benefit suggests that patients with a higher level of functioning may have a bigger benefit from RBC transfusion.
No serious adverse events were documented and we found a lower frequency of adverse events (4%), comparing with other studies which reported 12%  and 12,5%  adverse events. The low frequency of adverse events may reduce concerns about potential harms.
Compared to Neoh et al prospective study , in which, 32% of patients died within 30 days, we found an inferior 30-day survival-rate (57%). Compared to Sirianni et al retrospective study  we found a longer median time to death. Their analyses only considered the first TE date, while we have included each TE, as in our opinion each TE results from an independent decision. Nonetheless, deterioration is expected in this population given the progressive nature of disease and cannot be attributed to RBC transfusions.
Univariable and multivariable analyses results suggest that pre-transfusion ECOG-PS and Hb levels, as well as symptomatic benefit are independent predictors of survival in advanced cancer patients who received RBC transfusions.
Potential limitations of this study should be considered, as the retrospective design, and potential different criteria used by the assistant doctors to decide on transfusions. Moreover, as a single centre study it cannot be exclude possible selection biases accordingly to our transfusion practices.