Introduction. Although allogeneic blood transfusions have allowed surg
eons increased latitude in resecting advanced cancers, they can cause
significant morbidity or even death in rare instances. Potential side
effects may include transmission of infection and immunosuppression le
ading to an increased risk of cancer recurrence. Because patients have
become more reluctant to receive transfusions, they frequently reques
t preoperative autologous blood donation (PABD). In practice, however,
only 50% or less of the donated blood is ultimately transfused while
the remainder is discarded. Purpose. The purpose of this study was to
develop a transfusion prediction and risk assessment (TPRA model for p
redicting the need for perioperative blood transfusions in patients un
dergoing major head and neck oncologic surgical procedures. By knowing
the probability for blood transfusion, the physician and patient can
make an educated decision regarding the need for PABD. Patients and Me
thods. Over a 4-year period, 436 patients underwent major head and nec
k surgical procedures for neoplasms of the upper aerodigestive tract,
the thyroid gland, and the salivary glands. Data obtained prospectivel
y on each patient included age and gender, the TNM stage, primary dise
ase site, type of prior treatment, estimated intraoperative blood loss
, duration of surgery, transfusion requirements, preoperative and post
operative hemoglobin and hematocrit levels, type of procedure and meth
od of reconstruction. These variables were examined singly and in comb
ination both for descriptive purposes and to evaluate their interrelat
ionships. In order to develop the TPRA model, only the 12 variables av
ailable prior to the surgical procedure were examined. Variables assoc
iated with transfusion need were evaluated further in a multivariate a
nalysis. The logistic regression model allowed a linear expression of
patient characteristics to be related to a function of the probability
of transfusion need. Analyses of association between categorical vari
ables and transfusion status were based on chi-squared, Fisher's Exact
, and Mann-Whitney U tests. Results. Overall, 51 (11.7%) patients requ
ired blood transfusions. The median number of units transfused was 2.0
(range, 1 to 13 U). Univariate analysis demonstrated a higher probabi
lity for blood replacement in patients with oropharyngeal or hypophary
ngeal primary tumor sites, a preoperative hemoglobin level below norma
l, prior chemotherapy, composite resection, flap reconstruction, betwe
en 50 and 59 years of age, and T3 or T4 tumor stage. Logistic regressi
on analysis demonstrated that the need for flap reconstruction, a preo
perative hemoglobin below the normal level, and T3 or T4 primary stage
were the three factors most significantly associated with the need fo
r transfusion (P<.03). Based on eight combinations of these three vari
ables, transfusion risk predictions were obtained. The TPRA model pred
icted that patients with a normal hemoglobin level who did not require
flap reconstruction and did not have either a T3 or T4 primary stage
tumor had the lowest probability (.02) for requiring blood transfusion
. Patients at highest risk (.65) were those with less than a normal he
moglobin level, who required flap reconstruction, and had T3 or T4 pri
mary tumor stage. Based on the TPRA model, an algorithm was developed
which could serve as a guideline for preoperative transfusion planning
. Conclusion. By using the TPRA model to change guidelines for preoper
ative transfusion planning, costs can theoretically be reduced by 50%
without significantly increasing the risk of exposing patients to allo
geneic blood transfusion. If the TPRA model proves accurate in a follo
w-up study to test its validity, it may have clinical utility for aidi
ng the surgeon in more cost-effective transfusion planning.