Limits...
Body surface area: a predictor of response to red blood cell transfusion

View Article: PubMed Central - PubMed

ABSTRACT

A current focus of transfusion medicine is a judicious strategy in transfusion of blood products. Unfortunately, our ability to predict hemoglobin (Hgb) response to transfusion has been limited. The objective of this study was to determine variability of response to red blood cell transfusion and to predict which patients will have an Hgb rise higher or lower than that predicted by the long-standing convention of “one and three”. This was a retrospective chart review in a single hospital. Data for 167 consecutive patient encounters were reviewed. The dataset was randomly divided into derivation and validation subsets with no significant differences in characteristics. DeltaHgb was defined as posttransfusion Hgb minus pre-transfusion Hgb per red blood cell unit. We classified all the patients in both the subsets as “high responders” (DeltaHgb >1 g/dL) or as “low responders” (DeltaHgb ≤1 g/dL). In univariate analysis, age, sex, body weight, estimated blood volume, and body surface area were significantly associated with response category (P<0.05). Different multivariate regression models were tested using the derivation subset. The probability of being a high responder was best calculated using the logarithmic formula eH / (1 + eH), where H is B0 + (B1 × variable 1) + (B2 × variable 2). Bis are coefficients of the models. On validation, the model H=6.5–(3.3 × body surface area), with the cutoff probability of 0.5, was found to correctly classify patients into high and low responders in 69% of cases (sensitivity 84.6%, specificity 43.8%). This model may equip clinicians to make more appropriate transfusion decisions and serve as a springboard for further research in transfusion medicine.

No MeSH data available.


ROC curve to identify the best cutoff probability value for Model 2.Note: A cutoff probability of 0.5 provides maximal sensitivity + specificity.Abbreviation: ROC, receiver operating characteristic.
© Copyright Policy
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC5036545&req=5

f1-jbm-7-199: ROC curve to identify the best cutoff probability value for Model 2.Note: A cutoff probability of 0.5 provides maximal sensitivity + specificity.Abbreviation: ROC, receiver operating characteristic.

Mentions: Univariate analysis was performed to identify variables that were statistically significantly associated with response level (high vs low). These variables included age (P=0.007), sex (P<0.001), body weight (P<0.001), EBV (P<0.001), and BSA (P<0.001). We built a variety of binary regression models to predict transfusion response using different combinations of these variables (Table 2). The probability of being a high responder was calculated using the logarithmic formula eH / (1 + eH), where H is B0 + (B1 × variable 1) + (B2 × variable 2). Bis are coefficients of the regression models, as shown in Table 2. We compared the receiver operating characteristic curves of these different models. Model 2 had the best performance with a cutoff probability value of 0.5 (Figure 1). The equation for Model 2 is H = 6.5 – (3.3 × BSA). Figure 2 depicts the sequential steps involved in predicting high vs low response.


Body surface area: a predictor of response to red blood cell transfusion
ROC curve to identify the best cutoff probability value for Model 2.Note: A cutoff probability of 0.5 provides maximal sensitivity + specificity.Abbreviation: ROC, receiver operating characteristic.
© Copyright Policy
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC5036545&req=5

f1-jbm-7-199: ROC curve to identify the best cutoff probability value for Model 2.Note: A cutoff probability of 0.5 provides maximal sensitivity + specificity.Abbreviation: ROC, receiver operating characteristic.
Mentions: Univariate analysis was performed to identify variables that were statistically significantly associated with response level (high vs low). These variables included age (P=0.007), sex (P<0.001), body weight (P<0.001), EBV (P<0.001), and BSA (P<0.001). We built a variety of binary regression models to predict transfusion response using different combinations of these variables (Table 2). The probability of being a high responder was calculated using the logarithmic formula eH / (1 + eH), where H is B0 + (B1 × variable 1) + (B2 × variable 2). Bis are coefficients of the regression models, as shown in Table 2. We compared the receiver operating characteristic curves of these different models. Model 2 had the best performance with a cutoff probability value of 0.5 (Figure 1). The equation for Model 2 is H = 6.5 – (3.3 × BSA). Figure 2 depicts the sequential steps involved in predicting high vs low response.

View Article: PubMed Central - PubMed

ABSTRACT

A current focus of transfusion medicine is a judicious strategy in transfusion of blood products. Unfortunately, our ability to predict hemoglobin (Hgb) response to transfusion has been limited. The objective of this study was to determine variability of response to red blood cell transfusion and to predict which patients will have an Hgb rise higher or lower than that predicted by the long-standing convention of &ldquo;one and three&rdquo;. This was a retrospective chart review in a single hospital. Data for 167 consecutive patient encounters were reviewed. The dataset was randomly divided into derivation and validation subsets with no significant differences in characteristics. DeltaHgb was defined as posttransfusion Hgb minus pre-transfusion Hgb per red blood cell unit. We classified all the patients in both the subsets as &ldquo;high responders&rdquo; (DeltaHgb &gt;1 g/dL) or as &ldquo;low responders&rdquo; (DeltaHgb &le;1 g/dL). In univariate analysis, age, sex, body weight, estimated blood volume, and body surface area were significantly associated with response category (P&lt;0.05). Different multivariate regression models were tested using the derivation subset. The probability of being a high responder was best calculated using the logarithmic formula eH / (1 + eH), where H is B0 + (B1 &times; variable 1) + (B2 &times; variable 2). Bis are coefficients of the models. On validation, the model H=6.5&ndash;(3.3 &times; body surface area), with the cutoff probability of 0.5, was found to correctly classify patients into high and low responders in 69% of cases (sensitivity 84.6%, specificity 43.8%). This model may equip clinicians to make more appropriate transfusion decisions and serve as a springboard for further research in transfusion medicine.

No MeSH data available.