- Email: [email protected]
Quantifying risk of transfusion in children undergoing spine surgery
The Spine Journal 2 (2002) 166–172
Clinical Studies
Quantifying risk of transfusion in children undergoing spine surgery Michael G. Vitale, MD, MPHa,c,*, Douglas E. Levy, MPHa, Maxwell C. Park, MDb, Hyunok Choi, MPHa, Julie C. Choe, MPHa, David P. Roye, Jr., MDc a
International Center for Health Outcomes and Innovation Research, College of Physicians and Surgeons, and The Joseph L. Mailman School of Public Health, Columbia University and New York Presbyterian Hospital, 600 West 168th Street, 7th Floor, New York, NY, 10032 USA b New York Orthopaedic Hospital, New York Presbyterian Medical Center, 622 West 168th Street, 11th Floor, New York, NY, 10032 USA c Division of Pediatric Orthopaedics, Department of Orthopaedic Surgery, Columbia University College of Physicians and Surgeons, 3959 Broadway, 8th Floor North New York, NY, 10032 USA Received 22 September 2001; accepted 7 February 2002
Abstract
Background context: The risks and costs of transfusion are a great concern in the area of pediatric spine surgery, because it is a blood-intensive procedure with a high risk for transfusion. Therefore, determining the predictors of transfusion in this patient population is an important first step and has the potential to improve upon the current approaches to reducing transfusion rates. In this study, we reveal several predictors of transfusion in a pediatric patient population undergoing spine surgery. In turn, we present a general rule of thumb (“rule of two’s”) for gauging transfusion risk, thus enhancing the surgeon’s approach to avoiding transfusion in certain clinical scenarios. Purpose: This study was conducted to determine the main factors of transfusion in a population of pediatric patients undergoing scoliosis surgery. The goal was to present an algorithm for quantifying the true risk of transfusion for various patient groups that would highlight patients “at high risk” for transfusion. This is especially important in light of the various risks associated with undergoing a transfusion, as well as the costs involved in maintaining and disposing of exogenous blood materials. Study design/setting: This is a retrospective review of a group of children who underwent scoliosis surgery between 1988 and 1995 at an academic institution. Patient sample: A total of 290 patients were analyzed in this study, of which 63 were transfused and 227 were not. Outcome measures: No outcomes measures were used in this study. Methods: A retrospective review of 290 patients presenting to our institution for scoliosis surgery was conducted, with a focus on socioclinical data related to transfusion risk. Univariate analysis and logistic regression were used to quantify the determinants of transfusion risk. Results: Univariate analysis identified many factors that were associated with the risk of transfusion. However, it is clear that several of these factors are dependent on each other, obscuring the true issues driving transfusion need. We used multivariate analysis to control for the various univariate predictors of transfusion. Our logistic regression model suggested that the type of scoliosis (odds ratio [OR], 2.02; 95% confidence interval [CI], 1.07 to 3.82), degree of curvature (OR, 1.012/degree curve; 95% CI, 1.01 to 1.03), and use of erythropoietin (OR, 0.29; 95% CI, 0.14 to 0.62) were the main determinants of transfusion risk for our population. Conclusions: The main risk factors of transfusion were used to formulate a simple algorithm, which can be used to quantify transfusion risk and to guide efforts to avoid transfusion in children undergoing spinal surgery. Given a 10% baseline risk for transfusion, our “rule of two’s” indicates that each risk factor approximately doubles the chance of transfusion, whereas the administration of recombinant human erythropoietin roughly halves the risk of transfusion. © 2002 Elsevier Science Inc. All rights reserved.
Keywords:
Transfusion; Erythropoietin; Pediatrics; Scoliosis; Clinical decision rule; Probability of transfusion; Rule of two’s
FDA device/drug status: approved for this indication (erythropietin). Support in whole or in part was received from Ortho Biotech. The author (MV) acknowledges a financial relationship (grant research support from Ortho Biotech), which may indirectly relate to the subject of this manuscript. * Corresponding author. 600 West 168th Street, 7th Floor, New York, NY 10032, USA. Tel: (212) 305-5028; fax: (212) 305-4256. E-mail address: [email protected] (M.G. Vitale).
Introduction Despite efforts to avoid allogenic transfusion, children undergoing spine surgery often require exogenous blood products (Rothstein P, Jr., Roye DP, Verdisco L, Experience with preoperative administration of erythropoietin in
1529-9430/02/$ – see front matter © 2002 Elsevier Science Inc. All rights reserved. PII: S1529-9430(02)00 1 7 4 - 2
M.G. Vitale et al. / The Spine Journal 2 (2002) 166–172
adolescents undergoing scoliosis surgery, personal communication, 1997; [1–6]). Concerns over the safety of the blood supply have resulted in a dramatic decrease in the risk of transmission of viruses, including human immunodeficiency virus (HIV) and hepatitis. Nevertheless, it is unlikely that screening strategies will be able to decrease this risk to zero [3,5,7–11]. Furthermore, several studies have highlighted other potential negative effects of transfusion, including ABO mismatch as a result of labeling errors and adverse immunological effects, such as allergic reaction, graft-versus-host disease and alloimmunization [12]. For these reasons, there has been an evolution of techniques and strategies to avoid allogenic transfusion. Measures that can be taken to decrease the possibility of transfusion include the use of preoperative erythropoietin, preoperative autologous donation, preoperative hemodilution, intraoperative salvage or postoperative salvage of autologous blood, intraoperative normovolemic hemodilution and hypotensive anesthesia [2,13–25]. Although many of these techniques have been shown to effectively decrease transfusion requirements, the appropriate indications for use of these costly strategies have not been well described. Clearly, efforts to avert transfusion are more important and more cost effective in situations where transfusion is more likely. The decision to use cell saver, for example, is often based on the clinician’s implicit assessment regarding the risk of blood loss and risk of transfusion in a particular clinical setting. The setting can be predetermined by the patient (eg, the type of scoliosis or degree of curvature) and influenced by the physician (eg, use of cell saver or preoperative erythropoietin). An improved understanding of the specific factors and the quantitative effect that each of these factors has on transfusion risk can help guide the clinician in making the most informed decision possible in this setting. To date, there is a great need to better understand the factors that drive transfusion rates in children undergoing spinal surgery. Few studies have specifically examined the relationship between socioclinical factors and transfusion risk in this patient population [26–29]. In a recent retrospective review of 244 adult spine patients who underwent spinal instrumentation and fusion, Nuttall et al. [26] identified preoperative hemoglobin concentration, increased number of donated autologous red blood cells, pulmonary disease, surgery for tumor and greater number of spine levels fused as significant predictors for allogenic transfusion need. This group concluded that multiple factors determine the need for transfusion, suggesting that the approach to reducing this need should be multifaceted, yet integrated. Similarly, we analyzed the factors significant in affecting need for transfusion in the pediatric population undergoing multilevel spine surgery. However, we extended the analysis by producing a quantitative model, thus providing a practical means of gauging transfusion risk. Ideally, this will enhance the approach to preventing unnecessary transfusions in pediatric spine surgery cases. This current review
167
is the first such effort to quantitatively model the relative effect of various clinical factors on transfusion rates in a pediatric orthopedic population. Materials and methods Data This study was a retrospective review of the medical records of pediatric patients who underwent spine surgery with the principal investigator. Of 420 patients initially identified, records were available for 317 patients (75%). Furthermore, among these 317 patients, 290 had verifiable transfusion and recombinant human erythropoietin (rh-EPO) records. This limited the sample to 63 individuals who received intraoperative transfusion (defined as “transfused”) and 227 individuals who did not (defined as “nontransfused”). During the period of this review, we began to use rh-EPO routinely in an effort to avert transfusion. This has proven to exert a strong protective effect against transfusion, and we therefore included this in our model as an independent variable [13]. Multiple variables relevant to spine surgery were examined, including type of curve etiology, Cobb angle, location of primary curve, number of levels fused, preoperative hematocrit, primary versus revision surgery, use of bone graft (autograft), need for lumbrosacral fusion, use of preoperative rh-EPO and need for transfusion. In addition, demographic factors, including age and gender, were included in our analysis. Analysis Univariate analyses were conducted, by transfusion status, to quantify the number of patients within each variable category. Table 1 shows the demographic and clinical characteristics of our patient population, with frequencies for categorical data and averages for continuous data. Bivariate analyses were performed between the dependent variable (transfusion status) and all 12 independent variables of interest (age, gender, type of scoliosis, rh-EPO treatment, need for revision surgery, surgical approach, use of bone graft, level of fusion, lumbrosacral fusion, preoperative hematocrit, degree of curvature and location of the curve). Table 2 shows the odds of transfusion, using crude odds ratios (ORs), for each category. The reference group in each variable category was randomly chosen and designated a crude OR of 1.0. It is called the crude OR, because this type of analyses did not control for possible confounding and/or interaction between the variables. Therefore, multivariate analyses were conducted next to eliminate the possibility that interrelated variables (eg, the degree of curvature, the decision to use erythropoietin, type of scoliosis) were obscuring the true effects of another variable in predicting transfusions in this population. In this study, logistic regression was performed to predict which variables were significant in determining trans-
168
M.G. Vitale et al. / The Spine Journal 2 (2002) 166–172
Table 1 Demographic and clinical characteristics of transfused and nontransfused Transfused (n 63) Age Gender Male Female Type of scoliosis Idiopathic Kyphosis Congenital Neurogenic Other Erythropoietin treatment Yes No Surgery as revision Yes No Surgical approach Anterior Posterior Combined Bone graft Yes No Level of fusion Lumbrosacral fusion Yes No Hematocrit level at admission Degree of curvature Location of curve Upper (cervical, cervicothoracic, thoracic locations) Lower (thoracolumbar, lumbar, lumbrosacral locations)
15.0 3.0
Table 2 The odds of transfusion (crude odds ratio)
Not transfused (n 227) 13.0 5.0
25 38
63 164
26 1 1 35 0
115 12 28 65 2
10 52
84 141
2 61
22 205
10 33 20
30 137 60
24 39 10.5 4.2
115 110 9.0 4.4
13 50 40.5 5.2 61.0 27
15 212 40.0 5.0 50.0 25.0
15
84
47
142
fusion need, controlling for all variables considered. A hierarchical backward elimination paradigm was used in this modeling procedure, such that interaction terms were simultaneously tested for their statistical significance through a log-likelihood-ratio test. Multicollinearity analysis of the model with interaction terms revealed that there was no gross collinearity among the independent variables. In stratified analysis, Mantel-Haenzel summary odds ratio for the rh-EPO treatment was calculated by stratifying other independent risk factors into various levels of exposure. The Breslow-Day test of significant variation in odds ratio between the strata was noted as an indicator for the presence of interaction. At the same time, a noticeable change in estimate of the Mantel-Haenzel summary odds ratio for rhEPO, when compared with the crude odds ratio, was noted as a possible sign of confounding. Finally, fitness of the data to our final model was assessed using the HosmerLemeshaw test. As shown in Table 3, the final logistical regression model consists of four clinical variables: rh-EPO treatment, type of scoliosis, degree of curvature and whether the child had lumbrosacral fusion.
Age (years) 3 3–10 11–19 20 Gender Male Female Type of scoliosis Idiopathic Kyphosis Congenital Neurogenic Erythropoietin treatment No Yes Surgery as revision No Yes Surgical approach Anterior Posterior Combined Bone graft No Yes Level of fusion (number of levels) 11 11 Lumbrosacral fusion No Yes Hematocrit level at admission 41 41 Degree of curvature (degrees) 45 45–89 90 Location of curve Upper Lower
Crude OR
95% Confidence interval
p value
1.00 1.532 3.329 7.118
— (0.07,31.67) (0.18,61.19) (0.33,156.0)
— .431 .223 .113
1.00 1.667
— (0.932,3.018
— .084
1.00 0.369 0.170 2.317
— (0.049,2.758) (0.027,1.053) (1.281,4.190)
— .331 .057 .005
1.00 0.324
— (0.16,0.70)
— .002
1.00 0.305
— (0.075,1.237)
— .096
1.00 0.682 0.983
— (0.304,1.552) (0.406,2.378)
— .356 .940
1.00 0.604
— (0.34,1.07)
— .068
1.00 1.433
— (0.816,2.52)
— .21
1.00 3.679
— (1.708,7.925)
— .001
1.00 1.157
— (0.66,2.05)
— .615
1.00 1.335 6.179
— (0.708,2.517) (2.4,15.91)
— .373 .001
1.00 2.005
— (1.05,3.84)
— .036
OR odds ratio.
Clinical decision rule A simple clinical decision rule was developed from the final logistic regression model, as shown in Table 4. A total of 24 clinical states were determined from all possible combinations of the four variables that were determined to be statistically significant from the final model: 1) preoperative rh-EPO, 2) type of scoliosis, 3) lumbrosacral fusion, and 4) degree of spine curvature. Lumbrosacral fusion was included in the model as an indicator for an increased number of levels fused or the use of instrumentation, both of which may necessitate longer pro-
M.G. Vitale et al. / The Spine Journal 2 (2002) 166–172 Table 3 Final logistical regression model
Variable Erythropoietin treatment No Yes Type of scoliosis Idiopathic Neurogenic Degree of curve* Lumbrosacral fusion No Yes
Adjusted OR
95% Confidence Interval
p value
1.00 0.290
— (0.135,0.624)
— .0015
1.00 2.020 1.015
— (1.069,3.819) (1.005,1.025)
— .03 .004
1.00 2.174
— (0.910,5.192)
— .08
*Estimate for one-degree curve. OR odds ratio.
cedures and, therefore, increase the risk for blood transfusion. Furthermore, lumbrosacral fusion was a significant independent predictor for transfusion, even after controlling for possible confounding. Degrees of curvature were assigned at 50, 90 and 125 degrees to simplify the effect of this variable in our model. The probability of receiving a transfusion was calculated for 23 of these states, using odds ratios to estimate the likelihood of needing a transfusion relative to the reference case. In our clinical model, the reference case (OR1.0) was chosen as a child not treated with preoperative rh-EPO, having idiopathic scoliosis, no lumbrosacral fusion and a spine curvature of 50 degrees. Based on a review of the recent experience of the senior author of this article, which is in agreement with other reports in the literature, we have quan-
169
tified the probability of transfusion at 10% for the reference case. All other scenarios were compared with this case, such that an OR less than 1.0 indicates a smaller likelihood of requiring a transfusion, whereas an OR greater than 1.0 indicates a greater likelihood of needing a transfusion, relative to the reference case. Results Descriptive data Table 1 illustrates the demographic and clinical characteristics of our patient population. Among the 290 patients with verifiable transfusion and rh-EPO records, 63 received transfusion (22%) and 227 did not (78%). The overall patient sample consisted of 30% males and 70% females, with an average age of 15 years (range, 12 to 18 years) among those who were transfused and 13 years (8 to 18 years) among those who were not. Forty-nine percent of all patients were diagnosed with idiopathic scoliosis, 34% with neurogenic scoliosis, and the remaining 17% were diagnosed with kyphosis, congenital or other types of scoliosis. In terms of erythropoietin treatment, 10 of 94 patients who received rh-EPO preoperatively underwent a transfusion (11%), whereas 52 of 193 patients who did not receive rhEPO (27%) needed a transfusion. For those requiring revision surgery, 8.3% were transfused, whereas 22% were transfused who underwent a primary spine surgery. Thirtythree of 170 patients who underwent a posterior approach received transfusion (19%), whereas 10 of 30 patients who were operated in an anterior fashion (33%) and 20 of 80 patients who underwent a combined approach were transfused
Table 4 Clinical decision model: algorithm for risk (odds ratio) of transfusion Erythropoietin treatment
Type of scoliosis
Lumbrosacral fusion
Degree of curve
Odds ratio
95% Confidence interval
No No Yes Yes No No Yes Yes No No Yes Yes No No Yes Yes No No Yes Yes No No Yes Yes
Idiopathic Idiopathic Idiopathic Idiopathic Idiopathic Idiopathic Idiopathic Idiopathic Idiopathic Idiopathic Idiopathic Idiopathic Neurogenic Neurogenic Neurogenic Neurogenic Neurogenic Neurogenic Neurogenic Neurogenic Neurogenic Neurogenic Neurogenic Neurogenic
No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes
50 50 50 50 90 90 90 90 125 125 125 125 50 50 50 50 90 90 90 90 125 125 125 125
1.0 2.2 0.3 0.6 1.8 3.9 0.5 1.1 3.0 6.4 0.9 1.9 2.0 4.4 0.6 1.3 3.6 7.8 1.0 2.3 6.0 13.0 1.7 3.8
— 0.9–5.20 0.14–0.62 0.19–2.13 1.19–2.65 1.46–10.27 0.23–1.15 0.32–3.92 1.41–6.21 1.99–20.75 0.33–2.25 0.47–7.37 1.07–3.82 1.79–10.79 0.22–1.54 0.38–4.29 1.77–7.30 2.97–20.61 0.40–2.73 0.67–7.71 2.40–14.94 4.13–40.73 0.59–5.11 0.99–14.24
170
M.G. Vitale et al. / The Spine Journal 2 (2002) 166–172
(25%). Twenty-four of 139 patients who received bone graft required transfusion (17%). Of the 149 patients who did not receive bone graft, 39 were transfused (26%). The number of levels fused was slightly greater among those who were transfused, with an average of 10.5 levels fused (range, 6.3to 14.7 levels), whereas the average was 9.0 levels fused among nontransfused patients (range, 4.6 to 13.4 levels). Thirteen of 28 patients who underwent lumbrosacral fusion were transfused (46%). On the other hand, only 50 of 212 patients who did not receive this procedure were transfused (19%). The hematocrit level at the time of admission was approximately the same in both the transfused patients (40.55.2) and nontransfused patients (40.05.0). The degree of curvature was greater among the transfused group, with an average of 61 degrees (range, 34 to 88 degrees). The average curvature among the nontransfused group was 50 degrees (range, 25 to 75 degrees). Regarding curve location (upper versus lower spine), whereas only 15 of 99 patients with upper curvature (cervical, cervicothoracic and thoracic locations) of the spine were transfused (15%), 47 of 189 patients with lower curvatures (thoracolumbar, lumbar and lumbrosacral locations) were transfused (25%). Bivariate analysis The OR method of comparison was used to estimate the likelihood of needing a transfusion relative to a reference group, as indicated in Table 2. Five of these factors indicated statistically significant ORs: type of scoliosis, rh-EPO treatment, lumbrosacral fusion, degree of curvature and location of curve. In terms of scoliosis diagnosed, patients with idiopathic scoliosis were assigned as the reference group. Each of the other types of scoliosis that were considered in this analysis (kyphosis, congenital, neurogenic) were compared with this reference group. Of these, only the neurogenic scoliosis group produced a significant OR compared with the idiopathic scoliosis group (OR, 2.317; 95% confidence interval [CI], 1.281 to 4.190). Therefore, by this analysis, patients with neurogenic scoliosis were 2.317 times more likely than patients with idiopathic scoliosis to undergo a transfusion. Erythropoietin was found to have a protective effect on transfusion rates in this population. In comparison to those who were not given rh-EPO treatment, children who were administered rh-EPO were 0.324 times less likely to undergo a transfusion (95% CI, 0.16 to 0.70). Children who had lumbrosacral fusion were 3.679 times more likely to be transfused (95% CI, 1.708 to 7.925) than those who did not undergo this procedure. Patients who had a curvature greater than 90 degrees were 6.179 times more likely to be transfused than those with curvature of less than 45 degrees (95% CI, 2.4 to 15.91). Patients with curves in the lower spine were 2.005 times more likely than those with upper spine curves to need a transfusion (95% CI, 1.05 to 3.84). Multivariate analysis In comparing Table 2 with Table 3, the same variables, except for “location of curve,” were determined to be signif-
icant for predicting transfusion, although the odds ratios differed. Based on multivariate analyses, those who received rh-EPO were still less likely to need a transfusion than those who were not given this drug (OR, 0.29; 95% CI, 0.135 to 0.624). For type of scoliosis, patients with neurogenic scoliosis were 2.02 times more likely than patients with idiopathic scoliosis to undergo a transfusion (95% CI, 1.069 to 3.819). The degree of curvature was estimated for every one degree increase in curvature (OR, 1.015/degree; 95% CI, 1.0005 to 1.025). Therefore, for each increase in the curvature degree, the odds of needing a transfusion increases by 1.015 times. Finally, patients who underwent lumbrosacral fusion were more likely to undergo a transfusion (OR, 2.174; 95% CI, 0.910 to 5.192). Note, that although the p value of .08 for this last variable was greater than .05, the analysis from Table 3 yielded a p value of .001, warranting its inclusion here. Algorithm for probability of transfusion While multivariate analysis can usefully predict the probability of transfusion for children with various risk factors, a simple clinical model integrating the relevant variables would enhance clinical decision making. In Table 4, we present an algorithm with 24 scenarios that includes each possible combination of the four significant factors derived from our regression model. The reference scenario for this algorithm is a child who was not given rh-EPO preoperatively, has idiopathic scoliosis, no lumbrosacral fusion performed and a 50-degree curvature of the spine (OR, 1.0). The other 23 scenarios were compared with this reference case. In comparison to the reference case, for example, a child differing only by undergoing preoperative rh-EPO treatment had the smallest odds ratio (OR, 0.3; 95% CI, 0.14 to 0.62). This means that a child with this background is 70% less likely to need a transfusion than the reference case. At the other end of the spectrum, a child with neurogenic scoliosis, with a 125-degree curve of the spine, who was not treated with rh-EPO preoperatively but who underwent a lumbrosacral fusion had the largest odds ratio (OR, 13.0; 95% CI, 4.13 to 40.73). This indicates that a child with this set of clinical characteristics is 13.0 times more likely than the reference case to undergo a transfusion. A general, but practical rule can be extrapolated from Table 4. When compared with the reference case, each variable difference accounts for a factor of two increase in the transfusion likelihood. This obviously does not apply when the erythropoietin treatment variable is different or “yes,” in this case. When preoperative rh-EPO is administered, the transfusion rate, in the least, is halved. For curves that are 125 degrees, the addition of rh-EPO reduces the transfusion rate by a factor of three. As an example, consider the patient with neurogenic scoliosis not treated preoperatively with rh-EPO who had undergone lumbrosacral fusion for a 90-degree curve. Here, three variables, except rh-EPO treatment, are different with respect to the reference case. Because each variable differ-
M.G. Vitale et al. / The Spine Journal 2 (2002) 166–172
ence accounts for a twofold increase in transfusion risk relative to the reference case, one would expect an eightfold increase in risk (222). Table 4 indicates a 7.8-fold increase. The estimate, although rough, is well within the confidence intervals. If this patient were given preoperative rh-EPO, the risk for transfusion would more than 0.5 to 2.3 instead of 7.8. Note that for a patient with a 125-degree curve, the relative increase in risk is 4 (or 22) when compared with the reference case. In summary, a roughly twofold increase in the risk of transfusion is introduced for each additional risk factor that is added into the clinical scenario, except for the case when preoperative erythropoietin is given. When rh-EPO is administered, the likelihood for transfusion at least halves. Although this “rule of two’s” does not precisely match the odds ratio, the numbers are very close, falling well within the confidence intervals, and should serve as an easy method of identifying patients at high risk for transfusion. Discussion This study was conducted in order to determine what factors are most predictive for transfusion in children undergoing multilevel spine surgery. According to our logistic regression model, four significant transfusion predictors were recognized as significant factors: 1) preoperative erythropoietin, 2) neurogenic type of scoliosis, 3) degree of curve and 4) lumbrosacral fusion. From the algorithm discussion above, a “rule of two’s” was derived. For each additional variable change, relative to the reference case (no preoperative rh-EPO, idiopathic scoliosis, no lumbrosacral fusion and 50-degree curvature), the risk for transfusion roughly doubles, except when preoperative erythropoietin is given, in which case it is at least halved. This practical rule will not only identify pediatric spine patients at increased risk for transfusion, and quantify such risk, but it will allow surgeons to take measures to avoid transfusion and its attendant risks (eg, viral transmission). In addition, surgeons will be better able to apply blood-saving measures in those scenarios most likely to cause excessive bleeding, thus maximizing cost-effectiveness. Cost can be significant when considering collection, production, storage, distribution and waste of this supply [30]. In both adults and children undergoing major orthopedic procedures, rh-EPO has been shown to be an effective adjunct for minimizing the need for transfusion [2,14– 16,18,31,32]. A recent study by Vitale et al. [13] showed that among children with idiopathic scoliosis, preoperative rh-EPO can lower the rate of transfusion during surgery. In the current study, the use of preoperative erythropoietin was also found to be protective against transfusion. In addition to the use of preoperative rh-EPO, other techniques, such as preoperative autologous donation, preoperative hemodilution, intraoperative and postoperative salvage of autologous blood and hypotensive anesthesia [9,21,23– 25] have been shown to have a protective effect in the pedi-
171
atric orthopedic population. Recently, Copley et al. [24] conducted a case-control study comparing two groups of children (n43 in each group) and concluded that hemodilution was a safe method to satisfy the perioperative transfusion requirements of adolescents undergoing extensive spinal surgery; hemodilution, cell saver and hypotensive anesthesia were the relevant variables. In our patient population, neurogenic scoliosis proved to be a significant predictor of transfusion. Previously, our group hypothesized that there was erythropoietin “resistance” among children with neurogenic scoliosis [13]. Because we controlled for confounding between these variables (type of scoliosis and preoperative rh-EPO), it is less likely that the association of neurogenic scoliosis to increased transfusion rates is the result of rh-EPO resistance alone. Further investigation will be needed to detect the underlying predisposition among children with neurogenic scoliosis toward increased risk for transfusion. Degree of curve and lumbrosacral fusion also proved to result in increased risk for transfusion in our cohort. Nuttall et al. [27] recently published a retrospective review of 244 adults who had spine surgery with instrumentation and fusion and were transfused. Predictors of allogeneic blood transfusions for adult patients undergoing different types of multilevel spine surgery included preoperative hemoglobin concentration, increased number of donated autologous red blood cells, pulmonary disease, surgery for tumor and greater number of spine levels fused. Greater number of levels fused has previously been shown to be associated with increased bleeding and allogeneic red blood cell use among patients with idiopathic scoliosis [28]. In addition, Johnson et al. [29], in a prospective analysis of 55 adult patients undergoing lumbar fusions, showed that increasing levels of fusion and/or instrumentation correlated with more blood loss, suggesting more safety measures against allogenic transfusion (cell saver and/or red blood cell predeposit) are required, the more extensive the surgery. In the present series, lumbrosacral fusion may be a surrogate marker for an increased number of levels fused or the use of instrumentation. Fusion at any level may be a marker of severity, as is degree of spine curvature, thus necessitating longer procedures. Previous studies have shown an increase in blood transfusions with duration of surgery [27, 28]. Whether children with neurogenic scoliosis, higher degrees of spine curvature and necessity for lumbrosacral fusion require more operative time and, thus, have an increased risk for transfusion is a question that cannot yet be answered completely. Although this study is the first step toward quantifying the true risk of transfusion in a pediatric orthopedic setting, it was retrospectively conducted. Because of the retrospective nature of this study, there are clear biases inherent in the study design, including those related to selection of cases and reporting of data. Therefore, a prospective randomized trial, including, but not limited to, operative time (eg, stratified by number of levels involved or necessity for instrumentation versus bone graft), other modes of avoiding trans-
172
M.G. Vitale et al. / The Spine Journal 2 (2002) 166–172
fusion (eg, rh-EPO versus hemodilution, hypotensive anesthesia, aprotinin [21, 33]), surgical technique, preoperative hematocrit and underlying comorbidities (eg, pulmonary disease) will need to be conducted. The results, used in a predictive model, can give the surgeon an objective and practical means of assessing the true risk of transfusion in the pediatric population undergoing multilevel spine surgery. Conclusions The risk for transfusion is related to multiple factors and can be quantitatively measured using a simple “rule of two’s” algorithm that incorporates 1) use of preoperative erythropoietin, 2) presence of neurogenic scoliosis, 3) degree of spine curvature and 4) need for lumbrosacral fusion. As suggested by other researchers in this field, a multifaceted, integrated approach may be necessary to reduce the risk for transfusion [26]. The current study develops this approach and extends the analysis by providing a useful quantitative rule to gauge transfusion risk. This rule provides a practical means for surgeons to avoid unnecessary transfusion, and its attendant risks, among pediatric patients undergoing multilevel spine surgery. Furthermore, our model provides the surgeon with a method of identifying those most at risk for transfusion, allowing optimization of blood-product allocation in the most cost-effective manner. A future prospective trial will help further elucidate the true predictors for transfusion. References [1] Cowell H, Swickard J. Autotransfusion in children’s orthopaedics. J Bone Joint Surg Am 1974;56:908–12. [2] Faris P, Ritter M, Abels R. The effects of recombinant human erythropoietin on perioperative transfusion requirements in patients having a major orthopaedic operation. J Bone Joint Surg Am 1996;78:62–72. [3] Klein H. Allogeneic transfusion risks in the surgical patient. Am J Surg 1995;170:21S–6S. [4] MacEwen G, Bennett E, Guille J. Autologous blood transfusions in children and young adults with low body weight undergoing spinal surgery. J Pediatr Orthop 1990;10:750–3. [5] Roye DJ, Rothstein P, Rickert J, Verdisco L, Farcy J. The use of preoperative erythropoietin in scoliosis surgery. Spine 1992;17:204–5. [6] Simpson M, Georgopoulos G, Orsini E, Eilert R. Autologous transfusions for orthopaedic procedures at a children’s hospital. J Bone Joint Surg Am 1992;74:652–8. [7] Lackritz E, Satten G, Aberle-Grasse J. Estimated risk of transmission of the human immunodeficiency virus by screened blood in the United States. N Engl J Med 1995;333:1721–5. [8] Schreiber G, Busch M, Kleinman S, Korelitz J. The risk of transfusion-transmitted viral infections: the retrovirus epidemiology donor study. N Engl J Med 1996;334:1685–90. [9] Lemos M, Healy W. Blood transfusion in orthopaedic operations. J Bone Joint Surg 1996;78A:1260–70. [10] Dodd R. The risk of transfusion transmitted infection [editorial]. N Engl J Med 1992;327:419–21. [11] Holland P. Transfusion transmitted infectious diseases. J Fla Med Assoc 1993;80:33–6. [12] Shulman G, Solanski D, Hadijipavlou A. Augmented autologous transfusions in major reconstructive spine surgery. J Clin Apheresis 1998;13:62–8.
[13] Vitale M, Stazzone E, Gelijns A, Moskowitz A, Roye DJ. The effectiveness of preoperative erythropoietin in averting allogenic blood transfusion among children undergoing scoliosis surgery. J Pediatr Orthop Part B 1998;7:203–9. [14] Birkmeyer J, Goodnough L, Aubuchon J. The cost-effectiveness of preoperative autologous blood donation in total hip and knee replacement. Transfusion 1993;33:544–51. [15] Goodnough L, Price T. The EPO study group: a phase III trial of recombinant human erythropoietin therapy in non-anemic orthopaedic patients subjected to aggressive autologous blood phlebotomy: dose, response, toxicity, and efficacy. Transfusion 1994;34:66. [16] Goodnough L, Rudnick S, Price T. Increased preoperative collection of autologous blood with recombinant human erythropoietin therapy. N Engl J Med 1989;321:1163–8. [17] Goodnough L, Verbrugge D, Marcus RE, Goldberg V. The effect of patient size and dose of recombinant human erythropoietin on red blood cell volume expansion in autologous blood donors for elective orthopaedic surgery. J Am Coll Surg 1994;179:171–6. [18] Rollo V, Hozack W, Rothman R, Chao W, Eng K. Prospective randomized evaluation of blood salvage techniques for primary total hip arthroplasty. J Arthroplasty 1995;10:53–9. [19] Sparling E, Nelson C, Lavender R, Smith J. The use of erythropoietin in the management of Jehova’s Witnesses who have revision total hip arthroplasty. J Bone Joint Surg 1996;78A:1548–52. [20] Keating E. Canadian orthopaedic perioperative erythropoietin study group: effectiveness of perioperative recombinant human erythropoietin in elective hip replacement. Lancet 1993;341:1227–32. [21] Keating E. Current options and approaches for blood management in orthopaedic surgery. In: Price CT, ed. American Academy of Orthopaedic Surgeons Instructional Course Lectures, vol. 48. Orlando: American Academy of Orthopaedic Surgeons, 1999:655–65. [22] Sharock N, Mineo R. The effect of two levels of hypotension on intraoperative blood loss during total hip arthroplasty performed under lumbar epidural anesthesia. Anesth Analg 1993;76:580–4. [23] Murray D, Forbes R, Titone M, Weinstein S. Transfusion management in pediatric and adolescent scoliosis surgery: efficacy of autologous blood. Spine 1997;22:2735–40. [24] Copley L, Richads B, Safavi F, Newton P. Hemodilution as a method to reduce transfusion requirements in adolescent spine fusion surgery. Spine 1999;24:219–22. [25] Tate DJ, Friedman R. Blood conservation in spinal surgery. Review of current techniques. Spine 1992;17:1450–6. [26] Nuttall G, Horlocker T, Santrach P, Oliver WC, Dekutoski M, Bryant S. Predictors of blood transfusions in spinal instrumentation and fusion surgery. Spine 2000;25:596–601. [27] Bostman O, Hyrkas J, Hirvensalo E, Kallio E. Blood loss, operating time, and positioning of the patient in lumbar disc surgery. Spine 1990;15:360–3. [28] Guay J, Haig M, Lortie L, Guertin M, Poitras B. Predicting blood loss in surgery for idiopathic scoliosis. Can J Anaesth 1994;41:775–81. [29] Johnson R, Murphy M, Miller M. Fusions and transfusions: an analysis of blood loss and autologous replacement during lumbar fusions. Spine 1989;14:358–62. [30] Tretiak R, Laupacis A, Riviere M, McKerracher K, Souetre E, Group BR. Cost of allogeneic and autologous blood transfusion in Canada. Canadian cost of transfusion study group. Can Med Assoc J 1996;154:1501–8. [31] Goodnough L, Vizmeg K, Robecks R. Prevalence and classification of anemia in elective orthopaedic surgery patients: implications for blood conservation programs. Vox Sang 1992;63:90–5. [32] Goodnough L, Vizmeg K, Verbrugge D. The impact of autologous blood ordering and blood procurement practices in allogeneic blood exposure in elective orthopaedic surgery patients. Am J Clin Pathol 1994;101:354–7. [33] Bidstrup B, Royston D, Sapsford R, Taylor K. Reduction in blood loss and blood use after cardiopulmonary bypass with high dose aprotinin (Trasylol). J Thorac Cardiovasc Surg 1989;97:364–71.
Clinical Studies
Quantifying risk of transfusion in children undergoing spine surgery Michael G. Vitale, MD, MPHa,c,*, Douglas E. Levy, MPHa, Maxwell C. Park, MDb, Hyunok Choi, MPHa, Julie C. Choe, MPHa, David P. Roye, Jr., MDc a
International Center for Health Outcomes and Innovation Research, College of Physicians and Surgeons, and The Joseph L. Mailman School of Public Health, Columbia University and New York Presbyterian Hospital, 600 West 168th Street, 7th Floor, New York, NY, 10032 USA b New York Orthopaedic Hospital, New York Presbyterian Medical Center, 622 West 168th Street, 11th Floor, New York, NY, 10032 USA c Division of Pediatric Orthopaedics, Department of Orthopaedic Surgery, Columbia University College of Physicians and Surgeons, 3959 Broadway, 8th Floor North New York, NY, 10032 USA Received 22 September 2001; accepted 7 February 2002
Abstract
Background context: The risks and costs of transfusion are a great concern in the area of pediatric spine surgery, because it is a blood-intensive procedure with a high risk for transfusion. Therefore, determining the predictors of transfusion in this patient population is an important first step and has the potential to improve upon the current approaches to reducing transfusion rates. In this study, we reveal several predictors of transfusion in a pediatric patient population undergoing spine surgery. In turn, we present a general rule of thumb (“rule of two’s”) for gauging transfusion risk, thus enhancing the surgeon’s approach to avoiding transfusion in certain clinical scenarios. Purpose: This study was conducted to determine the main factors of transfusion in a population of pediatric patients undergoing scoliosis surgery. The goal was to present an algorithm for quantifying the true risk of transfusion for various patient groups that would highlight patients “at high risk” for transfusion. This is especially important in light of the various risks associated with undergoing a transfusion, as well as the costs involved in maintaining and disposing of exogenous blood materials. Study design/setting: This is a retrospective review of a group of children who underwent scoliosis surgery between 1988 and 1995 at an academic institution. Patient sample: A total of 290 patients were analyzed in this study, of which 63 were transfused and 227 were not. Outcome measures: No outcomes measures were used in this study. Methods: A retrospective review of 290 patients presenting to our institution for scoliosis surgery was conducted, with a focus on socioclinical data related to transfusion risk. Univariate analysis and logistic regression were used to quantify the determinants of transfusion risk. Results: Univariate analysis identified many factors that were associated with the risk of transfusion. However, it is clear that several of these factors are dependent on each other, obscuring the true issues driving transfusion need. We used multivariate analysis to control for the various univariate predictors of transfusion. Our logistic regression model suggested that the type of scoliosis (odds ratio [OR], 2.02; 95% confidence interval [CI], 1.07 to 3.82), degree of curvature (OR, 1.012/degree curve; 95% CI, 1.01 to 1.03), and use of erythropoietin (OR, 0.29; 95% CI, 0.14 to 0.62) were the main determinants of transfusion risk for our population. Conclusions: The main risk factors of transfusion were used to formulate a simple algorithm, which can be used to quantify transfusion risk and to guide efforts to avoid transfusion in children undergoing spinal surgery. Given a 10% baseline risk for transfusion, our “rule of two’s” indicates that each risk factor approximately doubles the chance of transfusion, whereas the administration of recombinant human erythropoietin roughly halves the risk of transfusion. © 2002 Elsevier Science Inc. All rights reserved.
Keywords:
Transfusion; Erythropoietin; Pediatrics; Scoliosis; Clinical decision rule; Probability of transfusion; Rule of two’s
FDA device/drug status: approved for this indication (erythropietin). Support in whole or in part was received from Ortho Biotech. The author (MV) acknowledges a financial relationship (grant research support from Ortho Biotech), which may indirectly relate to the subject of this manuscript. * Corresponding author. 600 West 168th Street, 7th Floor, New York, NY 10032, USA. Tel: (212) 305-5028; fax: (212) 305-4256. E-mail address: [email protected] (M.G. Vitale).
Introduction Despite efforts to avoid allogenic transfusion, children undergoing spine surgery often require exogenous blood products (Rothstein P, Jr., Roye DP, Verdisco L, Experience with preoperative administration of erythropoietin in
1529-9430/02/$ – see front matter © 2002 Elsevier Science Inc. All rights reserved. PII: S1529-9430(02)00 1 7 4 - 2
M.G. Vitale et al. / The Spine Journal 2 (2002) 166–172
adolescents undergoing scoliosis surgery, personal communication, 1997; [1–6]). Concerns over the safety of the blood supply have resulted in a dramatic decrease in the risk of transmission of viruses, including human immunodeficiency virus (HIV) and hepatitis. Nevertheless, it is unlikely that screening strategies will be able to decrease this risk to zero [3,5,7–11]. Furthermore, several studies have highlighted other potential negative effects of transfusion, including ABO mismatch as a result of labeling errors and adverse immunological effects, such as allergic reaction, graft-versus-host disease and alloimmunization [12]. For these reasons, there has been an evolution of techniques and strategies to avoid allogenic transfusion. Measures that can be taken to decrease the possibility of transfusion include the use of preoperative erythropoietin, preoperative autologous donation, preoperative hemodilution, intraoperative salvage or postoperative salvage of autologous blood, intraoperative normovolemic hemodilution and hypotensive anesthesia [2,13–25]. Although many of these techniques have been shown to effectively decrease transfusion requirements, the appropriate indications for use of these costly strategies have not been well described. Clearly, efforts to avert transfusion are more important and more cost effective in situations where transfusion is more likely. The decision to use cell saver, for example, is often based on the clinician’s implicit assessment regarding the risk of blood loss and risk of transfusion in a particular clinical setting. The setting can be predetermined by the patient (eg, the type of scoliosis or degree of curvature) and influenced by the physician (eg, use of cell saver or preoperative erythropoietin). An improved understanding of the specific factors and the quantitative effect that each of these factors has on transfusion risk can help guide the clinician in making the most informed decision possible in this setting. To date, there is a great need to better understand the factors that drive transfusion rates in children undergoing spinal surgery. Few studies have specifically examined the relationship between socioclinical factors and transfusion risk in this patient population [26–29]. In a recent retrospective review of 244 adult spine patients who underwent spinal instrumentation and fusion, Nuttall et al. [26] identified preoperative hemoglobin concentration, increased number of donated autologous red blood cells, pulmonary disease, surgery for tumor and greater number of spine levels fused as significant predictors for allogenic transfusion need. This group concluded that multiple factors determine the need for transfusion, suggesting that the approach to reducing this need should be multifaceted, yet integrated. Similarly, we analyzed the factors significant in affecting need for transfusion in the pediatric population undergoing multilevel spine surgery. However, we extended the analysis by producing a quantitative model, thus providing a practical means of gauging transfusion risk. Ideally, this will enhance the approach to preventing unnecessary transfusions in pediatric spine surgery cases. This current review
167
is the first such effort to quantitatively model the relative effect of various clinical factors on transfusion rates in a pediatric orthopedic population. Materials and methods Data This study was a retrospective review of the medical records of pediatric patients who underwent spine surgery with the principal investigator. Of 420 patients initially identified, records were available for 317 patients (75%). Furthermore, among these 317 patients, 290 had verifiable transfusion and recombinant human erythropoietin (rh-EPO) records. This limited the sample to 63 individuals who received intraoperative transfusion (defined as “transfused”) and 227 individuals who did not (defined as “nontransfused”). During the period of this review, we began to use rh-EPO routinely in an effort to avert transfusion. This has proven to exert a strong protective effect against transfusion, and we therefore included this in our model as an independent variable [13]. Multiple variables relevant to spine surgery were examined, including type of curve etiology, Cobb angle, location of primary curve, number of levels fused, preoperative hematocrit, primary versus revision surgery, use of bone graft (autograft), need for lumbrosacral fusion, use of preoperative rh-EPO and need for transfusion. In addition, demographic factors, including age and gender, were included in our analysis. Analysis Univariate analyses were conducted, by transfusion status, to quantify the number of patients within each variable category. Table 1 shows the demographic and clinical characteristics of our patient population, with frequencies for categorical data and averages for continuous data. Bivariate analyses were performed between the dependent variable (transfusion status) and all 12 independent variables of interest (age, gender, type of scoliosis, rh-EPO treatment, need for revision surgery, surgical approach, use of bone graft, level of fusion, lumbrosacral fusion, preoperative hematocrit, degree of curvature and location of the curve). Table 2 shows the odds of transfusion, using crude odds ratios (ORs), for each category. The reference group in each variable category was randomly chosen and designated a crude OR of 1.0. It is called the crude OR, because this type of analyses did not control for possible confounding and/or interaction between the variables. Therefore, multivariate analyses were conducted next to eliminate the possibility that interrelated variables (eg, the degree of curvature, the decision to use erythropoietin, type of scoliosis) were obscuring the true effects of another variable in predicting transfusions in this population. In this study, logistic regression was performed to predict which variables were significant in determining trans-
168
M.G. Vitale et al. / The Spine Journal 2 (2002) 166–172
Table 1 Demographic and clinical characteristics of transfused and nontransfused Transfused (n 63) Age Gender Male Female Type of scoliosis Idiopathic Kyphosis Congenital Neurogenic Other Erythropoietin treatment Yes No Surgery as revision Yes No Surgical approach Anterior Posterior Combined Bone graft Yes No Level of fusion Lumbrosacral fusion Yes No Hematocrit level at admission Degree of curvature Location of curve Upper (cervical, cervicothoracic, thoracic locations) Lower (thoracolumbar, lumbar, lumbrosacral locations)
15.0 3.0
Table 2 The odds of transfusion (crude odds ratio)
Not transfused (n 227) 13.0 5.0
25 38
63 164
26 1 1 35 0
115 12 28 65 2
10 52
84 141
2 61
22 205
10 33 20
30 137 60
24 39 10.5 4.2
115 110 9.0 4.4
13 50 40.5 5.2 61.0 27
15 212 40.0 5.0 50.0 25.0
15
84
47
142
fusion need, controlling for all variables considered. A hierarchical backward elimination paradigm was used in this modeling procedure, such that interaction terms were simultaneously tested for their statistical significance through a log-likelihood-ratio test. Multicollinearity analysis of the model with interaction terms revealed that there was no gross collinearity among the independent variables. In stratified analysis, Mantel-Haenzel summary odds ratio for the rh-EPO treatment was calculated by stratifying other independent risk factors into various levels of exposure. The Breslow-Day test of significant variation in odds ratio between the strata was noted as an indicator for the presence of interaction. At the same time, a noticeable change in estimate of the Mantel-Haenzel summary odds ratio for rhEPO, when compared with the crude odds ratio, was noted as a possible sign of confounding. Finally, fitness of the data to our final model was assessed using the HosmerLemeshaw test. As shown in Table 3, the final logistical regression model consists of four clinical variables: rh-EPO treatment, type of scoliosis, degree of curvature and whether the child had lumbrosacral fusion.
Age (years) 3 3–10 11–19 20 Gender Male Female Type of scoliosis Idiopathic Kyphosis Congenital Neurogenic Erythropoietin treatment No Yes Surgery as revision No Yes Surgical approach Anterior Posterior Combined Bone graft No Yes Level of fusion (number of levels) 11 11 Lumbrosacral fusion No Yes Hematocrit level at admission 41 41 Degree of curvature (degrees) 45 45–89 90 Location of curve Upper Lower
Crude OR
95% Confidence interval
p value
1.00 1.532 3.329 7.118
— (0.07,31.67) (0.18,61.19) (0.33,156.0)
— .431 .223 .113
1.00 1.667
— (0.932,3.018
— .084
1.00 0.369 0.170 2.317
— (0.049,2.758) (0.027,1.053) (1.281,4.190)
— .331 .057 .005
1.00 0.324
— (0.16,0.70)
— .002
1.00 0.305
— (0.075,1.237)
— .096
1.00 0.682 0.983
— (0.304,1.552) (0.406,2.378)
— .356 .940
1.00 0.604
— (0.34,1.07)
— .068
1.00 1.433
— (0.816,2.52)
— .21
1.00 3.679
— (1.708,7.925)
— .001
1.00 1.157
— (0.66,2.05)
— .615
1.00 1.335 6.179
— (0.708,2.517) (2.4,15.91)
— .373 .001
1.00 2.005
— (1.05,3.84)
— .036
OR odds ratio.
Clinical decision rule A simple clinical decision rule was developed from the final logistic regression model, as shown in Table 4. A total of 24 clinical states were determined from all possible combinations of the four variables that were determined to be statistically significant from the final model: 1) preoperative rh-EPO, 2) type of scoliosis, 3) lumbrosacral fusion, and 4) degree of spine curvature. Lumbrosacral fusion was included in the model as an indicator for an increased number of levels fused or the use of instrumentation, both of which may necessitate longer pro-
M.G. Vitale et al. / The Spine Journal 2 (2002) 166–172 Table 3 Final logistical regression model
Variable Erythropoietin treatment No Yes Type of scoliosis Idiopathic Neurogenic Degree of curve* Lumbrosacral fusion No Yes
Adjusted OR
95% Confidence Interval
p value
1.00 0.290
— (0.135,0.624)
— .0015
1.00 2.020 1.015
— (1.069,3.819) (1.005,1.025)
— .03 .004
1.00 2.174
— (0.910,5.192)
— .08
*Estimate for one-degree curve. OR odds ratio.
cedures and, therefore, increase the risk for blood transfusion. Furthermore, lumbrosacral fusion was a significant independent predictor for transfusion, even after controlling for possible confounding. Degrees of curvature were assigned at 50, 90 and 125 degrees to simplify the effect of this variable in our model. The probability of receiving a transfusion was calculated for 23 of these states, using odds ratios to estimate the likelihood of needing a transfusion relative to the reference case. In our clinical model, the reference case (OR1.0) was chosen as a child not treated with preoperative rh-EPO, having idiopathic scoliosis, no lumbrosacral fusion and a spine curvature of 50 degrees. Based on a review of the recent experience of the senior author of this article, which is in agreement with other reports in the literature, we have quan-
169
tified the probability of transfusion at 10% for the reference case. All other scenarios were compared with this case, such that an OR less than 1.0 indicates a smaller likelihood of requiring a transfusion, whereas an OR greater than 1.0 indicates a greater likelihood of needing a transfusion, relative to the reference case. Results Descriptive data Table 1 illustrates the demographic and clinical characteristics of our patient population. Among the 290 patients with verifiable transfusion and rh-EPO records, 63 received transfusion (22%) and 227 did not (78%). The overall patient sample consisted of 30% males and 70% females, with an average age of 15 years (range, 12 to 18 years) among those who were transfused and 13 years (8 to 18 years) among those who were not. Forty-nine percent of all patients were diagnosed with idiopathic scoliosis, 34% with neurogenic scoliosis, and the remaining 17% were diagnosed with kyphosis, congenital or other types of scoliosis. In terms of erythropoietin treatment, 10 of 94 patients who received rh-EPO preoperatively underwent a transfusion (11%), whereas 52 of 193 patients who did not receive rhEPO (27%) needed a transfusion. For those requiring revision surgery, 8.3% were transfused, whereas 22% were transfused who underwent a primary spine surgery. Thirtythree of 170 patients who underwent a posterior approach received transfusion (19%), whereas 10 of 30 patients who were operated in an anterior fashion (33%) and 20 of 80 patients who underwent a combined approach were transfused
Table 4 Clinical decision model: algorithm for risk (odds ratio) of transfusion Erythropoietin treatment
Type of scoliosis
Lumbrosacral fusion
Degree of curve
Odds ratio
95% Confidence interval
No No Yes Yes No No Yes Yes No No Yes Yes No No Yes Yes No No Yes Yes No No Yes Yes
Idiopathic Idiopathic Idiopathic Idiopathic Idiopathic Idiopathic Idiopathic Idiopathic Idiopathic Idiopathic Idiopathic Idiopathic Neurogenic Neurogenic Neurogenic Neurogenic Neurogenic Neurogenic Neurogenic Neurogenic Neurogenic Neurogenic Neurogenic Neurogenic
No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes
50 50 50 50 90 90 90 90 125 125 125 125 50 50 50 50 90 90 90 90 125 125 125 125
1.0 2.2 0.3 0.6 1.8 3.9 0.5 1.1 3.0 6.4 0.9 1.9 2.0 4.4 0.6 1.3 3.6 7.8 1.0 2.3 6.0 13.0 1.7 3.8
— 0.9–5.20 0.14–0.62 0.19–2.13 1.19–2.65 1.46–10.27 0.23–1.15 0.32–3.92 1.41–6.21 1.99–20.75 0.33–2.25 0.47–7.37 1.07–3.82 1.79–10.79 0.22–1.54 0.38–4.29 1.77–7.30 2.97–20.61 0.40–2.73 0.67–7.71 2.40–14.94 4.13–40.73 0.59–5.11 0.99–14.24
170
M.G. Vitale et al. / The Spine Journal 2 (2002) 166–172
(25%). Twenty-four of 139 patients who received bone graft required transfusion (17%). Of the 149 patients who did not receive bone graft, 39 were transfused (26%). The number of levels fused was slightly greater among those who were transfused, with an average of 10.5 levels fused (range, 6.3to 14.7 levels), whereas the average was 9.0 levels fused among nontransfused patients (range, 4.6 to 13.4 levels). Thirteen of 28 patients who underwent lumbrosacral fusion were transfused (46%). On the other hand, only 50 of 212 patients who did not receive this procedure were transfused (19%). The hematocrit level at the time of admission was approximately the same in both the transfused patients (40.55.2) and nontransfused patients (40.05.0). The degree of curvature was greater among the transfused group, with an average of 61 degrees (range, 34 to 88 degrees). The average curvature among the nontransfused group was 50 degrees (range, 25 to 75 degrees). Regarding curve location (upper versus lower spine), whereas only 15 of 99 patients with upper curvature (cervical, cervicothoracic and thoracic locations) of the spine were transfused (15%), 47 of 189 patients with lower curvatures (thoracolumbar, lumbar and lumbrosacral locations) were transfused (25%). Bivariate analysis The OR method of comparison was used to estimate the likelihood of needing a transfusion relative to a reference group, as indicated in Table 2. Five of these factors indicated statistically significant ORs: type of scoliosis, rh-EPO treatment, lumbrosacral fusion, degree of curvature and location of curve. In terms of scoliosis diagnosed, patients with idiopathic scoliosis were assigned as the reference group. Each of the other types of scoliosis that were considered in this analysis (kyphosis, congenital, neurogenic) were compared with this reference group. Of these, only the neurogenic scoliosis group produced a significant OR compared with the idiopathic scoliosis group (OR, 2.317; 95% confidence interval [CI], 1.281 to 4.190). Therefore, by this analysis, patients with neurogenic scoliosis were 2.317 times more likely than patients with idiopathic scoliosis to undergo a transfusion. Erythropoietin was found to have a protective effect on transfusion rates in this population. In comparison to those who were not given rh-EPO treatment, children who were administered rh-EPO were 0.324 times less likely to undergo a transfusion (95% CI, 0.16 to 0.70). Children who had lumbrosacral fusion were 3.679 times more likely to be transfused (95% CI, 1.708 to 7.925) than those who did not undergo this procedure. Patients who had a curvature greater than 90 degrees were 6.179 times more likely to be transfused than those with curvature of less than 45 degrees (95% CI, 2.4 to 15.91). Patients with curves in the lower spine were 2.005 times more likely than those with upper spine curves to need a transfusion (95% CI, 1.05 to 3.84). Multivariate analysis In comparing Table 2 with Table 3, the same variables, except for “location of curve,” were determined to be signif-
icant for predicting transfusion, although the odds ratios differed. Based on multivariate analyses, those who received rh-EPO were still less likely to need a transfusion than those who were not given this drug (OR, 0.29; 95% CI, 0.135 to 0.624). For type of scoliosis, patients with neurogenic scoliosis were 2.02 times more likely than patients with idiopathic scoliosis to undergo a transfusion (95% CI, 1.069 to 3.819). The degree of curvature was estimated for every one degree increase in curvature (OR, 1.015/degree; 95% CI, 1.0005 to 1.025). Therefore, for each increase in the curvature degree, the odds of needing a transfusion increases by 1.015 times. Finally, patients who underwent lumbrosacral fusion were more likely to undergo a transfusion (OR, 2.174; 95% CI, 0.910 to 5.192). Note, that although the p value of .08 for this last variable was greater than .05, the analysis from Table 3 yielded a p value of .001, warranting its inclusion here. Algorithm for probability of transfusion While multivariate analysis can usefully predict the probability of transfusion for children with various risk factors, a simple clinical model integrating the relevant variables would enhance clinical decision making. In Table 4, we present an algorithm with 24 scenarios that includes each possible combination of the four significant factors derived from our regression model. The reference scenario for this algorithm is a child who was not given rh-EPO preoperatively, has idiopathic scoliosis, no lumbrosacral fusion performed and a 50-degree curvature of the spine (OR, 1.0). The other 23 scenarios were compared with this reference case. In comparison to the reference case, for example, a child differing only by undergoing preoperative rh-EPO treatment had the smallest odds ratio (OR, 0.3; 95% CI, 0.14 to 0.62). This means that a child with this background is 70% less likely to need a transfusion than the reference case. At the other end of the spectrum, a child with neurogenic scoliosis, with a 125-degree curve of the spine, who was not treated with rh-EPO preoperatively but who underwent a lumbrosacral fusion had the largest odds ratio (OR, 13.0; 95% CI, 4.13 to 40.73). This indicates that a child with this set of clinical characteristics is 13.0 times more likely than the reference case to undergo a transfusion. A general, but practical rule can be extrapolated from Table 4. When compared with the reference case, each variable difference accounts for a factor of two increase in the transfusion likelihood. This obviously does not apply when the erythropoietin treatment variable is different or “yes,” in this case. When preoperative rh-EPO is administered, the transfusion rate, in the least, is halved. For curves that are 125 degrees, the addition of rh-EPO reduces the transfusion rate by a factor of three. As an example, consider the patient with neurogenic scoliosis not treated preoperatively with rh-EPO who had undergone lumbrosacral fusion for a 90-degree curve. Here, three variables, except rh-EPO treatment, are different with respect to the reference case. Because each variable differ-
M.G. Vitale et al. / The Spine Journal 2 (2002) 166–172
ence accounts for a twofold increase in transfusion risk relative to the reference case, one would expect an eightfold increase in risk (222). Table 4 indicates a 7.8-fold increase. The estimate, although rough, is well within the confidence intervals. If this patient were given preoperative rh-EPO, the risk for transfusion would more than 0.5 to 2.3 instead of 7.8. Note that for a patient with a 125-degree curve, the relative increase in risk is 4 (or 22) when compared with the reference case. In summary, a roughly twofold increase in the risk of transfusion is introduced for each additional risk factor that is added into the clinical scenario, except for the case when preoperative erythropoietin is given. When rh-EPO is administered, the likelihood for transfusion at least halves. Although this “rule of two’s” does not precisely match the odds ratio, the numbers are very close, falling well within the confidence intervals, and should serve as an easy method of identifying patients at high risk for transfusion. Discussion This study was conducted in order to determine what factors are most predictive for transfusion in children undergoing multilevel spine surgery. According to our logistic regression model, four significant transfusion predictors were recognized as significant factors: 1) preoperative erythropoietin, 2) neurogenic type of scoliosis, 3) degree of curve and 4) lumbrosacral fusion. From the algorithm discussion above, a “rule of two’s” was derived. For each additional variable change, relative to the reference case (no preoperative rh-EPO, idiopathic scoliosis, no lumbrosacral fusion and 50-degree curvature), the risk for transfusion roughly doubles, except when preoperative erythropoietin is given, in which case it is at least halved. This practical rule will not only identify pediatric spine patients at increased risk for transfusion, and quantify such risk, but it will allow surgeons to take measures to avoid transfusion and its attendant risks (eg, viral transmission). In addition, surgeons will be better able to apply blood-saving measures in those scenarios most likely to cause excessive bleeding, thus maximizing cost-effectiveness. Cost can be significant when considering collection, production, storage, distribution and waste of this supply [30]. In both adults and children undergoing major orthopedic procedures, rh-EPO has been shown to be an effective adjunct for minimizing the need for transfusion [2,14– 16,18,31,32]. A recent study by Vitale et al. [13] showed that among children with idiopathic scoliosis, preoperative rh-EPO can lower the rate of transfusion during surgery. In the current study, the use of preoperative erythropoietin was also found to be protective against transfusion. In addition to the use of preoperative rh-EPO, other techniques, such as preoperative autologous donation, preoperative hemodilution, intraoperative and postoperative salvage of autologous blood and hypotensive anesthesia [9,21,23– 25] have been shown to have a protective effect in the pedi-
171
atric orthopedic population. Recently, Copley et al. [24] conducted a case-control study comparing two groups of children (n43 in each group) and concluded that hemodilution was a safe method to satisfy the perioperative transfusion requirements of adolescents undergoing extensive spinal surgery; hemodilution, cell saver and hypotensive anesthesia were the relevant variables. In our patient population, neurogenic scoliosis proved to be a significant predictor of transfusion. Previously, our group hypothesized that there was erythropoietin “resistance” among children with neurogenic scoliosis [13]. Because we controlled for confounding between these variables (type of scoliosis and preoperative rh-EPO), it is less likely that the association of neurogenic scoliosis to increased transfusion rates is the result of rh-EPO resistance alone. Further investigation will be needed to detect the underlying predisposition among children with neurogenic scoliosis toward increased risk for transfusion. Degree of curve and lumbrosacral fusion also proved to result in increased risk for transfusion in our cohort. Nuttall et al. [27] recently published a retrospective review of 244 adults who had spine surgery with instrumentation and fusion and were transfused. Predictors of allogeneic blood transfusions for adult patients undergoing different types of multilevel spine surgery included preoperative hemoglobin concentration, increased number of donated autologous red blood cells, pulmonary disease, surgery for tumor and greater number of spine levels fused. Greater number of levels fused has previously been shown to be associated with increased bleeding and allogeneic red blood cell use among patients with idiopathic scoliosis [28]. In addition, Johnson et al. [29], in a prospective analysis of 55 adult patients undergoing lumbar fusions, showed that increasing levels of fusion and/or instrumentation correlated with more blood loss, suggesting more safety measures against allogenic transfusion (cell saver and/or red blood cell predeposit) are required, the more extensive the surgery. In the present series, lumbrosacral fusion may be a surrogate marker for an increased number of levels fused or the use of instrumentation. Fusion at any level may be a marker of severity, as is degree of spine curvature, thus necessitating longer procedures. Previous studies have shown an increase in blood transfusions with duration of surgery [27, 28]. Whether children with neurogenic scoliosis, higher degrees of spine curvature and necessity for lumbrosacral fusion require more operative time and, thus, have an increased risk for transfusion is a question that cannot yet be answered completely. Although this study is the first step toward quantifying the true risk of transfusion in a pediatric orthopedic setting, it was retrospectively conducted. Because of the retrospective nature of this study, there are clear biases inherent in the study design, including those related to selection of cases and reporting of data. Therefore, a prospective randomized trial, including, but not limited to, operative time (eg, stratified by number of levels involved or necessity for instrumentation versus bone graft), other modes of avoiding trans-
172
M.G. Vitale et al. / The Spine Journal 2 (2002) 166–172
fusion (eg, rh-EPO versus hemodilution, hypotensive anesthesia, aprotinin [21, 33]), surgical technique, preoperative hematocrit and underlying comorbidities (eg, pulmonary disease) will need to be conducted. The results, used in a predictive model, can give the surgeon an objective and practical means of assessing the true risk of transfusion in the pediatric population undergoing multilevel spine surgery. Conclusions The risk for transfusion is related to multiple factors and can be quantitatively measured using a simple “rule of two’s” algorithm that incorporates 1) use of preoperative erythropoietin, 2) presence of neurogenic scoliosis, 3) degree of spine curvature and 4) need for lumbrosacral fusion. As suggested by other researchers in this field, a multifaceted, integrated approach may be necessary to reduce the risk for transfusion [26]. The current study develops this approach and extends the analysis by providing a useful quantitative rule to gauge transfusion risk. This rule provides a practical means for surgeons to avoid unnecessary transfusion, and its attendant risks, among pediatric patients undergoing multilevel spine surgery. Furthermore, our model provides the surgeon with a method of identifying those most at risk for transfusion, allowing optimization of blood-product allocation in the most cost-effective manner. A future prospective trial will help further elucidate the true predictors for transfusion. References [1] Cowell H, Swickard J. Autotransfusion in children’s orthopaedics. J Bone Joint Surg Am 1974;56:908–12. [2] Faris P, Ritter M, Abels R. The effects of recombinant human erythropoietin on perioperative transfusion requirements in patients having a major orthopaedic operation. J Bone Joint Surg Am 1996;78:62–72. [3] Klein H. Allogeneic transfusion risks in the surgical patient. Am J Surg 1995;170:21S–6S. [4] MacEwen G, Bennett E, Guille J. Autologous blood transfusions in children and young adults with low body weight undergoing spinal surgery. J Pediatr Orthop 1990;10:750–3. [5] Roye DJ, Rothstein P, Rickert J, Verdisco L, Farcy J. The use of preoperative erythropoietin in scoliosis surgery. Spine 1992;17:204–5. [6] Simpson M, Georgopoulos G, Orsini E, Eilert R. Autologous transfusions for orthopaedic procedures at a children’s hospital. J Bone Joint Surg Am 1992;74:652–8. [7] Lackritz E, Satten G, Aberle-Grasse J. Estimated risk of transmission of the human immunodeficiency virus by screened blood in the United States. N Engl J Med 1995;333:1721–5. [8] Schreiber G, Busch M, Kleinman S, Korelitz J. The risk of transfusion-transmitted viral infections: the retrovirus epidemiology donor study. N Engl J Med 1996;334:1685–90. [9] Lemos M, Healy W. Blood transfusion in orthopaedic operations. J Bone Joint Surg 1996;78A:1260–70. [10] Dodd R. The risk of transfusion transmitted infection [editorial]. N Engl J Med 1992;327:419–21. [11] Holland P. Transfusion transmitted infectious diseases. J Fla Med Assoc 1993;80:33–6. [12] Shulman G, Solanski D, Hadijipavlou A. Augmented autologous transfusions in major reconstructive spine surgery. J Clin Apheresis 1998;13:62–8.
[13] Vitale M, Stazzone E, Gelijns A, Moskowitz A, Roye DJ. The effectiveness of preoperative erythropoietin in averting allogenic blood transfusion among children undergoing scoliosis surgery. J Pediatr Orthop Part B 1998;7:203–9. [14] Birkmeyer J, Goodnough L, Aubuchon J. The cost-effectiveness of preoperative autologous blood donation in total hip and knee replacement. Transfusion 1993;33:544–51. [15] Goodnough L, Price T. The EPO study group: a phase III trial of recombinant human erythropoietin therapy in non-anemic orthopaedic patients subjected to aggressive autologous blood phlebotomy: dose, response, toxicity, and efficacy. Transfusion 1994;34:66. [16] Goodnough L, Rudnick S, Price T. Increased preoperative collection of autologous blood with recombinant human erythropoietin therapy. N Engl J Med 1989;321:1163–8. [17] Goodnough L, Verbrugge D, Marcus RE, Goldberg V. The effect of patient size and dose of recombinant human erythropoietin on red blood cell volume expansion in autologous blood donors for elective orthopaedic surgery. J Am Coll Surg 1994;179:171–6. [18] Rollo V, Hozack W, Rothman R, Chao W, Eng K. Prospective randomized evaluation of blood salvage techniques for primary total hip arthroplasty. J Arthroplasty 1995;10:53–9. [19] Sparling E, Nelson C, Lavender R, Smith J. The use of erythropoietin in the management of Jehova’s Witnesses who have revision total hip arthroplasty. J Bone Joint Surg 1996;78A:1548–52. [20] Keating E. Canadian orthopaedic perioperative erythropoietin study group: effectiveness of perioperative recombinant human erythropoietin in elective hip replacement. Lancet 1993;341:1227–32. [21] Keating E. Current options and approaches for blood management in orthopaedic surgery. In: Price CT, ed. American Academy of Orthopaedic Surgeons Instructional Course Lectures, vol. 48. Orlando: American Academy of Orthopaedic Surgeons, 1999:655–65. [22] Sharock N, Mineo R. The effect of two levels of hypotension on intraoperative blood loss during total hip arthroplasty performed under lumbar epidural anesthesia. Anesth Analg 1993;76:580–4. [23] Murray D, Forbes R, Titone M, Weinstein S. Transfusion management in pediatric and adolescent scoliosis surgery: efficacy of autologous blood. Spine 1997;22:2735–40. [24] Copley L, Richads B, Safavi F, Newton P. Hemodilution as a method to reduce transfusion requirements in adolescent spine fusion surgery. Spine 1999;24:219–22. [25] Tate DJ, Friedman R. Blood conservation in spinal surgery. Review of current techniques. Spine 1992;17:1450–6. [26] Nuttall G, Horlocker T, Santrach P, Oliver WC, Dekutoski M, Bryant S. Predictors of blood transfusions in spinal instrumentation and fusion surgery. Spine 2000;25:596–601. [27] Bostman O, Hyrkas J, Hirvensalo E, Kallio E. Blood loss, operating time, and positioning of the patient in lumbar disc surgery. Spine 1990;15:360–3. [28] Guay J, Haig M, Lortie L, Guertin M, Poitras B. Predicting blood loss in surgery for idiopathic scoliosis. Can J Anaesth 1994;41:775–81. [29] Johnson R, Murphy M, Miller M. Fusions and transfusions: an analysis of blood loss and autologous replacement during lumbar fusions. Spine 1989;14:358–62. [30] Tretiak R, Laupacis A, Riviere M, McKerracher K, Souetre E, Group BR. Cost of allogeneic and autologous blood transfusion in Canada. Canadian cost of transfusion study group. Can Med Assoc J 1996;154:1501–8. [31] Goodnough L, Vizmeg K, Robecks R. Prevalence and classification of anemia in elective orthopaedic surgery patients: implications for blood conservation programs. Vox Sang 1992;63:90–5. [32] Goodnough L, Vizmeg K, Verbrugge D. The impact of autologous blood ordering and blood procurement practices in allogeneic blood exposure in elective orthopaedic surgery patients. Am J Clin Pathol 1994;101:354–7. [33] Bidstrup B, Royston D, Sapsford R, Taylor K. Reduction in blood loss and blood use after cardiopulmonary bypass with high dose aprotinin (Trasylol). J Thorac Cardiovasc Surg 1989;97:364–71.