Pediatric patients undergoing complex cranial vault reconstruction (CCVR) are at risk of significant blood loss and perioperative transfusion. Minimizing blood loss and blood product transfusion is an important goal given the risks and costs of allogeneic blood transfusion.1,2,3,4,5,6 Numerous blood management strategies have been employed in craniosynostosis surgery, including antifibrinolytics, cell saver, preoperative erythropoietin, acute preoperative normovolemic hemodilution, and the use of restrictive transfusion protocols. The strength of evidence supporting the efficacy and safety of these techniques varies, while knowledge of the impact of other patient and surgical factors on perioperative transfusion, such as patient age, American Society of Anesthesiologists (ASA) physical status, surgical duration, or intraoperative complications, is generally lacking.7

In 2011, a consortium of international collaborators from the United States and Canada formed the Pediatric Craniofacial Collaborative Group under the auspices of the Society for Pediatric Anesthesia. This group then established the Pediatric Craniofacial Surgery Perioperative Registry (PCSPR), which has become the largest pediatric craniofacial surgery registry worldwide. It was designed to collect data pertaining to the perioperative course, management, and outcomes in this patient population. A further goal was to foster multicentre research and quality improvement of pediatric anesthesia for reconstructive craniofacial surgery.

We conducted this study to evaluate the relationships between demographic and perioperative variables and transfusion outcomes of CCVR in the multicentre PCSPR; we have not previously examined variables associated with transfusion outcomes using this database. Three outcomes were evaluated: intraoperative red blood cell-containing product (RBC-CP) transfusion, total perioperative blood donor exposures, and the occurrence of an allogeneic transfusion-free hospitalization course. We hypothesized that patient-related variables (e.g., age, weight, ASA physical status), procedure-related variables (procedure type, surgical duration, redo operation) and specific blood management techniques (use of antifibrinolytics, cell saver, and transfusion protocols) would be independently associated with these transfusion outcomes. The primary aim of this study was to report independent predictors of the transfusion outcomes mentioned.

Methods

We searched the PCSPR for subjects undergoing CCVR. Complex cranial vault reconstruction was defined as fronto-orbital advancement/anterior cranial vault reconstruction, middle/posterior cranial vault reconstruction, or total cranial vault reconstruction. All procedures involved a craniotomy. Neuro-endoscopic procedures, spring-mediated cranioplasties, and modified Pi procedures were excluded. Data extracted included demographic and surgical data, fluid and transfusion data, perioperative management data, and complications. The study analysis plan was submitted to the Pediatric Craniofacial Collaborative Group scientific review committee and approved prior to the conduct of the analysis. Appropriate institutional review board (IRB)/research ethics committee approval was obtained at each study site as well as written informed consent when required; at some study sites, the requirement for written informed consent was waived by the IRB. The study was conducted in accordance with STROBE guidelines.8

Pediatric craniofacial surgery perioperative registry

Participating institutions began entering data following local institutional ethical review board or equivalent approval. Study data were collected and managed using REDCap electronic data capture tools hosted at the Children’s Hospital of Philadelphia.9 Data entry began on June 25, 2012. A de-identified dataset with data through to September 30, 2016 was analyzed. Individual institutions were represented using an institution code and were not identifiable.

Data collection, entry, and validation

Thirty-three institutions contributed data over the study time period. Sites were required to provide quarterly reports quantifying case capture rates to avoid reporting bias. The median (25th, 75th interquartile range) reported case capture rate across all participating institutions was 100% (98%, 100%). All participating institutions were required to report data collection and auditing processes to ensure accuracy.10 Perioperative blood management strategies captured by the registry included the preoperative use of erythropoietin, the intraoperative use of antifibrinolytics (tranexamic acid or aminocaproic acid), deliberate hypotension, cell saver utilization, and acute preoperative normovolemic hemodilution. Intraoperative and postoperative transfusion protocols were also used.

In addition to the above, the study’s principal investigator at the Data Coordinating Center audited data prior to analysis. This included scrutinizing cases for omissions of critical data (e.g., age, weight) and identifying outlier data. Cross-validation of data within individual records was performed to optimize data accuracy. Queries based on omissions, outliers, and discrepancies identified through this process were aggregated and sent to site investigators for rectification.

Outcomes and statistical analysis

Demographic data and data describing perioperative management were analyzed using descriptive statistics. Our primary objective was to use univariable and multivariable analysis to identify independent factors associated with three transfusion outcomes: total perioperative blood donor exposures, total volume of intraoperative RBC-CPs, and the occurrence of a transfusion-free hospital course. Red blood cell-containing products included packed red blood cells (PRBCs), whole blood, and reconstituted blood composed of PRBCs and fresh frozen plasma (FFP).11 Hemostatic blood products included non-reconstituted FFP, cryoprecipitate, and platelets.

Independent variables in the analyses included age, weight, sex, race, ethnicity, ASA physical status, preoperative tracheostomy presence, procedure category, cranial distractor placement, history of prior craniofacial surgery, preoperative evidence of elevated intracranial pressure, preoperative hematocrit, use of deliberate hypotension, use of intraoperative cell saver, antifibrinolytic administration (tranexamic acid or aminocaproic acid), use of a standardized protocol to guide either intraoperative or postoperative red cell transfusion, diagnosis of syndromic craniosynostosis or craniosynostosis with three or more sutures involved, duration of surgery, intraoperative administration of a vasopressor infusion, occurrence of intraoperative hypothermia (nadir temperature of < 35°C), hypothermia on arrival to the intensive care unit or recovery room (< 35°C), and whether or not one or more major intraoperative cardiorespiratory complications occurred. Major intraoperative cardiorespiratory complications were specific itemized data elements on the registry case report form that were identified prior to analysis. These complications included administration of an intravenous epinephrine bolus, cardiac arrest (defined as any case in which chest compressions were initiated), suspected venous air embolism with cardiovascular collapse, dural venous sinus bleed, suspected transfusion reaction, pulmonary edema, and re-intubation in the operating room.

Data measured on a continuous scale were presented as mean (standard deviation [SD]) or median [interquartile range (IQR)]; categorical variables were presented as frequencies and percentages. Univariable and multivariable analyses were performed for variables potentially associated with the three outcomes described above. For univariable analyses of intraoperative RBC-CP volume and total perioperative blood donor exposure outcomes, the Wilcoxon-Mann-Whitney test was used for dichotomous independent variables and the Kruskal-Wallis test was used for multi-level independent variables. Independent variables with a univariable significance of P < 0.1 were included in the multivariable model. Because three outcomes were evaluated, a corrected P value of < 0.0167 (0.05/3) was considered statistically significant for the multivariable analysis. For the volume of intraoperative RBC-CP and total perioperative blood donor exposure outcomes, multivariable analysis was performed using a generalized linear mixed effect negative binomial model accounting for the effect of institution with a random effect. For the transfusion-free hospital course outcome, the Chi-square test or Fisher’s exact test (when the expected counts were less than 5) was used in the univariable analysis. Logistic regression was used in the multivariable analysis for this binary outcome. For the multivariable analyses, case wise deletion was employed in cases of missing data (available as Electronic Supplementary Material; eTable). Statistical analysis was done with the statistical package SAS version 9.4 (SAS Institute, Cary, NC, USA).

Results

Demographics

Search of the PCSPR yielded 1,814 subjects from 33 institutions; 75% (1,377) of these were under 24 months of age (infant group). Detailed demographic data are presented in Table 1. Single-suture metopic and sagittal craniosynostosis were the most common diagnoses. Anterior cranial vault reconstruction/fronto-orbital advancements accounted for greater than 60% of cases. Craniofacial reoperation accounted for 20% of cases for all ages and 11% of cases in the infant group.

Table 1 Patient demographic data
Full size table

Intraoperative blood product transfusion

Intraoperative transfusion data for specific blood products are presented in Table 2. One thousand five hundred and ninety-two (88%) of CCVR patients were transfused a mean (SD) volume of 29.7 (25.9) mL·kg−1 of allogeneic RBC-CP intraoperatively. Fourteen percent of all patients received an autologous erythrocyte transfusion from intraoperative cell saver (similar in both age groups). With respect to hemostatic blood products, discrete FFP transfusion was the most common, being administered to 22% of all patients. Platelets and cryoprecipitate were transfused relatively infrequently (2% and 3% of cases, respectively).

Table 2 Intraoperative transfusion data
Full size table

Intraoperative RBC-containing product transfusion outcome

Univariable and multivariable analyses of selected independent variables and their relationship to intraoperative RBC-CP transfusion are presented in Table 3, with variables organized into three categories: patient variables, procedure variables, and blood conservation variables. Patient variables independently associated with increasing intraoperative RBC-CP transfusion included younger age, decreasing weight, and increasing ASA physical status. Procedure-related variables that were associated with increasing intraoperative RBC-CP transfusion in multivariable analysis included procedures not involving distractor placement, longer surgical duration, use of a vasoactive infusion intraoperatively, and the occurrence of a major intraoperative cardiorespiratory complication. Lack of antifibrinolytic administration was the only blood conservation variable independently associated with higher RBC-CP transfusion.

Table 3 Univariable and multivariable analysis for the outcome “intraoperative red blood cell-containing product transfusion (mL·kg−1)” among all age patients
Full size table

Blood product donor exposure outcome

Univariable and multivariable analyses for variables associated with total perioperative blood donor exposures are shown in Table 4. Younger patient age and the presence of a tracheostomy were the only patient variables that independently correlated with increased donor exposures. Procedural variables that were associated with increasing blood donor exposures in multivariable analysis included procedures not involving distractor placement, longer duration of surgery, vasoactive infusion administration, and total cranial vault reconstruction (vs mid/posterior cranial vault reconstruction). None of the blood conservation variables examined were independently associated with this outcome.

Table 4 Univariable and multivariable analyses for the outcome “total perioperative blood donor exposures” among patients of all ages
Full size table

Transfusion-free hospital course outcome

Univariable and multivariable analyses for the binary transfusion-free hospital course outcome are presented in Table 5. Patient-related variables independently associated with this outcome included age greater than 24 months (odds ratio [OR], 4.5; 95% confidence interval [CI], 2.2 to 9.14), weight greater than 11 kg (OR, 3.5; 95% CI, 1.6 to 7.9), ASA physical status III-IV (OR, 0.42; 95% CI, 0.25 to 0.72), and preoperative hematocrit greater than 36 (OR, 2.7; 95% CI, 1.7 to 4.2). Procedure variables that were significantly associated with the occurrence of a transfusion-free course in multivariable analysis included shorter duration of surgery, while total cranial vault reconstruction was associated with a decreased likelihood of this outcome (OR, 0.32; 95% CI, 0.14 to 0.75, vs mid/posterior cranial vault reconstruction). With respect to blood conservation variables, intraoperative red cell transfusion was associated with a lower likelihood of a transfusion-free course (OR, 0.10; 95% CI, 0.05 to 0.21), while intraoperative cell saver use (OR, 4.62; 95% CI, 2.8 to 7.7) and postoperative transfusion protocols (OR, 2.23; 95% CI, 1.3 to 3.8) were independently predictive of a transfusion-free hospital course.

Table 5 Univariable and multivariable analyses for the binary outcome “transfusion-free hospital course (yes/no)” among patients of all ages
Full size table

Discussion

Significant variability exists in the amount of bleeding and transfusion in pediatric CCVR, ranging from massive intraoperative transfusion to a transfusion-free hospitalization.10 We evaluated patient, surgical, and blood conservation variables and three perioperative transfusion outcomes in a large multicentre pediatric CCVR dataset to identify independent predictors of transfusion outcomes. The vast majority (88%) of patients were transfused with a mean RBC-CPs transfusion volume of 30 mL·kg−1. In multivariable analysis, patient age and surgical duration were the only independent variables associated with all three outcomes. Conversely, prior craniofacial surgery, elevated intracranial pressure, syndromic or multi-suture craniosynostosis, intraoperative hypothermia, sex, race, and ethnicity were not associated with these outcomes.

Intraoperative RBC-containing product transfusion

We found that younger age, decreasing weight, and higher ASA physical status were each independently associated with increased intraoperative RBC-CP transfusion. This inverse relationship between age and weight with transfusion is consistent with prior reports.12,13,14,15 The association between higher ASA physical status and increased transfusion has not been previously reported in CCVR. While age and weight are not directly modifiable, our data suggest the possibility that when appropriate, performing surgery in older children may be a strategy to reduce transfusion. Our findings suggest that increased transfusion requirements might be anticipated in younger, smaller, and more medically complex patients.

Surgical variables associated with increased intraoperative transfusion included longer operations, surgeries without distractor placement, and the occurrence of a major intraoperative cardiorespiratory event or intraoperative vasoactive infusion administration. The correlation of distractor placement with less transfusion is consistent with the less complex nature of these procedures. The association of longer operations with increased transfusion has been described.12,13,14 The underpinnings of this relationship are unclear; for example, it is unknown whether it relates to procedure complexity, or if the operations are longer because more time is required to achieve hemostasis.

Intraoperatively, both the occurrence of a major cardiorespiratory complication and administration of a vasoactive infusion were associated with increased transfusion. We suspect that this association was observed because patients with more hemorrhage (and therefore transfusion) were more likely to develop hypotension requiring a vasoactive infusion and/or suffer a related complication.

Antifibrinolytic administration was the only blood conservation variable associated with decreased intraoperative transfusion. Antifibrinolytic administration is the only readily modifiable variable we studied that was associated with transfusion requirements. In the context of the evidence supporting antifibrinolytic efficacy and safety in CCVR, our findings support more widespread use of antifibrinolytic agents in these children.16,17,18,19,20

Perioperative blood donor exposures

Because CCVR may entail postoperative transfusion, we studied total blood donor exposures to capture the entire perioperative course. Additionally, this outcome captures hemostatic blood products not included in the intraoperative RBC-CP outcome. Demographic variables associated with increased blood donor exposures included age over 24 months and preoperative tracheostomy. It is likely that tracheostomy presence is indicative of increased patient complexity in this outcome; similar to increased ASA status. Indeed, greater ASA status was associated with increased blood donor exposure in univariable analysis but fell out following multivariable analysis. Lower weight did not predict blood donor exposures despite its association with higher intraoperative transfusion volumes. This makes sense given that a standard unit of PRBC (i.e., one blood donor exposure) represents a progressively increasing volume (in mL·kg−1) as patient weight decreases.

Similar to the intraoperative RBC-CP transfusion outcome findings, longer operations and vasoactive infusion administration were associated with increased perioperative donor exposures. Total cranial vault reconstruction was associated with more donor exposures, an expected finding given the greater extent and complexity of these procedures. Interestingly, none of the blood conservation variables were associated with a greater number of blood donor exposures, suggesting that the effects of patient and procedural variables predominated.

Transfusion-free hospital course

Achieving a safe, transfusion-free hospital course is a laudable goal. We found that increasing age, increasing body weight, and higher preoperative hematocrit were each associated with an increased likelihood of a transfusion-free hospitalization. Children with a preoperative hematocrit above the median of 35% were nearly three times more likely to avoid transfusion. This highlights the potential value of optimizing preoperative hematocrit and treating anemia when present; a concept supported elsewhere in the adult and pediatric perioperative literature.21,22,23,24

More medically complex patients (ASA physical status III or IV) were less likely to experience a transfusion-free hospital course, similar to data previously reported for pediatric surgical patients.25 The only procedural variable associated with a transfusion-free hospitalization was shorter duration of surgery, consistent with our other findings.

Use of cell saver and transfusion protocols were predictive of a transfusion-free hospitalization. Interestingly, despite evidence of efficacy, transfusion protocols are underutilized.10 Counterintuitively, intraoperative transfusion protocols were associated with a reduced likelihood of a transfusion-free course. This may be because we did not account for the effect of institution in this outcome, as centres with greater transfusion may also have implemented intraoperative transfusion protocols.

Limitations

Limitations of this study include those inherent to clinical databases including inaccurate or missing data.26 Associations and not cause and effect relationships can be established with observational data. Additionally, the registry may not capture all management strategies that have an impact on outcomes. While the effect of institution was accounted for in the RBC-CP transfusion and blood donor exposure analyses, it was not accounted for in the transfusion-free hospitalization outcome because it is a binary variable with a low frequency of occurrence. For this outcome, the effects of variables for which there was little within-institutional heterogeneity could not be separated from institution-specific outcomes. Finally, the number of patients who experienced a transfusion-free hospitalization was small and so conclusions should be drawn cautiously regarding this outcome.

Conclusion

In this large, multicentre dataset we identified significant predictors of relevant transfusion outcomes in CCVR. The significant demographic variables may help physicians preoperatively identify patients with greater transfusion requirements. Patient variables independently associated with increasing blood transfusion included younger age, decreasing weight, and increasing ASA physical status. Procedure-related variables that were associated with increasing blood transfusion included the more complex total cranial vault remodeling procedures, longer surgical duration, use of a vasoactive infusion intraoperatively, and the occurrence of a major intraoperative cardiorespiratory complication. Lack of antifibrinolytic administration was the only blood conservation variable independently associated with higher blood transfusion. Our findings of increased blood donor exposures for longer procedures and total cranial vault reconstruction and reduced transfusion with procedures involving distractor application provide information to be considered in surgical planning. Some of the identified predictors are directly modifiable and can serve as substrates for future investigation of improvement efforts.