Risk Factors for 30-day Postoperative Surgical Site Hematoma Requiring Evacuation After Resection of Brain Metastases

Study design

A retrospective study of patients who underwent surgical resection of brain metastases at our institution between March 2011 and December 2019 was conducted. Institutional ethics approval was obtained from the institutional review board prior to study initiation (National Healthcare Group Domain Specific Review Board; Reference Number 2020/00358). A waiver of informed consent was granted as this study posed no more than minimal risk to participants.

Cohort selection

The operating theatre records database was accessed to retrieve the National Registration Identity Card numbers (NRICs, which act as the national identification number) of patients who underwent surgical resection of a brain tumor. The electronic medical records of the patients were then accessed and screened for inclusion in the study. Patients who (1) had a histologically verified metastasis to the brain, and (2) were 18 years old or above on the day of surgical resection were included.

Data collection

Clinical data were collected using a standardized data collection form. Variables collected include (1) demographics, including age, sex, ethnicity, preoperative height, weight, body mass index (BMI), smoking history, (2) medical history, including hypertension, hyperlipidemia, diabetes mellitus, history of stroke or transient ischemic attack, (3) details on the primary cancer, including presence of extracranial metastases preoperatively and site of the primary tumor (defined as the 5 commonest sites and others), (4) details on the brain metastasis, including presence of preoperative intratumoral hemorrhage (as reported by the radiologist in the radiology report), volume of the largest resected tumor on preoperative MRI (defined as AP x ML x CC/2, where AP, ML, and CC denote the greatest anteroposterior, medial-lateral, and craniocaudal diameters), presence of preoperative hydrocephalus, extent of midline shift, extent of resection (gross total or subtotal resection), presence of residual tumor on the first postoperative MRI as reported by the radiologist, (5) results of the preoperative full-blood count closest to the surgical resection, including white blood cell count, red blood cell count, hemoglobin level, mean corpuscular volume, hematocrit, platelet count, neutrophil count, lymphocyte count, monocyte count, eosinophil count, basophil count, (6) results of the preoperative clotting studies closest to the surgical resection, including the prothrombin time (PT), activated partial thromboplastin time (aPTT), INR, and (7) the ABO blood type of the patient.

Statistical analysis

The baseline characteristics of the patients were reported using mean and standard deviation for continuous variables and count numbers and percentages for categorical variables. For hypothesis testing against the occurrence of a 30-day POH, the Pearson’s χ2 test was used for categorical variables, and the student’s t-test was used for continuous variables. The Fisher’s exact test was used for variables with a sample size of lesser than 5 under any category. Statistical significance was defined as a p-value of 0.050 or lower.

Variables with a p-value of lesser than 0.200 on univariate analysis against the occurrence of a 30-day POH were included as independent variables in a multivariate logistic regression model, with the occurrence of a 30-day POH as the dependent variable. The independent variables that had a p-value of 0.050 or lower on multivariate logistic regression analysis were considered as risk factors for a 30-day POH. For the risk factors that were continuous variables, the threshold value that optimally discriminated a 30-day POH occurrence from non-occurrence was derived using receiver operating characteristic analysis.

To determine whether a 30-day POH was associated with overall mortality, univariate time-to-event analysis using the Kaplan-Meier method was performed. Hypothesis testing was performed using the log-rank test, and a p-value of 0.050 or lower was taken to be statistically significant. For patients who died during the follow-up period, time-to-mortality was defined as the duration between the first surgical resection of the brain metastases and the date of death. For patients who did not die during the follow-up period, time-to-mortality was defined as the duration between the first surgical resection of the brain metastases and the date of the latest follow-up. All data analyses were conducted using R Studio Version 1.2.5042.

Baseline characteristics of the study population

The baseline characteristics of the study population were reported in Table 1. A total of 158 patients were included in the analysis. The mean (SD) age of the study population was 59.3 (12.0) years, and 82 (53.2%) patients were female. The incidence of a 30-day POH was 8.2% (13 patients). The median overall survival of patients with and without a 30-day POH were 11.3 and 12.4 months respectively. There was no statistically significant association between the occurrence of a 30-day POH and overall mortality (p=0.100) (Figure 1).

Risk factors for a 30-day POH

On multivariate analysis, younger age (OR=0.91; 95% CI=0.83, 0.99; p=0.035), higher BMI (OR=1.61; 95% CI=1.16, 2.46; p=0.010), and blood type AB (OR=21.7; 95% CI=1.66, 552; p=0.031) were independently associated with a higher risk of a 30-day POH (Table 2).

On receiver operating characteristic analysis, a threshold BMI of 25.1 kg/m² gave the optimum balance of sensitivity (80.0%) and specificity (76.6%) in predicting the occurrence of a 30-day POH (AUC=0.74; 95% CI=0.56, 0.92). On the other hand, a threshold age of 57 gave the optimum balance of sensitivity (61.5%) and specificity (66.2%) in predicting the occurrence of a POH (AUC=0.66; 95% CI=0.52, 0.79).