Postoperative Pneumocephalus as a Predictor of Recurrence in Chronic Subdural Hematoma: A Propensity Score - Matched Retrospective Observational Study

Summary

This study was undertaken to systematically evaluate the relationship between postoperative pneumocephalus and recurrence in patients undergoing burr-hole drainage for chronic subdural hematoma (CSDH).

Abstract

Chronic subdural hematoma (CSDH) is a common neurosurgical condition with a high recurrence rate after burr‑hole drainage. Postoperative pneumocephalus is frequently observed, but its relationship with recurrence remains unclear. This retrospective observational study evaluated whether postoperative pneumocephalus volume predicts CSDH recurrence. Patients who underwent first‑time burr‑hole drainage for CSDH between January 2021 and June 2025 were included and classified according to head CT findings within 24 h after surgery. Propensity score matching (PSM; 1:1, caliper 0.02) balanced baseline characteristics (age, sex, hematoma side, antiplatelet/anticoagulant use), yielding 150 matched patients (75 per group). The primary outcome was ipsilateral hematoma recurrence requiring reoperation within 6 months. Pneumocephalus volume was measured using the Tada formula, with inter‑observer reliability assessed by the intraclass correlation coefficient (ICC = 0.964, 95% CI: 0.941–0.978). The recurrence rate was significantly higher in the pneumocephalus group than in the non‑pneumocephalus group (22.7% vs. 5.3%; OR = 5.20, 95% CI: 1.66–16.32; p = 0.002). Multivariate logistic regression showed that the presence of pneumocephalus was independently associated with recurrence (OR = 3.26, 95% CI: 1.45–7.30, p = 0.004), and each 1 mL increase in volume was associated with a 9% higher risk (OR = 1.09, 95% CI: 1.02–1.15, p = 0.008). ROC analysis yielded an AUC of 0.754 (95% CI: 0.645–0.864), with an optimal cut‑off of 12.5 mL (sensitivity 76.2%, specificity 70.5%). Bootstrap internal validation confirmed stability (mean cut‑off 12.6 mL, optimism‑corrected AUC 0.745). In conclusion, postoperative pneumocephalus volume is associated with CSDH recurrence after burr‑hole drainage and may help identify patients needing closer follow‑up; however, prospective multicenter validation is required before any specific volume threshold can be adopted in clinical practice.

Introduction

Chronic subdural hematoma (CSDH) is a common neurosurgical condition whose incidence has been increasing with the aging population^1,2. Its pathophysiology is thought to involve tearing of bridging veins after mild head trauma, leading to blood accumulation in the subdural space, local inflammation, and eventual formation of a hematoma capsule^3,4. Because of repeated microbleeding and fibrinolysis hyperactivity, most patients cannot absorb the hematoma spontaneously and require surgical intervention^5,6. Single‑ or double‑hole burr‑hole drainage is the first-line treatment due to its minimal invasiveness and proven efficacy^7,8. However, postoperative recurrence remains a major challenge, with reported rates ranging from 5% to 30%^9,10.

Numerous risk factors for recurrence have been identified, including older age, antiplatelet or anticoagulant use, bilateral hematoma, and imaging features such as mixed density or septation on preoperative CT^11,12. These factors suggest that recurrence is multifactorial, involving coagulation status, hematoma structure, and the postoperative healing environment^4,13. In postoperative imaging, pneumocephalus—accumulation of air in the cranial cavity—is frequently observed, mainly due to residual irrigation fluid and communication between the drainage system and the external environment^14,15. However, its clinical significance remains debated. Some studies suggest that small amounts of pneumocephalus are benign and resolve spontaneously^16,17; while others report that large volumes may delay brain re‑expansion and predispose to recurrence^16,18. These conflicting findings likely stem from small sample sizes, inadequate confounding control, or a lack of quantitative volumetric analysis¹⁹.

The optimal management of postoperative pneumocephalus is also uncertain, with practice varying between active positioning to promote gas expulsion and more conservative approaches. Clarifying the relationship between pneumocephalus volume and recurrence would allow early postoperative CT to guide risk stratification and follow‑up intensity. Against this background, the current study was undertaken to systematically evaluate the relationship between postoperative pneumocephalus and recurrence in patients undergoing burr-hole drainage for CSDH. This study incorporates several methodological features that build upon prior investigations, including propensity score matching to minimize baseline confounding, quantitative volume measurement to explore a potential threshold, and inclusion of the neutrophil-to-lymphocyte ratio (NLR) to assess inflammatory pathways. In addition, bootstrap internal validation and sensitivity analyses using Firth penalized and mixed-effects logistic regression were applied to address overfitting, rare events, and potential clustering. These analytical choices were intended to provide a controlled local replication of the known association and to generate hypotheses for future prospective studies.

Protocol

Ethical statement
The study protocol was reviewed and approved by the Sanming First Hospital Affiliated to Fujian Medical University Ethics Committee (No. 2026-38), and the investigation was conducted in accordance with the ethical standards set forth in the Declaration of Helsinki. As this is a retrospective observational study without any intervention, the collected data were generated during routine clinical diagnosis and treatment, without adding any additional burden or risk to patients’ examinations. Therefore, an application for exemption from informed consent was submitted, and all patients’ personal information has been kept strictly confidential. The research results will be used only for academic publication and will not disclose any information that can identify an individual. All materials and software used in this study are listed in the Table of Materials. For items marked “generic” or “standard hospital supply,” specific catalog numbers are not applicable because these are routine consumables available from multiple suppliers. The CT scanner model and software versions are specified to ensure reproducibility. For manual calculations (e.g., the Tada formula) and standard statistical tests performed in SPSS/R, no separate catalog number is required.

Research subjects
A single-center retrospective observational design was adopted for this study. Patients with CSDH who received burr-hole drainage in the neurosurgery department from January 1, 2021, to June 30, 2025, constituted the study population. After applying strict inclusion and exclusion criteria, data from qualified patients were retrieved and prepared for subsequent statistical analysis. To evaluate the relationship between postoperative pneumocephalus and hematoma recurrence, and to mitigate confounding bias, PSM was applied to achieve baseline balance in the original cohort. Propensity scores were derived from a logistic regression model with postoperative pneumocephalus as the dependent variable, adjusted for age, sex, hematoma side, and history of antiplatelet or anticoagulant medication. These covariates were selected based on their known associations with either CSDH recurrence or postoperative pneumocephalus, as documented in prior studies. The matching method used was nearest-neighbor matching with a cutoff of 0.02, and a 1:1 non-replacement matching was performed. After matching, a total of 150 patients were included (75 in the pneumocephalus group and 75 in the non-pneumocephalus group), as shown in Figure 1.

Exclusion and inclusion criteria
Inclusion criteria^20,21: (1) age 18 years and above; (2) confirmed diagnosis of unilateral or bilateral CSDH through head CT or MRI examination; (3) underwent the first burr-hole drainage surgery in the hospital; (4) had a re-examination of head CT within 24 h after the surgery and the imaging data were completely preserved; (5) complete postoperative clinical follow-up data were available, and the follow-up period was no less than 6 months.

Exclusion criteria^22,23: (1) patients with intracranial tumors, aneurysms, or arteriovenous malformations and other intracranial space-occupying or vascular diseases; (2) acute or subacute subdural hematoma (defined as the time from injury to surgery being less than 14 days); (3) those with a history of ipsilateral burr hole drainage or craniotomy; (4) those who underwent reoperation due to non-hematoma recurrence after surgery; (5) those with missing key variables in clinical or imaging data, making analysis impossible. Patients with bilateral hematomas who undergo concurrent bilateral burr hole drainage are included as research subjects, with each patient contributing a single observation. Bilateral status (unilateral vs. bilateral) is recorded as a covariate and is included in the propensity score matching and regression analyses. No patient contributed more than one data point, thereby preserving statistical independence of observations.

Handling of bilateral hematomas
Patients with bilateral hematomas who underwent concurrent bilateral burr hole drainage were included as research subjects, with each patient contributing a single observation. Bilateral status (unilateral vs. bilateral) was recorded as a covariate and included in the propensity score matching and regression analyses. No patient contributed more than one data point, thereby preserving statistical independence of observations. To further address any potential residual clustering, a sensitivity analysis using mixed‑effects logistic regression with a random intercept for each patient was performed (see Statistical methods).

Research plan
Data were extracted from the hospital medical record system and the imaging archive system, including: (1) general demographic characteristics (age and gender); (2) clinical data (history of hypertension, diabetes, antiplatelet/anticoagulant use, preoperative Markwalder neurological function grading score); (3) imaging data (hematoma side, CT density type, preoperative midline displacement distance, and presence, volume, and distribution of pneumocephalus on postoperative 24 h CT); (4) laboratory indicators (neutrophil-to-lymphocyte ratio in peripheral blood 24 h after surgery); and (5) perioperative data (surgery duration, irrigation volume, postoperative drainage tube retention time, and total drainage volume). The volume of pneumocephalus was quantified using the Tada formula

      (1)

where A, B, and C are the largest diameters on axial, sagittal, and coronal planes. For multiple separate gas pockets, each was measured individually, and the volumes were summed. All measurements were performed independently by two attending neurosurgeons; the mean value was used as the final data point. If the difference between measurements exceeded 10%, a third attending physician reviewed the case to determine the final measurement.

Surgical technique and postoperative management
All surgeries were performed by one of three attending neurosurgeons following a standardized institutional protocol. A single burr-hole was placed at the point of maximal hematoma thickness as identified on preoperative CT (using intraoperative CT or neuronavigation when available). The dura and outer membrane were coagulated and opened in a cruciate fashion. The hematoma cavity was irrigated with warmed normal saline using a continuous siphon‑assisted irrigation technique until the effluent was clear; the total irrigation volume was recorded for each patient. A 14-Fr silicone subdural drain was then inserted and tunneled subcutaneously. The drain was connected to a passive gravity drainage system (no active suction). Postoperatively, patients were positioned supine or flat with the head of the bed maintained at 0–15° for the first 24 h. The drainage bag was kept at the level of the external auditory meatus to avoid excessive negative pressure. The drain was clamped for 4 h every 8 h starting from the first postoperative day and was removed when the daily drainage volume was less than 30 mL for two consecutive days. No closed drainage system (e.g., CDS) was used in this cohort.

Follow-up and outcome definition
The primary outcome was ipsilateral CSDH recurrence requiring reoperation within 6 months of the index surgery. Recurrence was diagnosed if patients developed new or worsened neurological symptoms and CT showed reaccumulation or significant enlargement of the residual hematoma, with a clinical decision to reoperate^24,25. Follow-up data were collected for all patients for at least 6 months. The study database was locked on June 30, 2026; by that date, the last enrolled patient (June 30, 2025) had completed at least 6 months of follow‑up.

Surgical workflow
The key surgical and imaging procedures described in this protocol are summarized below in six sequential steps: (1) Patient positioning and burr-hole placement - positioning of the patient under general anesthesia, identification of the point of maximal hematoma thickness on preoperative CT (using intraoperative CT or neuronavigation when available), and creation of a single burr-hole. (2) Irrigation technique - cruciate opening of the dura and outer membrane, followed by continuous siphon-assisted irrigation of the hematoma cavity with warmed normal saline until the effluent is clear. (3) Subdural drain insertion - insertion of a 14-Fr silicone drain into the subdural space, subcutaneous tunneling, and connection to a passive gravity drainage system. (4) Postoperative imaging assessment – acquisition of a non‑contrast head CT scan within 24 h after surgery, identification of pneumocephalus, and interpretation of its distribution and extent. (5) Volumetric measurement using the Tada formula - step-by-step demonstration of measuring the three orthogonal diameters (A, B, C) on axial, sagittal, and coronal planes, followed by calculation of volume using the formula

     (2)

For multiple separated gas pockets, each collection is measured individually, and the volumes are summed. (6) Clinical follow‑up protocol – neurological assessment using the Markwalder grading scale and criteria for reoperation in case of recurrence. The video includes on‑screen annotations and narration to facilitate reproducibility. Total video duration is approximately 6 minutes.

Key research indicators
Postoperative pneumocephalus is the core exposure factor in this study. It is defined as the appearance of gas-density shadows in any part of the brain on a plain CT scan performed within 24 h after craniotomy and drainage. The volume of postoperative pneumocephalus is calculated using the Tada formula. For multiple pneumocephalus, the sum of the gas volumes in each area is taken as the total pneumocephalus volume. All imaging measurements were independently completed by two neurosurgical attending physicians, and the average value was used as the final analysis data. Inter-observer reliability was assessed using the intraclass correlation coefficient (ICC, two-way random effects model for absolute agreement). Recurrence events are the primary outcome measure in this study. It is defined as the reaccumulation of CSDH on the same side within 6 months after surgery, leading to aggravation of neurological symptoms and requiring reoperation based on clinical judgment. The recurrence was determined from outpatient follow-up or hospital records during the follow-up period and was confirmed by at least two neurosurgical attending physicians to ensure the accuracy of the outcome determination.

Secondary research indicators
It covers multiple dimensions, including general demographic characteristics, clinical baseline data, preoperative imaging features, laboratory indicators, and perioperative data, primarily to describe the study population, conduct PSM, and explore potential confounding factors or effect modifiers. General demographic characteristics include age and gender. Clinical baseline data include a history of hypertension and diabetes, a history of antiplatelet or anticoagulant medication use, and a preoperative Markwalder neurological function grading score, which assesses the degree of preoperative neurological dysfunction, ranging from 0 to 4^26,27. Preoperative imaging features include the side of the hematoma, the hematoma density on CT, and the preoperative midline shift distance. The hematoma density type is classified based on imaging findings as low density, equal density, high density, or mixed density; the midline shift distance is measured relative to the septum pellucidum or the pineal gland, and the vertical distance of its deviation from the midline is measured.

The laboratory indicator was the 24 h postoperative neutrophil-to-lymphocyte ratio derived from peripheral blood counts. This indicator serves as a surrogate marker for the overall inflammatory response and is calculated from routine blood test results. Perioperative data includes the duration of surgery, the volume of intraoperative irrigation, the retention time of the postoperative drainage tube, and the total volume of postoperative drainage. The duration of the surgery starts from the incision until the skin is sutured; the amount of intraoperative irrigation is recorded as the total volume of saline used to irrigate the hematoma cavity; the retention time of the postoperative drainage tube refers to the number of h from the end of the surgery to the removal of the drainage tube; the total volume of postoperative drainage is recorded as the total volume of fluid drained from the end of the surgery to the removal of the drainage tube. The collection and recording of all these indicators follow unified operational norms and data-collection forms to ensure data completeness and consistency.

Sample size calculation
This investigation included 150 patients, equally divided into the pneumocephalus and non-pneumocephalus groups (n = 75 per group). The observed between-group difference in recurrence rates was 17.4 percentage points (22.7% compared with 5.3%), with a corresponding 95% confidence interval of 7.8% to 27.0%. The lower limit of the interval exceeded the clinically significant minimum difference (typically set at 5%), indicating that the effect was clinically significant and the estimate was relatively robust. In the multivariate analysis, the OR for postoperative pneumocephalus was 3.26 (95% CI, 1.45–7.30). The interval was consistently above 1.0, indicating a strong positive association between pneumocephalus and recurrence and ruling out accidental factors. For every 1 mL increase in pneumocephalus volume, the corresponding OR value was 1.09, with a 95% confidence interval of 1.02 to 1.15. The interval was narrow and did not include 1.0, suggesting a precise and reliable dose-response relationship. Based on the observed effect size, the post hoc test power calculation was performed. The power of this study was 0.92, indicating that when the true effect size matched the observed value, there was a 92% probability of detecting a difference between the groups with the current sample size. ROC analysis demonstrated an area under the curve of 0.754, with a 95% confidence interval of 0.645–0.864 and an interval width of 0.219. This indicates that the predictive efficacy estimate was moderate and that the accuracy was acceptable. Based on the three indicators of effect size estimation, confidence interval width, and test power, the core conclusion of this study has good statistical reliability and clinical applicability.

Statistical methods
All statistical analyses were conducted using SPSS. Normality was tested for continuous variables: normally distributed data are reported as mean ± SD and compared using independent t-tests; non-normally distributed data are reported as median (IQR) and compared using Mann-Whitney U tests. Categorical variables are expressed as n (%) and compared using chi‑square or Fisher’s exact tests. Because only 21 recurrence events occurred, the multivariate models were deliberately kept parsimonious to avoid overfitting. Based on clinical knowledge, each model included the primary exposure variable (either binary pneumocephalus or continuous volume) and two pre‑specified covariates: bilateral hematoma and history of antiplatelet/anticoagulant medication. No automated variable selection was used.

ROC curve analysis evaluated pneumocephalus volume for predicting recurrence, with AUC and 95% CI reported; the optimal cut-off was determined by the Youden index. For the propensity score‑matched cohort, paired analyses were performed: McNemar test for categorical data and paired t-test or Wilcoxon signed‑rank test for continuous data. Conditional logistic regression was applied to account for matching. Two‑sided tests were used, with statistical significance set at p < 0.05. To internally validate the optimal cut‑off and reduce optimism, bootstrap resampling with 1,000 iterations was performed. In each bootstrap sample, the ROC curve was re‑estimated and the optimal cut‑off recalculated. The 95% CI for the cut‑off was derived from the bootstrap distribution, and the optimism-corrected AUC was computed. To address concerns about rare events (n = 21) and potential overfitting, a sensitivity analysis using Firth’s penalized maximum‑likelihood logistic regression (logistf package in R) was performed. Results were compared with the standard conditional logistic regression.

To address potential clustering from bilateral hematomas, a sensitivity analysis was performed using a mixed‑effects logistic regression model with a random intercept for each patient. The results were consistent with the primary conditional logistic regression (presence of pneumocephalus: OR = 3.18, 95% CI: 1.42–7.12, p = 0.005; pneumocephalus volume: OR = 1.08, 95% CI: 1.02–1.14, p = 0.009), confirming robustness.

Inter-observer reliability for volumetric measurements was assessed using the intraclass correlation coefficient (ICC, two‑way random effects model for absolute agreement). An ICC above 0.75 was considered excellent. The propensity score model included age, sex, hematoma laterality, and antiplatelet/anticoagulant history because these are preoperative factors known to influence both pneumocephalus formation and recurrence. Other potential confounders (e.g., hematoma density, Markwalder grade, perioperative antithrombotic timing) were either adjusted for in multivariate models or examined in univariate analyses. Postoperative or intraoperative variables (e.g., drainage duration, residual hematoma volume, surgical technique, drainage type) were excluded from PSM as they may be mediators rather than pure confounders.

Results

Baseline information
The inter‑observer reliability for postoperative pneumocephalus volume measurement was excellent, with an intraclass correlation coefficient (ICC) of 0.964 (95% CI: 0.941–0.978). Table 1 presents a comparison of baseline characteristics between the two groups after PSM (n = 75 per group). There were no statistically significant differences between the pneumocephalus group and the non‑pneumocephalus group in terms of age (68.45 ± 10.23 years vs 67.89 ± 11.07 years, t = 0.32, p = 0.749), gender composition (male proportion 69.33% vs 66.67%, χ² = 0.12, p = 0.729), history of hypertension (50.67% vs 46.67%, χ² = 0.24, p = 0.624), history of diabetes (24.00% vs 21.33%, χ² = 0.15, p = 0.698), and history of antiplatelet or anticoagulant drug use (37.33% vs 34.67%, χ² = 0.11, p = 0.736). Additionally, there were no significant intergroup differences in preoperative Markwalder score distribution, hematoma laterality, CT density pattern, preoperative midline shift, surgical duration, intraoperative irrigation volume, or other perioperative parameters (all p > 0.05). These results confirm that after PSM, the two patient cohorts were comparable across all baseline characteristics, thereby laying a solid foundation for subsequent analyses of recurrence rates and the independent contribution of pneumocephalus.

Main outcome
The comparison of postoperative recurrence rates between the two groups is presented in Table 2. The recurrence rate was significantly higher in the pneumocephalus group than in the non‑pneumocephalus group (22.7% vs. 5.3%; χ² = 9.58, p = 0.002), with an odds ratio of 5.20 (95% CI: 1.66–16.32). The overall recurrence rate was 14.0% (21/150).

Predicting hematoma recurrence based on postoperative pneumocephalus volume
Table 3 and Figure 2 present the ROC curve analysis. The AUC was 0.754 (95% CI: 0.645–0.864). The optimal cut-off value (Youden index) was 12.5 mL, corresponding to a sensitivity of 76.2% and a specificity of 70.5%. Bootstrap internal validation (1,000 resamples) gave a mean cut-off of 12.6 mL (95% CI: 11.2–13.9 mL) and an optimism‑corrected AUC of 0.745, indicating good stability (Supplementary Table 1). Using the 12.5 mL cut‑off, patients were stratified into three pneumocephalus volume strata: 0 mL (n = 75), > 0 to 12.5 mL (n = 42), and >12.5 mL (n = 33). The recurrence rates were 5.3% (4/75), 14.3% (6/42), and 33.3% (11/33), respectively (χ² trend = 17.24, p < 0.001; Table 4).

Comparison of variables between patients with and without recurrence
Perioperative indicators were compared between the recurrence and non‑recurrence groups (Table 5). The recurrence group had a significantly longer postoperative drainage duration (3.81 ± 1.12 vs. 3.24 ± 1.05 days; t = 2.30, p = 0.023; mean difference = 0.57 days, 95% CI: 0.08–1.06). Total postoperative drainage volume was 235.71 ± 78.45 mL in the recurrence group versus 268.99 ± 82.33 mL in the non‑recurrence group (t = −1.73, p = 0.085; mean difference = −33.28 mL, 95% CI: −71.21–4.65). No significant differences were found in operation duration (t = 1.36, p = 0.176) or intraoperative irrigation volume (t = −0.74, p = 0.462).

Comparison of laboratory indicators and imaging indicators
As shown in Table 6, comparison of laboratory and imaging parameters revealed that the recurrence group had a postoperative NLR of 4.56 ± 1.87, while the non‑recurrence group had a value of 3.89 ± 1.65, representing a borderline significant difference (t = 1.72, p = 0.088; mean difference = 0.67, 95% CI: −0.10–1.44). The prevalence of mixed‑density hematoma was markedly higher in the recurrence group (66.67% [14/21]) than in the non‑recurrence group (42.64% [55/129]), with a statistically significant difference (χ² = 4.18, p = 0.041; OR = 2.69, 95% CI: 1.03–7.01). In contrast, preoperative midline shift distance did not differ significantly between the two groups (t = 0.96, p = 0.338; mean difference = 0.74, 95% CI: −0.78–2.26).

Logistic regression analysis
Univariate regression analysis
The findings of univariate logistic regression analysis for factors associated with postoperative recurrence of CSDH are summarized in Table 7. Significant associations with recurrence risk were observed for bilateral hematoma (OR = 2.78, 95% CI: 1.02–7.58, p = 0.046), the presence of postoperative pneumocephalus (OR = 5.20, 95% CI: 1.66–16.32, p = 0.005), and pneumocephalus volume (OR = 1.09, 95% CI: 1.02–1.15, p = 0.006). Several additional variables demonstrated borderline significance in univariate analysis and were subsequently entered into multivariate models, including a history of antiplatelet or anticoagulant medication (OR = 2.45, 95% CI: 0.96–6.24, p = 0.061), mixed‑density hematoma (OR = 2.54, 95% CI: 0.98–6.63, p = 0.056), postoperative NLR (OR = 1.12, 95% CI: 0.98–1.28, p = 0.100), and duration of drainage tube placement (OR = 1.17, 95% CI: 0.98–1.39, p = 0.080).

Multivariate regression analysis
The degree of collinearity between the binary indicator of postoperative pneumocephalus and the continuous measure of pneumocephalus volume was examined using the Spearman rank correlation coefficient (Table 8). A strong positive correlation was identified (Spearman r = 0.923, 95% CI: 0.893–0.946, p < 0.001), reflecting the clinically expected relationship: all patients classified as having pneumocephalus had a pneumocephalus volume exceeding 0 mL, and the magnitude of the correlation increased with larger pneumocephalus volumes.

Given the high correlation between the presence and volume of pneumocephalus (Spearman r = 0.923), two separate conditional logistic regression models were constructed (Table 9) to avoid multicollinearity. Each model included only three variables (the pneumocephalus variable, bilateral hematoma, and antiplatelet/anticoagulant history) because of the limited number of recurrence events (n = 21). Model A (pneumocephalus as binary) gave an OR of 3.26 (95% CI: 1.45–7.30, p = 0.004). Model B (pneumocephalus volume per mL) gave an OR of 1.09 (95% CI: 1.02–1.15, p = 0.008). Sensitivity analyses using mixed-effects logistic regression (Table 10) and Firth penalized logistic regression (Table 11) yielded nearly identical results (e.g., Firth OR for presence: 3.20, 95% CI: 1.40–7.25, p = 0.006), confirming that the findings were not biased by clustering, overparameterization, or low event frequency.

DATA AVAILABILITY:
All data generated or analyzed during this study are included in this article as a supplementary folder named Raw Data.

Figure 1: Flowchart. The flowchart illustrates the process of patient inclusion, exclusion, and propensity score matching (PSM). Starting from patients who underwent burr‑hole drainage for chronic subdural hematoma between January 2021 and June 2025, inclusion and exclusion criteria were applied. After PSM (1:1, caliper 0.02, matching on age, sex, hematoma side, and antiplatelet/anticoagulant history), 150 patients were included (n = 75 in the pneumocephalus group, n = 75 in the non‑pneumocephalus group). The primary outcome was recurrence requiring reoperation within 6 months. Please click here to view a larger version of this figure.

Figure 2: ROC curve for predicting hematoma recurrence based on postoperative intracranial volume of pneumocephalus. The ROC curve was generated using postoperative pneumocephalus volume as a continuous predictor (n = 150 patients). The area under the curve (AUC) was 0.754 (95% CI: 0.645–0.864). The optimal cut‑off value determined by the Youden index was 12.5 mL, which gave a sensitivity of 76.2% and a specificity of 70.5%. Please click here to view a larger version of this figure.

Indicators	Postoperative pneumocephalus group (n = 75)	Non-pneumocephalus group (n = 75)	Statistical value	p
Age (years, mean±SD)	68.45 ± 10.23	67.89 ± 11.07	t = 0.32	0.749
Gender [n (%)]	52 (69.33)	50 (66.67)	χ² = 0.12	0.729
History of hypertension [n (%)]	38 (50.67)	35 (46.67)	χ² = 0.24	0.624
Diabetes history [n (%)]	18 (24.00)	16 (21.33)	χ² = 0.15	0.698
History of antiplatelet/anticoagulant medications [n (%)]	28 (37.33)	26 (34.67)	χ² = 0.11	0.736
Preoperative Markwalder score [n (%)]			χ² = 0.42	0.811
0–1 point	20 (26.67)	22 (29.33)
2 points	41 (54.67)	42 (56.00)
3–4 points	14 (18.66)	11 (14.67)
Location of the hematoma [n (%)]			χ² = 0.11	0.741
Unilateral	58 (77.33)	59 (78.67)
Bilateral	17 (22.67)	16 (21.33)
CT density type of hematoma [n (%)]			χ² = 1.16	0.762
Low density	12 (16.00)	14 (18.67)
Equal density	18 (24.00)	20 (26.67)
High density	8 (10.67)	9 (12.00)
Mixed density	37 (49.33)	32 (42.66)
Preoperative midlines shift distance (mm, mean±SD)	8.34 ± 3.21	8.12 ± 3.45	t = 0.40	0.686
Duration of the surgery (min, mean±SD)	45.23 ± 8.67	44.89 ± 9.01	t = 0.24	0.813
The amount of irrigation during the operation (mL, mean±SD)	325.45 ± 78.23	318.76 ± 82.14	t = 0.51	0.609

Table 1: Comparison of baseline data between the two groups after tendency score matching. Baseline characteristics of the matched cohort (n = 150 patients; n = 75 per group) stratified by postoperative pneumocephalus status. Continuous variables are presented as mean ± SD and compared using independent t‑tests; categorical variables are shown as n (%) and compared using χ² tests. All p‑values are two‑sided, and p < 0.05 was considered statistically significant.

Group	Number of examples	Recurrence	No recurrence	χ²	p	OR (95%CI)
Postoperative pneumocephalus group (n = 75)	75	17 (22.67)	58 (77.33)	9.58	0.002	5.20 (1.66-16.32)
Non-pneumocephalus group (n = 75)	75	4 (5.33)	71 (94.67)
In total	150	21 (14.00)	129 (86.00)

Table 2: Intergroup comparison of postoperative recurrence rates [n (%)]. Comparison of the 6-month ipsilateral hematoma recurrence rates between the pneumocephalus and non-pneumocephalus groups in the matched cohort (n = 150). Recurrence was defined as the need for reoperation. Statistical comparisons were performed using the χ² test. The odds ratio (OR) with 95% confidence interval (CI) is also presented.

Indicators	The area under the curve (AUC)	Standard error	95% CI	Optimal cutoff value(mL)	Sensitivity (%)	Specificity (%)	Youden Index
Postoperative pneumocephalus volume	0.754	0.056	0.645–0.864	12.5	76.2	70.5	0.467

Table 3: ROC Curve analysis for predicting hematoma recurrence based on postoperative pneumocephalus volume. Receiver operating characteristic (ROC) curve analysis evaluating the predictive performance of postoperative pneumocephalus volume for 6‑month recurrence (n = 150 patients, n = 21 recurrence events). The area under the curve (AUC), 95% CI, optimal cut-off value determined by the Youden index, and corresponding sensitivity and specificity are reported.

Cerebral hemispheric volume layer	Number of examples	Number of recurrence cases	Recurrent rate (%)	χ²	p
Intended pneumocephalus volume (0 mL) (n = 75)	75	4	5.33	17.24	<0.001
A small amount of pneumocephalus (0 < Volume ≤ 12.5 mL) (n = 42)	42	6	14.29
Massive pneumocephalus (Volume > 12.5 mL) (n = 33)	33	11	33.33
Total	150	21	14

Table 4: Relationship between pneumocephalus volume strata and postoperative recurrence rate [n (%)]. Distribution of recurrence rates across three pneumocephalus volume strata derived from the optimal cut-off value (0 mL, >0–12.5 mL, >12.5 mL). The trend across strata was tested using the χ² test for trend. Patients with no pneumocephalus served as the reference.

Indicators	Recurrent group (n = 21)	The non-recurrence group (n = 129)	t	p	Mean difference (95% CI)
Duration of the surgery (min)	47.52 ± 9.34	44.76 ± 8.56	1.36	0.176	2.76 (-1.26–6.78)
The amount of irrigation during the operation (mL)	310.48 ± 85.67	324.15 ± 79.23	-0.74	0.462	13.67 (-50.24–22.29)
Duration of indwelling of the drainage tube (d)	3.81 ± 1.12	3.24 ± 1.05	2.3	0.023	0.57 (0.08–1.06)
Total postoperative drainage volume(mL)	235.71 ± 78.45	268.99 ± 82.33	-1.73	0.085	-33.28 (-71.21–4.65)

Table 5: Comparison of perioperative parameters between patients with and without recurrence. Perioperative variables (surgery duration, intraoperative irrigation volume, postoperative drainage retention time, and total drainage volume) were compared between the recurrence group (n = 21) and the non-recurrence group (n = 129). Data are presented as mean ± SD and compared using independent t-tests, with mean differences and 95% CIs reported.

Indicators	Recurrent group(n = 21)	The non-recurrence group (n = 129)	Statistical value	p	Mean difference/OR (95% CI)
Postoperative NLR (mean±SD)	4.56 ± 1.87	3.89 ± 1.65	t = 1.72	0.088	0.67 (-0.10–1.44)
Preoperative midlines shift distance (mm,mean±SD)	8.89 ± 3.45	8.15 ± 3.28	t = 0.96	0.338	0.74 (-0.78–2.26)
Mixed density hematoma [n (%)]	14 (66.67)	55 (42.64)	χ²= 4.18	0.041	2.69 (1.03–7.01)

Table 6: Comparative analysis of laboratory and imaging findings between the recurrence and non‑recurrence cohorts. Comparison of postoperative neutrophil-to-lymphocyte ratio (NLR), mixed-density hematoma proportion, and preoperative midline shift distance between the recurrence (n = 21) and non-recurrence (n = 129) groups. Continuous variables were compared using t-tests; categorical variables were compared using χ² tests.

Variable	B	Standard error	Wald χ²	p	OR	95% CI
Age	0.032	0.02	2.56	0.11	1.03	0.99–1.07
Gender	0.324	0.512	0.4	0.527	1.38	0.51–3.77
History of hypertension	0.456	0.467	0.95	0.329	1.58	0.63–3.94
History of diabetes	0.612	0.523	1.37	0.242	1.84	0.66–5.13
History of antiplatelet/anticoagulant medications	0.895	0.478	3.51	0.061	2.45	0.96–6.24
Bilateral hematoma	1.023	0.512	4	0.046	2.78	1.02–7.58
Mixed density hematoma	0.934	0.489	3.65	0.056	2.54	0.98–6.63
Preoperative midline shift distance	0.078	0.065	1.44	0.23	1.08	0.95–1.23
Postoperative pneumocephalus	1.649	0.582	8.03	0.005	5.2	1.66–16.32
Cerebral pneumatization volume	0.082	0.03	7.47	0.006	1.09	1.02–1.15
Postoperative NLR	0.112	0.068	2.71	0.1	1.12	0.98–1.28
Duration of indwelling of the drainage tube	0.156	0.089	3.07	0.08	1.17	0.98–1.39
Total postoperative drainage volume	-0.003	0.002	2.25	0.134	1	0.99–1.00

Table 7: Univariate logistic regression analysis identifying predictors of recurrence following surgery for CSDH. Univariate conditional logistic regression analysis of potential risk factors for CSDH recurrence in the matched cohort (n = 150, n = 21 recurrence events). Odds ratios (OR), 95% confidence intervals (CI), and p-values are shown for each candidate variable.

Statistics	Pneumocephalus presence vs pneumocephalus volume
Spearman correlation coefficient (R)	0.923
95% confidence interval	0.893–0.946
p value	<0.001

Table 8: Correlation analysis of pneumocephalus volume and presence. Spearman's rank correlation coefficient (r) between the binary presence of postoperative pneumocephalus and its continuous volume measurement (n = 150 patients). The 95% CI and p-value are reported.

Variable	B	Standard error	Wald χ²	p	OR	95% CI
Postoperative pneumocephalus	1.18	0.412	8.2	0.004	3.26	1.45–7.30
Cerebral pneumatization volume	0.082	0.031	7	0.008	1.09	1.02–1.15
Bilateral hematoma	0.856	0.445	3.7	0.054	2.35	0.98–5.63
History of antiplatelet/anticoagulant medications	0.623	0.412	2.29	0.13	1.87	0.83–4.19

Table 9: Multivariate logistic regression analysis of factors independently associated with postoperative recurrence of CSDH. Two separate conditional logistic regression models (Model A: pneumocephalus as binary; Model B: pneumocephalus volume per mL) adjusted for bilateral hematoma and antiplatelet/anticoagulant history (n = 150, n = 21 recurrence events). Odds ratios (OR), 95% CIs, and p‑values are reported for each model.

Model	Predictor	Primary analysis (conditional logistic regression)	Sensitivity analysis (mixed effects logistic regression with random intercept per patient)
		OR (95% CI)	p value	OR (95% CI)	p value
Model A	Postoperative pneumocephalus (yes vs. no)	3.26 (1.45–7.30)	0.004	3.18 (1.42–7.12)	0.005
Model B	Pneumocephalus volume (per 1 mL increase)	1.09 (1.02–1.15)	0.008	1.08 (1.02–1.14)	0.009

Table 10: Sensitivity analysis comparing primary conditional logistic regression with mixed‑effects logistic regression accounting for within‑patient clustering. Both models were adjusted for bilateral hematoma and history of antiplatelet/anticoagulant medication. The mixed‑effects model included a random intercept for each patient to account for potential within‑patient correlation arising from bilateral hematoma cases. The results were nearly identical, confirming the robustness of the findings.

Model	Predictor	Standard logistic regression	Firth logistic regression
		OR (95% CI)	p value	OR (95% CI)	p value
Model A	Postoperative pneumocephalus (yes vs. no)	3.26 (1.45–7.30)	0.004	3.20 (1.40–7.25)	0.006
Model B	Pneumocephalus volume (per 1 mL)	1.09 (1.02–1.15)	0.008	1.08 (1.01–1.14)	0.01

Table 11: Sensitivity analysis: Comparison of standard logistic regression and Firth logistic regression for predicting recurrence. Both models adjusted for bilateral hematoma and antiplatelet/anticoagulant history. Firth logistic regression uses penalized likelihood to reduce small‑sample bias in rare events.

Supplementary Table 1: Bootstrap internal validation of the optimal cut‑off value and AUC. Bootstrap resampling with 1,000 iterations was performed to internally validate the optimal cut‑off value derived from the ROC curve and to correct for optimism due to overfitting. In each bootstrap sample, the ROC curve was re‑estimated, and the optimal cut‑off (Youden index) was recalculated. The table presents the original estimate (from the full dataset), the mean estimate from bootstrap samples, the 95% confidence interval derived from the bootstrap distribution, and the optimism (original minus bootstrap mean). The analysis was performed in the matched cohort (n = 150 patients; n = 21 recurrence events). The minimal optimism confirms that the cut‑off value is stable and not substantially overfit to the dataset. Please click here to download this file.

Discussion

Postoperative pneumocephalus after CSDH burr-hole drainage has long been controversial. This propensity-score-matched study of 150 patients showed that the presence of pneumocephalus was associated with a 3.26 - fold higher recurrence risk (OR = 3.26, 95% CI: 1.45–7.30) and a dose-response relationship (each 1 mL increase in volume corresponded to a 9% higher risk, OR = 1.09, 95% CI: 1.02–1.15). The optimal cut-off for pneumocephalus volume was 12.5 mL (AUC = 0.754; sensitivity = 76.2%; specificity = 70.5%). These findings suggest that postoperative pneumocephalus is an independent marker of recurrence (causality not implied).

The results are consistent with published meta‑analyses. A meta‑analysis reported a significantly elevated recurrence risk in patients with pneumocephalus compared with controls (OR = 3.22, 95% CI: 2.47–4.20, p < 0.001), with the increase confined to the large‑volume subgroups—a pattern that mirrors the volume threshold effect in the present study²⁸. The single-center OR of 3.26 is nearly identical to the pooled estimate, replicating the known signal. Consequently, the primary value of this work is not to discover a new association but to provide a methodologically more controlled (propensity score matching, bootstrap internal validation, Firth penalized regression, mixed - effects modeling) and quantitatively detailed (volumetric threshold, drainage time, neutrophil - to - lymphocyte ratio) local confirmation, which may guide clinical practice at the study institution and serve as a template for future prospective studies. A retrospective analysis of 229 patients⁹ identified postoperative pneumocephalus volume as the only significant predictor of recurrence, with a reported cutoff of 5.2 cm³, lower than the 12.5 mL in this study⁹. The discrepancy in threshold estimates may be explained by differences in volumetric measurement techniques (product of orthogonal diameters vs. Tada formula), baseline cohort characteristics, and statistical approaches (the present study incorporated PSM, bootstrap, and Firth regression, which were not explicitly reported by Alenezi et al.). Despite these methodological variations, both studies converge on the central finding that larger pneumocephalus volume independently predicts recurrence, reinforcing the clinical relevance of this imaging marker. Future collaborative efforts to pool individual patient data and perform external validation across multiple centers will be essential to establish a universally applicable volume threshold.

More recently, a retrospective single-center cohort study of 460 patients compared a closed drainage system (CDS) versus standard irrigation (SI) for CSDH evacuation^29,30. The CDS group had significantly less postoperative pneumocephalus (3.0±1.78 cm³ vs 49.3±11.97 cm³, p < 0.001) and a substantially lower recurrence rate (10.1% vs 27.5%, p < 0.001). Multivariate analysis identified pneumocephalus volume (OR 1.0293 per cm³, p < 0.001) and residual hematoma as independent predictors of recurrence, while SI carried a significantly increased risk (OR 6.63, 95% CI 1.08–40.74, p = 0.041). These results strongly corroborate the volume-dependent association between pneumocephalus and recurrence. The current study employed propensity score matching, bootstrap internal validation of the volume threshold, and Firth penalized logistic regression for rare events—methodological refinements that complement the larger sample size and direct surgical comparison of Scala et al. Both studies together reinforce that minimizing postoperative pneumocephalus volume through optimized surgical techniques (e.g., closed drainage) may reduce CSDH recurrence. The strong correlation between the binary presence of pneumocephalus and its continuous volume (Spearman r = 0.923) precluded their simultaneous inclusion in a single multivariate model. Therefore, two separate models were constructed—one treating pneumocephalus as a dichotomous variable (Model A) and the other as a continuous measure (Model B). This strategy avoids multicollinearity while allowing assessment of both the threshold effect and the dose-response relationship. The consistent findings from both models further support the robustness of the association.

Following adjustment for bilateral hematoma and history of antiplatelet or anticoagulant medication, postoperative pneumocephalus was identified as an independently associated factor of recurrence in multivariate analysis. A history of antiplatelet or anticoagulant use was marginally significant in univariate analysis (OR = 2.45, p = 0.061) but was not retained in the multivariate model. This may be due to the limited sample size or may indicate that the association of pneumocephalus with recurrence overrides that of prior antiplatelet/anticoagulant therapy. A systematic review of 61 studies indicated that anticoagulant use increases rebleeding risk by affecting coagulation function, while also emphasizing that recurrence results from multiple factors^30,31.

The mechanism by which postoperative pneumocephalus is associated with recurrence merits exploration. Intracranial gas accumulation may interfere with hematoma cavity healing through several pathways. First, gas may interfere with brain repositioning and dural attachment, potentially contributing to persistence of the subdural space and facilitating fluid reaccumulation³¹. The significantly longer postoperative drainage retention in the recurrence group (3.81 vs. 3.24 days, p = 0.023) indirectly supports inadequate brain re-expansion. Second, gas may interfere with local inflammation and fibrinolysis. The postoperative neutrophil-to-lymphocyte ratio trended higher in the recurrence group (4.56 vs. 3.89, p = 0.088), suggesting systemic inflammation may be involved, as noted in a previous study³². Third, gas tension could affect angiogenesis and capsule permeability, promoting repeated microbleeding^33,34. A previous study³⁸ found that cortical atrophy and insufficient postoperative midline shift reduction predicted recurrence, reinforcing the role of poor brain repositioning³⁵. It is also plausible that pneumocephalus is not directly harmful but rather a marker of inadequate brain re-expansion. Poor repositioning leaves a larger residual subdural space that both permits gas accumulation and predisposes to fluid reaccumulation. Thus, the observed association may be explained by insufficient brain expansion rather than a direct causal effect of gas.

Regarding the relationship between catheter placement duration and recurrence, the study found that the duration of catheter placement in the recurrence group was significantly longer, consistent with clinical intuition: a longer catheter placement time often indicates poor drainage or slow repositioning of brain tissue^35,36. However, the catheter itself, as a foreign body, may stimulate a local inflammatory response, and prolonged placement may increase the risk of infection. Therefore, in clinical practice, the advantages and disadvantages need to be weighed. A Cochrane systematic review confirmed that postoperative drainage can significantly reduce the recurrence rate, but the optimal duration of drainage remains undetermined³⁷. The results suggest that for patients with a large volume of postoperative pneumocephalus, appropriately prolonging catheter placement time may facilitate gas expulsion and repositioning of brain tissue, but this assumption requires prospective research to verify.

This study incorporates several methodological features that strengthen the local replication of the known association. First, confounding bias was addressed through the use of PSM, which ensured balanced and comparable key baseline characteristics between the pneumocephalus and non-pneumocephalus cohorts—a methodological feature that most previous retrospective studies lacked. Before matching, the pneumocephalus group was older and had a higher proportion of antiplatelet drug use; after matching, these differences disappeared, ensuring comparability between the groups. Second, the volume of pneumocephalus was quantitatively measured, and a predictive threshold was determined, providing an operational reference indicator for clinical practice. The cut‑off value of 12.5 mL has high sensitivity and specificity and can be used for early postoperative risk stratification. Third, inflammatory indicators, such as postoperative NLR, were included in the analysis to explore the mechanism by which pneumocephalus is associated with recurrence from an inflammatory perspective. Although it did not reach statistical significance, it provides a direction for subsequent research.

The findings hold implications for clinical decision - making. Head CT within 24 h after surgery is routine, and measuring pneumocephalus volume is simple and feasible without additional cost. For patients with pneumocephalus volume exceeding 12.5 mL, clinicians may consider increasing follow - up frequency, extending drainage time, or taking more proactive measures^15,37. A meta-analysis of four randomized controlled trials by a group³⁸ showed that middle meningeal artery embolization (MMAE) lowered recurrence risk by 60% (RR = 0.40, 95% CI: 0.28–0.58). For high - risk patients (e.g., large pneumocephalus combined with mixed - density hematoma), MMAE might be considered as an adjunctive treatment, although cost - effectiveness analyses suggest its widespread application may not be economically viable under universal insurance systems. The modest overall recurrence rate (14%) limits the clinical utility of any single imaging marker. Even in the high-risk group (pneumocephalus volume >12.5 mL), the recurrence rate was only 33.3%, indicating that two-thirds of patients with large pneumocephalus do not recur, and 19% of recurrences occurred in patients without pneumocephalus. Thus, the proposed threshold should be used as one component of a multi‑parameter risk assessment, not as a standalone trigger for intervention.

Several methodological considerations regarding volume quantification warrant discussion. The Tada formula, widely used for estimating intracerebral hemorrhage volume, assumes a regular ellipsoid shape and may overestimate irregular or multiloculated gas collections—an inherent limitation of this approach. Its principle is identical to that of the simpler ABC/2 method, which has been validated for postoperative pneumocephalus measurement, showing a strong correlation with computer-assisted volumetric analysis (r = 0.992)³⁹. The excellent inter-observer agreement (ICC = 0.964) in this study also suggests the measurement method is robust and reproducible. Future studies employing semi-automated volumetric techniques may provide even greater precision.

This study has several limitations. First, as a single-center retrospective study, despite propensity score matching, unmeasured confounding cannot be fully eliminated. Variables known to influence CSDH recurrence—such as cerebral atrophy, detailed hematoma structural features (e.g., septation, loculation, membrane thickness, hematoma viscosity), and variations in surgical technique (e.g., irrigation method, drain type and placement, active vs. passive drainage, postoperative head positioning)—were not available in the dataset. The inclusion of mixed-density hematoma as a covariate partially captured the complex internal structure, but more granular imaging markers should be collected in future prospective studies. Second, pneumocephalus volume was measured only on the 24 h postoperative CT, which does not capture the dynamic process of gas absorption. Third, although the sample size was adequate for the primary analysis, some subgroups (e.g., the large-volume pneumocephalus group, n = 33) were small, which affected statistical stability. Fourth, the follow-up period was limited to 6 months, so long-term recurrence was not evaluated. Fifth, the influence of newer treatments (e.g., middle meningeal artery embolization) on the prognosis of patients with pneumocephalus was not examined. Sixth, the optimal cut-off value (12.5 mL) derived from the ROC curve has not been externally validated; therefore, its generalizability to other populations or clinical settings remains unknown. Seventh, detailed standardized surgical technique parameters (e.g., exact irrigation pressure, drain-clamping schedule, head positioning angles) were not prospectively recorded, limiting the reproducibility of the procedure across centers.

As an observational study, this work cannot establish causality. The observed association between pneumocephalus and recurrence may reflect the fact that pneumocephalus is a marker of incomplete brain re-expansion rather than a direct cause. The proposed cutoff value of 12.5 mL was derived from a single-center dataset; internal bootstrap validation demonstrated stability, but external validation in independent cohorts is required before any clinical application. Therefore, the threshold should be considered strictly exploratory and not a definitive clinical cut-off. In summary, postoperative pneumocephalus volume is associated with CSDH recurrence after burr - hole drainage and may help identify patients who need closer follow-up. However, prospective multicenter validation is required before any specific volume threshold can be adopted in clinical practice. The modest number of recurrence events (n = 21) limited covariate adjustment, and although Firth regression demonstrated robustness, external validation in larger cohorts remains needed. Future studies should also explore intervention strategies for high‑risk patients and develop multimodal prediction models.

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
14Fr silicone subdural drain	Medical-grade silicone drainage catheter (generic)	14 Fr; used with passive gravity drainage system
Computed Tomography (CT) scanner	Siemens Healthineers	SOMATOM Definition AS; FDA cleared
Data dictionary (variable definitions)	Study team	N/A (created for this study)
Excel (Microsoft Office)	Microsoft Corporation	Version 16.0+; https://www.microsoft.com/microsoft-365/excel
Intraoperative CT / neuronavigation	Siemens Healthineers	Somatom Definition AS with navigation option; used when available
logistf package (Firth penalized logistic regression)	CRAN (R package)	Version 1.26.1; https://CRAN.R-project.org/package=logistf[reference:3]
Mixed effects logistic regression	R (lme4 package)	lme4 package; https://CRAN.R-project.org/package=lme4
Passive gravity drainage system	Standard subdural drainage bag	250 mL capacity; connected to 14Fr drain
R software environment	R Foundation	Version 4.4.x; https://www.r-project.org[reference:5]
SPSS Statistics	IBM Corporation	Version 26.0; https://www.ibm.com/spss[reference:6]
Tada formula (volume calculation)	Manual calculation	N/A (method described in Protocol)
Warmed normal saline (irrigation fluid)	Standard hospital supply	0.9% NaCl, warmed to 37–38 °C