Robot-assisted aspiration improves survival over conservative treatment in critical primary brainstem hemorrhage: a multicenter, volume-stratified cohort study

Introduction

Primary brainstem hemorrhage (PBSH) accounts for approximately 5% to 10% of all intracranial hemorrhages, yet it remains a formidable challenge in neurocritical care due to its extremely high mortality and disability rates (1). Currently, the prevailing international opinion still favors conservative management for brainstem hemorrhage, as rigorous conservative treatment can minimize iatrogenic injury (2). However, with increasing hematoma volume, the limitations of conservative therapy become progressively apparent. When the hematoma volume exceeds 5 mL, the early mortality rate of conservative treatment alone rises steeply. Multiple cohort studies have consistently shown that patients with hematoma volume >5 ml who receive only conservative treatment have a 30-day mortality rate generally ranging from 40% to 60%, and the long-term mortality rate remains persistently high at 70% to 80% (3–5). Evidently, under modern neurocritical care conditions, purely medical conservative management can no longer break the deadlock of high mortality in critical PBSH.

In recent years, frameless robotic-assisted neurosurgical systems, leveraging multimodal image fusion and submillimeter precision navigation, have significantly shortened preoperative preparation time, reduced surgical collateral damage, and improved 30-day survival rates, emerging as a promising interventional modality (6, 7). Nevertheless, controversy persists regarding the optimal hematoma volume range for which patients can truly benefit from this approach. It is widely accepted that patients with a hematoma volume < 5 mL are suitable for conservative treatment, whereas those with a volume >15 ml or presenting with deep coma upon admission often already have irreversible brainstem injury, leaving limited room for surgical salvage. Previous studies have indicated that when the hematoma volume crosses the threshold of 10 ml, the early mortality rate reaches an extremely dangerous peak (8). Based on this, we propose that 5–15 ml constitutes the key beneficial interval for minimally invasive surgery, and hypothesize that stratification at the 10 mL cutoff could more accurately reveal the true survival protective effect of robotic surgery across different severity levels. However, current clinical studies addressing this controversy mostly pool patients with 5–15 ml or even broader hematoma volumes together, generally lacking refined stratification based on specific hematoma volume. Consequently, the genuine differential efficacy of minimally invasive surgery across varying degrees of severity cannot be precisely identified.

Therefore, this multicenter stratified cohort study aims to clarify the value of robot-assisted puncture in patients with critical PBSH and hematoma volumes of 5–15 ml, elucidate the stratified survival differences across distinct hematoma volume grades, and verify whether this minimally invasive technique confers an independent survival protective effect in a multivariable model. The ultimate goal is to provide high-quality evidence-based medical evidence for stratified clinical decision-making and individualized minimally invasive intervention in critical PBSH.

Materials and methods

This retrospective multicenter cohort study analyzed data from 79 consecutive patients with critical PBSH who met the inclusion and exclusion criteria and were admitted to the neurosurgery departments of the Second Hospital of Hebei Medical University, Handan Central Hospital, and Jingxing County Hospital between May 2022 and October 2025. All patients had a Glasgow Coma Scale (GCS) score of 3–8 at admission and a hematoma volume of 5–15 ml. Patients were divided into two groups according to the treatment strategy: the robot-assisted aspiration group (aspiration group, n = 38) and the conservative treatment group (conservative group, n = 41). Based on the initial hematoma volume, each group was further stratified into two subgroups: the 5–10 ml subgroup and the >10 ml subgroup.

Inclusion and exclusion criteria

Inclusion criteria

• ①Imaging-confirmed primary brainstem hemorrhage with a hematoma volume of 5–15 ml.
• ②Admission GCS score of 3–8.
• ③Age ≥ 18 years.

Exclusion criteria

• ①Secondary brainstem hemorrhage caused by cerebrovascular malformation, aneurysm, tumor-related stroke, traumatic brain injury, or severe coagulation dysfunction.
• ②Concurrent multiple intracerebral hemorrhages or brainstem hemorrhage volume > 15 ml.
• ③Incomplete follow-up data.
• ④Presence of multiple underlying diseases or hepatic/renal insufficiency.

Treatment protocols

Robot-assisted aspiration group

Target planning

Emergency thin-slice head computed tomography (CT) was performed before surgery and the images were imported into the workstation of the Neurosurgical Robot System for three-dimensional reconstruction of the hematoma. The optimal puncture trajectory was planned according to the shape of the hematoma, following the principles of shortest path and avoidance of critical vessels and nerve nuclei.

Registration and catheter placement

After successful general anesthesia, the patient was placed in the prone position and the head was fixed in a head frame. Two registration methods were used: !1! Skull-pin registration – spatial registration was achieved by identifying cranial bone markers with a registered probe, ensuring a registration error < 1 mm; !2! Structured-light registration – a structured-light scanner non-contactedly scanned the facial surface, and the point-cloud data were fused with the preoperative CT facial images for registration. After successful registration, a burr hole was drilled under the guidance of the robotic arm, the dura mater was punctured, and a drainage catheter was smoothly inserted into the predetermined target. Upon outflow of old blood, manual slow aspiration was performed, with the first aspiration volume controlled at 50%−60% of the initial hematoma volume. The catheter was then secured and connected to a closed drainage bag.

Postoperative liquefaction and catheter removal strategy

Head CT was repeated within 24 h after surgery to assess catheter tip position and residual hematoma. Once active bleeding was excluded, urokinase (30,000 IU dissolved in 3 ml saline, injected over > 1 min) was administered through the drainage tube, followed by clamping for 2 h and then opening for drainage, twice daily. Catheter removal was considered when follow-up CT showed residual hematoma < 2–3 ml, brainstem compression was relieved, and the drainage volume was < 5 ml/24 h. The indwelling time was typically 5–7 days.

Conservative treatment group

Management of cerebral edema and blood pressure

The patient's head was elevated to 30°. Dehydration and intracranial pressure reduction were achieved by combined administration of 20% mannitol (every 6–8 h), furosemide (10–20 mg), and albumin. Systolic blood pressure was strictly controlled between 120 and 140 mmHg using continuous infusion of urapidil or nicardipine.

Airway management

Immediate endotracheal intubation was performed if the admission GCS score was < 8 and SpO₂ was < 90%. PaCO₂ was maintained at 35–40 mmHg. Early tracheostomy was performed within 3–7 days after onset if ventilation support was anticipated to exceed 7–14 days.

Systemic support

Enteral nutrition was initiated early. Routine acid-suppressive therapy was given for 3–7 days to prevent stress ulcers. Lower-extremity pneumatic compression was applied to prevent deep vein thrombosis.

Data collection

The data collected and recorded in this study primarily comprised baseline characteristics, perioperative parameters, and follow-up prognostic indicators. Baseline characteristics included age, sex, history of hypertension and diabetes, hemorrhage location, admission GCS score, and initial hematoma volume measured by three-dimensional reconstruction using the Huake Precision Neurosurgical Robot System. To ensure the accuracy of volume-based stratification, two board-certified neurosurgeons, blinded to treatment allocation and clinical outcomes, independently performed volumetric measurements using the built-in software of the Huake Precision Neurosurgical Robot Workstation. Inter-rater reliability was excellent, with an intraclass correlation coefficient (ICC) of 0.96 (95% CI: 0.93–0.98). In cases where discrepancies exceeded 1 mL, a third senior neurosurgeon reviewed the images and adjudicated the final volume. Perioperative and surgical data included operative duration, intraoperative hematoma evacuation volume, residual hematoma volume on postoperative 24-h head CT, and various perioperative complications. Follow-up and outcome measures included GCS score at discharge, 30-day and 90-day mortality rates, and the 90-day modified Rankin Scale (mRS) score for assessing long-term neurological functional recovery. The final follow-up endpoint was 90 days after onset.

Statistical analysis

Statistical analyses were performed using SPSS version 29.0 and GraphPad Prism version 10.0. Continuous variables were tested for normality using the Shapiro–Wilk test. Normally distributed continuous variables are presented as mean ± standard deviation and were compared using the independent-samples t-test. Non-normally distributed continuous variables are presented as median (interquartile range) and were compared using the Mann–Whitney U-test. Categorical variables are expressed as frequencies (percentages) and were compared using the chi-square test or Fisher's exact test. Ordinal variables, including the distribution of 90-day modified Rankin Scale grades, were compared between groups using the Mann–Whitney U test. The 90-day post-onset period served as the follow-up endpoint. Cumulative survival curves were generated using the Kaplan–Meier method, and survival differences between groups were compared using the log-rank test. To control for baseline confounding and identify independent prognostic factors for 90-day survival, age and sex were forced into the multivariable Cox proportional hazards regression model (Enter method) regardless of their statistical significance in univariable analysis, together with variables with a univariable P value < 0.05. Hazard ratios (HR) and their 95% confidence intervals (CI) were calculated. All tests were two-sided, and a p < 0.05 was considered statistically significant. Given the retrospective, non-randomized cohort design of this study, 1:1 propensity score matching (PSM) was performed as a pre-specified sensitivity analysis in order to effectively mitigate the selection bias that may have been present in treatment allocation. Propensity scores were estimated using a multivariable logistic regression model, and the covariates entered into the model covered the key baseline confounders: age, sex, admission GCS score, initial hematoma volume, and hemorrhage location. Nearest-neighbor matching without replacement was applied, with the caliper width set at 0.2 standard deviations of the logit of the propensity score. In the successfully matched cohort, between-group baseline balance, the main stratified efficacy outcomes (GCS score at discharge, 30-day and 90-day survival rates, and the distribution of 90-day mRS scores), and the multivariable Cox proportional hazards regression model were re-evaluated, in order to verify the robustness of the principal study conclusions after removal of selection bias.

Results

Baseline characteristics

No statistically significant differences were observed between the aspiration group and the conservative group in terms of age, sex, history of hypertension and diabetes, admission GCS score, or hemorrhage location, either in the overall cohort or in any of the hematoma volume subgroups (all p > 0.05), confirming the comparability of the two groups (Table 1).

Table 1. Baseline characteristics of patients with critical pbsh undergoing robot - assisted aspiration or conservative treatment, stratified by hematoma volume.

Data are presented as mean ± SD (for continuous variables) or n (%) (for categorical variables). GCS, Glasgow Coma Scale; PBSH, Primary brainstem hemorrhage. p- values were calculated using independent t- test (continuous variables) or χ² test (categorical variables).

Safety outcomes

There were no significant differences between the two groups in the rates of postoperative rebleeding, intracranial infection, perioperative pneumonia, or stress ulcer (all p > 0.05) (Table 2).

Table 2. Perioperative complications in patients with critical PBSH undergoing conservative treatment or robot - assisted aspiration.

Data are presented as n (%). p- values were calculated using χ² test for categorical variables.

Neurological improvement (GCS score)

In the overall cohort, the aspiration group exhibited a greater improvement in GCS score from admission to discharge compared with the conservative group. In the 5–10 ml subgroup, the discharge GCS score was significantly higher in the aspiration group than in the conservative group. In the >10 ml subgroup, neither group showed substantial improvement in GCS score at discharge (Figure 1).

30 -day survival rate

The 30-day survival rate was significantly higher in the aspiration group than in the conservative group in both the overall cohort (76.3% vs. 43.9%, P = 0.006) and the 5–10 ml subgroup (82.1% vs. 45.7%, p = 0.004). In the >10 mL subgroup, 30-day survival was 60.0% (6/10) vs. 33.3% (2/6), an absolute difference of +26.7 percentage points (95% CI: −19.9 to +59.8; Fisher's exact p = 0.61). Because this stratum comprised only 16 patients, the comparison had a post-hoc power of 0.28 and is reported as inconclusive; detecting a difference of this magnitude at α = 0.05 with 80% power would require approximately 60 patients per group (Figure 2).

90 -day survival rate

In the 5–10 ml subgroup, the 90-day survival rate was significantly higher in the aspiration group than in the conservative group (71.43% vs. 42.86%, p = 0.044). In the overall cohort (63.2% vs. 41.5%, p = 0.089) and in the >10 mL subgroup (40.0% [4/10] vs. 33.3% [2/6]; absolute difference +6.7 percentage points, 95% CI: −36.7 to +43.9; Fisher's exact p = 1.000), the aspiration group showed numerically higher survival rates. In the >10 mL subgroup this result should be regarded as inconclusive rather than as evidence of no difference (Figure 3).

Kaplan–Meier survival analysis

In the overall cohort and the 5–10 ml subgroup, the Kaplan–Meier survival curves of the two groups began to diverge markedly within the first 14 days after onset. The survival rate in the conservative group declined more rapidly during the first two weeks, whereas the curve for the aspiration group showed a significantly slower decline and entered a stable plateau phase after approximately 20 days. In the >10 mL subgroup (exploratory; 10 and 6 patients), both curves declined steeply within the first 14 days; thereafter, the aspiration group entered a relatively gradual phase, and the long-term survival curves of the two groups eventually converged. With only 16 patients and 10 events in this stratum, the shape of these curves should not be over-interpreted (Figure 4).

Functional outcome (mRS score at 90 days)

Compared with the conservative group, the aspiration group showed a lower proportion of patients with mRS score 6 (death) and a correspondingly higher proportion of patients with mRS score 5 (severe disability). In the >10 mL subgroup, only 4 of 10 patients in the aspiration group and 2 of 6 in the conservative group were alive at 90 days, and every survivor in both arms had an mRS grade of 5 (severe disability); no patient in this stratum achieved functional independence (Figure 5). This pattern indicates that the survival advantage was absorbed almost entirely by the mRS grade 5 category: patients who would otherwise have died instead survived with severe disability requiring continuous care, rather than achieving functional independence or moderate disability. The proportion of patients free of severe disability or death was therefore not demonstrably increased.

Cox proportional hazards regression analysis for 90-day survival

Univariable Cox regression analysis revealed that admission GCS score (HR = 0.713, p = 0.020), initial hematoma volume (HR = 1.165, p = 0.004), and treatment modality (HR = 0.494, p = 0.041) were significantly associated with 90-day survival, whereas age, sex, history of hypertension, and history of diabetes were not (all p > 0.05) (Figure 6).

After adjustment in the multivariable Cox model, initial hematoma volume remained an independent risk factor for 90-day mortality (HR = 1.173 per 1ml increase, 95% CI: 1.039–1.325, p = 0.010), while robot-assisted aspiration was confirmed as a significant independent protective factor, reducing the risk of death by 50.6% compared with conservative treatment (HR = 0.494, 95% CI: 0.245–0.997, p = 0.049) (Figure 7).

Propensity score-matched sensitivity analysis

To determine whether the survival benefit observed in the full cohort could be explained by baseline imbalances between the treatment groups, we repeated the primary analyses in the propensity score-matched cohort.

After 1:1 propensity score matching, 34 matched pairs of patients with highly balanced baseline characteristics were included (34 patients in the robot-assisted aspiration group and 34 in the conservative treatment group; 68 in total; Table 3). Between-group balance was achieved for age (51.2 ± 11.8 vs. 51.5 ± 13.3 years, p = 0.916), admission GCS score (4.15 ± 1.60 vs. 3.88 ± 1.34, p = 0.462), initial hematoma volume (8.24 ± 2.93 vs. 8.24 ± 2.50 mL, p = 0.993), and the distribution of hemorrhage location (p = 0.822) (all p > 0.05). No statistically significant differences were observed between the two groups in the incidence of pneumonia (76.47% vs. 73.53%), stress ulcer (29.41% vs. 26.47%), rebleeding (8.82% vs. 5.88%), or intracranial infection (0.00% vs. 2.94%) (all p > 0.05).

Table 3. Baseline characteristics and complication rates in the propensity score-matched cohort.

GCS, Glasgow Coma Scale; PSM, propensity score matching; SD, standard deviation.

In the matched cohort, all core volume-stratified outcome measures fully reproduced the findings obtained in the full cohort (Table 4, Figures 8, 9).

Table 4. Volume-stratified outcomes in the propensity score-matched cohort.

GCS, Glasgow Coma Scale; mRS, modified Rankin Scale; PSM, propensity score matching.

Typical cases

Case 1

A 54-year-old female patient with a 10-year history of hypertension and no prior diabetes mellitus presented with 1 day of sudden unconsciousness. On admission, her GCS score was 3T with endotracheal intubation. Preoperative emergency cranial CT scans were imported into a robotic workstation for three-dimensional reconstruction, confirming a brainstem hematoma with a volume of 6.2 mL. The patient received robot-assisted stereotactic puncture and drainage for hematoma evacuation. A 24-h postoperative cranial CT scan confirmed accurate placement of the drainage catheter, with the residual hematoma volume reduced to 4.0 mL and marked resolution of mass effect. The patient regained partial spontaneous breathing on postoperative day 2, indicating effective alleviation of brainstem compression by early minimally invasive decompression. During hospitalization, her neurological function improved progressively. Endotracheal intubation was removed on postoperative day 7, with complete recovery of consciousness and spontaneous breathing. The patient was discharged with a GCS score of 11 and stable vital signs, without severe perioperative complications, including rebleeding, intracranial infection, and severe pneumonia. The patient survived at the 30-day follow-up with stable neurological status and no hemorrhagic recurrence or secondary neurological deterioration. At the 90-day follow-up, the patient exhibited continuous functional recovery, presenting only mild contralateral limb weakness without consciousness disturbance or cognitive dysfunction. The 90-day mRS score was 3, indicating moderate disability with partial independence in daily living, which demonstrated favorable long-term functional outcomes after robot-assisted minimally invasive treatment (Figure 10).

Case 2

A 44-year-old male patient with a 30-year history of hypertension and no history of diabetes mellitus was admitted due to a 2-day duration of sudden unconsciousness. His admission GCS score was 3. Preoperative emergency cranial CT reconstruction identified an isolated brainstem hematoma with a volume of approximately 6 mL. After thorough communication and full informed consent from the patient's family, standardized conservative medical treatment was implemented. On hospital day 7, the patient presented occasional spontaneous eye opening, and follow-up cranial CT revealed gradual hematoma resorption with reduced lesion volume and decreased hematoma density. By hospital day 21, the patient was able to complete simple command-following movements of the right limb, and repeat CT scan demonstrated near-complete absorption of the brainstem hematoma. The patient was discharged on hospital day 21 to receive systematic rehabilitation therapy. At the 90-day follow-up, the patient achieved partial neurological functional recovery with a final mRS score of 3, presenting moderate disability and partial independence in daily activities (Figure 11).

Discussion

Safety assessment is the foremost prerequisite for any surgical intervention in the brainstem. The present study demonstrated that there were no statistically significant differences between the robot-assisted aspiration group and the conservative treatment group in terms of postoperative rebleeding rate or intracranial infection rate (both p> 0.05). Similarly, the incidences of perioperative pneumonia and stress ulcer were comparable between the two groups (p> 0.05). These findings strongly confirm that early robot-assisted puncture carries a favorable safety profile. The above data indicate that modern precision minimally invasive surgery can achieve effective physical decompression of the vital brainstem centers without incurring additional surgical risks, thereby establishing its reliable surgical feasibility (9, 10).

Multivariable Cox regression analysis confirmed that, after adjusting for age, sex, and admission GCS score, initial hematoma volume was a significant independent risk factor for 90-day mortality (11); each 1ml increase in hematoma volume independently raised the 90-day mortality risk by 17.3%. Conversely, early robot-assisted surgery emerged as a significant independent protective factor for 90-day survival, reducing the overall mortality risk by 50.6% (6). This protective effect exhibited a high degree of stratification specificity in the 5–10 ml hematoma subgroup, not only substantially lowering the 30-day and 90-day mortality rates (12), but also leading to a significant improvement in GCS score at discharge, which is of great importance for early neurological recovery and salvage.

Stratified analysis based on initial hematoma volume revealed distinctly different surgical benefit windows. For the 5–10 ml range, the 90-day Kaplan–Meier cumulative survival curves and log-rank test in this study provided robust statistical evidence of survival benefit (p = 0.010), confirming this range as the “golden window” for robot-assisted puncture. Precise volume reduction effectively interrupted the fatal compression of the vital brainstem centers caused by acute mass effect and secondary edema (6, 12, 13). For hematomas exceeding 10 ml, our data are insufficient to determine whether robot-assisted aspiration is beneficial. In the unmatched cohort, 30-day survival was 60.0% (6/10) vs. 33.3% (2/6) — an absolute difference of +26.7 percentage points — but the 95% confidence interval (−19.9 to +59.8) spans both appreciable benefit and appreciable harm, and in the propensity score-matched cohort the same comparison gave 37.5% (3/8) vs. 33.3% (2/6), a difference of only +4.2 percentage points. The apparent short-term advantage was therefore not reproduced after baseline characteristics were balanced, which suggests it was substantially influenced by between-group differences at baseline; the 90-day survival curves converged, and all survivors in both arms were severely disabled (mRS grade 5). One interpretation is that the high-pressure arterial flow generated instantaneously by massive hematomas causes devastating and irreversible mechanical destruction of the ascending reticular activating system and of vital brainstem nuclei that cannot be reversed by physical drainage alone (5, 14). however, with 16 patients this explanation cannot be distinguished from a simple loss of statistical power, and we present it as a hypothesis to be tested rather than as a conclusion supported by our data. Traditional pooled analyses tend to obscure the true surgical benefit (15); only through rigorous volume stratification can the statistical dilution of moderate-volume hematoma benefit data by extreme critical cases be avoided, which holds important discriminative value for guiding precise, stratified clinical decision-making.

The marked improvement in early survival rates coupled with the lack of substantial improvement in 90-day functional outcomes in the aspiration group reflects the irreversibility of primary neurological damage in critical PBSH (16, 17). Even though minimally invasive catheter drainage can maximally evacuate the hematoma mass and salvage the ischemic penumbra, surviving patients inevitably exhibit severe disability at 3 months post-onset (18, 19). A gain in survival that is not accompanied by a gain in independence cannot, however, simply be equated with benefit. Whether survival with severe disability requiring continuous care is preferable to death is a value judgment that our data cannot settle: some patients and families regard it as unacceptable, others as preferable to death, and preferences differ substantially between individuals, families and cultural settings. We therefore caution explicitly against reading the reduction in 90-day mortality as evidence of improved quality of life (20). Our findings are consistent with those of Hong and colleagues, whose systematic review and meta-analysis of robot-assisted surgery for primary brainstem hemorrhage likewise found that the survival advantage of the procedure is not matched by a comparable improvement in functional outcome; that analysis, like ours, was limited by small, retrospective, non-randomized primary studies with heterogeneous and short follow-up. Accordingly, the decision to operate in this population should be made through shared decision-making in which the realistic probability of survival with severe disability — not merely the probability of survival — is communicated explicitly, and in which best supportive and palliative care is presented on equal terms. Future trials and registries in this field should adopt functional independence, health-related quality of life and, ideally, patient- or proxy-reported outcome measures as co-primary endpoints, because mortality alone will systematically overstate the value of an intervention that converts death into severe dependency.

Several limitations of this study should be acknowledged. Selection bias is inherent in any retrospective, non-randomized comparison, and it is the principal caveat attaching to our findings. Treatment allocation was determined by the attending neurosurgeon in consultation with the patient's family rather than by randomization. Propensity score matching balanced the measured baseline characteristics, and the survival advantage persisted in the matched cohort; neither matching nor multivariable adjustment can, however, address factors that were not recorded. Such unmeasured clinical and psychosocial factors may include the rate of evolution of secondary peri-hematomal edema, the fragility of the deep microvasculature, and the willingness and degree of commitment of the patients' families with regard to the intensity of intensive care and to early tracheostomy and ventilator-weaning care. Each of these could plausibly have influenced both the decision to operate and the probability of death, and each may therefore have contributed residual confounding to the observed survival trajectories. The matching procedure itself also entailed a loss of 11 patients, from 79 to 68, and it is for this reason that the full-sample multivariable Cox model was retained as the primary analysis and the matched cohort used as a sensitivity analysis — an arrangement that reconciles adequate statistical power with the minimization of selection bias.

A second limitation is the modest sample size, which bears most heavily on the >10 mL stratum. That stratum contained only 16 patients, and only 14 after matching. For 30-day survival the absolute difference was +26.7 percentage points (95% CI: −19.9 to +59.8); the width of this interval is more informative than the post-hoc power of 0.28, because it shows the data to be compatible with a large survival gain, with no difference, and with clinically meaningful harm. Detecting a difference of this magnitude at α = 0.05 with 80% power would require approximately 60 patients per group, that is about 120 in this stratum alone, more than seven times the number enrolled here. In the matched cohort the same comparison gave 37.5% (3/8) vs. 33.3% (2/6), a difference of only +4.2 percentage points (95% CI: −39.6 to +43.9), so the point estimate observed in the unmatched cohort was not reproduced. We therefore regard the >10 mL stratum as inconclusive and exploratory: it provides no evidence of benefit and equally no evidence of futility, and it should not be cited in either direction.

The assessment of functional outcome was limited in two respects. First, follow-up was censored at 90 days, which is short for brainstem injury: neurological recovery, central compensation and rehabilitation gains frequently continue for 6 to 12 months or longer, and disability measured at 3 months may overstate permanent dependency. We do not assume that a longer window would necessarily favor either treatment arm, since it could reveal either further recovery or late deterioration and complications such as shunt-dependent hydrocephalus, or the longer-term impact of tracheostomy and gastrostomy dependence, which are highly relevant to how patients and families weigh the trade-off between survival and severe disability. Second, functional outcome is reported descriptively by mRS grade: although the overall distribution of mRS grades was compared between groups, we did not perform a formal comparison of the proportion of patients free of severe disability or death, nor an ordinal (proportional-odds) shift analysis yielding a common odds ratio. Identifying the minority of patients who can achieve independence after critical PBSH is a priority that our sample is too small to address.

Finally, all participating centers were located within a single province, and the findings may not generalize to settings with different surgical volumes, operator experience or end-of-life care practices. Prospective multicenter studies with larger cohorts, extended follow-up beyond one year incorporating standardized cognitive and quality-of-life evaluations, and finer hematoma volume stratification are warranted to validate these findings, refine surgical indications and further establish the therapeutic window for robot-assisted intervention in critical PBSH.

Conclusions

• Robot-assisted stereotactic puncture and drainage for critical PBSH demonstrates a favorable safety profile, without increasing the risks of rebleeding, intracranial infection, pneumonia, or stress ulcer.
• Early minimally invasive aspiration was associated with higher 30-day survival and facilitated early recovery of consciousness; the 90-day survival benefit was statistically significant in the 5–10 mL stratum (n = 63) but not in the overall cohort, and the >10 mL stratum (n = 16) was underpowered and inconclusive.
• Initial hematoma volume is an independent risk factor for 90-day mortality in patients with PBSH, whereas robot-assisted puncture was independently associated with lower 90-day mortality. This survival advantage was not accompanied by a demonstrated improvement in 90-day functional outcome, and longer follow-up with standardized quality-of-life assessment is required to determine whether it translates into a net benefit for patients.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Roberto Colasanti, Maurizio Bufalini Hospital, Italy

Reviewed by: Xinjie Hong, Naval Medical University, China

Ahmad Mohammad Elsharkawy, Zagazig University, Egypt

Footnotes

PBSH, Primary brainstem hemorrhage; GCS, Glasgow Coma Scale; mRS, Modified Rankin Scale; HR, Hazard ratio; CI, Confidence interval; CT, Computed tomography; SpO₂, Peripheral oxygen saturation; PaCO₂, Partial pressure of carbon dioxide in arterial blood; PSM, Propensity score matching.