Introduction

Malignant cerebral infarction (MCI) is associated with poor functional outcome and often death. For more than 35 years, surgical treatment of space-occupying cerebral edema has been suggested to relieve raised intracranial pressure [1]. Nevertheless, only since the seminal publication of Vahedi et al. [2] in 2007 has decompressive craniectomy (DC) been regarded as an effective treatment for MCI. Since then, several studies have confirmed the usefulness of DC [3,4,5,6,7,8]. Although no clinical trial has proven its usefulness, infratentorial suboccipital decompressive craniectomy for malignant cerebellar infarction is also likely associated with a greater survival and an improved outcome [9, 10]. However, there are still some hesitations and disparities among neurosurgeons and neurosurgical centers, regarding the possibility of leaving patients in a poor functional state [11]. Moreover, DC is a very invasive procedure, associated with a significant risk of complications [12]. If the patient survives, DC is best followed by a cranioplasty a few months later, which itself is also an invasive surgical treatment [13].

Almost all studies on DC published to date concern fairly small cohorts, often comprising less than 50 cases. Moreover, few have addressed long-term survival, with observational studies and randomized trials often having follow-up periods limited to up to one year [7,8,9].

Objectives

The purpose of this study was to assess the evolution of practices and long-term survival of patients after DC for brain ischemia and associated prognostic factors.

Materials and Methods

We performed a nationwide descriptive and analytical observational retrospective study. Data were extracted from the French medico-administrative Programme de Médicalisation des Systèmes d’Information (PMSI) national database, which collects discharge abstracts from all public and private hospitals in France. All adult patients (> 18 years) who were admitted primarily for an ischemic stroke (IS) and who underwent a DC between 2008 and 2017 were included. We used an algorithm which combines two variables to retrieve appropriate cases within the PMSI [14]. The first variable was associated with the surgical procedure DC (LAFA900) identified by the “Common Classification of Medical Acts” (CCAM) which describes medical and surgical interventions for both hospital and ambulatory care (http://www.atih.sante.fr/ccam-descriptive-usage-pmsi-2017). The second variable related to the primary diagnosis of cerebral infarction (I63; I63.6: venous thrombosis, nonpyogenic excluded) according to the International Classification of Diseases (ICD-10) [15]. We excluded patients with secondary cerebral ischemia after brain surgery, traumatism, subarachnoid hemorrhage, cardiac arrest, drowning, etc. Both cerebral (supra-) and cerebellar (infratentorial) MCI were considered together as they could not be differentiated. Therefore, suboccipital DCs for cerebellar infarction were also included. The same method was applied to identify patients who underwent a cranioplasty (CCAM code = LAMA009). In the PMSI, a patient is identified by a unique pseudonym, regardless of the institution where he or she was admitted or treated, or the year in which he or she was treated.

Statistical Methods

Continuous variables are reported as means and standard deviations or as medians and interquartile ranges (IQR) for non-Gaussian distributions; categorical variables are reported as frequencies and proportions. Survival analyses were conducted to delineate factors associated with survival. The primary outcome was defined as death (event). Survival statistics were based on time to death, which was measured from the date of DC to the date of the last follow-up or death [16]. We used the Kaplan–Meier method to estimate the event-free survival and the Mantel Cox log-rank test to compare survival curves. Cox proportional hazards regressions were used to identify predictors of death and to estimate hazard ratio (HR) with their 95% confidence intervals (95% CI) [17]. The PMSI database, created initially for health-care costing, registers all in-hospital medical acts and deaths. A patient who was not admitted into any medical institution was considered to be lost to follow-up and right-censored in the survival analysis at the time of his or her last hospital discharge or medical intervention. The total number of patients available at 1, 2, and 5 year follow-up are displayed as “number at risk” in survival figures. All tests were 2-sided, and statistical significance was defined with an alpha level of 0.05 (p < 0.05). The proportional hazards assumption was explored using statistical tests based on the scaled Schoenfeld residuals. Analysis was performed with the R programming language and software environment for statistical computing and graphics (R version 3.6.0 (2019-04-26)), specifically the survival package, among others [18, 19]. The statistical programme and workflow were written in R Markdown v2 with RStudio® for dynamic and reproducible research [20].

Compliance with Ethical Standards

This study was conducted according to the ethical guidelines for epidemiological research, Helsinki Declaration (2008), to the French data protection authority (CNIL authorization number: 2008538) and to the European the General Data Protection Regulation (EU) 2016/679. The data within the database were de-identified prior to analysis, and thus, informed consent from patients was not required. The study is reported in accordance with the RECORD and the SAMPL guidelines [21, 22].

Results

Incidences

A total of 1841 DCs for MCI were identified over the 10-year period. Fifty-one centers were involved with a median annual activity of 2.5 procedures IQR (1, 5.2). Eight centers performed only one procedure during the 10 years and, on the contrary, 16 centers performed at least one DC annually. Despite being variable, the number of centers involved rose significantly from 28 in 2008 to 37 in 2017 (p = 0.030). The "high activity centers," those performing four or more DCs per year (30.8%), accounted for 68.5% of procedures. University hospitals performed 84.3% of DCs. There has been a significant increase in the use of DC for IS over time (p < 0.001). Between 2008 and 2017, the annual number of procedures performed more than doubled (95/year vs. 243/year).

Population Description

Mean age at the procedure was 50.9 ± 10.5 years, and 18% were above 60 years (Table 1). A majority (64.4%) of the patients were male, and patients were older at the time of surgery compared to females (51.9 vs. 49.1, p < 0.001). Over the 10 years, the mean age at the time of surgery significantly increased for both genders (48 years vs. 51.5 years, p < 0.001). At stroke onset, 53.3% of the patients presented with aphasia or dysarthria, and 84.1% with mono- or hemiplegia. Consciousness was impaired in 66.3%, and 51.5% of the patients were comatose before the craniectomy (Glasgow Coma Scale [GCS] < 8). Median time between hospital admission and DC was 1 day, IQR [0, 1]. 76.5% of the patients had the DC performed within the 24 h after their admission and 88.3% within 48 h. Median follow-up was 1.3 years, IQR [0.1, 3.3].

Table 1 Patients characteristics
Full size table

,

Early Survival

A total of 371 patients (20.2%) died within the month following DC (Figs. 1, 2), and median time between DC and death was 8 days, IQR [4, 52]. Mean age at death was 53.7 ± 10.2 years. Regardless of the outcome status, most patients (79.8%) were admitted to intensive care unit for a median duration of 12 days, IQR [6, 22]. Early overall survival (OS) at one week, one and 6 months were 86%, 95%CI [84.5, 87.6], 79.7%, 95%CI [77.8, 81.5], and 74.8%, 95%CI [72.8, 76.8], respectively (Figs. 1, 2). Among the patients who survived, 58.1% were transferred directly to rehabilitation. Median hospital stay of the survivors was 35 days, IQR [24, 50.25].

Fig. 1
figure 1

Diagram of patients with DC for malignant brain infarction

Full size image
Fig. 2
figure 2

Five-year OS of patients after DC for malignant brain infarction

Full size image

Cranioplasty

A total of 1063 patients (57.7%) had a cranioplasty insertion. Median time between the craniectomy and the cranioplasty was 5.2 months, IQR [3.4, 7.8]. For those who survived the DC, the probability of having a cranioplasty at one year was 75.6%, 95%CI [77.9, 73.1] (Fig. 3d). Fifty-six patients (5.3%) who previously had calvaria reconstruction ultimately died after a median time of 1.6 years, IQR [0.6, 4.2] following the DC. Thirty-four other patients (3.2%) had the cranioplasty removed following surgical complications (infection). At the end of the study, 973 patients (52.9%) were alive with a cranioplasty.

Fig. 3
figure 3

Survival curves comparison. a Survival by category of age. b Survival by consciousness status. c Survival by center surgical activity. d Cumulative incidence plot for cranioplasty

Full size image

Long-Term Outcome

At one year following DC, 473 patients (25.7%) had died, 374 (20.3%) were in rehabilitation, and 32 still in hospital. At data collection, 529 patients (28.7%) were deceased. Only 60% who had been operated on in 2008 were still alive compared to 77% for those who underwent a DC in 2017. OS at 1, 2, and 5 years was 73.6%, 95%CI [71.6, 75.7], 71.9%, 95%CI [69.8, 74.1], and 68.9%, 95%CI [66.5, 71.4], respectively (Fig. 2). Seventy-four patients (5.6%) remained bedridden (modified Rankin Scale [mRS] = 5).

Prognostic Factors

In the multivariable Cox regression analysis, age at DC below 60 years (HR = 0.5; 95%CI [0.4, 0.7], p < 0.001), DC being performed in a center with a high surgical activity (HR = 0.8; 95%CI [0.6, 0.9], p = 0.002), and the patient having unimpaired consciousness (HR = 0.6; 95%CI [0.5, 0.8], p < 0.001) were identified as independent prognostic factors of OS (Table 2, Fig. 3).

Table 2 Univariate Cox regression for OS after DC for malignant brain infarction
Full size table

Discussion

Strengths and Limitations

The strengths of the PMSI database reside both in the high number of patients and in the exhaustivity of data. Previous works have demonstrated its value in the nationwide assessment of medical questions [14, 23,24,25,26]. However, several potential factors affecting survival are not recorded in the PMSI as well as the functional status (mRS) with the exception of bedridden patients. Neither the CCAM nor the ICD-10 includes an artery or an anatomy-based classification of stroke. In the PMSI, there is a lack of information regarding the affected vascular territory. Therefore, hemispheric and posterior fossa craniectomies are mixed together. Nevertheless, this nationwide analysis, the first of its kind, provides insight into treatment practices and outcomes of DC for MCI in France. The use of an exhaustive, nationally representative database reduces institutional bias, particularly of university medical centers, and increases the generalizability of our findings which may represent robust long-term survival data to serve as a reference for further research in stroke.

Incidence and Practices’ Evolution

Between 2008 and 2014, the number of patients hospitalized for stroke increased by 13.7% [26]. Our results, consistent with previous findings, also show a significant increase in the number of DC by 161.3% over the 10-year study period (p for linear trend < 0.001), confirming that there is a wider acceptance among surgeons to perform such procedure for patients with MCI [27, 28]. Walcott et al. [28] found a prevalence of DC for MCI of 38.6 for 100 000 hospitalizations in 1999–2000 versus 144.6/100 000 in 2007–2008. According to Lecoffre et al. [26], 82 912 patients were hospitalized for IS in France in 2014. Of those, only 218 (0.3%) had a DC according our findings versus 0.7% in Suyama et al. report [29].

Population Description

The 1841 patients included in our study had a mean age of 50.9 years which is much younger compared to those enrolled in the studies by Suyama et al. (67 years) and Dasenbrock et al. (55.9 years) [29, 30]. In the pooled analysis by Vahedi et al., median age at the time of surgery was lower (43.2–51.6 years). However, this is likely explained by the inclusion criteria being restricted to 55 or 60 years [6].

In France, for the 2008 to 2014 period, mean age at IS was 74 years [26]. This highlights that those who underwent a DC are about 13 years younger than the general IS population. We confirm that DC is a procedure more frequently performed in males 64.4% (sex ratio of 1.8) versus 59.5% and 59% as found in the previous studies [29, 30]. In most randomized trials, there was no gender imbalance apart from the HAMLET trial which had 64% male patients [6,7,8]. Men are more likely to suffer strokes with a sex ratio of 1.55 95%CI [1.48–1.61] with an age-standardized incidence of hospitalized patients of 153.9 per 100 000 inhabitants for males versus 102.0 for females [26, 31]. This likely explains the gender disparity observed in our analysis. Median hospital stay of the survivors of 35 days IQR [24, 50.25] was longer compared to Dasenbrock et al. findings of 17 days IQR [12, 25].

Outcome

Global cerebral infarction 30-day mortality has previously been estimated to be 11.1% [32]. Thirty-day mortality after DC was 18.6% for Suyama et al. compared to 20.2% in our study, i.e., a survival probability at one month of 79.7%, 95%CI [77.8, 81.5]. The survival curve shows that the risk of death was the highest within the first week following DC (Fig. 2). Our findings are consistent with the previous results which showed a high in-hospital mortality rate of 25.3% [30]. The effect of DC on OS observed at 6 months of 74.8%, 95%CI [72.8, 76.8] appears to be sustained long-term (5-year survival = 68.9%, 95%CI [66.5, 71.4]) [7]. In the pooled analysis by Vahedi et al., the absolute risk reduction for death was 49% in the DC cohort, with a probability of survival beyond 1 year after DC of 78% which is slightly better higher than our result of 73.6%, 95%CI [71.6, 75.7] [6].

Apart from bedridden patients, the long-term functional status is not recorded in the PMSI. It is tempting to extrapolate outcome results to confirm suggestions from non-randomized studies that DC reduces mortality and also increases the number of patients with a favorable outcome, i.e., 43% of patients mRS ≤ 3 [6]. However, Gul et al. suggested that there was no significant difference between DC and medical treatment in terms of proportion of survivors with an unfavorable outcome. Knowledge about the long‐term functional outcome after stroke is limited. Most reports agree on relatively stable levels of disability and activities of daily living dependency (mRS ≥ 4) over long-term follow-up being between 20 and 30% for survivors [33,34,35]. After IS, two-thirds of the patients were dead or functionally dependent at 5 years according to Sennfalt et al. [32]. No long-term comparison of DC versus medical treatment has been performed, but it raises several ethical questions as to the risk to leave the patient in a vegetative or severely disabled state. Thus, the decision to surgically treat a patient with a malignant IS needs to be made on an individual basis, according to the patient’s wishes.

Factors Affecting the Survival

Timing of Surgery

Optimal time for surgical decompression remains uncertain despite most studies showing that early DC reduces mortality [36]. In the pooled analysis by Vahedi et al. [6], all patients were treated within 48 h, and no difference in the outcome was found between patients operated on the first and those operated on the second day. Timing of DC was not associated with in-hospital mortality for Dasenbrock et al., but later surgery increased the odds of a poor outcome [37]. Suyama et al. [29] also found that early DC was not an independent factor for 30-day mortality. We did not find a positive association. On the contrary, in univariate analysis, the patients who underwent DC within 24 or 48 h had reduced survival (HR = 1.4, 95%CI [1, 1.9], p = 0.05, Table 2). However, there may be a complex interaction between the timing of DC and survival as the proportion of patients who died is not significantly different in both groups (χ2p = 0.11). The vast proportion of our patients (88.3%) had DC within 48 h versus 55.8% and 63.1% in Dasenbrock et al. and Suyama et al. reports, respectively. The fact that there is no favorable effect of early DC is presumably because the patients treated earlier had larger IS and faster swelling than those who were operated on later. In our study, there were significantly fewer missing data on the timing of DC in 2017 compared to those in 2008 (χ2p < 0.001). In addition, considering the significant level of missing data, the year of DC was predictive of an early intervention. 85.9% of the procedures were performed within 48 h in the 2008–2012 period versus 89.6% for the years 2013–2017 (χ2p = 0.03). As these cause both a selection and an information bias, the covariate “timing of DC” was not considered in the multivariable analysis. Patient age was not associated with the timing of surgery (χ2p = 0.07).

Age

Age at DC was found to be an independent factor affecting survival (HR = 0.5; 95%CI [0.4, 0.6], p < 0.001) [28, 30]. There was a trend for reduced survival alongside the increase in age, especially above 60 years (Fig. 3a). In the pooled analysis, DC was beneficial in all subgroups (< 50 or > 50 years) [6]. Patients > 60 years experienced higher odds of mortality and poorer outcomes compared with younger patients [30]. Only 56.8% of the patients above 60 years were still alive at data collection, compared to 74% for those under 60 (χ2p < 0.001). Even if the prognosis worsens as the patients get older, patients above 61 years may still benefit from a DC [8]. For Suyama et al. [29], older age was not found to be an independent risk factor for mortality. Thus, the appropriate age-group for surgical intervention remains debated. The patient’s age by itself should not be a limiting factor, but it should be taken into account as most survivors aged 60 years or older have major disability (mRS 4 or 5) after DC for MCI [8]. The possibility of end-of-life decision after DC may also have influenced the in-hospital mortality, especially for senior patients with prolonged stay in intensive care units.

Surgical Activity

Despite being a relatively simple neurosurgical procedure, DC is often performed on unstable and fragile patients who require specific management in dedicated intensive care units. OS was not greater in academic centers (HR = 1.1; 95%CI [0.9, 1.4], p 0.490) but greater in trained centers performing four or more DCs per year (HR = 0.7; 95%CI [0.6, 0.9], p < 0.001). Few data exist on DC activity by neurosurgical centers. For Suyama et al., the mean surgical activity was 2.9 procedures per center in 2011 vs. 2.9 procedures per center and per year in our study. Further efforts should be made to improve regional stroke schemes and prompt allocated management for patients suffering acute brain ischemia, especially from the neurosurgical point of view. As outcomes tend to be poorer in inexperienced centers, DC should be preferably centralized and performed in trained institutions.

Consciousness

Consciousness level at the time of surgery was established as independent prognostic factors of OS (HR = 0.6; 95%CI [0.5, 0.8], p < 0.001). This result is consistent with Suyama et al. [29] findings of greater mortality among patients in comatose status (GCS < 6). In IS patients, a progressive coma usually indicates raised intracranial pressure and an associated ongoing brain injury. Therefore, once the decision has been taken to perform a DC, the procedure must be performed rapidly regardless of the GCS of the patient, save for a fully awake and responsive patient (GCS 15), although this clinical presentation is very unlikely for a large brain infarction.

More than half (51.5%) of the patients were comatose before surgery although more than 3/4 of procedures were performed within 24 h. This is unusual for the course of an MCI and reflects the PMSI’s imprecision for assessment of timing of DC. This aligns with the information bias associated with the covariate “timing of DC” which is not reliable for determining the precise onset time of a stroke. Moreover, this is in contrast to the criteria of most randomized trials and DC guidelines, with a low GCS usually constituting an exclusion criterion. As favourable criteria for DC remain uncertain, a consensus on what clinical and radiological factors to take into account needs to be reached.

Cranioplasty After DC for Stroke

As for the long-term-survival, no study has addressed specifically the question of cranioplasty after DC for MCI. A growing number of survivors usually with several comorbidities and an impaired functional status require a cranioplasty at some point. Patients with a good functional status (mRS ≤ 3) may benefit from skull vault reconstruction, at least from the aesthetic and protective points of view. Despite a higher risk of surgical complications (wound healing), an early cranioplasty within 3 months of DC may improve the functional outcome [38]. 18.7% of the survivors had the cranioplasty inserted within 3 months, 57.8% in the 6 months, and the probability of having a cranioplasty at one year was 75.6%, 95%CI [77.9, 73.1] (Fig. 3d). For those who remain bedridden or in a poor functional status (5.6%), the necessity of a cranioplasty is debatable.

Implications of this research for practicing clinicians are that DC should be preferably assessed in dedicated stroke units and performed on patients below 60 years before they become comatose. When a potential candidate is identified, the patient’s relatives, and if possible, the patient him/herself must be aware and consent to this very invasive procedure—a procedure in which one-fourth of patients will not survive. Moreover, those who may survive are at risk to be left in a poor functional state.

Despite advances in stroke care, long-term prognosis remains a cause for concern [32]. Clinical trials performed in small populations and often across limited periods of time may produce different results compared to the “real-world clinical scenario.” Implementation of the Système National des Données de Santé (SNDS) and of the Health Data Hub (HDH) may provide an effective tool for assessing long-term comparison of DC versus medical treatment, which remains the foremost question to be answered. Moreover, linkage between the SNDS/HDH and clinical databases such as the DECIMAL may offer further opportunities to address such uncertainties [2].

Conclusion

In France over the past 10 years, DC has been increasingly performed for MCI regardless of age. However, in-hospital mortality remains considerable, as about one quarter of patients died within the first weeks. For those who survive beyond 6 months, the risk of death significantly decreases.

Early mortality is especially high for comatose patients above 60 years operated in inexperienced centers. Most of those who remain in good functional status tend to undergo a cranioplasty within the year following DC.