Introduction

Randomized controlled trials (RCTs) are considered the gold standard for investigating the relationships between medical or surgical interventions and outcomes as well as comparing different treatment modalities [21]. The main hallmark of a RCT is the randomization process, which is, if successfully accomplished, the most reliable and efficient way of forming groups without systematic differences in their baselines characteristics, i.e. free from selection bias. Another important step to reduce bias is to implement blinding of patients, medical personnel and outcome assessors. These measures can reduce performance and detection biases, although especially in surgical RCTs effective blinding may not be possible [9]. Additionally, modern RCTs have to be registered in publicly available databases, such as ClinicalTrials.gov, to meet law and publication requirements. Registration before beginning of a RCT assures transparency to participants and the research community, provides reliable information to potential participants and referring clinicians and helps reduce publication bias [29].

However, despite their inherently higher risk of bias, other forms of study (such as observational studies) are considered to have certain advantages over RCTs, such as lower cost, greater timeliness and a broader and thus more realistic range of patients [1]. Comparisons of results from RCTs and preceding non-randomized trials suggest consistency in at least 60% [22].

With the propagation of evidence based medicine, the available published data acquired through RCTs increased significantly [16]. Of note, the risk of study discontinuation and presumably also the non-publication rates of such results are reported to be substantial in surgical RCT compared to medical RCTs [28]. While this certainly indicates higher obstacles when exploring surgical treatments with RCT methodology, these findings also reveal a significant publication bias in surgery.

Despite these intricacies, medical, surgical, and interventional RCTs have significantly influenced the treatment of patients with ischemic stroke: While recanalization with intravenous, intraarterial or mechanical thrombolysis is strongly associated with improved functional outcomes and reduced mortality, decompressive craniectomy (DC) has been shown to reduce mortality in those patients who still suffer a space-occupying territorial cerebral infarction [5, 23]. Aim of the present study was to explore the extent and timeline of the translation from RCTs to routine patient care and the subsequent impact on neurosurgical practice.

Materials and methods

We have hypothesized that publication of a RCT would influence routine patient care, and that significant shifts in major ischemic stroke outcomes would ultimately be reflected in the number of patients requiring DC for space-occupying territorial cerebral infarction, as we assume the majority of those patients to represent failures of preceding recanalization attempts.

Since RCTs for intravenous and intraarterial thrombolysis were mainly published before the widespread implementation of DC and an association is thus undetectable, the present study focused on RCT investigating DC and mechanical interventional thrombolysis. The PubMed database (https://www.ncbi.nlm.nih.gov/pubmed, accessed on 22nd October 2017) was searched for “ischemic stroke” and “cerebral infarction” and search results were automatically filtered to include only “randomized controlled trials” published between January 2000 and October 2017. The search results were then manually screened for RCTs investigating DC and mechanical interventional thrombolysis. The reference lists of relevant meta-analyses and systematic reviews were searched for additional RCTs.

The annual number of decompressive craniectomies for ischemic stroke performed between January 2000 and October 2017 at our institution, which is a tertiary referral hospital delivering neurosurgical care to a large urban catchment area with approximately seven certified stroke units and ten to fifteen general neurological departments, was calculated in an anonymous, retrospective fashion. The electronic hospital reporting is based on the German version of the International Statistical Classification of Diseases and Related Health Problems (ICD-10; https://www.dimdi.de/static/de/klassi/icd-10-gm/kodesuche) and on the German version of the International Classification of Procedures in Medicine (ICPM; https://www.dimdi.de/static/de/klassi/ops/kodesuche/onlinefassungen/opshtml2018/index.htm). Employing these classification systems, annual patient numbers undergoing DC for ischemic stroke were retrieved by using the ICD-10 codes I63.0 to I63.9 and the ICPM codes 5-012.0, 5-010.1× and 5-010.10 to 14.

Results

Between 2008 and 2009, we have observed a sharp increase in annual decompressive craniectomies for stroke from 26 to 43, respectively (Fig. 1). The numbers remained relatively stable between 2008 and 2015, with a median annual number of 45 operations. Between 2015 and 2016, a sharp decline was evident from 45 to 33 operations per year.

Fig. 1
figure 1

Annual numbers of decompressive craniectomies for stroke at our institution, correlated with publication years of RCTs on DC and thrombectomy, blue boxes indicate publication of RCTs on DC and yellow boxes indicate publication of RCTs on mechanical thrombectomy

Full size image

We have identified three RCTs investigating the role of DC in space occupying cerebral infarction [10, 15, 31]. A previously published meta-analysis of the data from these trials suggested that in patients with malignant cerebral infarction, DC reduces mortality with a number needed to treat of two for survival irrespective of functional outcome [30]. The results were deemed highly consistent across these RCTs. We have identified nine RCTs investigating the role of mechanical thrombectomy in acute ischemic stroke [2,3,4, 7, 14, 18, 24, 25, 27]. These RCTs uniformly reported positive results and demonstrated superiority of endovascular clot retrieval over the previous standard of care. Mechanical thrombectomy is a highly effective treatment modality for ischemic stroke with a number needed to treat of less than three for improved functional outcome [8]. It has therefore been quickly implemented and made widely available as a new standard of care for acute stroke patients in several countries, including Germany [19].

The dynamics of annual decompressive craniectomies for stroke correlate with the publication of RCTs, with a presumed impact on neurosurgical practice detectable with a latency of approximately 1 year (Fig. 1): The first two RCTs demonstrating efficacy of DC for space occupying cerebral infarction were published in 2007, followed by a sharp increase in annual operation numbers between 2008 and 2009 [15, 31]. The first RCT on mechanical thrombectomy was published in December 2014, with four further RCTs being published in 2015 [2, 4, 7, 14, 27]. Subsequently, we observed a decline in annual decompressive craniectomies for stroke in 2016.

Discussion

The present study demonstrates an increase in annual numbers of decompressive craniectomies for ischemic cerebral infarction in temporal relationship with the publication of the first two RCTs suggesting a reduction in mortality, with a latency of approximately one year. Conversely, the publication of the initial five RCTs revealing superiority of mechanical thrombectomy over previous recanalization approaches was followed by a decline in annual decompressive craniectomies for stroke, suggesting a reduction of space occupying cerebral infarctions after implementation of mechanical clot retrieval. Similar time trends can also be observed in the data on DC published by Lammy and co-authors [20].

Our analysis suggests that results from positive or partially positive RCTs are quickly translated into clinical practice, similar to previous reports in different medical fields [6, 11, 26]. However, the latency of implantation into clinical routine appears to vary between few years, as reported by us as well as Hudgins et al., and several years, as described by Ketley and Woods [11, 17]. This variation might be related to the extent of change in outcomes ascribed to the “new” treatment, or in other words, whether the results of the RCT are truly “groundbreaking” advances. Other factors might be the increased promotion of products with proven benefit by manufacturers and the degree of doctors’ willingness to adopt a new treatment as opposed to abandon an “established” treatment in response to positive or negative RCT results, respectively [11]. Of note, negative results are less likely to be translated directly into clinical practice, in addition to the well-established publication bias regarding negative findings [12, 13]. Concerns have been raised for surgical RCTs in particular, with higher discontinuation and non-publication rates compared to medical trials [28]. Another factor could be the limited generalizability of RCT findings, which can be related to differences between patients enrolled in an RCT and those encountered in day-to-day practice as well as to limited availability of certain treatment options in the clinical routine setting [32]. However, our results suggest that these obstacles were overcome for both the DC RCTs as well as the mechanical thrombectomy RCTs.

A limitation of the present study is related to potential bias within the retrospective patient cohort undergoing DC: We assume that some patients presented too late for mechanical thrombectomy or other forms of thrombolysis or were not eligible for an interventional approach. This subgroup would not be affected by RCT results. However, presumably the proportion of patients presenting too late with already demarcated, unsalvageable cerebral infarction remains stable over time. Intravenous and intraarterial thrombolysis were implemented before DC became wide spread practice and thus no assumptions were possible concerning impact of these RCTs on operation numbers.

In summary, the present study demonstrates a close temporal relationship between publication of RCTs in acute ischemic stroke and neurosurgical practice at a tertiary referral hospital. The publication of surgical and interventional RCTs with significant positive or partially positive results was translated into clinical practice with a latency of 1 year, as reflected by annual numbers of decompressive craniectomies for cerebral infarction. These results suggest that neurosurgery, neurology, and neuroradiology in Germany offer a dynamic, evidence-based and up to date patient care with fast implementation of the latest study results into routine clinical practice.