Vicky Andreou Chalos

Ischemic stroke occurs when blood supply to a part of the brain is reduced or disrupted. It causes acute neurological deficits and results in high mortality and morbidity worldwide. In order to prevent irreversible damage of the brain tissue, patients should be treated as soon as possible to recover blood flow (‘time is brain’). For over two decades, intravenous thrombolysis (IVT) with alteplase had been the standard of care for all patients without contra-indications. In 2015, endovascular treatment (EVT) was proven effective and safe in patients with a large vessel occlusion of the anterior circulation. Although this resulted in a drastic improvement of functional outcome at group level, almost half of these patients do not recover to functional independence. Outcome prediction for individual patients remains challenging, as outcomes vary between patients and are based on multiple factors. Clinical prediction models combine multiple characteristics to estimate the probability of an outcome of interest for individual patients. They can be used by physicians to select which patients may benefit from EVT, but also to inform physicians, patients and their family members about the extent of recovery they can expect after EVT has been performed. In addition, it clearly recognized that EVT needs to be refined to further improve outcome. Optimizing macrovascular and microvascular reperfusion by modifying the pre-procedural intravenous alteplase and peri-procedural antithrombotic treatment strategies, could potentially improve outcomes. New evidence from clinical studies is needed to assess the potential benefit of these two approaches. This includes data from observational research, but ideally from randomized controlled trials (RCTs). Next to defining the primary objective of an RCT, important considerations in the design and execution of future RCTs regarding EVT include the selection of a primary outcome measure and the informed consent procedure.

The overall aim of the research presented in this thesis was to predict outcome of patients with ischemic stroke (Part I) and to find ways to further improve outcome of EVT by optimizing reperfusion (Part II) and by improving trial design and execution (Part III).

Specific research questions were:
1. Is sex predictive of outcome and of the effect of EVT?
2. One day after EVT, which post-procedural clinical and imaging characteristics predict functional outcome at three months?
3. How reliably and accurately can we predict outcome for individual patients after receiving EVT?
4. What is the potential benefit and safety of pre-procedural intravenous alteplase and peri-procedural antithrombotic agents on functional outcome after EVT?
5. How can we optimize the design and execution of future randomized controlled trials on EVT, specifically focusing on the choice of a primary outcome measure and on the informed consent procedure?

Part I. Prediction of outcome after endovascular treatment
Based on previous subgroup analyses, it remained unclear whether women and men benefit equally from EVT. In Chapter 2, I used individual patient data from seven RCTs (n=1762) on EVT within the HERMES collaboration to study the influence of sex on functional outcome and on the effect of EVT. Functional independence (modified Rankin Scale score [mRS] 0-2) at three months was reached by 39% of women and 39% of men. The effect of EVT on the ordinal mRS was also similar in women (adjusted common odds ratio [acOR] 2.13; 95% CI: 1.47–3.07) and men (acOR 2.16; 95% CI: 1.59–2.96), with a P for interaction of 0.926.

Chapter 3 presents the development and external validation of the post-procedural prognostic model MR PREDICTS@24H. It was developed using data from the HERMES collaboration (n=718) and includes: age, baseline National Institutes of Stroke Health scale (NIHSS), pre-stroke mRS, history of diabetes mellitus, occlusion location, collateral score, recanalization after EVT, NIHSS 24h after EVT, and symptomatic intracranial hemorrhage (sICH) after EVT. NIHSS at 24h was the strongest independent predictor of functional outcome at three months. External validation with data from MR CLEAN Registry (n=3260) showed excellent discriminative ability (c-statistic of 0.91 for functional independence; 0.89 for survival). In this cohort, the observed probability of functional independence was systematically higher than what was predicted by the model (41% vs 34%), while the observed and predicted probability of survival were similar (72% vs 75%). To take into account this difference, the online tool will show different predicted probabilities reflecting different settings. One day after EVT, MR PREDICTS@24H can then be applied to reliably and accurately predict functional outcome for individual patients at three months.

In Chapter 4 I addressed shortcomings in model development and clinical applicability of a post-procedural prognostic model that included gender, age, diabetes mellitus, follow-up infarct volume and smoking status. Prognostic models should be simple and comprising characteristics that are routinely available in clinical practice, to increase their clinical applicability. Follow-up infarct volume, requires follow-up MRI, which is not routinely available in many countries, such as the Netherlands. In addition, their findings should be interpreted with extreme caution, as timing of predictor and outcome assessment were unclear, a small sample size was used, and no internal nor external validation were performed.

Part II. Optimizing reperfusion at a macrovascular and microvascular level
In Chapter 5 baseline characteristics and outcomes including recanalization rate, sICH, and functional outcome, were compared between patients treated with IVT + EVT and those treated with EVT alone in the MR CLEAN Registry (n=1485). Baseline characteristics of both groups differed: patients treated with IVT + EVT had atrial fibrillation less often and had better pre-stroke mRS scores than those treated with EVT alone. Successful recanalization and sICH did not differ between both groups. In regression analysis with covariate adjustment IVT + EVT was associated with better functional outcome (acOR 1.47; 95% CI: 1.10-1.96), similar to the results obtained with propensity score matching. These results may have been a true benefit of IVT prior to EVT, but its interpretation was considered to be hampered by the possibility of residual confounding. Therefore, RCTs were necessary to properly assess the effect of IVT prior to EVT.

Chapter 6 provides a systematic review of existing literature to evaluate the potential safety and functional outcome of peri-procedural antiplatelet agents and heparin during EVT. The search yielded 837 studies, of which 19 were included; 14 on antiplatelet agents, 4 on heparin, 1 on the combination of both treatments. RCTs were lacking. Only four observational studies reported relative effects on sICH, of which two pointed towards an increased risk of sICH. The three studies presenting relative effects of treatment with antiplatelet agents on functional outcome showed neutral effects, while positive relative effects of heparin were found on functional outcome in two studies. We recommended that RCTs are warranted to address the question whether the potential effect of these two antithrombotic agents on functional outcome could outweigh the increased risk of sICH.

The findings from Chapters 5 and 6 were used to design new RCTs within the CONTRAST consortium, aiming to improve functional outcome after EVT. The research protocol and results from one of these RCTs, the MR CLEAN-MED, were described in Chapter 7 and 8 respectively. The primary objective of this RCT was to assess the safety and efficacy of peri-procedural intravenous acetylsalicylic acid, unfractionated heparin, both or neither during EVT in 1500 patients with ischemic stroke. The trial was prematurely stopped (n=633) because of safety concerns. Colleagues and I found that peri-procedural acetylsalicylic acid and unfractionated heparin during EVT were both associated with an increased risk of sICH (acetylsalicylic acid: aOR 1.95; 95% CI: 1.13-3.35, unfractionated heparin: aOR 1.98; 95% CI: 1.14-3.36), while there was no beneficial effect on functional outcome in either treatment group (acetylsalicylic acid: acOR 0.91; 95% CI: 0.69-1.21, unfractionated heparin acOR 0.81; 95% CI: 0.61-1.08).

Part III. Improving the design and execution of future randomized controlled trials on endovascular treatment
The most commonly used primary outcome measure in stroke treatment trials is the mRS at three months. However, it has some limitations, including its requirement for long-term follow-up interviews, which consume time and resources. In Chapter 9 I used data from two RCTs on EVT, a positive (MR CLEAN) and a neutral one (IMS III), and used a causal mediation model to evaluate whether the early postprocedural NIHSS at 24 hours and 5-7 days after EVT could serve as an alternative primary outcome measure. All requirements for a surrogate endpoint were satisfied. First, NIHSS was significantly correlated with mRS in both trials. Second, in IMS III, no treatment effect was found on NIHSS, while in MR CLEAN a significant treatment effect was found on the NIHSS. Third, after adjustment for NIHSS at 24h and 5-7 days in MR CLEAN, treatment effect of EVT on mRS decreased from cOR 1.68 to 1.36 and 1.21 respectively. I therefore concluded that NIHSS within one week may be used as a primary outcome measure in phase II(b) trials on EVT, which could reduce trial duration and costs.

Informed consent is a core ethical and legal prerequisite for enrolling patients in an RCT. However, in research concerning patients with ischemic stroke it is challenging to obtain informed consent as they might have neurological deficits that may interfere with their capacity to decide about trial participation. In Chapter 10 I assessed the proportion of patients undergoing EVT in the MR CLEAN Registry (n=1526) with neurological deficits that may impede the capacity to provide informed consent for participation in trials on EVT. Colleagues and I formulated a strict and mild capacity assessment rule using different cut points in the individual item scores in the NIHSS. Using the strict and mild rule, at hospital admission the majority of patients – respectively 96% to 80% – were deemed incapable for decision making, while this was 79% to 60% for patients 24-48 hours after undergoing EVT. In Chapter 11 I then analyzed which proportion of patients in the MR CLEAN trial (n=500) were included with patient vs proxy consent. Only 16% of patients who underwent EVT provided their own consent, meaning in that 84% of patients proxy consent was necessary. Patients enrolled with proxy consent were older, had more severe neurological deficits, and worse functional outcomes. In addition, time from CTA-to-randomization was approximately one hour in both groups, which suggests that written prior informed consent takes time and may delay the start of EVT.

Discussion
The aim of this thesis was to predict outcome and to find ways to further improve outcome after EVT for ischemic stroke. Subgroup analysis in an individual patient pooled meta-analyses showed that sex does not influence functional outcome after EVT and that EVT is equally effective in men and women. EVT should therefore not be withheld based on the single characteristic of sex. I further found that recanalization, NIHSS at 24h, and sICH at 24h are post-procedural characteristics one day after EVT, that are independently associated with functional outcome at three months. These three variables are included in the developed and externally validated prognostic model MR PREDICTS@24H, which gives reliable and accurate predictions of functional outcome. The online calculator may be used as a complementary tool in clinical practice to provide treating physicians, patients, and their family members with reliable outcome expectations, and assist physicians in personalizing their patients’ follow-up and rehabilitation plans. It should be further externally validated and updated in contemporary patients and different patient populations, settings, and countries to optimally support stroke care. As outcome remains poor in a substantial proportion of patients who underwent EVT despite rapid and successful recanalization, I also aimed to find ways to improve macrovascular and microvascular reperfusion. An observational study showed better functional outcomes in patients undergoing IVT + EVT than EVT alone, but these results may have been hampered by residual confounding. Consequently, over the past few years several RCTs outside this thesis were conducted and pooled together to acquire more reliable evidence on the use of IVT prior to EVT. No benefit of EVT alone in patients directly admitted to an EVT-center was shown. A systematic review including only observational data revealed that a potential effect of antithrombotic agents including acetylsalicylic acid and unfractionated heparin on functional outcome – possibly through improving incomplete microvascular reperfusion – could outweigh the increased risk of sICH. However, the MR CLEAN-MED, an RCT testing this hypothesis, was preliminary stopped because of increased risk of sICH for both treatments without better functional outcomes. Acetylsalicylic acid and unfractionated heparin should therefore not routinely be administered in patients during EVT. Next, I suggested that future phase II(b) RCTs may use the NIHSS at 24h or 5 to 7 days as a surrogate outcome measure for mRS at three months, in order to reduce trial duration and costs, as all requirements for a surrogate endpoint were met. A debate in the field with all parties is required to assess whether and under which conditions it could be used in phase III trials. Finally, the majority of patients eligible for EVT have neurological deficits that impede the capacity to provide informed consent for trial participation, and both written prior patient and proxy consent seem to delay the start of acute treatment. Therefore, alternative consent types, such as deferred consent, should be more often considered in future RCTs. This may reduce treatment delays, and increase generalizability of trial results, patient recruitment rates, and patient autonomy. Future research should focus on the necessity of deferred consent, its execution, and on its acceptance by treating physicians, researchers, patients and proxies.