Emergent aneurysm treatment compared with treatment on neurological improvement in patients with ruptured poor-grade aneurysmal subarachnoid haemorrhage: the TOPSAT2 RCT

Article history

The research reported in this issue of the journal was funded by the EME programme as project number 13/29/83. The contractual start date was in June 2016. The final report began editorial review in June 2020 and was accepted for publication in January 2021. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The EME editors and production house have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the final report document. However, they do not accept liability for damages or losses arising from material published in this report.

Copyright statement

Copyright © 2021 White et al. This work was produced by White et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This is an Open Access publication distributed under the terms of the Creative Commons Attribution CC BY 4.0 licence, which permits unrestricted use, distribution, reproduction and adaption in any medium and for any purpose provided that it is properly attributed. See: https://creativecommons.org/licenses/by/4.0/. For attribution the title, original author(s), the publication source – NIHR Journals Library, and the DOI of the publication must be cited.

2021 White et al.

Existing research

To assess patients systematically on the basis of their initial neurological status, various grading systems have been introduced, with the World Federation of Neurosurgical Societies (WFNS) grading system being the most widely used. 3 Patients with WFNS grades 1–3 are considered ‘good grade’. These patients mostly make a reasonable physical recovery and are usually managed aggressively with early coiling or clipping of their aneurysms. Patients with WFNS grades 4 or 5 are considered ‘poor grade’ and generally have considerably worse outcomes than those with grades 1–3. Traditionally, neurosurgical clipping of aneurysms in these patients has been deferred until their neurological status improves. This is because surgery at an early stage in this group of patients is thought to be associated with an unacceptably high risk of stroke. 4

In more recent years, intracranial aneurysms have been treated primarily by endovascular coiling [85% coiling rate in the 2013 National Confidential Enquiry into Patient Outcome and Death (NCEPOD) report]. 1 Packing the aneurysm with platinum coils through a minimally invasive endovascular route avoids the need for craniotomy and retraction/manipulation of an already oedematous brain. There is high-quality evidence favouring coiling in grade 1–3 patients. 5–7 However, there is no good-quality evidence to indicate whether coiling should be undertaken early or only after neurological improvement in grade 4–5 patients. Grade 4–5 aSAH patients are not usually considered for clipping unless they make substantial clinical improvement. The landmark, UK-led, international subarachnoid aneurysm trial (ISAT)6 compared coiling with clipping but included predominantly grade 1–3 patients. Just 5% of patients recruited into ISAT were grades 4 or 5 (most were grade 4). The small number of poor-grade patients enrolled, and probably the differential centre enrolment bias around grade, meant that no conclusions on poor-grade management could be drawn. Overall, 37% (46/123) of grade 4–5 patients enrolled in ISAT had a good outcome at 1 year (alive and independent), compared with 75% of grade 1–3 patients. 6 In the only other substantial trial of aneurysm coiling versus clipping of which we are aware, a more representative 19% (91/471) of patients were grades 4 or 5, but outcome data by individual clinical grade (on randomisation) were not presented; however, the odds ratio (OR) for poor outcome was 3.51 [95% confidence interval (CI) 2.21 to 5.68] for grade 3–5 patients combined, compared with grade 1–2 patients, on multivariable analysis. 7

The conventional management strategy for grade 4–5 patients [treatment on neurological improvement (TONI) (control) arm] incurs a risk of aneurysm rebleed. The patient outcome if a rebleed occurs prior to aneurysm treatment is dismal, as > 80% of patients have a poor outcome. 6,7 Grade 4–5 patients are also thought to have a higher aneurysm rebleed rate than grade 1–3 patients, and this risk is highest quite soon after the first bleed. 4

Therefore, reducing the chance of rebleeding by early aneurysm treatment may improve patient outcome. Based on this assumption, an early coiling strategy in grade 4–5 patients is being practised in many centres and some of the results are encouraging. 8–15 One larger (459 patients) heterogeneous population-based non-randomised prospective study found evidence that treatment of ruptured intracranial aneurysms within 24 hours of aSAH improves medium- and long-term clinical outcome. 16 The benefit of such ultra-early treatment was even more apparent for patients treated with endovascular coiling.

Review of prior literature on early coiling in poor-grade patients published from 2002 (when coiling became a proven aneurysm therapy) until 2015 identified eight relevant studies. 8–15 Unfortunately, none of these studies had a control group and most were small, retrospective studies carried out in a single centre, therefore suffering from inherent selection, review and recall bias. Overall, this was a very heterogeneous group of studies in terms of methodology, inclusion criteria, treatment timing and outcome measures used (and their timing).

Summary analysis of the eight studies identified a combined mortality rate of 36%, with good outcome in 52% (258/495). That represents an absolute 15% improvement over ISAT results, despite studies in the summary analysis including proportionately more WFNS grade 5 patients than were enrolled in ISAT. The primary and over-riding difference between ISAT and the subsequent acute aneurysm treatment literature is the timing of the aneurysm treatment.

A Chinese registry of poor-grade aSAH patients was published in early 2014. 17 This was an observational rather than a randomised study, examining outcomes rather than management strategy. A randomised poor-grade trial protocol for a single Chinese centre has also been published recently. 18 However, this proposed a trial of 99 patients examining timing of clipping in three groups of 33 (at < 3 days, 3–7 days and > 7 days), none of which is truly early aggressive aneurysm treatment. On both grounds (treatment modality and timelines), it is not comparable with the treatment of poor-grade subarachnoid haemorrhage trial 2 (TOPSAT2).

Rationale

There is genuine uncertainty about the optimal management strategy for grade 4–5 aSAH patients and a lack of high-quality research evidence in this area, confirmed by the NCEPOD report. 1

Although there is reasonable evidence that fewer rebleeds occur in patients with good clinical grades treated by coiling ultra-early, there are additional procedural risks in poor-grade patients. 23 Therefore, the management of poor-grade aSAH is based on individual or team experience, although there is a clear trend for these poor-grade patients being treated more aggressively, mostly with early coiling. This is mainly because, with the availability of coiling as a less invasive alternative to clipping, most clinicians are not comfortable with leaving a ruptured aneurysm unprotected at a stage when the risk of rebleed is greatest. However, this is not an evidence-based approach and potentially exposes health-care systems to the following considerable extra costs:

• A substantial additional demand on already stretched neurocritical care bed and staff resources.

• Long-term care costs if early coiling results in survival with major disability rather than improving the proportion of patients with a truly good outcome (alive and independent or with minor disability in the medium to long term at follow-up).

• Costs of possibly unnecessary aneurysm coiling (staff, infrastructure and consumables).

• Drive to deliver weekend coiling services locally rather than by potentially cheaper networking (networking between centres may be a good option for grade 1–3 aSAH patients but not so for grade 4–5 patients for whom extra transfer between centres may be risky, is very staff/equipment resource intensive, risks ITU overload at a centre and is costly).

Conversely, if an early-treatment strategy (primarily with coiling) of grade 4–5 aSAH patients was proven to be superior in a RCT, there is a compelling argument that it should be provided to all patients. Endovascular coiling services would need to be extended to cover 7 days, as it would not be logical to admit a critically ill patient to an intensive care bed from a peripheral hospital and then delay coiling treatment because of lack of endovascular service. In 2013, approximately one-quarter of UK NSCs offered a robust weekend coiling service [NCEPOD1 + UK Neurointerventional Group (UKNG) survey 2013, undertaken on behalf of the UKNG by Professor Phillip White (Newcastle University, 2013, personal communication)].

Crucially, there is a need for better understanding of the mechanisms involved in determining outcome in these patients. The current practice, which is reliant on crude clinical grading and to some extent the initial computerised tomography (CT) scan for risk stratification, needs to be refined. There is an urgent need for biomarkers for better understanding of the disease and, therefore, better selection of patients for aggressive management, as well as potentially stratifying long-term care and rehabilitation needs. This would help to ensure that appropriate individualised medicine is practised in an area where treatment/care costs for each patient are relatively high, but societal socioeconomic impact is also disproportionately high.

Patient and public involvement (PPI) was central to developing TOPSAT2. Feedback following presentations to the Newcastle SAH survivors’ group fed into the development of the process of obtaining assent from relatives and the way that follow-up would be conducted. The Trial Steering Committee (TSC) also included two PPI members, an aSAH survivor and a relative, who were involved at every stage of the study, including contributing to the Plain English summary and the final report.

For many neurosurgeons/interventional neuroradiologists in UK NSCs, surveys confirmed that there was genuine uncertainty (clinical equipoise) regarding whether or not to treat all grade 4–5 aSAH patients as soon as possible; this also has service provision implications. The main risk in TOPSAT2 was that more patients would undergo aneurysm treatment, mostly by coiling, with some attendant risks that would otherwise not be the case. Some of these patients would die after coiling but before neurological improvement. In relation to the early-treatment procedure, this was estimated to occur in no more than 10–15 out of 170 patients enrolled on the ET arm.

The risk of modern aneurysm coiling-related long-term morbidity/mortality is around 3–5%, although it may be slightly higher in grade 4–5 patients (unclear from existing RCT data). 13,24 However, we know that rebleeding from an aneurysm has an awful prognosis (82% poor outcome in ISAT across patients of all grades) and that the rebleed rate is also higher in grade 4–5 patients. 4,6

By contrast, the potential benefits of determining the optimum management strategy for aSAH were considerable. If early treatment was proved, service reconfiguration would be necessary but the outcome for grade 4–5 aSAH patients in the UK could be transformed. Studies on early treatment for grade 4–5 patients indicate good outcome rates around 50%, yet the NCEPOD report1 found that many grade 4–5 patients were not admitted to a NSC in the UK. Even when patients are admitted, very few NSCs provide treatment 7 days per week. Furthermore, delays to treatment are correlated with poor outcome. There are almost 1300 grade 4–5 aSAH patients per annum in the UK and NCEPOD data show that ≈ 20% currently have a good outcome, yet almost 50% might have a good outcome with consistent early aggressive (emergent) treatment (with substantial associated societal health benefit). Although an early intervention strategy for all grade 4–5 patients would be very expensive initially, it would carry substantial long-term care and social benefits savings. However, that expense provides a strong case for the care of poor-grade aSAH to be truly individualised, which the use of imaging biomarkers can potentially help to deliver.

If TONI was at least as good as ET, fewer patients would need NSC admission and some might avoid ITU admission; there would be savings on coils and other consumables that could be made immediately, and simpler options would be viable for coiling services out of hours.

Objectives

Design

The TOPSAT2 trial was a pragmatic, randomised, open-blinded, end-point (PROBE), controlled, parallel-group study conducted in the UK and Eastern Europe. Extending recruitment beyond the UK was necessary to obtain the desired sample size within a reasonable time frame. Eastern Europe was chosen because of established links from surgical trials in intracerebral haemorrhage (STICHs). 25–27 It was planned to open 20 centres in the UK and 10 outside the UK, recruiting 246 UK and 100 non-UK participants. The proposed MRI substudy was to recruit a total of 100 consecutive participants from the UK only, in a subset of sites able to participate in it.

Participants were randomised to one of two groups in a 1 : 1 ratio to receive the standard local treatment for aneurysm either (a) as soon as possible within 72 hours of ictus (ET arm) or (b) after neurological improvement to WFNS grades 1–3 (TONI arm).

Eligibility

Eligible participants were adults aged 18–80 years and had been diagnosed with aSAH grades 4 or 5 on the WFNS scale. For trial eligibility purposes, the grade was that recorded at their first medical assessment following hospital attendance and confirmation of the diagnosis of SAH by CT (or MRI) and/or lumbar puncture. The protocol stressed that before including a participant in the trial it must also be confirmed that it would be possible to treat them within 72 hours of ictus, should they be randomised to the ET arm.

The exclusion criteria were as follows:

• Age > 80 years (the prognosis of poor-grade aSAH patients in this subgroup is extremely poor and they are a very small percentage of NSC aSAH admissions).

• WFNS grades 1–3, or uncertain WFNS grade. However, patients with aneurysms of uncertain grade on transfer to a NSC where a formal sedation hold was undertaken and the patient was subsequently established to be truly grades 4 or 5 were eligible for the trial. This also applied to patients with aneurysms of uncertain grade undergoing sedation hold after insertion of an external ventricular drain (EVD) or other early intervention for hydrocephalus.

• Signs of coning or brain death on arrival at the NSC.

• Pure aneurysmal intraventricular haemorrhage (no SAH).

• Large intracerebral haematoma from an aneurysm requiring immediate surgical clot evacuation.

• Significant aSAH-related haemodynamic instability.

• Lack of clinical equipoise.

• Lack of assent/consent.

• Pregnancy.

• Pre-SAH mRS > 2.

• Pre-existing severe comorbidity such that clinical follow-up at 12 months was judged unlikely.

• Non-saccular, mycotic, giant or other atypical aneurysm (as these are conditions for which early simple coiling/clipping would not be appropriate).

Recruitment

Outcomes

Sample size

The outcome measure in TOPSAT2 was based on the mRS, a 7-point scale, with values 0–5 representing increasing disability and a value of 6 representing death. In the past it was common to dichotomise such a scale for outcome analysis, resulting in a loss of information as not all patients contribute to the detection of a treatment effect. The Optimising Analysis of Stroke Trials Collaboration and other authors have shown the benefit of ordinal analysis in the field of stroke. 29 Using an ordinal analysis achieves substantially greater statistical power to detect a treatment effect with equal sample size. The sample size calculation was therefore based on a proportional odds regression ordinal analysis of the mRS.

The best available literature was used to provide the expected distribution of mRS after grade 4–5 aSAH patients, although it was a non-randomised study. 30 Detecting a difference in clinical outcome (i.e. favouring one treatment strategy over the other) by an OR of > 1.5 would be compelling evidence to rapidly change to a uniform practice, as would a number needed to treat (NNT) of < 10. Expected mRS distribution data (0, no symptoms 17%; 1, minor symptoms 10%; 2, some restriction in lifestyle 6%; 3, significant restriction in lifestyle 19%; 4, partially dependent 11%; 5, fully dependent 10%; and 6, dead 27%)30 was entered into the ‘sample size for ordered categories’ routine of the Compare2 program in WinPepi version 11.43 during July 2014 and using two-sided significance of 5%, power of 80%, and 1 : 1 ratio of sample size, the sample sizes needed for different ORs (where the OR is assumed to be the same at all cutting points) were examined.

With 167 participants in each arm, a proportional odds model ordinal analysis of mRS gave a cumulative OR in favour of better mRS in one treatment arm of 1.7 at a 5% significance level and 80% power. A margin was built in to allow for losses to follow-up and crossovers, giving a total sample size of 346 (173 per arm). Loss to follow-up of < 2% was based on data from multiple UK-based aneurysm coiling trials. 6,31,32

Using the expected distribution from the best available evidence30 as the control event rate and calculating the expected number of patients to fall in each outcome for the treatment group given an OR of 1.7, the NNT was calculated as the inverse of the proportion of pairwise comparisons, with a better result in the treatment group minus the proportion with a worse result. This gave a NNT of 5.7.

A sample size of 346 was sufficient that some secondary outcome analyses, including a dichotomised mRS, could also reach statistical significance. For instance, for a mRS 0–2 (alive and independent) versus 3–6 (dead or dependent) comparison, a 15% absolute difference in treatment effect would be statistically detectable.

If equal numbers of grades 4 and 5 patients had been recruited (and they are approximately equal in proportion of aSAH patients), with 346 patients overall, the trial was powered to detect a statistically significant OR of 2.2 for improved clinical outcome, favouring one treatment strategy over the other by individual grade, particularly for grade 5. This is clinically relevant given the appreciable differences between grades 4 and 5 patients; the optimum management strategy may differ between grades.

Randomisation

Randomisation was carried out by the research team at sites through the use of an online system accessed through the trial website (www.topsat2.co.uk), and hosted by the Centre for Healthcare Randomised Trials, University of Aberdeen. Randomisation utilised a minimisation algorithm with an 80% chance of being allocated to the minimisation group to reduce differences in the two arms with respect to the following:

• WFNS grades 4 or 5 (so distributed equally between the two arms)

• participant age at the time of randomisation (age bands 18–50 years, 51–65 years and 66–80 years)

• presence of clinically significant hydrocephalus requiring CSF drainage procedure (yes/no)

• UK/non-UK site.

The remaining 20% followed a totally random allocation.

If the participant was randomised to the ET arm, the result of randomisation was communicated to the neurovascular team treating the aneurysm. They decided on the most appropriate treatment strategy but, as per the RCT evidence base, if the aneurysm was technically amenable to coiling, coiling was the initial therapeutic option. Not all Eastern European centres provided an endovascular service, so only neurosurgical treatment (clipping) was available at those centres. The usual institutional surgical consent form for the procedure was completed, and the aneurysm was treated within 24 hours of randomisation and 72 hours of the ictal bleed that led to hospital admission.

If the participant was randomised to the TONI arm, the result was communicated to the ITU and neurovascular team, who continued to manage the patient as per the established local protocol. Once the patient’s neurological status improved to WFNS grade 3 or better, the aneurysm was treated expeditiously. There was no specific time delay criterion for aneurysm treatment in this arm, but it was anticipated that treatment beyond 1 month post randomisation would be exceptional owing to both the lower risk of rebleeding after that time and the very guarded prognosis if a participant had not recovered sufficiently neurologically for treatment within 1 month.

Statistical methods

Safety

Monitoring for safety was undertaken in accordance with good clinical practice. Full details and definitions were provided in the trial protocol. The local investigator responsible for the care of the participant was asked to assign causality and expectedness of SAEs, in accordance with the trial protocol.

It was anticipated that most AEs that would occur during the trial, whether serious or not, would be expected because of the clinical indication and the standard treatment that the patient was receiving. A full list of expected SAEs related to acute SAH and aneurysm treatment is provided in Appendix 2.

All AEs related to the trial intervention (randomisation to either the ET arm or the TONI arm) were recorded in the case report form/electronic case report form. Those classified by the trial centre as SAEs were reported to the chief investigator, Clinical Trials Unit and sponsor on a SAE form, which was sent using either secure fax-to-e-mail transmission or secure e-mail. Details of the events were entered into the trial database. SAEs were followed up until resolution.

Participant flow

Figure 1 shows the participant flow through the study from screening to 12-month follow-up.

FIGURE 1.

The TOPSAT2 CONSORT flow diagram.

Screening

A total of 1269 grade 1–5 aSAH patients were screened (1189 from UK sites, 80 from non-UK sites), of whom 964 (936 UK, 28 non-UK) were immediately excluded because their aneurysm was not classified as poor grade (WFNS grades 4 or 5). Of the 305 poor-grade patients screened (286 UK, 19 non-UK), 23 were randomised; 16 were from UK sites and seven were from a single, non-UK site. Table 1 shows the number of patients screened at each site with the proportion of those recruited. See Tables 2 and 3 for more details of screened patients.

Recruitment

Recruitment opened on 20 September 2016 in Newcastle upon Tyne and a further 14 sites were opened over the subsequent 22 months, including three non-UK sites (Łódź, Poland; Riga, Latvia; and Timişoara, Romania).

The first participant was randomised on 2 March 2017, 6 months after screening started. A chart of actual cumulative recruitment by month (up until recruitment was paused in December 2018) against target recruitment over the same period is shown in Figure 2.

FIGURE 2.

Planned and actual recruitment for TOPSAT2.

Baseline characteristics

Twenty-three participants were recruited to TOPSAT2. Eleven were randomised to the ET arm and 12 to the TONI arm. The baseline characteristics of each of the two randomisation arms are shown in Table 4. Participants ranged in age from 44 to 75 years with a median age of 63 years. They were more likely to be female (65%) than male, and more were recruited in the UK (70%) than outside the UK. Most participants had an aneurysm of WFNS grade of 5 (74%) and Fisher grade 4 (87%), confirming their poor-grade status, although nearly all participants had a pre-stroke mRS of 0 (70%) or 1 (26%). The average time from ictus to randomisation was 20.9 hours. Randomisation resulted in two well-matched treatment arms.

A table of baseline characteristics by WFNS grade is provided in Appendix 3, and a table of baseline characteristics by location of site (UK vs. non-UK) is provided in Appendix 4.

Aneurysm treatment details

The aneurysm characteristics are shown in Table 5. The majority of aneurysms were located in the anterior communicating artery (43%) or posterior communicating artery (30%). The planned treatment of the aneurysm was endovascular (coiling) in 30% of patients and neurosurgical (clipping) in 52% of patients but was not stated for 17% of patients who were in the TONI arm. Although there did appear to be a difference in planned treatment between the two treatment arms, with 55% planned coiling in the ET arm and 8% in the TONI arm, it is likely that coiling would have been the eventual preferred treatment in the ‘not yet planned’ participants, as these were all UK patients where coiling was more prevalent. The time from ictus to securing the aneurysm in the ET arm varied between 3 and 71 hours, with a median of 26 hours, whereas for the TONI arm it varied between 15 and 434 hours, with a median of 163 hours (over 6 days) for the four patients treated.

Outcomes

The primary objective for TOPSAT2 was functional outcome at 12 months using the mRS. Outcomes for each arm at 6 and 12 months after randomisation are shown in Table 6.

Secondary outcomes

At 6 months, the Wilcoxon rank-sum test gives a value of p = 0.33 for the patients with known mRS. However, at 6 months there were four patients alive but for whom mRS was unknown (three ET patients and one TONI patient), and at this time their place of residence was also not known. Their best possible mRS was 0 and their worst possible mRS was 5. Assuming that the patients in the ET arm whose mRS were unknown had a value of 0 and the patient in the TONI arm whose mRS was unknown had a value of 5, and then assuming the opposite scores for those in each arm, the Wilcoxon rank-sum test produced p-values of 0.08 and 0.18, respectively.

Table 7 shows the outcomes for mortality and for dichotomised mRS, with the break between mRS 3 and mRS 4 and between mRS 2 and mRS 3, respectively. There is no evidence of a difference between the two treatment arms. Only one patient died after 30 days. Figure 3 shows the results of the survival analysis (Log-rank p = 0.32).

TABLE 7 - Secondary outcomes: mortality and dichotomised mRS

At 12 months, two patients were lost to follow-up (one in each group) but both were known to be living at home at that point, so for these analyses they were assumed to both have a mRS of 3.

Secondary outcomes	ET (N = 11), n (%)	TONI (N = 12), n (%)	Fisher’s exact test
Dead at 30 days	6 (55)	9 (75)	p = 0.40
Dead at 6 or 12 months	6 (55)	10 (83)	p = 0.19
12-month mRS 0–3 vs. 4–6a	3 (27)	1 (8)	p = 0.32
12-month mRS 0–2 vs. 3–6a	2 (18)	0	p = 0.22

FIGURE 3.

Kaplan–Meier plot.

Table 8 shows cause of death for all patients who died in the trial and demonstrates that the main cause of death in both arms for these poor-grade patients is subarachnoid haemorrhage.

TABLE 8 - Cause of death

Cause	ET (N = 6), n (%)	TONI (N = 10), n (%)	Total (N = 16), n (%)
Subarachnoid haemorrhage	3 (50)	6 (60)	9 (56)
Cardiac arrest	1 (17)	2 (20)	3 (19)
Pneumonia	1 (17)		1 (6)
Cerebral infarction	1 (17)	2 (20)	3 (19)

Table 9 shows outcomes at 30 days. By day 30, six ET patients and 10 TONI patients had died. Of the survivors, two ET patients had been discharged: one to be home alone (mRS = 2) and the other into rehabilitation (mRS = 3). The remaining three ET patients (all mRS 5) and two TONI patients (mRS 4 and 5) were still in hospital. There was no evidence of a difference in mRS at 30 days (Wilcoxon rank-sum test p = 0.33). Length of time in ITU by day 30 varied between 3 and 30 days with a median in the ET group of 13 days [mean 11.8 days; standard deviation (SD) 7.7 days] and a median in the TONI group of 10 days (mean 14.8 days; SD 9.2 days) (Wilcoxon rank-sum test p = 0.68).

TABLE 9 - Outcomes at 30 days

SD, standard deviation.

Randomised	ET (N = 11)	TONI (N = 12)	Randomised (N = 23)
Status by day 30, n (%)
Died before discharge	6 (55)	9 (75)	15 (65)
Discharged	2 (18)	0	2 (9)
Still in hospital	3 (27)	3 (25)	6 (26)
Discharged to, n (%)
Home alone	1	0	1 (50)
Home not alone	0	0	1 (50)
Residential home	0	0	21
Nursing home	0	0
Referring hospital	0	0
Rehabilitation unit	1	0
Other	0	0
Not applicable	9	12
Number of days in ITU by 30 days
Mean	11.8	14.8	13.4
SD	7.7	9.2	8.3
Median	13	10	10
Q1, Q3	5, 15	9.2, 24.8	8, 15
Minimum, maximum	3, 30	4, 30	3, 30
mRS score at 30 days, n (%)
0	0	0	0
1	0	0	0
2	1 (9)	0	1 (4)
3	1 (9)	0	1 (4)
4	0	1 (8)	1 (4)
5	3 (27)	1 (8)	4 (17)
6	6 (55)	10 (83)	16 (70)

Adverse events

There were 41 AEs recorded: 19 in the ET arm (mean 1.7 per participant) and 22 in the TONI arm (mean 1.8 per participant). The number of AEs for an individual participant ranged from 0 to 5, as shown in Figure 4.

FIGURE 4.

Number of AEs per participant.

The nature of AEs is shown in Table 10. There are no statistically significant differences between the arms, although infections were more likely to be reported in patients allocated to the ET arm (Fisher’s exact p = 0.08), whereas brain oedema (p = 0.35) and cardiorespiratory (p = 0.17) events were more likely to be reported in patients allocated to the TONI arm. Most of the ‘other’ events (four out of the five) were death (in the fifth a new infarction was reported).

Other analysis

Appendix 4 reports the baseline characteristics of UK and non-UK participants. Non-UK participants tended to be older and were more likely to be male and to have high blood pressure. The planned treatment to secure all aneurysms for non-UK participants was neurosurgical clipping. The mortality rate differed between UK (56%) and non-UK (100%) participants (Fisher’s exact p = 0.06).

Appendix 3 reports the baseline characteristics by WFNS grade. Participants with WFNS grade 4 tended to be older, but the other baseline variables showed little difference, and there was no evidence of a difference in mortality rate, with rates of 83% in grade 4 patients and of 65% in grade 5 (Fisher’s exact p = 0.62).

Summary of results

Of the 1269 aSAH patients screened, 305 were identified as poor grade. From this group, 23 patients were recruited and randomised to the ET group (11 patients) or TONI group (12 patients). Patients had a median age of 63 years, 65% were female and 74% were WFNS grade 5. All patients randomised to the ET group had their aneurysm clipped or coiled within 3 days and prior to any change in their WFNS grade. Only four patients randomised to the TONI group went on to have their aneurysm secured: three before neurological improvement and one after improvement to grade 2. As expected, the mortality rate was high among poor-grade patients: 55% for the ET group and 83% for the TONI group. There was no evidence of a statistically significant difference in the mRS between the two groups at 12 months, as per the Wilcoxon rank-sum test (p = 0.11).

Overview of recruitment and comparison of clinical results with other key studies

Recruitment to TOPSAT2 proved to be very challenging in both the UK and Eastern Europe, and the reasons are explored in detail in Reasons for early termination.

Reasons for early termination

The trial recruited only 23 patients over 25 months; therefore, feasibility could not be confirmed and recruitment had to be terminated early. Furthermore, although by then centres had been opened in four European countries, participants were actually recruited only in the UK and Romania, and only 9 out of 15 open centres recruited participants. It was clear that this patchy enrolment pattern applied across both UK and Eastern European centres so that opening additional European centres did not seem likely to overcome the very slow recruitment. Various measures were taken to attempt to stimulate recruitment: we conducted local training with ITUs/neurosurgery departments and we offered feedback calls for units, as required, following on from site initiation visits/issues arising. We presented TOPSAT2 at multiple relevant conferences including the UK Neurointerventional Group, the European Association of Neurological Societies, the European Society of Minimally Invasive Neurological Therapy, the Society of British Neurological Surgeons, the British Neurovascular Group and the British Society of Neuroradiologists. We ran stands at multiple national/international conferences to raise awareness/profile. We held trial webinars for after ‘conference season’ as an ongoing awareness raiser and to specifically address/discuss equipoise issues as well as to assist with site/PI training. We addressed any issues arising and focused on recruitment in the trial monthly newsletter.

In addition, through one of the co-applicants, Mr Paul Brennan, we engaged further with the British Neurosurgical Trainees Research Collaborative to assist with TOPSAT2 screening and enrolment. We developed training materials/a slide-set for ITU staff and neurosurgeons. We also agreed on a policy of preferential enrolment with the EME-funded Subcutaneous Interleukin-1 Receptor Antagonist (SCIL) trial run from Manchester (this recruited patients of all SAH grades), so that neither trial interfered with recruitment of the other. Both trials actively encouraged their centres to recruit any SAH case eligible into either TOPSAT2 or SCIL, as appropriate (see Appendix 5).

A total of 1269 patients were assessed by centre using screening logs. These findings are summarised in Table 2.

Data from screening logs indicate that the proportion of poor-grade SAH patients was within the anticipated range (expected 22% and actual 24%) and that the medically eligible proportion of poor-grade patients was also very close to expected, at 48% (50% expected). The reasons for screened poor-grade SAH patients not being eligible were summarised in Tables 2 and 3.

The reason for non-eligibility was known for 240 out of 305 patients. Definite (known) contraindications (to trial enrolment) account for a majority of reasons for non-eligibility (147/240). Of non-medical reasons for non-eligibility, the lack of clinical equipoise (74/240) is by far the largest factor and accounts for 28% of all poor-grade patients overall (where the reason for non-eligibility was recorded). Of the 42 grade 4–5 patients screened as non-eligible but with no reason given for non-eligibility, it is possible, indeed probable, that a fair proportion of these were also medically eligible but lack of equipoise was not declared and recorded. It was the unexpectedly high ‘lack of clinical equipoise’ rate in otherwise medically eligible poor-grade patients that undermined recruitment, not timelines (15 patients) or lack of assent (four patients). Centres had to confirm such clinical equipoise regarding the question of treat all early regardless of improvement versus TONI (irrespective of whether that improvement was early or late) to participate in the trial and, indeed, several large UK centres that declared they lacked equipoise declined to participate on those grounds. ‘Other’ (18/240) comprises a large number of reasons but can be aggregated into three main groupings (more than one reason applying in some cases): timeline related (delayed presentation or admission to NSC, etc.) for 15 patients, query over grade for five patients, and a combination of reasons for non-eligibility for three patients.

There was a good conversion rate, of 23 out of 45 for patients screened as eligible, in TOPSAT2 (51% of eligible enrolled vs. 20% expected). Therefore, had the lack of clinical equipoise rate been remotely similar to that seen in the TOPSAT1 pilot (4%), recruitment would have been increased by ≈ 35 patients. In other words, there would have been a < 150% increase in recruitment during the trial recruitment period, rendering it feasible to have completed the trial with more centres opened.

It is uncertain what part the UK national drive towards treating aSAH within 48 hours may have played in that equipoise shift. 1,35 Although the National Clinical Guideline for Stroke35 is clear that treatment within 48 hours refers to good-grade patients (as early coiling was proven in good-grade patients in the ISAT trial), it may well have influenced perceptions quite widely within NSCs, such that units may have been concerned that TOPSAT2 enrolment might make it appear that they were missing the ‘48 hours to secure aneurysm’ target repeatedly. That was one of the contributory factors stated for not joining TOPSAT2 by some of the centres that declined to participate in the trial. The equipoise change is probably also due to a number of other reasons; all consultants and senior nursing staff managing a case had to be content with randomisation for this to be feasible. It transpired that a team consensus to clinical equipoise in participating centres was not the prevailing situation, whereas it might have tended to be so in the past if agreement to participate in a trial had been given by all relevant consultant groups. Indeed, even the understanding of the concept of clinical equipoise as it applies to an individual patient may have worsened with time, and/or a mindset of defensive medicine may affect clinical care trials such as this. Anything potentially perceived (even erroneously) as ‘conservative management’ is perceived as risky, although, of course, randomisation into a clinical care trial like TOPSAT2 is actually a very positive decision. 36,37

Limitations

The TOPSAT2 trial aimed to compare a policy of securing a ruptured aneurysm in poor-grade patients early in all cases, with the aneurysm being treated expeditiously on neurological improvement (whether that was early or late was irrelevant). TONI had been the standard approach to management when aneurysm clipping was the main/only proven treatment (up until at least 2002). The trial PROBE design used in TOPSAT2 is a very well-validated and robust one that has proven successful in the neurointerventional field before. However, such an approach has been less successful in prior trials where neurointerventional treatment has been compared with ‘conservative’ management. 33,34

Although TOPSAT2 was not a trial of conservative management, it was clear that the actual policy applied in most trial centres was ET in all cases, even though they had stated equipoise regarding the approach in response to pre-trial enquiries and in centre-declared expressions of interest. In other words, it seems that many clinicians in participating centres viewed TOPSAT2 as a de facto conservative management trial, even though it was not, and, therefore, they felt that they (or at least one of the team) were not in personal clinical equipoise to randomise.

This lack of equipoise was not seen in the earlier feasibility TOPSAT1 trial. 19 This change is probably for a number of reasons discussed above.

Relatives’ assent was required for all TOPSAT2-screened eligible patients, and the relatives’ approach to this may also have shifted towards ET. This is unsurprising given that they were often told that their relative was being transferred to a regional NSC for the aneurysm to be coiled, raising expectations of definite and early treatment. These factors combined led to an unfeasibly low recruitment rate (relative to time and target necessary), even though the conversion rate of screened, eligible patients was somewhat better than predicted.

The very modest sample size recruited is the major limitation of TOPSAT2. The preponderance of WFNS grade 5 patients would have been only a relatively minor limitation of the trial (as grade was a minimisation criterion) if recruitment targets were able to have been met. The PROBE design and the multicentre/multinational recruitment with blinded outcome assessment are indeed strengths, but cannot negate the small sample size achieved.

Generalisability of trial findings

The low recruitment inevitably limits drawing any substantial conclusions about the management of poor-grade aSAH. However, the demographics in terms of sex and key comorbidities were as expected, although the population was, on average, older than is typical of aSAH patients. There were also some trends emerging within the population actually recruited into TOPSAT2 indicating that it may not reflect the generality of poor-grade aSAH. For instance, younger patients and those in WFNS grade 4 are under-represented in the TOPSAT2 RCT compared with the NCEPOD national sample.

The TOPSAT2 trial gives an indication of outcomes really only for WFNS grade 5 SAH patients and grade 4 patients aged > 60 years, neither group being well represented in either ISAT or the NCEPOD sample (regarding NSC admissions). Given that the clinical outcomes statistically are the same between arms, the results are likely to reflect the true clinical outcome for the sample population (WFNS grade 5 and older grade 4 patients).

Interpretation

Unfortunately, the low recruitment and early termination of TOPSAT2 preclude making many clear statements or recommendations. The subgroup of poor-grade SAH patients proved to be a difficult population in which to undertake a randomised trial, at least in the neurovascular field. The trial data do suggest that the care pathway for aSAH has improved in the 15 years since the ISAT trial, at least in part driven by a 48-hour target to treat aneurysms in good-grade patients. However, as a by-product, that 48-hour aspiration may have contributed to undermining equipoise to randomise patients into the trial. Younger patients, especially those in WFNS grade 4, were under-represented; this may well be because the team/unit preference to treat immediately may have been more compelling given that, on average, both worse grade and older age are independent poor prognostic factors following aSAH.

However, if there are differences between the ET and TONI approaches they would appear to be modest at best, as no strong trends let alone statistically significant differences between the two management approaches were detected for mortality or functional clinical outcome in TOPSAT2 (or indeed in the preceding small TOPSAT1 feasibility study).

Although any interpretation of TOPSAT2 data must necessarily be limited, it is reassuring that no evidence was demonstrated that supported the hypothesis that ET would reduce mortality but lead to a large proportion of highly dependent survivors. The combined mRS 4–5 rate at 12 months was 2/11 for the ET arm compared with 1/12 for the TONI arm, and mortality was 6 versus 10 and not statistically different.

Implications for health care

With regard to poor-grade aSAH management, at present, some units employ ET in nearly all patients routinely, some apply both ET and TONI on an ad hoc individual basis and others are (variably) selective in admissions regarding poor-grade SAH and some restrict admission to grade 4. 1 The TOPSAT2 data indicate no difference in outcome for the strategies of ET and TONI, albeit with the caveats discussed above (see Chapter 4). Therefore, the evidence suggests that TOPSAT2 is not able to provide data to indicate that the current heterogeneous approach to poor-grade management should change in one or another direction and, thus, that no service reconfiguration is necessary.

The use of MRI remains very challenging in aSAH patients, particularly in poor-grade patients. Therefore, caution should be employed when assessing any data on MRI in SAH – especially if generated on small sample sizes and/or extrapolating results from predominantly good-grade patients.

Future research implications

The RCT approach to poor-grade SAH management of TOPSAT2 proved unfeasible, and funding for a prospective observational study specific to this group was rejected by the National Institute for Health Research (NIHR) EME programme. It is also not likely to appeal to any other UK funder currently supporting studies in this clinical field given the scale of resources and study duration required, especially given the impact of the COVID-19 pandemic on charitable incomes. This leaves clinical audit approaches. A systematic comprehensive SAH patient audit has been proposed in the UK but does not yet exist. Anything less than a universal (all NSC units) and comprehensive (near 100% case ascertainment and inclusion) audit is inherently afflicted by many biases. These include admission and inclusion selection biases, data collection bias, transfer bias (losses to follow-up not random), interviewer bias, recall bias, outcome misclassification bias, confounding bias and performance bias. There is an ongoing voluntary UK and Ireland aSAH audit but it involves < 70% of NSCs; no measures exist to ensure 100% case ascertainment and all of the follow-up is undertaken and adjudicated by the clinical team looking after the patient, which is inherently open to major bias. Extension of the Stroke Sentinel National Audit Programme (SSNAP) to include SAH would probably be a better way to achieve that universal audit. However, no funding exists for such a development at present. Although it would not eliminate all the biases mentioned above, it could deal with some and ameliorate others. The linkage of SSNAP data to quality improvement markers, and thus NHS funding, has proven successful in the ischaemic stroke care setting.

Randomised controlled trials of proven design overcome most biases. However, RCTs of management pathways or so called ‘clinical care trials’, although often called for, have proven difficult, certainly in the neurovascular field. 38 If we are to provide patients with optimal, prudent care in the context of uncertainty and perform such trials then it does seem to require a major change in the mentalities of clinicians, other health-care staff and patients/relatives alike. Recognition that the answer to a care question was valuable and needed (with supporting survey and indeed feasibility study evidence) no longer seems to be enough. Innovative trial methodologies, such as step-wedge or cluster RCTs, may not map well onto the sort of specialised management care pathways that TOPSAT2, TEAM36 and ARUBA39 have attempted to investigate – where units that perform the treatment are relatively few and where key team leaders will frequently interact regionally, nationally and internationally. These trials’ collective experience does raise the question of whether or not it is appropriate that trials aimed at evaluating clinical care have a fixed, limited time period to provide an answer (recruitment ‘feasibility’) while potentially unjustified clinical care continues ‘unquestioned’ over an unlimited time period. In such clinical care questions, it is more likely than not that one approach is indeed somewhat better than the other, but if both approaches continue in the long term it is clearly likely to be to the detriment of some of those receiving the clinical care in question.

The use of MRI biomarkers looks promising in aSAH but the TOPSAT2 experience indicates that, at least in the subgroup of poor-grade patients, routine use of such MRI in the UK remains some way off and is associated with major logistical barriers in any health-care system. More traction might exist for the investigation and validation of MRI biomarkers in good-grade patients (not requiring a critical care support team for scanning), but studies would need to be correspondingly much larger, given that the number of (poor) outcome events is much lower in good-grade patients, amplifying the costs of such studies, which per patient are already high.

Contributions of others

Professor Peter Andrews (Consultant and Honorary Professor) was chairperson of the DMC and reviewed the final report.

Professor Jens Fiehler (Director of Diagnostic and Interventional Radiology) was a member of the TSC and reviewed the final report.

Professor Alex McConnachie (Professor of Clinical Trial Biostatistics) was an independent statistician on the DMC and reviewed the final report.

Dr Andrew Molyneux (Consultant Neuroradiologist) was a member of the DMC and reviewed the final report.

Contributions of authors

Philip White (https://orcid.org/0000-0001-6007-6013) (Translational and Clinical Research Institute, Newcastle University) was chief investigator for TOPSAT2, helped conceive the multicentre trial and develop the trial protocol, chaired the Trial Management Group (TMG), helped draft the report and edited it for important intellectual content.

Barbara Gregson (https://orcid.org/0000-0003-4868-9137) (Principal Research Associate, Neurosurgical Trials Group) was co-investigator and statistician for TOPSAT2, helped conceive the multicentre trial and develop the trial protocol, analysed the data, helped draft the report and edited it for important intellectual content.

Elaine McColl (https://orcid.org/0000-0001-8300-3204) (Professor of Health Services Research, Newcastle University) helped finalise the trial protocol, was a TMG member and edited the report for important intellectual content.

Paul Brennan (https://orcid.org/0000-0002-7347-830X) (Senior Clinical Lecturer, Neurosurgery, University of Edinburgh) contributed to neurosurgical engagement and recruitment.

Alison Steel (https://orcid.org/0000-0003-1279-1455) (Senior Trial Manager, Newcastle Clinical Trials Unit) helped finalise the trial protocol, was a TMG member and edited the report for important intellectual content.

Philippa Watts (https://orcid.org/0000-0003-3523-5101) (Trial Manager, Newcastle Clinical Trials Unit) set up and worked with study sites, conducted trial management and contributed to drafting and editing the final report for important intellectual content.

Ruth Wood (https://orcid.org/0000-0002-8296-1774) (Data Manager, Newcastle Clinical Trials Unit) designed and set up the trial-specific database, managed central data processes, reviewed study results and edited the final report.

Clare Bowes (https://orcid.org/0000-0002-9212-5315) (Professional and Administrative Services Officer, Faculty of Medical Sciences, Newcastle University) carried out data collection, data entry and revision of the report manuscript for important intellectual content.

Mohsen Javadpour (https://orcid.org/0000-0001-9836-1741) (Consultant Neurosurgeon, National Neurosurgical Centre, Beaumont Hospital) was chairperson of the TSC and had input in and final approval of the version of the report.

Amanda Weston (https://orcid.org/0000-0002-8119-1050) (independent layperson) contributed to public and patient involvement and reviewed and edited the final report.

Dipayan Mitra (https://orcid.org/0000-0002-1891-5415) (Consultant Neuroradiologist, Department of Neuroradiology, Royal Victoria Infirmary) contributed to the development of the study hypothesis, collection of pilot data, grant application, trial management, initiation of sites and writing of the manuscript.

Data-sharing statement

Fully anonymised data from this trial will be available to the scientific community subject to appropriate ethics approval and TSC agreement. Requests for anonymised data should be directed to the corresponding author.

Disclaimers

This report presents independent research. The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, the MRC, NETSCC, the EME programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the EME programme or the Department of Health and Social Care.

Adverse events related to acute subarachnoid haemorrhage (neurological or systemic)

Adverse events related to aneurysm treatment (coiling or clipping)