The durability of endovascular coiling versus neurosurgical clipping of ruptured cerebral aneurysms: 18 year follow-up of the UK cohort of the International Subarachnoid Aneurysm Trial (ISAT)

Summary

Background

Previous analyses of the International Subarachnoid Aneurysm Trial (ISAT) cohort have reported on the risks of recurrent subarachnoid haemorrhage and death or dependency for a minimum of 5 years and up to a maximum of 14 years after treatment of a ruptured intracranial aneurysm with either neurosurgical clipping or endovascular coiling. At 1 year there was a 7% absolute and a 24% relative risk reduction of death and dependency in the coiling group compared with the clipping group, but the medium-term results showed the increased need for re-treatment of the target aneurysm in the patients given coiling. We report the long-term follow-up of patients in this UK cohort.

Methods

In ISAT, patients were randomly allocated to either neurosurgical clipping or endovascular coiling after a subarachnoid haemorrhage, assuming treatment equipoise, between Sept 12, 1994, and May 1, 2002. We followed up 1644 patients in 22 UK neurosurgical centres for death and clinical outcomes for 10·0–18·5 years. We assessed dependency as self-reported modified Rankin scale score obtained through yearly questionnaires. Data for recurrent aneurysms and rebleeding events were collected from questionnaires and from hospital and general practitioner records. The Office for National Statistics supplied data on deaths. This study is registered, number ISRCTN49866681.

Findings

At 10 years, 674 (83%) of 809 patients allocated endovascular coiling and 657 (79%) of 835 patients allocated neurosurgical clipping were alive (odds ratio [OR] 1·35, 95% CI 1·06–1·73). Of 1003 individuals who returned a questionnaire at 10 years, 435 (82%) patients treated with endovascular coiling and 370 (78%) patients treated with neurosurgical clipping were independent (modified Rankin scale score 0–2; OR 1·25; 95% CI 0·92–1·71). Patients in the endovascular treatment group were more likely to be alive and independent at 10 years than were patients in the neurosurgery group (OR 1·34, 95% CI 1·07–1·67). 33 patients had a recurrent subarachnoid haemorrhage more than 1 year after their initial haemorrhage (17 from the target aneurysm).

Interpretation

Although rates of increased dependency alone did not differ between groups, the probability of death or dependency was significantly greater in the neurosurgical group than in the endovascular group. Rebleeding was more likely after endovascular coiling than after neurosurgical clipping, but the risk was small and the probability of disability-free survival was significantly greater in the endovascular group than in the neurosurgical group at 10 years.

Funding

UK Medical Research Council.

Introduction

Since the introduction of endovascular coiling of ruptured cerebral aneurysms in 1990 in the USA and in 1992 in Europe, concerns have been expressed about the durability of endovascular coiling and the degree of angiographic occlusion of the aneurysm (believed to be a surrogate marker for risk of recurrent subarachnoid haemorrhage [rebleeding from the aneurysm]). The International Subarachnoid Aneurysm Trial (ISAT) was a multicentre, randomised controlled trial, funded principally by the UK Medical Research Council, that enrolled 2143 patients with ruptured intracranial aneurysms in 43 centres. The trial aimed to assess the relative safety and efficacy of endovascular coiling treatment versus neurosurgical clipping in patients who were suitable for either treatment. The primary objective was to establish the clinical outcomes at 1 year. The secondary objective was to assess differences between the two treatments in the prevention of rebleeding from the treated aneurysm. The tertiary objective, reported here, was to obtain long-term follow-up data.

The trial methods and the 1 year primary outcomes have been published previously.^1,2 The primary outcome (death or dependency at 1 year) showed that endovascular coiling resulted in significantly fewer dead and dependent patients than did neurosurgical clipping. The medium-term results (secondary objective) showed that the endovascular group needed more frequent retreatment of the target aneurysm than did the neurosurgical group (late retreatment defined as later than 1 month after the original subarachnoid haemorrhage for neurosurgery and later than 3 months after the original subarachnoid haemorrhage for endovascular treatment³). However, these additional treatments had no detrimental effect on dependency status in patients or on clinical outcome.⁴ In this Article, we report long-term follow-up data (tertiary objectives) for the 1644 patients enrolled in UK centres with 19 216 patient-years of follow-up.

Methods

Patients

Between Sept 12, 1994, to May 1, 2002, ISAT investigators enrolled 2143 patients with ruptured intracranial aneurysms at 43 centres, mostly in the UK, the rest of Europe, and Canada. Patients with ruptured intracranial aneurysms were randomly allocated to either neurosurgical clipping (n=1070) or endovascular coiling (n=1073). Patients were enrolled who had an aneurysm anatomically suitable for either treatment, in whom clinical equipoise existed for the best treatment, and in whom appropriate consent of the patient or assent of relatives could be obtained. The UK cohort was enrolled from 22 participating neurosurgical centres that provided both neurosurgical and endovascular treatment for acute subarachnoid haemorrhage from ruptured aneurysms. The methods of the trial have been described in detail in the published protocol and in the previous reports.^1,2

From the very beginning of ISAT to adoption by the Medical Research Council and the publication of its results, the study has adhered to the strict standards and procedures for the conduct, design, and reporting of clinical trials, as laid out in the Consolidated Standards of Reporting Trials (CONSORT) guidelines.^5,6 After challenges to the published results, the investigators have remained true to the protocol and the analysis of the data by intention to treat, showing the everyday reality of working with patients who have a ruptured cerebral aneurysm.

Procedures

Every year, between the date of enrolment and May, 2012, we sent all surviving ISAT patients in the UK a postal questionnaire that assessed their dependency status, whether they had any further hospital admissions, and whether they had undergone any further treatment related to their subarachnoid haemorrhage or original aneurysm treatment. Patients who completed the annual questionnaire used the self-reported modified Rankin scale (mRS) and were followed up for a minimum of 10 years. All UK patients were flagged with the Office for National Statistics (ONS; now known as the Health and Social Care Information Centre). The Oxford Neurovascular and Neuroradiology Research Unit automatically received notification of the death of patients registered in the study. The Oxford unit collected data for all suspected instances of recurrent subarachnoid haemorrhage from information given by patients or their families on the questionnaires, after notification by the centre or the patient's family doctor, or from information on the ONS notifications. Data for recurrent subarachnoid haemorrhage were collected, including date, any hospital records, operative reports, imaging data, autopsy reports (when available), and death certificates. An experienced neurosurgeon and neuroradiologist (RSCK and AJM) reviewed these records. All UK centres obtained local ethical committee approval for the study before enrolling patients.

Outcomes

We aimed to assess rates of death, dependence (measured according to patient-reported modified Rankin scale scores), and recurrent subarachnoid haemorrhage after neurosurgical clipping or endovascular coiling as part of the tertiary objectives of ISAT.

Statistical analysis

Survival analyses were done for time to rebleeding and for time to death, which was measured from the time of the initial subarachnoid haemorrhage. We censored observations in the analysis of time to rebleeding when a patient was lost to follow-up or had died. For the analysis of time to death, we retrieved all deaths among the UK cohort to a maximum of 18·5 years via ONS flagging. We extracted data for survival on March 31, 2013. We extracted data for dependency and rebleeding analyses on May 31, 2012, which had a maximum of 17·6 years follow-up. We calculated Kaplan-Meier curves for each treatment group, and the groups were compared using the log-rank test.

We calculated the standardised mortality ratio as the ratio of deaths recorded to deaths expected based on age and sex stratification, as reported in the UK interim life tables from ONS,⁷ conditional on 1 year survival. A two-tailed probability value of less than 0·05 and non-overlapping confidence intervals were used to judge statistical significance.

To calculate the standardised mortality ratio, we included all deaths of patients from 1 year after subarachnoid haemorrhage until Dec 31, 2012. We used standard death rates for England and Wales, which covered 5 year age bands for individuals aged 15–24 years, 10 year age bands for individuals aged 25–85 years, and an age band for individuals older than 85 years, for all years from 1998 until 2012. We ascertained the number of patients alive on Jan 1, each year, at least 1 year from the date of their initial subarachnoid haemorrhage, and we counted the number of deaths during the year. The age of each patient was assessed as though at the mid-point of each year (July 1). We calculated the expected number of deaths for each year and summed them for all years.

To compare rates of low severity self-reported mRS scores (0–2) at 10 years between the treatment groups, we extracted the results of the tenth annual questionnaire, together with the absolute numbers of deaths before 10 years after subarachnoid haemorrhage. The probability of a good outcome at 10 years is the probability of a good score on the self-reported mRS multiplied by the probability of being alive at 10 years. These probabilities have different denominators; we calculated the death rate with the complete UK cohort and estimated the score on the Rankin scale only in individuals not lost to follow-up and who returned the annual questionnaire. We calculated the SE for the probability of a good mRS score and of being alive (a good outcome) with the delta method (not shown). Similarly, we calculated the SE and 95% CI for the odds ratio (OR) of a good outcome for the treatment groups with the delta method.

We did sensitivity analyses, replacing missing 10-year mRS scores with 11 year, 9 year, or 8 year scores, in that order, when available, for the calculation of the probability of a good mRS and of being alive at 10 years after subarachnoid haemorrhage. Finally, we repeated the calculation with the latest available mRS score for patients alive at 10 years after subarachnoid haemorrhage but who had a missing score at 8, 9, 10, and 11 years. We did all analyses with Stata version 12. This study is registered, number ISRCTN49866681.

Role of the funding source

The trial was designed, completed, and analysed independently of the funders. The chief investigators (RSCK and AJM) had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Results

As of March 31, 2013, 1256 (76%) of 1644 patients enrolled in ISAT from UK centres were still alive. We followed up these patients for a minimum of 10·0 years and a maximum of 18·5 years (19 216 patient-years). 10 years after their original subarachnoid haemorrhage, 1331 (80%) of the 1644 patients had survived (table 1). Ten patients were lost to follow-up after the 2 month questionnaire. 1283 people returned questionnaires at 5 years or later. Thus, we had good follow-up from 1283 (96%) of 1331 of the cohort who lived for 10 years and some follow-up from 1321 (99%) of 1331 of the whole UK cohort.

Table 1.

Death and dependency at 10 years

	Survival at 10 years			Independence at 10 years			Independent survival at 10 years
	Alive	Dead	OR (95% CI)	mRS score 0–2	mRS score 3–5	OR (95% CI)	Probability of independent survival*	OR (95% CI)
Endovascular coiling group	674/809 (83%)	135/809 (17%)	1·35 (1·06–1·73)	435/531 (82%)	96/531 (18%)	1·25 (0·92–1·71)	0·682	1·34 (1·07–1·67)
Neurosurgical clipping group	657/835 (79%)	178/835 (21%)	Ref	370/472 (78%)	102/472 (22%)	Ref	0·617	Ref

Data are n/N with available data (%), unless otherwise stated. OR=odds ratio. mRS=modified Rankin scale.

^*Calculated by multiplication of probability of being alive and probability of good mRS (0–2).

1003 (75%) patients returned a questionnaire at 10 years (table 1). The proportion of patients with a good mRS score (0–2) did not differ between endovascular and neurosurgery groups (table 1). The probability of being alive with a good outcome (mRS 0–2) compared with death or dependence was significantly better for the group allocated endovascular treatment than for those allocated neurosurgical clipping (table 1).

In our sensitivity analysis, we retrieved mRS values from questionnaires done at years 11, 9, and 8 to replace missing values in records at year 10. Data were available for 1200 (90%) of 1331 patients alive after 10 years. The proportion of patients with a good mRS score was 81% (503 of 620 individuals) in the endovascular group and 78% (450 of 580 individuals) in the neurosurgical group (OR 1·24, 95% CI 0·93–1·65). In this analysis, the probability of a good clinical outcome (being alive and mRS 0–2) compared with death or dependence for the two groups favoured the endovascular group (OR 1·33, 95% CI 1·08–1·65). Finally, we replaced missing values with the latest available mRS score, including the 2 month mRS score for the ten patients who were lost to follow-up before the first year questionnaire. The proportion of patients with a good mRS score was 81% (546 of 674) in the endovascular group, and 77% (504 of 657) in the neurosurgical group (OR 1·29, 95% CI 0·99–1·70). The probability of a good clinical outcome rather than death or dependence again favoured the endovascular group (OR 1·36, 95% CI 1·11–1·68).

All data for recurrent subarachnoid haemorrhage that occurred before 1 year of follow-up have been reported previously, therefore we only report data for patients who had a recurrent subarachnoid haemorrhage later than 1 year after their initial haemorrhage.² 33 patients had a recurrent subarachnoid haemorrhage (table 2). Rebleeding from the target aneurysm (the aneurysm identified as ruptured at the time of enrolment) occurred in 17 patients, whereas the other 16 patients rebled from another source. At end of follow-up, the cumulative risk of a rebleed from the target aneurysm was 0·0216 (95% CI 0·0121–0·0383) for patients in the endovascular group and 0·0064 (0·0024–0·0173) for patients in the neurosurgery group (figure 1).

Table 2.

Number of patients who had recurrent subarachnoid haemorrhage after 1 year

	Rebleeding from target aneurysm*(deaths)	Rebleeding from aneurysm known at baseline†(deaths)	De-novo aneurysm‡(deaths)	Aneurysm from unknown source (deaths)	Total
Endovascular coiling group (8351 patient-years)	13 (3)	4 (2)	3 (1)	1 (1)	21 (7)
Neurosurgical clipping group (8228 patient-years)	4 (2)	2 (2)§	6 (3)	0	12 (7)
Total	17 (5)	6 (4)	9 (4)	1 (1)	33 (14)

Numbers for UK patients as of May 31, 2012. Numbers in parentheses show deaths within 30 days of rebleeding.

^*Aneurysm identified at the time of enrolment in the trial (log-rank p=0·02; figure 1).

^†Other known aneurysm present on the first angiogram but not believed to have ruptured.

^‡Aneurysms not shown on the first angiogram.

^§One patient died at 48 days from recurrent subarachnoid haemorrhage.

Figure 1.

Kaplan-Meier plot of cumulative risk of rebleeding from target (treated) aneurysm later than 1 year after subarachnoid haemorrhage

13 patients in the endovascular group rebled from the target aneurysm (between 1·9 years and 13·2 years after the procedure), and three of these individuals died within 30 days (table 2, appendix). Of the remaining ten patients, seven were independent at end of follow-up. Incidence of recurrent subarachnoid haemorrhage was one in 641 patient-years (table 3). In the neurosurgical group, rebleeding occurred in four patients (timing between 6·1 and 11·4 years), resulting in two deaths within 30 days of the event. Incidence was one in 2041 patient-years. The risk of having a poor outcome (death or mRS score ≥3) from this recurrent subarachnoid haemorrhage was one in 1397 patient-years in the endovascular group and one in 2041 patient-years for the neurosurgery group.

Table 3.

Incidence of recurrent subarachnoid haemorrhage from target aneurysm and of resulting death or disability (mRS 3–5)

	Number of patients	Total patient-years	Incidence per 1000 patient-years
Number of rebleeds*
Endovascular group	13	8351	1·56 (0·69–2·40)
Neurosurgery group	4	8228	0·49 (0.01–0.96)
Death and disability due to rebleed†
Endovascular group	6	8384	0·72 (0·14–1·29)
Neurosurgery group	4	8228	0·49 (0·01–0·96)

Data in parentheses are 95% CI.

^*Rebleeds from target aneurysm assessed from 1 year up to 17·6 years' of follow-up in the UK population.

^†Rebleeds from target aneurysm resulting in death within 60 days or score of 3–5 on the modified Rankin scale.

Of the 16 patients that rebled from another source, 15 bled from another aneurysm (table 2, appendix). Six bled from a pre-existing aneurysm that had been identified on angiogram at the time of presentation with the original subarachnoid haemorrhage and nine bled from a new aneurysm that was not present or reported on the initial angiogram. In one patient, it was not possible to determine the source of the rebleed.

Investigators have previously summarised data for deaths and causes of deaths from 2 months after randomisation to 14 years, with a mean of 9 years follow-up.^2,3 During the 18·5 year follow-up period, 338 (24%) patients died in the UK cohort (figure 2). Of the six deaths due to recurrent subarachnoid haemorrhage from the target aneurysm after 1 year, four were in the endovascular groups and two in the neurosurgical group (table 4). Over the same time period (up to March 31, 2013), nine deaths were due to rupture of another aneurysm. 249 deaths occurred later than 1 year after the initial haemorrhage, of which 232 were not related to cerebral aneurysms or the previous subarachnoid haemorrhage; most were due to cancer and cardiovascular disease (table 5). There were significantly more deaths in the survivors in both groups than in the general age-matched population.

Figure 2.

Kaplan-Meier plot of cumulative mortality

Patients observed for 10–18·5 years in 22 UK centres.

Table 4.

Causes of death in all UK patients from 2 months after enrolment

	2–12 months		1–5 years		After 5 years
	Endovascular (n=762)*	Neurosurgery (n=768)*	Endovascular (n=753)†	Neurosurgery (n=752)†	Endovascular (n=733)‡	Neurosurgery (n=719)‡
Complication of severely dependent survival (chest or other infections)	6	13	0	2	0	0
Cancer	0	0	6	13	37	25
Cardiac¶	0	0	2	7	19	20
Recurrent SAH from treated target aneurysm	1	1	0	0	4§	2
Recurrent SAH from another known, unknown, or de-novo aneurysm	0	0	3	1	1	4
Other intracranial haemorrhage	0	0	1	0	2	3
Ischaemic stroke	0	0	2	3	3	5
Suicide	0	1	1	0	1	0
Renal failure	1	0	0	1	3	2
Infections not related to dependent survival	0	1	2	3	6	2
Respiratory complications	0	0	2	2	17	11
Other‖	0	0	1	0	10	16
Unknown	1	0	0	1	2	1
Total	9	16	20	33	105	91

Shows all UK deaths to March 31, 2013. SAH=subarachnoid haemorrhage.

^*Number of patients alive at 2 months.

^†Number of patients alive at 1 year.

^‡Number of patients alive at 5 years.

^§One patient died from residual complications 3 years after recurrent SAH.

^¶Cardiac causes are cardiac arrest, coronary artery disease, ischaemic heart disease, and myocardial infarction.

^‖Other includes trauma, perforated ulcer, pulmonary embolus, and neurodegenerative disease.

Table 5.

Standardised mortality ratios

	Deaths	Expected deaths	Standardised mortality ratio (95% CI)
All patients	246	175	1·40 (1·23–1·58)
Women	146	106	1·37 (1·15–1·59)
Men	100	69	1·45 (1·17–1·73)
Endovascular group	122	89	1·37 (1·13–1·62)
Neurosurgery group	124	87	1·43 (1·18–1·67)

Calculated using deaths of all UK patients up to Dec 31, 2012.

Discussion

In this study, we provide long-term data for clinical outcomes after treatment of subarachnoid haemorrhage (panel). Since the introduction of endovascular coiling of ruptured cerebral aneurysms, concern has been expressed about the long-term durability of the treatment, and its ability to prevent recurrent subarachnoid haemorrhage. Coiling compared with clipping improves short-term and medium-term clinical outcomes for selected patients with ruptured aneurysms; proven to class one level A evidence.⁸ Patients were enrolled in ISAT soon after the introduction of aneurysm coiling, first used in the UK in 1992. There is a growing body of evidence that suggests that the clinical outcome after coiling ruptured aneurysms has improved compared with that in the ISAT enrolment period. Its effectiveness to prevent early rebleeding also seems to have improved.^9,10

Panel. Research in Context.

Systematic review

We searched clinical trials databases (current controlled trials and ClinicalTrials.gov), the Cochrane Library, and the published work. We searched PubMed and Medline for publications in English using the search terms: “subarachnoid haemorrhage”, “intracranial aneurysm”, “endovascular coiling”, “neurosurgical clipping”, “randomised trial”, and “long-term outcomes”. To our knowledge, our study is the only prospective, longitudinal, multicentre randomised trial comparing clipping and coiling after aneurysmal subarachnoid haemorrhage that addressed very long-term clinical outcomes or protection from recurrent subarachnoid haemorrhage.

Interpretation

Findings from our International Subarachnoid Aneurysm Trial have changed worldwide clinical practice and the study is very unlikely to be repeated. Our report shows that the durability of coil treatment and the clinical benefit of coiling noted at 1 year are maintained at 10 years. In both groups, the risk of death and rebleeding from the treated aneurysm up to 18 years was very small and similar to the risk of recurrent haemorrhage from another aneurysm. All patients had an increased standardised mortality ratio (1·40, 95% CI 1·23−1·58) compared with the general population, but they were far more likely (40 times) to die as a result of causes other than from the treated aneurysm—mainly cancer or cardiovascular disease. This finding emphasises the importance of advice about lifestyle factors in these patients, particularly about smoking and control of hypertension.

The results of ISAT apply to those patients who were eligible for the study. The original criteria for enrolment were patients with proven ruptured aneurysms in whom clinical equipoise existed as to whether clipping or coiling was the best treatment and who had aneurysms that were judged anatomically suitable for endovascular coiling and surgical clipping. There was widespread criticism of the external validity of ISAT, particularly among clinicians in USA, based on the low proportion of patients enrolled (about 22% overall) in participating centres. This low number could have led to selection bias. In practice, the proportion of patients enrolled varied widely between participating centres (from about 50% to less than 2% of all patients), showing the wide variations in the clinical equipoise of individual clinicians, mostly neurosurgeons.¹¹ In ISAT, the patient population that was represented was the most common group of patients treated for ruptured aneurysms. 88% of enrolled patients were in good clinical grade (World Federation of Neurosurgical Societies grade 1 or 2) with anterior circulation aneurysms smaller than 10 mm. When compared with the demographics of patients treated after subarachnoid haemorrhage in the UK National Audit done by the Royal College of Surgeons,¹² 81% of patients in 2001–02 had similar demographics to patients in ISAT. In an audit of subarachnoid haemorrhage treatment in the UK over a 5 month period in 2011,¹³ 85% of patients treated for ruptured aneurysms had coiling. We think that the results apply to most patients with a ruptured cerebral aneurysm. With the improved coil devices and techniques that are now available, most ruptured cerebral aneurysms are anatomically suitable for coiling.

Concern has remained, quite reasonably, about whether benefits for the endovascular treatment might be lost in the long-term because of an increased risk of rebleeding and complications associated with further treatment. Our data lend support to a small excess risk of recurrent subarachnoid haemorrhage from the target aneurysm in the endovascular group up to 17·6 years after the initial haemorrhage; however, this excess did not translate to a significantly worse clinical outcome when compared with that of the surgically treated group. The overall risk of death or dependency from a rebleed did not differ between the groups.

In the UK we were able to examine the long-term mortality risk of patients by prospectively flagging patients with the ONS. This process has enabled reliable data for death to be collected for the entire UK ISAT cohort. Long-term causes of death in surviving patients were largely unrelated to the subarachnoid haemorrhage and were mainly due to other reasons (mostly cancer and cardiovascular disease). Patients were 40-times more likely to die from another cause than from the treated aneurysm. This figure emphasises the importance of lifestyle advice, particularly about smoking cessation and management of hypertension.

A substantial portion of the UK ISAT patients have continued to return their annual questionnaires; analysis of these data show that the noted short-term and medium-term clinical benefit of endovascular coiling remains significant in the long term. Furthermore, sensitivity analysis shows that the 10-year mRS was not biased by missing values, which seem to be random. Little difference existed between 5-year and 10-year dependency status for individuals.

The main strength of our study is that it represents a unique, prospectively enrolled population with complete ascertainment of death rates for up to 18 years, and very low loss of data for clinical outcome at 10 years. Thus, we believe that this dataset will be the best available for the long-term risks for patients treated for ruptured cerebral aneurysms. However, the long-term follow-up is limited to the UK cohort for cost and logistical reasons. Nevertheless, long-term durability of treatment will probably be similar between UK and non-UK patients who are given endovascular coiling. The coils and techniques and equipment are the same worldwide.

Another limitation of the study is the absence of imaging follow-up, which would have enabled correlation of angiographic findings after coiling with the rebleeding risk. We were also unable to assess how frequently and for how long patients should be assessed with imaging surveillance after coiling of an aneurysm. This question will be the subject of modelling in another study using these data, but, given the very low late risk of rebleeding, such long-term follow-up does not seem justified.

ISAT was a landmark international randomised controlled trial within the neurosurgical and neurointerventional communities. The results of the trial have changed clinical practice, with 85% of UK patients who have a ruptured aneurysm undergo coiling.¹³ Furthermore, our results also lead us to question the benefit of continued surveillance of treated aneurysms, in view of the extremely low event rate noted in the treated population, which has now been followed up for 18 years.