Delay May Reduce Procedural Risk, But at What Price to the Patient?

Introduction 

“Carotid intervention for recently symptomatic, severe carotid stenosis should be regarded as an emergency procedure in patients who are neurologically stable, and should ideally be performed within 48 hours of a transient ischaemic attack or minor stroke.”

The National Stroke Strategy.1

The UK Department of Health, 5th December 2007.

The American Academy of Neurology (AAN) and the American Heart Association (AHA) recommend that carotid endarterectomy (CEA) be performed within two weeks in patients presenting with a transient ischaemic attack (TIA) or a non-disabling stroke.2, 3 The available evidence, however, suggests that very few centres achieve this threshold.4 In a 2006 survey of practice in eleven centres in Holland, only 24% of patients presenting with a TIA/minor stroke and a 70–99% stenosis underwent endarterectomy within 6 months.5 In the recently completed GALA trial, the median delay to surgery was 80 days.4 In 1997, an audit of UK practice found that the median delay from index event to surgery was 189 days,6 falling to 45 days when repeated in 2007.7 While the latter example represents a considerable improvement in practice, it is nowhere near the AAN and AHA ‘two week’ recommendation and ‘light years’ from achieving the ‘48 hour’ aspiration of the United Kingdom Government.

So, why is there such a discrepancy between national/international recommendations and ‘real world’ practice? The answer is multifactorial but requires governments and health providers to consider cerebral vascular disease on a par (in terms of political and fiscal priority) with cancer and heart disease. More controversially, it requires surgeons/interventionists to recognise that the current pre-occupation with treating vast numbers of asymptomatic patients will do little to reduce the overall burden of stroke. It is an uncomfortable fact that even if one could identify every patient with an asymptomatic 60–99% stenosis and perform CEA/CAS with a 2.3% procedural risk, 97% of strokes destined to occur in the community will still happen.4, 8

This paper reviews the evidence underlying the current drive towards expedited intervention in patients presenting with TIA and minor stroke. It will also try to provide reassurance to the surgeon as to how such a strategy can also be reconciled with a recent report9 suggesting that “early CEA in symptomatic patients is associated with poorer perioperative outcomes.”

Level I, Grade a Evidence… 

Most interventionists and surgeons cite data from the international, randomised trials as being the level I, Grade A evidence base upon which their clinical practice is founded (Table 1). In effect, ECST, VA and NASCET showed that ‘recently symptomatic’ patients (defined as events occurring within the preceding 6 months) with 50–99% NASCET stenoses benefited from ‘intervention’, whilst ACAS and ACST found that surgery was of significant benefit in asymptomatic patients with 60–99% stenoses.10, 11, 12, 13, 14 Armed with such high quality data from more than 10,000 randomised patients, the only remaining issue is usually whether that ‘intervention’ should still be surgery or carotid angioplasty with stenting (CAS).

Table 1. Principle outcomes from the randomised trials

	(a) Carotid Endarterectomy Trialists Collaboration∗: Five year risk of ‘any stroke’ (including 30-day stroke/death) from the VA, ECST and NASCET trials10, 11, 12
	Trial	Stenosis	n	30-day CEA risk	Five Year Risk		ARR	RRR	NNT	Strokes prevented per 1000 CEAs
					Surgery	Medical
	CETC	<30%	1746		18.36%	15.71%	−2.6%	n/b	n/b	none at 5 years
	CETC	30–49%	1429	6.7%	22.80%	25.45%	+2.6%	10%	38	26 at 5 years
	CETC	50–69%	1549	8.4%	20.00%	27.77%	+7.8%	28%	13	78 at 5 years
	CETC	70–99%	1095	6.2%	17.13%	32.71%	+15.6%	48%	6	156 at 5 years
	CETC	String	262	5.4%	22.40%	22.30%	−0.1%	n/b	n/b	none at 5 years

	(b) ACST & ACAS13, 14: Five year outcomes (including 30-day death/stroke)
	Trial	Endpoint	30-day CEA risk	Five Year Risk		ARR	RRR	NNT	Strokes prevented per 1000 CEAs
				Surgery	Medical
	ACAS	ipsilateral stroke	2.3%	11.0%	5.1%	+5.9%	54%	17	59 at 5 years
	ACST	‘any stroke’	2.8%	11.8%	6.4%	+5.4%	46%	19	53 at 5 years

n/b=no benefit conferred by CEA, ARR=Absolute Risk Reduction, RRR=Relative Risk Reduction, strokes prevented per 1000 CEAs=number of strokes prevented at five years by performing 1000 CEAs at the risk quoted.

∗ data derived from the CETC10, 11, 12 with all pre-randomisation angiograms remeasured using NASCET method.

Not surprisingly, attempts to divert contemporary debate away from whether CEA or CAS is preferable and back towards identifying who benefits most from intervention is not always welcomed. For many, the rationale is simple. If, for whatever, reason, CEA/CAS cannot be performed within six months of the index event, then surely the results from ACAS and ACST justify any intervention thereafter? Why make it any more complicated?

At first sight, this interpretation of the trials seems logical, reasonable and difficult to counter. However, anyone who does read beyond the ‘headlines’ will realise that this is far too simplistic an interpretation. It is an uncomfortable fact that the system currently allows most practitioners of CEA and CAS to treat myriads of relatively ‘low risk for stroke’ patients within nationally accepted guidelines, while the really ‘high risk’ patients suffer strokes with little chance of undergoing any intervention. Just look again at Table 1 to see just how few strokes are prevented at five years by operating on 1000 patients with symptomatic 50–69% or 60–99% asymptomatic stenoses.

To this observer, the ongoing debate should not just be about whether ACAS/ACST justifies intervention should the 6 month ‘symptomatic’ threshold elapse (they probably do), nor whether CEA or CAS is ‘king’ (just do the randomised trials properly), but can health providers, physicians, surgeons and interventionists change attitudes and practice and then prioritise resources towards those who really need our help. Contrary to political and professional preferences, this does not mean expending resources on asymptomatic carotid screening programmes as the number one priority. As will be shown, this money would be far better being used to expedite the investigation and treatment of symptomatic patients within one/two weeks of presentation. Once that has been achieved, it is then reasonable to treat as many asymptomatic patients as you like!

Something Old… 

In common with many of you, I was taught that; (i) CEA should be delayed for 6–8 weeks after a stroke because of the increased risk of haemorrhagic transformation of the infarct, i.e. emergency surgery was dangerous, (ii) early/expedited surgery (in general) was probably associated with an increased rate of complications, so why expose yourself to unnecessary medico-legal risk, especially as (iii), the risk of suffering a stroke in the first few weeks after presentation was probably not really that high and (iv) early symptom resolution was generally taken to be a sign that urgent investigation/treatment was unnecessary.15 Interpretation: A little bit of delay in the system probably does no real harm and might, actually, be beneficial to the patient (and possibly to the surgeon regarding published risks). Outcome: No professional impetus to change the way the system works.

This nihilistic attitude persists into the modern era, is rarely mentioned in debates about developing the roles of CEA and CAS and was completely ignored in a recent survey of doctor's attitudes towards “treating stroke as a medical emergency”. In this study, Wang16 ascertained that there was; (i) room for improvement in recognising symptoms due to stroke/TIA (that in itself is a damning indictment of attitude and practice), (ii) scepticism regarding the role of CEA in asymptomatic individuals, which varied according to the doctor's specialty (no surprise there then) and (iii) 28% of physicians would not administer tissue plasminogen activator in acute stroke (but at least the majority knew it was beneficial if administered quickly). However not one single question was directed towards ascertaining whether doctors knew that rapid access to surgery was highly beneficial to the very recently symptomatic patient.

Something New… 

The recent drive towards promoting expedited investigation and intervention has not generally been led by surgeons (with a few notable exceptions)17, 18 but rather by Neurologists and Stroke Physicians. In a series of surprisingly ‘basic’ methodological studies and provocative editorials, they have challenged accepted dogma and thrown down a very large gauntlet for their surgical colleagues to respond to.19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31

The first myth to be challenged by the Neurologists was the conventional teaching that the early risk of stroke after TIA/minor stroke was only about 1–2% at 7 days and 2–4% at 30 days. These data were primarily derived from a collection of ‘cohort’ studies,19, 22, 23, 24 which had largely stood the test of time and which generally reassured surgeons that there was no real need to change existing practice and intervene urgently. However, these cohort studies tended to recruit their patients some time after the index event, patients suffering a stroke in the intervening period were excluded, as were those who suffered a stroke on the same day as their TIA. More importantly, most patients were recruited following attendance at the outpatient clinic or accident and emergency department and few were derived from prospective community/population based environments.19, 22, 23, 24 In short, these cohort studies were heavily biased in their design and they influenced practice in a manner more than was justified.

In a systematic review of 18 independent cohort studies which reported the early risk of stroke in 10,216 patients presenting with a TIA, Giles observed that the pooled risk of stroke was 5.2% at seven days, i.e. about three to four times greater than is traditionally quoted.31 However, this study encountered some of the limitations alluded to earlier (e.g. many of the constituent trials recruited from outpatient clinics or the accident and emergency department etc). A separate meta-analysis of three population based studies (which used ‘face to face follow-up’) observed that the risk of stroke at 2 days was 6.7% and was 10.4% at seven days.31 Population based studies are more likely to represent ‘real world’ practice and these figures suggest that the highest risk period for the TIA patient is the first seven days after the index event. How many medical and surgical units are ‘geared up’ to deal with TIA/minor stroke patients this quickly?

One of the three population based studies23 was particularly informative as it also reported early stroke risks for patients presenting with a minor stroke as well as TIA (Fig. 1). Note that there was no significant difference in the early risk of recurrent stroke at 7, 30 and 90 days relative to mode of presentation. ‘Conventional teaching’, however, dictates that patients suffering a minor stroke should wait 6–8 weeks for CEA. As was actually observed in Coull's study, by that time, most of those destined to suffer a stroke will have already done so.23 The surgeon is, in effect, left with a ‘safer’ cohort of patients whose long term risk of stroke is probably much less.

View Large Image Download to PowerPoint

Fig. 1 Cumulative risk of stroke after transient ischaemic attack (TIA) or minor stroke. Reproduced with permission from Coull and the British Medical Journal.23

In an alternative approach to evaluating ‘early stroke risk’, Rothwell combined four independent patient databases comprising 2416 patients who had presented with an ischaemic stroke.26 Approximately one quarter of these stroke victims (549 patients) reported a preceding TIA. Almost one fifth (17%) of the preceding TIAs occurred on the same day as the stroke, 9% happened on the day prior to the stroke, while 43% of TIAs (overall) occurred within the seven days prior to stroke onset.

Of course, not all strokes are secondary to thrombo-embolic disease from the carotid artery and this could confound meaningful interpretation of the data. Approximately 25% of strokes are due to small vessel intracranial disease, while 15% are cardioembolic. In order to correct for the effect of the underlying aetiology on the risk of recurrent stroke, Lovett et al. performed a meta-analysis of four studies which stratified for underlying aetiology.24 Table 2 summarises the principle findings. Early recurrent stroke was extremely rare after small vessel (lacunar) stroke. However, stroke due to ‘large artery’ disease (i.e predominantly involving the carotid artery) was associated with the highest rate of early recurrent stroke (4% at 7 days, 12.6% at 30 days and 19.2% at 3 months).

Table 2. Cumulative recurrent stroke risk stratified for underlying aetiology∗

		7 days	1 month	3 months
	‘large vessel disease’	4.0% (95%CI 0.2–7.8)	12.6% (95%CI 5.9–19.3)	19.2% (95%CI 11.2–27.2)
	cardioembolic	2.5% (95%CI 0.1–4.9)	4.6% (95%CI 1.3–7.9)	11.9% (95%CI 6.4–17.4)
	‘small vessel disease’	0.0%	2.0% (95%CI 0.0–4.2)	3.4% (95%CI 0.5–6.3)

∗ data derived from meta-analysis by Lovett.24

The studies detailed so far clearly suggest that the early risk of stroke after TIA/minor stroke is much greater than previously thought and most centres will logistically be unable to immediately change practice in order to investigate and treat these patients more quickly. It would, therefore, be helpful (in the interim) if ‘higher risk’ subgroups of patients could be identified for ‘fast-tracking’. A number of studies have tried to identify clinical/imaging parameters that are predictive of an increased risk of early recurrent stroke. These include presentation with; (i) motor deficit,19, 32 (ii) dysphasia or dysarthria,19, 32 (iii) contralateral occlusion, (iv) exhausted cerebral vascular reserve,21 (v) age >60 years19 and (vi) diabetes mellitus.19 This theme was developed further by Rothwell in the form of the ‘ABCD’ scoring system (Table 3), superseded by the ABCD2 version.27, 33 The origins and validation of this predictive scoring system has been discussed in greater detail in the March Carotid Masterclass.34 In summary, TIA patients with an ABCD score of 0–3 incurred a 0% seven day risk of stroke, increasing to 2.2% for a score of 4, 16.3% for a score of 5 and 35% for patients with a maximum score of 6.27 In the future, such scoring systems could become invaluable in identifying patients meriting emergency admission after suffering their TIA/minor stroke (eg those with an ABCD score of 5+).

Table 3. ABCD criteria for predicting very early stroke risk∗

				Score
	A	Age ≥60 years		1
		Age <60 years		0
	B	BP: systolic ≥140mmHg or diastolic ≥90mmHg		1
		BP: neither of the above		0
	C	Clinical features:	Unilateral weakness	2
			Speech disturbance, no weakness	1
			Other	0
	D	Duration of symptoms	≥60 minutes	2
			10–59 minutes	1
			<10 minutes	0
	Maximum Score			6

∗ details derived from Rothwell.27

Something Borrowed… 

Table 1 summarises the principle findings from the randomised trials that have influenced practice most of all. It will be noted that no attempt has been made to separate NASCET, ECST and VA outcomes in the symptomatic cohort of patients. This is because the Carotid Endarterectomy Trialists Collaboration (CETC) provided an invaluable service by combining all of these data together, having remeasured all of the pre-randomisation angiograms using the NASCET measurement method.10, 11, 12 This database of >6000 patients is now the largest and most comprehensive set of outcome data in symptomatic patients with carotid disease (despite being somewhat historical). More importantly, it's very large size has enabled a number of important subgroup analyses to be performed that would go on to challenge many preconceived notions about what really constitutes being ‘recently symptomatic’.

Table 4 details some of the CETC findings in a format slightly different to that used in Table 1. The absolute risk reduction conferred by surgery (ARR), the number needed to treat to prevent one stroke at 5 years (NNT) and the number of strokes prevented per 1000 CEAs at 5 years (CVA/1000) are detailed for three NASCET stenosis subgroups (50–69%, 70–99% and 50–99%) and stratified according to the delay between onset of symptoms and randomisation. In the trials, CEA was performed (on average) within 7 days of randomisation (P Rothwell, personal communication). As can clearly be seen, the benefit conferred by CEA was maximum if the delay between symptom onset and randomisation was <2 weeks. In symptomatic patients with 50–69% stenoses, the ARR conferred by surgery fell rapidly with time with no benefit being apparent after 12 weeks had elapsed (how many of you even consider this fact when discussing the rationales of surgery with some of your patients?).

Table 4. Absolute risk reduction conferred by CEA in the 5 year cumulative risk of ipsilateral carotid territory ischaemic stroke (including the peri-operative risk) in patients with a NASCET 50–99% stenosis, stratified for delay from index event to randomisation∗

	Time since randomisation	50–69% stenoses			70–99% stenoses			ALL 50–99% stenoses
		ARR	NNT	CVA/1000	ARR	NNT	CVA/1000	ARR	NNT	CVA/1000
	< 2 weeks	14.8	7	148	23.0	4	230	18.5	5	185
	2–4 weeks	3.3	30	33	15.9	6	159	9.8	10	98
	4–12 weeks	4.0	25	40	7.9	13	79	5.5	18	55
	>12 weeks	−2.9	nil	nil	7.4	14	74	0.8	125	8

nil=no benefit conferred by CEA, ARR=Absolute Risk Reduction (%) conferred by surgery over best medical therapy, NNT=number of CEAs needed to be performed to prevent one stroke at 5 years, CVA/1000=number of strokes prevented at five years by performing 1000 CEAs.

∗ data derived from the CETC10, 11, 12 with all pre-randomisation angiograms remeasured using NASCET method.

In symptomatic patients with 70–99% stenoses, the benefit conferred by surgery again diminishes with increasing delays to surgery (< 2 weeks, ARR=23% versus>12 weeks, ARR=7.4%), although the cumulative benefit was greater than that observed in similar time delayed patients with 50–69% stenoses. One important caveat to be borne in mind, however, is that the ‘70–99%’ and ‘50–99%’ outcome data in Table 4 does include patients with angiographic evidence of ‘near occlusion’. The CETC have shown that CEA does not confer significant benefit in patients with this pattern of disease.10 If these patients are excluded, the ARR conferred by CEA at 5 years increases to 30.2% for patients randomised<2weeks with 70–99% stenoses and to 17.6% (2–4 weeks), 11.4% (4–12 weeks) and 8.9% (>12 weeks).

These data were some of the first to really challenge surgeons to question how their own practice influenced long term patient benefit. Put quite simply, surgeons who performed CEA very rapidly prevented far more strokes than those whose practice allowed excessive delays to occur. But there was more…

One further very important issue emerged from the CETC secondary analyses; gender.12 Interestingly, not one national guideline includes any recommendation that gender might be an issue when discussing the rationales of intervention with symptomatic patients. Fig. 2 details the number of strokes prevented at five years by performing 1000 CEAs in patients with 50–69% and 70–99% stenoses, having now stratified for gender. As can be seen, males appear to gain continuing benefit from CEA (albeit cumulatively reduced with time) for both moderate (50–69%) and severe (70–99%) stenosis subgroups.4 Contrast this with what appears to be happening in females. Notwithstanding methodological criticisms regarding the validity of subgroup analyses and whether the trials were adequately powered, the CETC data suggest a very clear trend towards little benefit accruing to the female patient if surgery is deferred beyond four weeks, even if severe carotid disease was present. This author does not suggest that we should now stop operating on all symptomatic female patients should 4+ weeks elapse, but can anyone really look at the data in Table 4 and Fig. 2 and still say that ‘delay to surgery’ does not matter?

View Large Image Download to PowerPoint

Fig. 2 Number of strokes prevented at 5 years by performing 1000 CEAs. Effect of gender and timing from event to CEA on prevention of late stroke relative to degree of stenosis (recalculated from CETC data10, 11, 12 and reproduced with permission from AR Naylor).4

Something Blue… 

The rationale underlying early surgery in patients presenting with TIA and minor stroke includes; improving overall cerebral perfusion, reducing cumulative neuronal loss by restoring blood flow to the ischaemic penumbra, preventing early progression towards thrombosis and, of course, removing a source of ongoing embolism.

But doesn't early surgery increase the procedural risk and so negate any long term benefit to the patient? This is another very important and often confounding issue. There have been a number of publications citing 30-day outcomes in patients undergoing early/urgent carotid surgery, but very few centres have published median delay data for all of their symptomatic patients. In a recent review of the available literature,4 only Vancouver has published a median delay of 14 days from onset of symptoms to surgery for all patients.35 Interpretation: surgeons are aware of the AAN and AHA ‘two week’ recommendations, but most will say that they do not have the infra-structure to achieve it…

Or is there, perhaps, another important factor to be considered in this debate? I suspect that while most surgeons do accept that rapid intervention will capture some of the patients who would previously have suffered a stroke before any intervention could be offered, they remain concerned that any increased procedural risks incurred by implementing such a strategy might lead to unwarranted professional censure and criticism. After all, who wants to be publicly labelled as one of the worst performing carotid surgeons if a little bit of delay in the system conveniently avoided this? Anyone who doubts the ability/intention of the media to criticise individual surgeon's performances might like to read the July 12th edition of the San Diego Union Tribune36 which lists the ‘best’ and ‘worst’ performing cardiac surgeons in the state of California. League tables are fine provided they stratify for case-mix. To-date, no national guideline has ever considered ‘very early surgery’ within its ‘accepted’ risk recommendations.

The paradox of early intervention equating to increased risk was amply highlighted by a very important paper from New York which showed that symptomatic patients undergoing CEA within 4 weeks of the index event incurred a threefold increase in the 30-day death/stroke risk from 1.6% to 5.1%.9 The authors concluded (in their manuscript title) that “early endarterectomy was associated with poorer peri-operative outcomes”, thereafter recommending that it might be better to delay surgery in order to get better procedural results. Given the style of reporting in the San Diego Union Tribune, it might be hard not to agree with them!

So how can this paradox be reconciled, especially as the first half of this paper has provided compelling evidence that the highest risk period for a patient suffering a TIA or minor stroke is the first seven days? Two further sets of evidence and an alternative way of interpreting the literature may provide a solution, provided future ‘Guideline Makers’ recognise that a slight increase in the accepted early procedural risk is acceptable in selected patients.

Is Rockman's experience of increased procedural risk consistent with other published studies? Bond et al.37 performed a systematic review of the published data (11 trials, 4278 patients) and compared the 30-day risk of stroke/death in patients presenting with a minor stroke who underwent either early CEA (<3 to 6 weeks) versus late CEA (>3 to 6 weeks). The findings were consistent across all of the studies in the review and no excess risk was associated with early versus late CEA, provided the patient presented with stable neurological symptoms (Pooled Odds Ratio=1.13, 95% CI 0.79–1.62). This suggests that surgery within the previously accepted 6 week threshold was not associated with excess risk.

Few studies have, however, published outcomes when CEA was performed within the ‘two week’ recommendation laid down by the AAN and the AHA. Table 5 details 30-day outcomes and conversion rates to intracranial haemorrhage in 8 studies published since 2000 where the median delay from onset of events to surgery was <14 days.38, 39, 40, 41, 42, 43, 44, 45 In seven studies, the median delay to surgery was <7 days. There was no standardisation within studies regarding scoring the severity of a ‘stroke’ and only four studies documented that there was independent neurological assessment.

Table 5. 30-Day Death/Stroke rates following CEA when performed within a median of 14 days since onset of symptoms

	Reference	Year	n	Symptom	Median days from event to CEA	Stroke score	30-day death/stroke	ICHa
	Ricco38	2000	72	CVA	5.5 (range 2–15)	NESS=3e	2.8%	0.0%h
	Paty39	2004	72	CVA	<7	‘non-disabling’	2.8%	not stated
	Paty39	2004	131	CVA	<14	‘non-disabling’	3.1%	not stated
	Dorigo40	2007	70	CVA/TIA	2.0	‘non-disabling’	5.4%g	not stated
	Aleksic41	2006	50	CVA/TIA	4.5 (range 1–21)	Rankin <4	6.0%	4.0%i
	Karkos42	2007	42	TIAc	3.0 (range 1–7)	Rankin <4	7.0%	0.0%h
	Sbarigia43	2006	96	CVA	1.5 (range 1–11)	NIHSS <22f	7.3%g	0.0%h
	Rantner44	2006	226b	CVA	12 (IQR 8–19)	66% Rankin ≤2	8.4%g	0.4%h
	Huber45	2003	67	CVA/TIAd	2.0 (range 0–18)	48% Rankin 3–5	16.4%g	1.5%h

a prevalence of intracranial haemorrhage. b Rantner's paper combined data from Eckstein17, Woelfle46 and Rantner.47 c cohort had suffered ≥3 TIAs in preceding 7 days. d mixed cohort including completed strokes, crescendo TIAs and patients with carotid occlusion and severe (Rankin 3–5) neurological deficits. e Neurological Event Severity Score. f NIHSS=National Institute of Health Stroke Scale. g independent neurological verification of outcomes. h post-op CT scans only done in patients with worsening/new neurological deficit. None of these had CT evidence of new haemorrhage. i new haemorrhage present on CT scan, but not associated with symptoms.

The first key message from reviewing these studies was that the overall 30-day death/stroke rates probably were higher than that cited in the ‘popular’ literature. However, the highest 30-day risk (16.4%) was probably not typical of the type of patient normally being considered for early intervention. Huber's study included patients with a high proportion of residual severe neurological deficits (48%) and some with carotid occlusion.45 The second key message was that the rate of conversion to intracranial haemorrhage was reassuringly low, although there was no real consistency regarding the performance of routine post-operative CT scans unless there were complications.

Rantner's study44 (226 stroke patients) combined patients from three contributory databases17, 46, 47 and represents the largest published series to-date. Two thirds of Rantner's stroke patients were classified as ‘minor’ on the modified Rankin system (0–2) and the 30-day death/stroke rate of 8.4% probably represents ‘real world’ practice, especially as it included independent neurologist assessment.

Many of the constituent studies listed in Table 5 were too small to provide reliable guidance regarding which patients with stroke benefited most (and least) from early intervention, but a number of common themes emerged. The literature suggests that patient subgroups in whom one should ‘avoid’ early surgery include; (i) evidence of recent carotid occlusion,41 (ii) Rankin score ≥341, 44 (Table 6), (iii) ischaemic infarct >two thirds of the middle cerebral artery territory43 (iv) no neurological plateau44 (v) fluctuating conscious levels and (vi) evidence of intracranial haemorrhage on CT scan. The worst outcomes after early CEA (23% death/stroke rate at 30-days) were seen in stroke patients presenting with the combination of a Rankin score ≥3+an ASA score >2+CEA within 12 days.44 This excessively high procedural risk compares with a 30-day death/stroke rate of 6.4% in patients presenting without this combination.

Table 6. The modified Rankin scoring system48

	0	no deficit
	1	minimal deficit without any functional impairment
	2	minor deficit, slight functional impairment, patient independent
	3	moderate deficit, patient can walk alone, needs some assistance in daily life
	4	severe deficit, patient can walk only with assistance
	5	disabling stroke, patient confined to bed or wheelchair
	6	dead

If one accepts that Rantner's data from a large cohort of independently monitored patients undergoing surgery within 14 days of a stroke reflects ‘real world’ practice, is there any evidence to suggest that their 8.4% procedural risk represents poor surgical practice? I suspect that many surgeons will, intuitively, say yes. However, the data from Fig. 3 might convince them otherwise! Fig. 3 presents an alternative way of interpreting the CETC database. Because the CETC provided the ‘raw’ 5 year medical and surgical stroke risks as well as the procedural risks stratified for the effect of ‘delay to surgery’, it is possible to model the benefit conferred by CEA depending on different delays to surgery and procedural risks.4 For example, a surgeon operating within 2 weeks of the index event with a 10% procedural risk (which most of you I suspect will think unacceptable) will still prevent 150 strokes at 5 years per 1000 CEAs performed. Despite the very high initial risk used in this example, the CETC data suggest that this surgeon will still prevent more strokes in the long term than if he/she waited for >4 weeks and then operated with zero risk! Bear that in mind the next time you are asked to interpret published outcomes in the literature or are tempted to criticise surgeons/interventionists who operate very early!

View Large Image Download to PowerPoint

Fig. 3 Strokes prevented per 1000 CEAs at 5 years stratified for; (i) delay from last event to surgery and (ii) 30-day death/stroke risk (recalculated from CETC10, 11, 12 and reproduced with permission from AR Naylor).4

In Conclusion 

To this observer, the natural history data documenting the high early risk of stroke after suffering a TIA/minor stroke are quite compelling. The system currently ignores the very highest risk patients and commits far too many resources towards treating relatively low risk patients. The ‘48 hour’ aspirations of the UK Government are almost certainly beyond the reach of most centres, but a seven day target to treatment would be an absolutely massive improvement in practice and will prevent far more strokes in the long term than by treating large numbers of asymptomatic individuals. However, ‘Guideline Makers’ must recognise that a drive towards rapid intervention will probably come at the expense of a slightly increased procedural risk. Unless this is recognised, few surgeons will change their practice for fear of being labelled as ‘poorly performing’. As a (personal) suggestion, what about considering that a 30-day death/stroke risk of 8% is acceptable if CEA is performed within 2 weeks of the index event, reducing this threshold of acceptable risk to 6% if surgery is performed between 2–4 weeks and down to 4% if CEA is delayed beyond 4 weeks. If you are still moved to disagree with this suggestion, look again at Fig. 3 and see for your self which threshold is preventing the most strokes in the long term!