INTRODUCTION
Recent articles have extensively reviewed the literature on the
treatment of unruptured cerebral aneurysms (UIAs). The Stroke Council
of the American Heart Association provides guidelines for the
management of UIAs (Bederson et al., 2000),
but many of the complex issues associated with the decision to treat
UIAs remain controversial. This is readily apparent from the detailed
comments of a number of experts published in a recent issue of the
Journal of Neurosurgery (Dumont et al.,
2002; Juvela, 2002; Weir, 2002; Weir et al, 2002;
Wiebers et al., 2002). There is agreement regarding
the treatment and management of some patients with UIAs but minimal
consensus is found in the literature regarding the treatment of
patients in the 40 to 70 year age group with unsymptomatic UIAs less
than 10 mm in diameter, especially if they have not suffered a previous
subarachnoid hemorrhage (SAH).
The crucial question is whether a patient will be better off if an
aneurysm is left untreated, with the patient followed-up at regular
intervals and advised to reduce risk factors, or if the aneurysm is
occluded. The obvious risks of surgery include the manipulation of the
brain during clipping, and potential surgical complications. Coiling
carries a high risk that the aneurysm will not be completely occluded,
as according to a review by Brilstra et al. (1999) only 54% are completely occluded, in addition
to the uncertainty about the long-term effects of coiling. Rupture of
the aneurysm is a risk of both types of treatment.
Given that the patients for whom treatment decisions are most
difficult tend to be healthy middle-aged adults with many years of
active working and social life ahead of them, it is important to take
into account the long-term consequences of either harboring an UIA, or
having it treated. A 50-year-old patient with a 10 mm UIA on the
anterior communicating artery may well be advised to have the aneurysm
treated, but if the likely long-term consequences of this included even
a small decrease in cognitive functioning, then some patients might
prefer to leave well alone. On the other hand, patients who are
inclined to worry about their health, and perceive an UIA as a
“time-bomb” ticking in their head, might achieve a better
quality of life if their aneurysm were occluded, even if the downside
was a possible diminishing of some cognitive functions. Research
looking at the long-term cognitive, psychosocial and quality of life
outcome following treatment of UIAs is required in order to provide
patients with useful data that can inform their decisions about the
best treatment option for them. Such information is especially
pertinent for those patients in the “gray zone”; that is
when there is no clear-cut evidence for treatment versus
conservative management in their particular case.
Most of the existing studies have assessed outcome from UIA treatment
in terms of mortality and neurological morbidity. In comparison very
few studies exist which include measures of outcome such as cognitive
status, psychosocial functioning and quality of life. Where cognitive
outcome has been included, methodology differs across studies. In his
review Weir (2002) briefly covers outcome
research, but he does not critique these studies in any detail. Making
sense of studies that include measures of cognitive and psychosocial
outcome requires a good understanding of neuropsychological and
quality-of-life research. For example, a study using the Mini-Mental
State Examination (Folstein et al., 1975) at
6 weeks post treatment as the only outcome measure cannot be compared
with a study that uses a carefully constructed neuropsychological
battery of tests pre-treatment, and again at 6 to 12 months post
treatment. In this paper we will review and comment on outcome studies
published to date, and make suggestions for future research.
Unruptured Intracranial Aneurysms
An aneurysm is formed when a region of the blood vessel wall or
artery weakens and balloons out to form a sac-like structure (Figure 1). A UIA is classified as an aneurysm
without historical or pathological evidence of a breach through the
artery wall (Weir, 2002). The presence of an
UIA can come to clinical attention in several ways, including discovery
when investigating headaches, seizures or embolic events (Orz et al., 2000). UIAs can also be discovered when
treating a ruptured aneurysm, or they may be found incidentally during
the assessment of an unrelated problem (Orz et al.,
2000). Differences in the manner in which UIAs come to clinical
attention have led to variations in the terminology used to describe
these aneurysms. For example, the term “symptomatic
aneurysm” is used to describe aneurysms that rupture, or
aneurysms that produce neurological symptoms by space occupying
effects. “Asymptomatic aneurysms” are not responsible for
clinical presentations and are typically found in patients with an
additional symptomatic aneurysm, or when conducting investigations in
patients who are at risk of harboring an aneurysm. “Incidental
aneurysms” are those found unexpectedly in patients undergoing
investigation for other suspected pathology (Wardlaw
& White, 2000). In this paper the generic term
“unruptured intracranial aneurysm” (UIA) is used to
describe all three categories of unruptured aneurysm.
Radiographic image of an unruptured intracranial aneurysm.
Decisions to treat UIAs are said to depend on the relative risk of
subsequent spontaneous rupture and SAH in untreated patients
versus the risks of treatment (Orz et al.,
2000). Unfortunately, accurate knowledge regarding the natural
history of aneurysms is still lacking (Asari &
Ohmoto, 1993; Juvela et al., 1993),
although an annual rate of rupture of between 1–2% is in the main
well accepted (Juvela et al., 2000). Several
potential risk factors for aneurysm formation and rupture have been
discussed extensively in the literature and the interested reader is
referred to the excellent review article recently published in the
Journal of Neurosurgery (Weir,
2002). Briefly, increasing aneurysm size and aspect ratio have
been hypothesized to present a significant risk in terms of intrinsic
features of the aneurysm itself whilst cigarette smoking, alcohol
intake and high life event stress appear to present the greatest risk
in terms of modifiable risk factors. Age is also thought to be an
important risk factor with it suggested that the peak risk for aneurysm
rupture occurs between the ages of 40 to 60 years (Samson, 1996). Factors related to pregnancy and
hormone changes following menopause also appear to be important but at
this stage are largely uninvestigated.
Outcome following treatment for an UIA is also still largely open to
debate. Currently there are generally three main management techniques
employed. These are surgical clipping, endovascular therapy with
coiling and “best medical therapy,” otherwise known as
conservative management (lifestyle modifications and ongoing
monitoring). With both surgery and endovascular therapy (Figure 2) the aim is to exclude the aneurysm from
the blood circulation to prevent future rupture and/or to relieve
any symptoms of mass effect associated with the aneurysm (Molyneux, 2000), whilst at the same time not
producing adverse cognitive or neurological symptoms. Additional
factors that need to be considered when selecting the optimal treatment
method are the cost of the treatment, length of hospital stay and
duration of rest and recovery period (Martin,
1997).
Endovascularly coiled aneurysm (A) and surgically clipped aneurysm
(B) (Johnston et al., 2000).
Mortality and Neurological Morbidity Following
Surgical Treatment
Based on a review of 30 studies conducted between 1977 and 2001
(Table 1) the gross outcome following
clipping of an unruptured aneurysm appears to be generally good. The
range of reported morbidity and mortality values does, however, vary
widely, falling anywhere between 0–7% for mortality and between
5–25% for morbidity. These variations can largely be accounted
for by the differences in measures of morbidity used, time frame of
assessment of outcome, exclusion and inclusion protocols for patient
populations, and different demographics of patient populations.
Surgical and endovascular mortality and morbidity
For example, studies with lower morbidity and mortality
figures typically did not include larger aneurysms or poor risk
candidates (Wirth et al., 1983). Poor
surgical outcome is generally reported by studies that include patients
with larger aneurysms, aneurysms in difficult-to-treat locations and
aneurysms in the elderly (Solomon & Baker,
1992). Another variable affecting the mortality and morbidity
figures is hospital volume, with those hospitals more
frequently performing aneurysm operations having lower mortality (Solomon et al., 1996). These, and other
methodological differences, make comparisons between the various
studies problematic.
In order to overcome the limitations of many of the studies and the
difficulty in comparing results, several investigators have opted to
pool studies and conduct meta-analyses of the findings. Selection
protocols for meta-analysis studies are typically rigorous and ensure
that only studies with similar methodological protocols are compared.
One such meta-analysis has been conducted by King et al. (1994). In their study only patients who had
undergone surgery for incidental or multiple aneurysms were included,
and all patients with symptomatic aneurysms were excluded. They also
included only patients who had a good neurological status prior to
treatment. Surgical mortality was defined as any death within 30 days
of surgery. Morbidity was defined as a permanent significant
neurological deficit and was based on the longest reported follow-up
period. Their search of the literature produced a 28-case series
containing a total of 733 eligible patients. They reported a range in
morbidity rates from 0.0% to 16.7% and a range in mortality of 0.0% to
7.7%, with a combined morbidity of 4.1% and a mortality of 1.0%.
Raaymakers et al. (1998) have also
conducted a meta-analysis of studies based on surgery for unruptured
aneurysms, but with slightly different criteria to those of King et al.
(1994). Unlike King et al. (1994), Raaymakers et al. (1998) included in their analysis studies of
patients with symptomatic aneurysms and patients who had neurological
signs or symptoms preoperatively. They also included higher proportions
of posterior circulation aneurysms and giant aneurysms. The analysis of
Raaymakers et al. (1998) included 61 studies
with 2,460 patients. From these studies they reported a combined
mortality of 2.6% and a permanent morbidity of 10.9%. They noted that
morbidity percentages tended to be higher in studies they considered to
be of “high” quality.
Recently the International Study of Unruptured Intracranial Aneurysm
Investigators (ISUIA; 1988) conducted a large
scale multi-center study of UIAs in which they found the morbidity and
mortality figures for treatment of unruptured aneurysms to be much higher
than previously reported by King et al. (1994),
and slightly higher than those reported by Raaymakers et al. (1998). Morbidity and mortality figures were
based on 996 cases that proceeded to surgery. Mortality was 1.8% 30
days after surgery and 3.6% 1 year after surgery. The ISUIA (1988) study is interesting and, at the time it was
published, was the only study to have utilized cognitive as well as
neurological measures of outcome. The measures used by the ISUIA (1988) study were the Telephone
Interview for Cognitive Status (Brandt et al.,
1988) and the Mini-Mental State Examination (Folstein et al., 1975). When based on the Rankin
Disability Scale (Rankin, 1957) alone, a
brief scale of functional outcome following a stroke, morbidity 30 days
after surgery was 3.6%, and 1 year after surgery it was 2.9%. When
based on cognitive status alone, 30 days after surgery morbidity was
6.3% and 1 year after surgery it was 6.1%. When based on either the
Rankin scale, or cognitive status, or both, morbidity 30 days after
surgery was 15.0% and one year after surgery it was 12.0%. Although at
1 year the patients without a history of SAH had a slightly poorer
outcome in terms of overall morbidity, when it came to measures of
cognitive status alone, patients with a history of SAH had a poorer
outcome. The authors attributed this to the effect of three consecutive
cerebral events (1 SAH and 2 craniotomies) on mental status. These
findings raise several interesting points. First, basing measures of
outcome on either neurological status or cognitive status alone can be
misleading. Second, morbidity changes with time, and typically will
improve given time. Studies that report morbidity figures based
only on short follow-up periods may therefore result in overestimates
of morbidity outcomes. Third, the ISUIA (1988)
study fails to adequately control for the effect of other factors on
morbidity, such as previous SAH.
In summary, the generally favorable outcome following surgery for
unruptured aneurysms combined with extensive research demonstrating the
poor outcome following the rupture of an aneurysm (Ogden et al., 1990, 1993, 1994) have been used to argue in favor of treatment
of unruptured aneurysms to avoid the possibility of future rupture
(Eskesen et al., 1988; Heiskanen, 1986; Mizoi et al.,
1989). The ISUIA (1988) study concludes
that mortality and morbidity rates following surgery for unruptured aneurysms
are higher than the rate of spontaneous rupture if the UIA is left untreated.
However this conclusion has been criticized by several researchers
(Alexander & Spetzler, 1999; Debrun, 1999; Hashi,
1999; Kobayashi & Orz, 1999;
Solomon, 1999). The ISUIA study does nonetheless raise the possibility
that surgical risks might outweigh the benefits of occluding the
aneurysm (Roy et al., 2001), and hence the
controversy remains unresolved.
Mortality and Neurological Morbidity Following
Endovascular Treatment
In contrast to the information available about surgical clipping,
data on the effectiveness and risks of endovascular coil embolization
for UIAs is more tentative (Wardlaw & White,
2000). This technique has only been in widespread use since
1991, and is undergoing constant refinement. Thus long-term follow-up
is not yet well documented (Kahara et al.,
1999; Lot et al., 1999). In addition,
at this stage, endovascular coiling and surgical clipping tend to be
used in different patient populations (Viñuela et al., 1997). For example, in most
studies the indications for endovascular treatment remain that surgical
treatment is unsuitable because of anatomical considerations, patient
medical condition, or patient age. For this reason, how the two
techniques compare in equivalent patient populations is as yet unknown
(Johnston et al., 1999a). A more extensive
analysis of the long-term results of treatment with endovascular
coiling is considered to be particularly important as such a treatment
may provide a less invasive alternative to surgery (Wardlaw & White, 2000).
Many of the existing studies on the efficacy of endovascular
embolization using coiling have been conducted on SAH populations
(Byrne et al., 1999; Casasco et al., 1993; Lempert et
al., 2000; Vanninen et al., 1999;
Viñuela et al., 1997), or on a
combination of ruptured and unruptured cases where the two groups are
not separately analyzed (Cognard et al.,
1999; Raftopoulos et al., 2000; Steiger et al., 1999; Zubilaga
et al., 1994). In other studies the numbers of UIAs in the
various groups in the study are too small to make useful comparisons
(Kahara et al., 1999; Redekop et al., 1999; Regli et
al., 1999). Only a limited number of studies have reviewed
coiling in samples consisting of UIAs. It has, however, been argued
that studies of UIAs provide more useful information about treatment
outcome than SAH studies because patients are generally minimally
impaired and therefore any new deficits can be attributed to the
treatment rather than the hemorrhage (Johnston et
al., 2000). These UIA studies report a wide range of mortality
percentages, from 0–13%, and morbidity percentages from
5–20% (see Table 1).
A recent meta-analysis of studies of embolization suggested that
treatment-related complication rates are similar to those reported
following surgery, with a morbidity of 3.7% and a mortality of 1%
(Brilstra et al., 1999). Brilstra et al.
(1999) however noted that longer-term results
are needed before final conclusions can be drawn. This is particularly
important because of the possibility of recanalization and rebleeding
in incompletely occluded aneurysms. In addition it is unlikely that the
group of patients treated endovascularly and reviewed in the
meta-analysis by Brilstra et al. (1999) are
equivalent to patients reviewed in studies of surgical outcome.
Comparing studies of outcome following coiling and clipping is
problematical because the two treatment methods are not considered
equally appropriate for every case. In addition Brilstra et al. (1999) advises caution when comparing the two
techniques because of the widely heterogeneous nature of study designs,
patients and aneurysms. A few studies have, however, attempted to
compare outcome following the two procedures, although with varying
conclusions. As an example Leber et al. (1998) concluded that coiling with Guglielmi
electrolytically detachable coils (GDCs) is as safe and effective as
clipping in cases of UIAs. However, it should be noted that in their
study there were many more patients enrolled in the endovascular group
than in the surgical group, with the reason for this not given.
In another study that compares endovascular and surgical outcome in a
large multi-center cohort of patients with UIAs, Johnston et al. (1999a) conclude that adverse outcomes were
significantly more common in surgical cases when compared to
endovascular cases. Mortality, however, was not significantly affected
by treatment type. The authors do acknowledge several limitations of
their study. For example, an adverse outcome was defined as in-hospital
death or transfer to a nursing home or rehabilitation hospital at
discharge. As they note, recovery from surgery may be slower, but more
complete, and this would not be reflected in their measure of adverse
outcome. Further, although variables of gender, age, race and admission
source (emergency room, transfer or elective), were controlled for
statistically when comparing the two treatment groups, other important
variables, such as aneurysm characteristics, could not be controlled
for in the study (Johnston et al., 1999a). As
noted by Johnston et al. (2000) in a later
article, the previous study cannot rule out that better outcome
following coiling was a result of the selection of lower risk
candidates for this procedure. Nonetheless the authors concluded that
despite the limitations of the study, the findings are important and
provocative. Although they cannot provide a definitive basis for
recommending coil embolization they do consider their study provides
impetus for further studies comparing the two methods.
The later paper by Johnston et al. (2000)
reports on a similar study conducted at a single institution. From that
study the authors draw a very similar conclusion that endovascular
treatment of UIAs seems to be significantly safer than surgical
clipping. They found that surgery was associated with greater rates of
early and persistent disability, more procedure-related complications
and a greater delay in return of function. They do note, however, that
the rupture of three aneurysms in the endovascular patients during
follow-up compared with one in the surgical group highlights the need
for longer-term follow-up. As with the first study they also note that
there may be unmeasurable factors contributing to the differences
between outcomes in groups (Johnston et al.,
2000).
To summarize, it has been suggested that it might be too soon to make
durable comparisons between surgical and endovascular methods as it is
quite likely that the endovascular technique will continue to undergo
further modifications, with alternative devices possibly providing more
complete and stable aneurysm occlusion (Bavinzski et
al., 1999). Combined studies may be particularly useful in
evaluating the belief that the combination of surgical clipping and
coiling in management protocols may help to improve the overall outcome
of all cases (Johnston et al., 2000; Sturaitis et al., 2000). Combined approaches may
also be particularly useful in cases of complex intracranial aneurysms
often deemed too difficult for standard treatment (Hacein-Bey et al., 1997). As Martin (1998) comments, endovascular methods should be
considered as a complementary method to surgery, rather than as a
replacement method. They suggest endovascular options extend treatment
to patients with inoperable or high-surgical-risk aneurysms.
Mortality and Morbidity Following Aneurysm Rupture
As mentioned earlier “best medical therapy,” or
conservative management, is considered to be a viable option for the
treatment of UIAs. However, this treatment option in itself carries a
risk—the risk that the aneurysm may rupture at some point in the
future. As already noted, rupture rates of between 1–2% per year
are generally accepted. Unfortunately whilst the number of patients
experiencing good outcome following the rupture of an aneurysm have
increased in the last decade (Cesarini et al.,
1999; Hop et al., 1997; Le Roux et al., 1995), fatality rates are still
high. In the review conducted by Hop et al. (1997) case-fatality rates ranging between 32% to
67% were reported. Further, morbidity rates in those who survive a SAH
also remain unacceptably high and are said to range between
40–50% (Dorsch, 2000; Samson, 1996).
Adding to the already dismal picture Hop et al. (1997) note many of the patients who remain
“independent” after a SAH may experience difficulties
returning to their former level of functioning. Many studies of
reportedly “good” recovery patients, as determined by
neurological outcomes scales, have revealed that many of these patients
still demonstrate deficits in terms of cognitive functioning,
psychosocial functioning and quality of life (Bornstein et al., 1987; Hop et
al., 1998; Ljunggren et al., 1985;
Ogden et al., 1990, 1993, 1994; Säveland et al., 1992).
In a study conducted by Ogden et al. (1990)
100% of the patients were rated as having achieved a good neurological
recovery whilst only 37.5% were rated as having achieved a good
neuropsychological outcome. All patients had achieved full physical
independence but only 63% had returned to their previous level of
leisure activities and only 50% had returned to their premorbid work
levels. Other researchers have reported similar figures (Ropper & Zervas, 1984; Säveland et al., 1986; Tidswell et al., 1995). Further, a recent review of
health outcome one year following SAH (Hackett &
Anderson, 2000) likewise found that many patients continue to
experience significant reductions in quality of life, and in particular
in social role functioning. They also reported that between one-third
to one-half of the patients reported ongoing subjective problems of
memory and mood (Hackett & Anderson,
2000). Examples such as these, and the extensive experience in
the field of SAH research again reinforces the importance of
considering factors other than neurological outcome when evaluating the
management of a particular disease process.
Another important area debated in the existing SAH literature,
particularly in terms of predicting what we might expect to see
following treatment of a UIA, is that of what processes underlie the
observed poor cognitive outcome of many patients. SAH is considered to
be a complex disorder involving the initial hemorrhage and a number of
secondary events (Cesarini et al., 1999). It
has been suggested that the considerably unfavorable outcomes that have
been observed to follow SAH are to a large extent the consequence of
secondary insults sustained by the brain during the acute phase of the
disease (Cesarini et al., 1999).
Some evidence has been reported which suggests that observed
neuropsychological deficits are related to the diffuse brain damage as
disclosed on CT scan 1 year after SAH (Vilkki et
al., 1990). However other investigations have failed to find a
relationship between damage to the brain and cognitive disturbances in
patients with reportedly good neurological outcome (Berry et al., 1997). Other factors associated with
the surgical treatment of the aneurysm have also reportedly been found
not to be associated with evidence of poor cognitive outcome (Berry et al., 1997). Indeed, from a study that
reviewed the cognitive and psychological sequelae of uncomplicated
aneurysm surgery, Maurice-Williams et al. (1991) concluded that uncomplicated surgery itself
does not involve any risk of detectable impairment. Once again,
however, there remains some controversy with other researchers who
conclude that aspects of surgery to treat a SAH do correlate with
longer-term cognitive disability (Tidswell et al.,
1995).
A possible means of resolving some of this controversy may be to
compare surgically clipped SAH cases to matched endovascularly coiled
SAH cases. If differences in neuropsychological outcome are observed in
these cases then it may be possible to argue that these differences are
due to the treatment, assuming of course that the cases are well
matched. If there are no differences between the two groups then it may
in turn be appropriate to argue that any observed deficits are most
likely due to shared factors, such as the diffuse damage from the
original bleed, experience of hospitalization, anesthetic effects, etc.
Some evidence on this controversy has recently been provided by
Hadjivassiliou et al. (2001) who compared
pairs of subarachnoid hemorrhage patients matched on factors other than
treatment modality and found that coiling and clipping produced
statistically similar impairments across a wide range of cognitive
domains. In this study the authors concluded that such evidence
suggested that the impairments in cognitive outcome observed to follow
SAH were largely the result of the original hemorrhage. They do however
also note that there was some evidence that surgical clipping produced
more frontotemporal damage (Hadjivassiliou et al.,
2001). In addition, the recent International Subarachnoid
Aneurysm Trial (ISAT) study (2002) found the
neurological morbidity in clipped SAH cases to be significantly worse than
the neurological morbidity of endovascularly coiled SAH cases, suggesting
that clipping produces some additional damage above and beyond the effects
of the original bleed.
Another study design which can also shed useful light on the
controversy is a design which compares neuropsychological outcome in
well matched SAH and UIA cases. As Hillis et al. (2000) note such a study allows the
neuropsychological effects which result from the SAH to be separated
out from the neuropsychological deficits that result from the
treatment, treatment complications, anesthesia and other factors such
as psychological depression. From their study which utilized such a
design they concluded that some specific deficits appeared to be
related to the original SAH whereas other deficits were more likely
related to the general effects of treatment (Hillis
et al., 2000). Further investigations with study designs that
incorporate clipped and coiled cases and/or SAH and UIA cases are
clearly needed to shed further light on the relative contribution of
initial SAH and subsequent treatment to the profile of cognitive
deficits often observed to follow rupturing of an aneurysm.
In addition to the cognitive deficits experienced by many victims of
SAH a considerable number of recovered patients also report
psychological deficits. With regard to the processes that may underlie
these deficits it has been hypothesized that the poor social and
psychological outcome observed in SAH victims may result from the
subjective experience of suffering from a SAH (McKenna et al., 1989; Ogden et
al., 1990; Vilkki et al., 1990). An
interesting finding from the study conducted by Ogden et al. (1990) was that the group of patients for whom an
aneurysm could not be identified experienced significantly poorer
outcomes on several psychosocial variables. From this Ogden et al.
(1990) hypothesized that because these
patients could not be reassured that an aneurysm had been occluded,
psychological rather than neuropsychological processes may underlie
some aspects of recovery. If part of the outcome picture typically
observed to follow SAH without an identifiable aneurysm can be
attributed to psychological repercussions of having suffered a
life-threatening illness, then we might expect to see some deficits
following discovery of an UIA. Again, however, a thorough investigation
is needed to confirm this.
Neuropsychological and Psychosocial Outcome Following
Treatment of UIAs
Physicians are schooled in the tradition of “first do no
harm.” But what do we consider to be harm? To a large extent
measurement and significance of postoperative outcome depends on the
measure of outcome used (Raaymakers, 2000).
Raaymakers notes that outcome measures, or measures of harm, have
generally been restricted to measures of neurological impairment.
Typically assessments of neurological status are similar across studies
and include the monitoring of such deficits as hemiparesis, aphasia,
nerve palsy, motor deficits and speech disturbances. In most studies
the usual standard of practice is to group neurological outcomes into
four categories: excellent, good, poor, and
death. Excellent is usually defined as a return to
preoperative function level with no new neurological deficits.
Good is assigned to a patient considered to be ambulatory,
cognitively independent but with minor new neurological deficits, and
poor is a patient who is dependent or not ambulatory. These
categories are often labelled differently; for example as good
recovery, mild deficits, severe deficits, and
death. They still however represent similar groupings of
outcomes. Usually the patients who fall into the poor
categories are those represented in morbidity figures. Patients
who fall into the good or excellent categories are in
turn typically regarded as having had a “favorable”
recovery.
Commonly used measures such as the Glasgow Outcome Scale (Jennett & Bond, 1975), the Rankin Scale (Rankin, 1957), or even the Mini-Mental State
Examination (Folstein et al., 1975) do not
assess cognitive functioning in any detail. It is therefore possible
that many people classified as having “favorable” outcome
have not returned to their pre-surgery level of functioning (Orz et al., 2000). In addition, many studies
reporting morbidity figures do not make explicit their method
for classifying deficits or for defining morbidity (King et al., 1994). Finally, use of a single
measure of outcome such as neurological status on discharge from
hospital, or at 6 weeks post surgery, may also be misleading,
especially if the important information is outcome in the long term.
One particularly detailed system of outcome measurement is the scheme
proposed by the World Health Organization for the classification of the
consequence of any disease (World Health
Organization, 1980). As Wade (1996)
notes, although considered too complex for everyday use, the WHO model
does form a good conceptual framework for understanding neurological
deficits. The model notes the need to assess illness on three levels:
impairment, disability and handicap. Just recently a new version of the
World Health Organization model has been released (World Health Organization, 2001). This revised
model defines several domains of health, with these domains grouped
under two classifications: (1) body functions and structures; and (2)
activities and participation. These two new classifications replace the
terms impairment, disability and handicap from the original model, and
are said to extend their scope to include positive experiences. In
addition, the revised model has moved away from an emphasis on the
consequences of disease towards an approach that considers the various
components of health (World Health Organization,
2001). Although this revised model undoubtedly builds upon and
extends the original model, as it is not yet in widespread use the
original model is referred to here.
Impairments are defined as the direct neurophysiological consequences
of the underlying pathology (Wade, 1996) and
are typically measured by neurological outcome. Assessment of
impairment should also include assessment of psychological and
neuropsychological dysfunction, such as memory impairment. Disability
refers to a disturbance in aspects of a patient's behavior or
normal function and is essentially the external, behavioral
consequences of the pathology. Examples include slow walking, needing
help to dress, no longer being able to cook a simple meal, or
forgetting to take one's medication (Wade,
1996). Finally, handicaps are the social and societal
consequences of the pathology and are the most important determinant of
the severity of the illness from the patient's point of view
(Wade, 1996). In essence, handicap reflects
the consequences for the individual that stem from the presence of impairment
and disability (WHO, 1980). A common example is the
patient's inability to return to work full time because of
increased fatigue levels (Ogden et al., 1990,
1994). It is this particular aspect of
outcome that has in the past been neglected by most morbidity studies
of UIAs.
Health-related outcome has many facets, and its measurement should
reflect this. As Raaymakers (2000) notes,
different measures of outcome may be of differing importance for
professionals and patients. A neurosurgeon may be most interested in
neurological impairments whereas a patient may consider ongoing quality
of life to be the most relevant factor. Behavior is often the most
sensitive measure of cerebral function, and hence investigations,
including clinical neuropsychological assessments, which assess
behavior consistently find problems that are not revealed by other
measures of brain function (Kolb & Whishaw,
1996; Lezak, 1995). Neuropsychological
assessment can provide data that is useful for both the neurosurgeon
and the patient.
A particular difficulty with neurological measures is that common
outcome scales may not reveal deficits that can affect people in very
debilitating ways. As an example, Ogden et al. (1993) discuss the example of an architect whose
work involved a high degree of visuospatial ability. Such an individual
would be considerably disadvantaged by very subtle impairments of
visuospatial function. However, such deficits are only likely to be
detected if comprehensive neuropsychological testing is carried out. As
Ogden et al. (1993) note, if the patient is
unaware that they have a deficit, even a mild deficit, they may become
frustrated and irritable at their work performance. Such emotions may
negatively impact on friends, family and colleagues, thereby creating a
vicious cycle that ultimately results in a decline in well-being.
Providing patients and their families with simple information about
what they can expect with regard to their recovery can help to
alleviate many fears and misconceptions and allow patients and families
to make the necessary adjustment to their lives. Therefore from the
patient's perspective, the neuropsychological report is in many
cases likely to be more informative than the neurological report.
Neuropsychological assessment is also useful for defining baseline
levels of cognitive functioning for longitudinal comparisons with
follow-up data. Longitudinal follow-up data is particularly helpful
when attempting to identify the rate of improvement or deterioration in
cognitive status in response to treatment (Mitrushina et al., 1999). In addition to providing
the patient with useful information on their pattern of recovering,
neuropsychological data can be used to make inferences at a group level
concerning the overall “success” of a particular treatment
option.
Neurological status data, such as has been reported in the bulk of
the literature on UIA outcome, can also provide useful information
regarding the overall impact of a particular treatment option. This is
especially pertinent in studies that include large numbers of patients,
because of the time-consuming nature of neuropsychological assessment.
In addition, when evaluating the effect of a treatment involving
neurosurgery, the neurological outcome is clearly important. Therefore,
in the early stages of treatment evaluation standard measures of
neurological outcome may be most appropriate, with long-term
neuropsychological and psychosocial outcome assessed at 3 months, and
again at 6 to 12 months following treatment.
To date, very few studies of UIAs have gone beyond the measures of
neurological status or simple “screening” tests of
cognitive function. There are, however, a few exceptions where
researchers have included more comprehensive neuropsychological outcome
measures. Fukunaga et al. (1999) recently
reviewed the neuropsychological functioning of 30 patients prior to and
following surgery for UIAs. Patients who demonstrated a decline in
performance 1 month after surgery were tested again 2 months later. The
tests used in the study were the Mini Mental State Examination (MMSE),
a maze test, and a speeded letter recognition task
(“Kana-Hiroi” test). One month after surgery scores on the
MMSE decreased in 10 cases, scores on the “Kana-Hiroi” test
decreased in 17 cases and the time taken to complete the maze test was
increased in 30 cases. All three scores worsened in the case of 6
patients. At 3 months post-surgery all these 6 patients had recovered
to pre-operative levels. Patients with anterior communicating artery
aneurysms were significantly worse on measures of neuropsychological
deterioration following surgery than those with UIAs in other sites.
An interesting finding to emerge from the Fukunaga et al. (1999) study is that residual cognitive deficits
remained evident 1 month post surgery but had resolved by the time of
the 3-month follow-up assessment. This information can be useful in
planning rehabilitation strategies and, as the authors note, for
indicating when it is advisable for patients to return to normal life,
be this occupationally or socially. However, the Fukunaga et al. (1999) battery covers only a small number of
possible cognitive impairments, and therefore their findings are
limited in their applicability.
The ISUIA (1988) study also included measures
of cognitive change, using the Telephone Interview for Cognitive Status
(Brandt et al., 1988) and the MMSE (Folstein et al., 1975). One of the points of
interest from this study was the difference in morbidity
figures produced by basing outcome on neurological status alone
when compared to the morbidity figures derived from basing
outcome on cognitive status (2.9% vs. 6.1% at 1 year). Another
interesting finding from this study was that of the 90 patients who had
either impaired cognitive status or impaired neurological status at one
year, only 29 of these had impairments on both measures. Thus the ISUIA
(1988) study demonstrates that the assessment of
cognitive status detects impairments additional to neurological difficulties
and supports the inclusion of both types of outcome measures.
The ISUIA (1988) study has been criticized on
several grounds; however only those that relate to the measures of cognitive
outcome will be discussed here. First, the study's estimates of cognitive
changes were based on a short telephone interview of cognitive status
conducted 1 year after surgery. Second, the study also failed to take a
measure of pre-operative cognitive functioning, making it difficult to
ascertain whether cognitive impairment was due to the intervention or
instead reflected pre-morbid variability within the study population
(Hillis et al., 2000). In addition, cognitive
status was not measured in the patients who did not undergo surgery;
therefore there was no comparison group (Alexander
& Spetzler, 1999). Despite these difficulties, the ISUIA
study takes a positive step with its inclusion of cognitive testing,
particularly in such a large sample of patients.
More recently an excellent study of functional outcome and quality of
life in 18 patients who underwent elective operation for asymptomatic
UIAs was published by Raaymakers (2000).
Measures of patient-reported quality of life are considered to be an
important component of patient outcome as they give insight into
patients' subjective feelings of physical, psychological and
social well-being not otherwise detected by measures of functional or
cognitive abilities (Hop et al., 1998).
Raaymakers (2000) identified the 18 patients
who made up the subject pool for the study by screening relatives of
patients who had suffered a spontaneous SAH. All patients were assessed
in terms of impairments, disabilities, handicaps and quality of life.
Disability in activities of daily living were assessed by means of the
Barthel Index (Mahoney & Barthel, 1965),
handicap by a modified Rankin Scale (Rankin,
1957), and quality of life by the Sickness Impact Profile (SIP)
(Bergner et al., 1981) and the Medical
Outcomes Study Short Form 36 (Ware & Sherbourne,
1992).
Raaymakers (2000) reported that the
combination of angiography and the operation resulted in persistent
neurological sequelae in 5 of the 18 patients (28%) if anosmia was
included as a neurological deficit, and 3 of the 18 patients (17%) if
it was not, with these sequelae present at three months and one year
after the operation. In terms of other measures of outcome, 3 months
after surgery 2 patients had disabilities, as measured by the Barthel
Index and 15 had less than optimal Rankin handicap scores. At 1 year
after operation all disabilities had resolved. However, 8 patients
still had sub-optimal Rankin handicap scores. In terms of the quality
of life findings, Raaymakers (2000) reports
that according to the SIP at 3 months, 16 patients had a lower quality
of life than before their operation, and according to the SF-36, 11
patients showed a decrease in quality of life. At 1 year, according to
the SIP 10 patients still reported a decrease in quality of life
whereas on the SF-36 8 patients reported a decrease in quality of life.
Raaymakers (2000) concluded that elective
surgery for unruptured aneurysm has a short-term negative impact on
quality of life in most patients. However, 1 year after operation,
outcome had considerably improved, although recovery was not complete.
Raaymakers (2000) noted that these results
could only be generalized to patients with incidentally found, mostly
small aneurysms, who have one relative with SAH. Whether these same
outcomes also apply to other categories of patients treated for
unruptured aneurysms is not known.
Hillis et al. (2000) have also recently
published a study of cognitive impairments following surgery for UIAs.
In addition to being valuable for its analysis of cognitive functioning
following treatment for UIAs, the study is also notable because it
conducts pre-surgery and post-surgery testing. Such testing allows
patients' performances following treatment to be evaluated against
their own pre-operative performance level.
This is a more reliable method than comparing post-treatment
performance to published test norms to assess outcome. In the Hillis et
al. (2000) study, a subset of 12 patients
were administered a comprehensive battery of neuropsychological tests
prior to surgery and approximately 3 months after surgery. As a group,
poorer post-operative performance was reported on tests of verbal
fluency, immediate and delayed verbal recall, and “frontal
lobe” or executive abilities. The authors comment that as
impairments in these areas are also commonly found to follow repair of
ruptured aneurysms, a proportion of the long-term cognitive sequelae of
SAH may result from the effects of the neurosurgery per se,
rather than from the hemorrhage. This study does suffer from small
patient numbers, and a 3-month follow-up time is quite short for
assessment of neuropsychological function. It is possible that many of
the impairments may have resolved given a longer follow-up period and
that deficits may have been underestimated due to possible test-retest
effects given such a short follow-up period.
In addition to the possibility of cognitive changes or quality of
life deterioration following UIA treatment, it is also necessary to
consider the costs of not having the aneurysm treated, and carrying the
ongoing risk of the aneurysm rupturing at any point in time. In their
study on the costs and benefits of treating UIAs Johnston et al. (1999b) estimated that the cost of living with a low
risk, but high consequence, condition would produce a mildly impaired
emotional state (Johnston et al., 1999b). As
the risks associated with the aneurysm increased, or the UIA produced
painful or compressive symptoms, the emotional costs of living with the
untreated aneurysm increased. The ongoing emotional disturbance
associated with living with an untreated UIA is an important variable
to take into consideration when evaluating treatment options, but it is
at this stage largely unmeasured.
Neuropsychological and Psychosocial Assessment: Future
Recommendations
Although the few neuropsychological and psychosocial outcome studies
published make an excellent contribution and their preliminary findings
are an important step forward, further comprehensive studies are
needed. Some recommendations of how these studies might be conducted
follow.
Cognitive and psychosocial domains
Individual tests and tasks included in a neuropsychological
assessment battery need to be chosen primarily on the basis of their
proven ability to detect impairments that may result from the diffuse
or focal brain damage that can follow aneurysm treatment. To this end
the existing SAH literature can provide a useful starting point with
regard to which areas of functioning may deserve special attention.
Studies by some researchers (Hadjivassiliou et al.,
2001; Ljunggren et al., 1985; Ogden et
al., 1990, 1993;
Richardson, 1989) have reported evidence of
persisting memory deficits, both of a verbal and non-verbal nature. In
addition, subjective memory problems and reports by relatives of
problems in real life memory functioning have also been commonly noted
(Ogden et al., 1993, 1997). Deficits in the areas of attention and
concentration and visuospatial functioning have also been reported
(Ljunggren et al., 1985; Ogden et al., 1990, 1993) as have
problems with verbal fluency (Mavaddat et al.,
1999; Richardson, 1991) and reaction
time (Hütter et al., 1995). From these
findings it can be recommended that any investigation of outcome
following UIA treatment would need to include tests that tap memory,
including verbal, nonverbal and real-life memory functioning, attention
and concentration, visuospatial functioning, verbal fluency and
reaction time.
In addition to the above mentioned cognitive deficits, SAH outcome
studies have noted deficits in psychosocial functioning in the areas of
irritability, fatigue, noise sensitivity, and reduced quality of life
(Hop et al., 1998; Hütter et al., 1995; Ljunggren et al., 1985; Ogden et al., 1990, 1994; Vilkki et al., 1990). Interestingly, these
persisting deficits have not been found to be related to premorbid
factors such as IQ, occupation, depression, stress or
smoking/drinking habits, nor do these deficits appear to be related
to demographic factors such as age, gender or marital status (Ogden et al., 1994). Instead, it was hypothesized
that persisting deficits following SAH are linked to diffuse brain
damage from the SAH and to the psychological consequences of suffering
from a life-threatening disease (Ogden et al.,
1994).
Unfortunately whilst psychosocial functioning is an area that is
often of crucial importance to the well being of individuals, it is one
not easily tapped by standardized assessment measures. It is
acknowledged that there are many methodological difficulties in the
assessment of psychosocial function, with it having been suggested that
this may account for the relatively limited number of trials that
attempt to assess aspects of quality of life or psychosocial
functioning (Sanders et al., 1998).
Psychosocial functioning is usually assessed through a variety of
questionnaires, self-report measures and semi-structured interviews. In
some cases psychosocial variables can be measured in terms of
traditional quantitative data, such as responses on an anxiety
inventory converted to a numerical score that is then used to rate the
magnitude of an individual's anxiety. In other cases the data
provided is qualitative, such as when participants are simply asked
about their experiences and their responses are recorded verbatim. As
Hop et al. (1998) note, it is important to
include validated quality of life instruments in order to facilitate
comparison to other studies. In addition, less standardized measures
such as semi-structured interviews that include questions that simply
ask patients about their recovery can also provide useful information
(Dorman et al., 2000; Lindley et al., 1994).
When considering the use of quantitative versus qualitative
data it is important to note that most psychosocial assessments are not
undertaken for diagnostic purposes but rather to describe the
individuals' neuropsychological and psychosocial status (Lezak, 1995). When quantitative and qualitative
methods of assessing variables are combined they provide useful data
with regard to the complexity, variability and subtleties of behavior
(Lezak, 1995). If researchers are to be
encouraged to include measures of psychosocial function in future
treatment outcome studies then it is important that standards are
developed further with regard to practices for assessing and measuring
quality of life and psychosocial functioning (Sanders et al., 1998).
Retrospective versus prospective study design
A particularly important decision regarding the design of any outcome
study is the question of whether to adopt a retrospective or
prospective study design. Retrospective designs have the advantage of
allowing for a larger number of participants to be included in a study
than is usually possible with prospective studies (Aseltine et al., 1995). Rather than waiting for new
cases to occur and then attempting to recruit these cases into the
study, retrospective studies can draw from a large pool of patients who
have already been treated, potentially going back as many years as
necessary to collect the required participant numbers. Following on
from this, the data collection period for a retrospective design is
often much shorter, particularly for rare medical conditions.
Unfortunately retrospective study designs also have several
disadvantages, including difficulties in estimating pre-treatment
functioning and the significant problems of selection bias. Failure to
accurately measure pre-treatment functioning is particularly
problematic for data interpretation. Given that many patients with UIAs
have already suffered a SAH this is especially relevant in UIA
treatment outcome studies. In the absence of pre-treatment data it can
be difficult to conclude whether changes in cognitive performance are
the result of: (1) the treatment; (2) premorbid variability amongst the
study population (Hillis et al., 2000); or
(3) the affect of cerebral injuries sustained prior to the treatment
(Maurice-Williams et al., 1991).
Retrospective studies therefore are often considered unsatisfactory for
monitoring change across a period of time (McKenna
et al., 1989).
Another concern with regard to retrospective designs is the issue of
cognitive dissonance. It has been suggested that patients who have
undergone a treatment, and particularly a stressful treatment, will be
motivated to exaggerate the benefits of this treatment when looking
back on the experience (Aseltine et al.,
1995). In their study Aseltine et al. (1995) found that in evaluations of overall health,
retrospective assessments of change were more positive than indicated
by prospective studies. Although there may be alternative explanations
for this effect, the authors concluded that, at best, retrospective
reports were only weakly related to change (Aseltine
et al., 1995).
With regard to selection bias, a major problem is posed by the
substantial heterogeneity in the patient population eligible for
enrolment in a study (Fleming, 1982). An
example of this bias is reported by Ogden et al. (1993) in a SAH outcome study in which they report
that in some cases, patients with high levels of education or good
recovery may be more likely to be motivated to participate in
time-consuming studies, hence potentially biasing results in the
positive direction. The opposite problem is also potentially true. For
example, patients still under medical care or who have not been able to
return to work may also be more likely to participate in studies due to
their increased time availability, hence again biasing the results but
this time in the negative direction (Ogden et al.,
1993).
Despite these disadvantages, more often than not studies of UIA
treatment have used retrospective study designs. From a meta-analysis
review of UIA treatment morbidity and mortality studies Raaymakers et
al. (1998) reports that of the 61 studies
included in their review only eight were prospective, with the other 53
being either retrospective or unspecified. Raaymakers et al. (1998) suggests that well-designed prospective
studies are needed in order to gather UIA outcome data not limited by
the problems of retrospective designs discussed above.
To randomize or not to randomize?
With the introduction of coiling technology in the early 1990s
patients now have two main options for the “treatment” of
their UIAs. To help physicians and patients choose between these two
options outcome information for each of these treatments is needed.
Several types of experiments can be designed to provide such
information. Cook et al. (1992) suggest that
studies can provide five levels of clinical decision-making
information. At the top of the list are randomized control trials
(RCTs).
Although the use of a RCT seems ideal for evaluating the two UIA
treatment options, given the evolving nature of UIA treatment
technology (Johnston et al., 2001; Sturaitis et al., 2000) such studies are considered
both impractical (Kassell, 2001) as well as
potentially unethical. For patients to be randomly allocated to
treatment options it is ethically important that the two options be of
equal potential value and equal potential risk. At the current stage in
the development of coiling technology the two treatment options are not
equal for all patient groups (Broderick,
2000). For example, UIAs of the posterior circulation are said
to be a high-risk category for clipping, but a relatively low risk for
coiling (Broderick, 2000). In contrast, UIAs
of the middle cerebral artery are typically considered a low risk group
for clipping but a high-risk group for coiling (Broderick, 2000).
One possible alternative to conducting a fully randomized control
trial is to conduct a RCT study in a restricted population where the
two treatment options produce comparable outcomes in terms of treatment
success and patient morbidity. Broderick (2000) proposes that we are quickly approaching the
stage where we will have sufficient data to justify the use of such a
restricted randomized trial. The type of randomized clinical trial
Broderick (2000) suggests is not fully
randomized but instead involves allocating participants to the
treatment best suited to their aneurysm and then randomly assigning
participants to either the best treatment option for their aneurysm or
to a “best medical therapy” treatment group. In the design
proposed by Broderick (2000) a truly random
three arm design comparing coiling, clipping and “best medical
therapy” would only be conducted in those participants with
aneurysms considered amenable to either clipping or coiling.
Where only small numbers of participants are available to be included
in a study, as is the case for rare medical conditions such as
identified unruptured aneurysms, restricting study inclusion to a
select group of UIA patients would reduce participant numbers to an
untenable level. In addition, although a study such as proposed by
Broderick (2000) would provide useful data
for those aneurysms of the size and location included in the study, it
is questionable whether results could be generalized to other aneurysm
groups.
Possibly because of the generally accepted unequal status of the two
treatment methods for some aneurysm groups there are currently no fully
randomized controlled trials comparing endovascular coiling to surgical
clipping in the UIA literature. Further, it is questionable whether UIA
treatment will ever be a suitable candidate for completely randomized
trials. For a condition to be suitable for such a trial it should be
common, have clear endpoints and occur within a relatively short period
of time (Caplan, 1998). In addition, it is
often necessary in RCTs to combine various subgroups of patients, and
to exclude other groups such as patients who are old, frail, pregnant,
or have coexisting conditions (Caplan, 1998).
Thus it is questionable whether information from such trials are
generalizable to individual cases (Caplan,
1998). Given that it is perhaps unlikely that RCTs will be
carried out in the future (Bederson et al.,
2000) it is necessary to consider lower levels of evidence from
the classification system of Cook et al. (1992).
Evaluating neuropsychological and psychosocial change
An important question to address when assessing the outcome of any
treatment is how best to measure change. This is a surprisingly
difficult question, but an important one that needs to be carefully
considered when drawing conclusions about treatment outcomes, as
clearly indicated in an excellent series of articles and commentaries
in a recent issue of Neuropsychology (Chelune, 2002; Keith &
Puente, 2002; Keith et al., 2002;
Millis, 2002; Sawrie,
2002; Smith, 2002).
In order to assess the extent to which a patient has changed as a
result of a treatment, both pre-treatment and post-treatment
assessments are needed. However, it is not enough to simply evaluate
the absolute amount of difference between these two assessments (Sawrie et al., 1996). To determine whether an
individual has been significantly affected by a treatment it is
necessary to establish whether the observed changes are reliable and
beyond what would be expected given normal change over time and
practice (Chelune et al., 1993). A
particularly useful approach to analyzing this change involves
application of one of the methods from a group of techniques commonly
referred to as reliability of change indices (RCI). RCI can be used to
determine whether an individuals' score on a particular task or
tasks is different from what might be expected given the effects of
normal change and repeat testing (Sawrie,
2002). It is generally the goal of RCI to differentiate change
due to statistical artifact from change that is statistically rare, due
to the treatment, and both reliable and clinically significant (Chelune, 2002).
RCI are a particularly useful way of identifying clinically
significant change, which at the same time also address the problems of
the practice effect and other sources of measurement error. They offer
additional advantages in that they have been developed to meet both lay
persons and professional expectations regarding outcome measurement and
they allow for the classification of patients into
“changed” or “unchanged” groups (Jacobsen & Truax, 1991) while avoiding the need
for defining arbitrary criteria for indicating significant change has
occurred (Sawrie et al., 1996). Because most
RCI measures include an adjustment for practice effects the use of
alternative forms can be avoided.
However, although RCI have been put forward as a useful method for
defining “true” clinically significant change some authors
argue they do not do so reliably. Keith et al. (2002) argue that it is difficult to find a
quantitative solution to defining which changes are clinically
meaningful and which are not. As they note, small changes in
performance on one cognitive domain may produce greater functional and
quality of life changes for an individual than larger changes in
another domain, with these differences not reflected in traditional
quantitative solutions. Keith et al. (2002)
also disagree that the ability to group participants into
“changed” and “unchanged” categories is an
advantage. They argue that the presentation of data in terms of
categories results in a loss of information and, further, that
traditional parametric statistics should be preferred as they allow the
expression of performances on cognitive tests as continuous variables,
with both small and large performances contributing to overall effect
size.
Although their point is a good one, so too is Chelune's (2002) argument regarding the importance of
presenting outcome data in a form that can be easily utilized by the
end user, usually the clinician but also often the patient. When the
question the neurosurgeon wants answered is: “Are patients
treated for UIAs impaired cognitively?” then grouping
participants into impaired and not-impaired categories as a result of
treatment surely best addresses this question. Further, as Chelune
(2002) notes, patients are more likely to be
interested in knowing what their chances are of having a memory
impairment following treatment, than in knowing, on average, how many
points they will drop on a measure of immediate verbal memory.
Finally it should be noted that, although one of the main advantages
put forward for RCI is their potential to account for practice effects,
this also raises one of the major problems with the method. As Keith et
al. (2002) note, RCI methods assume that the
practice-related performance improvements in a treatment group are
equivalent to those of the control group. This indeed appears to be a
difficult assumption to meet. However, one potential way of reducing
the impact of varying rates of practice effect is to select a
well-matched control group, with the aim that the two groups
(experimental and control) are drawn from the same population and with
the only difference of interest between the two groups being the
treatment (Chelune, 2002). As Sawrie (2002) notes, “RCI methodology is only as good
as the control group on which the actual RCI are derived” (p.
430).
Unfortunately the question of the best control group to include is
not a straightforward issue to address. A seemingly excellent choice of
control group for a UIA outcome study would be a group of untreated UIA
participants. It would be easy to argue that a group of untreated UIA
participants was drawn from the same population as a group of treated
UIA participants. As such it would be possible to have confidence in
concluding that any observed difference between the treated UIA group
and the untreated UIA control group was due to the effects of the
condition of interest, the treatment. However, as Keith and Puente
(2002) note, while suggestions of including a
disease-matched control group is a good one, in practice it can be
difficult to achieve. Keith and Puente (2002)
did attempt to include such a control group in their study but
encountered many difficulties in recruiting this group. In addition,
they also note that there were greater baseline differences on
cognitive testing between a group of disease-matched participants and
their treated group than there was between their healthy control group
and the treated group.
Another potential difficulty with using a control group of untreated
UIA participants is that such a group would not have experienced a
recent surgery and hospitalization and therefore cannot control for any
of the associated effects of such hospitalization. Many researchers
therefore suggest that the most appropriate control group is not a
disease-matched group, but is instead a group of other surgery or
treatment patients (Millis, 2002; Smith, 2002). While it is usually assumed that
other treatment groups are a better control group as they have also
experienced the impact of acute hospitalization this has not been
clearly tested. A third option for a control group is the use of
healthy controls. Healthy controls have been criticized on several
grounds, but as noted by Slade et al. (2001)
when they are drawn from friends or relatives of the experimental group
they provide an optimal method of controlling for practice effects.
Perhaps an ideal solution to the problem of which control group to
include in the study would be to include more than one control group.
As suggested by Slade et al. (2001) an ideal
combination would be a group of healthy controls drawn from friends of
the treated participants and a group of other treated and hospitalized
patients. In reality however, identifying and recruiting this second
group may be more difficult.
Another particularly important question in the assessment of change
is when to conduct follow-up assessments. In selecting a time frame for
the follow-up assessment of neuropsychological and psychosocial outcome
it is important to consider what questions you are seeking to answer.
If the research question concerns whether there are any short-term
predictors of long-term outcome, then acute assessments, preferably
when patients are still in hospital, need to be conducted.
Alternatively, if the research question asks whether treatment results
in any long-term cognitive or psychosocial deficits then the time frame
of final assessment needs to be long enough to allow for any deficits
that may resolve to have done so. Evidence from the SAH literature
suggests whilst some deficits do persist, most patients have recovered
fully, or at least almost fully, within 1 year (Ogden et al., 1993, 1994, 1997). As such, an end-point assessment time frame
of 12 months should be sufficient, although where feasible longer time
frames of follow-up would be valuable. Finally, if the aim of the
research is to map a recovery course in order to plan for
rehabilitation strategies, then assessment at regular intervals may be
needed; for example, acute, 3 months, 6 months, and 12 months.
