Content uploaded by Nyika D Kruyt
Author content
All content in this area was uploaded by Nyika D Kruyt
Content may be subject to copyright.
HIGH MEAN FASTING GLUCOSE LEVELS
INDEPENDENTLY PREDICT POOR
OUTCOME AND DELAYED CEREBRAL
ISCHEMIA AFTER ANEURYSMAL
SUBARACHNOID HEMORRHAGE
Journal of Neurology, Neurosurgery & Psychiatry, 2008, PMID: 18403438
Kruyt ND
Roos YWBEM
Dorhout Mees SM
van den Bergh WM
Algra A
Rinkel GJE
Biessels GJ
CHAPTER 7
78
Hyperglycemia has been related to poor outcome and delayed cerebral ischemia (DCI) aer
aneurysmal subarachnoid hemorrhage (SAH).
We aimed to assess whether in paents with aneurysmal SAH, levels of mean fasng
glucose within the rst week predict poor outcome and DCI beer than single admission
glucose levels alone.
Data on non-diabec paents admied within 48 hours aer aneurysmal SAH with at least
two fasng glucose assessments in the rst week were retrieved from a prospecve database
(N=265). The associaon of admission glucose or mean fasng glucose, dichotomized at
the median levels, with outcome was assessed with logisc regression and with DCI with
Cox regression. To explore whether the associaon between glucose levels and outcome
was mediated by DCI, we adjusted for DCI.
The crude and mulvariable adjusted odds raos and 95% condence intervals for poor
outcome were 1.9 (1.1 to 3.2) and 1.6 (0.9 to 2.7) for high admission glucose and 3.5 (2.0
to 6.1) and 2.5 (1.4 to 4.6) for high mean fasng glucose. The crude and adjusted hazard
raos for DCI were 1.7 (1.1 to 2.5) and 1.4 (0.9 to 2.1) for high admission glucose and 2.0
(1.3 to 3.0) and 1.7 (1.1 to 2.7) for high mean fasng glucose. Aer adjusng for DCI, the
odds raos on poor outcome for high mean fasng glucose decreased only marginally.
Compared with high admission glucose, high mean fasng glucose, represenng impaired
glucose metabolism, is a beer and independent predictor of poor outcome and DCI. DCI is
not the key determinant in the associaon between high fasng glucose and poor outcome.
FASTING GLUCOSE LEVELS AND OUTCOME IN ANEURYSMAL SAH
79
Hyperglycemia occurs oen in paents with aneurysmal subarachnoid hemorrhage (SAH),
and has in these paents been related to poor clinical condion on admission and to poor
clinical outcome.
203-207;217;218;220;222;228
This relaon may in part be explained by an increased
risk of delayed cerebral ischemia (DCI) in paents with hyperglycemia on admission.
203;205;218
However, the relaon between hyperglycemia and poor outcome or DCI is not consistent
across studies. Whereas some studies found that the associaon remained aer correcng
for other explanatory variables,
203;206;217
others did not.
204;205;228
An explanaon for these
inconsistent ndings may be that admission glucose levels can uctuate rapidly due to stress
reacons and variable nutrional intake. Furthermore, glucose is likely to be assessed more
frequently in case of clinical deterioraon, or in paents with established hyperglycemia,
which introduces the risk of sampling bias. Instead, repeated fasng glucose measurements
are less inuenced by these factors, especially if sampled at standard intervals during
hospital stay, and provide a more accurate reecon of glucose metabolism and actual
glucose control than single admission or randomly assessed, non-fasng glucose levels.
The purpose of this study was to assess whether in paents with aneurysmal SAH, levels of
mean fasng glucose are a beer predictor of poor outcome and DCI than single admission
glucose levels alone, and whether these associaons are independent from other baseline
characteriscs. Furthermore, we aimed to assess whether the potenal relaon between
glucose levels and outcome was mediated by the occurrence of DCI.
Data on paents were retrieved from a prospecvely established database of all paents
with aneurysmal SAH admied to the stroke unit of the University Medical Centre in Utrecht.
This database includes data on medical history, clinical condion on admission, radiology and
laboratory tests, clinical course and complicaons, and funconal outcome at 3 months. From
the period 1997 to 2003 we retrieved all non-diabec paents admied within 48 hours aer
aneurysmal SAH, with at least two fasng glucose assessments before DCI onset, and with a
complete data set on admission. Paents with pre-existent diabetes mellitus (DM) were not
included because we intended to relate acute, not previously known abnormalies of glucose
metabolism, to outcome and also because glucose-lowering treatment prior to and aer the
SAH might confound the results. The diagnosis of SAH was made by the presence of extravasated
blood in the basal cisterns on a CT scan or, if the CT- scan was negave by xanthochromic
cerebrospinal uid. Paents with a perimesencephalic or other non-aneurysmal causes for SAH
were excluded. The clinical condion on admission was assessed with the World Federaon of
Neurological Surgeons (WFNS) scale, a ve-point scale based on the Glasgow Coma Scale and
the presence or absence of focal decits.
241
A dichotomy was made between good neurological
condion (WFNS I to III) and poor neurological condion (WFNS IV or V) on admission. We
assessed the amount of cisternal and ventricular blood on the inial CT scan according to the
method described by Hijdra and collegues.
242
The sumscores of blood in the cisterns (range 0 to
30) and ventricles (0 to 12) were dichotomized at their median value (25 for cisternal blood; 1 for
CHAPTER 7
80
ventricular blood). We quaned the size of the frontal horns by means of the bicaudate index
(BCI). To calculate age adjusted relave sizes, the BCIs were divided by the corresponding upper
limit per age group. Hydrocephalus was dened as an age adjusted relave BCI greater than 1.
243
During hospitalizaon, all paents were under connuous observaon for at least 2 weeks
and treated according to a standardized protocol that consisted of absolute bed rest unl
aneurysm treatment, oral doses of nimodipine, cessaon of anhypertensive medicaon, and
intravenous administraon of uid with the aim of normovolemia.
According to our protocol admission glucose was assessed on the day of admission, and fasng
glucose levels were assessed on the day aer admission. During the clinical course fasng
glucose levels were assessed every 2 days up to 21 days aer the event or unl discharge if this
occurred within the rst 3 weeks. Admission glucose was dened as the rst available glucose
assessment. Fasng glucose was calculated by the mean of all fasng glucose assessments unl
day 7 for each paent. This me interval was chosen because this was the median day of DCI
onset. Guidelines for the treatment of hyperglycemia aer SAH were based on expert opinion.
The protocol recommended treatment of hyperglycemia with subcutaneous insulin every 4
to 6 h on a sliding-scale dosing schedule once glucose levels > 12 mmol/L on two consecuve
occasions. However, the actual decision to start treatment was le to the discreon of the
treang physician.
Clinical outcome was assessed with the modied Rankin Scale (mRS),
212
a six point scale dening
the degree of the paent’s funconal disability. Poor outcome aer 3 months was dened
as death or moderately severe disability (mRS >3). The mRS was assessed through telephone
interviews by trained research nurses with the paents or the paent’s representaves. These
assessments were blinded for laboratory data. DCI was dened as clinical features of DCI
(gradually developing focal decit, or a decrease in level of consciousness or both) with a new
hypodense lesion on CT, or, in case no new hypodensity was found, exclusion of other causes
of the deterioraon by means of CT and appropriate laboratory examinaons.
244
The primary
endpoints of the present study were poor clinical outcome at 3 months and the occurrence
of DCI.
Admission glucose and mean fasng glucose levels were dichotomized at their median into
high and normal values. The primary aim of the analysis was to esmate the relave risk of
poor outcome and DCI in relaon to glucose levels. For poor outcome, odds raos (OR) and the
corresponding 95% condence intervals (CI) were calculated by means of logisc regression.
Potenal confounders in the relaon between glucose and outcome were idened in a
stepwise approach. First, baseline characteriscs that were prospecvely assembled in our
database (age, gender, and hydrocephalus) and well established predictors of poor outcome
aer SAH (i.e. clinical condion on admission and sumscores of extravasated blood)
198;245;246
were entered in bivariable analysis together with glucose values. If the bivariable OR diered
more than 5% from the univariable OR for the relaon between glucose values and outcome,
the entered variable was considered a potenal confounder in the relaon between glucose
FASTING GLUCOSE LEVELS AND OUTCOME IN ANEURYSMAL SAH
81
levels and outcome and subsequently entered into a mulvariable model together with glucose
levels. In order to assess the inuence of DCI outcomes intermediate in the pathway to poor
outcome, addional adjustment for this variable was done. A similar analycal approach was
used for the associaon between glucose values and DCI, with proporonal hazards regression
analysis, resulng in hazard raos (HR) and corresponding 95%CI. A secondary analysis was
performed that only included fasng glucose values before the onset of DCI (i.e. excluding
fasng glucose values assessed aer the occurrence of DCI). Addionally, to evaluate the
glucose proles in paents with DCI, we calculated the mean fasng glucose one day before
and one day aer the occurrence of DCI.
A total of 299 consecuve paents fullled the inclusion criteria of our study. We could not
include 34 paents because the inial CT scan could no longer be retrieved, thus 265 paents
were analyzed (table 1). The main reason for non retrieval of CT scans was that they had been
sent back to the referral hospital and could not be retraced. These paents did not dier in
baseline characteriscs or outcome compared to the included paents. Median admission
glucose was 6.9 mmol/L (inter quarle range (IQR): 6.1 to 7.8) and median fasng glucose
before day 7 was 7.0 mmol/L (IQR: 6.2 to 7.5). Eighty-two paents (34%) had a poor outcome.
DCI occurred in 86 paents (33%) with a median onset at day 7 (IQR 4.5 to 9.5) aer the ictus.
Compared with paents with admission glucose values below or equal to the median, paents
with glucose values above the median were older, and more oen had a poor clinical condion
on admission. Paents with mean fasng glucose (but not admission glucose) values above the
median also more oen had higher sumscores of extravasated blood than paents with mean
fasng glucose levels below the median.
Table 2 lists the associaon of high levels of admission glucose or mean fasng glucose with
poor outcome. The crude OR for poor outcome associated with high levels of admission glucose
Male (%) 46 (32) 44 (36) 47 (35) 43 (33)
Age (years ± sd) 53 ± 14 59 ± 12 51 ± 14 60 ± 12
WFNS scale IV-V (%) 17 (12) 32 (26) 16 (12) 33 (25)
Sumscore ventricles >1 (%) 59 (42) 58 (47) 50 (37) 67 (51)
Sumscore cisterns >25 (%) 63 (44) 64 (52) 52 (39) 75 (57)
Hydrocephalus (%) 55 (39) 43 (35) 46 (34) 52 (40)
Sd: standard deviaon; WFNS: World Federaon of Neurological Sociees.
CHAPTER 7
82
was 1.9 (1.1 to 3.2). In bivariate analyses, this crude OR changed more than 5% for age, clinical
condion on admission, and hydrocephalus. Aer mulvariable adjustment for these baseline
characteriscs the OR for poor outcome associated with high levels of admission glucose was
1.6 (0.9 to 2.7).
The crude OR for poor outcome associated with high levels of mean fasng glucose was 3.5
(2.0 to 6.1). This crude OR changed more than 5% for age, clinical condion on admission, and
extravasated blood sumscores in bivariate analyses.
Univariable model 1.9 (1.1 – 3.2) 3.5 (2.0 – 6.1)
§
Male
Age > 56 yrs
WFNS scale IV-V
Sumscore ventricles >25
Sumscore cisterns >1
Hydrocephalus
DCI
1.9
1.7
*
1.7
*
1.8
1.8
2.0
*
1.6
(1.1 – 3.2)
(1.0 – 3.0)
(1.0 – 2.8)
(1.1 – 3.2)
(1.1 – 3.1)
(1.2 – 3.4)
(0.9 – 2.8)
3.5
3.3
*
3.2
*
3.2
*
3.2
*
3.5
3.0
(2.0 – 6.1)
(1.9 – 5.8)
(1.8 – 5.6)
(1.8 – 5.7)
(1.8 – 5.5)
(2.0 – 6.1)
(1.7 – 5.4)
1.6† (0.9 – 2.7) 2.5‡ (1.4 – 4.6)
*: 5% change in crude OR; †OR adjusted for: age, WFNS and hydrocephalus; ‡ OR adjusted for: age, WFNS,
ventricle and cisternal blood sumscores; §: the bivariable models include glucose level in combinaon
with each of the listed variables; DCI, delayed cerebral ischemia; WFNS, World Federaon of Neurological
Sociees.
Univariable models 1.7 (1.1 – 2.5) 2.0 (1.3 – 3.1)
§
Male
Age > 56 yrs
WFNS IV-V scale
Sumscore ventricles>25
Sumscore cisterns >1
Hydrocephalus
1.7
1.5
1.5
*
1.6
1.6
*
1.7
(1.1 – 2.6)
(1.0 – 2.4)
(1.0 – 2.3)
(1.1 – 2.5)
(1.1 – 2.4)
(1.1 – 2.6)
2.0
1.9
1.8
*
1.9
*
1.9
*
1.9
(1.3 – 3.1)
(1.2 – 3.0)
(1.2 – 2.9)
(1.2 – 2.9)
(1.2 – 2.8)
(1.2 – 3.0)
1.4† (0.9 – 2.1) 1.7‡ (1.1 – 2.7)
CI: condence interval; WFNS: World Federaon of Neurological Sociees; DCI: Delayed Cerebral Ischemia.
*: >5% change of crude Hazard rao; §: the bivariable models include glucose level in combinaon with
each of the listed variables; † Hazard Rao adjusted for: age; WFNS; and cisternal blood sumscore; ‡
Hazard Rao: adjusted for: WFNS; ventricle- and cisternal blood sumscores.
FASTING GLUCOSE LEVELS AND OUTCOME IN ANEURYSMAL SAH
83
Aer mulvariable adjustment for these baseline characteriscs the OR for poor outcome
associated with high levels of mean fasng glucose was 2.5 (1.4 to 4.6). Also shown in table 2
are the bivariable analyses with the ORs for poor outcome in relaon to glucose levels adjusted
for DCI only. The adjusted ORs changed marginally relave to the crude ORs. Table 3 lists the
associaon of high levels of admission glucose or mean fasng glucose with DCI. The crude HR
associated with high levels of admission glucose was 1.7 (1.1 to 2.5). Aer bivariable adjustment,
this HR changed more than 5% for clinical condion on admission and extravasated cisternal
blood sumscore. Aer mulvariable adjustment for these baseline characteriscs the adjusted
HR for DCI associated with high levels of admission glucose was 1.4 (0.9 to 2.1). The crude
HR for DCI associated with high levels of mean fasng glucose was 2.0 (1.3 to 3.1). This HR
changed more than 5% for clinical condion on admission and extravasated blood sumscores
aer bivariable adjustment. Aer mulvariable adjustment for these baseline characteriscs
the adjusted HR for DCI associated with high levels of mean fasng glucose was 1.7 (1.1 to
2.7). The secondary analysis that only included glucose values before the onset of DCI provided
similar results (crude HR for DCI associated with high levels of mean fasng glucose: 1.9 (1.2
to 3.0). Levels of fasng glucose did not dier substanally before or aer DCI onset. One day
before DCI onset mean fasng glucose was 7.4 ± 1.8 mmol/L (± standard deviaon) on the
day of DCI this was 7.3 ± 1.9 mmol/L and one day aer DCI this was 7.4 ± 2.3. This indicates
that DCI occurrence as such had no noceable eect on fasng glucose levels. As indicated by
the proporonal hazards regression analysis, the mean fasng glucose was signicantly higher
in the rst week in paents with DCI compared to paents without DCI (7.4 vs. 6.9 mmol/L,
p<0.001). The analyses on outcome and the occurrence of DCI were repeated with (fasng)
glucose values as connues variables, which yielded similar results as the analyses presented
above.
We have demonstrated that high mean fasng glucose levels during the rst week aer
aneurysmal SAH are strongly and independently associated with poor clinical outcome and the
occcurence of DCI. In fact, the associaon with poor outcome was at least as strong as for other
predictors of poor outcome, such as clinical condion on admission and the amount of blood on
the inial CT scan.
198;245;246
We could not comrm the hypothesis that the associaon between
high mean fasng glucose levels and poor clinical outcome was mainly mediated by DCI. As can
be seen from the bivariable analysis, the associaon between high mean fasng glucose levels
and poor clinical outcome was marginally aected by other paent characteriscs.
The mechanism for the relaon between poor glycemic control and poor outcome and DCI is
unclear. Hyperlgycemia on admission and high levels of fasng glucose could be due to the
stress reacon and therefore only reect the magnitude of the inial insult. Another possibility
is that high levels of mean fasng glucose during the rst weeks aer SAH reect an acute
metabolic response including insulin resistance persisng during the acute ilness. Such a state
has been described in other groups of crically ill paents,
231;232;247
and has been associated with
cardiovascular complicaons and poor clinical outcome, but mostly on the long term and in
paents who already had abnormalies of glucose metabolism.
248-250
Although previous studies
did not report a relaon between DM and aneurysmal SAH occurrence,
236;237;251
it is possible
CHAPTER 7
84
that high fasng glucose levels aer aneurysmal SAH represent pre-existent, but previously
unrecognized abnormalies in glucose metabolism. Indeed, several factors that are associated
with chronic (subclinical) abnormalies in glucose metabolism such a history of hypertension, a
high body mass index, and DM, predispose to poor clinical outcome aer aneurysmal SAH.
197;222
Unfortunately, we do not have sucient data to analyse wether the included paents had pre-
existent signs of a metabolic syndrome.
The associaon between high levels of admission glucose and outcome was weaker than that of
mean fasng glucose and did not persist aer mulvariable assessment. Most probably, single
admission glucose levels are also aected by the severity of the ictus due to a subsequent stress
reacon. This observaon is in line with previous reports that showed a similar weak associaon
between admission glucose levels and outcome that was not sustained aer mulvariable
adjustment.
204;205;228
Our study has some limitaons that should be adressed. Although we used
a prospecvely collected database, which was built specically to study (fasng) glucose in
paents with aneurysmal SAH, the analyses were performed retrospecvely, and paents with
incomplete data were excluded. Because the main reason for missing data was non-retrieval of
CT scans and these paents did not dier in baseline characteriscs or outcome compared to
the included paents, selecon bias was probably minimal.
In retrospect, we would have liked to collect more detailed informaon on nutrion to more
accurately assess glucose metabolism. Paents not on tube feeding, however, always had their
breakfast aer assessments of blood glucose levels and paents fed via a nasogastric tube
(around 15% of paents) received feeding in porons; assessments of blood glucose levels were
always performed in the morning, around 6 to 8 hours aer the previous poron. Furthermore,
we do not have exact data on the use of insulin to control for hyperglycemia. Current literature,
however, indicates that glycemic control and not insulin dosage per se correlates with improved
outcome.
181;218;252
Furthermore, insulin use was shown not to be associated with symptomac
vasospasm aer aneurysmal SAH.
218
Finally, no causal associaons can be infered from this
observaonal study. Despite these limitaons, this study shows that aer aneurysmal SAH,
high levels of mean fasng glucose, represenng impaired glucose control, is a beer and
independent predictor of poor outcome and DCI than high levels of admission glucose, which
probably only represent the extent of the cerebral insult. Furthermore, we have shown that DCI
is not a major determinant in the associaon between high fasng glucose and poor outcome.
Although the presented data do not prove any causal relaon between persistent high mean
fasng glucose levels and poor outcome, they may oer leads for therapy. Tight glycemic
control, if performed over a longer period (days), has previously demonstrated to improve
clinical outcome for paents with various medical emergencies,
28;29;177;253
although these
observaons are not unequivoval.
35,36
It remains unclear if this is also true for paents with
aneurysmal SAH. Thus far, only one randomized pilot trial tested the applicabity of intensive
insulin therapy in the intensive care unit seng for this paent group with unclear results.
37
Ecacy of such treatment in paents with aneurysmal SAH eventually needs to be addressed
in a suciently large randomized clinical trial.