Ada Anaeslhesiol Scand 1994:
38:
64665
Prinled
zyxwvutsrqpon
n
Denmark
zyxwvutsrqponml
all
zyxwvutsrqponm
ighrr
reserved
opyriahl
zyxw
da
Anacsthesiol
Scand
z
994
z
Acta Anaesthesiologica Scandinavica
ISSN
00015172
A
zyxw
ultifactorial analysis
of
the spread
of
epidural analgesia
M.
CURATOLO ,
.
ORLANDO*,
. M.
ZBINDEN ,
P.
SCARAMOZZIN03
and F.
s.
VENUTI*
'Institute of Anaesthesiology and Intensive Care, Inselspital, University of Bern, Switzerland, the 'Institute
of
Anaesthesiology and Intensive Care, University of Messina, Italy and the 3Department of Economics, SOAS, University of London, London, United Kingdom The controversies about the factors determining the spread of epidural analgesia are partly due to inappropri ate methodology
or
sample size of previous studies. We performed a multivariate regression analysis on
8 3
ASA class
12
nonatherosclerotic adults, undergoing lumbar epidural anaesthesia according to a predefined standardised procedure. The spread of epidural analgesia is more accurately studied by analysing dose/ segment
(R2=
0.671) instead of spread
(R2=0.27
)
as dependent variable. The impact of local anaesthetic
(2
lidocaine
CO,
or
0.5
bupivacaine) and addition of adrenaline is not significant. Spread significantly increases with increasing age, weight, bodymass index, dose of local anaesthetic, addition of fentanyl, higher site of injection, and decreasing body height. The impact of age and dose is higher under the age of
40
and at doses lower than
20
ml. Increasing the total dose increases the dose needed to block one spinal segment. Unknown idiosyncratic factors still determine a certain proportion of the sample variance. The addition of adrenaline to lidocaine and the use of bupivacaine improve the predictability of spread. In conclusion, we found clinically significant correlations between a group of factors and epidural spread. Alternative anaesthetic solutions lead to different degrees of predictability.
Received I4
October
1993, accepted
zyxwvutsr
or
publication
1
March 1994 Key
words:
Anesthesia: epidural; spread; anesthetics, local: bupivacaine; lidocaine; fentanyl; statistics: multiple regression; vasoconstrictors: epinephrine.
According to the findings of Bromage
(1)
the dose of local anaesthetic needed to block
a
spinal segment is correlated
to
various physiological factors such as age, height, presence of atherosclerosis and pregnancy. However, other studies have found either no relation ship or a very weak correlation between the level of sensory blockade and the abovementioned factors, including the dose
(24).
Moreover, it has been shown that the dose/segment also depends
on
the volume of local anaesthetic injected: the higher the volume, the higher the dose requirement
(3,
4).
In most studies the factors influencing the spread of epidural anaesthesia have been analysed separately. This methodology is likely to lead to significant statisti cal bias, as it does not consider simultaneously all the potential explanatory variables
(5).
In fact, the estimated impact
of
each variable could be seriously distorted if the others are not included in the same analysis. Furthermore, this approach does not provide a general model to predict epidural spread. The aim of our study was to perform
a
multivariate analysis of the spread of epidural analgesia, in order both to determine the real influence of the single factors and to improve the predictability of epidural spread. The
cta Anaesthesiologica Scandinavica
38
1994)
analysis is performed on both spread (number of spinal segments blocked) and dose/segment (dose needed to block a spinal segment), in order to determine which functional form more accurately analyses the spread of epidural analgesia. MATERIAL AND
METHODS
The study was prospectively conducted on
807
ASA class
12
pa tients,
14
to
81
years old, who underwent lumbar epidural anaesthesia for elective orthopaedic surgery of the lower limbs. Pregnant, diabetic and patients suspected to be atherosclerotic on the basis
of
their history were excluded. Two experienced anaesthesiologists performed the epidurals, at the same hospital, from
1988
to
1991.
The epidural blocks were performed according to the standardised routine pro cedure of the hospital, and consent was obtained from each patient. Every epidural block was performed at the lumbar level in the lateral position, with the median approach, using a 17G Tuohy needle. The site of puncture was chosen according to the kind of surgery
6):
he second lumbar interspace for hip and femur surgery, the third for knee and tibia surgery, the fourth for ankle and foot surgery. The lumbar interspace was identified by drawing a line between the highest point of the iliac crests, which corresponds to the third lumbar interspace. The epidural space was identified by the
loss
of resistance technique, injecting no more than
2
ml of
0.9
saline. A test dose of
3
ml of the chosen anaesthetic solution was first administered, followed after
zyx
min by the remaining dose. The
ANALYSIS OF SPREAD OF EPIDURAL ANALGESIA
647
anaesthetic solution was injected through the epidural needle, at the speed
of
I
ml/second, with the bevel of the needle pointing cranially. The local anaesthetic was chosen according to the predicted dur ation and the site of surgery. Lidocaine
CO,
2 was used for oper ations lasting
less
than 30 min. Adrenaline 1:200,000 was freshly added to lidocaine
CO,
2 for operations lasting 30 to 90 min and for every foot
or
ankle operation
(7).
Patients undergoing surgery lasting longer than 90 min received 0.5 plain bupivacaine. The local anaesthetic was administered according to age, height and body mass index
zyxwvutsrqp
1,
2), applying the following formulae: amount in ml=
40

.5
x
age (years) for patients under 30 years old, ml= 33 .3
x
age for patients between 30 and 60 years old and ml= 30

.25
zyxwvu
age for patients over 60. The calculated dose was decreased by
0

0 in patients shorter than 165 cm
or
with a bodymass index (weight [kg]/height* [ml]) higher than 27, and increased by 030 in pa tients taller than 180 cm. The maximum and minimum allowed doses were 30 ml and 10 ml respectively. This formula was also used to calculate the dose of the youngest patients, as they were considered biologically adult on the basis of height and body weight. Epidural fentanyl was used to potentiate analgesia in surgery believed to be associated with a higher algogenic input (8), namely: total hip and knee arthroplasty, reconstruction of anterior and posterior cruciate ligaments, tibia1 osteotomy, ankle ligament reconstruction and cor rection of hallux valgus. Fentanyl, when used, was added to the last 5 ml
of
the anaesthetic solution injected, at the dose
of
100 pg for patients under 50 years old and
50
pg for patients over 50. Immedi ately after the administration of the anaesthetic solution the patient was turned to the supine position. The level
of
sensory blockade was determined by absence of pain to pinprick with a
2
1
G sharpbevel needle. Determinations were made at 5 and 10 min and then every 10 min, up to
60
min after the injection of the anaesthetic solution.
In
case of asymmetry of the block, the mean of the number of segments on both sides was con sidered. Patients displaying an asymmetry of more than two dermat omes were not included in the analysis.
zyxwvutsrq
Statistical methodology
The data were analysed by multivariate regression analysis, which simultaneously studies the influence
of
various factors (independent or explanatory variables)
on
a dependent numerical variable. The dependent variable investigated in our study was either the spread of analgesia (number of dermatomes) or the doselspinal segment (ml). A backward stepwise procedure, considering a
zyxwvu
value<0.05 as significant, was used. Initially, all the potential explanatory vari ables were included in the analysis. Then the parameters not signifi cantly correlated with the dependent variable were, step by step, removed from the model. This procedure led to a final model which included only parameters significantly correlated with the dependent variable. The following independent variables were initially analysed: age (years), height (cm), weight (kg), bodymass index (weight [kg]/ height' [m]), dose of local anaesthetic (ml), local anaesthetic used (carbonated lidocaine 2 or bupivacaine 0.5 ), site of injection, addition of adrenaline, addition of fentanyl. The dose of local anaes thetic was included in the regression performed
on
dose/spinal seg ment in order both to study the relationship between the total dose of local anaesthetic injected and the doselsegment, and to perform an unbiased analysis
of
the impact of the other factors.
In
fact, since the dose was administered according to age, height and weight, the real impact of these variables
on
doselsegment would probably be distorted if the dose of local anaesthetic were not included in the regression. The possibility of nonlinearities in the impact
of
the numerical variables was explored by partitioning the ranges of age, height, weight and dose into intervals and by defining dummy vari ables corresponding to those intervals. For example, the range of dose was partitioned into intervals of 5 ml (1014, 1519, 2024 and 2530). The set
of
dummies was included in the multiple regression.
RESULTS Four patients displayed an asymmetry of the block of more than two dermatomes and were excluded from the analysis. Sacral analgesia reached
zy
5
in every pa tient. Summary statistics for numerical and categorial variables analysed in our study are presented in Tables
1
and 2, respectively. The initial multiple regression, including all the par ameters investigated, showed that the impact of the type of local anaesthetic and the addition
of
adrenaline on both the spread and the doselsegment were not statistically significant. The final model, which in cludes significant variables, is shown in Table 3. The impact of the body mass index was also highly signifi cant (P<O.OOl) when included instead of height and weight separately, but the overall
fit
of the regression slightly deteriorated. Splitting the sample according to age groups (2029,3039, etc.) showed a break at around the age of
40,
indicating that the impact of age is noticeably higher below the age of 40 (Fig.
1
)
.
The correlation of the dose with the spread is also nonlinear, as the impact for doses
of
1019 ml is higher than for doses of 2030 ml (Table 3). On the other hand height and weight showed a linear correlation with both spread and doselsegment. The doselspinal segment significantly increased with increasing volume of local anaesthetic injected (Fig.
A
much higher proportion of the sample variance
is
explained when doselsegment instead of spread is analysed as dependent variable (R2= .671 and 0.271, respectively). Moreover, the ratio of the root mean square error (MSE) to the standard deviation
of
the sample is much lower in the former than in the latter model
(0.58
and
0.88,
respectively), indicating a higher degree of predictability (see discussion for ex planation). Table
4
shows that the overall
fit
of the multiple regression improves if adrenaline is added to lidocaine,
2).
Table
1
Numerical variables included in the analysis. Spread and dose/seg ment are the dependent variables. Mean s.d. Median Range Age (years) 38.3 15.4 36 1481 Height (cm) 170.6 8.0 171 144193 Weight (kg) 72.0 11.8 72 41105 Bodymass index (kg/m2) 24.7 3.6 24.4 14.537.5 Dose (ml) 21.0 4.1 20 1030 Spread
(no.
segments) 14.4 2.8 14 822 Doselsegment (ml) 1.53 0.44 1.49 0.633.33
648
M.
CURATOLO ET AL. Table
2
Categorial variables included in the analysis. Local anaesthetic Site
of
injection Lidocaine Bupivacaine
co,
2 0.5 L2 L3 L4 Adrenaline Fentanyl
No.
65 152 148 574 81 196 190
zy
81
19 18.4 71.5 10.1 24.4 23.7 Table 3 Final multiple regression including only the explanatory variables significantly correlated to the spread or the doselsegment. Two alternative values of constant intercept, according to the age, are given. Dependent variable
~ ~~
Spread DoselSpinal segment Coefficient
zyxwvutsr
.e.
zyxwvu
Coefficient
zyxw
e
P
Age 140
y.
Age
>
40
y.
Height (cm) Weight (kg) Dose <20 ml Dose 220 ml Site
of
injection L2L3 (reference) L3L4 L4L5 Fentanyl 0.136 0.057 0.034 0.250 0.207 0.071

.909

.045 0.668 0.018 0.017 0.014
0.009
0.046 0.036 0.231 0.348 0.207
10 001 0 001
<
0.001 <0.001
<0 001
<0 001
<0.001
<O.OOI
0 001
0.017

.003 0.007 .004 0.045 0.049 0.097 0.240

.069 0.002 0.002 0.001 0.001 0.005 0.004 0.024 0.036
0.021
<0 001
0.054
<0.001
<
0.001
<0.001
<
0.001
KO.001
<0.001
<0 001
Intercept Age <40 15.520 2.273
<0.001
0.036 0.236 0.878 Age
>
40 18.299 2.394
<0.001
.450 0.249 0.07
I
Rsquare
=
0.27
1
s.d.
zyxwvuts
f
sample= 2.8 Root MSE/s.d. =0.88 Rsquare
=
0.67
zy
s.d.
of sample=0.44 Root MSE/s.d.
=
0.58 Root MSE
=
2.456 Root MSE= 0.255 s.e.: standard error. MSE: mean square error. s.d.: standard deviation of the dependent variable.
and is even better in the bupivacaine group (higher
R2
and lower root MSE). The analysis of variance did not show significant differences in the age, height, weight, dose administered, spread and doselsegment among the three groups. DISCUSSION
Overall statistical jit
We analysed the influence of various factors on the spread of epidural analgesia by making use of two different specifications, namely: (spread),
=
a,+aIxli+a2x2,+
..+
kxk;+&;
(1)
=)
=
o
+
PIXli
+
p2x2,
+
.
+
pnxni
+
Ti
(2)
spread where
spread);
and
dose/spread),
are the number of segments blocked and the dose
of
local anaesthetic required to block a segment in the patient
“i”,
respec tively; xil to x;k and xil to
xi,
the explanatory variables;
zyxw
a
to
ak
and
Po
to
zyx
k
constant parameters;
Ei
and
Ti
random disturbances. Equation
2
can be rewritten as: dosei
(3)
pread)i
=
o
+
PIxli
+
Zi
**.
+
Pnxni
+
i
which characterises a hyperbolic function. This means that the additional effects on
spread
of
increases in an explanatory variable
zyx
tends to become smaller as
x
gets larger.
A
comparison between the two functional forms shows a much better statistical fit of the regression performed on dose/segment (Table
3).
In fact, the parameters investigated can explain 67.1 of the vari ability of dose/segment and only
27.1
of the vari ability of spread. The root mean square error
(MSE)
is an index of predictability of the model: the lower its value, the higher the predictability. Conversely, the closer the root MSE to the standard deviation of the dependent variable, the lower the predictive value. The ratio root MSE to standard deviation is 0.88 and
ANALYSIS
OF
SPREAD
OF
EPIDURAL ANALGESIA
zyxw
49
. .
s
zyx
,,t
.*
.
....
*
*
. . ..
..
.
,
.
..
.
*:
**
:*
..
*.
zy
. .
*.
.'
..
zy
I
10
2
zyxw
4
6
6
7
8
Age [years1
Fig.
1
Scatter
of
predicted values
of
the spread
of
epidural analgesia to age, according
to
Table
3.
The impact
of
age decreases above the age
of
40
0.58
in the regressions performed on spread and dose/ segment, respectively, indicating a higher predictive value of the latter model.
A
possible interpretation of this finding is that the specification with doselsegment, characterised by
a
hyperbolic function, is able to cap ture the decreasing effects on spread of increases in the explanatory variables. Thus, this functional form seems to be a more accurate approach to analyse the spread of epidural analgesia.
zyxwvuts
atients characteristics
The correlation between age and both spread and spinal dose requirement is highly significant and nonlinear (Fig.
1).
A
10year increase in age extends the spread by about 1.4 segments in patients younger than 40 and by 0.6 segments in patients over 40. Previous studies found either no correlation (2) or
a
much lower quanti tative impact of age on spread (3,4,9), compared to our results. Inadequate sample size (24) and omission of factors which were likely to interact with age (3), par ticularly the dose (9), might have biased the results of these studies. The higher spread with increasing age is the result of different physiological factors:
a
progressive narrowing of the intervertebral foramina, which re duces the loss of anaesthetic solution from the epidural space and enhances its longitudinal spread; an alter ation and decrease in the number of myelinated fibres;
a
deterioration in the nerve sheaths
(6).
The reduced impact of age on spread in older patients could be due, at least in part, to an increased systemic absorption of local anaesthetic (lo), which partially offsets the above mentioned physiological changes. The impact of height is highly significant, although quantitatively modest: a 10 cm increase in height de creases spread by
0.7
segments. Previous studies found a significant correlation with height, but with only less than half of the quantitative impact, compared to our results (4,
9).
One would expect height
to
play a more relevant role, as the local anaesthetic has to spread longitudinally in order to block a certain number of segments. One should consider, however, that the epi dural space is not a closed space, as leakage occurs through the intervertebral and sacral foramina. Body weight is positively related to spread and nega tively to dose/segment, in
a
linear fashion.
No
corre lation was found in earlier studies (2,
4 ,
probably because of the smaller size of the samples, whereas
a
significant influence
of
body mass index had been demonstrated
(2,
11). The quantitative impact
of
weight, however, is of limited clinical relevance, as a
3
kg' increase in weight increases spread by only one segment.
650
M. CURATOLO ET AL.
zyxw
zyxwvuts
1
o
E
8
1.5
zyxw
zyxwvu
r
tD
zyxwvutsrq
8
1.1..
0.71
.
e
.
z
zyxwvutsrqpon
.
i
1
. .
.
.
I
14 18 22
26
3
Total Dose
[ml]
Fig.
2.
Scatter of predicted values of doselsegment to dose of local anaesthetic administered, according to Table
3.
Increasing the volume
of
local anaesthetic injected increases the dose required to anaesthetise a spinal segment. Table
4
Overall tit
of
multiple regressions performed on both spread and dose/segment in three subgroup, defined according to the local anaesthetic administered and the addition of adrenaline. The parameters of Table
3
were
considered. Depend. Lidocaine CO, Lidocaine
CO,
Bupivacaine Variable Plain Adrenaline Plain Rsquare Spread
0.204 0.356 0.393
ml/segm.
0.632 0.675 0.800
RootMSE Spread
2.533
ml/segm.
0.262 2.431 2.242 0.268 0.214
s.d. of sample Spread
2.808 2.947 2.781
ml/segm.
0.427 0.458 0.462
ml/segm.
0.61 0.59 0.46
RootMSE/s.d. Spread
0.90 0.82 0.8
1
No.
patients
46
1
190 152
MSE: mean square error. s.d.: standard deviation of the dependent variable.
dural pressure, which increases the leakage of anaes thetic solution through the intervertebral and sacral foramina and reduces its longitudinal diffusion. Con sistent with previous findings
(3,
4 ,
our study shows that increasing the volume
of
local anaesthetic in creases the dose required to block a spinal segment (Fig. 2), probably because of the abovementioned enhanced leakage
of
anaesthetic solution.
naesthetic technique
The relationship between dose and spread is highly significant and nonlinear, as the impact decreases with increasing dose: a 5 ml increase in dose increases spread by 1.25 and
1
segments when administering less or more than 20 ml, respectively. Higher volumes of local anaesthetic are likely to determine higher epi