Simons SHP, van Dijk M, van Lingen RA, Roofthooft D, Duivenvoorden HJ, Jongeneel N, Bunkers C, Smink E, Anand KJS, van den Anker JN, Tibboel D. Routine Morphine Infusion in Preterm Newborns Who Received Ventilatory SupportA Randomized Controlled Trial. JAMA. 2003;290(18):2419-2427. doi:10.1001/jama.290.18.2419
Author Affiliations: Departments of Pediatric Surgery (Drs van Dijk and Tibboel and Mr Simons) and Pediatrics (Dr van den Anker), and Division of Neonatology (Dr Roofthooft, Mr Simons, and Ms Jongeneel), Erasmus MC-Sophia, and Department of Medical Psychology, Erasmus-MC (Dr Duivenvoorden), Rotterdam, the Netherlands; Department of Pediatrics, Division of Neonatology, Isala Clinics, Zwolle, the Netherlands (Dr van Lingen and Mss Bunkers and Smink); Department of Pediatrics, University of Arkansas for Medical Sciences and Arkansas Children's Hospital, Little Rock (Dr Anand); and Division of Pediatric Clinical Pharmacology, Departments of Pediatrics and Pharmacology, George Washington University, Washington, DC (Dr van den Anker).
Context Newborns admitted to neonatal intensive care units (NICUs) undergo a
variety of painful procedures and stressful events. Because the effect of
continuous morphine infusion in preterm neonates has not been investigated
systematically, there is confusion regarding whether morphine should be used
routinely in this setting.
Objective To evaluate the effects of continuous intravenous morphine infusion
on pain responses, incidence of intraventricular hemorrhage (IVH), and poor
neurologic outcome (severe IVH, periventricular leukomalacia, or death).
Design, Setting, and Patients A randomized, double-blind, placebo-controlled trial conducted between
December 2000 and October 2002 in 2 level III NICUs in the Netherlands of
150 newborns who had received ventilatory support (inclusion criteria: postnatal
age younger than 3 days and ventilation for less than 8 hours; exclusion criteria:
severe asphyxia, severe IVH, major congenital malformations, and administration
of neuromuscular blockers).
Interventions Intravenous morphine (100 µg/kg and 10 µg/kg per hour) or
placebo infusion was given for 7 days (or less because of clinical necessity
in several cases).
Main Outcome Measures The analgesic effect of morphine, as assessed using validated scales;
the effect of morphine on the incidence of IVH; and poor neurologic outcome.
Results The analgesic effect did not differ between the morphine and placebo
groups, judging from the following median (interquartile range) pain scores:
Premature Infant Pain Profile, 10.1 (8.2-11.6) vs 10.0 (8.2-12.0) (P = .94); Neonatal Infant Pain Scale, 4.8 (3.7-6.0) vs 4.8 (3.2-6.0)
(P = .58); and visual analog scale, 2.8 (2.0-3.9)
vs 2.6 (1.8-4.3) (P = .14), respectively. Routine
morphine infusion decreased the incidence of IVH (23% vs 40%, P = .04) but did not influence poor neurologic outcome (10% vs 16%, P = .66). In addition, analyses were adjusted for the use
of additional open-label morphine (27% of morphine group vs 40% of placebo
group, P = .10).
Conclusions Lack of a measurable analgesic effect and absence of a beneficial effect
on poor neurologic outcome do not support the routine use of morphine infusions
as a standard of care in preterm newborns who have received ventilatory support.
Follow-up is needed to evaluate the long-term effects of morphine infusions
on the neurobehavioral outcomes of prematurity.
Morphine has been one of the most frequently used drugs to relieve pain
in many age groups. Nevertheless, debate continues about whether morphine
and analgesic therapy should serve as standard of care for preterm newborns
who have received ventilatory support,1 despite
the recognition that all preterm neonates feel pain.
Lack of a gold standard to assess neonatal pain, fear of adverse effects,
and uncertainty about the long-term effects of opioids in the neurodevelopmental
outcome of newborns contribute to this clinical conundrum. Although numerous
neonatal pain instruments are available, they have been based and validated
on models of acute pain.2 It is difficult,
therefore, to measure the analgesic effect of morphine in neonates. Suggested
adverse effects of morphine are hypotension,3- 6 seizures,7 bradycardia, decreased gastrointestinal motility,8 intestinal obstruction, urinary retention, and respiratory
depression.9,10 Although a few
long-term effects of neonatal morphine exposure have been suggested from animal
studies,11- 13 the
effects seem to be minimal at 5 to 6 years in a cohort of former preterm infants.14 On the other hand, morphine administration may decrease
morbidity, such as intraventricular hemorrhage (IVH) and periventricular leukomalacia
(PVL).15 We hypothesized that continuous morphine
infusions may improve outcomes and diminish pain responses of nonsurgical
neonates who have received ventilatory support in stressful conditions. Furthermore,
repeated pain exposure may cause hypersensitivity and lower pain threshold
in preterm neonates,16- 20 and
morphine administration might protect preterm neonates from the harmful effects
of pain on their short- and long-term outcomes.21,22
Consensus statements on the analgesic treatment of neonatal pain23,24 have suggested the use of continuous
opioid infusions for preterm neonates who have received ventilatory support.
Studies25- 28 investigating
intravenous opioids in neonates who have received ventilatory support do not
provide conclusive evidence. Therefore, double-blind randomized controlled
trials have been suggested as a means to resolve the uncertainty over whether
and when to administer analgesics to critically ill neonates.1,29
Based on the protocol of a multicenter trial (NEOPAIN study [Neurologic
Outcomes and Pre-emptive Analgesia In Neonates]), we performed a randomized,
double-blind, placebo-controlled trial to evaluate the effect of continuous
intravenous morphine infusion on pain responses, the incidence of IVH, and
poor neurologic outcomes (severe IVH, PVL, or death) in preterm neonates who
had received ventilatory support. We tested the hypothesis that continuous
morphine administration in neonates who had received ventilatory support would
reduce both the degree of pain experienced and the incidence of poor neurologic
outcome and IVH (all grades).
All neonates admitted to the neonatal intensive care unit (NICU) who
required mechanical ventilation were eligible for inclusion. Other inclusion
criteria were postnatal age younger than 3 days, artificial ventilation for
less than 8 hours, and indwelling (peripheral or umbilical) arterial catheter.
Excluded were neonates with severe asphyxia (Apgar score after 5 minutes of
<4 or cord blood pH <7.0),30 severe IVH
(grade III or IVH plus apparent periventricular hemorrhagic infarction), major
congenital malformations and facial malformations (eg, cleft lip and palate),
neurologic disorders, or receiving continuous or intermittent neuromuscular
Patients were recruited from 2 level III NICUs in the Netherlands: Erasmus
MC-Sophia, Rotterdam (center 1), a university hospital, and the Isala Clinics
in Zwolle, a nonuniversity hospital (center 2). Seventy-four percent of neonates
admitted to the NICUs were born in the study hospital. The local ethics committees
of the participating centers approved the study protocol.
The parents of eligible patients were asked to give written informed
consent within 8 hours after endotracheal intubation. If possible, parents
were informed about the study before the birth of their child. If consent
was refused, information about morphine use of the patient involved was collected
retrospectively and compared with information on the participants. Data from
nonenrolled patients were not incorporated into other outcome analyses or
pooled with that from any other patients. Enrolled patients were randomly
allocated to receive a loading dose (100 µg/kg) followed by a continuous
infusion (10 µg/kg per hour) of either morphine hydrochloride or placebo
(sodium chloride), both dissolved in 5% glucose. To prevent possible overdosing,
the study medication loading dose was not given if a preintubation morphine
loading dose had been given less than 3 hours before the start of the study.
The use of masked study medication was continued for 7 days or less, as required
by the patient's clinical condition. After 7 days, study medication was weaned
and stopped or replaced by open-label morphine infusion.
If patients from either group were judged to be in pain or distress
during masked study medication use, they were given additional morphine based
on decisions of the attending physician (independent of the study). Allowed
additional doses were 50 µg/kg followed by 5 to 10 µg/kg per hour
of continuous open-label morphine.
Primary outcomes were defined as the analgesic effects of morphine,
assessed by validated pain measurement instruments at baseline, before study
medication, 30 minutes after the loading dose, and twice daily at a standardized
time point before, during, and after endotracheal suctioning. At each time
point, we videotaped the infants for 2 minutes with 2 cameras: one obtaining
a whole-body image and the other focused on the patient's face. Simultaneously,
the caregiving nurse applied the visual analog scale (VAS) for pain at bedside.
The VAS score ranges from 0 to 10 on a horizontal, continuous line with "no
pain" on the left and "extreme pain" on the right; observers indicated the
level of pain by marking the line. All nurses had been trained to assess neonatal
pain. The videotapes were analyzed afterward using the Neonatal Infant Pain
Scale (NIPS)31 and the VAS during all moments
and the Premature Infant Pain Profile (PIPP)32 during
suctioning. Videotapes were assessed by 2 researchers (N.J. and S.H.P.S) with
acceptable interrater reliability (intraclass correlation coefficient of 0.70
and 0.73 for the NIPS and PIPP, respectively, and 0.67 for the VAS score).
Secondary outcome measures were poor neurologic outcome defined as severe
IVH, PVL, or death within 28 days and the incidence of all grades of IVH.
Other clinical outcome measures were also compared between the morphine and
placebo groups, including duration of artificial ventilation, length of NICU
stay, incidence of comorbidity, and number of painful procedures. Regarding
duration of artificial ventilation, we distinguished between the first ventilation
period (including further periods of ventilation if the infant was extubated
in between for <24 hours) and the second ventilation period (all further
periods of artificial ventilation, after extubation for >24 hours). During
the first 14 days of a patient's NICU admission, we recorded all painful procedures.
A power analysis showed that 75 patients per group were needed to achieve
a medium effect size (Cohen d = 0.55), with an α
error of .05 (2-tailed) and a power of 90%. Neonates had an equal probability
of being assigned to either condition. The randomization code was developed
using a computer random-number generator to select random permuted blocks.
These blocks of 10 were stratified into 5 groups of gestational age ranges
(<27, 27-30, 31-33, 34-36, and ≥37 weeks) to obtain a balanced number
of infants within each stratum.
Using the computer-generated randomization list, independent pharmacists
placed ampules of either 1 mL of morphine hydrochloride or 1 mL of placebo
into boxes. These boxes were numbered with the study numbers and stored with
increasing numbers for the different gestational age groups in a locked closet
accessible only to the researchers. At a patient's enrollment, the next box
in line for the specific group was taken out by one of the researchers. All
research and clinical staff, as well as the parents of the infants, were blinded
Data were analyzed using SPSS statistical software version 10.1 (SPSS
Inc, Chicago, Ill). Nonparametric tests were used and results are shown as
medians and interquartile ranges (IQRs) when variables deviated from the normal
distribution. Background characteristics between the 2 treatment groups were
compared using nonparametric Mann-Whitney U tests
or Fisher exact tests (in case of low incidences). Characteristics of the
nonparticipating patients were compared with data from study infants using
Pain Scores. Multiple regression analyses were
performed with VAS-bedside and NIPS (scored 30 minutes after study medication
loading dose) as outcome variables predicted by treatment group, having received
a morphine dose before intubation, gestational age, Clinical Risk Index for
Babies (CRIB) score, center, sex, and postnatal age in hours corrected by
the pain scored before the bolus was given. Pain scores were log 10 transformed
to approximate a normal distribution.
Across all assessments, mean PIPP, NIPS, and VAS scores, scored during
endotracheal suctioning, were calculated for each patient and used as outcome
variables in multiple regression analyses. Summary statistics (mean scores
for each patient) were used to increase reliability and to take repeated measures
into account during analyses. Predictors were treatment group, mean amount
of additional morphine, center, sex, and duration in study. The importance
of the predictors is shown by unstandardized coefficients.
Clinical Outcome. Logistic regression analyses
were used with poor neurologic outcome (death within 28 days, IVH grade III
or IVH plus apparent periventricular hemorrhagic infarction, and/or PVL) and
IVH (all grades) as outcome variables; treatment condition and additional
morphine use as predictor variables; and center, gestational age, sex, CRIB
score, deviation from mean birth weight for gestational age,33 prenatal
corticosteroid use, preeclampsia and/or HELLP (hemolysis, elevated liver enzymes,
low platelets) syndrome, and the use of indomethacin as covariates.
Collinearity for the logistic regression analyses was checked by performing
a multiple regression analysis instead of the logistic regression analyses
to calculate the variance inflation factors, which were all well below 2.0.
The same was true for the multiple regression analyses. The risk of overfitting
was controlled by using a ratio of 1:10 at least for the number of explanatory
variables and sample size. To assess overfitting more precisely, the patients
in these 2 groups were split into deciles. To cross-validate, the training
sample was composed of 9 of the 10 deciles; the validation sample contained
the remaining decile. The predicted values for the remaining decile were obtained
by the parameters of the logistic regression analysis performed on 9 of the
other deciles. This procedure was repeated 10 times because each decile functioned
as a validation sample. The overall mean obtained from the 10 mean values
and the pooled SD derived from the 10 SDs of the validation samples for each
condition separately were compared with the overall mean and SD of the predicted
values of the total sample. A high level of agreement between the overall
solution and the cross-validation samples indicates high stability. Stepwise
procedures were used.
Comorbidity (eg, chronic lung disease, necrotizing enterocolitis, duration
of artificial ventilation) was compared using the Mann-Whitney U test and Fisher exact test. Missing values were excluded listwise
during all analyses in the sense that all cases that had any values missing
on any of the variables used in the analyses were excluded. In all analyses,
the intention-to-treat principle was used and involved all included infants
who were randomly assigned to the morphine and placebo groups.
A total of 210 infants were eligible between December 2000 and October
2002; the parents of 60 newborns refused informed consent and 150 were randomized
(Figure 1). The percentage of nonenrolled
patients was 36% in center 1 (n = 51) and 13% in center 2 (n = 9). Seventy-three
patients were allocated to receive continuous morphine infusion (44 in center
1 and 29 in center 2), and 77 patients were assigned to receive placebo (48
in center 1 and 29 in center 2). Median duration of study medication infusion
was 48 hours (IQR, 19-96 hours). Use of the medication was stopped for the
following reasons: extubation (n = 106), 7 days in study (n = 24), hypotension
(n = 6), continuous use of neuromuscular blockers (n = 5), death (n = 4),
surgery (n = 2), the need for too much additional morphine (n = 2), and overdosing
(n = 1).
Patient characteristics for both treatment groups are shown in Table 1. All patient characteristics were
comparable between the groups. Demographic characteristics of the nonparticipants
also showed no significant difference compared with the participating infants.
Painful procedures were counted for a median duration of 6 days (IQR, 3-10
days). The number of daily painful procedures was similar in the morphine
group (median, 13; IQR, 10-16) and placebo group (median, 13; IQR, 9-16) (Mann-Whitney U test, 2479; P = .66).
At baseline, median NIPS scores in the morphine and placebo groups were
0.0 (IQR, 0.0-0.0) and 0.0 (IQR, 0.0-0.8) and median VAS scores were 0.6 (IQR,
0.3-2.2) and 0.7 (IQR, 0.3-1.5), respectively. Thirty minutes after study
medication administration, median NIPS scores in the morphine and placebo
groups were 0.0 (IQR, 0.0-0.0) and 0.0 (IQR, 0.0-1.0), and median VAS scores
were 0.6 (IQR, 0.3-1.6) and 0.6 (IQR, 0.2-1.4), respectively.
During suctioning, median PIPP scores in the morphine and placebo groups
were 10.1 (IQR, 8.2-11.6) and 10.0 (IQR, 8.2-12.0) (P =
.94), median NIPS scores were 4.8 (IQR, 3.7-6.0) and 4.8 (IQR, 3.2-6.0) (P = .58), and median VAS scores were 2.8 (IQR, 2.0-3.9)
and 2.6 (IQR, 1.8-4.3) (P = .14), respectively (Table 2). There were no significant differences
between groups for pain scores. Of the 2530 VAS scores, only 293 values indicated
moderate pain34 by exceeding 4 (69% were scored
during suctioning), with 146 and 147 values noted in the morphine and placebo
groups, respectively. Table 2 shows
pain scores at the different time points for the morphine- and placebo-treated
infants. The mean SDs of pain scores for those patients who underwent multiple
procedures were 2.5 for the PIPP, 2.2 for the NIPS, and 2.2 for VAS scores.
Multiple regression analyses revealed that VAS and NIPS scores after
the loading dose of study medication did not significantly differ between
the 2 groups (unstandardized regression coefficient [B] = −0.019; 95%
confidence interval [CI], −0.071 to 0.032; P =
.46; and B = 0.031; 95% CI, −0.053 to 0.12; P =
.47) and were not influenced by withholding the loading dose (B = −0.014;
95% CI, −0.075 to 0.047; P = .65; and B = 0.022;
95% CI, −0.13 to 0.080; P = .67). These pain
scores were significantly predicted, however, by the pain scores before bolus
administration (B = 0.65; 95% CI, 0.53 to 0.78; P<.001;
and B = 0.54; 95% CI, 0.34 to 0.73; P<.001). VAS
scores were higher in girls compared with boys (B = −0.057; 95% CI,
−0.11 to −0.005; P = .03) and higher
in center 2 compared with center 1 (B = −0.065; 95% CI, −0.12
to −0.010; P = .02). Pain scores tended to
be higher when no morphine was given before intubation (B = −0.054;
95% CI, −0.11 to 0.002; P = .06; and B = −0.11;
95% CI, −0.20 to 0.018; P = .02).
The PIPP, NIPS, and VAS scores during suctioning were not predicted
in multiple regression analyses by treatment group or by the amount of additional
morphine used (Table 3). Mean
NIPS and VAS scores decreased with increasing length of study, and VAS scores
were lower in center 1 compared with center 2. Spearman ρ correlation
coefficients between the different pain scores were 0.44 (NIPS vs PIPP, P<.001), 0.31 (NIPS vs VAS, P<.001),
and 0.22 (PIPP vs VAS, P = .02).
Table 4 lists the clinical
outcomes and incidences of morbidity and mortality for the 2 groups. Overall,
11 infants died within 28 days, and 48 were diagnosed as having IVH, 10 of
which had the severe type (grade III or IVH plus apparent periventricular
hemorrhagic infarction). Four infants had PVL. Logistic regression analysis
showed that the incidence of poor neurologic outcome was not related to treatment
group or to additional morphine use (Table
5). It was, however, associated with lower gestational ages (P = .005) and higher CRIB scores (P =
.004) and was more apparent in boys compared with girls (P = .003).
The incidence of IVH (all grades), also evaluated with logistic regression
analysis, was significantly higher in the placebo group compared with the
morphine group (adjusted odds ratio, 2.36; 95% CI, 1.05-5.28; P = .04). Furthermore, the incidence of IVH was associated with lower
gestational ages (P = .006) and was higher in those
born small for gestational age (P = .05) and in infants
born outside the study hospital (P = .04). Median
duration of the first period of artificial ventilation, median total duration
of ventilation, and median length of NICU stay did not significantly differ
between groups (P = .72, P =
.81, and P = .92, respectively).
Open-label morphine was administered to 20 infants (27%) in the morphine
group and 31 (40%) in the placebo group (χ21 = 2.76, P = .10) (Table 6),
with comparable median dosages of 3.0 µg/kg per hour (IQR, 1.3-6.8 µg/kg
per hour) and 4.3 µg/kg per hour (IQR, 1.6-7.7 µg/kg per hour)
in the morphine and the placebo groups, respectively (Mann-Whitney U test, 282.5; P = .60). Of the 60 eligible
but nonenrolled patients, 55% received morphine, with a median dose of 3.6
µg/kg per hour (IQR, 1.7-6.7 µg/kg per hour). These infants received
additional morphine more frequently than the study infants (Kruskal-Wallis
test: χ22= 10.4, P = .005). Among the 2 centers, nonenrolled patients received morphine
more frequently in center 2 (Mann-Whitney U test,
94.0; P = .03).
We hypothesized that continuous morphine infusion in preterm neonates
would reduce pain experience and incidences of poor neurologic outcome and
IVH. However, pain measurements validated for this age group did not reveal
any analgesic effects of morphine. Although routine morphine infusions did
not affect poor neurologic outcomes or any other clinical outcome measure,
preemptive morphine analgesia significantly decreased the incidence of IVH.
These findings suggest that routine morphine infusion in preterm newborns
who have received ventilatory support neither improves pain relief nor protects
against poor neurologic outcome. The impact of decreased IVH in the morphine-treated
neonates, however, should be evaluated with their long-term neurobehavioral
Overall, we found that pain scores did not significantly differ between
the 2 randomized groups. Although the results of pain scores should be viewed
with some caution, the PIPP and NIPS have both been validated for the assessment
of procedural pain in preterm neonates.31,32,35,36 The
sensitivity and specificity of these methods for measuring acute or chronic
pain in preterm infants remain unknown. The VAS has not been specifically
validated for this group of patients but appears to reflect the intensity
of pain.34 In this study, the VAS was applied
by experienced NICU nurses who were specifically trained for assessing neonatal
pain. Measuring the effect of morphine on the pain experienced by preterm
neonates remains difficult because of the lack of a gold standard to assess
neonatal pain. The absence of a measurable analgesic effect of morphine, as
established by these validated pain scores, may be explained by several reasons.
Our patients seemed to experience only minor pain. Most patients showed
no evidence of pain before or 30 minutes after the loading dose. Taking the
limited time span from birth to study enrollment (median, 8 hours; IQR, 5-12
hours) into consideration, the low pain scores may be explained by release
of endorphins, resulting from birth37- 39 and
postnatal stress.40 Since severe pain was mostly
absent, it need not be relieved by morphine.
Pain scores were obtained during an invasive, presumably noxious procedure.
Endotracheal suctioning was the only repetitively, frequently, and routinely
performed invasive procedure during our study. Heel lances were not performed
routinely because all patients had arterial catheters. Furthermore, previous
studies have shown that tracheal suctioning is related to increased pain scores15,41,42 and stress responses43 and is considered painful.44- 47 In
our study, tracheal suctioning was associated with a median PIPP score of
10, NIPS score of 4.8, and VAS score of 2.7, indicating mild to moderate pain.
These physiologic and behavioral responses are indicators of neonatal pain,
but they are also influenced by factors such as gestational age, severity
of illness, and time from the previous painful procedure.48 Previous
studies using these measures have reported large interindividual variability.49
The low correlation between the different pain scores also underlines
the difficulty of pain assessment in this group of patients, as was recently
reviewed by our group.2,34 However,
multivariate analyses, adjusting for these covariates, did not show any statistically
or clinically significant decrease in pain scores resulting from continuous
morphine administration. The explained variance of these analyses was low,
probably the result of low variability of pain scores. The few previous studies
on this subject present conflicting findings. The decrease in pain that resulted
from higher morphine doses compared with the ones used in our study during
endotracheal suctioning and heel lances15,50 was
not confirmed in another study using morphine doses of the same magnitude.51 The samples sizes in our study were considerably
larger and the amounts of morphine used in our study conformed to internationally
Despite the low pain scores, a number of infants were given additional
morphine (27% in the morphine group and 40% in the placebo group). Because
this study aimed to evaluate the effect of routine continuous morphine infusion
in newborns who received ventilatory support on primary and secondary outcome
measures, placebo-treated infants received open-label morphine if deemed to
be in pain. By reflecting variations among patients that occur in real clinical
practice, this study is a pragmatic trial that aimed to inform choices between
treatments (routine morphine administration or no routine morphine infusion).
In pragmatic trials, the treatment response is the total difference between
2 treatments, including both treatment and associated placebo effects, since
this will best reflect the likely clinical response in practice. Because the
intention-to-treat principle was used in our study, patients in both groups
receiving open-label morphine were not dropped out but included in the analyses.
In daily practice, a newborn in pain who receives ventilatory support needs
to receive analgesic treatment, independent of any routine morphine administration.
If an infant was in pain, morphine was given. In this way, our study was a
realistic reflection of 2 different strategies of daily NICU practice.
Clinical bias was minimized via randomization of patients and blinding
of physicians, parents, and investigators. The attending physicians and nurses
obviously considered these infants to be uncomfortable and in need of extra
pain relief, although this was not reflected in their pain scores. The use
of extra morphine was not significantly different between the randomized groups,
as reported previously.15 The nonparticipating
infants received open-label morphine somewhat more frequently than those in
the study group, suggesting that participation in this trial was not a causative
factor for additional morphine prescription. Furthermore, additional morphine
could be used only according to the protocol. Therefore, physicians were allowed
to administer additional doses of 50 µg/kg followed by 5 to 10 µg/kg
per hour continuous open-label morphine. The nonparticipants, however, often
received standard morphine boluses of 100 µg/kg. Additional morphine
use in nonparticipants differed between the 2 centers perhaps due to different
prescribing policies or to different patient characteristics.
Our results are indicative of nonstandardized pain management under
which lack of decision rules results in prescribing analgesics on the basis
of personal clinical experience. This is not only the case in our centers
but also representative of clinical practice in most NICUs worldwide.53 Implementation of pain scores (ie, using cutoff points
for prescribing additional analgesics that are integrated in clinical algorithms
or flowcharts) may be required for rationalizing the use of opioid analgesics
in the NICU. The development of new techniques, such as functional magnetic
resonance imaging and positron emission tomographic scans, might be useful
in the near future to further objectify the analgesic effects of opioids in
newborns, but they are not applicable in daily NICU care.
Morphine use might decrease the fluctuations in cerebral blood volume
and intracranial pressure caused by neonatal reactions to pain and painful
procedures. Morphine may thus protect against the development of venous hemorrhage
in the germinal matrix or brain parenchyma or against the extension of a small
previous IVH.54,55 High pain scores
were not related to the incidence of IVH or poor neurologic outcome. Oberlander
et al56 also found that parenchymal brain injury
did not cause a difference in pain response in premature neonates. Significantly
fewer neonates in the morphine-treated group were found to have IVH compared
with the placebo group. This effect of morphine can be partly explained by
a decreased incidence of low-grade IVH. The impact of routine morphine administration,
by reduction of low-grade IVH, on long-term outcome is hard to predict. Both
PVL and IVH were diagnosed and staged from cranial ultrasounds by staff neonatologists,
using standard criteria.57,58 It
is difficult to determine the neurobehavioral outcome in infants with IVH
because other confounding criteria, such as comorbidity, are involved. Mortality
and major neurologic sequelae are generally related to the degree of hemorrhage59- 63 and,
to a greater extent, to the degree of associated parenchymal injury.58 Infants with IVH grade I and II, without venous infarction,
seem to have little increased risk of adverse outcome compared with those
without IVH.58,60,64- 66 When
we studied the impact of morphine infusion on poor neurologic outcomes (eg,
death, PVL, IVH grade III, or IVH and apparent periventricular hemorrhagic
infarction), there were no differences between the 2 groups.
The neurologic condition of our patients, however, needs to be reevaluated
at older ages. A study by Quinn et al67 also
showed comparable clinical outcomes between placebo- and morphine-treated
neonates. A pilot study by Anand et al,15 with
a slightly different study design, showed decreased poor neurologic outcomes
on account of morphine compared with midazolam hydrochloride and placebo.
Relatively small groups, numbering approximately 20, in those studies, as
well as differences in morphine dose regimen, might explain the differing
results. Further results of that study should conclusively show whether routine
use of morphine reduces the incidences of IVH and poor neurologic outcome.
Overall, our results show a lack of measurable analgesic effect and
absence of a beneficial effect on poor neurologic outcome from routine continuous
morphine infusion in preterm neonates. Future research is needed to establish
cutoff points and an algorithm for the administration of analgesic agents
in this specific age group of children, which should be included in consensus
better understanding of individual differences in responses to morphine and
pain is necessary to improve neonatal pain management.
Our findings suggest that morphine infusion in preterm newborns who
receive ventilatory support should not be used as a standard of care. The
long-term consequences of reduced IVH incidence in the morphine-treated neonates
should be evaluated at predetermined time points at older ages, using validated
assessment instruments for neurodevelopmental outcome.