Taddio A, Shah V, Gilbert-MacLeod C, Katz J. Conditioning and Hyperalgesia in Newborns Exposed to Repeated Heel Lances. JAMA. 2002;288(7):857–861. doi:10.1001/jama.288.7.857
Author Affiliations: Department of Pharmacy, The Hospital for Sick Children, Graduate Department of Pharmaceutical Sciences, University of Toronto, Toronto (Dr Taddio), Department of Paediatrics, Mt Sinai Hospital, Toronto (Dr Shah), and Department of Anesthesia and Pain Management, Toronto General Hospital and Mt Sinai Hospital, Departments of Public Health Sciences and Anesthesia, University of Toronto (Dr Katz), Ontario; and Maritime Psychiatry, IWK Grace Health Centre, Halifax, Nova Scotia (Dr Gilbert-MacLeod).
Context Hospitalized infants undergo repeated invasive procedures. It is unknown
whether cumulative experiences with pain lead to anticipatory pain behaviors
Objectives To determine whether newborns who are born to mothers with diabetes
and undergo repeated pain learn to anticipate pain and exhibit more pain during
a painful procedure than normal infants.
Design, Setting, and Participants Prospective cohort study of 21 full-term newborns born to mothers with
diabetes and 21 born to mothers with uneventful pregnancies, at a university
teaching hospital between August 1999 and October 2000. Infants of diabetic
mothers underwent repeated heel lances in the first 24 to 36 hours of life
for monitoring of blood glucose concentrations. Pain responses of all infants
undergoing a venipuncture on the dorsum of the hand to obtain blood for the
newborn screening test after the first day of life were compared. In addition,
from September through November 2001, 12 infants of diabetic mothers and 12
normal infants were compared for pain reactions to intramuscular vitamin K
injection after birth.
Main Outcome Measures Percentages of time observed grimacing and crying and visual analog
scale (VAS) scores.
Results Raters were blinded to exposure group. Median baseline scores for grimacing,
crying time, and VAS did not differ significantly between groups (P = .27, P = .32, and P = .32, respectively). Median scores (interquartile range) for grimacing
during skin cleansing were higher in infants of diabetic mothers (22.2% [77.5%]
vs 0% [15%]; P = .03). The VAS scores for both groups
were zero, but the distribution of the scores was significantly different
(86% of normal infants vs 52% of infants of diabetic mothers had scores of
zero) (P = .04). During venipuncture, infants of
diabetic mothers had higher median scores for grimacing (81.7% [32.5%] vs
40% [73.4%]; P = .01), VAS (69% [27.5%] vs 5% [60.5%]; P = .002), and crying (40.2% [77%] vs 0% [54.8%]; P = .03) compared with normal infants. There were no differences
between groups on any pain measure in response to intramuscular injection.
Conclusions Newborns who had diabetic mothers and were exposed to repeated heel
lances in the first 24 to 36 hours of life learned to anticipate pain and
exhibited more intense pain responses during venipuncture than normal infants.
Ten percent to 15% of newborns require hospitalization after delivery
for medical reasons including prematurity, congenital anomalies, sepsis, jaundice,
and feeding difficulties.1 Hospitalized newborns
are subjected to many invasive procedures as part of their medical care. There
are anecdotal reports of infants learning to anticipate pain and exhibiting
altered pain responses as a consequence of their cumulative exposures to pain.2
The objective of this study was to prospectively examine the hypothesis
that infants who undergo repeated painful procedures as part of their medical
care learn to anticipate pain and exhibit more intense pain during a subsequent
painful procedure compared with control infants.
This was a prospective cohort study conducted in the postnatal ward
of a university-affiliated hospital. The study was approved by the local ethics
board, and written consent was obtained from parents of all participating
infants. There were 2 groups of full-term neonates: those born to mothers
with diabetes (type 1, type 2, or gestational) and those born to mothers with
uneventful pregnancies. We excluded infants who were admitted to the neonatal
intensive care unit, those receiving intravenous or intramuscular antibiotics,
or those who were scheduled for circumcision during the study. Posters were
placed in the delivery suite, and nurses and ward clerks were asked to contact
the investigator on call when a potential study participant was admitted in
Infants of diabetic mothers underwent repeated heel lances in the first
24 to 36 hours of life for monitoring of blood glucose concentrations. When
the study was conducted, the clinical guidelines for glucose concentration
monitoring involved obtaining a blood sample within the first hour after delivery
and then every 2 to 4 hours during the first 24 hours of life. The frequency
of blood sampling was conditional on prior glucose concentration readings
and infant symptoms of hypoglycemia. Blood was collected for glucose monitoring
by heel lance with automated Microtainer Safety Flow lancets (Becton, Dickinson
and Co, Franklin Lakes, NJ). Normal infants were matched to those of diabetic
mothers on birth weight (±500 g), sex, vaginal or cesarian delivery,
and anesthesia (yes or no) during labor and delivery.
After the first 24 hours of life, all infants underwent a venipuncture
to obtain blood for the newborn screening test. Venipunctures were performed
in the treatment room on the postnatal floor. The infants were optimally positioned
for the procedure (ie, swaddled and lying supine on the examination table).
The site chosen for all venipunctures was the dorsum of the hand. A 23-gauge
butterfly needle was used. Venipunctures were performed by a variety of nurses
who were unaware of the study hypothesis. To standardize observations, the
procedure was divided into 3 phases: baseline, skin cleansing and preparation,
and needle puncture and collection of the blood sample. Infants' faces were
videotaped for the entire procedure.
Pain was rated later from videotapes by using 3 validated behavioral
measures: grimacing (based on the Neonatal Facial Coding System),3 visual analog scale (VAS), and crying time percentage.
Three trained research assistants who were unaware of infant exposure group
viewed the videotape and rated infant pain by using 1 of the 3 pain measures.
The order in which infants appeared on the videotape was random. The videotape
contained no information or features that could identify an infant as belonging
to either group.
Grimacing has undergone extensive validity testing and is considered
the gold standard for assessment of pain in infants.4,5
For the grimacing scores, we used 3 facial actions that are particularly sensitive
for indicating pain: bulging the brow, squeezing the eyes closed, and furrowing
the nasolabial area.6 Each facial action was
scored as present or absent in 2-second intervals for the first 20 seconds
of each phase of the procedure (or less if the phase lasted fewer than 20
seconds). The percentage of the total time that each facial action was observed
during each of the 3 phases was calculated. An overall grimacing pain score
was computed for each phase by summing the individual facial action scores
and dividing by 3. Interrater reliability was assessed for each facial action
in 19% of participating infants; the κ values ranged from 0.86 to 0.91.
Pain ratings using the VAS were based on global facial actions as well
as other observable signs of pain, such as body movements and crying. Visual
analog scale pain scores were rated on an unmarked 100-mm pain ruler, where
0 was no pain and 100 was the worst possible pain, expressed as a percentage.
The VAS was scored across the same period as the grimacing scores. Crying
time percentage was calculated across the entire period for the baseline and
cleansing phases and across the first 2 minutes after venipuncture.7 Crying and VAS pain ratings correlate with grimacing
ratings and provide social validation of the pain scores obtained from the
grimacing scoring system for infant caregivers,8,9
allowing the pain scores to be understood in common terms.
To rule out the possibility that the infants exposed to multiple heel
lances may react differently to venipuncture as a consequence of their medical
condition and not repeated experience with pain, we compared the responses
of a separate cohort of infants of diabetic mothers with those of a group
of normal infants during intramuscular vitamin K injection, which is the first
painful procedure performed on all newborns. This protocol was carried out
in the same institution, with ethical approval and informed parental consent.
Infants were matched on type of delivery.
Rating of pain in response to the intramuscular vitamin K injection
was done from videotaped records and was divided into 3 phases: baseline,
skin cleansing and preparation, and needle puncture and injection. Grimacing,
VAS scores, and crying were calculated as described for venipuncture: during
baseline, skin cleansing, and the first 20 seconds after vitamin K injection.
There were no studies on which to base a sample-size calculation. In
a study of venipuncture pain in neonates, we observed a mean pain score of
3 (SD, 1) on a behavioral scale that ranges from 0 to 7.10
A sample size of 20 infants per group was chosen because it was feasible and
large enough to achieve a difference of 0.8 in the pain score, assuming an
SD of 1 (ie, moderately large effect size), with 70% power and α = .05.
Demographic data were compared between groups with the Fisher exact
test (categorical data) or t test (continuous data).
Pain scores were compared between groups by using the nonparametric Mann-Whitney
test because of the distribution characteristics of the data. Linear regression
analysis was used to determine the difference in venipuncture scores between
groups after adjustment for the cleansing phase. P≤.05
was considered significant, and there was no adjustment for multiple comparisons.
All analyses were performed with Statistical Package for the Social Sciences
version 10 (Chicago, Ill).
The study was carried out in 2 phases. The first phase was conducted
between August 1999 and October 2000. The investigator approached the parents
of 66 infants eligible to participate: parents declined participation for
11 infants (6 infants of diabetic mothers and 5 normal infants). Thus, 55
infants were recruited into the study. Seven infants of diabetic mothers were
excluded from the analysis because consent was later withdrawn by parents
(n = 2), the infant was admitted to a neonatal intensive care unit (n = 2),
the venipuncture was unsuccessful (n = 2), or the infant was crying inconsolably
before the procedure (n = 1). In the normal infant group, 6 infants were excluded
because of videotaping errors that prevented identification of infants or
the procedure phases (n = 3), the venipuncture was unsuccessful (n = 1), or
the infant was crying before the procedure (n = 2).
Data from the remaining 42 infants (n = 21/group) who completed the
study were included in the analysis (Table
1). In 1 infant of a diabetic mother–normal infant pair, there
was a mismatch on infant sex and delivery; however, data from these infants
were retained for analysis. Gestational age was statistically significantly
higher in normal infants (P = .006). Ninety percent
of mothers in both groups received epidural or spinal local anesthesia during
delivery. The mean number of heel lances was 9.8 (SD, 2.1) for infants of
diabetic mothers, and the last monitored mean blood glucose concentration
before venipuncture was 65 (SD, 9) mg/dL (3.6 [SD, 0.5] mmol/L).
Videotaping errors precluded analyses of infant responses during the
cleansing phase in one infant and during venipuncture in another infant in
the infants of diabetic mothers. Median baseline scores for grimacing, VAS
pain scores, and crying times did not differ significantly between groups
(P = .27, P = .32, and P = .32, respectively) (Table 2). During the skin-cleansing phase, grimacing scores were
significantly higher in the infants of diabetic mothers than in normal infants
(22% vs 0%; P = .03). Although the baseline median
scores for the VAS and crying time were 0% in both groups, the distribution
of scores differed statistically for the VAS (P =
.02). Eighty-six percent of normal infants had VAS scores of 0% compared with
52% of infants of diabetic mothers (P = .04). During
venipuncture, the infants of diabetic mothers had significantly higher pain
scores than the normal infants for grimacing (81.7% vs 40%; P = .01), VAS (69% vs 5%; P = .002), and crying
(40.2% vs 0%; P = .03). Differences in venipuncture
pain between groups remained significant for grimacing (P = .05) and VAS scores (P = .007) but not
for crying (P = .23) after adjustment for pain scores
during skin cleansing.
The second phase of the study was conducted throughout a 3-month period,
September through November 2001. Thirty-one infants were eligible to participate:
15 infants of diabetic mothers and 16 normal infants. Two infants of diabetic
mothers and 3 normal infants were excluded because their parents declined
to participate. Data from 2 of the 26 infants who participated in the study
were excluded from the analysis, leaving 12 infants in each group (Table 3). Reasons for exclusion included
inconsolable crying (n = 1) and venipuncture (n = 1) before vitamin K injection.
Infant characteristics and pain responses during vitamin K injection were
not significantly different between infants of diabetic mothers and normal
infants (Table 4).
In this study, newborns exposed to repeated heel lances in the first
24 to 36 hours of life learned to anticipate the pain of an impending venipuncture
compared with normal infants who had not undergone repeated painful procedures.
This finding was demonstrated by the significantly greater VAS and grimacing
pain scores they exhibited compared with the normal infants during the skin-cleansing
phase that immediately preceded the venipuncture. Without repeated exposure
to painful procedures, skin cleansing in neonates is a nonpainful event and
induces little response. However, after an average of about 10 heel lances
(preceded by skin cleansing), infants responded to cleansing of the skin with
behaviors indicative of pain. We suggest that throughout the course of repeated
exposure to skin cleansing followed by heel lancing, skin cleansing became
a conditioned stimulus, reliably signaling the impending painful event and
inducing anticipatory pain behaviors and possibly even pain.
Not only did these infants learn to anticipate pain but also the intensity
of the pain they experienced in response to venipuncture was greater than
that of normal infants. The difference in venipuncture pain scores between
the infants exposed to multiple heel lances and normal infants was maintained
when we adjusted for pain response during skin cleansing. In addition, we
chose a venipuncture site on the dorsum of the hand, in uninjured tissue,
remote from the site of repeated heel lancing to avoid testing in a region
that had been rendered hyperalgesic because of injury and inflammation.11
Conditioning and hypersensitivity responses have not been consistently
demonstrated. In one study of the cortisol responses of newborns,12 increased salivary cortisol concentrations were observed
after a second heel lance. This result was in contrast to a second experience
with a handling stressor (mock discharge examination), which produced a lower
cortisol response than the first exposure. In another study,13
infant crying response was not different during the preparatory phases of
a heel lance when there was one previous experience with a heel lance.
Learning and memory have been demonstrated in preterm infants as young
as 32 weeks of gestation by using paradigms of exposure learning, classic
conditioning, and habituation.14 The gestational
age at which newborns can begin to learn the associations between events involved
in painful procedures is unknown. In a study that assessed preterm and full-term
infant responses during the phases of painful medical procedures,15 the magnitude of heart rate change between the baseline
and procedure phase increased significantly over time with increasing conceptional
age. Investigators did not discuss whether there were similar changes in infant
responses between the baseline and preparatory phases. Thus, whether conditioning
contributed to the observed hyperalgesia is unknown.
We do not know the extent to which the anticipatory pain behaviors and
heightened pain responses persist beyond the first days of life and what impact,
if any, this early exposure might have on subsequent behavior, learning, and
memory. Cumulative experience with procedural pain may have long-term effects.
In a recent longitudinal investigation of preterm infants' responses to heel
lance, differences were observed in behavioral and physiologic responses to
pain at 32 weeks' postconceptional age that were correlated with the number
of previous invasive procedures. Changes in pain reactivity were normalized
if infants had received morphine, suggesting that analgesia ameliorated the
effects of cumulative pain on later pain reactivity.16
The number of invasive procedures in the preceding 24 hours significantly
predicted grimacing in a separate study of very low-birth-weight infants undergoing
medical procedures.17 However, in another study
by the same group of investigators, only subtle differences in pain response
lingered at 4 months' corrected age between former very low-birth-weight infants
and full-term controls during finger-lance blood collection.18
We previously showed that a single painful experience in the neonatal period
was associated with long-term effects.19,20
In 2 studies, neonatal circumcision was associated with increases in pain
following routine 4- or 6-month vaccination. Preoperative treatment with a
local anesthetic cream attenuated vaccination response, suggesting that vaccination
pain response was affected by circumcision pain experienced months before.19
There are limitations to this study. Because the study could not be
randomized, the differences in pain we observed between the groups may be
explained by factors other than the effects of repeated exposure to painful
procedures. We matched infants on birth characteristics. Infants of diabetic
mothers were on average 1 week younger than normal infants. This difference
in gestational age has not been associated with significant differences in
pain response in full-term infants.21 Moreover,
we matched on infant birth weight because weight (or size) of the infant may
be related to the ease of venipuncture.
Because randomization was impossible, the results can possibly be explained
by preexisting differences between infants of diabetic mothers and control
infants, specifically with respect to their medical condition. To evaluate
this possibility, glucose concentrations were monitored closely, and the value
before venipuncture was in the normal range. It is therefore unlikely that
low glucose concentrations (hypoglycemia), which may be associated with clinical
symptoms such as jitteriness,22 account for
the observed differences in pain behaviors. To further evaluate this possibility,
we compared pain responses during the first painful medical procedure performed
after birth (intramuscular injection of vitamin K) in a different cohort of
normal infants and infants of diabetic mothers. The latter did not demonstrate
an increased response during either skin cleansing or intramuscular injection.
Although the number of infants in this cohort was small and a significant
difference may not have been detected, it is reassuring that on nearly every
pain measure, infants of diabetic mothers scored lower than normal infants.
Therefore, preexisting differences in pain responses between groups probably
do not account for the significant differences we observed during venipuncture.
A final limitation is that parents and videographers were aware of infant
exposure status. We attempted to minimize bias during data collection by ensuring
that all infants were settled at baseline, using a variety of nurses to collect
blood samples, and using raters who were unaware of infant exposure group
to score pain from videotaped records. Videotaped records did not contain
any information that could identify the infant's exposure group. Raters were
unaware of any physical features that might have distinguished infants of
diabetic mothers from normal infants, and since birth weight was similar in
both groups, it is unlikely that infant appearance differed as a function
of group. Furthermore, if parents were present during the venipuncture, they
did not interfere with the procedure or speak about the study. In addition,
we excluded infants who were admitted to the neonatal intensive care unit
after delivery. All study infants stayed with their mothers, so separation
from the mother could not explain observed differences between groups.
In summary, full-term infants exposed to repeated heel lances demonstrated
heightened pain in response to skin cleansing and venipuncture compared with
a group of normal infants. This finding has important implications. First,
pain in hospitalized newborns may be classically conditioned by the unwitting
pairing of a neutral stimulus with a painful procedure. Thus, pain and pain
behaviors in neonates will be elicited by the previously neutral stimulus
after conditioning has occurred. Second, previous pain experience may be important
in determining analgesic efficacy. Because sick preterm and full-term infants
are subjected to many invasive procedures as part of their medical care, there
may be significant cumulative effects of pain that might lead to administration
of higher-than-normal doses of analgesics.
Our data suggest that effective management of neonatal pain during invasive
medical procedures might prevent the development of hyperalgesia, conditioned
pain, and conditioned pain behaviors. Taken together, these data provide further
evidence that infant pain is modulated by experiences with pain, as it is
in children and adults.