Pain and neurodevelopmental outcomes of infants born very preterm

To investigate the impact of the level of pain experienced by infants born preterm on neurodevelopmental outcomes during their stay in a neonatal intensive care unit.

painful stimuli during their hospital stay, 15 stimuli occurring per day with two-thirds being performed without any analgesia. 8 Now, Carbajal and colleagues have confirmed that neonates still suffer a considerable number of painful procedures in the NICU. 9,10 These numbers are particularly striking as early exposure to repetitive procedural pain in neonates born very preterm may disrupt the development of regions involved in somatosensory processing, leading to poor functional outcomes. [11][12][13][14] The number of procedures may constitute only one piece of this puzzle, because the severity of the pain event may also play an important role in explaining neurodevelopmental outcomes. To follow this up, Laudiano-Dray et al. tried to quantify the level of severity of these procedures. 15 Furthermore, susceptibility in neuroplasticity seems to be greater in the neonatal period, and a considerable number of studies have linked neonatal pain to health problems in later life. 16 Adverse neonatal experiences have been reported to lead to emotional and behavioural problems such as anxiety, depression, fear, and altered pain perception later in childhood. 17,18 Although well-established in animal models, 7,19 consistent human data on this topic are limited. 17,18,20 A physiological stress response associated with a noxious stimulus results in the activation of the hypothalamic-pituitary-adrenal axis and the subsequent release of cortisol, a key hormone for mediating certain adaptive behaviours in humans. 21 Repetitive exposure to pain during the neonatal period has been reported as a risk factor for dysregulated cortisol secretion 22,23 and could be at the root of later stress-induced psycho-emotional problems. 17 In this regard, it has recently been demonstrated that perinatal stress moderates the link between early and later emotional skills in infants born preterm. 20 Although the effect of repetitive pain on short-and longterm outcomes has already been investigated, 11,18,24 some of these studies focused on behavioural outcomes 18 rather than mental and motor development, 24 by considering the number of procedures rather than the level of pain experienced. 24 Therefore, our study aimed to investigate the effect of the level of pain on neurodevelopmental outcome in a carefully selected group of infants born preterm who had not suffered major morbidities of preterm birth or surgery. Assessment of neurodevelopmental outcome was done by using the Bayley Scales of Infant Development, which is a criterion standard in the examination of cognitive (mental) and motor abilities within the first years of development. 25 We hypothesized within this study that pain would have a negative impact on neurodevelopmental outcomes as measured by the Bayley Scales of Infant and Toddler Development, Third Edition (Bayley-III).

M ET HOD
This single-centre retrospective study was approved by the local ethics committee of the Medical University of Vienna (number 1556/2017). Considering the retrospective nature of the study and the collection of data that already existed, parental informed consent was not needed.
Starting from the assumption that neonates cannot verbalize their feeling of pain and discomfort, we implemented the Vienna Protocol for Neonatal Pain and Sedation (VPNPS) at two NICUs. 26 The VPNPS is a protocol developed by a multidisciplinary team for the regular assessment and management of pain and sedation. 26 The protocol is the result of a structured implementation process where best possible practice is decided according to published evidence. 26 The VPNPS provides strategies for both pharmacological and non-pharmacological interventions based on the ranges given in the Neonatal Pain, Agitation, and Sedation Scale (N-PASS). 26 The N-PASS is systematically scored for patients in pain (e.g. postoperative patients) at every care round within our intensive care units; otherwise the scores are event-related for patients in a stable clinical condition.
The N-PASS was developed as a clinically relevant tool to assess prolonged pain and sedation, as well as acute procedural pain in infants. The N-PASS has been shown to be a valid clinical tool. [27][28][29] It moves on a continuum between pain and sedation, including five criteria (crying-irritability, behaviour state, facial expression, extremities tone, vital signs) graded 0, +1, or + 2 for pain/agitation and 0, −1, or − 2 for sedation. The total score results by adding the scores for each criterion. In our NICU's computerized patients' documentation system (IntelliSpace Critical Care and Anaesthesia -Philips Clinical Healthcare), total N-PASS scores, sedation scores only, and pain scores only can be visualized and extrapolated. Following the experience gained during the implementation of the VPNPS, we focused on the N-PASS pain subscale, particularly on ranges resulting in an intervention or re-evaluation of the actual infants' situations. We therefore considered ranges between 5 and 7 for moderate pain and 8 to 10 for severe pain, according to the VPNPS. 26

What this paper adds
• Pain exposure had a measurable negative effect on neurodevelopment outcomes of infants born preterm. • Focuses on pain measured by the Neonatal Pain, Agitation, and Sedation Scale rather than considering the number of procedures during hospitalization. • Differences between the groups were found on the Bayley Scales of Infant and Toddler Development, Third Edition for both mental and motor development. • This was the case after excluding major morbidities of preterm birth and correcting for other important medical conditions.

Patients
All infants born preterm between 23 and 32 weeks gestational age from 2011 to 2015 who were assessed with the N-PASS and examined at 1 year corrected age with the Bayley-III were included in this study. We excluded all patients with severe intraventricular haemorrhage, cystic periventricular leukomalacia, post-haemorrhagic hydrocephalus, severe retinopathy of prematurity, congenital malformations of the central nervous system, necrotizing enterocolitis, and those who had any major surgical intervention. The level of pain experienced (pain vs no pain) was based on N-PASS categories associated with the VPNPS. 26 All patients who never had a score greater than 4 were categorized in the 'no pain group'; while all infants who experienced at least one pain event were categorized in the 'pain group'. This decision was supported by evidence reported in the literature about the impact of one pain event on outcome. 7,30 Furthermore, subgroup analyses were organized following the VPNPS categorization. In fact, according to our protocol, infants scoring between 8 and 10 are considered in pain and should be immediately treated; while infants with a score between 5 and 7 can be reassessed twice before pharmacological treatment is considered as a valid alternative to non-pharmacological intervention. Proceeding with this reasoning, we performed an exploratory sub-analysis where infants born preterm with at least one score between 8 and 10 and/or more than two scores between 5 and 7 were categorized in the 'severe pain group'; while all others, within the main 'pain group', constituted the 'moderate pain group'. A time-course of the N-PASS values of an exemplary patient is shown in Figure 1.

Data collection
Data from infants born preterm at the Vienna General Hospital between 2011 and 2015 were included in our analyses. This specific time-frame was chosen because the patient computerized documentation system was launched in 2011 and, for children born up to 2015, the 1-year assessments in our follow-up outpatient clinic had been completed at the time of ethical application. Demographic characteristics, NICU data, and N-PASS data were collected from the NICU's computerized patients' documentation system (IntelliSpace Critical Care and Anaesthesia -Philips Clinical Healthcare). Follow-up data including neurodevelopmental outcomes were collected at the follow-up clinic for infants born very preterm at the Medical University of Vienna where they had a routine visit at 1 year corrected age. The children's development was assessed using the Bayley-III.
The Bayley Scales of Infant Development were developed by Nancy Bayley and first published with a US standardization in 1969. 25 It is now the most widely used instrument to assess cognitive and motor abilities in infants and toddlers. It is an individually administered instrument to measure the developmental level of children between 16 days and 42 months of life. 25 It consists of five subtests: cognition, receptive communication, expressive communication, fine motor, and gross motor functions. Scaled scores are calculated for each subtest and range from 1 to 19 with a mean of 10 and a standard deviation of 3, using normative data for the child's age group. These scores can be converted into cognitive, language, and motor composite scores with a mean of 100 and a standard deviation of 15. 25 In this study, considering the age at follow-up of our participants, we mostly focused on mental and motor abilities, where scores between 85 and 115 were considered normal, scores between 70 and 84 indicated mild impairment, and a score lower than 70 moderate to severe impairment.
The Bayley-III cognitive scale evaluates several types of ability: sensory/perceptual discrimination, memory, learning, problem-solving, vocalization, and beginning of verbal communication. The motor scale assesses the degree of body control, coordination, and fine motor manipulation. German F I G U R E 1 Example of longitudinal Neonatal Pain, Agitation, and Sedation Scale (N-PASS) documentation. The x-axis provides information on the timeline (month-day); the y-axis provides information on the level of pain measured through the N-PASS. Each dot on the timeline represents a single N-PASS score and its severity can be contextualized through the corresponding N-PASS ranges normative values were used for Bayley-III as they were more comparable to Bayley Scales of Infant Development, Second Edition scores than US values. 25 Information on maternal education was also considered and categorized according to the Austrian school system: mandatory school, apprenticeship, vocational middle school, school-leaving examination, university. If none of these categories was representative, maternal education was coded as 'other'.

Statistics
Statistical analysis used SPSS Statistics for Mac version 21.0 (IBM Corp., Armonk, NY, USA), while graphical data representations were done with R (R Foundation for Statistical Computing, Vienna, Austria) and JASP version 0.14 (2021; jasp-stats.org/). Qualitative data (describing the attributes or properties of a variable, categorizing it into classes) are shown as counts and percentages, while continuous data are shown as means and standard deviations. A multiple linear regression model was used to isolate the effect of pain on neurodevelopmental outcomes, controlling for the following variables: gestational week, birthweight, sex, low-grade intraventricular haemorrhage, low-grade retinopathy of prematurity, bronchopulmonary dysplasia, mechanical ventilation, patent ductus arteriosus, sepsis, pain, and maternal education. Collinearity was controlled through the values of variance inflation factor and its inverse (tolerance). Variance inflation factor cut-off values were set at 5. 31 All values greater than 5 were considered a sign of collinearity. An unstandardized β was used to understand the slope of the line between the predictive variable and the dependent variable. R 2 values were used to contextualize model fit for mental and motor outcomes. A t-test was used to compare means within the two different pain groups, while a χ 2 test was used to compare categorical variables. An analysis of variance was used for subgroup analysis of the three pain groups, while a Tukey test was used as post hoc analysis to control the interaction between pain and analgosedation. A p-value of 0.05 or less was considered statistically significant.

R E SU LTS
Between January 2011 and December 2015, a total of 887 infants born preterm before 32 weeks of gestation were admitted to our unit ( Figure S1). All children with significant neurological problems, congenital malformation, any kind of surgical interventions, death or lost to follow-up, having fewer than three N-PASS assessments, and having been assessed with the Bayley Scales of Infant Development, Second Edition were excluded from this study ( Figure S1), leaving 196 patients for further analysis. Most of the study participants were European (87.7% in the no pain group vs 84.6% in the pain group), followed by Serbian native (2.9% in the no pain group vs 7.7% in the pain group), and Turkish native (1.9% in the no pain group vs 4.4% in the pain group).
A total of 6307 N-PASS scores were analysed with a median of 25 scores per patient. Most (n = 5958) scores were between the categories 0 and 4; 331 of the assessments were between 5 and 7, and only 10 were between 8 and 10. A significant mild correlation was found between the number of pain events and the number of N-PASS scores (r = 0.44; p < 0.001), emphasizing the fact that an infant being in pain was also frequently scored. No differences were found between males and females for the total number of pain events (p = 0.17). According to our selection algorithm based on the VPNPS, 105 infants born preterm were included in the 'no pain group', while 91 were included in the 'pain group'. The descriptive characteristics of the study group are presented in Table 1. Despite inclusion only of patients without any major medical complications, the two groups differed for some descriptive characteristics (Table 1). Also, as expected, patients in the 'pain group' received significantly more analgesic drugs (Table 1). Significant differences were also found for medical conditions such as low-grade retinopathy of prematurity and bronchopulmonary dysplasia, but not for low-grade intraventricular haemorrhage (grade 1) ( Table 1). Information about maternal education was available for 105 patients (44 in the pain group and 61 in the no pain group), and no differences could be detected between the groups (p = 0.31). Seven (11.4%) mothers in the 'no pain group' versus five (11.4%) in the 'pain group' had an elementary school degree; while 22 (36%) in the 'no pain group' versus 10 (22.7%) in the 'pain group' had a university degree. Furthermore, four (6.5%) mothers in the 'no pain group' versus one (2.2%) in the 'pain group' were not classifiable according to the Austrian school system, while 12 mothers in both groups (19.6% vs 27.2% respectively) had an apprenticeship; five (8.1%) mothers in the 'no pain group' versus nine (20.4%) mothers in the 'pain group' had a vocational middle school degree; finally, 11 (18%) mothers in the 'no pain group' versus seven (15.9%) in the 'pain group' had a school-leaving examination.
Follow-up data showed no differences in demographic variables. However, differences were found in mental and motor development with adverse effects seen in the 'pain group' (Table 2). A significant, yet weak correlation was found between the number of pain events and mental (r = −0.22; p = 0.002) and motor (r = −0.24; p = 0.001) outcomes. Figure 2 presents the distribution of mental and motor indices in the two groups. According to a statistical model controlling for gestational week, birthweight, sex, low-grade intraventricular haemorrhage, low-grade retinopathy of prematurity, bronchopulmonary dysplasia, mechanical ventilation, patent ductus arteriosus, sepsis, and maternal education (Table 3), pain was found to have a statistically significant effect on both mental and motor development (p = 0.001 for mental and p = 0.010 for motor development), with an impact of 12.53 points on the Bayley-III for mental and 9.23 points for motor development. In the regression model, no signs of collinearity were found, since none of the considered variables showed a variance inflation factor greater than 3. Moreover, the model fit expressed by the R 2 value was 0.303 and 0.216 respectively for mental and motor development. By performing subgroup analyses, looking at different levels of pain, differences could be found between the 'no pain group' and 'severe pain group' for both mental (p = 0.029) and motor scores (p = 0.030). Also, when controlling for the interaction between different levels of pain and administration of analgosedative drugs, exposure to analgosedation was not associated with adverse neurodevelopmental outcomes (p = 0.72 for mental scores, and p = 0.41 for motor scores) (Figure 3).

DISCUS SION
The current retrospective data analysis, featuring a carefully selected collective of infants born preterm with an uncomplicated course of hospitalization, provides a unique opportunity to understand the effect of the level of pain experienced in the NICU on neurodevelopmental outcomes in these highly vulnerable patients. According to our results, pain was associated with lower mental and motor performance at 1 year corrected age using the Bayley-III. In particular, an N-PASS indicating pain was associated with adverse neurodevelopmental outcomes when also controlling for other important medical conditions. Pain is a delicate topic within every intensive care unit. In a NICU setting, infants born preterm are particularly affected by certain painful experiences that are also clearly associated with intense behavioural and physiological responses. 1,32 Intense physiological responses can be caused by the immaturity of the nociceptive system, reduced thresholds for pain stimuli, and weakness of circuits related to pain modulation. 3 Also, the organization and interaction of neurotransmitters, as well as the role and distribution of receptors, in neonates are far from resembling those of adults, playing a different role in the processes of excitation and inhibition. 3,33 It is very likely that thalamic connections with the somatosensory cortex underlie cortical pain input in infants born preterm. 3,4 Thus, the anatomical substrate for interactions between the periphery and the cortex seems to be accessible in early treatment procedures. In fact, studies by Bartocci et al. 5 and Slater et al. 6 were both able to demonstrate a cortical reaction to noxious stimuli in infants born preterm as early as the 25th week of gestation.
In recent years, convergent results of clinical trials of infants born preterm exposed to pain during their stay in the NICU have been emerging. 7,12,24,34 Thus, the relation between neonatal pain exposure and neurodevelopmental outcome seems to be explained mostly by altered brain development. Brummelte et al. reported that greater exposure to procedural pain was associated with reduced volumes of white and grey matter. 13 Duerden et al. were able to demonstrate that early procedural pain was associated with specific thalamic alterations; 12 meanwhile, Doesburg et al. reported alterations in spontaneous brain activity at school age. 34 Congruently, in animal models, pain induced by formalin injection was shown to induce neural apoptosis and altered expression of proteins essential for brain development. 19 Thanks to major improvements in the field of neonatology, it is possible today to observe a major reduction in severe outcomes in infants born extremely preterm and very preterm. 32 However, even children born preterm with F I G U R E 3 Interaction between mental and motor scores stratified for level of pain and exposure to analgosedation a normal IQ, without a history of brain lesion, seem to report difficulties in specific neuropsychological domains potentially related to the areas of self-regulation, arousal, attention, and executive control. 32 A link between brain alterations and repeated painful stimulation has already been discussed and could represent a first-line explanation for poor cognitive and motor outcomes in infants born preterm. Interestingly, according to Doesburg et al., alteration in thalamocortical brain connections, critical for cognition and perception, were associated with visual problem-solving abilities. Similarly, Vinall et al. 14 reported that the interaction between the number of procedures and fractional anisotropy was associated with infants' IQ. Although many studies have concentrated on the impact of sedative and analgesic drugs on the developing brain, 35,36 less is known about the direct impact of pain on mental and motor outcomes. Similarly to our study, Grunau et al. reported that greater exposure to neonatal pain was associated with inferior motor and cognitive outcomes at the corrected age of 18 months. 24 Although both studies provide congruent findings, and focus on a well-selected collective (e.g. exclusion of patients with severe brain lesion), results are not directly comparable since the study by Grunau et al. focused on exposure to procedural pain operationalized as the sum of every skin-breaking procedure. Indeed, our study approached pain from an underinvestigated perspective, linking its level of severity to neurodevelopmental outcomes. Interestingly, Grunau et al. also controlled their results for other important factors potentially having an effect on the outcomes. Again, like our study, they did not find an association with gestational age, and only a trend towards maternal education was shown in their study. Nonetheless, according to Patra et al., 37 maternal education seems to be the strongest predictor of neurodevelopmental outcomes, compared also with other neonatal morbidities or sociodemographic factors, having an impact on motor and language outcome in addition to cognitive function.
Although pain can alter brain development and impact neurodevelopmental outcomes, the same is true for exposure to sedative and analgesic drugs. In fact, most of these substances act as NMDA (N-methyl-d-aspartate) receptor antagonists as well as GABA (γ-aminobutyric acid) receptor agonists. Preclinical studies have demonstrated these substances as having neurotoxic properties affecting the developing brain, in particular during synaptogenesis between the third trimester of gestation and the third year of life. 38,39 Adverse neurodevelopmental outcomes related to analgesic drug exposures have been reported not only in animal models but also in clinical studies. According to Grunau et al., investigating the effect of procedural pain in 137 infants born preterm on mental and motor development according to the Bayley Scales of Infant Development, Second Edition, opiate exposure was considered a risk factor for adverse motor development at the age of 8 months. 24 Furthermore, according to secondary outcomes deriving from the NEOPAIN (Neurological Outcomes and Pre-emptive Analgesic in Neonates) trial, pre-emptive morphine analgesia was distinctly associated with problems in short-term memory tasks (delayed matching to sample), and with longer latencies of choice response, than in comparison infants. 40 Also, De Graff et al. found that scores on the IQ subtest 'visual analysis' were significantly lower for children who had received morphine and open-label morphine during the neonatal period. 36 However, these effects dissipated when looking at neuropsychological outcomes at the age of 8 to 9 years. 41 Compared with these results, we were previously able to demonstrate that systematic assessment of pain and management of sedation has no impact on neurodevelopmental outcome, even if associated with an increased use of analgesic drugs. 42,43 Moreover, in this study, when controlling the interaction between different pain levels and analgosedation, we found no significant impact on mental and motor development.
As stated by Dan, 16 improving neonatal pain assessment and management is the starting point. This is in line with our conclusion, as also recommended by the American Academy of Pediatrics. 44 Today, much evidence supports the fact that neonatal pain can alter brain development and affect neurodevelopmental outcomes. However, there are few studies at preschool and school age, and less is known about possible behavioural dysregulation and cognitive performance at this stage of development.
There are some limitations to our study. It has provided information about the level of pain exposure in a well-selected collective of infants born preterm with no major morbidities of preterm birth and no history of surgery. Nevertheless, judgements about the level of pain depended on the NICU's staff, as neonates cannot verbalize their feelings. Considering the inclusion criteria of our selected collective, the frequency of documentation was almost event-related and may have had an impact on group assignments. Moreover, we recognize that, given the intensive care setting, it is not completely terminologically correct to use the term 'no pain'. We refer in these cases to infants who were not documented as being in a painful state according to the expert opinion of our caregiving team. Finally, even if subgroup analysis were performed in agreement with VPNPS cut-off values, these should be interpreted as explorative.

CONCLUSION
In our study, pain exposure had a measurable negative effect on neurodevelopment at 1-year corrected age in a group of infants born very preterm with no major morbidities of preterm birth. This observation highlights the importance of adequate pain management to reduce the risk of poor longterm neurodevelopmental outcomes in these highly vulnerable patients.

C ON F L IC T OF I N T E R E S T
The authors have stated that they had no interests that might be perceived as posing a conflict or bias.

DATA AVA I L A BI L I T Y S TAT E M E N T
All data generated or analysed during this study are included in this article [and/or] its supplementary material files. Further enquiries can be directed to the corresponding author.