A litany of milestones—the first heartbeat of a fetus, the first movements, and finally, birth—mark the early development of a human being. And, somewhere along the way, comes the ability to feel pain. A new study looks in the brain to see when that might happen. Researchers led by Lorenzo Fabrizi, Rebeccah Slater, and Maria Fitzgerald at University College London in the U.K. studied the electroencephalogram (EEG) activity of premature and full-term newborns, and found that at approximately 35 to 37 weeks of gestation, the brain’s response to a painful stimulus shifts from nonspecific neuronal bursts to adult-like activity. Their paper was published online September 8 in Current Biology.
The findings suggest that this timeframe, right at the end of gestation, may represent a critical period in the development of somatosensory processing in the human brain when infants begin to distinguish between painful and painless experiences. Further, because burst activity is an important driver of nervous system development, the observation that noxious stimuli strongly evoke burst activity in premature infants supports the idea that early painful events have the potential to disrupt development at a vulnerable time.
In previous studies, the team found that in full-term infants, painful or gentle contact evoked distinct, adult-like patterns of brain EEG activity (Slater et al., 2010). By placing electrodes on the babies’ scalps, they could measure electrical activity immediately after a gentle touch (a light tap on the heel) or noxious stimulus (a heel lance required for a blood sample). Evoked potentials in electrodes near the midline of the brain increased in response to both types of stimuli, but part of the signal extracted from one electrode was specifically elevated by the lance. The researchers say that the lance-evoked signals observed in infants line up with pain-related brain activity in adults (Bromm and Scharein, 1982).
In the new study, Fabrizi, Slater, and coworkers looked at brain activity in premature versus full-term newborns. They found that in premature newborns, light tap or heel lance evoked stimulus-specific activity less frequently in premature infants than in babies who had been born at full gestation: Touch-specific potentials occurred in only 7 percent of premature infants, compared to 30 percent of those at full term, and nociceptive-specific potentials occurred in 33 percent of preemies versus 63 percent of full-term infants.
While nociceptive-specific potentials were less apparent at early gestational ages, noxious stimuli did boost nonspecific neuronal bursting at widespread locations around the brain. In premature infants, bursts occurred spontaneously in 13 percent of background measurements, heel tap increased the frequency to 27 percent, and noxious heel lance increased it further, to 57 percent. (In full-term infants, noxious stimuli only marginally increased the burst rate compared to background.)
Comparing EEG activity from newborns of a range of gestational ages (24 to 45 weeks at the time of testing) revealed that over time, nonspecific burst behavior was gradually replaced by the more adult-like, stimulus-specific activity. For both touch and nociceptive responses, the switch happened at approximately 35 to 37 weeks, a moment the authors call “a critical crossover period in brain development.” Interestingly, this transition in the processing of mechanical stimuli coincides with a similar switch recently identified in visual processing (Colonnese et al., 2010).
One interpretation of the work is that the crossover period is when babies acquire the ability to tell painful stimuli from benign touch, but that is difficult to prove. “We cannot be sure that the brain does not distinguish noxious from gentle stimulation early in development, but we can say they are not distinguished in the same way we distinguish them later in life,” Fabrizi told PRF.
The work also suggests that nociceptive activity in premature newborns could have lasting effects, Fabrizi and his coauthors said. During normal brain development, spontaneous burst activity drives the formation of neuronal circuits. The new findings show that the same type of activity is also strongly evoked by a painful procedure. Thus, tissue-damaging stimuli—which premature infants experience frequently during medical treatment—has the potential to interfere with normal cortical development, and could account for the abnormal pain perception reported among children born premature (e.g., see Hohmeister et al., 2010).
The observation that premature babies’ brains respond to noxious stimuli with neuronal burst activity shows just how special their brains are. “The central nervous system [of premature newborns] is not just a scaled version of the one in term infants; it’s different. We know very little about the term ones, but we know almost nothing about the premature ones,” Fabrizi said. “We need to know more.” The goal, he said, is to combine brain activity, behavior, and reflex responses into an integrated framework that could be used to assess infants’ pain in the clinic.
Image credit: L. Fabrizi, University College London
Comments
Ruth E. Grunau, University of British Columbia & Child and Family Research Institute
There has been long-standing speculation that pain in preterm infants may negatively affect brain development, given the rapid maturation of microstructure, macrostructure and connectivity occurring during the last trimester of human fetal life, e.g. [1]. Fabrizi et al report a ‘delta brush’ response to touch as well as skin-break in preterm - but less so in fullterm - neonates. This finding of non-specific evenly dispersed neuronal bursts of EEG responses to touch and pain in the preterm infants appears to underscore the hypersensitivity of the preterm neonate to both mechanical and nociceptive stimulation. This phenomenon may contribute to understanding why diaper change can evoke as much behavioral and autonomic reactivity as blood collection, in preterm neonates in the neonatal intensive care unit [2].
This work provides a foundation to begin to understand the way pain might contribute to altered spontaneous neuromagnetic activity [3] and atypical long-range task-dependent magnetoencephalographic synchronization [4], as well as a host of atypicalities in structural and functional brain development revealed by MR imaging in children born very preterm (see [5] for review).
Given the important role of sleep for brain development (see [6] for review), more complete exploration of sleep-states in future EEG studies of neonatal pain is needed. Fabrizi et al reported that the occurrence of the noxious-specific potentials was not state-dependent, since the proportion of awake infants was the same for those who did and did not exhibit the noxious-specific EEG response. However, very preterm infants typically are in a state of arousal ranging from drowsy/active sleep/deep sleep while in the NICU.
References
Celeste Johnston, McGill University
This article is interesting in that the there are many questions that arise from presented results. The inherent assumption in this article is that sensory cortical activity as measured by EEG over the sensory cortex is reflective of nociception. As infants mature, their EEG responses appear to distinguish ‘pain’ from touch. While the maturation of responses to touch and tissue-damaging stimuli is of interest, the conclusions drawn by the authors go beyond the data.
Beginning at least two decades ago, it was repeatedly demonstrated that preterm neonates responded in both behavioural and physiological ways differentially to tissue damaging events such as heel lance or injection compared to touch, in neonates as young as 26 weeks gestational age. So, if they are showing differential behaviours regardless of age, how can they not be differentiating the type of stimulus? This leads to another point.
The EEG bursts are really only degrees of response measured in frequency which the authors label pain or not pain. By using Principal Components Analysis, they claim that specific ranges are indicative of specific pain/not pain responses. This is based on a small sample of full term infants. To then extrapolate these patterns to a different population of preterm neonates is unsubstantiated.
Pain is more complex than nociception and at the very least, the authors should discuss nociception and not pain. A decade and a half ago it was first shown that different areas of the brain were active in response to the emotional/evaluative aspect of pain and numerous studies have further documented this. The authors in this article have only examined one area of the brain. It would be interesting to see reports of other areas, including subcortical responses.
There are further methodological considerations that make us take the results with some caution. The postnatal age range was enormous and was not taken into account. It has been well established that postnatal age, particularly when associated with prior pain exposure, can greatly influence pain response and also probably perception.
The authors are to be congratulated in tackling the issue of cortical development in response to noxious stimuli. Conducting such a study in a clinical population is challenging and they have shown creativity in doing so. The conclusions though need to be tempered. We should look forward to further research into brain activity of preterm neonates who undergo many painful events, with tighter samples but more areas of the brain examined. Further work on the relationship of brain activity and visible responses should also be included in future studies.
K. J. S. (Sunny) Anand, Le Bonheur Children's Hospital
Note: This comment is an edited version of text that was originally posted 19 Sep and subsequently amended at the request of the moderator.
I am happy to comment on the attached paper and the Supplementary Online material, which recently appeared online in Current Biology. On the basis of EEG recordings, this paper concludes that preterm babies don’t perceive pain until 35-37 weeks of gestation. I have serious misgivings about the paper, its methodology, results, and conclusions—which fly in the face of our daily clinical observations as well as a huge body of literature on neonatal pain. Here are my initial thoughts:
1. Neonates were studied at various times after birth, and age was “not considered in the analysis,” thereby ignoring the effects of postnatal brain maturation and experience-dependent plasticity. These are likely to be very different among subjects of different ages, ranging from 0-110 days. This analysis indicates a serious lack of appreciation for the momentous maturational changes occurring in the (preterm or term) neonatal brain just after birth.
2. I am also concerned about their reductionist approach. The EEG signals were obtained using scalp surface electrodes (12-20 electrodes), but the signals from only one electrode (CPz or Cz) were analyzed. These signals were broken down into epochs (variables) and time periods (observations); on this association matrix they performed a Principal Components Analysis (PCA). They took the first principal component of signals between 50-300 ms as the tactile potential, and the second principal component of signals between 300-700 ms as the nociception potential. They discarded all EEG data obtained from all the other EEG electrodes and across all the other time periods. Pain is a multilayered sensory and emotional experience widely distributed and processed in multiple brain areas—this reductionist analysis focusing on miniscule time epochs does not really represent a study of pain or nociception in the developing human brain.
3. Throughout the paper, the authors equate cortical EEG signals as representative of processing in the entire brain—this appears to me as a huge leap of faith! Pain and touch are perceived based on subcortical and brainstem processing, with a rich interplay of inputs from the limbic system, insula, thalamus, and hypothalamus. Based on cellular Fos expression in the infant rat, we had previously noted a shift from subcortical to cortical processing that occurs from P0 to P7, and is completed by P14. In terms of neurological maturity, this postnatal period in rats corresponds to the developmental maturation occurring from 28 to 37 weeks of human gestation.
4. Another error in their reasoning is that “perception” only occurs in the cortex…. This is a common misconception, which perhaps stems from the “cognitivism” era in neuropsychology. On the basis of functional MRI (fMRI) scans, my colleagues at the Karolinska Institute and others have shown the rich functional processing of both cortical and subcortical areas in the neonatal brain. The current study did not examine subcortical processing--absence of evidence should not constitute evidence for the absence of subcortical processing.
5. The EEG responses of preterm neonates are judged based on the framework of typical responses in term neonates, selecting those which correspond to similar responses in adults. However, I feel that the preterm cortical responses should be analyzed from the viewpoint that they are unique and are not correlated with the pattern of responses in term neonates. Using this framework, what they define as tactile potentials or nociception-specific potentials were seen in variable numbers of preterm and term neonates—of the preterms, 7 percent show the tactile potentials and 33 percent the pain potentials; of the term babies, only 30 percent show the tactile potentials and 63 percent show the pain potentials. Thus, even though the EEG signatures associated with touch and pain “perception” were developed from term neonates, still they do not occur consistently in that population—and much less in the preterm population.
6. Neuronal bursts occurring in various cortical regions seem to be characteristic of the EEG patterns in preterm neonates. In this study, the occurrence of neuronal bursts was doubled by tactile stimulation and quadrupled by painful stimulation in the preterm neonates. Such increases in neuronal bursts with stimulation don’t seem to occur in the term neonates. Dismissing these increases in neuronal bursts as non-sensory just because similar EEG patterns appear during early brain development does not seem justified.
7. An alternative explanation of these data could be that the stimulation-induced changes in neuronal bursts represent the perceptual content of tactile or painful stimulation (perhaps related to subcortical processing), whereas these are inhibited in the term neonates, perhaps to enable more precise localization of these stimuli via cortical processing. We must be careful not to assign the values of “causation” to an “association” between stimulus and response.
8. Lastly, I’m not comfortable with the PCA and other statistical analyses. Factor analysis is a technique that amounts to a black box. I have used it reluctantly in the past, and then, almost invariably combined with another data reduction technique. Several errors occur frequently with principal components and other factor analyses. The first is giving each factor a biologically representative name. Then, the assumption is that the factor really conveys information about the arbitrary name. The second error is to fail to appreciate the loss of redundant information that results. Both errors are widespread according to scholarly descriptions of this technique (Ringner 2008; Shlens 2005; Hatcher and Stepanski 1994; Bro et al., 2008).
I sincerely hope that my opinions will be available to all the neonatologists, neonatal nurses, and other clinicians working daily with preterm neonates. There are immeasurable entities in this universe, one of which is the experiential content of individual consciousness—science will always fall short on this frontier; to take surface EEG signals as the representative of what the infant experienced is another example of the over-interpretation of measured data.
References:
What is principal component analysis?
Ringnér M. Nat Biotechnol. 2008 Mar;26(3):303-4.
A Tutorial on Principal Component Analysis
Jonathon Shlens.5 Dec 2005. Available at http://www.snl.salk.edu/~shlens/pca.pdf
A step-by-step approach to using the SAS system for univariate and multivariate statistics.
Hatcher, L. & Stepanski, E. (1994). Cary, NC: SAS Institute Inc.
Cross-validation of component models: a critical look at current methods.
Bro R, Kjeldahl K, Smilde AK, Kiers HA. Anal Bioanal Chem. 2008;390(5):1241-51.