Cerebral palsy (CP) is a broad term for disorders of the development of movement and posture, causing activity limitation, which are attributed to non-progressive disturbances in the developing fetal or infant brain (1, 2). CP is one of the most common causes of physical disability in children, and the majority of these children have impaired hand function that makes them experience difficulties in performing daily activities (3, 4). It occurs in 2–3:1,000–1:2,500 live births (5, 6). Depending on the timing of the lesion during fetal development, different types of lesions occur; cortical maldevelopments (first and second trimester of gestation), periventricular white matter (PVL) lesions (early third trimester) or cortical and subcortical lesions/ middle cerebral artery infarctions (MCA) (around term age).

The compensatory motor capabilities following early focal brain injury are intriguing, and these are reported to be superior to those of the adult brain. This observation, known as the Kennard principle, is based on a study of recovery following experimental lesions of the motor cortex in monkeys (7–9). Both lesion timing and cortical spinal tract (CST) wiring patterns have been shown to relate to upper limb function in children with unilateral CP (10). However, a large part of the variability in upper limb function still remains unexplained (7, 11).

Aside from motor impairment, somatosensory impairment is also observed in children with CP. Somatosensory impairment is a broad term used for tactile deficits as well as for impairments in the processing of sensory information such as vibration, stereognosis, and two-point discrimination. In addition, proprioception can be considered one of the subsystems within the somatosensory system (12, 13). Proprioception consists of kinesthesia and joint position sense. Kinesthesia is the sense of extremity movement without visual input, and position sense is characterized by static limb position (14).

Thalamocortical projections start to reach the somatosensory cortex at the beginning of the third trimester. Conversely to CST wiring, developing thalamocortical somatosensory projections can still bypass even large periventricular brain lesions during this period (10). This tends to lead to sprouting to a broader area in the somatosensory cortex (15). Somatosensory processing is located in the primary somatosensory cortex (S1) in the postcentral gyrus of the parietal lobe and the secondary somatosensory cortex, located on the parietal operculum (16, 17). Median nerve stimulation, as well as object recognition, have been shown to activate the secondary somatosensory cortex (S2) bilaterally, regardless of the hand being stimulated, but only the contralateral S1 (18, 19). According to Auld et al., children with UCP performed worse in sensory tasks with their impaired hand compared to the unimpaired hand. However both hands performed worse than either hand of typically developing children. Over 75% of children with UCP have tactile deficits (20).

Tactile deficits account for approximately 30% of the variance in upper-limb motor function in children with UCP (21). One example of this is the necessary sensory feedback in the modulation of fine motor tasks such as precision grip (15, 22). The majority of existing studies focus on especially motor performance, and there is only limited information about the involvement of the sensory system on functional outcome. However, understanding the extent and impact of sensory function on upper limb motor function is essential to improve rehabilitation approaches and functional outcomes (23).

In this systematic review we focus on all children and young adults with CP. However, experience shows that most studies on sensorimotor function concentrate on children and young adults with unilateral CP. Therefore, in the current study, we aimed to synthesize information on the consequences of brain damage (white matter characteristics, brain lesion types, functional connectivity) on somatosensory impairment and its impact on upper limb function in children and young adults with CP.

Based on the current literature, somatosensory impairment seems to be an important factor in upper limb dysfunction in children and young adults with CP (23). Therefore, we hypothesize a potential relation between neuroanatomical lesions, functional connectivity, and somatosensory impairment.

A systematic review was designed. The protocol has been registered in the National Institute for Health Research (NHS) on PROSPERO (International Prospective Register of Systematic Reviews) database: ID 342570.

A literature search was performed in six online databases: PubMed; Cochrane; Web of Science; Embase; CINAHL, and PEDro, from inception to March 13, 2021. Each search contained three main concepts: CP, MRI (magnetic resonance imaging), and sensation. The following Medical Subject Headings (MESH) terms and text words were used; ((((((Sensation) OR “Somatosensory Disorders”[Mesh]) OR “Sensation”[Mesh]) OR Sensory) OR Sensory function)) AND ((Cerebral palsy) OR Cerebral palsy [MeSH])) AND (((MRI) OR “Magnetic Resonance Imaging”[Mesh]) OR Magnetic Resonance Imaging). All results were uploaded to Rayyan Systems inc. (https://www.rayyan.ai/), an online tool to screen and select articles by multiple reviewers. Indicated duplicates were removed after checking by one of the first authors (CS). First, two review authors (CS, RV) independently screened titles and abstracts to remove irrelevant articles. Dissertation articles and conference abstracts were removed. Second, both authors reviewed the full text of potentially relevant studies to determine their eligibility. Consensus was reached on all articles.

The following inclusion criteria were applied: (1) human participants with spastic CP; (2) mean age of the participants was not older than thirty years of age, since the focus of the study included children and young adults; (3) MRI imaging available; (4) assessment of somatosensory function; (5) published in English, Dutch, French or German; (6) original research papers (exclusion of study protocols, reviews and conference abstracts).

Our study objective was to investigate somatosensory deficits in relation to specific MRI abnormalities in children and young adults with spastic CP. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were followed to extract the articles included in the review (24).

Data extracted included (1) first author and year of publication; (2) number and age of participants; (3) type of cerebral lesion; (4) clinical assessment of participants (motor and/or somatosensory assessment or categorization according to the Surveillance of Cerebral Palsy in Europe); (6) whether a control group was involved; (7) additional neuroimaging/neurophysiological method (e.g., electroencephalography, transcranial magnetic stimulation, somatosensory evoked potential, functional magnetic resonance imaging) (8) outcome of the studies.

Study quality was assessed (YJ) using the Standard quality assessment criteria for evaluating primary research papers from a variety of fields (25).

After removing duplicates, the database search (last updated on March 13, 2021) yielded 573 citations. Twenty-two records were eventually included in the review. Figure 1 displays the flowchart of the study selection process according to the PRISMA guidelines (24).

The extracted information from the included articles is summarized in Table 1.

These twenty-two articles reported a total of 905 observations, 388 with a PVL-type lesion, 163 with an MCA-type lesion, 326 typically developing children (TD), and 25 with other lesions (Table 1). Most studies included children and young adults with a clinical unilateral cerebral palsy, except for the study of Wingert et al., which analyzed children with spastic diplegia, and Hoon et al., which analyzed children with mainly spastic diplegia (21/28) and also quadriplegia (4/28), hemiplegia (2/28) and ataxic CP (1/28) (30, 44). Next to MRI, five additional neuro-imaging/neurophysiological techniques were used. Sensory function was assessed in all studies. A combination of thirteen sensory assessments, i.e., tests, protocols as well as evaluation criteria, were used (see Table 1). For instance, stereognosis assessment protocols differed. Some studies used identification of six out of twelve familiar objects; three of six objects matched in pairs (pencil/pen, coin/button, paperclip/safety pin), and three of six differing objects (key, clothespin, marble, comb, spoon, ball) (7, 35, 42, 43). Whereas others used the number of correct responses out of a possible maximum of ten (28), nine (27, 40), five (29), six (37), or three objects(nickel, key, paperclip) (31–33). Two other studies did not include a detailed description of the stereognosis assessment (38, 39).

In almost all studies, motor function was reported, though not all studies reported on GMFCS or MACS. Some of the papers are from the same research groups (26, 31–33, 35, 37, 41–43, 45), potentially resulting in 312 overlapping observations. The results of the reports will be discussed according to sensory function.

Using the standard quality assessment criteria, most studies were of strong quality, with a total score above 0.80; one study was rated adequate and one good (25). The quality assessment is shown in Table 2.

Tactile registration has been studied to a lesser extent. Hence the classification into “lesion characteristics” and “functional connectivity” has been discarded here. Feys et al. reported normal touch sensation in all children with PVL-type lesions, compared to approximately 75% of the children with cortical-subcortical brain lesions (7). Touch sensation could be significantly predicted by lesion timing, lesion location, and lesion extent (37, 38). In patients with PVL-type lesions, injury severity in thalamocortical pathways is related to sensation of touch (30, 37).

Reorganization of the sensory system after early brain lesions is a complex and intriguing process. Although information on reorganization of sensory functions in children and young adults with CP is increasing, this reorganization process is still not fully understood. Understanding the pathophysiology of this reorganization process, and its relationship to sensory outcomes, and its relationship to its impact on functional outcomes might ultimately lead to different rehabilitation strategies, as shown schematically in Figure 2.

However, a comprehensive comparison of the evidence in the literature on sensory function in relation to anatomical lesions is difficult because different test batteries, sensory test protocols, and evaluation criteria are used, as well as different outcome measures for neuro-imaging, as shown in Table 1. Moreover, when the available sensory information is associated with lesion timing, lesion location, lesion extent, and functional connectivity, studies tend to focus on one particular tract or lesion type. In addition, we need to take into account that currently used protocols for sensory assessment most likely underestimate the sensory deficits in patients with CP (22). Therefore, international consensus on comprehensive sensory test batteries, protocols, and evaluating criteria is necessary to allow comparison of somatosensory function in relation to brain injury characteristics and ultimately influence functional outcome with personalized rehabilitation strategies.

Most of the literature on sensorimotor function in CP focuses on children and young adults with unilateral CP. The results of our search confirmed this; while the intent was to include all children with CP, most studies included patients with clinical unilateral cerebral palsy and mild impairments. The study of Wingert et al., and Hoon et al., included a large/solely group of participants with spastic diplegia. The results of these studies were included in this systematic review because they were in our intended patient category. Even in the unilateral cerebral palsy groups, especially in case of white matter damage, the abnormalities were often bilateral, meaning even though the patients have a clinical unilateral CP the groups are more similar in lesion type than clinically expected. Further, the results of the study of Hoon et al. are in line with the study of Mailleux et al. and the study of Wingert et al. uses a different study protocol (grating discrimination), making the risk of bias low (30, 35, 44).

There is substantial evidence that patients with PVL-type lesions have significantly less sensory deficits as opposed to patients with cortical-subcortical/MCA-type lesions (7, 27, 28, 31, 32, 34, 37, 38, 41). Lesion extent, type of CST wiring pattern, and lesion location significantly further impact sensory function (7, 37).

In the next paragraphs, we will elaborate on the relationship between specific neuroanatomical abnormalities and specific sensory deficits and the potential consequence for rehabilitation in more detail.

Damage to cortical and subcortical structures reduces the likelihood of the CST trajectory developing in the typical contralateral pathway. In case of ipsilateral CST wiring, the association between sensory and motor functions is disrupted. This points toward a different mechanism of sensorimotor integration in patients with CP as a probable cause for the association between CST wiring and increased sensory deficits (39). The relation between CST wiring and stereognosis is found across multiple studies (28, 35, 37). Some studies found a correlation between CST wiring and poorer performance of the 2PD test, while others did not find this correlation (35, 37). However, there was a large spread of 2PD test scores in this group, which possibly explains the lack of a statistical significant difference (28).

Ipsilesional reorganization of the S1 area appears to be the primary compensation mechanism after a unilateral early brain lesion, regardless of the timing of the lesion. Developing thalamocortical somatosensory projections can still bypass even large periventricular brain lesions during the third trimester (10). This tends to lead to sprouting to a broader area in the somatosensory cortex, and the wider distances correlate with impaired sensory function (15). Somatosensory deficits in patients with PVL-type lesions are better explained by loss of integrity of these thalamocortical pathways than by loss of grey matter volume in ipsilesional S1 and/or S2. The loss of grey matter volume in S1 and S2 in these patients was not related to sensory outcome (36). In patients with MCA-type lesions and subsequent volume loss of the grey matter in S1 and S2, the reorganization capabilities of these thalamocortical pathways are less, resulting in a more severe deficit. The observed inter-hemispheric reorganization of S2 (34) could be related to the observed bilateral activation of S2 after tactile stimulation, seen in healthy volunteers (16, 19). However, high contralesional activation was associated with a severe impairment of sensory function, making this compensatory mechanism inadequate to say the least (34).

Diffusion tractography shows a positive correlation between the ascending sensory tract (AST) axial diffusivity (AD) of the more affected hemisphere and sensory test outcomes. Increased axial diffusivity may indicate gliosis and structural abnormalities in the integrity of the AST (15, 33, 40). These differences are more extensive in patients with MCA-type lesions compared to patients with PVL-type lesions, implying that damage to the ascending sensory tracts is more extensive in patients with MCA-type lesions (33). This is consistent with the finding that patients with MCA-type lesions have more severe sensory deficits. Posterior thalamic radiation injury also correlates with sensory impairment (30, 35).

Cortico-cerebellar circuits, measured using functional MRI (fMRI), were well preserved in almost all patients. Despite this well-preserved cortico-cerebellar circuitry, no correlations were found with sensory deficits. Thus, this intact circuitry did not compensate for sensory deficits (26, 41, 43).

When assessing proprioceptive deficits in relation to lesion types, this relationship is less clear. In the study of Feys et al. no significant differences in proprioception between lesion types were found (7). In contrast, Kuczynski et al. found deficits in position sense to be more common and also more severe in children with MCA-type lesions (31, 32). Guzetta et al. included only one patient with an impaired position sense; this patient had the most severe sensory impairment (29). It should be noted that in all studies a substantial number of patients with normal proprioception were included, which potentially caused a bias in the conclusions (7, 29, 31–33, 42, 43). Another potential explanation lies in the assessment itself, since sensory deficits are most likely to be underestimated in patients with CP (22).

Functional connectivity between the non-lesioned S1 and thalamus/SMA in patients with cerebral palsy is inversely correlated to position sense; higher functional connectivity is associated with better performance. Whereas in typically developing children, position sense is positively correlated with connectivity between the thalamus and bilateral sensorimotor regions; increased connectivity is associated with poorer performance. Overall, the thalamus showed decreased connectivity in children with PVL-type lesions compared to controls, suggesting that early lesions can disrupt sensory network components, and connectivity between these areas is related to tactile perception deficits (45).

There is limited evidence of interhemispheric reorganization of the somatosensory functions, only two studies reported on patients with an interhemispheric reorganization, and these patients had the most severe sensory deficit. So when an interhemispheric reorganization is observed, this reorganization does not lead to improvement of sensory function (29, 34).

In recent years, rehabilitation programs have paid more attention to enhancing sensory functions during CIMT/HABIT(ILE) programs (23, 46–49) and in study protocols (50–52). As in our systematic review, these studies show a large variability in study design, patient characteristics, and sensory assessment methods used. This makes it difficult to compare the findings. No distinction was made based on lesion type across these studies/ in the study protocols. However, all studies found an improvement in one or more sensory domains, making it worthwhile to explore these differences and the effect of lesion characteristics on the ability to achieve these differences (23, 46–49). In a study on adult stroke patients, different altered patterns of cortical activation were observed following touch discrimination training when patients with thalamic/capsular lesions were compared with patients with S1/S2 cortical somatosensory lesions. These changes were different despite common training and similar improvement (53). If the ability to change cortical activation in relation to lesion type and in relation to sensory improvement could be unraveled, a foundation could be laid for a more individualized training program.

There are several limitations to be considered. Due to the large variability in study design, patient characteristics, neuroimaging/neurophysiological techniques and outcome parameters, and sensory assessment methods used, only a partial synthesis of evidence was possible. In addition, some of the papers included in this systematic review used the same study population. Ten of the twenty-two articles included described at least one patient over eighteen. Not wanting to exclude a large portion of the studies and thus missing essential observations, led to the selection of papers including both children and young adults in this systematic review. The populations are by nature small and often heterogeneous. Several papers included children with relatively minor sensory deficits, which potentially caused a bias in the conclusions (29, 42, 43).

When reviewing the literature, another paper was discovered, which should have been included in the original search. This paper by Chu et al. researched the reorganization of hand somatosensory cortex in children with CP (54). Fortunately, we did not miss any essential information because this study shows similar results to the papers included in the original search. Key words in the paper of Chu were “perinatal brain injury” as opposed to “cerebral palsy” used in our original search terms. This might be the reason this paper was not included in the original search.

As we included only original research papers, the reviews of Brun et al. (12) and Poitras et al. (13), in which somatosensory deficits in children with cerebral palsy were discussed, were not selected for this review. Although neural correlates were mentioned, the main focus in these reviews was on different sensory domains. The review of Garberova et al. focused on somatosensory function in relation to fMRI, (functional magnetic resonance imaging), disregarding other modalities (55). This makes our review complementary to these reviews.

In conclusion, it is hard to draw definite inferences on the relationship between the reorganization of the sensory network following early brain damage and sensory function in children with CP because of the large variability in study design, patient characteristics, neuroimaging/neurophysiological techniques, and parameters and sensory assessment methods used. In general, lesion timing, lesion location, lesion extent, integrity of the ascending sensory tract, and structural abnormality of the somatosensory areas have an impact on sensory function. In line with these observations, patients with cortical MCA lesions have more severe sensory deficits across all sensory modalities as opposed to patients with white matter (PVL) lesions. Intrahemisferic reorganization is the most common type of reorganization of the sensory system. In addition, an interhemisferic reorganization of the sensory system was associated with poor sensory function. International consensus on a clinically relevant sensory test battery is needed to enhance understanding of the intriguing compensatory mechanisms of sensory network following early brain damage and potential consequences for rehabilitation approaches.