The effects of sensory re-education on hand function recovery after peripheral nerve repair: A systematic review

2.1 Search strategies

This systematic review was limited to articles published from 1970 to 23 December 2020. We performed an electronic search in CINAHL, Embase, PubMed, Web of Science, and Medline databases using the following keyword combinations: (sensory re-educati^*) or (sensory reeducat^*) or (sensory re-learn^*) or (sensory relearn^*) AND (peripheral nerve) or (median nerve) or (ulnar nerve) or (radial nerve) AND (nerve repair) or (nerve injur^*) or (nerve surger^*).

We included studies that met all of the following criteria: (1) randomized or quasi-randomized control trials (RCT or quasi-RCT); (2) adult patients after upper-limb peripheral nerve repair; (3) use of SR and sensory re-learning training programs as an intervention; and (4) published in English. We excluded studies that met any of the following criteria: (1) no clinical measurements related to hand function, (2) sample size less than 10, and (3) conference abstracts.

First, two reviewers independently selected all records based on titles and abstracts to exclude studies obviously irrelevant to our topic. Second, full texts of the remaining articles were reviewed for the eligibility criteria. Then, the two reviewers had a face-to-face discussion to reach an agreement on study inclusion, and any discrepancies were resolved by a third reviewer.

2.2 Assessment of methodological quality

Two reviewers independently assessed the methodological quality of the included studies using the Physiotherapy Evidence Database (PEDro) scale (Maher, Catherine, Herbert, Moseley & Mark, 2003). This scale contains 11 items and each item is judged by “yes” or “no” responses. The total number of “yes” responses for all items, except item one, is the final PEDro score, which ranged from 0 to 10 points. Studies with a final score lower than 4 points are described as having poor quality; 4–5 points as fair quality; 6–8 points as good quality; and 9–10 points as excellent quality. Another face-to-face discussion was carried out to achieve an agreement on quality assessment and any discrepancies were resolved by a third reviewer.

2.3 Data extraction and synthesis

Two reviewers independently conducted data ex-traction. A customized form was produced for data extraction regarding the general characteristics and results of the included studies. The extracted information consisted of authors, study design, characteristics of patients, experimental interventions (starting time, frequency and duration of training), outcome measures, and results. To perform a comprehensive qualitative synthesis, we considered both the consistency of results across included studies and the heterogeneity in methodological quality and sample size.

3.1 Literature search and study characteristics

The flowchart of study selection is presented in Fig. 1. The initial search identified 241 records, from which 16 articles were finally included in the current review. Nine of these articles were RCTs (Cheng, 2000; Hassan-Zadeh, Lajevardi, Esfahani & Kamali, 2009; Lundborg, Bjorkman & Rosen, 2007; Paula et al., 2016; Rosén, Björkman & Lundborg, 2006; Rosen & Lundborg, 2007; Rosen et al., 2015; Saberi et al., 2018; Vikström, Rosén, Carlsson & Björkman, 2018) and seven were nonrandomized controlled trials (Antonopoulos et al., 2019; Imai, Tajima & Natsuma, 1989, 1991; Mavrogenis, Spyridonos, Antonopoulos, Soucacos & Papagelopoulos, 2009; Mendes et al., 2018; Priya, 2012; Shieh et al., 1995). In the included articles three pairs of articles by Imai et al. (1989, 1991), Lundborg et al. (2007) and Rosén et al. (2006), Rosen et al. (2015) and Vikström, Rosén, et al. (2018), respectively, were reported based on the same dataset of three trials. Considering that pairs of articles of each trial focused on different time effects or outcomes, we assessed all of these articles for quality. Characteristics of the 16 studies are presented in detail in Table 1. The individual sample size ranged from 13 to 52 and a total of 374 adults with median nerve injury, ulnar nerve injury, or both at the digital, wrist, distal forearm, or forearm level were included.

OPEN IN VIEWER

Fig. 1

PRISMA flow diagram of studies through the review.

The result of methodological quality assessment is presented in Table 2. The PEDro scores of all studies ranged from 2 to 6 points, indicating that four were good-quality studies (Lundborg et al., 2007; Rosén et al., 2006; Rosen et al., 2015; Vikström, Rosén, et al., 2018) and five were fair-quality studies (Hassan-Zadeh et al., 2009; Mendes et al., 2018; Paula et al., 2016; Rosen & Lundborg, 2007; Saberi et al., 2018). A poor-quality rating was given to the remaining seven studies (Antonopoulos et al., 2019; Cheng, 2000; Imai et al., 1989, 1991; Mavrogenis et al., 2009; Priya, 2012; Shieh et al., 1995).

Interventions used in the studies varied in modality and starting time and were divided into four categories.

4.1 Underlying mechanism

Zink and Benjamin (2020) identified four types of cortical plastic changes that occur after upper extremity peripheral nerve injury, they are “Unmasking or removal of inhibitory controls in the affected region”, “Axonal sprouting and misdirection of regenerating peripheral nerves”, “Alteration of bilateral somatosensory and supplementary motor cortex maps”, “Invasion and expansion of adjacent cortical maps of affected nerve.” The therapeutic possibilities of SR were provided by this brain plasticity (Duffau, 2006; Rosén & Lundborg, 2004; Vikström, Carlsson, Rosén & Björkman, 2018). In SR, different techniques would be used in different phases to guide brain plasticity to support and enhance peripheral functions. In the early phase, cortical reorganization occurs in the somatosensory cortex after nerve transection immediately, and the region previously innervated by the injured nerve is occupied by territories of adjacent normal body parts (Lundborg, 2000; Rosén & Lundborg, 2004). In this phase, an early SR program uses alternative senses, such as vision or hearing, as a substitute for the loss of tactile stimuli to activate and maintain the hand map at the cortical area (Jerosch-Herold, 2011; Vikström, Carlsson, et al., 2018).

The mechanism underlying these substitutional stimuli is based on the hypothesis of cross-modal plasticity of the brain, which possibly creates an interaction between senses or sensory substitution (Bavelier & Neville, 2002; Mendes et al., 2013). According to the theory of long-term potentiation, applying consecutive pairings of cross-modal stimuli can strengthen the dendritic branching increments, which consequently increase the quantity of synaptic buttons and reduce neurotransmitter recapture and intracellular cascade facilitation. As a result, the relationship between nodes of a multi-scale network is promoted, and learning and adaptive behaviors are improved (Mendes et al., 2013).

In the studies reported by Mendes et al. (2018) and Lundborg et al. (2005) researchers used functional magnetic resonance imaging to investigate early cortical responses after a one-week early SR training with SGS. They found that an effective connection was established between the cortical hand map (Brodmann 3 area) and associative areas (auditory area), which proved that the use of SGS helped promote audio-tactile interaction and successfully built cross-modal plasticity.

It has been reported that MVF, which also is used in early SR, is a more effective method to induce higher activation in motor neurons (i.e., the “mirror neurons” in the sensorimotor cortex and supplementary motor area). The mirror neuron system plays a key role in motor action understanding and imitation by listening or reading action words and observing the MVF of action (Bai, Fong, Zhang & Hu, 2019; Svens & Rosén, 2009). Thus, researchers hypothesized that “sensory” observing or listening would active the somatosensory cortex, and they applied sensory MVF in early SR training (Svens & Rosén, 2009).

In the second phase, axons regenerate and the hand is re-innervated. Because of the misdirection of axonal growth, the cortical reorganization is unstructured (Rosén & Lundborg, 2004) and irreversible. The classical SR program, which inputs sensory stimuli to the injured nerve territory with visual feedback, memory, and learning, could be applied to help patients learn this “new language” spoken by the brain to adapt to the new synaptic organization. The cut-off time point between the two phases or the time point at which early and classical SR should be applied, however, is not specific. According to researchers who developed classical SR, SR training of the second phase commonly started once touch could be perceived by the hand (Dellon et al., 1971; Parry & Salter, 1976). In recent years, researchers suggested that the cut-off time was the time when some re-growing axons had reached the end of the hand (Rosen et al., 2015; Svens & Rosén, 2009). It is known that the sequence of sensory recovery is first the perception of pain and temperature that relates to unmyelinated and thinly myelinated fibers recovery, and then it is followed by the 30 Hz vibration, moving touch, static touch, and 256 Hz vibration that are related to large myelinated fibers recovery (Dellon, 1997c). Thus, a gap may exist between the time when axons re-grow to the hand and the time when touch is perceived by the hand. This nonidentical standard may explain why the starting time of interventions in the reviewed studies varied under these two principles. Nevertheless, by summarizing these studies, the cut-off time point may be identified between 3 to 4 months after nerve repair, if the injury is at the wrist level.

In summary, second phase (or classical) SR program and early phase SR with MVF were based on all four types of cortical plasticity mechanism that mentioned above. And early phase SR with SGS used cross-modal sensory substitution strategies related to the “Unmasking or removal of inhibitory controls in the affected region” mechanism. This was also proved by Zink & Benjamin (2020) in their current review.

4.2 Treatment influence factors

The treatment response to SR correlates with many factors, such as age, peripheral nerve status, and cognitive capacity (Miller et al., 2012; Rosén, Lundborg, Dahlin, Holmberg & Karlson, 1994; Vikström, Carlsson, et al., 2018). Previous studies have shown that 54% of the variance can be explained by the first two factors, whereas the specific cognitive capacities (verbal learning and visuo-spatial capacity) explained about 15% of the variation (Rosén et al., 1994). SR training program includes abstract components that is challenging to the cognitive capacities of trainers, especially, in early SR program. Generally, most trials included in this review had collected participants’ data on age and limited the criteria of eligibility in the peripheral nerve status, such as type of injury, timing of repair, and surgical technique. None of the trials, however, had considered or collected the data on the education level or cognitive capacities of patients who were included at baseline to determine whether the groups were equivalent.

A qualitative methodology study (Vikström, Carlsson, et al., 2018) was done to investigate the pers-pective of patients on early SR training, and three factors, explaining 45% of the variance, had been identified. Two of them, explaining 19% of the variance, were factors that descripted as “Do not get an illusion of touch easily and need support in their SR” or “Are not motivated, manage to get an illusion of touch but do not complete the sensory relearning”. This might cause some patients did not adhere to their long-term SR training.

As far as the methodological quality of included studies is concerned, the high attrition rate weakens the validity of findings. In 11 out of the 16 reviewed studies, more than 15% of participants did not undergo measurement of a key outcome at one of those time points when the outcomes were assessed. This finding meant that most of the trials could not obtain an adequate follow-up and some participants had dropped out of the studies. High drop-out rates decrease the statistical power, especially when there is a differential pattern of attrition between groups (Mahaniah & Rao, 2004). A significant difference between groups after treatment could be due to the bias related to attrition, not necessarily due to the effectiveness of treatment. A clear explanation of requirements to subjects and blinding of all subjects can minimize this attrition. Moreover, researchers should consider the drop-out rate when calculating the sample size. With such a drop-out rate, however, most of the reviewed studies did not clearly describe whether the subjects received treatment or control conditions as allocated, and also did not report about the use of intention to treat (ITT) analysis. The CONSORT statement (Schulz, Altman & David, 2011) recommends the use of ITT and flow charts to record in detail how subjects are included or excluded in each phase of the trial. Only Rosen et al. (2015) clearly pointed out that ITT was used in their research, and their participant follow-up chart was exhaustive.

4.3 Practice recommendation

The latest research by Zink and Benjamin (2020) highly recommended Classical SR in occupational therapy, and early SR with MVF or SGS were recommended for secondary or situationally use. However, all above recommendations are based on mechanism without considering the interventions’ efficacy. Our current study focus on effects of SR could be a supplement for this limitation. We find two RCTs with good quality reported the effects of early SR with using MVF (Rosen et al., 2015; Vikström, Rosén, et al., 2018). Therefore, applying early SR with MVF can be an effective intervention for people with sensibility dysfunction after peripheral nerve injury. Successful SR program should follow several principles, that is, 1) well-informed participant with activity and motivation, 2) and a long-term, repeated training sessions (Rosen et al., 2015; Vikström, Carlsson, et al., 2018). In detail, therapists or clinicians who conduct SR should make sure that patients are well educated with the concept of SR and fully understand it, so that, they are able to get an illusion of touch and complete the whole SR program (Vikström, Carlsson, et al., 2018).

One of the strengths of this review is that it focused not only on the evaluation of the effect of SR in functional sensibility but also on the whole related hand function. Therefore, we focused on the study of peripheral nerve injuries at upper extremity in adults. The outcome measurements of studies were not limited to sensibility and functional sensibility, and self-reported questionnaires of activities and participation also were applied. Another strength of this review is that the PEDro scale, which was used during the methodological quality assessment, was specially developed to identify an RCT or quasi-RCT if it seemed to be externally and internally valid and had sufficient statistical information to show the results. The limitation of this review is that various treatment techniques of SR made it difficult to perform a quantitative synthesis to investigate every SR technique.