Influence of vibrotactile random noise on the smoothness of the grasp movement in patients with chemotherapy-induced peripheral neuropathy

Patients with chemotherapy-induced peripheral neuropathy (CIPN) often suffer from sensorimotor dysfunction of the distal portion of the extremities (e.g., loss of somatosensory sensation, numbness/tingling, difficulty typing on a keyboard, or difficulty buttoning/unbuttoning a shirt). The present study aimed to reveal the effects of subthreshold vibrotactile random noise stimulation on sensorimotor dysfunction in CIPN patients without exacerbating symptoms. Twenty-five patients with CIPN and 28 age-matched healthy adults participated in this study. To reveal the effects of subthreshold vibrotactile random noise stimulation on sensorimotor function, participants were asked to perform a tactile detection task and a grasp movement task during random noise stimulation delivered to the volar and dorsal wrist. We set three intensity conditions of the vibrotactile random noise: 0, 60, and 120% of the sensory threshold (Noise 0%, Noise 60%, and Noise 120% conditions). In the tactile detection task, a Semmes–Weinstein monofilament was applied to the volar surface of the tip of the index finger using standard testing measures. In the grasp movement task, the distance between the thumb and index finger was recorded while the participant attempted to grasp a target object, and the smoothness of the grasp movement was quantified by calculating normalized jerk in each experimental condition. The experimental data were compared using two-way repeated-measures analyses of variance with two factors: experimental condition (Noise 0, 60, 120%) × group (Healthy controls, CIPN patients). The tactile detection threshold and the smoothness of the grasp movement were only improved in the Noise 60% condition without exacerbating numbness/tingling in CIPN patients and healthy controls. The current study suggested that the development of treatment devices using stochastic resonance can improve sensorimotor function for CIPN patients.


Background
Cancer is a major contributor to the global disease burden. The number of long-term cancer survivors is increasing because of improvements in early diagnosis and treatment (Siegel et al. 2019). Mitigating disability associated with the long-term effects of cancer treatment is important for improving quality of life (QOL) in cancer survivors. Chemotherapy-induced peripheral neuropathy (CIPN) is one of the most common complications of cancer treatments, with an incidence of 30-70% in cancer patients receiving chemotherapy (Cavaletti and Marmiroli 2010). CIPN typically persists beyond several years after initiating chemotherapy (Tanabe et al. 2013). Most patients with CIPN suffer from somatosensory symptoms (e.g., loss of somatosensory, numbness/tingling, and neuropathic pain) which spread to distal areas, such as the hand and foot (i.e., the so-called Communicated by Bill J Yates. "the stocking-glove" pattern), rather than proximal areas. These somatosensory symptoms can directly affect precise finger movements in daily living (e.g., typing on a keyboard, undoing and doing up buttons) (Gentilucci et al. 1997). In a previous study, we kinematically quantified precise finger movements using a three-dimensional measurement system and revealed impaired smoothness of thumb-index grasp movements in CIPN patients (Osumi et al. 2019). Several studies reported that exercise programs including resistance training, balance training, and cardiovascular exercise provided benefits for patients suffering from CIPN (McCrary et al. 2019;Andersen Hammond et al. 2020). Although simple exercise programs have been found to improve disability and QOL, they have demonstrated limited efficacy on somatosensory dysfunction derived from CIPN (Bland et al. 2019). In addition, there is little evidence regarding the efficacy of physical and occupational therapies for precise finger movements, although these are recommended for improving disability and QOL (Kim et al. 2015). Therefore, novel treatment strategies may be useful for improving somatosensory dysfunction and subsequent precise finger movements in patients with CIPN.
Precise finger movements result from feedback loops that operate simultaneously. For example, adaptable visuo-motor and somatosensory-motor loops have been reported to operate in the somatosensory and motor cortices and parietal cortex (Körding and Wolpert 2006;Todorov 2004). Therefore, impairment of somatosensory processing disturbs the feedback control of finger movement. From a rehabilitation perspective, complementary somatosensory input may improve grasping movements. Examining the influence of artificial sensory feedback in upper extremity amputees may be useful for elucidating the effects of somatosensory input on grasping movements. Previous studies reported that somatosensory feedback with electrical stimulation using implanted peripheral nerve cuff electrodes not only evoked artificial touch sensation but also improved the ability to manipulate delicate objects with a prosthesis (Tan et al. 2014;Ortiz-Catalan et al. 2014). Additionally, even transcutaneous electrical nerve stimulation enhanced tactile sensation and improved prosthesis control (D'Anna et al. 2017). The effects of somatosensory stimulation on motor control have been reported in various diseases. For example, suprathreshold electrical stimulation in the affected upper extremity has been reported to improve paralysis (Conforto et al. 2018). Considering such suprathreshold somatosensory stimulation causes discomfort and a tingling sensation (Golaszewski et al. 2004), suprathreshold somatosensory stimulation might exacerbate numbness and allodynia in CIPN patients. To compensate for such adverse effects, we alternatively focused on subthreshold somatosensory stimulation to improve somatosensory function, which is known as the stochastic resonance phenomenon. Stochastic resonance is a phenomenon in which the presence of a small amount of noise in a non-linear system improves sensitivity toward an external signal compared with when there is no noise (McDonnell and Ward 2011). Adding subthreshold noise to an input signal can cause the resultant signal to surpass the threshold of a given neuron or neural circuit and provide useful information for the central nervous system (McDonnell and Ward 2011). However, adding an excessively large noise signal (suprathreshold noise) would be expected to interfere with the non-linear threshold detection system (McDonnell and Ward 2011). The concept of stochastic resonance, in which the addition of subthreshold noise improves signal detection and feedback control system performance, has been demonstrated theoretically (Ma et al. 2013) as well as in a range of biological systems (Wiesenfeld and Moss 1995;Collins et al. 1996;Moss et al. 2004;Fertonani et al. 2011). For example, subthreshold vibrotactile noise directly applied to the index fingertip has been shown to immediately improve fingertip tactile sensation in stroke survivors (Liu et al. 2002) and healthy adults (Kurita et al. 2013). Clinical application of subthreshold random noise stimulation was reported to improve not only somatosensory sensitivity in older adults and patients with diabetic peripheral neuropathy or stroke (Collins 2004;Wells et al. 2005;Enders et al. 2013;Lakshminarayanan et al. 2015;Seo et al. 2019b), but also to improve hand manual dexterity in stroke patients using the Nine Hole Peg Test and the Box and Block Test (Seo et al. 2014). The present study aimed to investigate the clinical usefulness of subthreshold random noise stimulation for treating impaired somatosensory function and precise movements in CIPN patients. We hypothesized that somatosensory function and grasping movements in CIPN patients would be impaired, and that subthreshold random noise stimulation would increase the somatosensory sensitivity of the fingertip, thus improving the smoothness of grasping movements in CIPN patients.

Participants
Twenty-five patients with CIPN (age, 64.8 ± 9.4, mean ± standard deviation; 15 females) and 28 agematched adults with no hand-related problems (age, 65.8 ± 9.5; 16 females) participated in this study. Most CIPN patients reported numbness, neuropathic pain, and difficulties with precise hand movements after receiving chemotherapy. CIPN patients included ten patients with breast cancer, three with lymphomas, two with lung cancer, two with prostate cancer, two with ovarian cancer, two with gastric cancer, one with pancreatic cancer, one with uterine cancer, one with esophageal cancer, and one 1 3 with biliary tract cancer. They had been treated with one or a combination of anti-neoplastic agents such as taxanes, platinum compounds, and vincristine, which are all known to be strongly neurotoxic. The CIPN disease duration of those in the CIPN group was 27.8 ± 19.2 months. Although overall functional level was not assessed, all patients were independent in daily life. Clinical assessments of CIPN were conducted using part of the FACT/GOG NTX Ver.4.0 (https:// www. facit. org/ measu res/ FACT-GOG-NTX), the Patient Neurotoxicity Questionnaire (PNQ), the Japanese version of the Neuropathic Pain Symptom Inventory (NPSI) (Matsubayashi et al. 2015), and an 11-point numerical rating scale. The FACT/GOG-NTX Ver.4.0 measured symptoms of peripheral neuropathy, including sensory and motor problems, with a 5-point Likert scale (0 = Not at all, 1 = A little bit, 2 = Somewhat, 3 = Quite a bit, 4 = Very much). Among items of the FACT/GOG-NTX Ver.4.0, we used the NTX subscale comprising 11 items concerning neurotoxicity (NTX1, NTX2, NTX3, NTX4, NTX5, NTX 6, NTX 7, NTX 8, NTX 9, HI12, An6) (Calhoun et al. 2003;Huang et al. 2007). The PNQ includes two questionnaire items: one asking about sensory neurotoxicity and the other asking about motor neurotoxicity. Answers were coded 0-4, with a higher score indicating more severe CIPN. The NPSI is a self-administered questionnaire specifically designed to evaluate the different positive and negative symptoms of neuropathic pain. The NPSI comprises 10 items to identify the presence of respective neuropathic pain sensations on an 11-point (0-10) numerical scale, as well as two items to identify the duration of spontaneous pain and the frequency of pain attacks. To calculate the NPSI score, the total score of the former 10 items is typically calculated. In addition, pain intensity in the experimental hand was assessed before all experimental procedures using a numerical rating scale in which the patient was asked to grade the actual pain level experienced on a scale from 0 to 10 (0 = no pain, 10 = worst pain imaginable). The ethical review board of our institution approved this study (Approval Number: H28-42). We explained the protocol of this study to all participants, and obtained written informed consent. Patient characteristics are demonstrated in Table 1.

Vibrotactile random noise stimulation
Vibrotactile random noise stimulation was applied using two small actuators (length: 10 mm; width: 18 mm; height: 2 mm; Vibration Actuator Sprinter α; Nidec Seimitsu, Nagano, Japan) attached to the volar and dorsal wrist using adhesive tape. White noise signals that were low-pass filtered at 500 Hz drove the actuators, as described in a previous study (Enders et al. 2013). To confirm that the hand actually received random noise stimulation, we recorded the vibration signal of the hand using a vibration measuring instrument (Yubi-recorder, Tec Gihan, Kyoto, Japan). The results of the power spectral analysis confirmed that the hand received random noise stimuli, although the vibration signals were not perfectly uniform (Fig. 1  bottom). A digital amplifier (AP15d; FOSTEX, Tokyo, Japan) and digital electronic voltmeter were used to set the individual intensity of vibrotactile random stimulation. The experimenter turned the dial on the digital amplifier to gradually increase the intensity of vibrotactile noise stimulation. When the participant became able to detect the stimulation, the experimenter checked the displayed value of the digital electronic voltmeter and used the value as the somatosensory detection threshold (mean ± standard deviation: 280 ± 291 mV). The intensity levels of the vibrotactile random stimulation were set to 60% (Noise 60% condition) or 120% (Noise 120% condition) of the somatosensory detection threshold, which was defined at the beginning of this study. The intensity level used in the Noise 60% condition was previously reported to be a suitable intensity for stochastic resonance (Seo et al. 2014). In contrast, the intensity level in the Noise 120% condition was determined as the intensity to not cause stochastic resonance and evoked pain (i.e., allodynia). In the baseline condition, no tactile random noise stimulation was presented (Noise 0% condition). All participants performed a tactile detection task and a grasp movement task while receiving each level of vibrotactile stimulation intensity (Noise 0, 60, and 120% conditions), as described below. During the tactile detection task and grasp movement task described below, the noise stimulation lasted from the start to the end of each task in the Noise 60, and 120% conditions.  21.0 ± 6.2 2.8 ± 0.6 2.4 ± 0.8 1.6 ± 2.5 1.1 ± 1.7 0.7 ± 1.5 2.0 ± 2.3 4.7 ± 2.7 3.5 ± 3.1

Tactile detection task
Participants performed the tactile detection task using a Semmes-Weinstein monofilament (log of force: 1.65-6.65) based on standard testing measures (Bell-Krotoski et al. 1993). We applied each filament until it bent onto the volar surface of the index-fingertip at least five times, in an ascending fashion to minimize participants' fatigue. Starting with a 2.36 (0.02 g) filament as the baseline, the evaluation was continued using filaments of increasing weight until participants reported perceiving a touch, or until 6.65 (300 g) was reached. The detection tasks were randomized. We interpreted the tactile detection threshold as the force value of the filament weight in each experimental condition. The noise stimulation lasted from the start to the end of the tactile detection task in the Noise 60 and 120% conditions. The random stimulation was applied for approximately 5 min for each condition, and the inter-condition interval was approximately 3 min in duration.

Grasp movement task
The findings of our previous study clarified that only grasp movement was impaired in CIPN patients, whereas reaching movement was not impaired (Osumi et al. 2019). In the current study, we focused on grasp movement and used a device (UB-2, Maxell, Ltd. Tokyo) only for recording the aperture length between the thumb and index finger, and not the movement of the elbow and shoulder. The grasp movement task assessed the precise movements of the hand.
Participants were asked to grasp a target object (5 mm diameter, 35 mm height) located 30 cm away from the starting position ( Fig. 1) with the index finger and thumb. None of the participants grasped the object with the whole hand. Participants conducted this task at an arbitrary speed and repeated 10 trials in each Noise condition in a blocked fashion. The blocks were randomized. The random stimulation was applied for approximately 5 min for each condition, and the inter-condition interval was approximately 3 min. To record the grasp movements, magnetic sensors were attached to the tips of the thumb and the index finger, and the distance (i.e., aperture length) between the two sensors was extracted using a sampling rate of 100 Hz (UB-2, Maxell, Ltd. Tokyo) (Sano et al. 2016). Using the data for the aperture length, the smoothness of the grasping movements was quantified using jerk measurement. The jerk is mathematically defined as the third time derivative of the aperture length variable, and low values indicate smooth movement (Hogan and Sternad 2009). Importantly, the value of jerk depends on the movement length and duration (Hogan and Sternad 2009;van Kordelaar et al. 2014). For example, in cases in which the movement duration time increased because of movement freezing, the integrated squared jerk is dramatically lower despite such freezing movement not being smooth (Hogan et al. 2009;van Kordelaar et al. 2014). These previous findings suggest that the jerk value should be normalized with movement length (L) and movement duration (MD). A previous study recommended that the jerk value should be normalized with L 2 /MD 5 when quantifying the smoothness of grasping movement (Hogan and Sternad 2009;van Kordelaar et al. 2014). Therefore, grasp jerk needs to be normalized. Normalized jerk (NJ grasp ) was calculated using the following equations, which normalized the dependence on the movement duration (MD) and length (L). In this study, MD represents the time between the start and end of the movement. L grasp represented the difference in grasp aperture between the start and end of the movement. The

Statistical analysis
To verify our hypothesis that somatosensory function and grasping movements are impaired in CIPN patients and improved by subthreshold random noise stimulation, the following statistical procedures were conducted. First, to verify sensorimotor dysfunction in CIPN, patient's tactile detection threshold and log of NJ in the Noise 0% condition as the baseline were compared with those of age-matched controls.
In the comparison of the tactile detection threshold in the Noise 0% condition, the results of Levene's test indicated that the data were not normally distributed (p = 0.03). Thus, the Mann-Whitney U test was used for comparisons of the tactile detection threshold. In the comparison of the log of NJ in the Noise 0% condition, the results of Levene's test indicated that the data were normally distributed (p > 0.05). Thus, an independent t test was used for comparisons of the log of NJ. Second, to identify the effect of the vibrotactile random noise stimulation, the experimental data were compared using a two-way repeated-measures analysis of variance (ANOVA) for two factors including subject as a random variable: condition (Noise 0, 60, 120%) × group jerk 2 grasp (t)dt × MD 5 ∕L 2 grasp (health controls, CIPN patients). The Bonferroni method was used for post hoc testing. A difference was considered significant at p < 0.05. In addition, correlations between the effect of random noise stimulation on the tactile detection threshold and the log of NJ and CIPN clinical scores (FACT/ GOG NTX, PNQ sensory, PNQ motor, NPSI burning, NPSI pressing, NPSI paroxysmal, NPSI evoked, NPSI dysesthesia, and pain intensity NRS) were analyzed using Spearman's rank correlation coefficients. In the correlation analysis, the level of significance was set at p < 0.05.

Tactile detection threshold
The baseline tactile detection threshold in CIPN patients was higher than that in controls (U = 203.0, p = 0.009). No patients complained of worsening pain or numbness while undergoing each experimental condition. Two-way repeatedmeasures ANOVA revealed significant main effects of group (F = 20.55, p < 0.001) and condition (F = 6.44, p = 0.002). There was no significant interaction effect between group and intervention session (F = 0.40, p = 0.961). Post hoc tests revealed that the tactile detection threshold in controls in the Noise 60% condition was significantly lower than that in the Noise 0% condition (t = 3.69, p = 0.001) and the Noise 120% condition (t = − 3.76, p = 0.01). The tactile detection threshold in CIPN patients in the Noise 60% condition was Fig. 2 Comparisons of the tactile detection threshold and the log normalized jerk of grasp movement between three stimulus conditions (Noise 0, 60, and 120% condition). *Significant difference between stimulus conditions. #Significant difference between healthy controls and CIPN patients in the Noise 0% condition also significantly lower than that in the Noise 0% condition (t = 3.91, p = 0.001) and Noise 120% condition (t = − 3.88, p = 0.001). In addition, the effect of random noise stimulation on the tactile detection threshold was not significantly correlated with each clinical score (p > 0.05; FACT/ GOG NTX r = − 0.09, PNQ sensory r = 0.09, PNQ motor r = − 0.02, NPSI burning r = 0.04, NPSI pressing r = 0.04, NPSI paroxysmal r = 0.01, NPSI evoked r = 0.08, NPSI dysesthesia r = 0.04, pain intensity NRS r = 0.23) (Fig. 2).

NJ of grasp movement
The log of the NJ in CIPN patients was significantly higher than that in controls (t = − 2.99, p = 0.004). Two-way repeated-measures ANOVA revealed significant main effects of group (F = 49.93, p < 0.001) and condition (F = 10.27, p < 0.001). There was no significant interaction effect between group and condition (F = 0.001, p = 0.99). Post hoc tests revealed that the log of NJ in controls in the Noise 60% condition was lower than that in the Noise 0% condition (t = 3.15, p = 0.01) and Noise 120% condition (t = − 2.99, p = 0.01). The log of the NJ in CIPN patients in the Noise 60% condition was also significantly lower than that in the Noise 0% condition (t = 4.65, p < 0.001) and Noise 120% condition (t = − 3.64, p < 0.001). In addition, the effect of random noise stimulation on the log of the NJ was not significantly correlated with each clinical score (p > 0.05; FACT/ GOG NTX r = 0.22, PNQ sensory r = 0.15, PNQ motor r = 0.21, NPSI burning r = 0.01, NPSI pressing r = 0.09, NPSI paroxysmal r = − 0.14, NPSI evoked r = 0.09, NPSI dysesthesia r = 0.14, pain intensity NRS r = 0.04). (Fig. 2)

Discussion
The present study demonstrated that the tactile detection threshold and NJ values of grasp movements in CIPN patients were higher than those in age-matched controls. The findings quantitatively indicated that somatosensory function and precise finger movements of CIPN patients were profoundly impaired. The impaired tactile detection threshold and smoothness of grasping movements in CIPN significantly improved when vibrotactile random noise stimulation was applied with 60% intensity of the tactile detection threshold, but not when stimulation was applied at 120 or 0% of the threshold. Thus, the results indicated that the stochastic resonance phenomenon (i.e., subthreshold stimulation) was effective for improving somatosensory dysfunction and dysfunction of precise movements in patients with CIPN. Interestingly, sensorimotor function of the fingertip was improved when subthreshold noise stimulation was applied to the wrist. This effect of remote sub-threshold stimulation was similar to a previously reported finding in patients with stroke (Seo et al. 2014). As a potential mechanism underlying this effect, vibrotactile random noise at the wrist may have increased the synchronization of sensory neuron firing between the spinal cord and the somatosensory cortex (Manjarrez et al. 2002(Manjarrez et al. , 2003. The increased synchronization may facilitate neural communication between the spinal and cortical levels, thereby enhancing the ability to detect tactile stimulation of the fingertip, resulting in improvement of grasping movement. Another possible mechanism is that, at the level of the supra-spinal central nervous system, the sensorimotor representation of the fingertip is sharply outlined. Anecdotal evidence was reported in a previous study in which the subthreshold noise stimulation applied to the wrist enhanced fingertip touch-evoked potentials observed in the somatosensory, motor, and premotor cortices . In addition, a recent study reported that functional connectivity between such sensorimotor areas is associated with the effect of remote random noise stimulation . In the case of sensory loss, such as that in patients with somatosensory disorder, grasp movement is reported to become jerky because the function of feedback motor control in the central nervous system is impaired (Gentilucci et al. 1997(Gentilucci et al. , 1998Osumi et al. 2018). In the current study, NJ grasp values in CIPN patients were significantly disturbed compared with those of healthy controls. The results indicated that CIPN patients exhibited the loss of smoothness of grasp movement. Such disturbance of NJ grasp values in CIPN patients were improved in the Noise 60% condition. Considering these findings, the improvement of NJ grasp values in the Noise 60% condition suggests that subthreshold random noise stimulation may have improved the afference signal detection and feedback-controlled system in the central nervous system, improving the jerky grasp movement. A previous study using electroencephalography provides support for the suggestion that the subthreshold random noise stimulation enhances sensorimotor cortical activity (Seo et al. 2019a;Manjarrez et al. 2002). For example, subthreshold random noise stimulation was reported to increase mechanical-stimulation evoked EEG responses in sensorimotor areas (Manjarrez et al. 2002). Additionally, subthreshold random noise stimulation was reported to improve sensorimotor performance, such as the control of finger force (Mendez-Balbuena et al. 2012). However, suprathreshold random noise stimulation interfered with processing of the afferent signal itself (McDonnell and Ward 2011), and the current results revealed that grasp movement was not improved in the Noise 120% condition. These findings may suggest that subthreshold random noise stimulation improved the feedback motor control system in the central nervous system, and thus improved the smoothness of grasp movement in CIPN patients.
As an alternative to the present subthreshold somatosensory stimulation, several previous studies reported improved sensory and motor symptoms of CIPN patients using suprathreshold sensory stimulation (e.g., cold stimulation, transcutaneous electrical nerve stimulation, vibratory stimulation) (Mendez-Balbuena et al. 2012;Schönsteiner et al. 2017;Bailey et al. 2021, Childs et al. 2021). In one study, an extensive approach combining suprathreshold whole-body vibration stimulation with standard physical therapy was applied for symptoms of CIPN (Schönsteiner et al. 2017). However, a clinical study reported that applying suprathreshold somatosensory stimulation to the affected hands and feet of CIPN patients can exacerbate sensations of numbness and discomfort (Şimşek and Demir 2021). In contrast, subthreshold vibrotactile random noise stimulation is not perceived when applied, potentially minimizing the risk of exacerbating numbness and allodynia, both of which are characteristic symptoms of CIPN. Thus, future studies should confirm whether long-term subthreshold random noise stimulation reduces the risk of exacerbating patients' complaints of sensations of discomfort and annoyance.
Moreover, because subthreshold random noise can enhance motor skill learning (Prichard et al. 2014), application of subthreshold noise stimulation might improve the effects of physical and occupational therapy. In one study, rehabilitation with random noise stimulation for 2 weeks was reported to enhance motor learning in chronic stroke survivors (Seo et al. 2019b). Interestingly, such enhancement of motor learning with random noise was reported to be associated with EEG connectivity in sensorimotor areas . Considering that the cortex in CIPN patients is not directly injured and sensorimotor connectivity is critical for motor learning, delivering random noise stimulation to CIPN patients may enhance functional connectivity in sensorimotor areas and thus improve motor function in the long term. Although the present results revealed only the short-term effects of random noise, the longterm effects on CIPN patients should be observed in future studies referencing previous clinical protocols with stroke survivors (Seo et al. 2020.
The current study involved several limitations. First, regarding the method for determining vibrotactile thresholds, we measured the simple increase in stimulation intensity. The methods should be modified to increase the reliability of the results using psychophysical methods such as the staircase method. Second, the mechanisms underlying the effect of remote random noise stimulation on somatosensory and motor function of the fingertip are unclear. It would be useful to record and analyze physiological data in future studies.

Conclusion
The "silent" nature of subthreshold random noise stimulation may enable patients to keep applying the stimulation in their daily lives and consequently improve the smoothness of specific movements, such as typing on a keyboard and undoing and doing up buttons, ultimately improving their QOL.