Wireless transcutaneous electric nerve stimulation (TENS) for chronic chemotherapy-induced peripheral neuropathy (CIPN): a proof-of-concept randomized clinical trial.

Abstract

Chemotherapy-induced peripheral neuropathy (CIPN) affects approximately 30%−60% of people who receive neurotoxic chemotherapy. CIPN is associated with impaired quality of life and function and has few effective treatments. This 6-site, subject and assessor blinded RCT was designed to assess (1) preliminary efficacy (i.e., alpha pre-specified at 0.2) of a wearable, app-controlled, TENS device for chronic CIPN and (2) feasibility of conducting a confirmatory trial within the NCI Community Oncology Research Program (NCORP) (NCT 04367480). The primary outcome was the EORTC-CIPN20. The main secondary outcomes were individual symptoms assessed daily (via 0 – 10 numeric rating scales). The primary analysis was an ANCOVA (outcome: EORTC-CIPN20, fixed effect: arm, covariates: baseline EORTC-CIPN20 and site). Secondary analyses used similar ANCOVA models (excluding site) for each symptom on subgroups of subjects with ≥4 out of 10 for that symptom at baseline. 142 eligible subjects were randomized and received a device; 130 (91%) completed the study. The difference between groups in the EORCT-CIPN20 at endpoint (placebo-active) was 1.05 (95% CI: −0.56, 2.67; p = 0.199). The difference between groups for the individual symptoms were as follows: hot/burning pain: 1.37 (−0.33, 3.08; p=0.112), sharp/shooting pain: 1.21 (−0.37, 2.79; p=0.128), cramping: 1.35 (−0.32, 3.02; p=0.110), tingling: 0.23 (−0.61, 1.08; p=0.587), numbness: 0.27 (− 0.51, 1.05; p=0.492). A RCT of an app-controlled TENS device for chronic CIPN with excellent retention is feasible in NCORP. Preliminary efficacy evidence suggests that TENS is promising for pain and cramping from CIPN. A confirmatory RCT of TENS for painful CIPN is highly warranted.

Perspective

Daily, home-based TENS therapy demonstrates promising efficacy for painful CIPN symptoms in this proof-of-concept randomized clinical trial. Future confirmatory trial is warranted.

Introduction

Approximately 60% of patients who receive neurotoxic chemotherapeutic agents (e.g., platinums, taxanes, and vinca alkaloids) develop chemotherapy induced peripheral neuropathy (CIPN) and it becomes chronic (i.e., lasts at least 3 months after termination of chemotherapy) in approximately 50% of those individuals.¹ Neurotoxic agents are used to treat many common cancers including breast, gastrointestinal, lung, ovarian, and myeloma, which together were estimated to make up ~40% of new U.S. cancer cases in 2023.² CIPN presents as a variable combination of neuropathic symptoms, commonly including burning and shooting pain, tingling, cramping, and numbness. Data from our group and that of others suggest that chronic CIPN leads to functional limitations such as impaired balance,^3,4 increased falls,^5–8 limitations in activities of daily living (e.g., shopping),⁵ and interference with walking.^9,10 Multiple studies demonstrate that CIPN is associated with decreased quality of life in cancer survivors.¹¹ No therapies are approved by the FDA to treat CIPN. Duloxetine improved chronic painful CIPN in one randomized clinical trial (RCT),¹² and analgesics that are approved for other pain conditions are often prescribed clinically for patients with CIPN. However, only approximately 50% of patients with CIPN experience moderate to severe pain and no treatments have been shown to mitigate the functional limitations that are associated with CIPN, such as impaired balance or walking. Furthermore, patients often prefer non-pharmacologic treatments, especially after receiving many toxic pharmacologic treatments for cancer.

Transcutaneous electric nerve stimulation (TENS) is the application of electrical stimulation through the skin. Evidence suggests that TENS reduces central excitability and activates descending inhibition,¹³ both of which have been implicated in CIPN in animal studies and one human study. In animal models, TENS reduces the enhanced central excitability of nociceptive neurons in the spinal cord induced by inflammation and nerve injury¹⁴ and decreases the release of the excitatory neurotransmitter glutamate in the dorsal horn of the spinal cord.¹⁵ TENS also produces analgesia in animal models by activating descending inhibition in the brainstem and spinal cord through the activation of endogenous opioid, GABA, and muscarinic receptors.^16–19 Human studies confirm opioid receptor activation by TENS²⁰ and show that TENS restores descending inhibition in individuals with fibromyalgia during a single TENS session.²¹ RCTs in individuals with diabetic peripheral neuropathy show improvements in discomfort and a composite neuropathy score that includes ratings of pain, numbness, and prickling sensations,^22,23 suggesting that TENS can reduce non-painful neuropathy symptoms. Results from our single-arm pilot trial suggest that use of daily home-based TENS is feasible and improves symptoms of CIPN including pain, cramping, and tingling.²⁴

Methods

Objectives

The purpose of this proof-of-concept RCT was to evaluate preliminary efficacy of TENS in individuals with chronic CIPN compared to a sham TENS device and demonstrate feasibility of conducting a confirmatory RCT through the National Cancer Institute (NCI) Community Oncology Research Program (NCORP). Multiple sites are required in order to recruit sufficient numbers of participants with chronic CIPN for a confirmatory trial.¹² The NCORP provides the ideal platform to recruit large numbers of geographically and clinically representative participants for cancer symptom management trials. This trial aimed to determine recruitment rates for participants with chronic CIPN within the network, whether we could obtain reasonably high adherence with a home-based, App-controlled TENS device that was delivered via a multi-site study conducted in the community, and test our blinding strategy.

Study Design

This study is randomized, patient and assessor blinded, 2-arm (1:1), parallel group trial conducted at 6 geographically diverse NCORP sites. Approximately halfway through recruitment, the protocol was amended to allow 1 coordinator to perform both the blinded- (i.e., assessor) and unblinded-coordinator tasks due to staffing issues in the context of COVID-19. Sites were asked to only use this option if absolutely necessary. Only six participants from 1 NCORP site underwent study visits in which a single coordinator performed the blinded and unblinded coordinator duties. The study was coordinated by the University of Rochester NCORP research base. The NCI Central Institutional review board (IRB) and the local IRBs of each site approved the study. All participants provided informed consent prior to participation. The study is registered on clinicaltrials.gov (NCT 04367480)

Study population

Eligible participants had received treatment for cancer with a chemotherapy in one of the following classes: taxane, platinum agent, vinca alkaloid. They had to have completed their neurotoxic chemotherapy at least 3 months prior to enrollment in the study. In order to be eligible for the study, participants were required to have a CIPN diagnosis from their physician or physician designee, have bilateral abnormal sensations and report at least 1 non-painful symptom in their feet and legs. They also had to report the worst severity (over the past week) of at least two of the following symptoms as at least 4 out of 10 on a 0–10 numeric rating scale (NRS): hot/burning pain, sharp/shooting pain, numbness, tingling, or cramping. They had to be willing and able to not start new pain medications or change dosages of any pain medications that they are currently taking (other than acetaminophen of non-steroidal anti-inflammatory drugs (NSAIDs)) for the duration of the study. Finally, they had to be at least 18 years of age, able to read English, and have access to a smartphone or device that could run the TENS device App. Potential participants were excluded if they met the following criteria: (1) had a preexisting neuropathy documented in their medical chart prior to starting chemotherapy or responded “yes” to the question “Did you have frequent numbness, tingling, hot/burning, of sharp/shooting pain, or cramping in your feet before you started your chemotherapy?, (2) had taken, within the past 3 months, medications known to cause neuropathy, (3) had started a new prescription pain medication or changed dosages of a pain medication in the past 2 weeks, (4) were currently using a TENS device for any reason, (5) had an acute and symptomatic deep vein thrombosis, lower extremity edema greater than 2+ on the pitting scale, wounds or ulcers on the lower extremities, a cardiac pacemaker, or epilepsy, and (6) be pregnant or planning to become pregnant.

Sample size estimation

A sample size of 126 study completers was planned to provide 80% power to detect a 0.4 effect size at the alpha level of 0.2. Because very few treatment options are available for individuals with CIPN, we reasoned that our study should prioritize minimizing a false negative result (i.e., not advancing the study of a potentially useful and safe treatment) and accepting a slightly higher potential for type 1 error (or false positive conclusions) considering the results would be confirmed in a phase III trial. This approach is recommended for proof-of-concept clinical trials by the Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials (IMMPACT) group and multiple statisticians.^25–30 Assuming 15% withdrawal, we aimed to enroll 150 subjects. The effect size of 0.4 was chosen based on the range of effect sizes observed in a systematic review of chronic pain trials (i.e., approximately 0.3 to 0.42³¹). Thus, 0.4 was likely attainable, but would also leave room for a decrease in effect size that is often observed when transitioning an intervention from phase II to III.

Intervention

The TENS device used in this study was the Quell device manufactured by Neurometrix. It is a wireless device that is positioned with an elastic band directly below the knee. The elastic band holds an electrode strip with 4 evenly spaced electrode pads in place around the leg. All participants wore the device in the same location regardless of the distribution of their symptoms. delivers a high frequency (i.e., 60–100 Hz) stimulation (waveform is biphasic with alternating leading phase). The device is controlled by an App, which on the continuous setting, alternates between a 1-hour treatment period and a 1-hour rest period. Participants were asked to wear the device for 5 hours per day on the continuous setting every day for 6 weeks. They were asked to alternate the leg they wore it on each day in order to minimize potential skin reactions. The active device had the normal 1-hour on, 1-hour off stimulation paradigm, resulting in 3 hours of total stimulation in 5 hours of wear per day. The identical looking placebo device emitted 2 minutes of stimulation with a 30-second ramp down of intensity followed by 57.5 minutes of no stimulation in the treatment session, which resulted in 7.5 minutes of total stimulation in 5 hours. The placebo dose was designed to minimize TENS exposure but enhance blinding. This paradigm was informed by previous research demonstrating that brief stimulation provided by a placebo TENS device blinded participants better than a device with no stimulation.³²

The intensity of the stimulation was set during the baseline visit with the help of the study coordinator using the device’s calibration process, which sets the intensity at approximately twice the sensation detection threshold. After the calibration process was complete, the study coordinator instructed participants in the active group to increase the intensity until it was uncomfortable or painful and then decrease the intensity once. The goal of this process was to set the intensity to the strongest setting that was comfortable for the participant. The device App remembers the last intensity setting used by a participant, so unless the participant modified the intensity settings, subsequent sessions used the same intensity set at the baseline visit. Participants in the placebo group were not prompted to increase the stimulation until it is strong but comfortable because the stimulation did not continue sufficiently long after the calibration process to do so.

Randomization

Participants were randomized (1:1) in blocks of 2 and 4, stratified by study site. Randomization was performed using a centralized system operated by the URCC NCORP Research Base that assigned a study group to each participant at time of randomization.

Blinding

To maximize blinding the placebo device delivered 2.5 minutes of stimulation per treatment cycle (see intervention section for details). The coordinators who introduced the study device to the participants and taught them how to use it, assessed AEs, and addressed adherence and troubleshooting were unblinded. To minimize the influence of staff unblinding on the outcomes, the second coordinators served as assessors who were blinded to treatment and trained to instruct and remind participants not to discuss their experience with the device with them (the assessors). Participant and assessor blinding was assessed at the end of the trial (see outcomes).

Outcomes

Primary

The primary outcome was the European Organisation for Research and Treatment of Cancer-CIPN20 (EORTC-CIPN20) total score of 18 items that apply to all people.³³ The total score excluded questions regarding the ability to drive and get an erection because they do not apply to all people. The EORTC-CIPN20 includes items that assess CIPN-related symptoms and functional interference in daily activities. Multiple studies support the content and construct validity of the EORTC-CIPN20.^34–36

Secondary

The main secondary outcome was a daily diary of CIPN symptoms (hot/burning pain, sharp/shooting pain, tingling, cramping, and numbness). Participants were asked to rate each symptom on a 0–10 scale [0=no symptom, 10= worst symptom imaginable]. The instructions asked participants to think of the worst their legs and feet have felt over the past 24 hours when rating each symptom. Participants were to rate the symptoms in the evening while the TENS device was not actively in use. This measure was developed specifically to include relevant sensory symptoms from our pilot, feasibility trial,²⁴ worded similarly to the EORTC-CIPN20, and based on the structure of MD Anderson Symptom Inventory³⁷ a tool commonly used in studies of cancer patients. This diary was completed for 1 week at baseline and at endpoint (6 weeks); the average of the daily values was taken for each symptom to be assessed separately and the sum of the individual symptom averages was used for the total score.

Exploratory

Patient-reported outcome measures:

CIPN interference with general activity, mood, walking, normal work, relationships, sleep, and enjoyment of life will be measured using the CIPN Interference Questionnaire, which uses 0 – 10 NRSs [0 = does not interfere, 10 = completely interferes] adapted from the Brief Pain Inventory (BPI).³⁸ Physical function was measured using the PROMIS-Physical Function Short Form 8b.³⁹ Overall impression of change in CIPN and overall impression of change in ability to be active were measured using single-item questions adapted from the patient global impression of change with a 7-point Likert scale [very much improved, much improved, minimally improved, no change, minimally worse, much worse, very much worse].

Functional assessments:

Balance was measured using the balance subscale of the Short Physical Performance Battery (SPPB). The SPPB balance subscale is a composite measure of timed feet together, semi-tandem, and tandem foot stands.⁴⁰ Lower limb sensation was evaluated using a monofilament test threshold test.⁴¹ The participants were asked to close their eyes while each monofilament was applied (or not) to the dorsum of the big toe in random order 10 times (i.e., trials). With each trial, the participant was asked to state whether they think the filament touch their skin the first or second time. If the participant guessed correctly ≥ 9 times the next weaker monofilament was used. If the participant guessed correctly 7 or 8 times this filament was documented as the participant’s sensation threshold. If the participant guessed correctly ≤ 6 times the next stronger filament was trialed. The participant’s sensation threshold is the outcome of the test. This test was used previously to quantify sensation in our pilot trial.²⁴ Descending inhibition was measured using a Conditioned Pain Modulation (CPM) test. In this case, the test stimulus was pressure and the conditioning stimulus was brief (up to approximately 30 sec −1 minute) immersion in cold water. Specifically, the participant’s pressure pain threshold (PPT) (i.e., the point at which a pressure stimulus first became painful) was measured on the trapezius muscle 3 times (i.e., the test stimulus). Subsequently, the participant was asked to immerse their right hand in ice water (~4°C) for 15 seconds (i.e., the conditioning stimulus). After the 15 seconds, 3 subsequent pressure pain threshold measures were taken on the trapezius muscle while the hand was held in the water, if possible. Participants were told that they could remove their hand at any time if they could not tolerate the cold. If they removed their hand prior to the 15 seconds the post-conditioning stimulus PPT measurements were taken at that time. The CPM value was calculated as the [(PPT_{post-immersion} - PPT_pre-immerson)/PPT_{pre-immersion}] X 100% (i.e., percent increase in PPT). See Appendix A for the coordinator manual used to standardize the test across sites.

Feasibility outcomes

To assess the feasibility of conducting this trial using the NCORP method, we used the following outcomes: (1) the proportion of screened people who enrolled in the study, (2) the proportion of randomized participants who adhered to the treatment and completed the primary assessment, and (3) the proportion of randomized and completed participants who adhered to treatment using data collected from the device. The device registers a treatment session if at least 30 minutes of an hour session is completed. Thus, we used the number of treatment sessions and the total stimulation time each day to determine adherence. Participants who adhered exactly to the treatment would have registered 3 treatment sessions and 3 hours of stimulation per day. To allow for some flexibility, we calculated a second outcome in which participants who completed 3 sessions and logged at least 2.5 hours of stimulation, would be considered adherent.

Blinding

Blinding was assessed by asking participants to guess if they were assigned to the active or placebo device and then on what that guess was based (i.e., how much the device affected their symptoms, how it felt when it was on, or how it looked). Blinded coordinators were also asked to guess the group to which each participant was assigned and upon what they based their guess.

Adherence / usage

Stimulation times and intensities for each treatment session were assessed using data from the TENS device. The TENS device records the maximum intensity per treatment session. The average of the maximum intensities was calculated for each participant and summarized using the median.

Rescue analgesia

Use of acetaminophen and non-steroidal anti-inflammatory (NSAID) drugs was monitored in the daily diary. These data were summarized as the number of days in which acetaminophen or NSAIDs were used by the participant in the baseline and endpoint weeks.

Statistical Methods

Descriptive statistics were used to summarize the characteristics of the study population and the feasibility and blinding outcomes. The primary analysis was an analysis of covariance (ANCOVA) that included the EORTC-CIPN20 score at Week 6 as the outcome, treatment group as the factor, and study site and the baseline EORTC-CIPN20 score as covariates. The primary analysis used what is often termed a modified intention to treat (ITT) analysis and excluded 3 participants. The first 2 were excluded for reasons routinely cited in modified ITT analyses.⁴² One participant was found to have been ineligible after they were randomized, and thus their data could provide inaccurate information regarding the effect in the target treatment population. For another, an issue with their phone occurred on the day of randomization, leaving them randomized without actually receiving the device or completing any baseline assessments. Because the reason for exclusion had no relationship to their condition or the treatment (since they never received it), their exclusion should not introduce any bias in the analyses. Finally, because extreme outliers can drastically skew the data in studies with modest sample sizes, it was pre-specified in our protocol that extreme outlying data would be removed prior to unblinding. When reviewing the distribution of the primary outcome prior to unblinding, an extreme outlier in the difference between baseline and endpoint of the primary outcome (i.e., 6 SDs from the mean) was identified. It was found that this participant answered “not at all” for all of the items on the first page of the EORTC-CIPN20 at baseline, while rating the same symptoms as 8 out of 10 on the daily diary at baseline. Because the low score on the baseline EORTC-CIPN20 was the cause for the extreme outlier in the change from baseline and because the baseline EORTC-CIPN20 was also inconsistent with all other measures of similar constructs for this participant at baseline, we concluded that these data were a mistake and excluded this participant from the primary analysis as outlined in the protocol. Data from this participant were included in all analyses that did not involve the EORTC-CIPN20. Available data from all remaining participants were included. To account for missing data, a multiple imputation (MI) analysis using a fully conditional specification method ⁴³ (SAS PROC MI: FCS, with 100 imputations) was conducted. Secondary analyses used ANCOVA models similar to the primary analysis with the covariate of the baseline score of the outcome for each model replacing the baseline value of the EORTC-CIPN20. Missing data were not imputed in the secondary and exploratory analyses given the low drop-out rate (12 participants; 8.4%) and that 6 of those participants withdrew for reasons definitely unrelated to the treatment assignment (e.g., device distribution error, death, App glitch). The outcomes of the secondary analyses included the individual symptoms from the CIPN symptom inventory. Two models were applied for each symptom, one on the total completer population (N=130) and one on the subgroup of participants who reported at least 4 out of 10 on the particular symptom at baseline to avoid a floor effect in the analyses. The subgroup analyses did not include site as a covariate due to the smaller sample sizes. Although the subgroup analyses were not specifically specified in the protocol, the cut-off of 4 was based on the typical pain severity entry criteria for a clinical trial that is used to avoid a floor effect44 and was the only cut-off tested in secondary analyses. Additional post-hoc analyses were designed after imbalances in baseline characteristics were identified, including time since last neurotoxic chemotherapy, sex, cancer type, and chemotherapy type. Sex, cancer type, and chemotherapy type are highly correlated because breast cancer occurs almost exclusively in females and is often treated with taxane-based chemotherapy. Therefore, to minimize the number of variables added to the models, we chose to adjust for sex and logarithm of time since chemotherapy (referred to as time since chemotherapy). Logarithm was used to better adhere to the normality assumption.

Analyses of continuous exploratory outcomes (i.e., CIPN symptom inventory total score, PROMIS Physical function, CIPN symptom interference, CPM test, monofilament test) used similar ANCOVA models as the secondary analyses. The analysis of the CPM test and the monofilament test included site as a covariate because of the potential influence of the assessor on these outcomes. The proportion of participants who reported, their impression of change in “CIPN symptoms” or “ability to be active as “very much improved”, “much improved”, “minimally improved”, “no change”, “minimally worse”, “much worse”, and “very much worse” was compared between treatment groups using a Cochran Armitage Trend Test as indicated by the protocol. Additionally, post-hoc analyses logistic regression analyses were used to compare the proportion of participants who rated their symptoms or ability to be active as very much, much, or minimally-improved vs. no change or any amount worse. All regression analyses were run with and without sex and time since neurotoxic chemotherapy as covariates. The between group differences are represented by the difference in marginal means from ANCOVA models and by odds ratios in logistic regression models. This study was designed to evaluate preliminary efficacy, rather than confirm efficacy. The p-value for the primary efficacy analysis was set to 0.2 to prioritize increasing power while accepting an increase in potential type 1 error rate to prevent prematurely stopping a line of research that could provide an effective treatment option for a condition that has few effective treatments.^25–30 We did not adjust for multiple comparisons in the secondary analyses due to the exploratory nature of the study.

Results

Participant flow and baseline characteristics

Study enrollment was conducted between September 2020 and September 2022. In total, 380 patients were screened for the study, 223 (59%) of whom were eligible; 151 subjects (40% of the total screened and 68% of those who were eligible) enrolled in the study. The majority of participants were enrolled at 4 sites, with 2 sites enrolling 50 participants each and 2 sites enrolling 31 and 17 participants, respectively. Prior to randomization, 3 became ineligible, 2 were lost to follow-up, and 2 changed their mind, leaving 144 participants randomized. Three participants were removed after randomization (see Methods section and CONSORT diagram for reasons). Thus, the primary analysis cohort included 141 subjects. Of the 142 subjects randomized and included in any analysis, 130 (92%) completed the study (Figure 1). Randomized subjects (included in any analysis) were on average 63 years old and 65% female; 85% were white and 10.6% were Black or African American; 1.4% were Hispanic. The majority of subjects received taxane- (37%) or platinum-based (37%) chemotherapy or a combination of both (16%). The median time since neurotoxic chemotherapy was 454 days (IQR: 191, 956). The placebo group had a higher percentage of female participants (73% female vs. 57% female in the active group). The median time since neurotoxic chemotherapy was longer in the placebo group (667 days in placebo vs. 303 days in active). The type of cancer and chemotherapy were also somewhat imbalanced, with more breast cancer patients and patients receiving taxanes only in the placebo group. See Table 1 for complete depiction of demographic and clinical baseline characteristics.

Figure 1.

CONSORT diagram

Table 1.

Demographic and clinical characteristics

	All (142) N (%); Mean (SD);	Placebo (70) N (%) or Mean	Active (72) N (%) or Mean	p-value
Characteristic	Median [IQR]	(SD)	(SD)
Age	62.9 (9.7)	63.3 (10.7)	62.6 (8.6)	0.666
Sex				0.047
Female	92 (65%)	51 (73%)	41 (57%)
Male	50 (35%)	19 (27%)	31 (43%)
Race				0.372
White	120 (85%)	60 (86%)	60 (83%)
Black or AA	15 (11%)	7 (10%)	8 (11%)
Multiple races	5 (3.5%)	1 (1.4%)	4 (5.6%)
Chose not to answer	2 (1.4%)	2 (2.9%)	0 (0.0%)
Ethnicity				0.719
Not Hispanic/Latinx	133 (94%)	67 (96%)	66 (92%)
Hispanic/Latinx	2 (1.4%)	1 (1.4%)	1 (1.4%)
Chose not to answer/	7 (4.9%)	2 (2.9%)	5 (6.9%)
unknown
Education level				0.981
Less than HS	3 (2.1%)	2 (2.9%)	1 (1.4%)
HS graduate	53 (37%)	26 (37%)	27 (38%)
College degree	57 (40%)	27 (39%)	30 (42%)
Graduate degree	28 (20%)	14 (20%)	14 (19%)
Chose not to answer	1 (0.7%)	1 (1.4%)	0 (0.0%)
BMI	31 (7.0)	32 (7.9)	30 (5.8)	0.089
Cancer type				0.259
Breast	51 (36%)	31 (44%)	20 (28%)
GI	49 (35%)	20 (29%)	29 (40%)
Hematologic	14 (9.9%)	5 (7.1%)	9 (13%)
Gynecologic	13 (9.2%)	6 (8.6%)	7 (9.7%)
Other	15 (10.6%)	8 (11%)	7 (9.7%)
NTX chemo class				0.091
Platinum	53 (37%)	22 (31%)	31 (43%)
Taxane	52 (37%)	32 (46%)	20 (28%)
Platinum and taxane	23 (16%)	11 (16%)	12 (17%)
Bortezomib	7 (4.9%)	4 (5.7%)	3 (4.2%)
Vinca alkaloid	7 (4.9%)	1 (1.4%)	6 (8.3%)
Time since NTX chemo	454 [191, 956]	667 [349, 1129]	303 [170,735]	0.006

AA: African American; IQR: interquartile range, NTX: neurotoxic

Treatment adherence

Adherence data were available for 140 of the randomized participants. Of those 140 participants, 122 had at least 40 days of adherence data available. For the remaining 18 participants, 28 days was the median number of days for which adherence data were available. Depending on how missing adherence data were handled or whether considering randomized participants who received the intervention or completers only, the proportion of participants who received at least 3 treatment sessions per day on at least 70% of the days ranged from 71% to 88%. Supplemental Table 1 outlines the adherence results for different scenarios (including whether using 2.5 or 3 hours of stimulation per day was used as the minimum cut-off for 3 treatment sessions). No apparent differences in adherence were observed between the active and placebo groups. The median of the average maximum pain intensity per session for each participant was 21.0mAmps in the placebo group and 28.2mAmps in the active group.

Blinding

Sixty-three of the placebo group and 67% of the active group guessed their group assignment correctly. Supplemental Table 2 illustrates the proportion of subjects in each group who based their guess on how the TENS device felt when it was on, how it affected their symptoms, or how it looked. Of those who guess correctly, 11 (26%) in the active group based the guess on how it affected their symptoms and 30 (70%) based it on how the device felt. For the placebo group, 25 (61%) based the guess on how the device affected their symptoms and 16 (39%) based it on how the device felt. The coordinators guessed the treatment assignment correctly in 40 (66%) of cases in the placebo group and 41 (64%) of cases in the active group. The majority of instances that the coordinators guessed correctly, they reported basing their guess on things that the participant had told them about the device (52% for placebo group and 71% for active group).

Primary efficacy analysis

The EORTC-CIPN20 score decreased by 8.7% in the placebo group and 11.6% in the active group. The difference in marginal means between groups at endpoint (placebo-active) was 1.05 (95% CI: −0.56, 2.67; p = 0.199) (Table 2). The result without imputation for missing data (N=129) was very similar (mean difference: 1.08 (95% CI: −0.53, 2.68; p = 0.186). The result with adjustment for time since neurotoxic chemotherapy and sex showed a slightly higher effect size (mean difference: 1.43 (95% CI: −0.27, 3.12; p = 0.098).

Table 2.

ANCOVA analysis of primary and secondary outcomes as per original protocol

Outcome	% Change from baseline		Marginal mean group difference [placebo-active] (95% CI)	p-value
Primary analysis population (N=141)
	Placebo	Active
EORTC-CIPN20	−8.7%	−11.6%	1.05 (−0.56, 2.67)	0.199
Analyses including all completers
	Placebo	Active
Hot/burning pain	−23.8%	−37.3%	0.37 ( −0.34, 1.08)	0.302
Sharp/shooting pain	−33.0%	−43.4%	0.33 (−0.37, 1.02)	0.351
Cramping	−13.5	−38.4%	0.64 (−0.02, 1.31)	0.058
Tingling	−20.8%	−25.0%	0.19 (−0.56, 0.94)	0.622
Numbness	−11.7%	−22.5%	0.53 (−0.22, 1.27)	0.166
CIPN symptom diary total score*	−19.2%	−29.7%	2.03 (−0.54, 4.60)	0.120
Subgroup analyses including only those with >=4 at baseline in the particular symptoms
	Placebo	Active
Hot/burning pain (N: P=22; A=22)	−28.3%	−49.2%	1.37 (−0.33, 3.08)	0.112
Sharp/shooting pain N: P=24; A=23)	−34.3%	−52.5%	1.21 (−0.36, 2.79)	0.128
Cramping (N: P=18; A=18)	−30.2%	−52.1%	1.35 (−0.32, 3.02)	0.110
Tingling (N: P=50; A=56)	−23.7%	−27.5%	0.23 (−0.61, 1.08)	0.587
Numbness (N: P=50; A=61)	−18.4%	−24.0%	0.27 (−0.51, 1.05)	0.492

^*Total of the weekly averages of each symptom.

Secondary efficacy analyses

Table 2 illustrates the changes in the individual CIPN symptoms reported via the 0–10 NRS daily diary in the full sample of study completers and the subsets of participants who reported at least 4 out of 10 at baseline. When considering the subgroup populations, the between group differences ranged from 1.21 to 1.37 NRS points for hot/burning pain, sharp/shooting pain and cramping, with p-values considerably lower than the 0.2 threshold preset for this exploratory trial. However, the between group difference for numbness and tingling were considerably smaller and the p-value was well above the 0.2 threshold (Table 2). After adjustment for sex and time since neurotoxic chemotherapy, the effect sizes were similar (Supplemental Table 3). The full model results for the EORTC-CIPN20 and individual symptom analyses are presented in Supplemental Table 4.

At the end of the trial, 41 (61%) participants in the active group more frequently rated their CIPN symptoms as very much, much, or minimally improved compared to 26 (42%) in the placebo group. The odds of reporting improvement in CIPN symptoms were 2.18 (95% CI: 1.08, 4.41) times higher in the active vs. placebo group (p=0.03). Twenty-eight (42%) participants in the active group reported that their ability to be active was very much, much, or minimally improved compared to 16 (26%) in the placebo group (Table 3). After adjustment for time since neurotoxic chemotherapy and sex, the odds of reporting improvement in CIPN symptoms and ability to be active were 2.68 (95% CI: 1.25, 5.76; p=0.012) and 2.27 (95% CI: 1.02, 5.07; p=0.046), respectively (Supplemental Table 5). Granular distributions for all of the responses in these two questions and the results of the Cochran-Armitage trend test are reported in Supplemental Table 6.

Table 3.

Exploratory analyses of dichotomous outcome measures

	Improved	No change or worse	OR (95% CI)*	p-value*
Impression of change in neuropathy
Placebo	26 (42%)	36 (58%	2.18 (1.08 – 4.41)	0.030
Active	41 (61%)	26 (39%)
Impression of change in ability to be active
Placebo	16 26%	46 (74%)	2.06 (0.98 – 4.4)	0.058
Active	28 (42%	39 (58%)

^*OR (active vs. placebo) and p-value from Logistic regression with outcome improved vs. no change or worse (neuropathy symptoms and activity).

A floor effect was observed with the SPPB balance subscale, with almost all participants able to perform the side-by-side (98%) and semi-tandem (96%) tests successfully at baseline. The distribution of the results on the tandem test at baseline was more variable. Therefore, we assessed the effects of the TENS intervention only on the tandem test. After adjusting for baseline performance on the SPPB tandem stand test, the odds of perfectly performing the SPPB tandem stand test were 3.93 (95%CI: 1.13, 13.68; p=0.031) times higher in the active than the placebo group. This observed effect was smaller once the model was adjusted for sex and time since chemotherapy (OR: 2.83 (95% CI: 0.76, 10.59) p = 0.123).

We observed a small effect on self-reported physical function measured using the PROMIS short form, with the placebo and active groups improving by 3.4% and 7.0%, respectively. The mean difference was −1.01 (95% CI: −2.39, 0.37, p=0.150) in the model that adjusted only for the baseline PROMIS physical function score (Table 4). The observed effect after adjusting for sex and time since neurotoxic chemotherapy was −1.54 (−2.98, −0.10; p=0.037) (Supplemental Table 7). The PROMIS Physical function measure has items related to a range of function items, so we performed post-hoc analyses to identify whether the results on the Physical Function measure were due to effects on specific items. Two items seemed to be affected most by the TENS intervention (i.e., ability to “walk for at least 15 minutes” and ability to do “heavy work around the house”). These items are measure on a 1–5 scale and a larger number indicates better function. The mean difference between groups (placebo-active) were as follows: (1) ability to walk: −0.39 (95% CI: −0.66, −0.11; p=0.005); and (2) ability to perform heavy work: −0.32 (95% CI: −0.64, −0.001; p=0.049). The results were similar with adjustment for sex and time since chemotherapy (Supplemental Table 8).

Table 4.

Exploratory analyses of continuous outcome measures.

Outcome	% Change from baseline		Marginal mean group difference [placebo-active] (95% CI)	p-value
	Placebo	Active
CIPN-int. total score	−31.7%	−37.2%	1.13 (−2.30, 4.60)	0.517
Physical function	3.4%	7.0%	−1.01 (−2.39, 0.37)	0.150
CPM	−14.4%	−1.6%	−0.91 (−8.19, 6.36)	0.804
Monofilament sensation test	−1.39%	1.88%	−0.03 (−0.49, 0.43)	0.907

^*a higher score is better on the PROMIS Physical Function and CPM test.

No noticeable treatment effect was observed using on the monofilament sensation test, the conditioned pain modulation test, or the CIPN interference in daily life score (Table 4, Supplemental Table 7).

Rescue medication

Sixty-four percent of participants (14 placebo group; 25 active group) reported taking NSAIDs or acetaminophen during the baseline or endpoint weeks of the study. Seven (50%) of those in the placebo group reduced the number of days that they reported taking NSAIDs or acetaminophen by at least 30% compared to 8 (32%) in the active group (p = 0.268).

Adverse events

Three serious adverse events occurred in 3 participants; 1 death and 2 hospitalizations. All serious AEs were unrelated to the intervention. Table 5 presents all AEs that were reported for at least 2 participants. The most common AEs were new paresthesias or dysesthesias, muscle cramps, and skin reactions. All instances of muscle cramps had resolved by the end of the study. All skin reactions in the active group had resolved by the end of the study, where as those in the placebo group were documented as unresolved. Sixty-seven percent and 33% of paresthesias/dysethesias were resolved or resolving by the end of the study in the active and placebo groups, respectively. Four subjects (2 in each group) had paresthesia/dysethesia-related AE documented as resolved with sequelae or unresolved at the end of the study period. The dose of the TENS stimulation was decreased in 3 participants reporting paresthesias or dysesthesias and in 1 participant reporting muscle cramps.

Table 5.

Adverse events reported for at least 2 subjects.

AE	Placebo N (% of total (N=70))	Resolved/resolving N (% of participants with the AE)	Active N (% of total (N=72))	Resolved/resolving N (% of participants with the AE)
Paresthesias / Dysethesias	4 (5.7%)	1 (25%)	6 (8.3%)	4 (67%)
Muscle cramps	3 (4.3%)	3 (100%)	5 (6.9%)	5 (100%)
Skin reaction*	1 (1.4%)	1 (100%)	4 (5.5%)	3 (75%)
Itching	1 (1.4%)	--	2 (2.8%)	2 (100%)
Pain	2 (2.9%)	1 (50%)	1 (1.4%)	1 (100%)

^*Contact dermatitis/eczema/rash

AE numbers are presented as number of participants reporting each AE

Discussion

This is the first randomized, placebo-controlled trial of an app-controlled wireless TENS device for chronic CIPN. The feasibility of conducting such a study through the NCORP Network was demonstrated, with 151 participants recruited in 2 years at just 6 of the 31 available NCORP sites and an excellent retention rate of 90%. Study coordinators and subjects were able to navigate the app and adherence to the prescription of TENS usage was high, allowing for a dosage of TENS that is rarely achieved in clinical trials. Ninety percent completed the daily symptom diaries and over 90% of participants completed all of the physical assessments. The blinding strategy was fairly successful, with 67% and 63% of participants in the active and placebo groups, respectively, correctly identifying their treatment assignment. This blinding for the active group was similar to a placebo-controlled RCT of daily TENS for fibromyalgia in which 70% of participants in the active group guessed their group assignment correctly.²¹ The assessor blinding was not completely successful, with the most likely explanation for unblinding being the participants discussing their experience with the device. These data suggest that a future study should include more frequent training for assessors to reinforce expectations that the participants not discuss their device or symptoms with them. However, it is unlikely that coordinator blinding biased the treatment outcomes given that 2 of the 3 outcomes on which the assessor would have the most influence (i.e., monofilament and CPM tests) showed no treatment effect.

Preliminary efficacy of the TENS device was also supported in this proof-of-concept clinical trial, with the primary analysis assessing improvement in CIPN using the EORTC-CIPN20 reaching the pre-specified significance threshold of 0.2. The magnitude of change in the EORTC-CIPN20 outcome was relatively small (Cohen’s d 0.12). However, the EORTC-CIPN20 includes multiple items that are not specific to pain. Because we hypothesized that TENS would likely have the largest effects on painful symptoms, our main secondary analyses focused on individual CIPN symptoms. These secondary analyses of individual symptoms in the sub-groups of participants who would be included in a future trial (i.e., those with ≥4 out of 10 for the symptom) confirmed our hypotheses and demonstrated that the largest effects of the TENS device are on sharp/shooting and hot/burning pain, and cramping. The differences between active and placebo groups in these subgroup analyses ranged from 1.21 to 1.37 NRS points. These observed differences are considerably larger than the 0.73 NRS-point difference observed in a RCT of duloxetine for painful CIPN¹² (i.e., the result that supports the ASCO recommendations for duloxetine as the only effective treatment for CIPN⁴⁵). Additionally, the p-values for these analyses ranged from 0.11 to 0.13, which are well below the 0.2 cut-off we pre-specified for the primary analysis, even with small sample sizes in these subgroups (i.e., N=36 to 47). Finally, the larger improvements in pain and cramping than numbness and tingling was highly consistent between this trial and our open-label pilot trial ²⁴. The consistent results between these 2 studies further supports the potential for the treatment effect to be repeated in a confirmatory trial.

Fifty-six percent of the study sample reported ≥4 out of 10 severity for at least 1 painful symptom, demonstrating that a large portion of cancer survivors with CIPN would benefit from improvement in these symptoms. Although TENS is often used to treat pain, the evidence to support its effects in neuropathic pain are extremely limited and it is not included in treatment guidelines. Future RCTs definitively demonstrating the effects of daily TENS for painful symptoms of CIPN are important to promote the dissemination of TENS in this area. Although outside the scope of this article, future publications using these data will compare differing entry criteria and primary outcome measures, including personalized outcomes, that would be most effective for future trials. These data will provide variability estimates for these outcomes and allow for well-informed sample size calculations for a confirmatory trial.

A recently published compared TENS to scrambler therapy in patients with chronic CIPN⁴⁶. Scrambler therapy is an electrical stimulation-based therapy that uses 5 paired stimulation channels and is “designed to replace endogenous pain information with synthetic “non-pain” or “normal self” electrical signals.”⁴⁶ In this study, a 50% improvement in pain or tingling with 2 weeks of treatment was more frequently observed with scrambler therapy than with TENS. However, the duration of TENS usage was only 30 minutes per day (i.e., much lower than in our trial) and the type and intensity of the TENS was not reported in the article. Prior studies show that TENS parameters, particularly intensity, are critical to effectiveness.^13,47–49 Furthermore, the Scrambler study did not include a control group for TENS so the effects of TENS could not be evaluated. While scrambler therapy is a promising treatment, it requires daily trips to the clinic for 2 weeks and the long-term effects were smaller than those observed directly after treatment, suggesting a single application may not be sufficient. A home-based, TENS intervention is also promising and would provide an important, low-cost alternative.

The most common AEs observed in this trial were new paresthesias or dysesthesias, muscle cramps, and skin reactions. These AEs are consistent with those observed in our pilot trial ²⁴ and similar to that reported by Dailey et al.⁵⁰ in individuals with fibromyalgia. All AEs were resolved or resolving by the end of the trial except for 3 participants with new paresthesias / dysesthesias and 2 participants with skin reactions. These data suggest that TENS is a low-risk intervention with few side effects that are very likely to resolve, either with discontinuation, decreased dosing, or time. However, it is important that individuals who are offered TENS are made aware of these possible side effects and instructed to decrease TENS intensity or duration if improvements in their CIPN symptoms do not outweigh the presence of new symptoms. Individual dose optimization could improve this overall efficacy/side effect balance. Considering that pharmacologic treatments that are used to treat neuropathic pain all have significant side effects and risks this relatively low risk intervention could be a highly valuable option to treat CIPN.

We hypothesized that TENS would improve CIPN symptoms by improving the efficiency of central descending inhibitory pathways and used a CPM test to evaluate these pathways. We did not detect any change in CPM after treatment with TENS. These results are inconsistent with a study that demonstrated improvement in CPM with active, but not sham TENS in patients with fibromyalgia.²¹ This study performed the CPM test during the active TENS stimulation, whereas we performed CPM in the absence of the TENS stimulation. Although TENS and CPM both activate descending inhibitory pathways, animal models suggest that the signaling originates in different areas of the central nervous system and utilizes different mechanisms,^51–53 thus it is possible that the CPM test cannot detect TENS-induced changes in descending inhibitory pathways.

Strengths and limitations.

Strengths of this study include that it was conducted in geographically diverse community oncology centers and used a randomized, placebo-controlled design with largely successful blinding, given the nature of the intervention. The retention rate was excellent, with most discontinuations having a clear cause that was unrelated to the study intervention, limiting potential biases from missing data. This study was designed with an alpha cut-off of 0.2. This choice allowed us to decrease the chance of falsely rejecting a potentially effective, safe, convenient, and relatively inexpensive therapy for a condition with very few treatment options within the constraints of a feasible sample size for an early phase study. This choice also increases the possibility that our results may not be confirmed in a larger trial than if we had chosen a lower alpha. However, the facts that the primary analysis met this pre-specified significance level and key secondary analyses of painful symptoms demonstrated large improvements with active compared to the placebo TENS provide strong support for a future confirmatory trial in the painful sub-population. The randomization process failed to balance some key baseline characteristics, namely sex and the time since completion of neurotoxic chemotherapy. This limitation is less concerning because sensitivity analyses that adjusted for these baseline characteristics resulted in very similar results to the pre-planned analyses without these covariates. Participants can develop CIPN after chemotherapy completion so it is possible that eligible participants had CIPN for shorter than 3 months. We did not collect detailed descriptions of cramping (other than localization to the lower limbs). It is possible that cramping had causes other than CIPN in some participants.

Conclusions

This is the first placebo-controlled, RCT of daily TENS for CIPN. The study supports the preliminary efficacy of the TENS device for painful CIPN symptoms. Future studies of TENS for CIPN should focus on patients with painful symptoms, which our data suggest makes up approximately half of the population of patients with chronic CIPN of sufficient severity to join a trial. Such studies, if successful, will confirm efficacy and promote dissemination of this safe, non-pharmacologic treatment for painful CIPN, a highly unmet medical area.

Highlights.

• This trial provides evidence for preliminary efficacy of TENS for painful CIPN symptoms

• This trial demonstrates feasibility of conducting a confirmatory RCT of TENS for painful CIPN

• No effect of TENS on numbness or tingling associated with CIPN was observed