Mechanistic and thermal characterization of acupuncture for chemotherapy-induced peripheral neuropathy as measured by quantitative sensory testing

Chemotherapy-induced peripheral neuropathy (CIPN) is a common and debilitating side effect of chemotherapy. Acupuncture is a promising non-pharmacological intervention for CIPN. However, the physiological effects of acupuncture treatment remain poorly understood. We examined the effects of acupuncture on CIPN using semi-objective quantitative sensory testing (QST). We conducted a randomized controlled trial of real acupuncture (RA) and sham acupuncture (SA) compared to usual care (UC) in cancer survivors with moderate-to-severe CIPN. Treatment response was assessed with QST measures of tactile and vibration detection thresholds in hands and feet, thermal detection, and pain thresholds at weeks 0, 8, and 12. Constrained linear mixed model (cLMM) regression was used for statistical analysis. 63 patients completed QST testing. At week 8, vibrational detection thresholds in feet were significantly lower in RA and SA (p = 0.019 and p = 0.046) than in UC, with no difference between RA and SA (p = 0.637). Both RA and SA also showed significantly higher cool thermal detection than UC (p = 0.008 and p = 0.013, respectively), with no difference between RA and SA (p = 0.790). No differences in tactile detection, vibrational detection in hands, warm thermal detection, and thermal pain thresholds were detected among the three arms at weeks 8 and 12. QST demonstrated different patterns in RA, SA, and UC. After eight weeks of RA, we observed significant improvements in the vibrational detection threshold in feet and cool thermal detection threshold in hands compared to UC. No significant differences were seen when compared to SA. Trial Registration: ClinicalTrials.gov (NCT03183037); June 9, 2017.


Introduction
Chemotherapy-induced peripheral neuropathy (CIPN) is a painful and debilitating consequence of neurotoxic chemotherapeutic agents (i.e., taxanes, platinums, vinca alkaloids, bortezomib). In up to 30% of patients, CIPN symptoms such as pain, paresthesia, sensory loss, poor dexterity, and increased risk of falls can persist months to years beyond chemotherapy completion, resulting in low quality of life [1][2][3][4][5]. Unfortunately, there are no options for prevention and limited options for pain control. Most clinical research on CIPN interventions has been based on neurotoxicity-specific patient-reported outcomes (PROs) and National Cancer Institute-Common Terminology Criteria for Adverse Events (NCI-CTCAE).
Quantitative sensory testing (QST) is a clinical and translational tool for the assessment of small sensory fiber impairments [6]. It provides quantifiable information regarding gains and losses in somatosensory function by measuring sensory detection and pain thresholds in response to standardized sensory inputs [7]. Multiple studies have demonstrated abnormal QST findings in patients with CIPN [8][9][10]. Previously, our group reported that QST detected significantly worse tactile, vibration, and thermal detection thresholds in patients with CIPN when compared to patients without CIPN; QST correlates with PROs and can complement them in evaluation of CIPN [11,12]. Thus, QST could 1 3 be a useful measure to describe the CIPN phenotype and changes in response to an intervention.
Acupuncture is a widely used non-pharmacological Traditional Chinese Medicine technique in which acupuncturists apply fine needles to specific acupoints throughout the body to suggestively modulate the levels of neurotropic factors and neurotransmitters such as enkephalins, betaendorphin, and dynorphins [13,14]. Rising evidence has highlighted the effectiveness of acupuncture in reducing neuropathic pain [15]. However, both clinical research and practice suggest that response to acupuncture is variable among patients with persistent CIPN [16][17][18][19][20]. In this threearm placebo-controlled acupuncture study for patients with moderate-to-severe CIPN, we prospectively performed QST testing to determine somatosensory characteristic changes.

Study participants
We conducted a three-arm, randomized controlled trial comparing real acupuncture (RA), sham acupuncture (SA), and usual care (UC) at Memorial Sloan Kettering Cancer Center (MSK) in New York City. Participants were recruited from July 2017 to June 2018. The study protocol was approved by the MSK Institutional Review Board and the clinical trial was registered at ClinicalTrials.gov (NCT03183037). Eligible study participants were solid tumor cancer survivors who had completed neurotoxic chemotherapy at least three months before enrollment, with persistent moderateto-severe CIPN by subjectively reporting numbness, tingling, or pain rated greater or equal to four out of a 0-10 numeric rating scale, and were either not on neuropathy medication or on a stable regimen for the previous three months. We excluded participants who had a pacemaker or who had received acupuncture within five years of enrollment. We obtained informed consent from all patients prior to enrollment.

Study design
Using computer-generated randomization conducted by the MSK Clinical Research Database, participants were randomly assigned (1:1:1) to receive RA for a total of ten treatments in eight weeks, SA for a total of ten treatments in eight weeks, or UC using randomly permuted blocks of random length stratified by most bothersome CIPN symptom (tingling versus numbness versus pain) and severity of that symptom (moderate 4-6 versus severe 7-10 on the 0-10 CIPN symptom numeric rating scale [NRS]). We previously reported the detailed protocol and the primary endpoint for this study, NRS at week 8, and secondary quality-of-life endpoints [21,22].

Quantitative sensory testing
A research study assistant (RSA) administered QST assessments in a quiet room with minimal environmental stimuli at baseline and weeks 8 and 12. Participants were asked not to take any pain, stimulant, or sedative medications at least 12 h prior to testing session. They were assessed with their eyes closed to minimize influence by visual input on sensory responsiveness.

Tactile detection threshold (TDT)
Tactile detection threshold (TDT) was determined using a set of 20 Von Frey filaments (Touch Test Sensory Evaluators, North Coast Medical, Inc.) calibrated to generate a force in grams (g) within a 5% standard deviation (SD). TDT was assessed at the dorsum of the distal interphalangeal joint of the right and left third fingers and the right and left first toes. Starting with the smallest filament size, an RSA applied the filament to the testing site at a 90-degree angle until the filament bent. The RSA then repeated application with ascending filament size until the patient-reported tactile sensation at the test site. The force at which the filament was first perceived was recorded as the tactile threshold [23].

Vibration detection threshold (VDT)
Vibration detection threshold (VDT) was assessed by a hand-held biothesiometer (Bio-Medical Instrument Company; Newbury, Ohio) at the dorsum of the distal interphalangeal joints of the right and left index fingers and the right and left first toes. The amplitude of device vibration was gradually increased (1 V/second) until participants first perceived vibration (perception threshold) and then decreased at the same rate (1 V/second) until participants indicated that vibration had disappeared (disappearance threshold). The average of three paired measurements was recorded as the VDT at each of the four test sites.

Thermal detection and thermal pain thresholds
Thermal Detection and Thermal Pain Thresholds were assessed by a thermal neurosensory analyzer (TSA-II; Medoc Ltd.) that measured cool and warm thermal detection threshold (THDT), as well as cold and hot pain thermal threshold (THT) within a range of 0-52 °C. Cool THDT, warm THDT, cold pain THT, and hot pain THT were assessed at one test site, the thenar eminence of the dominant hand. The thermode was first attached to the surface of the participant's skin at the test site to allow temperature acclimation to a baseline skin temperature of 32 °C. For the cool THDT and warm THDT assessments, the thermode temperature decreased or increased at a rate of 1 °C/s until the participant perceived the first cool or warm sensation, respectively. For the cold pain THT and hot pain THT assessments, the thermode temperature decreased or increased at a rate of 2 °C/s until the participant indicated the transition of the cool and warm sensations to painful cold and hot sensations, respectively. We repeated each test three times with mean values used for analysis.

Statistical analysis
To estimate potential treatment effects and provide insight into QST trajectories over time while also including participants with missing follow-up scores in the analysis per the intention-to-treat principle, we analyzed each QST outcome measure using a constrained linear mixed model (cLMM). We constrained the treatment arms to have a common baseline mean [24], reflecting the pre-randomization timing of the baseline assessment. The dependent variable vector included the pre-randomization baseline (week 0) assessment, as well as the post-randomization assessments at weeks 8 and 12. The independent variables were treatment arm, week (categorical), and arm-by-week interaction. A patient-level random intercept was included in the model to account for the repeated within-patient outcome measurements over time. All randomized patients with at least one outcome assessment were included in the model. Results are reported as least-squares means, mean differences, and confidence intervals (CIs), with inferences regarding differences between arms and within-arm change based on model coefficients from the arm-by-week interaction and contrasts of model-adjusted means. Analyses were conducted using R (v4.1.0) [25] with models and marginal means estimated using the lme4 [26] and emmeans [27] packages, respectively.

Results
From July 2017 to June 2018, we enrolled 75 participants, with 27 in RA, 24 in SA, and 24 in the UC group. As QST testing was only available at the main campus site, 12 participants opted out or were unable to travel to complete QST testing. As such, a total of 63 participants (23 participants in the RA group, 22 participants in the SA group, and 18 participants in the UC group) completed the QST assessment at least once at weeks 0, 8, and/or 12. The consort diagram was published in the primary paper [21].

Baseline patient characteristics
We previously described the baseline characteristics of the total 75 participants enrolled in this study [22]. The characteristics of the 63 patients who completed QST testing are shown in Table 1. These characteristics are comparable to the overall study population. In brief, mean (range) age was 60.2 (36.3-85.9) years, mean body mass index (BMI) was 27.9 (19.8-44.7), 81% were female, 27% were non-white, 56% were breast cancer survivors, and mean time since cancer diagnosis was 4.2 (0.3-40.8) years. 54% received taxane-based chemotherapy only, 25% received platinumbased chemotherapy only, and 21% received both taxane and platinum combined chemotherapy. Patients were balanced among the three arms (Table 1).

Tactile detection threshold (TDT)
Using cLMM analysis, there were no significant differences of TDT means and mean changes in hands among RA, SA, and UC arms at weeks 8 and 12 (Tables 2, 3). TDT reduction, suggesting increased perception to tactile stimuli, was numerically larger in RA than in SA and UC at week 8, but it was not statistically significant. In feet, there were no statistically significant TDT differences between the three arms (Tables 2, 3).

Vibration detection threshold (VDT)
VDT means and mean change over time in both hands and feet are shown in Table 2. In hands, there were no statistically significant differences between RA, SA, and UC at weeks 8 and 12 ( Fig. 1; Table 3). In feet, compared to the UC arm at week 8, the RA arm had a VDT 6.88-points lower (p = 0.019), and the SA arm had a VDT 5.61-points lower (p = 0.046). VDT in RA was 1.27 points lower than SA at week 8 but not statistically significant (p = 0.637).

Thermal detection threshold (THDT)
The cool and warm THDT are shown in Table 2 as modelestimated means and mean changes from baseline in each arm and in Table 3 as model-estimated differences in changes between arms. There were no significant changes from baseline in cool THDT within the RA and SA arms, but the UC arm had significantly lower cool THDT at week 8 compared to baseline (p = 0.003, see Table 2 and Fig. 2), indicating a significant decline in cool thermal sensitivity in UC at week 8. Both RA and SA showed significantly higher cool THDT than UC at week 8 (p = 0.008 and p = 0.013 ,respectively, Fig. 2 and Table 3), indicating better cool thermal sensitivity. However, cool THDT did not differ between RA and SA at week 8 (p = 0.790). No significant differences were observed between the three arms at week 12. Warm THDT did not significantly change from baseline within the RA and SA arms, but significantly increased in the UC arm from baseline to week 8 (p = 0.046), indicating a decline in warm thermal sensitivity in UC arm at week 8.
There were no significant between-arm differences in warm THDT.
Thermal induced pain was measured under both cold and heat conditions. Within the RA and SA arms, neither the cold-induced nor heat-induced pain thresholds significantly changed from baseline, but both thresholds significantly increased within the UC arm from baseline to week 12 (p = 0.008 and p = 0.048, respectively), suggesting increased perception to cold-induced pain and decreased perception to hot-induced pain in UC. In between-arm comparisons, there was a significantly lower cold pain threshold in RA when compared to UC at week 12 (p = 0.021, Table 3) due to increased cold pain threshold in UC arm, but there were no other statistically significant between-arm differences in the cold-induced and heat-induced pain thresholds.

Discussion
In this manuscript, we present QST assessment at baseline, after eight weeks of acupuncture treatment, and at a fourweek post-treatment follow-up in a randomized, sham-and usual care-controlled acupuncture for CIPN clinical trial.
Our results demonstrate different QST patterns in RA, SA, and UC. After eight weeks of RA, we observed significant improvements in the vibrational detection threshold in feet and cool thermal detection threshold in hands compared to UC. However, there are no statistically significant differences between real and sham acupuncture in QST outcomes. QST has been used to assess pain sensitivity and mechanistic differences in several clinical pain populations, including peripheral neuropathy pain [7,[28][29][30][31][32][33][34][35][36][37][38][39][40]. Hershman et al. demonstrated that increases in CIPN symptoms were significantly associated with a worsening vibration detection threshold, suggesting correlation between CIPN and sensory loss to vibration [9]. Our previous retrospective study suggested that patients with persistent CIPN had more severely impaired sensory perception when compared to patients without CIPN, and those QST outcomes correlated with PROs, suggesting QST's potential to evaluate CIPN [11,12].
To the best of our knowledge, this is the first study using QST to characterize CIPN phenotype during an acupuncture intervention. The baseline QST values in all three arms are consistently within the ranges of our prior QST measurements, suggesting QST reproducibility in cancer survivors with persistent CIPN. Our results did not detect significant changes in tactile detection, either in hands or feet, in any of the three arms at any time point. Tactile detection is controlled by the mixed function of nociceptive afferent Aβ, Aδ, and C fibers. Preclinical animal studies suggest that acupuncture works by stimulating Aβ, Aδ, and C fibers via needle insertion, but the QST tactile detection threshold measurement is not very sensitive to the smaller C fiber function changes [41].
Vibration thresholds were not different in the hand but were significantly lower in the feet in both RA and SA when compared to UC (p = 0.019 and p = 0.046, respectively) at week 8; significant differences were not observed between RA and SA (p = 0.637). There were no significant differences between the three arms at week 12 in the feet. It is For each outcome, estimates are derived from a linear mixed model with baseline means constrained to be equal across study arms, reflecting the pre-randomization timing of the baseline assessment. The dependent variable vector included the pre-randomization baseline (week 0) assessment, as well as the post-randomization assessments at weeks 8 and 12. The independent variables were treatment arm, week (categorical), and the arm-by-week interaction. A patient-level random intercept was included in the model to account for the repeated outcome measurements within patients over time CI confidence interval, TDT tactile detection threshold, VDT vibrational detection threshold, THDT thermal detection threshold, THT thermal threshold +, p < 0.10, *p < 0.05, **p < 0.01, ***p < 0.001 important to note that the significant difference at week 8 between UC and the other arms is due as much to the UC increase (3.15 points) as it is to the decrease in the RA (3.7 points) and SA (2.5 points) arms. At week 12, the UC and SA arms returned to baseline, but the decrease in RA was maintained. This finding is consistent with our prior observation and literature indicating that the lower extremities are more sensitive to vibration loss, which might be due to lengthdependent Aβ fiber damage in CIPN [42]. This would explain why VDT differences were seen in the feet, but not in the hands. Furthermore, results in the UC arm suggest that without intervention, sensory thresholds might continue to worsen over time, but that an acupuncture intervention could potentially slow the decline, if not improve the thresholds.
Most importantly, this is the first time that we have reported QST assessment in both real and sham acupuncture in cancer survivors with persistent CIPN with the aim to differentiate true effects from real acupuncture to sham acupuncture. Despite finding no significant differences between RA and SA at week 8, we did observe slightly better VDT in RA alone that was maintained at four weeks post-acupuncture follow-up (Fig. 1). The placebo effect phenomenon has been well described as a key limitation in acupuncture research, and objective measurement such as QST might be useful to assess RA treatment response. These changes observed in the UC arm may represent hyperalgesia and dysregulation to thermal stimuli due to continued nerve damage. Certainly, the statistical changes in the UC arm are also possibly due to our small sample size, and the absolute changes Table 3 Between-arm differences in changes from baseline For each outcome, the estimates of between-arm differences in changes from baseline are derived from the constrained linear mixed models presented in Table 2 RA real acupuncture, UC usual care, SA sham acupuncture, CI confidence interval, TDT tactile detection threshold, VDT vibrational detection threshold, THDT thermal detection threshold, THT thermal threshold +, p < 0.10, *p < 0.05, **p < 0.01, ***p < 0.001

Outcome
Week RA-UC Mean (95% CI) in these thermal measures were quite small. These results are inconclusive and further studies are needed to further elucidate the QST pattern in patients with CIPN, as well as QST changes corresponding to RA or SA interventions. Our study has several limitations. First, the sample size is small, and it is a single-center study. One third of participants opted out of QST, which might lead to selection bias in our analysis. Moreover, in all 63 participants who completed QST assessments at any timepoint, 52 (82.5%) completed the assessments at two timepoints, and 42 (67%) completed all three assessments, which may also lead to selection bias in our analysis. The parent trial primary outcome is CIPNrelated symptom changes at week 8, but not other CIPN symptoms such as numbness and tingling, which might be better characterized by QST. In addition, our study has a heterogeneous group of patients who received a variety of neurotoxic chemotherapeutics. This makes it challenging to interpret the QST data since it could vary among different races and chemotherapy agents. Lastly, QST assessments were exploratory endpoints of the primary study, and the statistical analyses are inadequately powered to evaluate them.
Despite these limitations, this is the first placebo-controlled randomized trial assessing the efficacy of acupuncture in alleviating persistent CIPN symptoms with the incorporation of QST. These preliminary differences in QST measurements reveal the perceived value and benefits of acupuncture from an objective perspective of sensorimotor perception loss. Our study is strengthened by well-balanced baseline characteristics between groups and consistency in assessment fidelity, as the same RSA administered all QST sessions. Here, we found that QST is a feasible and reliable tool to assess CIPN severity and may provide valuable information on treatment response to acupuncture. These preliminary findings can inform future clinical trials of adequate power to delineate QST's role as a correlative biomarker in CIPN research.

Conclusion
Our exploratory study showed that QST may be utilized to assess acupuncture treatment responses in addition to patientreported outcomes in patients with symptomatic CIPN. We Fig. 1 Model-estimated changes in Vibration Detection Threshold (VDT) means and 95% confidence intervals in feet in Real acupuncture, Sham acupuncture, and usual care arms found that real and sham acupuncture may hinder continuous declination of vibrational perception and cool thermal detection in patients with chronic CIPN. Further study of QST in assessing the effect of acupuncture treatment on CIPN-induced sensory loss and in differentiating real and sham acupuncture is needed.
Acknowledgements The authors would like to extend their gratitude to Patricia Chen for her assistance in conducting this study. The authors would also like to thank all the cancer survivors in the study for their participation.
Author contributions WIZ and TB contributed to conceptualization/ design. TB contributed to provision of study material or patients. DT and YZM contributed to collection and/or assembly of data. WIZ, REB, and TB contributed to data analysis and interpretation. WIZ, REB, DT, YZM, AK, SEH, and TB contributed to manuscript writing. WIZ, REB, DT, YZM, SEH, and TB contributed to final approval of manuscript.
Funding This work was supported in part by the National Institutes of Health/National Cancer Institute Cancer Center Support Grant (grant number P30 CA008748), the Translational and Integrative Medicine Research Fund at Memorial Sloan Kettering Cancer Center, and the Frueauff Foundation. The funding sources were not involved in the study design; collection, analysis, and interpretation of the data; writing of the report; or decision to submit the article for publication. Ting Bao is supported by the National Cancer Institute (grant numbers R37CA248563, R01CA240417, R01CA251470). W. Iris Zhi is supported by the Gateway for Cancer Research (grant number G-22-1200).

Data availability
The data underlying this article will be shared on reasonable request to the corresponding author.

Declarations
Conflict of interest We certify that there are no affiliations with or involvement in any organization or entity with any financial interest or other equity interests or non-financial interests that influenced the design, outcome, and submission of this study.
Ethical approval This study was approved by Memorial Sloan Kettering's Institutional Review Board and the clinical trial was registered at ClinicalTrials.gov (NCT03183037). All procedures performed in this study involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.
Informed consent Informed consent was obtained from all individual participants included in the study.