This single center, observational and cross-sectional study was designed to assess CIPN in patients who were treated by bortezomib-based chemotherapy for multiple myeloma. The primary objective was the assessment of prevalence and severity of sensory CIPN. The secondary objectives were the severity of motor CIPN, the prevalence of neuropathic pain, the use of pain medications, the prevalence of anxiety and depression, and the HRQoL. Patients were assessed once and no longitudinal assessment was performed.

The study was designed according to the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines (von Elm et al., 2007). The study protocol was registered on the ClinicalTrials.gov website NCT03344328. The study was anonymous and approved by a local ethics committee (Comité de Protection des Personnes sud-est 5, IRB: 6705, No. 2017-A00651–52, april 25, 2017). The participants’ consent was obtained orally by contacting each eligible patient by phone.

The study design, the recruitment of patients and the data analysis were performed by the authors M.S., N.K. and D.B., none of whom were involved in the management of the patients included.

This single center study was conducted in the rheumatology and clinical hematology departments of the University Hospital of Clermont-Ferrand (CHU Clermont-Ferrand). The inclusion of patients and data collection were carried out from March 13, 2019 until April 02, 2019.

Inclusion criteria were patients having been treated by bortezomib-based chemotherapy for multiple myeloma. Exclusion criteria were patients <18 years, and neurological disease (stroke, Parkinson disease, Alzheimer disease, fibromyalgia).

Eligible patients were identified from the database of the chemotherapy prescription software (CHIMIO^®, Computer Engineering, France) of the University Hospital of Clermont-Ferrand. A specific algorithm was prepared for the systematic extraction of patient data. Thereafter, this first selection of patients was checked by the oncologists in charge of patients, to select patients according to inclusion and exclusion criteria. Patients were phoned to inquire whether they would participate in the study. After patient acceptance, a paper questionnaire and a stamped envelope for the response were sent to the patient. Patients returned their questionnaires to the University Hospital of Clermont-Ferrand, where their responses were recorded and analyzed.

The primary endpoints were the sensory score of the QLQ-CIPN20 rated from 0 (least) to 100 (worst) (Postma et al., 2005) (for scoring see: https://www.eortc.org/app/uploads/sites/2/2018/02/SCmanual.pdf), used as a quantitative variable, and the sensory CIPN defined as a sensory QLQ-CIPN20 score of ≥30/100 (Alberti et al., 2014; Selvy et al., 2020b), used as a qualitative variable. The secondary endpoints were the motor and vegetative scores of the QLQ-CIPN20 rated from 0 (least) to 100 (worst) (Postma et al., 2005). Present pain was reported by the patient (yes/no). Screening for pain was assessed with a visual analogue scale (VAS, 0 = no pain and 10 = maximum imaginable pain), and defined for a threshold ≥4/10. Among patients with a positive screening for pain (VAS score ≥4/10), neuropathic pain was assessed with DN4 interview (French abbreviation: Douleur Neuropathique 4, for neuropathic pain 4) questionnaire and detected for a DN4 interview score ≥3/7 (Bouhassira et al., 2005). The history of neuropathy and neuropathic pain before the diagnosis of multiple myeloma was also recorded. Ongoing pain medications in the past month were recorded, based on an established list of pain medications. Anxiety and depression were assessed with the Hospital Anxiety and Depression Scale (HADS) at the time of the answer, considering the following thresholds, normal (total score ≤7), borderline or suggestive of possible anxiety/depression (total score of 8–10) and indicative of anxiety/depression (total score ≥11) (Zigmond and Snaith, 1983). The HRQoL was assessed with the QLQ-C30 and QLQ-MY20 questionnaire (EORTC) (Aaronson et al., 1993; Cocks et al., 2007). The scoring of QLQ-C30 and QLQ-MY20 was done according to EORTC recommendations. The QLQ-C30 was divided into 3 subscales with a Global health status (0 worst to 100 best), the functional scales (0 worst to 100 best for physical functioning, role functioning, emotional functioning, cognitive functioning and social functioning) and the symptom scales (0 least to 100 worst for fatigue, nausea and vomiting, pain, dyspnea, insomnia, appetite loss, constipation, diarrhea and financial difficulties). The QLQ-MY20 was divided into 2 subscales with symptom scales (0 least to 100 worst for disease symptoms and side effects of treatment) and functional scales (0 worst to 100 best for body image and future perspective).

The oncological characteristics were recorded such as cumulative dose (mg/m²), route of administration (intravenous, subcutaneous), the date of the last bortezomib administration, the duration of bortezomib treatment, the administration of thalidomide, the date of the last thalidomide treatment, and the hematopoietic stem cell transplantation.

The socio-demographic characteristics of patients were also recorded such as gender, age, daily use of cigarettes, occasional alcohol use and hazardous alcohol use (>10 alcohol units per week) (France, 2017). Tobacco use has been associated to the severity of oxaliplatin-related CIPN in a previous study of our group (Selvy et al., 2020b). Alcohol use is a debated risk factor of CIPN (Molassiotis et al., 2019a).

Data assessing CIPN, neuropathic pain, ongoing pain medications, anxiety, depression, and HRQoL were obtained from the completed questionnaire. Oncological data and patient characteristics were obtained from the software of chemotherapy prescription and patient medical records. All the data were recorded and managed using REDCap™ electronic data capture tools hosted at the University Hospital of Clermont-Ferrand (Harris et al., 2009).

The sample size was determined to ensure that the confidence interval (CI) of the sensory score of the QLQ-CIPN20 had an accuracy of around 5 points for a standard-deviation at 20. The calculation showed that at least 65 patients were necessary to ensure a two-sided type I error of 5%.

The internal consistency of the QLQ-CIPN20 sensory scale was assessed and determined using Cronbach’s α coefficient, with a minimum accepted value of 0.70. Then, the categorical data were presented using number of patients, percentage, and appropriate 95% CI. Continuous data were expressed as mean and standard-deviation. The normality of the data was assessed using the Shapiro–Wilk test. Continuous data were compared between independent groups (such as no sensory CIPN vs. sensory CIPN) using the Student’s t-test or the Mann–Whitney U test when the assumptions of the t-test were not met. The homoscedasticity of the data was assessed using the Fisher–Snedecor test. The results were expressed using Hedge’s effect-size (ES) and 95% CI, and were interpreted according to the rules of thumb reported by Cohen (Cohen, 1988), who defined the ES bounds as small (ES = 0.2), medium (ES = 0.5), and large (ES = 0.8). Categorical data were compared between groups (independent proportions) using the chi-squared test or Fisher’s exact test, and McNemar test for paired proportions. To analyze the relationships between continuous parameters, Pearson and Spearman correlation coefficients were estimated according to the statistical distribution of variables and by applying Sidak’s type I error correction, and interpreted as: <0.2 negligible, 0.2–0.4 weak, 0.4–0.7 moderate, >0.7 strong (Altman, 1999). To determine factors associated with the sensory CIPN (dependent variable), multivariable analysis (i.e. generalized linear logistic regression) was performed, including patients’ characteristics (gender, age, tobacco and alcohol consumptions) and characteristics of chemotherapy (time since last bortezomib administration, cumulative dose of bortezomib, and thalidomide treatment). Particular attention was paid to the study of multicollinearity and to the interactions between covariates: 1) studying the relationships between the covariables, and 2) evaluating the impact of adding or deleting variables on a multivariable model. The results are expressed as odds-ratios and 95% CI, and forest plots were used to present the results. Statistical analyses were performed using Stata 15 (StataCorp, College Station, US). All the tests were two-sided, with a type I error set at 5%. In accordance with the literature (Rothman, 1990; Bender and Lange, 2001; Feise, 2002), we reported all individual p-values without systematically applying any mathematical correction to the aforementioned tests comparing groups. Specific attention was given to the magnitude of differences (i.e., ES) and clinical relevance.

One hundred and fifteen patients were screened by oncologists for inclusion in the study. Among them, 74 patients accepted to participate to the study and 67 sent back a filled questionnaire (response rate: 90.5%) (Figure 1). The characteristics of these 67 included patients are presented in Table 1.

The 20 items of the QLQ-CIPN20 indicated an excellent level of internal consistency (Cronbach α = 0.92), the 9 items of the sensory scale a good level (Cronbach α = 0.82), the 8 items of the motor scale indicated a good level (Cronbach α = 0.86) and the 3 items of the vegetative scale indicated a poor level (Cronbach α = 0.65). Thereafter, the vegetative scale of the QLQ-CIPN20 was not used for the analysis.

Among the patients analyzed, 26.9% 18) (95% CI 16.7; 39.1) had a sensory CIPN (sensory QLQ-CIPN20 score ≥30/100). The distribution of the sensory scores of the QLQ-CIPN20 over the years after the end of chemotherapy is presented in Figure 2. Sensory scores were not different between males and females (20.1 ± 19.9 vs. 17.8 ± 20.8, p = 0.51) and not correlated with the age of patients (Table 2).

Characteristics of bortezomib treatments (cumulative dose, duration, and time since last administration) were not different between patients with or without a sensory CIPN (Table 1). Likewise, sensory scores were not correlated with the cumulative bortezomib dose, with the duration of bortezomib treatment, or with the time since last bortezomib administration (Table 2 and Figure 2). The proportions of sensory CIPN were significantly different according to the route of bortezomib administration (p = 0.003) (Table 1). Post hoc analysis revealed a higher proportion of sensory CIPN in patients with both routes of bortezomib administration (intravenous + subcutaneous) compared to the subcutaneous route only (p < 0.05). Interestingly, the proportions of sensory CIPN were not different between intravenous and subcutaneous routes. Similarly, the sensory scores of the QLQ-CIN20 were different according to the route of bortezomib administration (intravenous vs. subcutaneous vs. both routes: 21.7 ± 28.9 vs. 15.8 ± 17.7 vs. 35.2 ± 18.8, p = 0.02).

Proportions of patients with a sensory CIPN and sensory scores (20.0 ± 22.4 vs. 17.5 ± 18.3, p = 0.95) were not different between patients treated or not with thalidomide (Table 1). The durations of thalidomide treatment were not different between patients with or without a sensory CIPN (Table 1), and were not correlated with the sensory scores (Table 2). Times since last thalidomide administration were longer for patients with a sensory CIPN than patients without a sensory CIPN (Table 1), but were not correlated with the sensory scores (Table 2). Moreover, proportions of patients with a sensory CIPN and the sensory scores were not different between patients having received a hematopoietic stem cell transplantation or not (for proportion see Table 1, and 17.8 ± 21.6 vs. 20.2 ± 18.5, p = 0.42, respectively). Finally, among patients with a sensory CIPN, tingling and numbness were proportionally higher in feet than in hands (p < 0.05) (Figure 3).

In parallel, a multivariable analysis of the sensory CIPN was performed on associated factors (male, age, tobacco, alcohol, time since last bortezomib administration, bortezomib routes, and thalidomide treatment). Bortezomib administration via intravenous + subcutaneous routes was associated with a higher proportion of patients with a sensory CIPN, compared to other routes (intravenous route only or subcutaneous route only). Thalidomide treatment was associated with a higher proportion of patients with a sensory CIPN compared to no thalidomide treatment (Figure 4).

Motor scores of the QLQ-CIPN20 were strongly correlated with sensory scores (Spearman coefficients: 0.74, p < 0.05), were higher for patients with a sensory CIPN than patients without a CIPN (36.7 ± 23.9 vs. 10.9 ± 14.2, p < 0.001), were not different between males and females (15.1 ± 19.7 vs. 20.4 ± 21.5, p = 0.3), and were not correlated with age of patients (Table 2). Motor scores were weakly correlated with the duration and the cumulative dose of bortezomib treatment, but not with the time since last bortezomib administration (Table 2). Motor scores were not different between patients having been treated by thalidomide or not (18.7 ± 20.4 vs. 17.0 ± 21.9, p = 0.65), and were not correlated with the time since last thalidomide administration, or the duration of thalidomide treatment (Table 2). Lastly, motor scores were not different between patient having received a hematopoietic stem cell transplantation or not (15.2 ± 20.0 vs. 21.9 ± 21.8, p = 0.11).

History of neuropathy and neuropathic pain was similar between patients with or without sensory CIPN. However, this data must be interpreted cautiously because patients tended to mix past neuropathy history, actual CIPN and multiple myeloma symptoms.

Among the patients analyzed, 59.7% (40) declared having present pain (yes/no), 25.4% 17) were screened positively for pain (pain VAS ≥4/10), and 14.9% (10) for neuropathic pain (pain VAS ≥4/10 and DN4 interview ≥3/7). Among patients with a sensory CIPN, 50.0% 9) were screened positively for pain, and 44.4% 8) for neuropathic pain; the latter proportion was higher than for patients without a sensory CIPN (4.1%, p < 0.001). Sensory and motor scores were higher for patients with a neuropathic pain than patients without (Sensory: 46.5 ± 21.7 vs. 14.1 ± 15.7, p < 0.001; Motor: 38.8 ± 19.8 vs. 14.2 ± 18.7, p < 0.001).

Finally, 52.0% of all the patients analyzed and 66.7% of patients with a sensory CIPN declared they took pain medications (Table 3). Most of the patients both in the sensory CIPN and neuropathic pain groups, declared to take paracetamol (75.0 and 62.5%, respectively), which was also similar to patients without sensory CIPN or neuropathic pain. No patient received duloxetine, one patient with a sensory CIPN received gabapentin and another one pregabalin. It is noteworthy that two patients without sensory CIPN took pregabalin (Table 3).

Proportions of anxiety and depression were higher in patients with sensory CIPN than those without (Figure 5). Sensory scores were higher among patients with anxiety or depression disorders (normal vs. suggestive vs. indicative scores of anxiety: 13.2 ± 15.4 vs. 34.2 ± 22.2 vs. 37.1 ± 27.3, p < 0.001; normal vs. suggestive vs. indicative scores of depression: 11.4 ± 15.9 vs. 34.5 ± 22.0 vs. 27.1 ± 18.8, p < 0.001). The same results were observed for motor scores (normal vs. suggestive vs. indicative scores of anxiety: 12.0 ± 16.2 vs. 34.1 ± 23.3 vs. 35.9 ± 24.8, p < 0.001; normal vs. suggestive vs. indicative scores of depression: 9.9 ± 15.4 vs. 26.5 ± 20.2 vs. 38.2 ± 23.3, p < 0.001).

The scores of sensory CIPN were moderately correlated with the scores of quality of life of the QLQ-C30 questionnaire for several dimensions (global health status, physical functioning, role functioning, emotional functioning, cognitive functioning, social functioning, pain and insomnia) (Table 4). Moreover, the scores of these dimensions were significantly different between patients with a sensory CIPN and those without (Table 4). The scores of sensory CIPN were moderately to strongly correlated with the scores of all the dimensions of the QLQ-MY20 (disease symptoms, side effects of treatment, body image and future perspective) and the scores of these dimensions were different between patients with a sensory CIPN and those without (Table 4).

The scores of motor CIPN were moderately correlated with the scores of each dimension of the QLQ-C30 questionnaire (except: nausea and vomiting, dyspnea, constipation, diarrhea, and financial difficulties, which had weak correlations) (Table 4). The scores of motor CIPN were moderately to strongly correlated with the scores of all the dimensions of the QLQ-MY20 (Table 4).

The use of the QLQ-CIPN20 questionnaire to assess the prevalence of CIPN may be controversial. In the present study, based on a work by Alberti et al., we used QLQ-CIPN20 scoring to approximate the prevalence of sensory CIPN, considering a QLQ-CIPN20 threshold of ≥30/100 to approximate a grade ≥2 sensory CIPN (Alberti et al., 2014). Alberti et al. showed a close relation between QLQ-CIPN20 scores and NCI-CTCAE sensory grade (p < 0.001). QLQ-CIPN20 scores between 30 and 40 (median ≈ 35, interquartile range ≈ 26–50, and mean ≈ 39) were associated with an NCI-CTCAE sensory neuropathy grade 2 and QLQ-CIPN20 scores >40 (median ≈ 59, interquartile range ≈ 39–62, and mean ≈ 57) were associated with an NCI-CTCAE sensory neuropathy grade 3/4 (Alberti et al., 2014). The QLQ-CPIN20 scoring was able to discriminate NCI-CTCAE neuropathy grade 1 vs. grade 2 (p < 0.001), and grade 2 vs. grade 3/4 (p < 0.001). However, the QLQ-CIPN20 scoring was not able to discriminate neuropathy grade 0 vs. grade 1 (p = 0.53). Le-Rademacher et al. concluded that there are no QLQ-CIPN20 score ranges that correspond directly with NCI-CTCAE grading levels (Le-Rademacher et al., 2017). However, they also emphasized that the QLQ-CIPN20 provided detailed information, distinguished more subtle degrees of neuropathy, and that it was more responsive to change over time than the NCI-CTCAE (Le-Rademacher et al., 2017). Importantly, the sensory score of the QLQ-CIPN20 of the present study (18.9 ± 20.3) were very close to those of the studies of Mendoza et al. (17.2 ± 4.7) (Mendoza et al., 2020) and Beijers et al. (15.3 ± 16.7) (Beijers et al., 2017). In our study, the self-administered questionnaire was particularly useful to assess CIPN severity using a paper questionnaire sent to patients. Neuropathy history was initially recorded but data were excluded from the analysis because the patients tended to mix past neuropathy history and ongoing CIPN. However, neuropathy history would have been of interest, because it has already been described as a risk factor of CIPN (Molassiotis et al., 2019b). Selection bias may be present in the study, since the patients came from a single center. However, the patients were managed in two different medical departments of the university hospital of Clermont-Ferrand. It is also possible that neuropathic patients are over-represented because these patients have felt compelled to respond to the questionnaires. Information bias was probably present, since the patients’ answers were subjective and unsupported by clinical assessment (such as neurological examination, nerve conduction studies, quantitative sudomotor axon reflex test or skin biopsies). Moreover, PRO measures may overestimate CIPN prevalence as they include symptoms that may have pre-existed the chemotherapy (Molassiotis et al., 2019b). Although the oncological data came from the medical prescription software of the university hospital.