Design, sample, and setting
The data for this secondary analysis were derived from a two-phase, longitudinal trial designed to explore the impact of a clinician decision support algorithm on clinicians’ documentation of CIPN assessment and management [22]. The sample consisted of 142 English speaking patients with breast or gastrointestinal cancer, or multiple myeloma who had received ≥ one infusion of neurotoxic therapy (e.g., taxanes, platinums, or proteasome inhibitors) at the time of consent and were scheduled to receive ≥ three more cycles of neurotoxic therapy. Patients were excluded if they had pre-existing neuropathy unrelated to cancer therapy. All participants were recruited from a National Cancer Institute-Designated Cancer Center. Approximately half of the sample (n = 70) participated in the usual care phase and the other half of the sample participated in the algorithm phase (n = 72). Clinicians received a CIPN assessment and management algorithm prior to each clinical visit for all participants during the algorithm phase. The following describes the measures, procedures, and statistical analyses pertinent to this secondary analysis of the measurement properties of the two PRO-CTCAE numbness and tingling severity and interference items.
Measures
PRO-CTCAE Numbness and Tingling Severity and Interference Items The PRO-CTCAE Measurement System is comprised of an item library with 124 PRO items that evaluate the presence, frequency, severity, or interference of 78 cancer treatment-related symptomatic adverse events [16]. The two PRO-CTCAE items that pertain to CIPN evaluate the severity at its worst and the associated interference of numbness and tingling in the hands and feet over the past seven days. PRO-CTCAE numbness and tingling item responses are scored from 0 to 4 with higher scores reflecting greater severity and interference, respectively [17,18,19].
5-Item Total Neuropathy Score – Clinical (TNSc©) The 5-item TNSc© [23,24,25] is a measure that includes patient self-report of sensory (i.e., numbness, tingling, and pain [burning, aching, stabbing]) and motor (e.g., difficulty buttoning or climbing steps) neuropathy symptom severity and/or location questions. Both of these questions overlap with the numbness and tingling severity and interference items of the PRO-CTCAE, respectively. These self-reports are integrated with examiner-administered tests of vibration sensibility, strength, and deep tendon reflexes. Items are scored from 0 to 4 with total scores ranging from 0 to 20 (higher scores reflect more severe neuropathy). Several studies support the reliability and validity of the TNSc© to capture peripheral neuropathy in adults undergoing neurotoxic cancer treatment [26,27,28].
0–10 Numerical Rating Scale (NRS) of Worst CIPN Pain Intensity Worst CIPN pain intensity over the past seven days was quantified using a 0–10 NRS [29, 30]. Higher scores on the NRS represent more severe CIPN pain intensity.
European Organization of Research and Treatment of Cancer QLQ-CIPN20 Sensory and Motor Subscales The QLQ-CIPN20 sensory subscale consists of nine self-report items that measure on a four-point scale the severity of neuropathic symptoms in the hands or feet over the past seven days, while the motor subscale consists of eight-items that measure on a four-point scale CIPN-related functional deficits over the past seven days. Each subscale is summed and linearly transformed to a score that can range from 0 to 100, where higher scores represent greater CIPN symptom severity [31]. The QLQ-CIPN20 sensory subscale contains four questions (i.e., numbness or tingling in the hands and/or feet, respectively) that are similar in nature to the one-item PRO-CTCAE numbness and tingling severity item. The QLQ-CIPN20 motor subscale contains five questions that are similar in nature (e.g., ask about specific functional limitations associated with CIPN such as walking, holding a pen, opening jars) to the one-item PRO-CTCAE numbness and tingling interference item. Evidence supports the internal consistency reliability (e.g., Cronbach’s alpha: 0.88 for sensory and motor subscales, respectively), concurrent validity, and responsiveness to change of the QLQ-CIPN20 sensory and motor subscales [10, 32]. In this study, the three-item autonomic symptom subscale was not administered due to its low item-item correlations with the rest of the QLQ-CIPN20 [10].
Procedures
Participants completed the PRO-CTCAE numbness and tingling items, 0–10 NRS of worst CIPN pain intensity, and QLQ-CIPN20 using an iPad at the cancer center before each of three consecutive clinical visits (i.e., T1, T2, T3) during neurotoxic therapy or up to approximately one month after neurotoxic cancer therapy completion. In addition, a trained member of the study staff administered the TNSc© at T3. Four study staff members, including the principal investigator (RK) administered the TNSc©. On the same day as the T3 visit, the PRO-CTCAE numbness and tingling items were administered again (T4) to evaluate test–retest reliability. The T4 surveys were administered after the provider visit while the participants were in the waiting room prior to receiving chemotherapy or when the participants were actively receiving chemotherapy. At the conclusion of study-related procedures, study staff abstracted cancer treatment-related information from participants’ medical records.
Statistical analyses
Data from participants in both study phases were pooled for this psychometric analysis. Due to the dose-dependent nature of CIPN, unless otherwise specified, all analyses were evaluated using data from T3, the time point when participants were anticipated to have received the greatest cumulative dosage of neurotoxic therapy. Descriptive statistics were calculated, and distributions were inspected for proportions at floor and ceiling. Floor and ceiling effects were calculated by frequencies and proportions of respondents reporting the lowest or highest possible scores.
Test–retest reliability of PRO-CTCAE numbness and tingling items between T3 and T4 was calculated in a subset of participants using the intraclass correlation coefficient (ICC) based on a two-way mixed effects analysis of variance model with interaction for the absolute agreement between single scores [33]. Values less than 0.50, between 0.50 and 0.75, between 0.75 and 0.90, and greater than 0.90 were interpreted as poor, moderate, good, and excellent reliability, respectively [34]. Longitudinal validity refers to how sensitive a measure is in detecting the real underlying change in symptom severity over time [35]. To evaluate longitudinal validity, a Cohen’s d effect size was calculated reflecting changes from T1 to T3 in PRO-CTAE numbness and tingling items, 0–10 NRS of worst CIPN pain intensity, and QLQ-CIPN20 sensory and motor subscale scores.
Concurrent validity between the PRO-CTCAE numbness and tingling items, and the QLQ-CIPN20 sensory and motor subscales, the 0–10 NRS of worst CIPN pain intensity, and the TNSc© was assessed using Spearman’s correlation.
Construct validity of the PRO-CTCAE numbness and tingling item response categories was evaluated with a one-sided Jonckheere-Terpstra Test [36] for ordered differences. The purpose of this analysis was to determine if there was monotonic ordering of scores on the QLQ-CIPN20 sensory and motor subscale, 0–10 NRS of worst CIPN pain intensity, or TNSc© across the item response categories for the PRO-CTCAE severity (0, 1, 2, and ≥ 3, respectively) and interference (0, 1, and ≥ 2, respectively) items. PRO-CTCAE severity item scores ≥ 3 and PRO-CTCAE interference item scores ≥ 2 were collapsed as few individuals had scores at or near the top of the scoring range (0–4).
Sensitivity refers to a screening measure’s ability to accurately identify a patient with the disease, while specificity refers to a screening measure’s ability to accurately identify a participant without the disease (higher values of both metrics are desirable) [37]. The sensitivity and specificity of the PRO-CTCAE numbness and tingling items were evaluated at cutoff levels of zero for each item in comparison to the reference definition of “probable sensory peripheral neuropathy” that is used for diabetic peripheral neuropathy [38]. The definition of probable sensory peripheral neuropathy requires participants to have two of the following signs or symptoms: neuropathic symptoms (e.g., numbness, tingling, or pain), decreased distal sensation, and/or decreased ankle reflexes [38]. Thus, participants with TNSc© Sensory scores ≥ 1 and either TNSc© Sensory Function: Vibration Sensibility scores ≥ 1 or TNSc© Reflexes scores ≥ 1 were classified as exhibiting probable sensory peripheral neuropathy. The PRO-CTCAE numbness and tingling items also were evaluated for their sensitivity and specificity to detect painful CIPN at cutoff levels of zero for each item. For the reference measure, 0–10 NRS of worst CIPN pain intensity scores ≥ 4/10 were interpreted as indicating the presence of painful CIPN, while scores < 4/10 indicated the absence of painful CIPN [39]. All analyses were conducted with R Statistical Software and evaluated at a significance level of 0.05.