Concordance between Self-reported Symptoms and Clinically Ascertained Peripheral Neuropathy among Childhood Cancer Survivors: the St. Jude Lifetime Cohort Study

Five-year survival after treatment for childhood cancer exceeds 85% (1). However, children with cancer experience acute toxicities during treatment. Some toxicities, including chemotherapy-induced peripheral neuropathy (CIPN; ref. 2), are not always transient, and may persist throughout life (3, 4). Clinical evaluation of a large group of survivors of pediatric onset hematologic and solid malignancies as well as central nervous system tumors identified peripheral neuropathy among 21.9% of survivors who were, on average, 25 years from diagnosis (5). When investigating self-reported symptoms in the same cohort, 17% endorsed impaired motor function and movement problems, and 34.2% sensory abnormalities (6). Persistent peripheral neuropathy is concerning because of its impact on physical function, limiting mobility, and participation in daily activities, which can eventually impact quality of life (QOL; refs. 7–10). Thus, identification of signs and symptoms of motor or sensory deficits among childhood cancer survivors (CCS) is important, as referral for interventions to remediate or teach survivors to compensate for impairments has potential to impact participation in daily activities and improve QOL.

The best way to determine the presence of CIPN among survivors is a comprehensive clinical evaluation that includes motor and sensory testing. However, such evaluations are not always feasible in the context of an overall medical visit because they require specific resources and are time consuming (11, 12). Simple screening measures are needed to identify those in need of referral for specialized testing and potential treatment. While studies have been conducted to assess concordance between self-report and clinical testing–based diagnoses of neuropathy in breast cancer survivors (13, 14), comparison studies to determine agreement between self-reported symptoms and clinically ascertained CIPN among adult survivors of childhood cancer have not been published. The overall goal of the current study was to evaluate the concordance between self-reported peripheral neuropathy compared with clinically ascertained neuropathy in adult survivors of childhood cancer, to see if self-reporting of symptoms accurately identified persons with clinically ascertained neuropathy. Further, we explored the host characteristics associated with misclassification of peripheral neuropathy by self-report among adult survivors of childhood cancer to determine if there is a subset of survivors who can only be identified with comprehensive clinical measurement.

Participants were members of the St. Jude Lifetime Cohort Study (SJLIFE), a study that aims to facilitate longitudinal evaluation of health outcomes among adult survivors of childhood cancer treated at St. Jude Children's Research Hospital (SJCRH; ref. 15–17). Participants complete an in-person clinical assessment with laboratory testing and imaging, and five questionnaires that query sociodemographic characteristics and health status, psychosocial well-being, health behaviors, sexual and reproductive health, and diet (Fig. 1). For these analyses, survivors were ≥5 years from their initial diagnosis, ≥18 years of age, and completed both self-reported and clinical assessments characterizing neuropathy by June 30, 2017. Persons eligible who died prior to participating and those invited but who had not yet had an opportunity to participate were not included. This analysis also excluded participants with diabetes mellitus, amputation, hemiplegia, and peripheral vascular disease. Study documents were approved by the SJCRH institutional review board. Participants provided written informed consent prior to assessment.

Outcomes: Clinically ascertained peripheral neuropathy was considered present in persons whose examination indicated a grade 2 or higher adverse event according to the NCI Common Terminology Criteria for Adverse Events (CTCAE) version 4.03 (18). Signs and symptoms of neuropathy were assessed with the Modified Total Neuropathy Score (mTNS), a measure that evaluates both sensory and motor integrity with scores ranging from 0 (no neuropathy) to 24 (severe neuropathy; ref. 19). These data were combined with performance data to determine CTCAE Grades. Grade 2 sensory neuropathy [peripheral sensory neuropathy (PSN)] was assigned to participants with at least one mTNS-based sign or symptom of sensory impairment and either a sensory organization test composite score from computerized dynamic posturography of less than 70 (indicative of impending fall risk; ref. 20), or taking longer than 20 seconds to write a short sentence (indicative of fine motor impairment) from the physical performance test (21). Grade 2 motor neuropathy [peripheral motor neuropathy (PMN)] was assigned to participants with at least one mTNS-based sign or symptom of motor impairment, and either ankle dorsiflexion or hand grip strength values at least 1.3 SDs below age- and sex-predicted values. Isokinetic ankle dorsiflexion was measured at 60 degrees per second with the participant sitting with full back, thigh, and knee support (Biodex System 4, Shirley, NY; ref. 22). Hand grip was measured with a hand-held dynamometer (Sammons Preston Roylan) with the participant sitting, the shoulder in neutral, the elbow in flexion, and the forearm midway between pronation and supination. (23).

Self-reported symptoms of peripheral neuropathy were determined using questionnaires modeled after the National Health Interview Survey (24), and used previously to characterize neuromuscular function in CCSs (25). Symptoms were quantified by a score generated from the answers on specific sensory and motor queries. To identify symptoms of sensory neuropathy, participants were asked: “Have you ever been told by a doctor or other health care professional that you have or have had: (i) decreased sense of touch or feeling in hands, fingers, arms, or legs; (ii) prolonged pain arms, legs, or back?; (iii) abnormal sensation in arms, legs, or back?”. If the participant answered “yes, and the condition is still present” then the participant was classified as having sensory symptoms. To identify symptoms of motor neuropathy, participants were asked: “Have you ever been told by a doctor or other health care professional that you have, or have had: (i) weakness or inability to move arm (s)?; (ii) weakness or inability to move legs?”. If the participant answered “yes, and the condition is still present” then the participant was classified as having motor symptoms.

Diagnosis and treatment information were abstracted from medical records by trained abstractors, and included diagnosis, age at diagnosis, chemotherapy exposures and doses, radiation exposures and doses, and surgery. Participant sex, age at assessment, race and ethnicity, smoking status, and alcohol use were obtained from self-report questionnaires. Smoking status classified as never, ever, or current smoker, and ever smokers were those who reported smoking at least 100 cigarettes in their lifetime (26). Risky/heavy drinkers were males who reported 5 or more drinks in a single day, or an average of 14 drinks per week, or females who reported 4 or more drinks in a single day, or an average of 7 drinks per week in the year prior to the questionnaire (27). Height and weight were measured with a wall mounted stadiometer (SECA) and an electric scale (Scale-Tronix), and obesity defined as a body mass index (BMI) of ≥30 kg/m²; ref. 28). Emotional distress was classified according to participant responses to the 18 item Brief Symptom Inventory (BSI-18), which asks respondents to rate their level of distress within the previous seven days. The BSI includes three subscales, anxiety, depression, and somatization (bodily symptoms not fully explained by any known medical condition), and a summary score, the Global Severity Index (GSI). The GSI and subscale scores were converted to t scores using sex-specific normative data. T scores ≥63 were considered relevant emotional distress (29). Final analysis included the three subscales. Cognition was evaluated with the Wechsler Abbreviated Scale of Intelligence (full-scale IQ; ref. 30), with those whose age-adjusted z score was ≤-2 considered impaired. Prevalent pain was defined as answering “yes” to: “Do you have pain in legs in past 4 weeks?” or “do you have pain in hands/arms in the past 4 weeks?”.

Descriptive statistics were used to characterize the study population, differences between participants and excluded survivors, and between male and female participants, were determined by χ² tests for categorical measures and t tests for continuous measures. Sensitivity (correctly reported presence PSN or PMN among those with clinically ascertained PSN or PMN), specificity (correctly reported absence PSN or PMN among those who did not have clinically ascertained PSN or PMN), and accuracy/total agreement (correctly reported positive and negative PSN or PMN from total reports of both tools) were estimated (31). Variations in the results of the two tests (self-report and clinical test) were evaluated by calculating Kappa coefficients (32). A kappa value less than 0.4 was considered poor-to-fair agreement, a kappa value of 0.41 to 0.60 moderate agreement, a kappa value of 0.61 to 0.80 substantial agreement, and a kappa value of 0.81 to 1.00 excellent agreement (33). Analyses were conducted for each reported symptom and the overall sensory and motor scores and stratified by sex. Generalized linear models with a binomial distribution and a log-link function (34) were used to estimate the extent that self-reported symptoms were associated with the likelihood of having clinically ascertained PSN or PMN, and to identify factors associated with reporting neuropathy symptoms when neuropathy was not clinically present, or with not reporting neuropathy symptoms when it was clinically present. Results are presented as Prevalence Rate Ratio (PRR) and 95% confidence intervals (CI). Analysis were done with and without inverse probability weights to account for the differences between participants and nonparticipants by demographic, diagnosis, and treatment factors (35, 36). Adding weights to the analysis did not change the results. Thus, the unweighted results are presented.

Among 3,509 potentially eligible study participants, 2,933 (83.6%) were included in the final analysis. Excluded survivors were 136 (3.9%) persons with amputation, 101 (2.9%) with peripheral vascular disease, 50 (1.4%) with diabetes, 5 (0.14%) with paralysis, 23 (0.65%) with missing clinical data, and 58 (1.6%) with missing questionnaires. An additional 203 (5.8%) were excluded from analysis because a proxy completed their questionnaires (Study Diagram). Compared with excluded survivors, participant survivors were more likely to be female (49.0% vs. 40.0%), younger at diagnosis [mean (SD), 8.4 (5.6) vs. 9.0 (6.0)], younger at assessment [mean (SD), 33.4 (8.9) vs. 36.4 (10.9)], and survivors of leukemia (37.4% vs. 26.6%). Participants were also less likely than excluded survivors to have been treated with carboplatin (5.5% vs. 8.3%), cisplatin (7.4% vs. 17.4%), and high-dose methotrexate (27.0% vs. 28.8%). Participants were less likely to be cognitively impaired (full scale IQ 3.2% vs. 20.7%, verbal impairment 6.3% vs. 26.6%, nonverbal impairment 4.3% vs. 21.6%) than excluded survivors. Participants compared with excluded survivors were less likely to report pain in past 4 weeks in legs (29.8% vs. 36.2%, respectively) or hands/arms (13.2% vs. 17.7%, respectively) (Table 1).

Female survivors were more likely than male survivors to have clinically ascertained peripheral motor (PMN; 8.4% vs. 5.6%, P = 0.004) but not sensory (PSN; 4.7%, 4.2%, P = 0.591, respectively) neuropathy. Females were also more likely than males (22.1% vs. 17.7%, P = 0.003) to report prolonged pain in arms, legs, or back weakness or inability to move arms (6.1% vs. 3.4%, P < 0.001; Table 2).

Tables 3a and 3b show the degree of concordance between clinically ascertained peripheral neuropathy and self-reported symptoms of peripheral neuropathy. Among females, for PSN the most sensitive, specific, and accurate self-reported symptom was prolonged pain in arms, legs, or back (42.4%, 78.9%, and 77.2%, respectively). For PMN, the most sensitive, specific, and accurate self-reported symptom was weakness or inability to move legs (38.1%, 96.5%, and 91.7%, respectively). For having both PSN and PMN the most sensitive, specific, and accurate self-reported symptom was weakness or inability to move legs (58.6%, 94.7%, and 94.0%, respectively). For having either PSN or PMN the most sensitive, specific, and accurate self-reported symptom was prolonged pain in arms, legs, or back (45.2%, 80.7%, and 76.9%, respectively). After accounting for chance, agreement between self-reported, and clinically ascertained peripheral neuropathy was poor (Kappa: 0.057–0.389; Table 3a).

Among males, for PSN, the most sensitive, specific, and accurate self-reported symptom was prolonged pain in arms, legs, or back (44.3%, 83.4%, and 81.8%, respectively). For PMN, the most sensitive, specific, and accurate self-reported symptom was weakness or inability to move legs (35.4%, 96.8%, and 93.5%, respectively). For having both PSN and PMN the most sensitive, specific, and accurate self-reported symptom was weakness or inability to move legs (52.4%, 95.7%, and 95.1%, respectively). For having either PSN or PMN the most sensitive, specific, and accurate self-reported symptom was prolonged pain in arms, legs, or back (45.8%, 84.8%, and 81.6%, respectively). Agreement between self-report and clinically ascertained neuropathy was poor (Kappa: 0.058–0.334) among males after accounting for chance (Table 3b).

In models, stratified on sex and adjusted for age at assessment and race, survivors who endorsed decreased sense of touch or feeling in hands, fingers, arms or legs, weakness or inability to move arm(s), or weakness or inability to move legs, were more likely than those who did not endorse these symptoms to have clinically ascertained PSN or PMN. Survivors who endorsed two or more symptoms were six times more likely than those who endorsed fewer than two symptoms to have clinically ascertained PSN or PMN (PRR, 6.54, 95% CI, 4.50–9.51, P < 0.001, females; PRR, 5.24; 95% CI, 3.45–7.95, P < 0.001, males; Table 4).

The proportion of survivors classified incorrectly (false positives and false negatives) are shown in Supplementary Fig. S1 and Table 5,Table 5B. Females and males were at similar risk for misclassification (false positive 6.3% vs. 5.3%, P = 0.22, false negative 6.9% vs. 5.4%, P = 0.09). Among females, factors associated with risk for being classified as positive when neuropathy was not clinically present were older age at diagnosis, and somatization. Factors associated with risk for being classified as not having neuropathy when neuropathy was clinically present were older age at assessment, being black, somatization, and depression.

Among males, factors associated with risk for being classified as positive when neuropathy was not clinically present were current smoking, somatization, and having pain in legs in past 4 weeks. Factors associated with risk for being classified as not having neuropathy when neuropathy was clinically present was only older age at assessment.

This study provides a novel comparison of concordance between patient self-reported symptoms (by questionnaire) of peripheral neuropathy and clinically ascertained PMN and PSN among adult survivors of childhood cancer. We found poor concordance between self-report and clinically ascertained neuropathy. Survivors who endorsed ≥2 symptoms on the self-report questionnaire were six times more likely than those who endorsed fewer than two symptoms to have clinically ascertained either PSN or PMN. Age at diagnosis, emotional distress, and reporting pain in legs in the past 4 weeks were associated with an increased risk for symptom-based positive neuropathy classification when neuropathy was not clinically present. Race (White), age at assessment, and emotional distress were associated with increased risk for symptom-based negative neuropathy classification when neuropathy was clinically present.

The current study underscores the need to develop a more valid tool to accurately assess symptoms of peripheral neuropathy. Our finding of moderate sensitivity (range 23.8%–58.6%) indicates that the symptoms questions we used for self report do not accurately identify those with either PSN or PMN. This may result from questions that are too general, reflecting symptoms associated not only with neuropathy, but also with other chronic health conditions experienced by survivors (37), or because survivors whose symptoms have been present for years no longer recognize their impairments (8). On the other hand, our symptom questions accurately identified survivors without either PSN or PMN (specificity range 78.9%–97.7%). Because of the low overall accuracy of the symptom questions, excellent specificity here likely identifies only survivors with a lower overall burden of chronic disease.

In this cohort, we observed that female survivors who reported emotional distress were more likely than those without emotional distress to be misclassified as having either PSN or PMN when not clinically present. Studies have shown that anxiety and depression are highly prevalent among patients with chronic pain and neuropathic pain (38). For example, a study among colorectal cancer survivors an average of 5.6 years after diagnosis showed that survivors with high CIPN report higher rates of depression and anxiety (39). Our cross-sectional design limited our ability to determine temporal association between reasons for misclassification. The nature of our discovered associations is potentially bidirectional. That is, peripheral neuropathy can cause emotional distress, or emotional distress can result in somatosensory symptoms and/or limit motor and physical function. Among females, age at diagnosis was associated with increased risk for reporting symptoms of either PSN and PMN when the condition was not clinically present. This could be explained by differential recall bias by age as older survivors are more likely to remember the treatment experience and its effect on activities of daily living compared with those who were at younger at diagnosis. White female survivors were less likely to report symptoms of either PSN or PMN when clinically present. This can be explained by the higher prevalence of functional impairment among non-White cancer survivors and older adults (40, 41), that is not reported among White survivors. Older women were also more likely to deny the presence of either PSN or PMN when clinically present. Increased age is associated with decreased ability to perform daily activities and increased risk for falling (42). Thus, survivors may misclassify the neuropathy symptoms by attributing increase in functional limitation to the normal aging process.

Male survivors with somatization and leg pain were more likely than those without these symptoms to report neuropathy symptoms when not clinically present. This finding is likely due to the lack of specificity of our questions to detect neuropathy and may reflect the overall burden of pain and/or disease. Pain remains difficult to manage among cancer survivors and may be a symptom of more than one underlying cause (43). In addition, survivors may have difficutly discerning differences between symptoms of sensory or motor neuropathy and symptoms of emotional distress. These data support a need for better self-report questionnaires that consider the impact of emotional distress and pain on self-reported sensory or motor neuropathy among long-term survivors of childhood cancer.

This study is subject to potential limitations. First, we excluded subsets of survivors from this analysis based on functional limitations (amputation) or potential confounding comorbidities (peripheral vascular disease, diabetes, paralysis). The excluded survivors were more likely to be male, non-White, treated with platinum analogues and high-dose methotrexate, and cognitively impaired. Even those applying inverse probability weighting to account for differences did not impact results, these differences may have biased our estimates. Second, all participants in our study were diagnosed and treated at a single institution. Although the characteristics and treatment exposures for this population are similar to national protocols for patients with pediatric cancer, our finding may not be generalizable to individuals treated in other centers and with different treatment protocols. Third, participants who were included in the current study survived ≥5 years after diagnosis, survival bias may influence the prevalence for both self-report and clinically ascertained neuropathy and may not apply to populations of survivors closer to diagnosis. Fourth, these questions were not originally designed with sufficient specificity to determine the presence of either PSN or PMN in survivors of childhood cancer.

In conclusion, the symptom questions used in this study are a poor proxy for identification of CCS with clinically ascertained peripheral neuropathy, likely because our questions were not specific enough. Older female survivors who are black, or who have emotional distress, are at high risk for not being classified with peripheral neuropathy when it is truly present. These results emphasize the need for the development of a comprehensive questionnaire to more accurately assess peripheral neuropathy among adult survivors of childhood cancer.

S. Hayek reports grants from NIH and American Lebanese Syrian Associated Charities (ALSAC) during the conduct of the study. R. Dhaduk reports grants from NIH and ALSAC during the conduct of the study. W.E. Evans reports grants from NCI during the conduct of the study; in addition, W.E. Evans serves as Chair of Scientific Advisory Board Princess Máxima Centre and is a board member of BioSkryb. K. Bjornard reports grants from NIH and other support from ALSAC during the conduct of the study. C.L. Wilson reports grants from NIH during the conduct of the study. L.L. Robison reports grants from NCI during the conduct of the study. K.R. Krull reports grants from NCI during the conduct of the study. K.K. Ness reports grants from NIH and other support from ALSAC during the conduct of the study. No disclosures were reported by the other authors.

S. Hayek: Conceptualization, formal analysis, methodology, writing–original draft, writing–review and editing. R. Dhaduk: Formal analysis, methodology, writing–review and editing. Y. Sapkota: Writing–review and editing. W.E. Evans: Conceptualization, writing–review and editing. B. Diouf: Conceptualization, writing–review and editing. K. Bjornard: Writing–review and editing. C.L. Wilson: Writing–review and editing. M.M. Hudson: Resources, funding acquisition, writing–review and editing. L.L. Robison: Resources, funding acquisition, writing–review and editing. R.B. Khan: Conceptualization, investigation, methodology, writing–review and editing. D.K. Srivastava: Conceptualization, formal analysis, methodology, writing–review and editing. K.R. Krull: Writing–review and editing. K.K. Ness: Conceptualization, formal analysis, validation, investigation, methodology, writing–review and editing.

This study was supported by research funding from the NCI (U01CA195547, to M.M. Hudson; P30CA021765, to Charles Roberts) and by the ALSAC.

Tracie Gatewood (St. Jude Children's Research Hospital) assisted with manuscript preparation; there was no financial compensation outside of salary.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.