Risk prediction model for taxane-induced peripheral neuropathy (TIPN) in patients with early-stage cancer receiving taxane therapy: SWOG S1714.

person Meghna S. Trivedi Herbert Irving Comprehensive Cancer Center, Columbia University Medical Center, New York, NY info_outline Meghna S. Trivedi, Joseph M. Unger, Norah Lynn Henry, Amy Darke, Daniel Louis Hertz, Thomas Brannagan, Stephanie Smith, Bryan P. Schneider, William Johnson Irvin, Amanda Redden Hathaway, Amy C. Vander Woude, Vinay K. Gudena, Paula Anel Cabrera-Galeana, Mary Orsted, Michael Leo LeBlanc, Michael Jordan Fisch, Dawn L. Hershman

Authors person Meghna S. Trivedi Herbert Irving Comprehensive Cancer Center, Columbia University Medical Center, New York, NY info_outline Meghna S. Trivedi, Joseph M. Unger, Norah Lynn Henry, Amy Darke, Daniel Louis Hertz, Thomas Brannagan, Stephanie Smith, Bryan P. Schneider, William Johnson Irvin, Amanda Redden Hathaway, Amy C. Vander Woude, Vinay K. Gudena, Paula Anel Cabrera-Galeana, Mary Orsted, Michael Leo LeBlanc, Michael Jordan Fisch, Dawn L. Hershman Organizations Herbert Irving Comprehensive Cancer Center, Columbia University Medical Center, New York, NY, SWOG Statistics and Data Management Center/Fred Hutchinson Cancer Research Center, Seattle, WA, University of Michigan Rogel Cancer Center, Ann Arbor, MI, University of Michigan College of Pharmacy, Ann Arbor, MI, Columbia University Medical Center, NY, NY, Lewis Cancer & Research Pavilion at St. Joseph's Candler/GA NCORP, Savannah, GA, Indiana University, Indianapolis, IN, Bon Secours Saint Francis Medical Center Cancer Institute/Southeast Clinical Oncology Research (SCOR), Midlothian, VA, University Cancer & Blood Center, Athens, GA, Cancer & Hematology Centers of Western Michigan, Grand Rapids, MI, Cone Health Cancer Center, Greensboro, NC, Instituto Nacional de Cancerología, Mexico City, DF, Mexico, HealthPartners Cancer Center at Regions Hospital, St. Paul, MN, The University of Texas MD Anderson Cancer Center; Carelon Medical Benefits Management, Houston, TX Abstract Disclosures Research Funding NIH/NCI/NCORP Background: TIPN can impact the ability to complete cancer treatment as well as patient quality of life and functional status. To create a risk prediction model for TIPN, a prospective observational cohort study was conducted. Methods: S1714 enrolled patients > 18 years old with Stage I-III primary lung, breast, or ovarian cancer starting treatment with a taxane-based regimen. The primary endpoint was the occurrence of TIPN, defined as an increase of >8 points over baseline in the EORTC QLQ-CIPN20 sensory subscale score at any follow-up assessment through week 24. A 2-step training/test approach was used to develop a TIPN risk prediction model. With a 60% random sample of evaluable patients, best subset selection identified a best model from among a set of demographic, baseline comorbid, and treatment factors. Model building was based on logistic regression using K-fold cross-validation to minimize predictive error, assessed using logistic model deviance and concordance (c-Statistic). From the best identified model, a risk model was built by summing adverse risk factors and creating high vs low- risk groups by splitting at the median. The derived model was tested in the remaining 40% of evaluable patients. The aim was to detect a 15% absolute difference in TIPN between high vs low-risk groups. Results: Among 1336 enrolled patients, N = 1278 were evaluable: median age 55.2 years (range 23-85), 98.6% female, 11.9% Black/4.6% Asian/12.0% other race/11.3% Hispanic/Latino, 90.2% with breast cancer. Paclitaxel was administered to 60.2% and docetaxel to 39.8%; 98.5% planned full dose of taxane. The rate of TIPN was 62.0%. In the training set of N = 768 patients, adverse risk factors were receipt of paclitaxel (vs docetaxel); stage 2/3 (vs 1) disease; planned duration of taxane > 12 weeks (vs <12 weeks); diabetes, autoimmune disease, or moderate kidney disease (>1 vs none); and self-identification as Black, Native American, Pacific Islander, multiple race, or Hispanic ethnicity (vs non-Hispanic White or Asian). Patients with >2 factors (high risk; n = 501, 65.2%), compared to patients with 0 or 1 factor (low risk; n = 267, 34.8%), were more likely to experience TIPN (70.9% vs 48.7%, p < .001). In the test set (n = 510), TIPN was more common in the high vs low-risk groups (68.3% vs 50.9%; OR = 2.08, 95% CI, 1.43-3.04, p < .001), exceeding the target difference of 15%. In all patients, TIPN proportions by quartile of risk score were 39.5% (Q1), 53.3% (Q2), 65.3% (Q3), and 76.3% (Q4), with a nearly 5-fold increased risk of TIPN for those in the highest vs lowest quartiles (Q4 vs Q1, OR = 4.93, 95% CI, 3.17-7.68, p < .001). Conclusions: A limited set of demographic, baseline comorbid, and treatment factors can be used to predict TIPN risk and may help guide treatment decision making. Future work to refine risk prediction using biomarkers is ongoing. Funding: NIH/NCI/NCORP grant UG1CA189974 Clinical trial information: NCT03939481.

Clinical