Associations between self-reported diet during treatment and chemotherapy-induced peripheral neuropathy in a cooperative group trial (S0221)

Data were obtained from the DELCaP study, an ancillary study to a phase III therapeutic trial (S0221) led by the South West Oncology Group (SWOG) (ClinicalTrials.​gov NCT00070564). DELCaP was developed to assess diet and lifestyle data at multiple time points including baseline, during, and post treatment. Detailed recruitment procedures, randomization procedures, and inclusion criteria for participation in both S0221 and DELCaP have been described previously [11, 16]. Briefly, women enrolled in S0221 had a confirmed diagnosis of stage II or III invasive breast cancer and were randomized to one of four treatment arms. Each patient received either doxorubicin plus cyclophosphamide every 2 weeks with pegfilgrastim or weekly doxorubicin plus daily cyclophosphamide with filgrastim. Patients then either received 12 weekly cycles of paclitaxel or paclitaxel every 2 weeks with pegfilgrastim for six cycles. Upon enrollment into S0221, participants were contacted for participation in the DELCaP study. Specifically, informed consent was obtained during the consent for S0221, allowing research staff to contact potential participants for inclusion in the DELCaP study. This study was approved by the Institutional Review Board at Roswell Park Cancer Institute and all participating institutions that enrolled patients.

A total of 1468 participants participated in the DELCaP study; 1460 participants completed the baseline questionnaire including foods usually eaten over the last 12 months prior to cancer diagnosis, and 1234 patients completed the 6-month follow-up regarding diet during treatment. In order to allow for the consideration of repeated measures, 226 participants who did not provide information on at least 50% of foods in both the baseline and follow-up questionnaires or had a difference > 10 in the number of foods missing between baseline and follow-up were excluded from this analysis. An additional 340 participants who did not complete or had missing values for the baseline or follow-up Functional Assessment of Cancer Treatment Gynecologic Oncology Group—Neurotoxicity (FACT/GOG-Ntx) questionnaire were not included. The total study sample included 900 women in this analysis.

DELCaP questionnaires were administered at four time points throughout the study: at baseline, following completion of treatment, and annually for the following 2 years. A baseline questionnaire (Q1), administered at study enrollment but before treatment, included questions on race and ethnicity, menopausal status, height and weight, smoking history, alcohol consumption, a 110-item food frequency questionnaire (FFQ) including detailed questions regarding vitamin and dietary supplement use, and the FACT/GOG-Ntx scale to assess neuropathic conditions. The FFQ was adapted from the validated Vitamins and Lifestyle (VitaL) study [17]. A second questionnaire (Q2) was mailed to participants 6 months after registration to the trial, when chemotherapy should have been completed. For the purpose of the current analyses, we compared baseline to Q2 neuropathy symptoms to determine the change in neuropathy symptoms resulting from chemotherapy treatments and FFQ data from Q2 to determine diet during treatment (6-month recall).

For quality control purposes, data entry was performed twice by different research staff and compared for accuracy and resolution of discrepancies.

In the baseline questionnaire, participants were asked to indicate how often each food and beverage was usually consumed over the last 12 months prior to diagnosis. Food consumption frequency included never, 1 per month, 2–3 per month, 1 per week, 2 per week, 3–4 per week, 5–6 per week, 1 per day, and 2+ per day, with additional options for a small, medium, and large serving size for all foods. Beverage consumption frequency included never, < 1 per month, 1–3 per month, 1 per week, 2–4 per week, 5–6 per week, 1 per day, 2–3 per day, 4–5 per day, and 6+ per day with small, medium, and large serving size options. Standard medium serving sizes were provided for reference for both food and beverages. Missing food frequency values among those not missing > 50% of foods were treated as a value of 0 (not eaten). Missing serving sizes were assigned a default medium value. An aggregate monthly total for each food was computed by converting small, medium, and large serving sizes into values of 0.5, 1, and 1.5, respectively, which was then multiplied by the total monthly frequency. The monthly sums of specific foods were totaled to create food groups based on the University of Minnesota Nutrition Data System for Research and MyPlate food groups (Table 1) [18, 19]. Food groups included citrus fruits, other fruits, dark green vegetables, red/orange vegetables, starchy vegetables, cruciferous vegetables, beans/bean dishes, other vegetables, fish, poultry, red meat, processed meat, other protein, dairy, grains, sweets, fried foods, added fats, and alcohol. The follow-up questionnaire included the same FFQ and asked participants to indicate their diet over the previous 6 months (presumably reflecting diet during treatment) during the time period in which neuropathy symptoms would have developed or worsened. Data from this FFQ were used in analysis and modeled as ordinal variables.

At baseline and in the follow-up questionnaire, participants completed an 11-item FACT/GOG-Ntx scale to assess the severity of neuropathy symptoms during the previous 7 days. Symptoms including numbness in hands and feet, discomfort in hands and feet, joint and muscle pain, hearing and ear trouble, trouble feeling, and trouble walking were assessed on a 5-point scale ranging from 4 = “Not at all”, 3 = “A little bit”, 2 = “Somewhat”, 1 = “Quite a bit”, to 0 = “Very much”. Total baseline and 6-month follow-up scores were computed and used to determine the percentage decrease in neuropathy score, indicating a worsening of symptoms. Previous literature has determined that a 10% or greater decrease in FACT/GOG-Ntx score is clinically meaningful for assessment of neuropathy [10, 20, 21]. Scores were grouped into approximate tertiles based on whether the participant experienced no to minimal increase (< 10%), moderate increase (10–30%), or severe increase (> 30%) in severity of neuropathy symptoms and were modeled as an ordinal variable.

Demographic characteristics obtained from baseline questionnaires included age, race, height, weight, menopausal status, smoking status, highest education obtained, and marital status. Height and weight at baseline and follow-up were used to determine the BMI and whether the participant had changed weight during treatment. A chi-square test for independence and one-way analysis of variance were performed to test for differences in neuropathy score frequencies across groups. Food groups were categorized as tertiles of monthly servings during chemotherapy treatment and compared using Pearson’s chi-square. Food frequencies at baseline and follow-up were not independent and were predictive of the other time point (0.36 ≤ r ≤ 0.58), which is commonly observed in other studies of long-term reproducibility and considered to fall within an acceptable range for FFQs [22, 23]. Therefore, baseline FFQ data were not included in the analysis.

Ordinal regression was used to estimate odds ratios (ORs) and 95% confidence intervals (95% CIs) for the associations between each food group and change in neuropathy score. This modeling approach takes into account the natural order of the multiple category outcome, change in reported neuropathy score, or increasing severity. The ordinal model also assumed that the odds ratio of each outcome category was independent of other categories (proportional odds assumption) [24]. The proportional odds assumption was first tested for associations between the outcome and each food group as well as the final model. Study arm, age in years, self-identified race or ethnicity (non-Hispanic White, Spanish/Latino/Hispanic, Black/African American, other), BMI calculated from self-reported height and weight, change in weight from diagnosis to post treatment (lost weight, maintained weight, gained weight), menopausal status (pre, post), smoking status (never, former, current), highest level of education, and marital status were assessed as potential covariates. Details regarding randomization of participants and the dose and schedule of treatment have been described previously [16]. Variables that were significantly associated with both neuropathy category and at least two food groups at p ≤ 0.20 were included in the final adjusted model (age, race, BMI at baseline, and smoking status). The final adjusted model included a single OR for CIPN severity associated with each food group meeting the proportional odds assumption and was interpreted as the odds of reporting worse neuropathy at follow-up for increased consumption of each food group (in tertiles) [24]. Statistical tests performed were two-sided at α = 0.05. All analyses were performed using SAS 9.4 (SAS Institute, Cary, NC, USA).