• 2019-07
  • 2019-08
  • 2019-09
  • 2019-10
  • 2019-11
  • 2020-03
  • 2020-07
  • 2020-08
  • br Cytokine Measurement br Blood samples were drawn in


    2.2. Cytokine Measurement
    Blood samples were drawn in EDTA test tubes on the morning of the assessment visit, usually between 0700 and 0900 h. All samples were centrifuged, aliquoted and frozen at −70 °C within 60 min of collection. Cytokine immunoassays tests were performed in duplicate on singly-thawed Epirubicin HCl samples. Human V-plex electrochemiluminescence im-munoassays from Meso Scale Discovery (Gaithersburg, Maryland), all from a single lot number, were read on a SECTOR® Imager 2400. Upper and lower limits of detection (LLOD) as well as coefficients of var-iation of each cytokine assay are shown in Supplemental Table 1.
    2.3. Sample Size Calculation
    No formal sample size calculation was performed for the cytokine substudy since it was not the primary endpoint of the larger study [29,30] and there was limited published literature to allow a proper calculation.
    2.4. Statistical Analysis
    Our primary outcomes were change in fatigue (from the FACT-F) and global QOL (from the QLQ-C30). Cytokine values were rescaled such that a one-unit change in the specific cytokine represents a one-standard deviation (SD) change in that cytokine. Cytokines with 20% or more missing data (due primarily to plate failure) were excluded from the analysis (IL-1α, IL-2, IL-4, and IL-13). Remaining cytokine values below the LLOD were imputed with the LLOD for that cytokine. In a sensitivity analysis we assumed those values were zero and reran our models.
    Please cite this article as: S.M.H. Alibhai, H. Breunis, J. Matelski, et al., Age-related cytokine effects on cancer-related fatigue and quality of life in acute myeloid leukemia, J Geriatr Oncol,
    For each outcome, a bi-directional Akaike Information Criterion Table 1
    (AIC) step selection procedure was applied to a multiple regression Baseline patient demographics and disease characteristics.
    model including all available covariates (all measured cytokines, age, Baseline characteristic
    gender, hemoglobin, visit number, Charlson Comorbidity Index score,
    body mass index, and race). This multiple regression model focused
    Range, y
    on examining the association between cytokines and PROMs, and Gender, n (%)
    allowed us to include patient data at a given time point as long as Male
    both cytokines and PROMs were available at that time point. Standard Female
    English as First Language, n (%)
    errors and p-values were computed using robust standard errors to ac-
    count for repeated measurements per patient. Parameter estimates for
    continuous predictors are reported Epirubicin HCl as expected change in outcome per Married
    one-SD change in the predictor. Overall model quality was assessed Divorced/Separated
    using the adjusted R2 statistic, which assesses the amount of variability Single
    in the outcome variable (e.g. CRF) Triassic Period is explained by the variability in Widowed
    the predictor(s). The proportion of outcome variability explained by White
    each selected cytokine was assessed using the metric described by Black
    Lindemann et al. [44]. Multicollinearity (particularly among individual Oriental
    South Asian
    cytokines with similar mechanisms of action, e.g. pro-inflammatory)
    was assessed first by examining correlations between individual cyto-
    Working situation, n (%)
    kines and subsequently using Variance Inflation Factors during multi- Still at Work
    variable modelling, with a threshold of ten to identify high Retired
    multicollinearity [45]. To explore the variability in specific cytokines Sick Leave
    driving behavioral responses over time, we included cytokine-time in-
    teraction terms for the top three cytokines. Other
    In addition to reporting parameter estimates, the final models were Smoking status, n (%)
    used to compute expected CRF and QOL for different combinations of Current Smoker
    Quit Smoking
    cytokine values. First, we varied the most influential cytokine from its
    Never Smoked
    mean value to 10% of its mean value while holding the remaining cyto-
    ECOG performance status, n (%)
    kines and other covariates at their means. This allowed us to predict
    what might happen if an intervention (e.g. pharmacological agent, exer- 1
    cise) could substantially reduce levels of a single cytokine. Second, we 2
    co-varied all step-selected cytokines over their respective distributions,
    Mean Karnofsky Score % (SD)
    taking into account the direction of association with the outcome (e.g. if Cytogenetic risk group, n (%)
    cytokine A had a positive association with CRF and cytokine B had a neg- Favorable
    ative association with CRF, cytokine B was reverse-coded). This allowed