3.1. Participant Characteristics
At baseline, the 136 participants were primarily married, non-Hispanic Caucasians with a mean age of 60.4 years. The majority of participants had advanced stage and high grade disease (). As seen in , levels of IL-6 dropped substantially from pre-surgery to one year (p<.001). Mean levels of fatigue also decreased from pre-surgery to one year (p=.03). However, global sleep disturbance decreased only minimally over this time period (p=.41), with means at one year remaining above the cutoff of 5, indicating sleep disturbance. Means for the mood composite also were relatively stable over time (p=.56). There were no differences between levels of global sleep or fatigue at pre-surgery in individuals with vs. without ascites (fatigue: F(1,130)=.52, p=.47; sleep: F(1,130)=1.98, p=.16). At pre-surgery, no significant differences were seen between women included in the one year analyses and those not included with respect to IL-6 (p=.84), global sleep (p=.33), fatigue (p=.90), or the modified sleep index (p=.20), controlling for stage, age, and BMI.
Means and standard deviations of biological and psychosocial measures at pre-surgery and one year.
3.2. Pre-Surgical Sleep and Fatigue
Prior to surgery, higher IL-6 levels were significantly correlated with greater sleep disturbance, worse sleep index scores, and greater fatigue (all p values <.01), though IL-6 was not significantly associated with the mood composite (r=.04, p=.67). Regression models indicated that peripheral IL-6 was associated with significantly higher global sleep scores (β=0.21, p=.015) and greater fatigue (β=.18, p=.032), adjusting for covariates. Secondary analyses indicated that IL-6 significantly predicted modified sleep scores (β=0.19, p=.03). ().
Regression Models predicting Global Sleep and Fatigue from IL-6 – Pre-surgery
Because of the significant positive associations between sleep, fatigue, and IL-6 and a significant positive association between the modified sleep index and fatigue (β=.29, p<.001) we assessed whether the relationship between IL-6 and fatigue was mediated by sleep, as lack of sleep might affect daytime fatigue. When the modified sleep index was included in the regression model for IL-6 and fatigue, the association between IL-6 and fatigue became non-significant (β=.13, p=.13), although the drop in standardized beta weights was not particularly large, suggesting that sleep partially mediated the association between IL-6 and fatigue. ().
Regression Model predicting Fatigue from IL-6 controlling for Revised Sleep - Pre-surgery
3.3. Sleep and Fatigue at One-Year Follow-up
One year analyses were conducted on 63 participants who had completed adjuvant treatment, did not have documented disease progression, and were not on chemotherapy at the time of follow-up. IL-6 levels, global sleep, the sleep index score, and fatigue were all significantly associated at one year (all r ’s > .30, all p values < .05), whereas IL-6 was not significantly associated with the mood composite (r=.06, p=.70). All analyses controlled for pre-surgery values of the dependent variables (n=63 for sleep; n=60 for fatigue). At one year, regression models indicated that higher levels of IL-6 were significantly associated with greater sleep disturbance (β=.23, p=.02), adjusting for covariates. (). Secondary analyses with the revised sleep index showed similar results, though the relationship was non-significant (β=.18, p=.09). IL-6 was positively related to fatigue (β=.12, p=.20) adjusting for covariates including pre-surgical fatigue, but this relationship was not significant.
Regression Models Predicting Fatigue and Sleep from IL-6 – One Year
3.4. Associations of Changes in Sleep, Fatigue, and IL-6 Over Time
To examine the relationships between changes in IL-6, fatigue and sleep over time, analyses were conducted on the subset of patients who had completed measures of fatigue, sleep, and had levels of IL-6 at both time points. Controlling for stage, age, BMI, baseline sleep and one year mood, percent change in IL-6 over time was significantly associated with change in sleep over time (β=.27, p=.01). This indicates that, for example, a 10% decrease in IL-6 would correspond with a 0.13 point improvement in the global sleep change score, suggesting a minimal effect, although statistically significant. (). Secondary analyses excluding 2 outliers were consistent with these findings (β= .21, p=.049). In parallel analyses, including similar covariates and pre-surgery fatigue, percent change in IL-6 over time was associated with change in fatigue over time but findings did not reach significance (β= .22, p=.06). (). Secondary analyses excluding the two IL-6 outliers were consistent with these findings (β= .22, p=.053).
Graph of relationship between Percent Change in IL-6 and Change in Global Sleep from pre-surgery to one year. Y-axis values reflect unstandardized residuals controlling for stage, age, BMI, pre-surgical sleep, and mood at one year.
Regression Models Predicting Change in Sleep and Fatigue from Percentage Change in IL-6 Over Time