br Limitations There are some limitations to take into accou
Limitations There are some limitations to take into account when interpreting these results. Firstly, direct causation cannot be established in this cross-sectional study, and it may be the downstream effects of TNFα, such as increased HOAt production, which are more proximally related to cognitive dysfunction in BDI (Rosenblat et al., 2015). Second, as mentioned above, the study may have been underpowered to detect between group differences in serum cytokine levels due to the size of the healthy control group (n = 20). However, this between group comparison was not the main outcome of interest, which was instead the association between serum cytokine levels and cognition within early stage BDI patients. For this, we had a sample size of 46 patients. Sample sizes for previous studies examining the relationship between inflammatory cytokines and cognition in BDI patients with longer duration of illness range from n = 21 to n = 117 (see Rosenblat et al., 2015 for review). This suggests that our sample size of early stage BDI patients is in line with previous studies investigating similar issues in patients with later stage illness. Third, cytokine levels were not taken on the same day as neuropsychological testing for all patients. Previous studies have found, however, that inflammatory cytokine levels display good intra-individual reliability over periods of up to 2 years in both healthy and medically ill populations, and that a single measure is reasonably reflective of the level over several months (Ho et al., 2005; Hofmann et al., 2011; Lee et al., 2007; McKay et al., 2017; Navarro et al., 2012). Additionally, the fact that TNFα levels were negatively associated with multiple cognitive measures (global cognition, processing speed and working memory) in our sample suggests that this is not a spurious finding; as well, when BDI patients who were in a different mood state at the time of cognitive testing and serum cytokine sampling were excluded, the association of TNFα with global cognition (p = 0.024) and processing speed (p < 0.001) remained significant, and that with working memory remained at trend level (p = 0.095) significance. Previous clinical studies linking TNFα with decreased cognitive functioning in BDI patients with longer duration of illness, and preclinical studies finding that elevated levels of TNFα exert neurotoxic effects (see Discussion), also lend clinical and biological plausibility to the results of this study. Fourth, while cytokine samples were taken in the early afternoon, circadian rhythms influence inflammatory cytokine production and this diurnal variation may have been an additional confounding factor (Lange et al., 2010). Fifth, the significant positive association between YMRS scores and processing speed was an unexpected finding, but may be explained by some patients with subsyndromal manic symptoms displaying accelerated thought form and consequently faster processing speed. Lastly, while we attempted to control for practice effects in the analysis, repeat testing in some patients may be a confounding variable. However, this is to our knowledge the first examination of the association between inflammatory cytokines and cognitive functioning in a sample of early stage BDI patients. Thus, despite the limitations of the analysis and need for reproduction, the current finding that TNFα is independently associated with decreased cognitive functioning could serve as a starting point for future investigations in strategies to preserve cognition in newly diagnosed BDI patients.
Role of funding source The data for this manuscript was generated as a part of the Systematic Treatment Optimization Program for Early Mania (STOP-EM), which was supported by an unrestricted grant from AstraZeneca, Canada Study # DC-990-205. The sponsor was not involved in the design of the study, data collection, analysis/interpretation of data or manuscript preparation.