The sex-specific influence of body mass index on the association between serum lipids and cognitive function in Chinese elderly population: data from 2009-2015 CHNS

Objectives: Previous studies reported that the association between lipid levels and cognitive function related with sex, age and specific cognitive domains, but the influence of body mass index (BMI) on this association is limited. This triggered our interest in exploring how serum lipids relate to cognitive function in different BMI levels. Methods: Data was collected from 2009 wave and 2015 wave of China Health and Nutrition Survey (2009-2015 CHNS). Multivariable linear regression analyses examined serum lipids level as predictors of gender-specific measure of cognitive function in different BMI levels, which were adjusted for age, nationality, BMI, SBP, DBP, smoking status, alcohol consumption and education level. Results: Cognitive function score have different concentration curves in serum lipids quartile levels in different BMI categories. After adjustment for confounding factors, serum TG was positively associated with cognitive function score in underweight (β±SE: 2.06±0.88, P=0.023) and obese (β ±SE: 1.44±0.71, P=0.045) male group, and serum HDL-C was positively associated with cognitive function score in overweight (β±SE: 1.89±0.92, P=0.041) and obese female group (β±SE: 5.04±1.62, P=0.002). Conclusions: The associations of serum lipids with cognitive function differed between different gender groups and were specific to BMI levels. This result indicated that better nutritional status has superior cognitive function performance. 3.54 kg/m 2 , were included in this study. We observed sex-specific associations between serum lipids and cognitive function in different BMI categories and found that serum TG was positively associated with cognitive function score in underweight (β ± SE: 2.06 ± 0.88, P = 0.023) and obese (β ± SE: 1.44 ± 0.71, P = 0.045) male group, and serum HDL-C was positively associated with cognitive function score in overweight (β ± SE: 1.89 ± 0.92, P = 0.041) and obese female group (β ± SE: 5.04 ± 1.62, P = 0.002).

Compared with men, women had a higher prevalence rate of mild cognitive impairment (Wang et al. 2019) and a worse ability of naming (Li et al. 2017). In a study among of Chinese older adults aged 90 years and over (Huang et al. 2009), the prevalence of cognitive impairment was 57.8% overall and up to 67.2% among women. Such inconsisitencies might be attributable to sex differences in the lipid profiles, etiology of dementia and methods of cognitive assessment (Lu et al. 2017).
Previous studies speculated that associations of serum lipids with cognitive performance is inversely U-shaped or J-shaped (Lu et al. 2017). These associations implied that participants with extremely low or high levels of serum lipids would be expected to show poorer cognitive performance. The possible pathophysiological mechanisms of high serum lipids with the risk of cognitive decline is that high level of cholesterol can contribute to an overproduction and accumulation ofβ-amyloid in the brain (Hughes et al. 2014). And a possible explanation for low serum lipid levels with the risk of cognitive decline is the nutritional status of participants. Participants with malnutrition show alterations in the energetic profile as weight loss, reduced caloric, increased energy requirement, low lipid levels and cognitive function impairment (Oreopoulos et al. 2009; Reitz et al. 2004).
Two reviews concluded that high serum cholesterol associates with cognitive function and the association is strongly age-dependent(van Vliet 2012; van Vliet et al. 2009). But the associations between serum triglyceride (TG) and cognitive function were complicated and related with BMI and sex (Table 1). Yin, Z. X. et al. (Yin et al. 2012) found that high normal plasma TG was association with preservation of cognitive function while lower concentrations were not in the Chinese oldest-old. Two studies reported that lower TG was associated with higher cognitive scores in most cognitive domains (Corley et al. 2015) and better short-term memory (Kanoski and Davidson 2011). Lv, Y. B. et al. (Lv et al. 2016) found that TG was associated with MMSE score, but after adjustment for central obesity and other confounding factors, it was not associated with the risk of cognitive impairment. Most of research only focused on the total population and did not considered the influence of BMI on these associations, leading to some ambiguity with respect to potential BMI differences in serum lipids and cognitive function. So these inconsistent results and potential relationships triggered our interest in exploring the sex-specific influence of BMI on the association between serum lipids and cognitive function in Chinese elderly population.

Methods Setting
In 1989, the CHNS started and was intended to represent a range of economic and demographic tests for testing memory performance. The first question asked about memory status: "How is your memory?" The response categories were "very good", "good", "OK", "bad" and "very bad". Those who reported "bad" or "very bad" were defined as having a poor memory. The second question asked about changes in memory status: "In the past twelve months, how has your memory changed?" The response categories were "improved", "stayed the same" and "deteriorated". Those who reported "deteriorated" were defined as self-reported memory decline.
The following four tests were related to cognitive function on specific memory tasks. Four tests were administered in the following order: 1) the first was a word list memory test for immediate memory, in which an examiner read a list of 10 unrelated words at 2-second intervals and immediately asked the participant to repeat to them as many words as possible in any order (score 10); 2) the following two tests were mind control ability tests in which an examiner counted backward from 20 to 1 (score 2) and calculated 100 minus 7 and subtracted 7 again and again (score 5); and 3) the last test was a test for delayed memory, in which a list of words were repeated to an examiner after a period of time (score 10). An orientation test was not included in the analysis as it was only assessed in 2015 wave.
The cognitive function score was used to assess memory performance, which was the sum of the scores of the four tests and could range from 0-27 points. The Cronbach alpha internal consistency coefficient of this scale was 0.73, which is above the acceptable cutoff of 0.70.

Study population
In

Baseline characteristics of the study participants
The baseline characteristics of the participants were described separately for different gender groups ( Other characteristics were not significantly different. The influence of BMI on the association between serum lipids and cognitive function Sex-specific associations between serum lipids and cognitive function in different BMI categories were shown in Fig. 2. Participants were divided into four groups according to quartile values of serum lipids. Cognitive function scores have different curves in different BMI categories. Multivariable linear regression analyses were performed separately to examine the sex-specific association between serum lipids and cognitive function in different BMI categories (Table 3). After adjustment for confounding factors, serum TG was positively associated with cognitive function score in underweight (β ± SE: 2.06 ± 0.88, P = 0.023) and obese (β ± SE: 1.44 ± 0.71, P = 0.045) male group, and serum HDL-C was positively associated with cognitive function score in overweight (β ± SE: 1.89 ± 0.92, P = 0.041) and obese female group (β ± SE: 5.04 ± 1.62, P = 0.002).

Discussion
This study investigated the sex-specific influence of BMI on the association between serum lipids and cognitive function in Chinese elderly population for the first time. A total of 2246 participants (1120 men and 1126 women), aged 65.64 ± 7.52 years (range, 55-94 years) with a BMI of 23.98 ± 3.54 kg/m 2 , were included in this study. We observed sex-specific associations between serum lipids and cognitive function in different BMI categories and found that serum TG was positively associated with cognitive function score in underweight (β ± SE: 2.06 ± 0.88, P = 0.023) and obese (β ± SE: 1.44 ± 0.71, P = 0.045) male group, and serum HDL-C was positively associated with cognitive function score in overweight (β ± SE: 1.89 ± 0.92, P = 0.041) and obese female group (β ± SE: 5.04 ± 1.62, P = 0.002).
Accumulating evidence suggests that lipid levels are associated with cognitive function and dementia, but the results showed that the association between lipid levels and cognitive function might be In this study, we observed sex-specific associations between serum lipids and cognitive function.
Serum HDL-C was positively associated with cognitive function score in overweight and obese female group. Two prospective studies on metabolic syndrome have examined cognitive decline in women specifically, finding no significant association between low HDL-C or high TG and decline on global We also found that serum TG was positively associated with cognitive function score in underweight and obese male group. Serum lipids are considered to be biomarkers of malnutrition (Dana et al. 2009) and serum TG may be a meaningful indicator of nutrition status (Weir et  The strengths of this study relate to large sample size of Chinese elderly population to focus on the sex and BMI-specific associations of serum lipids with cognitive function. There were several limitations of our study. First, the number and type of cognitive function tests performed at this study were limited and were not likely to provide particularly precise individual cognitive function score, such as an orientation test was not included in the analysis as it was only assessed in 2015 wave.
Second, we had only 1 time point of measurement of serum lipids and assessment of cognitive function. Longitudinal analysis of the temporal relationship between serum lipids levels and cognitive function was not possible. Third, we adjusted the data for multiple confounding factors but cannot exclude the possibility of residual bias due to unmeasured confounders. Fourth, because this was an observational study, our findings could have been confounded by indication bias and cannot show direct causality of BMI and serum lipids with cognitive function. Last, although participants who use lipid-lowering agent were excluded, but we cannot exclude participants with hormone treatment.

Conclusions
In conclusion, we found the relationship between serum lipids and cognitive function was related with BMI levels. Serum TG was positively associated with cognitive function score in underweight and obese male group, and serum HDL-C was positively associated with cognitive function score in overweight and obese female group. Such analytical method needs to be further replicated and it could benefit from corresponding analysis.   The influence of BMI on the association between serum lipids and cognitive function