Niacin, lutein and zeaxanthin and physical activity have an impact on Charlson comorbidity index using zero-inflated negative binomial regression model: National Health and Nutrition Examination Survey 2013–2014

The study sample consists of one continuous cycle (2013–2014) of NHANES, which used a stratified multistage probability cluster design to be representative of the civilian, noninstitutionalized U.S. population, conducted by the National Center for Health Statistics, under the Centers for Disease Control and Prevention [27, 28]. A total of 10,175 individuals participated in NHANES during 2013–2014, but 5769 adults who were 20 years of age or older were restricted to our study. We excluded participants who were pregnant or lactating (n = 104). Besides, adults with caloric intakes of < 500 kcal or > 5000 kcal per day were excluded (n = 60). Then, we excluded 1219 participants due to missing covariate information, leaving 4386 eligible subjects for the study. Study protocols for NHANES were approved by the National Center for Health Statistics ethnics review board (https://​www.​cdc.​gov/​nchs/​nhanes/​irba98.​htm). All the participants signed the informed consent before participating in the study. The dataset is available at https://​wwwn.​cdc.​gov/​nchs/​nhanes/​continuousnhanes​/​default.​aspx?​BeginYear=​2013.

The study included a number of covariates evaluated as potential confounding factors, such as age (20–39; 40–59; 60 and above); gender (male, female); race/ethnicity (Hispanic, Non-Hispanic White, Non-Hispanic Black, Other Race); educational attainment (less than high school; high school, including general equivalent diploma; college or higher); income ($0–$24,999, $25,000–$54,999, $55,000–$74,999, $75,000 or above); smoking (yes, no); drinking (yes, no).

In addition, physical activity was also incorporated into the paper. The Physical Activity Questionnaire is based on the Global Physical Activity Questionnaire [29], and the questions were asked using the Computer-Assisted Personal Interview software. The main outcomes of physical activity were defined using the following questions: (1) Sedentary activity: “How much time do you usually spend sitting on a typical day?”; (2) Vigorous work activity: “Does your work involve vigorous-intensity activity that causes large increases in breathing or heart rate like carrying or lifting heavy loads, digging or construction work for at least 10 minutes continuously?” (yes, no); (3) Moderate work activity: “Does your work involve moderate-intensity activity that causes small increases in breathing or heart rate such as brisk walking or carrying light loads for at least 10 minutes continuously?” (yes, no); (4) Moderate recreational physical activity: “Do you do any moderate-intensity sports, fitness, or recreational activity that cause a small increase in breathing or heart rate such as brisk walking, bicycling, swimming, or volleyball for at least 10 minutes continuously?” (yes, no); (5) Vigorous recreational physical activity: “Do you do any vigorous-intensity sports, fitness, or recreational activity that cause large increases in breathing or heart rate like running or basketball for at least 10 minutes continuously?” (yes, no).

NHANES subjects were asked for two averaged 24-h recalls of dietary intakes using the USDA’s Automated Multiple-Pass method [30]. The first dietary recall was conducted by trained dietary interviewers, they gathered detailed information on all the foods and beverages participants consumed in the past 24 h, and a second recall was administered by telephone 3 to 10 days later. The total dietary nutrients taken through food and beverages are averaged over two days, nonetheless, if one did not complete the second dietary interview, only the first interview was used. Dietary supplements were also evaluated by the information on the number of days they were taken and the amount that were taken in the past 30 days. What is more, the average intake for 30 days is the total amount of supplement. Furthermore, total dietary nutrients intake was evaluated by foods and supplements.

Forty-nine chronic diseases (including infectious diseases, like HBV and HCV) were included in the study, which were self-reported using the NHANES questionnaire. Chronic HBV and chronic HCV are considered as a global public health problem, which usually lead to liver cirrhosis and HCC [9‐11], so they were included in this study. Diagnosed diseases were defined by positive responses to one or more questions such as: “Have you ever been told that you have the illness?” and “Are you now taking some pills to control your disease?”. People who do not report whether they suffer from a certain disease are considered healthy, which means the disease condition is assigned a value of zero. And we did this for the following reasons. Firstly, those who “do not know” the answer are most likely to have never been diagnosed by patients, because they do not have the symptoms of certain disease. Thus, we consider them do not suffer from such disease. What is more, the occurrence of chronic diseases is a slow progress. And it takes a long time from the change of body to the appearance of symptoms, which brings us to keep a conservative estimate. Meanwhile, this also avoid reducing the sample size by excluding the participants with missing values, so as to improve the representativeness and reliability of the study. The comprehensive health status of a participant was measured by the CCI [31], which is shown in Additional file 1. Derived from adding up the scores of forty-nine diseases, CCI evaluates a participant’s severity of comorbidity and the mortality in 10 years. The 10-year mortality (Y) is calculated as follows: Y = e ^ (CCI ∗ 0.9), which can be obtained at website https://​fpnotebook.​com/​prevent/​Exam/​ChrlsnCmrbdtyInd​x.​htm#fpnContent-panel-id_​6.

Figure 1 demonstrates that the distribution of CCI is non-normal count data and the variance is greater than the mean, which could be deemed as over-dispersed data. Simultaneously, there are too many zero counts presented in Fig. 1, and NB model unable to model CCI well. CCI was assumed as an over-dispersed and zero-inflated count variable, so we utilized ZINB model to appropriately model it [24, 32]. The model examined the association of the exposure variables, in this case dietary nutrients and physical activity, with the outcome variable (CCI) in two parts. Specifically, the logit part models CCI equals zero (structural zero), and count part models CCI equals to zero (sampling zero) and positive counts. Seven covariates including age, gender, race, education, income, smoking and drinking were adjusted for both in the logit part and count part.

In our study, we found the higher the total niacin intake, the higher the risk of death. Although there are limited studies that directly look at the association between niacin and CCI, the relationship between niacin and certain disease plays the role of indirect explanation. Similar to our findings, Park et al. [33] reported that there is a positive correlation between niacin intake and basal cell carcinoma risk, the authors also indicated that the higher the total niacin intake, the higher the risk of melanoma risk in men, but not in women. This also provides new ideas for our future research, because we did not analyze gender differences. Niacin promote the synthesis of sex hormones, and affect the development of melanocytes, and higher estrogen exposure increases the risk of skin melanoma [34, 35]. Although high doses of niacin is used as a cholesterol-lowering drug clinically, side effects on the body are also obvious, such as causing vasodilation, skin flushing, headache and hypotension [36, 37]. However, the intakes of niacin always accompanied with other B group vitamins or other nutrients, and this might mask the opposite relationship. In addition, in this study, niacin intake is 37.1 ± 1.34 mg averagely presented in Table 2, which marginally exceeds recommended allowances of nutrients [38]. And it is easily accumulated in the body because of the properties of fat-soluble, which might be partially reasons of adverse relationship between niacin and mortality. So, the results may be due to chance. Thus, future studies are need in this area.

Our study showed a negative correlation between lutein and zeaxanthin intake and 10-year mortality. Although it’s hard to find studies that directly look at the correlation between lutein and zeaxanthin and CCI, the relationship between lutein and zeaxanthin and certain disease plays the role of indirect explanation. In this way, our result is in agreement with plenty of earlier studies. One recent research suggested that the more lutein and zeaxanthin consumed, the lower risk of colorectal cancer [39]. In addition, Heinen et al. [40] indicated that lutein and zeaxanthin also plays a crucial role in skin cancer. Moreover, researchers demonstrated that lutein and zeaxanthin may reduce the risk of cancer types especially breast and lung cancers, they also indicated its function of suppressing heart disease and stroke [41]. Its impact on preventing and controlling diseases might attribute to the following mechanisms. Firstly, acting as anti-oxidant, they reduce oxidative damage indirectly by absorbing light [42, 43]. Secondly, early studies indicated that lutein and zeaxanthin may protect against inflammation by inhibiting the increase of oxidation-induced cytokines and upregulating the expression of inflammation-related genes [44]. Besides, as anti-oxidants, zeaxanthin effectively remove water-soluble and lipid-soluble peroxyl radicals [45]. Furthermore, animal studies suggested that lutein and zeaxanthin could reduce the adverse effects of inflammatory cytokines and low density lipoprotein on blood vessels [46, 47]. These reports all support our finding.

Our study did not find any associations between other nutrients intake and CCI. Consistent with our report, researchers of a cohort study reported that vitamin K intake was not associated with mortality [48], and Vinceti et al. [49] indicated that increasing selenium intake through diet and supplement may not prevent cancer. Besides, a systematic review and meta-analysis of prospective cohort studies demonstrated that dietary choline intake have no effect on incident CVD, and authors also find the association between dietary choline intake and CVD mortality was not statistically significant [50]. Table 3 shows the 95% confidence interval (CI) of OR and RR values for dietary fiber, vitamin E, copper, etc. is very large. This may be attributed to the insufficient sample size, leading to the instability of results. However, some of our results differed from several others. Authors of a prior study of middle and older-aged adults aged 40 and older years (NHANES, 1988–1994) found the relationship between vitamin D status and cardiometabolic mortality [51]. Our results may differ from others due to age difference, and dietary recall is less accurate than a biological sample. In addition, in contrast to a cohort study of Spanish graduates’ findings of correlations between vitamin C intake and cardiovascular mortality [52], we found no relationship between vitamin C intake and mortality, which may be attributed to ethnic factors. Furthermore, the differences in our study results and those of other studies might be partially due to the potential confounding factors and measurement differences. Furthermore, statistical model plays a crucial role as well.

We concluded that individuals who have higher sedentary time and do not have vigorous work activity will be more likely to have higher rate of death. And people who do not have vigorous recreational activity or moderate recreational activity have the risk of death. Previous studies have reported that there is a negative correlation between health status and sedentary time, which is associated with chronic diseases and overall mortality [53, 54]. For adults and the elderly, inactivity in leisure time and high television time increase the incidence of chronic diseases [55]. Moreover, a research from NHANES 2009–2012 pointed out that increasing time spent on physical activity will lead to lower risk of chronic diseases [56]. However, some researchers suggested that high activity level could weaken the increased risk associated with excessive sedentary time [57]. Importantly, this provide us with a novel idea that our study could consider the interaction between physical activity and sedentary time in further studies.

There are several strengths of this work. Firstly, including a representative sample with a large size, and NHANES employs thorough quality assurance procedures, which ensure the validity of the data. Secondly, wide ranges of potential confounders such as demographic characteristics and behavior elements were controlled to provide a better estimate of the association between nutrient intake and physical activity and CCI. Thirdly, we applied CCI to systematically evaluate one’s 10-year mortality, for one may have various diseases at the same time. Lastly, this study is the first to explore the associations between nutrients intake and physical activity and CCI by using ZINB model. However, several limitations of this study should be noted. Firstly, some of the participants were excluded due to the missing data, and we have a conservative assignment for CCI. Secondly, a dietary recall is less accurate than a urine sample, which may result in a false estimate of nutrients intake. Furthermore, we cannot infer a causal relationship because this was a cross-sectional study. Lastly, the count part includes sampling zero, and the logit part includes structural zero, but ZINB model cannot distinguish the causes of their generation well and may result in biased estimates.