Examining the Associations of Smoking Behavior and Obesity Among Older Adults in China: Should We Consider Food Consumption Behaviors?

Abstract

Objective: This research investigates the associations between smoking frequency and body mass index (BMI) among older adults in China with and without the inclusion of food consumption behaviors. Method: Applying three waves of the Chinese Longitudinal Healthy Longevity Survey (2009, 2012, and 2014), with 12,312 observations from 4,104 participants, gender-stratified panel ordered logistic regressions were performed. Food consumption included intake frequency and types of fruits, vegetables, staple food, cooking oil, and meat. Results: Among male older adults, more frequent smoking behavior was associated with lower BMI with the inclusion of food consumption behaviors. However, more frequent smoking behavior was not associated with BMI among female participants with the inclusion of food consumption behaviors. Former smoking status was not associated with BMI in all models. Discussion: The findings suggest the need of food consumption behaviors when researchers study the associations between smoking frequency and obesity. Gender gaps also should be considered.

Keywords

China obesity older adults smoking food consumption behaviors

Introduction

China is facing imminent public health concerns due to rapid social changes including the trends of escalating obesity and its growing elderly population. In fact, Chinese dietary patterns have shifted dramatically since the late 20th century (Du, Wang, Zhang, Zhai, & Popkin, 2014). With Westernized food options increasingly available in China, healthier cooking methods such as steaming and boiling have been replaced by frying and dining away from households (Zhai et al., 2014; Zhang, Zhai, Du, & Popkin, 2014), leading to higher consumption of fast food and rapid growth of the fast food industry in China (Wang, Wang, Xue, & Qu, 2016).

In addition, healthier food options have shown steady decreases in China. For instance, in urban regions, daily fruits and vegetables consumption declined by approximately 100 grams between 1982 and 1992 (Du et al., 2014). In the same vein, the prevalence of adult overweight has increased steadily in China. During an 18-year span from 1991 to 2009, the prevalence of adult overweight increased from 11.7% to almost 30% (Gordon-Larsen, Wang, & Popkin, 2014). Taken together, the shifts of dietary patterns, such as the increasing consumption of high protein food and beverages with higher level of calories intake were directly associated with higher risk of obesity among Chinese adults (Zou et al., 2017). Based on these facts, China faces an urgent need to provide more effective health promotional strategies to adjust people’s lifestyle.

The growing elderly population is another concern in China. An estimated 30% of the population will be at least 60 years of age in China by 2050 (Banister, Bloom, & Rosenberg, 2010). The growth of the aging population goes hand-in-hand with longer life expectancy, which increased from 40 to 73 years between the 1950s and the 2000s (Banister et al., 2010). However, the aging population also brings higher incidence of chronic diseases to Chinese older adults (Banister et al., 2010; Wu et al., 2013; Zheng et al., 2016). One study using a nationally representative database of the Study on Global AGEing and Adult Health (SAGE-China Wave 1) found that nearly 19% of older adults reported two chronic conditions and approximately 6% reported three chronic diseases (Wu et al., 2013). Chronic diseases account for 80% of deaths in China (Xiao, Long, Tang, & Tang, 2014), and the burden of chronic diseases also challenges the Chinese health care system and results in an increasingly larger burden. For example, it has been estimated that health expenditures for diabetes will skyrocket from 11 million Renminbi (RMB) to more than 36 million RMB between 2005 and 2030 (Zheng et al., 2016), respectively. Thus, it has become more important to emphasize healthy lifestyle and healthy aging among older adults in China (Banister et al., 2010).

In addition to these challenges, smoking is another public health concern in China. In fact, China is a country with a long smoking history, alongside with alcohol consumption behaviors (Lee, Chang, Liu, & Shelley, 2018). Approximately 26% of the Chinese population are daily smokers, who constitute the largest smoking population in the world (Bloom et al., 2013). Furthermore, the prevalence of cigarette users increased rapidly between the 1950s and 1990s and there have not been significant changes since then (Bloom et al., 2013; Liu et al., 1998).

Pepino and Mennella (2014) argued that smoking and poor dietary behaviors are strongly related, in that smokers tend to crave both cigarettes and high-fat foods, while Mackay, Gray, and Pell (2013) observed that current smoking status was associated with lower likelihood of being overweight than never smokers among adults (Mackay et al., 2013; Pepino & Mennella, 2014). The craving for carbs and high-fat foods (Mackay et al., 2013), on the contrary, might be directly associated with higher incidence of obesity. Furthermore, changes in the dietary landscape and the prevalence of obesity in China have been examined previously (Mu, Xu, Hu, Wu, & Bai, 2017; Popkin, 2001; Shang et al., 2012). A study conducted in Zhejiang province used older adults’ dietary patterns and current smoking status to investigate their relationships with obesity (Zou et al., 2017). However, the authors used only smoking status as a covariate to capture individuals’ smoking behavior (Zou et al., 2017). Other smoking-related variables, such as smoking frequency, are needed to capture further smoking patterns and study together with food consumption behaviors.

With the rapid dietary shifts in China (Zhai et al., 2014; Zhang et al., 2014), currently, research using nationwide data to assess the relationships of smoking frequency and food consumption with obesity among older adults in China within a single research effort remains limited. Food consumption behavior could be an important confounding variable, given that diet is a key to health and disease prevention in everyday life (Sleddens et al., 2015; Tomiyama, Mann, & Comer, 2009). In addition, as poor dietary and health behavioral choices including smoking are major risk factors for chronic diseases—a major burden in China especially among older adults—it is necessary to address this knowledge gap. We investigated Chinese older adults’ smoking frequency and the obesity-related measurement of body mass index (BMI), together with basic food consumption behaviors. This research included five major food consumption behaviors: fresh fruits, fresh vegetables, types of staple food, types of cooking oil, and meat. Further policy implications based on the empirical findings from this research are discussed.

Method

Theoretical Framework

The framework of micro- and macro-environmental determinants, rearticulated by Swinburn and Egger (2002), was applied to guide variable selection for the present study. The main purpose of this framework originally was to investigate obesity-related topics. However, it has been widely used as an alternative mechanism to explain the ecological approach to various health-related research topics (DiClemente, Salazar, & Crosby, 2013). Micro-environments include individual and social determinants (e.g., age, education background, and personal behaviors including smoking frequency and food consumption). Macro-environments are more distal determinants (e.g., community of residence).

Study Sample

Longitudinal data were extracted from three waves of the Chinese Longitudinal Healthy Longevity Survey (CLHLS): 2009 (collected between 2008 and 2009), 2012 (collected between 2011 and 2012), and 2014. The first wave was conducted in 1998 and additional waves were collected every 2 or 3 years. The methods of data collection included face-to-face interviews with a self-reported questionnaire and a cognitive assessment test. This database is the result of an ongoing international collaboration between the Center for the Study of Aging and Human Development at Duke University and other global partnerships (the Hong Kong Research Grants Council, the Max Planck Institute for Demographic Research, the China Social Sciences Foundation, and others).

The CLHLS investigators collected data from Chinese older adults to examine activities of daily living arrangements, chronic diseases (e.g., diabetes and arthritis), smoking/alcohol consumption behaviors, dietary behaviors, family relationships, overall health, mental health, and physical/sedentary lifestyles. Informed consents were obtained by the CLHLS investigators. The surveyed geographical areas cover almost 1.1 billion residents, 85% of China’s total population. The survey included mainly older individuals, but some middle-aged participants were added. As of the latest 2014 wave, major provinces and several mega cities were included: (a) Northern areas: Tianjin, Beijing, Hebei, and Shanxi; (b) Northeastern regions: Jilin, Heilongjiang, and Liaoning; (c) Eastern areas: Jiangsu, Shanghai, Zhejiang, Anhui, Jiangxi, Shandong, and Fujian; (d) Central and Southern regions: Hunan, Hubei, Guangdong, Henan, Hainan, and Guangxi; and (e) Western areas: Sichuan, Chongqing, and Shaanxi. Other information regarding CLHLS can be found elsewhere (Zeng, 2012). As only de-identified individual-level data were used, this research was exempted from the Institutional Review Board review.

Only survey participants who were at least 60 years old were used in this research. We included only individuals who fully answered all questions of interest and removed all missing values; furthermore, we only included older adults who participated in three occasions of data collection between 2009 and 2014. We considered the 2009 wave as the baseline wave. With our inclusion and exclusion criteria, 4,104 participants were retained in the final study sample with a total of 12,312 observations for statistical analyses. Each older adult provided three observations.

Measurements

Outcome variable

Participants’ height (cm) and weight (kg) were selected to calculate BMI = weight (kg) / height (m²). For purposes of data analysis, three categories for BMI were employed: “underweight” (BMI less than 18.5), “normal” (BMI 18.5-25), and “overweight/obese” (BMI higher than 25.0). This ordinal variable was coded 1 = underweight, 2 = normal, and 3 = overweight. All obesity-related values were measured by interviewers.

Predictors

This study included two major components for predictors: smoking frequency and food consumption behaviors. Smoking frequency was categorized as non-smoker, current smoker who smoked less than 15 times a day, current smoker who smoked more than 15 times a day, and former smoker. We chose five dietary behaviors, important food items in the Chinese basic diet, to describe participants’ basic food consumption frequencies: fruit and vegetable intakes (daily, quite often but less than daily, occasionally, and rarely or never), types of staple food (rice, corn, wheat, half rice–half wheat, and others), types of cooking oil (vegetable/sesame, or animal/lard), and meat consumption (daily, weekly, monthly, less than monthly, and rarely or none). Wheat included products of noodles and breads. These measurements were all self-reported.

Covariates

We selected the following biological and sociodemographic variables as covariates for statistical analyses: age (“60-70,” “71-80,” “81-90,” and “above 90”), years of formal education (“none,” “1-5,” “6-10,” and “above 11”), community of residence (urban or rural), provinces (North, Northeast, East, Central/South, and West), physical activity (no or yes), and current alcohol use (no or yes). Urban residents included those who resided in cities and towns, while rural included participants who lived in outlying areas. Year of survey (2009, 2012, and 2014) was included in all analyses to estimate the fixed effect of panel.

Statistical Analyses

We estimated six ordered logistic regression models of participants’ BMI, with BMI measured as an ordinal outcome (Agresti, 2002). The ordered logistic regression models in this present research were proportional odds logistic regressions. The regression equation is

\log \frac{P (Y \leq j | X)}{1 - P (Y \leq j | X)} = α_{j} - β^{'} X, j = 1, 2,

where

Y = {\begin{cases} 1, if BMI < 18.5 \\ 2, if 18.5 \leq BMI < 25 \\ 3, if BMI \geq 25 \end{cases},

and X is the vector of independent variables.

First, we performed two ordered logistic regression models of participants’ BMI for the total sample, with and without food consumption behaviors. Next, gender-stratified models were conducted due to the gender gaps in obesity (Jones-Johnson, Johnson, & Frishman, 2014; Kanter & Caballero, 2012). The first two gender-stratified models, without food consumption behaviors, investigated the associations between older adults’ smoking frequency and BMI among male and female study participants. The later gender-stratified models examined the associations of participants’ smoking frequency and food consumption behaviors with BMI. To ensure the robustness of hypothesis testing, we used the method of false discovery rate (FDR; Benjamini & Hochberg, 1995). This method made it possible to control the small p values because some statistically significant associations could happen by chance, in which case the result might reject the null hypothesis inaccurately. The FDR method can help avoid Type 1 errors. In addition, we used the adjusted generalized variance inflation factor (AGVIF) to detect issues related to multicollinearity that variables included in the statistical models should have AGVIFs smaller than 2. Preliminary statistical analysis showed that models adjusted using food consumption behaviors were free from multicollinearity (all AGVIFs < 2). Furthermore, previous studies have indicated the potential issues related to endogeneity of smoking and obesity (Baum, 2009; Cawley, Markowitz, & Tauras, 2006; Fang, Ali, & Rizzo, 2009). We performed the Hausman–Taylor estimator for error components and examined potential endogeneity between smoking and obesity among older adults included in this present research. Preliminary results did not find evidence of endogeneity in our statistical models.

Adjusted odds ratios (AORs; reference level = 1.00) and 95% confidence intervals (95% CIs) were reported as the results of ordered logistic regression. All analyses and ordered logistic regression were conducted on R (version 3.4.3) by using its package “pglm” (Croissant, 2017).

Results

Descriptive Statistics

Table 1 provides descriptive statistics of the study sample by respondents’ gender. A majority of participants were female. Most older adults were aged 71 years and above. Most participants had normal BMI, regardless of their gender. Approximately 17% of males and 24% of females were underweight. Nearly 15% of male older adults and 18% of female older adults were overweight or obese. Among males, nearly 40% did not have any smoking experiences and almost 12% of male older adults smoked more than 15 times a day. For female participants, nearly 90% were non-smokers. Female participants received less education than male participants. Most older adults resided in rural areas and lived in Central-Southern and Eastern regions. The majority were not physically active. More than 33% of male older adults reported current alcohol use, but only 8% of female older adults responded that they used alcohol.

Table 1.

Descriptive Statistics of the Variables of CLHLS Used in the Analysis, 2009-2014 (Total N = 12,312).

Variables	Measurements	Overall (N = 12,312)		Male (n = 5,850)		Female (n = 6,462)
Variables	Measurements	n	%	n	%	n	%
Outcome variable
BMI	Underweight	2,554	20.74	995	17.01	1,559	24.13
	Normal	7,738	62.85	3,973	67.91	3,765	58.26
	Overweight or obese	2,020	16.41	882	15.08	1,138	17.61
Primary predictor
Smoking frequency	Non-smoker	7,946	64.54	2,162	36.96	5,784	89.51
	Former smoker	1,913	15.54	1,600	27.35	313	4.84
	Smoked less than 15 times a day	1,703	13.83	1,398	23.9	305	4.72
	Smoked more than 15 times a day	750	6.09	690	11.79	60	0.93
Food consumption behaviors
Fresh fruits consumption	Daily	1,687	13.70	799	13.66	888	13.74
	Quite often	3,148	25.57	1,526	26.09	1,622	25.1
	Occasionally	4,416	35.87	2,052	35.08	2,364	36.58
	Rarely or none	3,061	24.86	1,473	25.18	1,588	24.57
Fresh vegetables consumption	Daily	7,801	63.36	3,764	64.34	4,037	62.47
Fresh vegetables consumption	Quite often	3,373	27.40	1,587	27.13	1,786	27.64
	Occasionally	856	6.95	380	6.5	476	7.37
	Rarely or none	282	2.29	119	2.03	163	2.52
Types of staple food	Rice	7,632	61.99	3,587	61.32	4,045	62.6
	Corn	424	3.44	186	3.18	238	3.68
	Wheat half and rice half	2,542	20.65	1,243	21.25	1,299	20.1
	Wheat	1,682	13.66	822	14.05	860	13.31
	Others	32	0.26	12	0.21	20	0.31
Types of cooking oil	Vegetable	10,545	85.65	5,037	86.1	5,508	85.24
Types of cooking oil	Animal	1,767	14.35	813	13.9	954	14.76
Meat consumption	Daily	4,045	32.85	2,031	34.72	2,014	31.17
	Weekly	5,034	40.89	2,452	41.91	2,582	39.96
	Monthly	1,311	10.65	598	10.22	713	11.03
	Less than monthly	907	7.37	392	6.7	515	7.97
	Rarely or none	1,015	8.24	377	6.44	638	9.87
Covariates
Age	60-70	1,744	14.17	915	15.64	829	12.83
	71-80	4,630	37.61	2,349	40.15	2,281	35.3
	81-90	3,621	29.41	1,730	29.57	1,891	29.26
	Above 90	2,317	18.82	856	14.63	1,461	22.61
Years of formal education	None	6,346	51.54	1,641	28.05	4,705	72.81
	1-5	3,294	26.75	2,222	37.98	1,072	16.59
	5-10	2,165	17.58	1,585	27.09	580	8.98
	11 and above	507	4.12	402	6.87	105	1.62
Community of residence	Urban	5,788	47.01	2,774	47.42	3,014	46.64
Community of residence	Rural	6,524	52.99	3,076	52.58	3,448	53.36
Provinces	North	507	4.12	234	4	273	4.22
	Northeast	744	6.04	408	6.97	336	5.2
	East	4,716	38.30	2,298	39.28	2,418	37.42
	Central-South	4,575	37.16	2,064	35.28	2,511	38.86
	West	1,770	14.38	846	14.46	924	14.3
Physical activity	No	7,672	62.31	3,343	57.15	4,329	66.99
Physical activity	Yes	4,640	37.68	2,507	42.85	2,133	33.01
Current alcohol consumption	No	9,896	80.38	3,939	67.33	5,957	92.19
Current alcohol consumption	Yes	2,414	19.62	1,911	32.67	505	7.81
Year of survey	2009	4,104	33.33	1,950	33.33	2,154	33.33
	2012	4,104	33.33	1,950	33.33	2,154	33.33
	2014	4,104	33.33	1,950	33.33	2,154	33.33

Note. CLHLS = Chinese Longitudinal Healthy Longevity Survey; BMI = body mass index.

Smoking Frequency, Food Consumption Behaviors, and Obesity

Table 2 provides the results of ordered logistic regression examining the associations between smoking frequency and BMI among all older adults, with the inclusion and exclusion of food consumption behaviors. In the first model without food consumption behaviors, we observed that smoking either less or more than 15 times a day was associated with lower odds of reporting obesity, compared with non-smoking. However, more frequent smoking behavior (smoked more than 15 times a day) was not associated with BMI after we included food consumption behaviors in the model.

Table 2.

Results of Ordered Logistic Regression Examining the Associations Between Smoking Frequency and BMI Adjusting for Food Consumption Behaviors in the Total Sample: CLHLS, 2009-2014 (N = 12,312).

	Without food consumption behaviors	With food consumption behaviors
	AOR (95% CI)	AOR (95% CI)
Variables
Age
60-70	—^a	—
71-80	0.63 [0.54, 0.74]**b	0.65 [0.55, 0.76]
81-90	0.34 [0.28, 0.42]**	0.36 [0.30, 0.44]
Above 90	0.18 [0.14, 0.22]**	0.19 [0.15, 0.23]
Years of formal education
None	—	—
1-5	1.28 [1.09, 1.49]**	1.29 [1.11, 1.51]**
6-10	1.81 [1.50, 2.19]**	1.72 [1.43, 2.08]**
Above 11	1.37 [0.98, 1.92]	1.33 [0.95, 1.86]
Community of residence
Urban	—	—
Rural	0.77 [0.69, 0.85]**	0.77 [0.69, 0.85]**
Provinces
North	—	—
Northeast	0.51 [0.33, 0.78]**	0.63 [0.41, 0.96]*
East	0.39 [0.27, 0.55]**	0.46 [0.32, 0.65]**
Central-South	0.19 [0.14, 0.28]**	0.29 [0.20, 0.42]**
West	0.30 [0.20, 0.43]**	0.45 [0.31, 0.66]**
Physical activity
No	—	—
Yes	1.29 [1.16, 1.43]**	1.25 [1.13, 1.39]**
Current alcohol consumption
No	—	—
Yes	1.06 [0.92, 1.23]	1.08 [0.94, 1.24]
Years of survey
2009	—	—
2012	1.63 [1.47, 1.81]**	1.66 [1.49, 1.84]**
2014	2.26 [2.02, 2.52]**	2.32 [2.07, 2.60]**
Major predictor
Smoking frequency
Non-smoker	—	—
Smoked less than 15 times a day	0.67 [0.57, 0.80]**	0.67 [0.56, 0.79]**
Smoked more than 15 times a day	0.68 [0.54, 0.87]**	0.69 [0.54, 0.87]
Former smoker	0.94 [0.81, 1.11]	0.92 [0.78, 1.08]
Food consumption behaviors
Fresh fruits consumption
Daily	—	—
Quite often but not daily		0.95 [0.81, 1.12]
Occasionally		0.83 [0.71, 0.98]*
Rarely or none		0.80 [0.68, 0.96]*
Fresh vegetables consumption
Daily	—	—
Quite often but not daily		0.85 [0.76, 0.95]**
Occasionally		0.69 [0.57, 0.84]
Rarely or none		0.56 [0.43, 0.77]**
Types of staple food
Rice	—	—
Corn		1.26 [0.96.1.66]
Wheat		1.91 [1.63, 2.25]**
Half rice–half wheat		1.30 [1.11, 1.52]**
Others		1.64 [0.67, 4.00]
Types of cooking oil
Vegetable/sesame	—
Animal oil		0.73 [0.62, 0.87]**
Meat consumption
Daily	—	—
Weekly		1.05 [0.94, 1.18]
Monthly		1.00 [0.84, 1.19]
Less than monthly		1.15 [0.94, 1.40]
Rarely or none		1.04 [0.85, 1.26]

Note. BMI = body mass index; CLHLS = Chinese Longitudinal Healthy Longevity Survey; AOR = adjusted odds ratio; CI = confidence interval.

Reference level of each category (—): AOR = 1.00.

All statistical significances were adjusted by Benjamini–Hochberg procedure of false discovery rate (FDR).

p < .05. **p < .01.

Table 3 provides the results of ordered logistic regression examining the associations between smoking frequency and BMI for male and female older adults, without food consumption behaviors. In both male and female sex-stratified models, daily smokers, regardless of number of times smoked a day, had lower odds of being overweight or obese, compared with older adults who did not smoke at all (p < .05). Former smoking status was not associated with BMI.

Table 3.

Results of Ordered Logistic Regression Examining the Associations Between Smoking Frequency and BMI Without Food Consumption Behaviors in Gender-Stratified Models: CLHLS, 2009-2014 (N = 12,312).

	Male	Female
	AOR (95% CI)	AOR (95% CI)
Variables
Age
60-70	—^a
71-80	0.60 [0.47, 0.75]**^b	0.68 [0.55, 0.85]**
81-90	0.35 [0.27, 0.47]**	0.35 [0.27, 0.46]**
Above 90	0.22 [0.16, 0.31]**	0.16 [0.12, 0.22]**
Years of formal education
None	—
1-5	1.25 [0.98, 1.58]	1.21 [0.95, 1.53]
6-10	1.84 [1.41, 2.41]**	1.77 [1.29, 2.44]**
Above 11	1.02 [0.67, 1.56]	3.78 [1.88, 7.57]**
Community of residence
Urban	—
Rural	0.82 [0.69, 0.96]*	0.74 [0.64, 0.85]**
Provinces
North	—
Northeast	0.52 [0.28, 0.97]	0.51 [0.28, 0.91]*
East	0.34 [0.20, 0.59]**	0.42 [0.26, 0.66]**
Central-South	0.18 [0.11, 0.31]**	0.21 [0.13, 0.33]**
West	0.23 [0.13, 0.42]**	0.35 [0.21, 0.58]**
Physical activity
No	—
Yes	1.16 [0.99, 1.35]	1.38 [1.20, 1.60]**
Current alcohol consumption
No	—
Yes	1.07 [0.90, 1.27]	1.03 [0.80, 1.34]
Years of survey
2009	—
2012	1.46 [1.25, 1.71]**	1.77 [1.54, 2.04]**
2014	1.95 [1.65, 2.31]**	2.49 [2.14, 2.89]**
Major predictor
Smoking frequency
Non-smoker	—
Smoked less than 15 times a day	0.69 [0.55, 0.86]**	0.62 [0.43, 0.88]*
Smoked more than 15 times a day	0.72 [0.55, 0.96]*	0.48 [0.24, 0.97]*
Former smoker	1.01 [0.82, 1.24]	0.87 [0.62, 1.22]

Note. BMI = body mass index; CLHLS = Chinese Longitudinal Healthy Longevity Survey; AOR = adjusted odds ratio; CI = confidence interval.

Reference level of each category (—): AOR = 1.00.

All p values were adjusted by Benjamini–Hochberg procedure of false discovery rate (FDR).

p < .05. **p < .01.

Table 4 provides the results of ordered logistic regression examining the associations between smoking frequency and BMI for male and female older adults, controlling for food consumption behaviors and other covariates. The results of the model for males were similar to the model without food consumption behaviors. However, more frequent smoking behavior (more than 15 times a day) was not associated with BMI in the female elderly population, controlling for food consumption behaviors and covariates. The relationship between former smoking status and BMI remained statistically insignificant.

Table 4.

Results of Ordered Logistic Regression Examining the Associations Between Smoking Frequency and BMI Adjusting for Food Consumption Behaviors in Gender-Stratified Models: CLHLS, 2009-2014 (N = 12,312).

	Male	Female
	AOR (95% CI)	AOR (95% CI)
Variables
Age
60-70	—^a
71-80	0.63 [0.50, 0.79]**b	0.69 [0.56, 0.87]**
81-90	0.38 [0.29, 0.50]**	0.36 [0.28, 0.47]**
Above 90	0.24 [0.17, 0.34]**	0.17 [0.13, 0.24]**
Years of formal education
None	—
1-5	1.27 [1.00, 1.60]	1.21 [0.96, 1.53]
6-10	1.74 [1.34, 2.27]**	1.65 [1.20, 2.27]**
Above 11	0.99 [0.65, 1.50]	3.72 [1.86, 7.42]**
Community of residence
Urban	—
Rural	0.81 [0.69, 0.95]*	0.74 [0.64, 0.86]**
Provinces
North	—
Northeast	0.73 [0.39, 1.35]	0.60 [0.33, 1.07]
East	0.44 [0.26, 0.75]**	0.47 [0.29, 0.75]**
Central-South	0.30 [0.17, 0.51]**	0.30 [0.18, 0.48]**
West	0.39 [0.22, 0.69]**	0.51 [0.31, 0.85]*
Physical activity
No	—
Yes	1.14 [0.97, 1.33]	1.34 [1.16, 1.55]**
Current alcohol consumption
No	—
Yes	1.06 [0.89, 1.27]	1.05 [0.81, 1.36]
Years of survey
2009	—
2012	1.45 [1.24, 1.70]**	1.82 [1.59, 2.10]**
2014	1.92 [1.62, 2.28]**	2.63 [2.26, 3.07]**
Major predictor
Smoking frequency
Non-smoker	—
Smoked less than 15 times a day	0.68 [0.54, 0.85]**	0.60 [0.42, 0.83]*
Smoked more than 15 times a day	0.72 [0.55, 0.95]*	0.49 [0.24, 0.99]
Former smoker	0.97 [0.79, 1.19]	0.84 [0.60, 1.18]
Food consumption behaviors
Fresh fruits consumption
Daily	—
Quite often but not daily	1.04 [0.81, 1.34]	0.90 [0.72, 1.12]
Occasionally	0.81 [0.64, 1.04]	0.87 [0.70, 1.08]
Rarely or none	0.88 [0.68, 1.14]	0.76 [0.60, 0.97]*
Fresh vegetables consumption
Daily	—
Quite often but not daily	0.85 [0.72, 1.01]	0.84 [0.73, 0.98]*
Occasionally	0.77 [0.57, 1.03]	0.63 [0.49, 0.82]**
Rarely or none	0.74 [0.45, 1.23]	0.46 [0.30, 0.70]**
Types of staple food
Rice	—
Corn	1.19 [0.78, 1.83]	1.29 [0.90, 1.84]
Wheat	2.43 [1.91, 3.10]**	1.64 [1.32, 2.04]**
Half rice–half wheat	1.51 [1.19, 1.92]**	1.17 [0.95, 1.45]
Others	1.67 [0.38, .7.36]	1.63 [0.54, 4.95]
Types of cooking oil
Vegetable/sesame	—
Animal oil	0.78 [0.61, 1.00]	0.70 [0.56, 0.87]**
Meat consumption
Daily	—
Weekly	0.98 [0.82, 1.16]	1.10 [0.94, 1.29]
Monthly	0.79 [0.60, 1.03]	1.19 [0.94, 1.50]
Less than monthly	1.02 [0.75, 1.39]	1.25 [0.96, 1.63]
Rarely or none	0.73 [0.53, 1.00]	1.27 [0.99, 1.63]

Note. BMI = body mass index; CLHLS = Chinese Longitudinal Healthy Longevity Survey; AOR = adjusted odds ratio; CI = confidence interval.

Reference level of each category (—): AOR = 1.00.

All statistical significances were adjusted by Benjamini-Hochberg procedure of false discovery rate (FDR).

p < .05. **p < .01.

Discussion

To the best of the authors’ knowledge, this is the first research available to investigate the associations of smoking frequency and older adults’ BMI, controlling for food consumption behaviors and other sociodemographic variables. The inclusion of food consumption behavior in such a study might serve as a new concept, given that diet is necessary in daily human health, in which case dietary behaviors could be important confounders of the associations between smoking and obesity. A nationally representative sample with panel information was used to examine this topic of interest. Furthermore, to test our research hypothesis robustly, we used the Benjamini and Hochberg (1995) FDR method.

In this research, smoking more than 15 times a day was not associated with BMI after controlling for food consumption behaviors in the total sample. However, for both models with or without basic food consumption behaviors, current smoking frequencies were associated with lower levels of BMI among male older adults, compared with those who did not smoke at all. Interestingly, the results were different for female participants before and after including food consumption behaviors in the models. More frequent daily smoking was not associated with BMI for females. The findings with female older adults were more consistent with results from the total sample. With the empirical evidence from this study, further research should continue to examine the gap between males and females on this topic.

Our findings regarding older adults’ current smoking status and BMI were similar to previous findings that current smokers had lower odds of being overweight or obese, compared with non-current smokers (Dare, Mackay, & Pell, 2015; Ginawi et al., 2016). In fact, nicotine is an appetite suppressant (Courtemanche, Tchernis, & Ukert, 2018) and could reduce people’s food consumption (Mineur et al., 2011) and appetite. Nevertheless, previous studies also pointed out that quitting tobacco use conveys health benefits, but tobacco cessation is associated with post-cessation weight gain or a new condition of obesity (Bush, Lovejoy, Deprey, & Carpenter, 2016; Courtemanche et al., 2018). However, this was not the case for the present research, given that we did not observe statistically significant associations either with or without food consumption behaviors among both male and female participants for former smoking status.

There are two potential explanations of this knowledge gap. First, older adults are known for unintentional weight loss (Stajkovic, Aitken, & Holroyd-Leduc, 2011). Even when older adults decide to quit smoking, our findings could be affected by older adults also losing weight unintentionally at the same time due to their age. Accordingly, this potential research gap in participants’ age should be taken into consideration when others may attempt to establish conclusively a relationship between smoking behavior and obesity. Second, older adults’ appetites tend to decline as they age (Pilgrim, Robinson, Sayer, & Roberts, 2015). Older adults trying to quit smoking may not have sufficient appetite to consume more food, although it is known that nicotine is a type of appetite suppressant (Courtemanche et al., 2018). Hence, the inclusion of food consumption behaviors could be critical covariates when researchers examine the associations between smoking behavior and obesity among older adults.

Nicotine reduces food intake and consumption (Mineur et al., 2011) but has been recommended as an effective substance for weight control (Audrain-McGovern & Benowitz, 2011). However, most common nicotine exposures were delivered by cigarette smoking (Audrain-McGovern & Benowitz, 2011). As we observed that more frequent smoking (more than 15 times a day) was associated with lower BMI only among male older adults after controlling for food consumption, it is important to be careful in supporting claims that smoking behaviors are associated with lower prevalence of obesity or that higher exposure to nicotine is associated with leaner populations adjusted for food consumption and other sociodemographic variables, especially for older adults.

In addition, the percentage of smokers among female older adults was quite low in our analyses, due to the overall low smoking rate among women in China, just 3.4% of whom have ever smoked and 1.5% of whom are current smokers (Lee, Wang, Chiang, & Liu, 2017; Liu et al., 2017). As previously mentioned, the observed differences could be participants’ age and gender. We suggest that further research should continue to investigate the associations between smoking behaviors and obesity among older adults in China, controlling for dietary patterns or food consumption behaviors. The results of older adults should be compared with findings for younger individuals as well. This comparison could help support public health practitioners to provide more effective strategies to reduce obesity and smoking behavior, alongside with the potential differences based on the gender gap observed in our results.

The present research has several strengths, including a nationally representative sample and a panel dataset. Besides, the inclusion of a substantial number of the oldest old (participants who were 80 years old or above) provides more generalizable observations in the demographic transition characterized by the escalating size of the aging population in China (Zheng et al., 2016). Furthermore, compared with cross-sectional design, the advantage of using a panel approach enables the present work to incorporate variability within the dataset across multiple dimensions (Hsiao, 2007), including the changes in food consumption and smoking behaviors over time.

We should state some study limitations as caveats. First, the fact that the CLHLS questionnaire includes only frequency measurements for all food items reported by participants raises the risk of self-reported bias. However, response bias is common in most survey-based research due to social desirability response patterns on normative behaviors (Brenner & DeLamater, 2016). Thus, the risk of self-reported bias does not appear to be a major challenge to this research. Second, we did not have serving information on food consumption behaviors to identify the amount of consumption, given that the questionnaire includes only self-reported consumption frequencies. Third, we did not have information on older adults’ use of specific tobacco products. Different tobacco products could lead to different conclusions based on the use of electronic tobacco products and higher incidence of obesity (Lanza, Pittman, & Batshoun, 2017). Nevertheless, cigarettes remain the most common type of tobacco product used in China (O’Connor et al., 2017); therefore, the lack of detailed tobacco products should not be a critical concern in interpreting our results. Finally, the results of Hausman–Taylor estimator did not reveal issues related to endogeneity of smoking and obesity in our analyses; however, with the evidence from previous research regarding endogeneity between smoking and obesity (Baum, 2009; Cawley et al., 2006; Fang et al., 2009), we cannot entirely eliminate such concern. As we only attempted to observe the confounding effects of food consumption behaviors in our research, further studies should use other statistical methods to control endogeneity such as instrumental variables (Bascle, 2008).

Conclusion

We investigated the associations of smoking frequency and food consumption behaviors with obesity, focused on older adults in China. After controlling for food consumption, male older adults’ smoking frequency was associated with a lower level of BMI, compared with male non-smokers. However, more frequent smoking behavior, more than 15 times a day, was not associated with BMI among female older adults or in the total sample. Former smoking status was not associated with BMI, with or without food consumption behaviors. Public health researchers and practitioners should continue to investigate the long-term impact of older adults’ food consumption and smoking patterns on obesity. Potential gaps based on participants’ gender should be assessed. Despite the effect of smoking behavior on lower risk of obesity observed in this study, we do not encourage older adults to smoke due to other harmful chemicals from cigarette products such as tar or phenol. Smoking behavior together with obesity are associated with premature mortality from chronic diseases (Li et al., 2017). Although nicotine has been suggested for weight management (Audrain-McGovern & Benowitz, 2011), further research targeting older adults should continue to investigate whether the associations between smoking and obesity are due to poor appetite among older adults or the true effect of nicotine on weight control. Public health practitioners should be fully aware of these associations. Finally, it is important for researchers to adjust for dietary patterns or food consumption when studying the relationships of smoking behaviors with obesity, given that patterns of dietary consumption are critical correlates with daily human behaviors.

Footnotes

Acknowledgements

The authors sincerely thank the research participants and investigators of the latest Chinese Longitudinal Healthy Longevity Survey. Data used for this research were provided by the Chinese Longitudinal Healthy Longevity Survey (CLHLS) managed by the Center for Healthy Aging and Development Studies, Peking University. CLHLS is supported by funds from the U.S. National Institutes on Aging (NIA), the China Natural Science Foundation, the China Social Science Foundation, and the United Nations Population Fund.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Yen-Han Lee

Ching-Ti Liu

References

Agresti

(2002). Categorical data (2nd ed.). Hoboken, NJ: John Wiley.

Audrain-McGovern

Benowitz

N. L.

(2011). Cigarette smoking, nicotine, and body weight. Clinical Pharmacology and Therapeutics, 90, 164-168.

Banister

Bloom

D. E.

Rosenberg

(2010). Population aging and economic growth in China (Program on the Global Demography of Aging). Retrieved from https://core.ac.uk/download/pdf/6518505.pdf

Bascle

(2008). Controlling for endogeneity with instrumental variables in strategic management research. Strategic Organization, 6, 285-327.

Baum

C. L.

(2009). The effects of cigarette cost on BMI and obesity. Health Economics, 18, 3-19.

Benjamini

Hochberg

(1995). Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society. Series B (Methodological), 57, 289-300.

Bloom

D. E.

Cafiero

E. T.

McGovern

M. E.

Prettner

Stanciole

Weiss

. . . Rosenberg

(2013). The economic impact of non-communicable disease in China and India: Estimates, projects, and comparisons (NBER Working Paper Series). Retrieved from http://www.nber.org/papers/w19335.pdf

Brenner

P. S.

DeLamater

(2016). Lies, damned lies, and survey self-reports? Identity as a cause of measurement bias. Social Psychology Quarterly, 79, 333-354.

Bush

Lovejoy

J. C.

Deprey

Carpenter

K. M.

(2016). The effect of tobacco cessation on weight gain, obesity, and diabetes risk. Obesity, 24, 1834-1841.

10.

Cawley

Markowitz

Tauras

(2006). Obesity, cigarette prices, youth access law and adolescent smoking initiation. Eastern Economic Journal, 32, 149-170.

11.

Courtemanche

Tchernis

Ukert

(2018). The effect of smoking on obesity: Evidence from a randomized trial. Journal of Health Economics, 57, 31-44.

12.

Croissant

(2017). Package “pglm.” Retrieved from https://cran.r-project.org/web/packages/pglm/pglm.pdf

13.

Dare

Mackay

D. F.

Pell

J. P.

(2015). Relationship between smoking and obesity: A cross-sectional study of 499,504 middle-aged adults in the UK general population. PLoS ONE, 10(4), e012357920.

14.

DiClemente

R. J.

Salazar

L. F.

Crosby

R. A.

(2013). Health behavior theory for public health: Principles, foundations, and applications. Burlington, MA: Jones & Bartlett Learning.

15.

Wang

Zhang

Zhai

Popkin

B. M.

(2014). China in the period of transition from scarcity and extensive undernutrition to emerging nutrition-related non-communicable diseases, 1949–1992. Obesity Reviews: An Official Journal of the International Association for the Study of Obesity, 15(Suppl. 1), 8-15.

16.

Fang

Ali

M. M.

Rizzo

J. A.

(2009). Does smoking affect body weight and obesity in China? Economics & Human Biology, 7, 334-350.

17.

Ginawi

I. A.

Bashir

A. I.

Alreshidi

Y. Q.

Dirweesh

Al-Hazimi

A. M.

Ahmed

H. G.

. . . Ahmed

M. H.

(2016). Association between obesity and cigarette smoking: A community-based study. Journal of Endocrinology and Metabolism, 6(5), 149-153.

18.

Gordon-Larsen

Wang

Popkin

B. M.

(2014). Overweight dynamics in Chinese children and adults. Obesity Reviews: An Official Journal of the International Association for the Study of Obesity, 15(Suppl. 1), 37-48.

19.

Hsiao

(2007). Panel data analysis—Advantages and challenges. TEST, 16(1), 1-22.

20.

Jones-Johnson

Johnson

W. R.

Frishman

(2014). Race and gender differences in obesity and disease. Sociology Mind, 4, 233-241.

21.

Kanter

Caballero

(2012). Global gender disparities in obesity: A review. Advances in Nutrition, 3, 491-498.

22.

Lanza

H. I.

Pittman

Batshoun

(2017). Obesity and cigarette smoking: Extending the link to E-cigarette/vaping use. American Journal of Health Behavior, 41, 338-347.

23.

Lee

Y.-H.

Chang

Y.-C.

Liu

C.-T.

Shelley

(2018). Correlates of alcohol consumption and alcohol dependence among older adults in contemporary China: Results from the Chinese Longitudinal Healthy Longevity Survey. Journal of Ethnicity in Substance Abuse. Advanced online publication. doi:10.1080/15332640.2018.1456388

24.

Lee

Y.-H.

Wang

Chiang

T. C.

Liu

C.-T.

(2017). Beverage intake, smoking behavior, and alcohol consumption in contemporary China—A cross-sectional analysis from the 2011 China Health and Nutrition Survey. International Journal of Environmental Research and Public Health, 14(5), E493.

25.

Zeng

Liu

Yin

. . . Zhou

(2017). Can China achieve a one-third reduction in premature mortality from non-communicable diseases by 2030? BMC Medicine, 15, Article 132.

26.

Liu

Peto

Chen

Boreham

. . . Chen

(1998). Emerging tobacco hazards in China: 1. Retrospective proportional mortality study of one million deaths. British Medical Journal, 317, 1411-1422.

27.

Liu

Zhang

Yang

Wang

Huang

. . . Zhou

(2016). Prevalence and patterns of tobacco smoking among Chinese adult men and women: Findings of the 2010 national smoking survey. Journal of Epidemiology and Community Health, 71, 154-161.

28.

Mackay

D. F.

Gray

Pell

J. P.

(2013). Impact of smoking and smoking cessation on overweight and obesity: Scotland-wide, cross-sectional study on 40,036 participants. BMC Public Health, 13, Article 348.

29.

Mineur

Y. S.

Abizaid

Rao

Salas

DiLeone

R. J.

Gündisch

. . . Picciotto

M. R.

(2011). Nicotine decreases food intake through activation of POMC neurons. Science, 332, 1330-1332.

30.

L. F.

Bai

M. J.

(2017). Dietary patterns and overweight/obesity: A review article. Iranian Journal of Public Health, 46, 869-876.

31.

O’Connor

R. J.

Stephens

W. E.

Hammond

Elton-Marshall

Cummings

K. M.

. . . Fong

G. T.

(2010). Cigarettes sold in China: Design, emissions and metals. Tobacco Control, 19(Suppl. 2), i47-i53.

32.

Pepino

M. Y.

Mennella

J. A.

(2014). Cigarette smoking and obesity are associated with decreased fat perception in women. Obesity, 22, 1050-1055.

33.

Pilgrim

A. L.

Robinson

S. M.

Sayer

A. A.

Roberts

H. C.

(2015). An overview of appetite decline in older people. Nursing Older People, 27(5), 29-35.

34.

Popkin

B. M.

(2001). The nutrition transition and obesity in the developing world. The Journal of Nutrition, 131(3), 871S-873S.

35.

Shang

Liu

Zhang

. . . Ma

(2012). Dietary pattern and its association with the prevalence of obesity and related cardiometabolic risk factors among Chinese children. PLoS ONE, 7(8), e43183.

36.

Sleddens

E. F.

Kroeze

Kohl

L. F.

Bolten

L. M.

Velema

Kaspers

. . . Brug

(2015). Correlates of dietary behavior in adults: An umbrella review. Nutrition Reviews, 73, 477-499.

37.

Stajkovic

Aitken

E. M.

Holroyd-Leduc

(2011). Unintentional weight loss in older adults. Canadian Medical Association Journal / Journal de l’Association medicale canadienne, 183, 443-449.

38.

Swinburn

Egger

(2002). Preventive strategies against weight gain and obesity. Obesity Reviews: An Official Journal of the International Association for the Study of Obesity, 3, 289-231.

39.

Tomiyama

A. J.

Mann

Comer

(2009). Triggers of eating in everyday life. Appetite, 52, 72-82.

40.

Wang

Xue

(2016). A review of the growth of the fast food industry in China and its potential impact on obesity. International Journal of Environmental Research and Public Health, 13(11), Article 1112.

41.

Guo

Kowal

Jiang

. . . Xu

(2013). Prevalence of major chronic conditions among older Chinese adults: The Study on Global AGEing and adult health (SAGE) wave 1. PLoS ONE, 8(9), e74176. doi:10.1371/journal.pone.0074176

42.

Xiao

Long

Tang

(2014). A community-based approach to non-communicable chronic disease management within a context of advancing universal health coverage in China: Progress and challenges. BMC Public Health, 14(Suppl. 2), S2.

43.

Zeng

(2012). Towards deeper research and better policy for healthy aging—Using the unique data of Chinese longitudinal healthy longevity survey. China Economic Journal, 5, 131-149.

44.

Zhai

Wang

Zhang

Popkin

(2014). Dynamics of the Chinese diet and the role of urbanicity, 1991–2011. Obesity Reviews: An Official Journal of the International Association for the Study of Obesity, 15(Suppl. 1), 16-26.

45.

Zhang

Zhai

Popkin

B. M.

(2014). The China health and nutrition survey, 1989–2011. Obesity Reviews: An Official Journal of the International Association for the Study of Obesity, 15(Suppl. 1), 2-7.

46.

Zheng

Pang

Chen

Huang

Liu

Zhang

(2016). Challenge of population aging on health. In Islam

M. N.

(Ed.), Public Health Challenges in Contemporary China (pp. 43-54). Berlin, Germany: Springer.

47.

Zou

Zhang

Xia

Huang

Meng

Fang

Ding

(2017). Dietary patterns and obesity among Chinese adults: Results from a household-based cross-sectional study. International Journal of Environmental Research and Public Health, 14(5), E487.