Effectiveness of diet and physical activity interventions among Chinese-origin populations living in high income countries: a systematic review

Background This review examines the effectiveness of diet and physical activity interventions to reduce cardiometabolic risk among Chinese immigrants and their descendants living in high income countries. The objective of this review is to provide information to help build future interventions aimed at improving diet and increasing physical activity levels among Chinese immigrants. Methods Outcomes included BMI, weight, waist circumference (WC), waist-hip ratio (WHR), cholesterol (LDL, HDL), systolic and diastolic blood pressure (SBP, DBP), hemoglobin A1c (HgbA1c), fasting blood glucose (FBG), and HOMA-IR. Six databases were systematically searched from database inception to date of search (February 2020). Meta-analyses used random effect models to estimate pooled effects of outcomes with 95% confidence intervals. The outcomes assessed were changes in mean outcomes (post-intervention versus baseline) among the intervention group versus control groups. Results Twenty-one articles were included for synthesis, and eight of these were included in the meta-analysis. Among children/adolescents, there were no significant effects of intervention for any of the outcomes having sufficient data for meta-analysis (BMI, WHR, SBP, and DBP). Among adults, the pooled effect including three studies showed significant changes in BMI (effect size = − 1.14 kg/m2; (95% CI: − 2.06, − 0.21), I2 = 31%). There were also significant effects of intervention among adults in terms of changes in SBP and DBP, as the pooled effect across three studies was − 6.08 mmHg (95% CI − 9.42, − 2.73), I2 = 0% and − 3.81 mmHg (95% CI: − 6.34, − 1.28), I2 = 0%, respectively. Among adults there were no other significant effects among the meta-analyses conducted (weight, WC, LDL, HgbA1c, and FBG). Conclusions This review is the first to summarize the effectiveness of diet and physical activity interventions specifically designed for Chinese immigrants living in high income countries. There were clinically meaningful changes in BMI and blood pressure among adults, but evidence was weak for other cardiometabolic outcomes (weight, WC, LDL, HgbA1c, and FBG), and among children, there was no evidence of effect for any cardiometabolic outcomes. Given our mixed findings, more work is needed to support the design of successful interventions, particularly those targeting children and their families. Trial registration The systematic review protocol was registered in PROSPERO on December 17, 2018, the international prospective register of systematic reviews (registration number: CRD42018117842).


Background
People of Chinese origin make up one of the fastest expanding groups in high-income countries such as the United States, Australia and Canada [1]. The cardiometabolic disease profile for this group is generally positive [1], but there are concerns about a high prevalence of type 2 diabetes identified in some studies [2,3] and about increasing adiposity. While measures of adiposity such as BMI and waist circumference are generally low in Chinese-origin populations in high-income countries in comparison with other ethnic groups [3,4], there is evidence that it increases with time living in a highincome country [4], that it is higher in those born to Chinese-origin parents in the United States than in migrants from China [4][5][6], and some evidence that it has been increasing faster amongst Chinese ethnic groups than amongst others [7]. People of Asian origin have a higher risk of cardiovascular disease at a given BMI relative to other ethnic subgroups [3], suggesting that strategies to improve diet and physical activity behaviors may be particularly important for those of Asian origin, including Chinese immigrants and their descendants [8].
There is good evidence of differences in physical activity and dietary practices between Chinese migrant groups and the rest of the population in a number of countries with the largest Chinese-origin populations. There was a higher prevalence of inactivity among Chinese Australians than non-Chinese Australians [3], Canadians of South-East Asian origin (including people with Chinese origins) were more likely to be physically inactive than the White population of Canada [9], those of Chinese origin reported lower levels of physical activity compared with the general population in the UK [10], and not only were Non-Hispanic Asians in New York City less likely to meet physical activity guidelines than non-Hispanic Whites or Blacks, but Chinese Americans were less likely to meet physical activity guidelines than other Asian subgroups [11]. Similarly, in New Zealand those of Chinese ethnicity were less likely to achieve physical activity recommendations [8].
Dietary differences are harder to characterize. Those of Chinese origin ate greater amounts of fruit and vegetables than the general population in the UK and fat intake was relatively low [12], while studies in the United States and in New Zealand found that those of Chinese ethnicity were less likely to meet recommendations for consumption of vegetables than the general population [13,14]. Dietary patterns change with length of residence amongst migrants from China, with migrants to Canada and the United States showing negative changes such as reduced consumption of fruit and vegetables, increased portion sizes and greater consumption of convenience foods [14,15] and a survey of Chinese immigrant mothers living in NYC reported several changes in diet after immigration including a decrease in family meals [7]. Thus interventions to promote physical activity and healthful diets could be particularly beneficial for those of Chinese-origin.
Considerations for developing interventions for Chinese migrants and/or their descendants include: 1) language (whether the intervention was offered in Cantonese, Mandarin, English, etc.); 2) health literacy; 3) traditional Chinese diet; 4) migration and acculturation; and 5) traditional Chinese medicine [16]. Successful interventions may encourage maintenance of healthful dietary practices, incorporate traditional and cultural beliefs, and provide information that would enable the participants to make healthful dietary modifications [17]. Adaptations at a surface level may involve the use of vernacular phrases, role models that represent the targeted group, identifying suitable media channels and settings for recruitment, and employing ethnically matched staff to administer the program [18]. At the deep structure level, adaptations may address the opposing cultural dimensions of collectivism and individualism [18].
In the context of some concerns about diet and physical activity in those of Chinese origin living in highincome countries, and evidence that this group may benefit from tailored interventions, this review examines the effectiveness of interventions designed to modify dietary and physical activity behaviors to reduce cardiometabolic risk in this group. The objective of this review is to provide information to build future interventions aimed at improving the diet and increasing physical activity levels among Chinese immigrants.

Methods
The review was conducted following the PRISMA Protocol for Systematic Reviews (PRISMA) [19] and the protocol was registered in PROSPERO, International prospective register of systematic reviews (CRD4201811 7842).

Information sources and search strategy
In February 2020, co-author (TR), an experienced Medical Librarian, searched PubMed Central, Ovid Medline, Ovid Embase, CABI, Food Science Technology Cinahl and the Cochrane Central Register of Controlled Trials. The Ovid Medline Search is included as supplementary material (Supplementary Table 1) to this article. The search was not limited by language or publication date. Additionally, the citations of included articles were checked and, if relevant, were included in the review.

Eligibility criteria
This review examined diet and physical activity interventions to reduce cardiometabolic risk among Chinese immigrants living in high income countries outside of China. To this end, studies were included in the review if 1) they quantitatively described the effect of an intervention designed to modify dietary and/or physical activity behaviors on cardiometabolic risk factors (BMI, weight, waist circumference (WC), waist-hip ratio (WHR), LDL and/or HDL cholesterol, systolic and diastolic blood pressure (SBP and DBP), hemoglobin A1c (HgbA1c), fasting blood glucose (FBG), and HOMA-IR), and 2) the recipients of the intervention were of Chinese origin and living in a high-income economy, as defined by the World Bank [20]. Exclusion criteria were: studies involving institutionalized populations (as individuallevel control over diet and physical activity behaviors may be restricted), and studies whose samples included residents of Hong Kong, Taiwan, and Macau (as these high-income economies are special administrative regions within China). Interventions could be at any level (individual, community, policy). The only types of studies to be excluded were observational studies in which no intervention was tested. Systematic reviews and meta-analyses on related topics were tagged for review of individual studies, but the review paper itself was not included to avoid double counting of studies. Control groups were comprised of alternative combinations of diet and physical activity interventions, attention control, cross-over designs, or before/after studies.

Study selection and data extraction
Titles and abstracts were screened by four independent reviewers (JB, JW, TP, NA), with each citation receiving two votes. The full-texts of studies with relevant abstracts were assessed for eligibility by two screeners independently (JB, JW). Any conflicts were discussed and resolved through consensus of all four reviewers.
Data from studies eligible for inclusion were extracted using a data extraction form adapted from published sources such as the Cochrane review [21,22]. If pre-and post-intervention means were not provided in the manuscript, the corresponding author was contacted to request the data. Quality assessment was determined using the Cochrane Review's Risk of Bias tool [21], and guidelines provided in the Cochrane handbook for systematic reviews of interventions were used to assess risk of bias [23]. Two reviewers (JB, JW) independently extracted outcomes by reading the full articles, tables, figures and interpretations for the findings and assessed the quality of papers to ensure consistency and to minimize individual bias. Discrepancies were resolved by consensus (TP, NA, JB, JW).

Synthesis of results
A narrative synthesis was used as it allows the compilation of data despite potential differences in research questions, design, or context in order to find a common underlying pattern. If at least two studies included the same outcome variable and pre-and post-intervention values were reported for both the intervention and control group, a meta-analysis was conducted. In cases where multiple post-intervention measurements were available, we extracted the measure that corresponded most closely to the endpoint of the intervention. We stratified analyses by age group (children/adolescents and adults).

Statistical analysis
Where meta-analysis was possible (e.g. pre-post measures were available for intervention and control groups), the analyses involved two steps. The first step was to assess mean differences (MD) in outcomes for both the intervention and control group by comparing changes in the mean as the difference between postintervention and baseline measures. For calculating MD, available adjusted or unadjusted means as reported in the included studies were used. The corresponding changes in standard deviation (SD) were not directly reported in most studies, and therefore was estimated using the formula suggested by the Cochrane handbook for systematic reviews of interventions [23]. A correlation of 0.6 between pre-and post-intervention values was assumed. The second step involved estimating the pooled effect for outcomes, where at least two randomized, controlled trials (RCTs) reported on the same outcome variables. The pooled effects as gain in the intervention group against the change in control group was reported as the pooled effect estimate with 95% CIs. The study weights were equal to the inverse of the variance of effect estimate of each study as suggested by DerSimonian and Laird [24,25]. The overall effect was interpreted as statistically significant if the 95% CIs did not include the null value of 0 (no difference) in their range. Sensitivity analyses were performed to assess whether correlation of 0.5 or 0.8 affected the interpretation of the pooled effect. Heterogeneity, i.e. variation in the intervention effects observed in the included studies, was quantified using the I 2 statistic. Results are to be interpreted with caution where there is significant heterogeneity (I 2 > 50%). Meta-analyses were performed in R software using the 'meta' package.

Study selection
After duplicates were removed, 4443 articles were identified (Fig. 1). The initial screening of titles and abstracts removed 4335 articles, leaving 107 full text articles to be screened by two reviewers independently (JB, JW). Of the full text articles reviewed, 86 articles were excluded for the reasons listed in Fig. 1. Twenty-one articles were included for synthesis, including one study reporting outcomes for both children and adults [26]. Of these, eight provided the pre-and post-intervention means for intervention and control groups, allowing for inclusion in the meta-analysis [26][27][28][29][30][31][32][33].

Study characteristics
Among children/adolescents, the first study was published in 2008 [34] and the most recent study was 2019 [30] ( Table 1). The range of publication dates was wider among adults (1998-2019) ( Table 2). All eight studies conducted among children/adolescents were conducted in San Francisco, CA, USA [26][27][28][29][30][34][35][36], and all but one [26] were led by the same principal investigator (Chen) ( Table 1). Among adults, one study was set in Australia [37], one in Canada [33], and one in South Korea [38], while all others were conducted in the United States [31,32,[39][40][41][42][43][44][45][46][47] ( Table 2). The average sample size was 60 and 63 among studies conducted in children/adolescents and adults, respectively (Tables 1  and 2). The average proportion of female participants was 50 and 64.5% among studies conducted in children/ adolescents and adults, respectively (Tables 1 and 2). The age range for interventions among children/adolescents was three to 18. Among children/adolescents, all interventions included both diet and physical activity components, while among adults, two interventions focused on diet exclusively while three interventions focused on physical activity exclusively (Tables 1 and 2). Among children/adolescents, intervention duration was 2 months for six studies and 6 months for two studies (Table 1). Among adults, intervention duration ranged from 5 weeks to 1 year, with most common duration of 6 months in four studies ( Table 2).

Risk of bias within studies
Among studies conducted in children/adolescents ( Fig. 2  and b), only Chen 2018 [30] had low risk of bias for all criteria. Four of the studies were not evaluated for random sequence generation, allocation concealment, or blinding, as they were not randomized controlled trials. Four studies had a high risk of bias for incomplete outcome data (attrition bias).
Among studies conducted in adults ( Fig. 3a and b), all of the studies had at least one criterion with a high risk of bias. Six of the studies were not evaluated for random sequence generation, allocation concealment, or blinding, as they were not randomized controlled trials. Common criteria rated with a high risk of bias was blinding of outcome assessment (six studies), incomplete outcome data (ten studies), and selective reporting (five studies).

Intervention characteristics
Among children/adolescents, four studies were randomized controlled trials, three studies were pre-post singlearm interventions, and one study included a historical control group (Table 3). The most common intervention was iStart Smart, which was adapted for Chinese American children based on the National Institute of Health's WeCan! program (educational play-based activities teaching self-efficacy, critical thinking, and problem solving skills related to nutrition, physical activity, and coping) [29,30,35,36]. Intervention components included short video clips with hands-on activities to reinforce concepts; interactive dietary software (The Wok); and 60 min exercise classes (basketball, dodge ball, badminton) weekly for 8 sessions. Study participants were provided with a pedometer, activity diary, and books related to physical activity. A one-hour parent workshop was also included to provide reinforcement and social support. Theoretical models included the Ecological Model of Childhood Obesity, Social Cognitive Theory (five studies), Transtheoretical model, and Information-Motivation-Behavior Models (Table 3).
Among adults, three studies were randomized controlled trials, nine studies were pre-post single-arm interventions, and two studies were two-group repeated measures quasi-experimental design (Table 4). Interventions included adaptations of the Diabetes Prevention Program [32,37,39,45] DASH diet [33], a cancer survival program [41], diabetes management programs [43,46,47], walking programs [38,40], community-based programs [42], tai chi [44], and an intervention to incorporate more brown rice in the diet [31]. Theoretical models included Transtheoretical Model, Culture Care Theory, Chronic Care model, Theory of reasoned action, Orem's theory of self care, Empowerment model, RE-AIM, Social Cognitive Theory, and traditional Chinese Medicine principles (Table 4).

Intervention effectiveness
Among children/adolescents, sufficient data were available for meta-analysis for BMI, WHR, SBP, and DBP. The pooled effect including five studies did not show significant changes in BMI (effect size = − 0.27 kg/m 2 ; (95%CI -0.91, 0.36) (Fig. 4a). For WHR, there were also no significant changes over time between groups, (two pooled studies with an effect size − 0.01 (95%CI -0.03, 0.00). There was also no significant effect of intervention in terms of changes in SBP or DBP as the pooled effect across three studies was − 3.41 mmHg (95%CI -9.40, 2.58) and − 4.58 mmHg (95%CI -9.56, 0.41), respectively. Results did not substantively change in sensitivity analyses using 0.5 and 0.8 as the correlation between baseline and follow-up measures (data not shown). For the other outcomes of interest (WC, LDL, HDL, and FBG) ( Table 5), just one study reported findings, and statistically significant differences were only reported for HDL.   Intervention Duration exercise physiologist that were trained in health coaching, group program delivery and standardised data collection used for evaluation. Chinese individuals were screened and referred to this study by their GP. As part of the screening and referral process, GPs administered the AUS-DRISK assessment tool to determine the person's risk of developing diabetes within five years. All individuals at high risk had blood tests to exclude undiagnosed diabetes. Those without undiagnosed diabetes who were medically cleared by their GPs were referred to the study.          RCT For each study arm (brown and white rice), all subjects were provided free rice. Subjects were encouraged to prepare rice items in their daily meals with the food items provided for the duration of the study and they were also advised not to change their usual patterns of cooking and eating.
For each study arm, the supplies provided were enough to meet the calculated total energy requirements for a 4-week period. No rice was provided for the family or other household members.
NR NR Significant decreases in weight and systolic and diastolic blood pressure among brown rice (intervention) group only. Insulin and HOMA, serum AGEs and 8-   For the three single group design studies, Chen 2008 only reported changes in BMI stratified by overweight status [34], while the other two reported minor improvements in BMI and blood pressure (Table 5) [35,36].
For the eleven studies that were not randomized controlled trials (Table 6), minor improvements were documented in BMI, weight, LDL, SBP, DBP, FBG, and HgbA1c. However, without a rigorous comparison group, the effects cannot be attributed to the interventions delivered with certainty. Data from one of the studies was not included in Table 6 due to incompatibility of the scales used to measure outcomes [45].

Discussion
As of February 2020, there were 21 published studies describing behavioral diet and physical interventions in Chinese migrants living in high-income countries. The majority were conducted in adults (n = 13), and just  Eight were conducted in children/adolescents; of these, seven were conducted by the same research group in San Francisco. There were clinically meaningful changes in BMI [48] and blood pressure [49] among adults, but evidence was weak for other cardiometabolic outcomes (weight, WC, LDL, HgbA1c, and fasting glucose), and among children, there was no evidence of effect for any cardiometabolic outcomes. The intervention having the largest change in BMI among adults (− 2.19) had a much smaller effect on the offspring (− 0.29) [26]. Several explanations may help explain the differences in effects observed between adults and children in this study and others. First, postintervention measures were collected 3 months later in children, while mothers' BMI was collected immediately following the intervention. Second, BMI z-scores, which better account for growth stage compared to BMI among children, were not reported by the authors. Furthermore, most of the adult intervention periods were longer-term (6-12 months) whereas most of the studies conducted among children were 2 months in duration.
This report fills a gap in our understanding of the evidence base for behavioral diet and physical activity interventions conducted in Chinese migrants and their descendants living in high-income countries. Other reviews have examined diet and physical activity behaviors among African [50] and South Asian [51] migrants to high-income countries. For example, a review of the effects of diet and physical activity interventions on weight, BMI, and waist circumference among South Asian migrants including 29 studies also observed no significant differences among children but a significant improvement in weight only among adults (mean difference − 1.8 kg, 95% CI − 2.5 to − 1.2 kg) [51].
Limitations must be acknowledged in interpreting these findings. Despite searching seven databases and reference lists for all identified articles, it is possible that relevant studies were missed, if for example, the title or abstract didn't describe analyses specific to Chinese migrants. Although the characteristics of each intervention as are described in this review in order to help identify which intervention components might be effective, given the small sample size and heterogeneity of the studies, the review cannot definitively summarize successful strategies for behavioral diet and physical activity interventions targeted at Chinese-origin groups [52][53][54][55].
Most studies conducted a complete case analysis rather than accounting for loss to follow-up incorporating missing data methods such as multiple imputation. Complete case analyses would overestimate any effect of the intervention if, for example, participants who dropped out lost less weight compared to those who completed the study. We did not make any adjustment for how studies accounted for attrition in our analysis, but attrition bias was accounted for in the quality assessment. In summary, a major limitation of our analyses was having a relatively small number of controlled trials that were suitable for meta-analyses. We only included controlled trials, as opposed to single arm pre-post studies, in the meta-analyses to minimize the likelihood that observed changes in cardiometabolic outcomes were due to factors other than the intervention, particularly in growing children.
Suggestions for improvement include increased attention to (1) how interventions are culturally adapted; (2) the types of behavior change techniques and theories that are used to underpin interventions; (3) loss to follow-up by study arm; (4) variability within the Chinese-origin population, particularly with respect to generational differences that may be important for the design of interventions; and (5) contextual factors, such as whether the setting is rural or urban. These recommendations would enable reviewers to assess how behavior change techniques and theories moderate effectiveness, to assess the equity impacts of interventions, and to examine explanations for heterogeneity between interventions.

Conclusions
Given our mixed findings, more work is needed to support the design of successful interventions, particularly those targeting children and their families. The