Migration from developing countries to affluent, Westernised countries is associated with an increased risk of chronic disease, especially obesity, diabetes and chronic heart diseases(1–Reference Landman and Cruickshank3). For example, the Ghanaian Health and Nutrition Analysis in Sydney found an increase in self-reported BMI (23·8 (sd 2·2) kg/m2 for men and 23·4 (sd 4·5) kg/m2 for women pre-migration v. 27·2 (sd 2·8) kg/m2 and 27·4 (sd 4·7) kg/m2, respectively, post-migration) among Ghanaian migrants(Reference Saleh, Amanatidis and Samman4). The study also found a high prevalence of both type 2 diabetes mellitus (T2DM) and impaired fasting glucose. The prevalence of T2DM was 20 % in men and 11 % in women, and impaired fasting glucose was 22 % and 20 %, respectively. Only 23 % of study participants were aware of their T2DM status, indicating the lack of awareness and the increased risk of medical conditions related to uncontrolled diabetes. These findings were consistent with those reported in studies of the African Diaspora of West African origin(Reference Cooper, Rotimi and Ataman5, Reference Cooper, Rotimi, Kaufman, Owoaje, Fraser, Forrester, Wilks, Riste and Cruickshank6) and Ethiopian migrants to Israel(Reference Cohen, Stern, Rusecki and Zeidler7, Reference Bursztyn and Raz8).
The pattern observed among African migrants has also been reported for other migrant sub-populations. For example, the latest data reported by the Australian Institute of Health and Welfare(1) found that, overall, migrants born in Southern and Eastern Europe, North Africa and the Middle East experience rapid weight gain following migration and have a prevalence of overweight that is significantly higher than the Australian-born population. People whose main language spoken at home is a Middle Eastern, Northern African, Southern Asian, other South-West Asian or East Asian language were found to have higher rates of T2DM than Australian-born people(1). In the USA, Hispanic (Mexican, Cuban and Puerto Rican) Americans have higher levels of overweight and obesity than do white Americans(Reference Pawson, Martorell and Mendoza9). Similarly, in The Netherlands, the prevalence of overweight/obesity among Turkish and Moroccan immigrants is very high at 57–89 %(Reference Dijkshoorn, Nierkens and Nicolaou10). The risk of the escalation of obesity and T2DM has been found to follow a gradient as migrants become more affluent and urbanised, and as nutrition and lifestyle behaviours change(Reference Misra and Ganda11).
Despite the evidence for increased risk of obesity and chronic diseases post-migration, it is not clear whether public health interventions geared toward addressing obesity and obesity-related chronic diseases among migrants have been effective. The aims of the present study were: (i) to determine whether interventions tailored specifically to particular immigrant groups from developing to developed countries decrease the risk of obesity and obesity-related diseases; and (ii) to delineate and examine process indicators that describe why and how a particular intervention has worked.
Methods
Search strategy
MEDLINE (1966–September 2008), CINAHL (1982–September 2008) and PsychINFO (1960–September 2008) were searched. Other databases such as Sociological Abstracts, Science Direct, Web of Knowledge and Google Scholar were also searched. The search strategies utilised are given in Table 1. However, no additional relevant or different studies were found through these additional databases. In addition, a lateral approach involving a review of reference lists in relevant papers/reviews was undertaken. Two more studies were identified but only one met our inclusion criteria.
Search strategies for all of the search engines utilised
/ = or.
Selection of interventions and data extraction
All potentially relevant studies were screened by two reviewers (A.M.N.R. and K.B.) for eligibility using the following criteria: (i) the intervention targeted immigrants from developing to developed countries; (ii) the intervention’s objective was the prevention of obesity, T2DM and/or CHD; (iii) the intervention design was a randomised controlled trial, a ‘quasi-randomised’ trial or a controlled before-and-after study; (iv) at least 40 % of the study sample were non-African American migrants where a multi-ethnic sample was used; (v) an intervention group had at least forty subjects; and (vi) the study was published in a peer-reviewed journal. Papers selected for inclusion (see Fig. 1 for flow chart) were then independently reviewed by two reviewers (A.M.N.R. and K.B.). The review extracted information describing the intervention and outcomes.
Flow chart of study selection
Poor-quality studies, for the purposes of the present review, were defined as those that had any of the following design flaws: inadequate sample size; poor definition of control groups; no clear measurement or control for confounders; no clear or specific definition of outcome measures; or analyses unadjusted for confounders(Reference Egger, Smith and Altman12). Studies without these flaws were categorised as being of good quality and suitable for inclusion.
Exclusion criteria
Studies were excluded if they included non-immigrants, focused mainly on African-Americans, were medical and/or drug interventions, did not define the migrant group, did not define the intervention, did not clearly outline the methodological approaches, and did not include obesity or BMI, T2DM or heart disease as an outcome measure.
Types of outcome measures
Outcome measures included change in BMI; weight, waist and hip measurements; lean tissues mass; body fat; nutritional targets for patients with diabetes; self-reported nutrition adherence for diabetes self-management; diabetes knowledge; self efficacy and management; dietary fat consumption; glycaemic control; and cardiovascular risk factors.
Data synthesis
Due to the heterogeneity of study characteristics and impact indicators, undertaking a meta-analysis was not appropriate. The analysis focused on the narrative, describing the target population, the setting, type and reported impact of intervention, and where possible the effect size. The interventions were assessed for cultural competence and cultural leverage. Culturally competent interventions were those that used community participants’ expertise and social structures to culturally tailor strategies and activities in order to address culture-related barriers; and used cultural practices, products, philosophies or environments in order to make the intervention culturally acceptable and as the means to facilitate behaviour change among participants.
Results
The initial electronic database searches identified 599 studies, of which only thirteen met the inclusion criteria (Fig. 1). Of the thirteen studies, ten (77 %) focused primarily on T2DM and three (23 %) focused on cardiovascular risk factors such as blood pressure and serum lipids. Six out of the thirteen studies (46 %) included anthropometric measures such as BMI and waist:hip ratio. The included studies and their limitations are summarised in Table 2. A summary of intervention outcomes by setting and by number of intervention components is presented in Table 3.
Study characteristics of articles included in the present systematic review
RCT, randomised controlled trail; T2DM, type 2 diabetes mellitus; GP, general practitioner; PBC, perceived behavioural control; HbA1c, glycated haemoglobin; WC, waist circumference; HC, hip circumference; WHR; waist:hip ratio; SBP, systolic blood pressure; DBP, diastolic blood pressure; BP, blood pressure.
Summary of intervention outcomes by setting and number of intervention components
HbA1c, glycated haemoglobin; 0 = intervention did not make a difference to outcome measures; + = intervention had an impact on outcome measures.
Interventions reporting significant effects
Gucciardi et al.(Reference Gucciardi, DeMelo, Lee and Grace13) conducted a 3-month randomised controlled trial to evaluate the impact of culturally competent diabetes education intervention on nutrition adherence and glycaemic control among sixty-one Portuguese Canadian adults with T2DM. The intervention had two components: (i) individual counselling only; and (ii) individual counselling (n 25) in conjunction with group education (n 36). After adjusting for baseline measures, the study found that regardless of the education format, participants in both treatment conditions improved significantly from pre to post intervention. However, the improvement was greater in the multiple intervention (mean difference between multiple v. single intervention) in scores for attitude (0·20 (sd 0·11), P = 0·09), subjective norm (0·25 (sd 0·13), P = 0·06), perceived behavioural control (0·16 (sd 0·11), P = 0·03), intention to adhere to nutrition recommendations (0·20 (sd 0·11), P = 0·03) and self-reported nutrition adherence (0·42 (sd 0·14), P = 0·001), except for glycated haemoglobin (HbA1c; 0·08 (sd 0·34), P = 0·75). The study limitations are summarised in Table 2.
In New Zealand, Simmons et al.(Reference Simmons, Fleming, Voyle, Fou, Feo and Gatland14) conducted a 2-year non-randomised, controlled, church-based intervention programme to reduce diabetes risk factors among Western Samoans. Two Seventh Day Adventist churches were selected for the study through their pastors. In one church, a 2-year nutrition and physical activity education programme was implemented as part of church attendance. After 2 years of intervention the study found that, compared with the control group, those in the intervention group had: (i) reduced waist circumference (change over time, diff = −4 (sd 10) cm v. +2 (sd 7) cm, P < 0·001); (ii) reduced waist and hip circumference (diff = −4 (sd 10) cm v. +2 (sd 7) cm and −5 (sd 9) cm v. +1 (sd 5) cm respectively, P < 0·001); (iii) increased open (diff = +46 (sd 26) % v. +4 (sd 17) %, P < 0·001) and closed (diff = +12 (sd 24) % v. +3 (sd 22) %, P < 0·05) knowledge scores; (iv) increased physical activity (+0·5 (sd 2·5) v. −0·5 (sd 2·4) d/week, P < 0·05; exercising 3 d/week: +22 % v. −8 %, P < 0·05; stopping exercising: 10 % v. 24 %, P < 0·05; initiating exercise: 31 % v. 15 %, P < 0·05; no change in exercise pattern: 59 % v. 61 %, P < 0·001); and (v) reduced dietary fat intakes (seven-item fat score: −14 (sd 28) v. −3 (sd 24), P < 0·001; do not cut fat off meat: −47 % v. −7 %, P < 0·001; do not cut skin off chicken: −9 % v. −4 %, P < 0·05). However, the two groups did not differ in terms of BMI or waist:hip ratio. Despite these positive outcomes, the study suffered some serious limitations (Table 2). In addition, the two churches (intervention and control) were only 3 km away from each other and there was a spill-over effect in that the control group initiated its own exercise programme during the intervention.
Simmons et al.(Reference Simmons, Voyle, Fou, Feo and Leakehe15) subsequently undertook a non-randomised controlled church-based study to compare the impact on weight and exercise of a 2-year diabetes risk reduction programme in four churches in New Zealand. There were two Tongan Latter Day Saints churches with two different bishops; and two Samoan Seventh Day Adventist churches with the same pastor. The intervention and control churches in each congregation were matched as closely as possible for denomination, socio-economic status and organisation. Both churches used leaflets in the first language of their congregation and a specially designed video in English (translated piecemeal by a bilingual church member). The intervention included a 2-year nutrition and physical activity education programme, with sessions customised for Samoan and Tongan people, respectively. The study found that diabetes knowledge increased significantly in both intervention churches when compared with the control church, but more so among Samoans (2-year change in open knowledge score: +46 (sd 26) % v. +4 (sd 17) % for Samoans; +19 (sd 24) % v. +8 (sd 25) % for Tongans; P < 0·001). While there was an improvement in weight, waist circumference and physical activity in the Samoan intervention, no significant change was seen in either control church or the Tongan intervention church. In this study, baseline characteristics were not adjusted for when establishing the effect of the intervention. As in Simmons et al.’s previous study(Reference Simmons, Fleming, Voyle, Fou, Feo and Gatland14), this study also suffered from a small sample size and spill-over effects.
Brown et al.(Reference Brown, Blozis, Kouzekanani, Garcia, Winchell and Hanis16) compared the effect of two culturally competent diabetes self-management interventions on the health beliefs of Mexicans in America. Analyses of covariance in which the baseline values and the 3-month values were used as covariates were conducted to examine the 3- and 12-month outcomes, respectively. After 12 months of intervention, health belief scores improved from baseline for both the compressed intervention (control of effect of diabetes: +0·3, P < 0·01) and the longer and extended intervention (control of effect of diabetes: −0·4, P < 0·01; barriers to taking medication: +0·2, P < 0·05; benefit of therapy: +0·2, P < 0·05). There was an improvement in HbA1c for the extended programme (−1·0, P < 0·05). Individuals with higher intervention attendance reported greater control over diabetes, and high control scores were associated with low HbA1c, explaining 13·1 % of the variance in HbA1c (P < 0·001). The sample size was acceptable and the study used robust power analysis, adjusting for baseline differences.
Griffiths et al.(Reference Griffiths, Motlib and Azad17) conducted a 4-month randomised controlled trial to examine the effectiveness of a culturally adapted, lay-led, self-management programme for Bangladeshi adults with diabetes and other chronic diseases (CVD, asthma and arthritis) in the UK. The intervention comprised six weekly education sessions, each of 3 h duration. The intention-to-treat analysis showed that the programme improved self-efficacy (effect size, i.e. adjusted difference in means: 0·67; 95 % CI 0·08, 1·25; P = 0·025) and self-management behaviour (effect size: 0·53; 95 % CI 0·01, 1·06; P = 0·047). Attending three or more sessions had greater impact than attending fewer sessions, with greater improvements in self-efficacy (effect size: 1·47; 95 % CI 0·50, 1·82; P < 0·001), self-management behaviour (effect size: 1·16; 95 % CI 0·50, 1·82; P = 0·047) and depression scores (effect size: −0·34; 95 % CI −1·22, −0·07; P = 0·028). The sample size was acceptable and the study used robust power analysis, adjusting for baseline differences.
Bell et al.(Reference Bell, Swinburn, Amosa and Scragg18) undertook a study to promote weight loss in Samoan church communities through an exercise programme and nutrition education. Using a quasi-experimental design, the study involved three non-randomised Samoan church communities (two intervention churches, n 365; one control church, n 106) in Auckland, New Zealand. The intervention churches received aerobics sessions and nutrition education over a year. No changes were observed over time in the intervention churches in terms of knowledge about how to reduce the fat content of meals (e.g. diluting coconut milk, removing fat from meat or skin from chicken or eating vegetables frequently). However, the proportion of participants undertaking vigorous physical activity increased by 10 % in the intervention churches but declined by 5 % in the control church (P = 0·007). BMI decreased by 0·2 kg/m2 (from 34·8 to 34·6 kg/m2) in the intervention churches compared with an increase of 0·4 kg/m2 (from 34·3 to 34·7 kg/m2) in the control group (P = 0·046). No group difference was found over time for waist circumference and blood pressure after adjusting for baseline characteristics.
The study by Hawthorne(Reference Hawthorne19) assessed the effect of culturally appropriate health education on glycaemic control and diabetes knowledge among Pakistani migrant women (n 105) in Manchester, UK, using a randomised controlled trial. The intervention arm received one-to-one education on diabetes care for 6 months. Diabetes knowledge increased significantly in both the control and intervention groups. The proportion of women in the intervention with knowledge related to diabetes complications increased from 23 % to 86 % for eye problems, from 28 % to 82 % for heart-related issues and from 7 % to 74 % for feet-related issues, while food knowledge score increased from 71 % to 85 % (all P < 0·05). Changes were also evident in the control group in terms of knowledge about diabetes complications (eye problems, +4 %; feet-related complications, +17 %) and food knowledge score (+4 %). Changes in the intervention group were significantly higher than those in the control group (P < 0·05). In the intervention group, the mean HbA1c varied from 8·9 (se 0·3) % to 8·3 (se 0·3) % and such change was more pronounced in literate than illiterate women. In contrast, HbA1c remained constant in the control group (from 8·8 (se 0·3) % to 8·7 (se 0·3) %), P > 0·05.
Brown et al.(Reference Brown, Lee and Oyomopito20) used a non-equivalent control group to examine the impact of a minimal intervention heart health programme in a sample of Greek migrant women (n 48) attending a Greek Orthodox church in Australia. The intervention consisted of exercise and dietary modification. For the intervention group, there was a significant reduction in BMI (from 29·4 to 27·6 kg/m2; P < 0·001), total skinfold thickness (from 115·4 to 199·9 mm; P < 0·001), proportion of body fat derived from skinfold measurements (from 41·8 % to 40·1 %; P < 0·001), diastolic blood pressure (from 87·1 to 79·3 mmHg; P < 0·001) and aerobic fitness as measured by the exercise heart rate (from 114·2 to 100·2 beats/min; P < 0·001). There was no effect on food habits scores, serum lipids, systolic blood pressure and waist:hip ratio. There was no change in any of the outcome measures over the follow-up time for the comparison group.
Lujan et al.’s(Reference Lujan, Ostwald and Ortiz21) randomised controlled trial aimed to determine the effectiveness of an intervention led by promotoras (community lay workers) on the glycaemic control, diabetes knowledge and diabetes health beliefs of Mexican-Americans (n 149) with T2DM living in a major city on the Texas–Mexico border. The intervention was delivered through Catholic faith-based clinics and behaviour change postcards. Mean HbA1c decreased in the intervention group and increased in the control group at 3 months post baseline, but the differences were significant and more pronounced at the 6-month assessment (decreased by 46 %, from 8·21 (sd 2·2) % to 7·76 (sd 1·87) %, in the intervention group v. increased by 30 %, from 7·71 (sd 1·49) % to 8·01 (sd 1·8) %, in the control group; P < 0·001). The diabetes knowledge mean score change did not differ at the 3-month assessment between the intervention and control groups, but was significantly higher in the intervention group (from 69·1 (sd 13·6) to 77·2 (sd 14·4)) than in the control group (from 66·9 (sd 15·2) to 65·1 (sd 21·0)) at the 6-month assessment (P < 0·001). The diabetes health belief score decreased in both groups from baseline; the difference was not significant at the 3-month assessment but significantly different at the 6-month assessment (from 56·4 (sd 12·2) to 54·6 (sd 8·4) in the intervention group v. from 57·0 (sd 10·6) to 50·8 (sd 13·6) in the control group; P < 0·01).
Features of interventions that work
In summary, seven successful interventions out of ten (70 %) showed significant improvement in addressing diabetes and diabetes-related behaviours (glycaemic control). Six out of the thirteen studies (46 %) incorporated anthropometric data, physical activity and healthy eating as ways to minimise weight gain and diabetes-related outcomes. Of the six interventions that included anthropometric data, only two (33 %) reported improvement in BMI Z-scores, total skinfold thickness or proportion of body fat. Only one in three (33 %) of the studies that included cardiovascular risk factors reported improvement in diastolic blood pressure after adjusting for baseline characteristics.
Interventions that showed improvement had something in common: the application of a cultural competence framework and cultural leverage. That is, they used community participants’ expertise and social structures both to define strategies for addressing culture-related factors and to shape the intervention. For example, in the study by Gucciardi et al.(Reference Gucciardi, DeMelo, Lee and Grace13), the control and intervention emphasised cultural and linguistic competence. Regardless of education format, there was an improvement in outcome measures because patients were more comfortable with the classes being taught in their language and incorporating cultural norms. Similarly, Brown et al.(Reference Brown, Blozis, Kouzekanani, Garcia, Winchell and Hanis16) used group sessions offered in Spanish and facilitated by Mexican-American nurses, dietitians and community workers as well as Spanish-language educational materials. In the same line of reasoning, Griffiths et al.(Reference Griffiths, Motlib and Azad17) successfully utilised Bangladeshi tutors of the same sex as the programme group in a lay-led self-management programme. The culturally adapted programme improved participant self-efficacy to control chronic disease and increased the use of self-management skills. Lujan et al.(Reference Lujan, Ostwald and Ortiz21) also used trained, bilingual lay workers to implement a culturally competent diabetes education programme which resulted in decreased HbA1c and increased diabetes knowledge.
In contrast, Simmons et al.(Reference Simmons, Fleming, Voyle, Fou, Feo and Gatland14) focused on community social structures where the community church was utilised as the intervention setting because the church is recognised as the social centre for the Western Samoan and Tongan cultures. The intervention was culturally tailored and held as part of church services, utilising translated and culturally relevant and tested educational tools. The follow-up study(Reference Simmons, Voyle, Fou, Feo and Leakehe15) found that participation in the intervention by the Tongan participants was less than that of the Samoan participants, suggesting that the role of community social structures as mediating influences in population-based interventions is important for success. Use of the culturally tailored and appropriate nutrition and exercise intervention resulted in decreased BMI and increased intensity of leisure-time activity, but had little impact on nutrition-related behaviours(Reference Bell, Swinburn, Amosa and Scragg18). Programmes that apply practical tools and seek to overcome socio-cultural (using existing community social groups and networks and run in familiar settings) and linguistic barriers maximise participation and adherence to activities, and result in greater health outcomes(Reference Brown, Lee and Oyomopito20).
Interventions not reporting significant effects
Four studies that reported no significant effect on obesity-related health issues were reviewed. Baradaran et al.(Reference Baradaran, Knill-Jones, Wallia and Rodgeres22) conducted a study with South Asians in Scotland to determine the effectiveness of a diabetes education programme. Participants, who were mainly Pakistani (72 %) and Indian (23 %) with diabetes and over the age of 30 years, were randomly allocated to two groups: one ethnic intervention group (n 59) and one ethnic control group (n 59). A third group of white people (n 27) served as a control group. Since patients had different religions and came from different cultures, three programmes were implemented for women only, two programmes for men only, and one programme for mixed gender. A total of eighteen educational sessions on diabetes management were offered in each programme as part of the intervention. On completion of the intervention all mean scores for both the intervention group and the control group were significantly higher. The study found that the intervention group showed significant improvements in scores for knowledge (+12·5 %), attitudes towards seriousness of diabetes control (+13·5 %), complications (+8·1 %), and health and diabetes practice (+20·0 %). However, there were also changes in the ethnic and white control groups but the differences in improvement between these two control groups were not significant. The study had a small sample size and the analysis did not adjust for baseline characteristics.
In The Netherlands, Uitewaal et al.(Reference Uitewaal, Voorham, Bruijnzeels, Berghout, Bernsen, Trienekens, Hoes and Thomas23) conducted a prospective controlled study to examine the effect of a diabetes education programme on glycaemic control and cardiovascular risk factors in Turkish T2DM patients. Participants were offered seven individual educational sessions and three group sessions or routine care, and glycaemic control and cardiovascular risk factors were used as outcome measures. After 1 year of the intervention there were no significant differences in glycaemic control (HbA1c and fasting plasma glucose) between the intervention and control groups. A significant decrease in HbA1c was observed in women with HbA1c > 7 % at baseline but not in the other subgroups studied. Serum lipid concentrations, blood pressure and BMI remained unchanged in the intervention group. Although the study adjusted for baseline characteristics, the sample size was too small and inadequate to detect any differences between the two groups.
Davis et al.(Reference Davis, Ventura, Alexander, Salguero, Weigensberg, Crespo, Spruijt-Metz and Goran24) conducted a study of twenty-three female 12- to 17-year-old Latinos in the USA. Two versions of a 12-week carbohydrate nutrition intervention were compared in relation to dietary, physiological and metabolic effects. One version of the intervention was disseminated via an individualised home-based format (n 11) while the other was delivered in a group classroom-based format (n 12). Repeated-measures analyses of covariance were used after controlling for covariates and baseline characteristics. Mixed modelling showed no significant differences in changes in dietary intake between intervention groups, but both groups significantly reduced their intake of added sugar, sugary beverages and refined carbohydrates by 33 %, 66 % and 35 %, respectively, while dietary fibre consumption increased significantly by 44 % (P < 0·01) throughout the 12 weeks. There was a significant time effect for BMI Z-scores within each intervention group (P < 0·05), exhibiting significant improvements (Z-score decreased by 0·1 in each of the interventions). There was no significant time by intervention group interaction for any of the physiological or metabolic variables, indicating that change over time was not significantly different between intervention groups. The lack of difference between the two groups may have been due to the small sample size.
Spruitj-Metz et al.(Reference Spruijt-Metz, Nguyen-Michel, Goran, Chou and Huang25) undertook a school-based intervention to increase physical activity and decrease sedentary behaviours using a pair-matched cluster randomised design in predominantly Latino middle-school girls (n 459; mean age 12·5 years). The intervention involved media-based physical activity in class for five (longer sessions) to seven (shorter sessions) consecutive school days. Students received materials on physical activity (print or verbal, e.g. physical activity fact sheets distributed and discussed) and sedentary behaviour learning activities. Students in the intervention also developed animated Public Service Announcements (created by a team of seven to ten children in each intervention school). The control group received no intervention. Outcome measures included BMI, total energy expenditure and percentage body fat. Hierarchical regression model using changes in outcome measures over the follow-up period suggested that the intervention had a significant effect on reducing time spent on sedentary behaviours (−0·27 (se 0·14), P < 0·05). The intervention did not have any effect on BMI percentiles or Z-scores (P = 0·27–0·49), percentage body fat (P = 0·86) or total energy expenditure (P > 0·05).
Discussion
We searched for evidence that interventions tailored specifically to particular immigrant groups from developing to developed countries decrease the risk of obesity and obesity-related diseases. We found thirteen studies that met our inclusion criteria. The analyses of these studies suggested that a number of factors may contribute to successful outcomes. Interventions that were tailored to the cultural needs of the target population in terms of cultural norms, attitude and beliefs(Reference Gucciardi, DeMelo, Lee and Grace13), and that were implemented within the community setting and utilised translated and culturally relevant educational tools, and community-based trained diabetes and physical activity educators(Reference Simmons, Fleming, Voyle, Fou, Feo and Gatland14–Reference Brown, Blozis, Kouzekanani, Garcia, Winchell and Hanis16, Reference Uitewaal, Voorham, Bruijnzeels, Berghout, Bernsen, Trienekens, Hoes and Thomas23), showed greater improvements in diabetes-related behaviours and glycaemic control, and in making changes in dietary and exercise patterns. The interventions’ effect on anthropometric outcomes such as BMI or waist:hip ratio was negligible or non-existent, and this may have been due to the relatively short follow-up periods. Notwithstanding this, the impact on diabetes was greater in interventions that encompassed multiple strategies, and studies that did not find any impact on diabetes had sample sizes too small to be able to detect any difference between the intervention and the control groups.
The findings suggest that community social structures are important mediating influences in migrant population-based interventions for diabetes and diabetes-related problems. The emphasis on culturally relevant group interventions and community structures in addressing diabetes is consistent with the literature(Reference Utz, Williams, Jones, Hinton, Alexander, Yan, Moore, Blankenship, Steeves and Oliver26, Reference Norris, Nichols and Caspersen27), and the positive impact could be due the fact that participants were more comfortable with the classes being taught in their language and incorporating cultural norms. However, given that three out of the seven studies showing an impact used small sample sizes and failed to adjust for baseline differences, the observed evidence can be described as limited. Although most of the studies with small sample sizes and failure to control for baseline characteristics found an intervention effect on diabetes regardless of the education format, in a relatively unbiased group intervention offered in Spanish and facilitated by Mexican-American nurses, dietitians and community workers, Brown et al.(Reference Brown, Blozis, Kouzekanani, Garcia, Winchell and Hanis16) reported a different pattern. These authors incorporated increased nutritional content and Spanish-language educational materials, as well a larger number of hours in instructional and group sessions. They found that, when the interventions were implemented over an extended period of time and with increased content, the dose effect of the intervention seemed to underlie the programme success.
That the present review is based only on a small number of studies that met the inclusion criteria underlines the paucity of data in this very important research area, and hence the need for more investigations. Future research must involve large numbers of participants from immigrant backgrounds and should ensure that adequate control groups are included, and that spill-over effects are minimised. Interventions that include anthropometric outcomes also need to be of longer duration and intensity to be able to observe a trend in outcomes. However, the reviewed papers provide a base on which to build studies that will identify low-cost, effective health promotion programmes that can be included in large community-based interventions. What has been adequately described in these studies is the format and the intensity of the intervention delivery, the uptake of the intervention and the training of the individuals involved with intervention delivery. What is missing and requires further evaluation includes the most appropriate method for recruiting migrant populations, the duration of the intervention, and follow-up approaches.
Study limitations
The present review has some limitations worth outlining. Although our main interest for inclusion was randomised control trials, ‘quasi-randomised’ trials or controlled before-and-after studies, the included studies varied considerably in terms of design, follow-up period, outcome measures, target population and setting, making synthesis of results and comparison of outcomes across studies difficult. Hence, it was impossible to measure whether observed changes in diabetes behaviours, glycaemic control, and dietary and exercise patterns were of clinical, not just statistical, significance.
The current systematic review did not include articles published a language other than English because the research team did not have adequate linguistic skills and competence in other languages. It is possible that there may be other studies published in other languages that could shed further light on in the issue investigated herein. Further information might also be elicited through expert opinion. It is also important to note that studies without positive outcomes tend to be harder to get published, so the literature available may not represent all of the knowledge generated by research.
Conclusion
Although the literature on interventions tailored specifically to immigrant groups targeted at obesity and related health concerns is limited, there are clear messages from the articles reviewed. Culturally tailored and language-specific educational programmes are more likely to engage participants and result in more efficacious outcomes if designed well. The findings that culturally tailored and facilitated interventions aimed at diabetes among immigrants provide increased outcome measures in the target culture compared with generalised interventions, and that intervention content is more important than the duration or venue of the intervention, require further investigation. Any obesity and chronic disease-related intervention to be aimed at a particular immigrant group needs to be culturally competent and research should be done into the cultural expectations, beliefs, behaviours and practices of the target group, which should be taken into account when designing the intervention programme.
Acknowledgements
The authors acknowledge a grant assistance provided by the Victorian Health Promotion Foundation. The authors declare that there has not been any conflict of interest. All authors contributed to the conceptualisation of the study. A.M.N.R. and K.B. reviewed included papers and drafted the manuscript. D.M. and B.S. reviewed the manuscript.
