• Open Access

Outcome results for the Ma'alahi Youth Project, a Tongan community-based obesity prevention programme for adolescents


KF Fotu, Tonga Health System Support, Ministry of Health, Vaiola Hospital, Nuku'alofa, Tonga. E-mail: sitafotu@yahoo.com


Tonga has a very high prevalence of obesity with steep increases during youth, making adolescence a critical time for obesity prevention. The Ma'alahi Youth Project, the Tongan arm of the Pacific Obesity Prevention in Communities project, was a 3-year, quasi-experimental study of community-based interventions among adolescents in three districts on Tonga's main island (Tongatapu) compared to the island of Vava'u. Interventions focused mainly on capacity building, social marketing, education and activities promoting physical activity and local fruit and vegetables. The evaluation used a longitudinal design (mean follow-up duration 2.4 years). Both intervention and comparison groups showed similar large increases in overweight and obesity prevalence (10.1% points, n = 815; 12.6% points, n = 897 respectively). Apart from a small relative decrease in percentage body fat in the intervention group (−1.5%, P < 0.0001), there were no differences in outcomes for any anthropometric variables between groups and behavioural changes did not follow a clear positive pattern. In conclusion, the Ma'alahi Youth Project had no impact on the large increase in prevalence of overweight and obesity among Tongan adolescents. Community-based interventions in such populations with high obesity prevalence may require more intensive or longer interventions, as well as specific strategies targeting the substantial socio-cultural barriers to achieving a healthy weight.


Overweight and obesity are major public health issues globally, and are especially important among Pacific nations, including the Kingdom of Tonga. Tonga, a nation of just over 101,000 people, has the fourth highest prevalence of overweight/obesity in the world (1) with recent studies showing prevalence rates of 84% among male adult and 93% among female adult (2). For adolescents, approximately one-third of male and approximately one half of female adolescents (3) are overweight/obese. This sharp weight increase during young adulthood (4) makes it critical to prevent unhealthy weight gain among adolescents. Obesity and its chronic disease consequences of diabetes and cardiovascular diseases are a large health burden in Tonga with non-communicable diseases accounting for 10.4% of hospital admissions but, a disproportionate, 19.6% of the health budget (5). Tonga, like other Pacific Island countries, has limited resources and this amplifies the medical and social burden of obesity.

The nature of food, transport and built environments impact on everyday eating and physical activity patterns (6). The Tongan food environment has become increasingly ‘obesogenic’, as traditional foods (e.g. root crops, fish and green vegetables) have been replaced with less healthy imported foods (e.g. corned beef and carbonated drinks) that are high in fat, salt and/or sugar (7). This nutrition transition has been coupled with a rise in sedentary behaviours in the Pacific (8). Furthermore, strong socio-cultural factors support preferences for a larger body sizes relative to Western ideals (9), as well as the provision of ample quantities of food to express love, care and respect (10–12). In contrast to many Western societies, where healthy choices are more common in those with high incomes and educational attainment, health is not a primary consideration in food selection. Indeed, Tongans with high incomes purchase higher quantities of unhealthy foods relative to Tongans on lower incomes (13). Together, the food environment and underlying socio-cultural and socioeconomic factors create an obesogenic environment that promotes unhealthy weight gain and predisposes the population to non-communicable diseases at increasingly younger ages.

Evidence-based strategies are essential for reducing unhealthy weight gain and, in Tonga, the rapid weight gain during adolescence makes this a crucial life stage to focus on. Interventions aiming to change attitudes, behaviours and obesogenic environments need to be holistic and reflect environmental, socio-cultural and socioeconomic factors that contribute to energy-dense diets and sedentary lifestyles (14). For community-based interventions, priority needs to be given to multi-strategy, multi-setting solutions, particularly when interventions target children and adolescents (15). Controlled obesity prevention trials in childhood and adolescence are few in number, mostly short-term (<1 year), focused on just a few strategies (e.g. education or social marketing) and have generally not arrested unhealthy weight gain (16–18). To date, there have been no well-evaluated obesity prevention interventions in the Pacific, despite the magnitude of obesity as a health problem in the region. This paper presents the results of the Ma'alahi Youth Project (MYP), the first community-based intervention to target adolescent obesity in the Kingdom of Tonga. The MYP was part of the Pacific Obesity Prevention in Communities (OPIC) project, which included 3-year community-based interventions in adolescents in four countries (19–21).



Ma'alahi Youth Project was the Tongan arm of the Pacific OPIC project described elsewhere in this supplement (22). MYP aimed to build the capacity of communities and schools to create their own solutions for promoting healthy eating, physical activity and healthy weight gain in adolescents aged 11–19 years and their families. MYP used social marketing approaches, community capacity building and grass roots activities to promote healthy behaviours, including eating breakfast, increasing water, fresh fruit and vegetable consumption, participation in organized sports and physical activity during and after school, and reducing sweet drink consumption and sedentary activities. The details of the intervention are described elsewhere (19). While the MYP objectives were common to the three Tongan intervention sites, the implementation processes were contextualized for each school and community, and the programmes were implemented from 2006 to 2008.

Study design

Ma'alahi Youth Project was evaluated using a quasi-experimental design with a longitudinal cohort follow-up of the intervention and comparison groups. The intervention group comprised Tongan secondary school students aged 11–19 years in the districts of Houma, Nukunuku and Kolonga on the main island of Tongatapu. The sites were selected on the basis of having: an adequate adolescent population, sufficient settings for interventions (e.g. schools, churches, community organizations), the presence of ‘champions’ for change, well-demarcated boundaries, a single administrative area and ease of access for the research team. The intervention targeted 22 villages in the three districts and seven schools, four of which were located outside the intervention area, but had a high percentage of students who resided in the intervention districts. The comparison group was drawn from all six secondary schools on the island of Vava'u, the second largest island in the Kingdom.

Data collection

The methods for the OPIC project are outlined elsewhere (22) but, briefly, anthropometric data, including height, weight, waist circumference and body fat percentage, were collected by trained researchers and behavioural and quality of life survey data were collected at school during school hours or in the villages after hours. The questionnaire was translated into Tongan, back-translated and piloted to ensure accuracy and cultural validity. Health-related quality of life was measured using two instruments: the Assessment of Quality of Life instrument (AQoL-6D) developed by Hawthorne et al. (23) in Australia which was specifically calibrated for the Tongan adolescent population (24) and the Pediatric Quality of Life Inventory 4.0 (PedsQL; generic module for 13- to 18-year-olds) developed by Varni and colleagues (25,26).

Baseline data were collected between September 2005 and March 2006 and a second wave of new participants entering form 1 (aged 11–12), in February and March 2007 to reach target recruitment numbers. Follow-up data were collected between April and December in 2008, either at school or in communities if participants had left school. At follow-up, questionnaire survey data were not collected among those who had left school because the behavioural questions related to school days. For this and reasons of time pressures in follow-up data collections (anthropometry was done as the priority), the numbers of participants with behavioural and quality of life outcomes are much less than those with anthropometry outcomes.

The 2007 WHO Reference standards for age/gender-specific body mass index (BMI) centiles and cut-offs (27) were used to define standardized body mass index scores (BMI-z) and to classify adolescents' weight status as either normal, overweight or obese. Adolescents were measured using a portable stadiometer (Surgical and Medical PE87) for height to the nearest 0.1 cm. A Tanita Body Composition Analyser (Model BC 418, Wedderburn Australia) was used to collect body-weight and bioelectrical impedance. Percentage body fat was derived from bioelectric impedance measures and equations validated for multi-ethnic adolescent populations (28). Trained researchers performed single measures in all cases according to detailed protocols (29).

The Tonga National Health Ethics Research Committee and Deakin University Human Research Ethics Committee both provided ethics approval for the study. Written informed consent was obtained from parents, teachers, principals and participants prior to the study. The study was registered as a Clinical Trial ACTRN12608000346370 with the Australian New Zealand Clinical Trials Registry.


All variables were checked for missing and out-of-range values. Cases with outlying (>3 SD from mean) values on the anthropometric variables at baseline or follow-up were removed from analyses (30). Demographic data were analysed using descriptive statistics, independent groups' t-tests or, where appropriate, chi-squared tests. Differences between follow-up (participants who were measured twice) and non-follow-up (those who were measured once) were tested with t-tests or chi-squared tests as appropriate, and where significant effects were discerned these variables were then entered into a logistic regression model for further testing. The difference from baseline to follow-up in prevalence of overweight/obesity within condition and gender was tested for significance using Newcombe's paired differences (31). Differences in follow-up anthropometry and quality of life were determined by separate linear regression models with group (intervention or comparison) entered into the model with the following covariates: baseline variable, age at follow-up, height at follow-up (weight), gender and duration between measurements. Differences in follow-up weight status and behaviours (categorical measures) were also determined by separate logistic regression models with group (intervention or comparison) entered into the model with the following covariates: baseline variable, age at follow-up, gender and duration between measurements. Analyses were conducted using Stata release 11.0 (StataCorp, College Station, TX, USA; 2009), and in all cases, P < 0.05 was considered statistically significant.


Figure 1 shows the flow diagram for participants of the two groups. There were 2,610 adolescents living in the intervention communities, of whom 41% were measured at baseline and, of those, 75% (n = 815) were re-measured at follow-up. In the comparison group, 2,182 adolescents were eligible to participate, and 64% were measured at baseline and, of those, 66% (n = 897) were followed up. The majority of participants lost to follow-up had moved from the area to attend senior secondary schools or tertiary institutions or to live elsewhere.

Figure 1.

Flow diagram of participants.

The demographic characteristics of participants and differences between characteristics of follow-up and those lost to follow-up are indicated in Table 1. The intervention and comparison groups were well matched at baseline. At baseline, intervention participants were younger (and consequently, lighter and shorter) than the comparison group. The followed-up group were more likely to be younger (OR 1.36; P < 0.001), female (OR 0.71; P < 0.001) and with a lower BMI-z (OR 1.36; P = 0.04) compared to the lost to follow-up group. The main reason for loss to follow-up was that those participants had moved away, as shown in Fig. 1. Older male adolescents in Tonga often move out of the district so the composition of the followed-up group was not surprising. Additionally, there was a rapid increase in the prevalence of overweight and obesity during adolescence so the higher mean BMI-z of the group who was not followed up was also unsurprising. The association found between BMI-z and follow-up is therefore unlikely to be related to refusal by heavier participants. There were no differences between intervention and comparison groups within those not followed up.

Table 1.  Unadjusted baseline and follow-up characteristics of the participants
  • *

    Follow-up and non-follow-up groups differ at baseline (P < 0.5).

  • Intervention and comparison groups differ at baseline (P < 0.5).

  • Includes three participants in the thin category.

  • BMI, body mass index; BMI-z, standardized body mass index; SD, standard deviation.

Total male n (%)815 (45.9)897 (41.4)1712 (43.5)1002(50.1)*
Age (SD) (years)14.4 (2.0)16.8 (2.2)15.2 (1.8)17.54 (1.7)14.8 (1.9)17.2 (2.0)15.8 (2.2)*
Height (SD) (cm)162.1 (9.2)168.0 (7.9)164.5 (8.6)168.8 (7.8)163.4 (9.0)168.4 (7.9)166.4 (9.4)*
Weight (SD) (kg)61.1 (14.6)71.6 (13.8)64.5 (15.1)73.7 (13.3)62.9 (14.9)72.7 (13.668.4 (15.6)*
BMI (SD) (kg m−2)22.9 (4.1)25.2 (4.2)23.6 (4.4)25.8 (4.2)23.3 (4.3)25.5 (4.2)24.5 (4.4)*
BMI-z score (SD)0.9 (0.90)1.1 (0.90)0.9 (1.0)1.1 (0.90)0.9 (1.0)1.1 (0.90)1.1 (0.90)*
Normal weight (%)53.643.553.841.353.742.446.0*
Overweight (%)30.836.632.239.931.538.337.0*
Obese (%)15.619.914.018.814.819.317.0*
Overweight/obese (%)46.456.546.258.746.357.654.0*
Fat percentage (SD)28.9 (9.7)28.7 (11.8)27.2 (10.8)29.8 (12.4)28.0 (10.3)29.3 (12.2)28.0 (10.9)
 Male23.8 (8.8)18.6 (8.0)18.7 (8.1)17.0 (6.9)21.3 (8.8)17.8 (7.5) 
 Female33.3 (8.0)37.4 (6.6)33.2 (8.0)38.7 (6.0)33.3 (8.0)38.1 (6.3) 
Time between measures (SD) (years) 2.4 (0.70) 2.4 (0.40)   

Anthropometric outcomes

Differences (unadjusted and adjusted) in anthropometric outcomes between the intervention group and the comparison group from baseline to follow-up are presented in Table 2. At follow-up, the intervention group recorded less body fat percentage than the comparison group and this finding was statistically significant when controlling for baseline variable, age, height, gender and duration between measures. There were no statistically significant differences in outcomes in weight, BMI and BMI-z, or prevalence of overweight/obesity between the intervention and comparison groups.

Table 2.  Unadjusted and adjusted* differences in outcome measures between intervention and comparison (reference) groups and by subgroups
MeasureDifference (unadjusted)Difference (adjusted)SEP
  • Bolding indicates significance P < 0.05.

  • *

    Adjusted for baseline variable, age at follow-up, height at follow-up (where applicable), gender and time between measurements using linear and logistic regression.

  • Percentage points.

  • Odds ratio.

  • BMI, body mass index; BMI-z, standardized body mass index.

 Weight (kg)
 Body fat percentage−2.4−1.460.21<0.001
 Proportion overweight/obese−2.5−
 Weight (kg)
 Body fat percentage−3.36−1.210.36<0.001
 Proportion overweight/obese−0.08−
 Weight (kg)1.22−0.040.470.93
 Body fat percentage−1.40−1.670.21<0.001
 Proportion overweight/obese−1.670.060.190.74

A visual representation of the gender differences in the prevalence of overweight/obesity between baseline and follow-up is shown in Fig. 2. The combined prevalence of overweight and obesity increased substantially over the mean follow-up duration of 2.4 years in both intervention and comparison groups (approximately 11% points while controlling for duration between measures) with intervention and comparison male adolescents increasing by +3.2 and +7.7% points, respectively, and female adolescents by +19.9 and +13.1% points, respectively.

Figure 2.

This graph is a visual representation of the within-group change in prevalence of overweight/obese adolescents from baseline to follow-up while controlling for duration between measures. The far left of the graph is change in prevalence of overweight/obese in intervention (11.1, CI 7.9, 14.2) and comparison (10.8, CI 7.8, 13.7), the centre is the male point estimates (intervention 3.2, CI −0.6, 7.0; comparison 7.7, CI 3.2, 12.3) and the right is the female point estimates (intervention 19.9, CI 13.0, 22.7; comparison 13.1, CI 9.1, 16.9). The point estimate for the intervention is denoted by the filled marker and for the comparison by the unfilled marker.

Behavioural outcomes

The (unadjusted) proportions of adolescents from baseline and follow-up from the intervention and comparison groups reporting relevant eating and physical activity behaviours and the adjusted odds ratios are shown in Table 3. Overall, the findings suggest that the intervention had few positive effects on these targeted behaviours among adolescents. Relative to the comparison group at follow-up, significantly more intervention participants brought their lunch from home and fewer bought snack foods from a shop or takeaway store after school. However, these positive findings were offset by a number of negative trends in the intervention group. Specifically, intervention participants reported reductions in regular breakfast consumption and fruit and vegetable consumption and increases in lunchtime inactivity and drinking sugar-sweetened soft drinks. Consumption of several after-school snacks and the resulting outcomes were more negative than those observed in the comparison group (Table 3). There were also some negative outcomes relating to physical activity. At follow-up, a smaller proportion of intervention participants walked/rode to or from school and engaged in after-school activity than comparison group participants. The two measures of quality of life were higher at follow-up for both groups but the increase was smaller for the intervention group than for the comparison group. The adjusted relative difference of approximately 4.5% points for the PedsQL measure was a moderately large difference.

Table 3.  Unadjusted proportions (95% CI) for behavioural and environmental measures at baseline and follow-up for intervention and comparison (reference) groups and adjusted* odds ratios (OR)
 InterventionComparisonOR (adjusted)P-value
  • Participant numbers for intervention (I) and comparison (C) groups for each item are included.

  • *

    Adjusted for baseline variable, age at follow-up, gender and time between measurements.

  • Among students living within 30-min walk from school; maximum trips were 10 per week to or from school.

  • Assessment of Quality of Life (AQoL), range is from 0 to 1.

  • §

    Pediatric Quality of Life Inventory (PedsQL), range is from 0 to 100.

Eating and diet      
 Breakfast (4–5 d) (I: 175, C: 392)62.9 (55.7, 70.1)43.4 (36.0, 50.8)68.4 (63.7, 73.0)58.2 (53.3, 63.1)0.630.02
 School lunch from home (I: 222, C: 437)14.4 (9.8, 19.1)24.3 (18.7, 30.0)4.8 (2.8, 6.8)7.8 (5.3, 10.3)2.880.001
 Combined fruit and vegetables (≥5 serve per d) (I: 191, C: 437)38.2 (31.3, 45.1)28.8 (22.3, 35.2)29.1 (24.8, 33.3)30.9 (26.5, 35.2)0.800.30
 Sugar-sweetened soft drink (≥3 d) (I: 227, C: 437)59.0 (52.6, 65.5)67.4 (61.3, 73.5)54.9 (50.2, 60.0)57.2 (52.6, 61.9)1.690.005
 Fruit drink/cordial (≥3 d) (I: 227, C: 437)65.2 (59.0, 71.4)65.2 (59.0, 71.4)81.5 (77.8, 85.1)68.9 (65.2, 73.9)0.930.70
 Snack food bought from shop or takeaway after school (≥3 d) (I: 224, C: 437)52.2 (45.7, 58.8)35.3 (30.0, 41.6)49.2 (44.5, 53.9)50.3 (45.6, 55.0)0.550.001
 Eat fruit after school (most days) (I: 215, C: 436)17.2 (12.1, 22.3)23.3 (17.6, 28.9)11.0 (8.1, 14.0)15.4 (12.0, 18.8)1.290.27
 Eat bread, toast, buns or sandwiches after school (most days) (I: 172, C: 437)62.6 (56.2, 69.0)61.3 (54.8, 67.7)73.2 (69.1, 77.4)72.8 (68.6, 77.0)0.690.05
 Eat biscuits, potato chips or snacks after school (most days) (I: 172, C: 436)15.1 (9.7, 20.5)23.8 (17.4, 30.2)8.9 (6.3, 11.6)12.2 (9.1, 15.2)2.210.001
 Eat pies, takeaways or fried foods such as French fries after school (most days) (I: 227, C: 437)10.1 (6.2, 14.1)23.3 (17.8, 28.9)5.9 (3.7, 8.2)10.1 (7.2, 12.9)2.790.001
Activity and inactivity      
 Walk/cycle to/from school (more than 5 times per week) (I: 149, C: 302)43.6 (35.6, 51.6)24.8 (17.9, 31.8)51.0 (45.3, 56.7)54.0 (48.3, 59.6)0.290.001
 Recess (mostly inactive – sat down) (I: 216, C: 437)20.8 (15.4, 26.3)36.1 (29.7, 42.5)26.3 (22.2, 30.5)32.3 (27.9, 36.7)1.170.41
 Lunch time (mostly inactive – sat down) (I: 215, C: 437)20.0 (14.6, 25.4)44.7 (38.0, 51.3)27.2 (23.0, 31.4)29.5 (25.2, 33.8)2.010.001
 Active after school (3 to 5 d per week) (I: 227, C: 437)49.0 (42.4, 55.4)27.3 (21.5, 33.1)43.7 (39.0, 48.4)33.9 (29.4, 38.3)0.580.008
 Average time watching TV, videos, DVDs per d (% ≤2 h) (I: 163, C: 437)73.0 (66.2, 80.0)82.8 (77.0, 88.6)88.1 (85.1, 91.1)80.4 (76.3, 83.8)1.290.32
 Average time playing video games, electronic games or used computer (not for homework) per d (% ≤1 h)81.6 (75.6, 87.6)81.6 (75.6, 87.6)84.9 (81.5, 88.3)75.8 (73.7, 81.5)1.190.48
Quality of life      
 AQoL (I: 130, C: 378)0.60 (0.28)0.65 (0.28)0.63 (0.26)0.69 (0.24)−0.040.12
 PedsQL§ (I: 215, C: 406)68.8 (16.6)70.6 (13.1)69.6 (15.2)74.3 (14.8)−4.660.001


This study reports on key outcomes for the MYP, which is the first community-based intervention to target reductions in unhealthy weight gain among Tongan adolescents, a population group undergoing rapid age-related increases in obesity prevalence. The MYP targeted selected behaviours in village and school settings and aimed to build the capacity of communities and schools to promote healthy eating, physical activity and healthy weight in adolescents. While there was a 1.5% relative reduction in body fat percentage in the intervention group, there was no difference between the intervention and comparison groups for the other anthropometric outcomes. Furthermore, the intervention group showed no consistent patterns of improved dietary and physical activity patterns and indeed showed a relative reduction in quality of life.

Changes in body composition and behaviours

The improvements in body fat percentage in the absence of improved outcomes in BMI and BMI-z is difficult to interpret. These findings could be explained by sufficient increases in physical activity to ensure that weight gain comes mainly from increases in lean mass rather than fat mass (32). While the MYP intervention activities did have a strong focus on increasing physical activity, the relevant behavioural indicator variables we measured (self-reported activity during school, after school and on the weekends, transport to and from school and screen-based behaviours) did not show any suggestion of increased physical activity in the intervention group relative to the comparison group. Differences in pubertal development between the groups may have been another explanation for the observed outcomes but this was not directly measured. However, body fat percentage was calculated using gender-specific equations that included age, height and weight for male adolescents and height and weight for female adolescents (28). Additionally, analyses were adjusted for age, gender and duration to control for differences in pubertal developments.

Also, we did not observe any clear pattern of improved eating patterns over the intervention period. While it was not known what types of food and drink students brought from home for lunch, it is likely that lunches brought from home would have been more healthy than the energy-dense, nutrient-poor food and drink that could be purchased at school (33), especially the commonly available deep-fried pancakes, noodles and sweetened drinks. Likewise, intervention students reported decreases in the purchase of snack foods on the way home from after school, but overall this was offset by increased consumption of soft drinks, biscuits/chips/snacks, takeaway foods, and chocolates and sweets at home. Taken together, these results indicate that while fewer students purchased unhealthy food on the way home from school, more of these items were available at home. Further examination of the purchasing patterns of Tongan adolescents is needed, given the high consumption of soft drinks and the very high prevalence of obesity.

There are a number of factors that may have contributed to the minimal findings with regard to the anthropometric and behavioural changes. First, whole-of-community intervention programmes have long establishment phases and the real momentum in the MYP programme was occurring at the end of the intervention period (19). Undertaking measurements at the start and end of the intervention period truncated the total duration of intervention time – three school years is shorter than three calendar years because of the school holidays when adolescents are harder to access. Second, the intensity of the dose delivered by the MYP may have been insufficient to effect beneficial anthropometric and behavioural changes. Certainly, the intervention dose was reduced when intervention activities were severely disrupted during two significant national events: the 80-day mourning period following the death of the King in September 2006 and a subsequent period of major civil unrest that erupted 2 months after the death of the King. Some intervention activities were not implemented at all as a result of the mourning period. Third, policy-based interventions during the MYP were limited, and while some efforts were made to implement school food policies, the enforcement of these was minimal. The fourth, and probably most important reason for the lack of beneficial anthropometric and behavioural outcomes, relates to socio-cultural factors. Tonga is a society with very strong socio-cultural influences that underpin food-related behaviours (34), physical activities and body size perceptions (35). These socio-cultural factors include the strong values of love, nurturing and respect that underpin food selection, distribution, volumes and consumption (10,12), the high status afforded certain obesogenic foods like confectionery in contemporary Tonga (3,13) and the low status of fruit and vegetables, which are not seen as an essential component of a meal (12). In terms of physical activity, there are strong expectations that adolescent girls in Tonga should be engaged in home-based activities (chores and homework) after school and recreational physical activity is discouraged (36). The relative energy expended in chores is likely to be significantly less than in recreational physical activities and sports (37). A greater integration of strategies to address these and other socio-cultural factors into the intervention may have strengthened the ‘dose’ of the overall intervention and led to more beneficial outcomes. While preliminary socio-cultural interviews informed the intervention, the short lead-in time prior to intervention activities limited the full extent to which socio-cultural factors could be conducted and integrated into the intervention (38). It must also be highlighted that the lower levels of weight gain observed in male adolescents compared to the female adolescents during the project indicates the importance that gender plays in values, behaviours and lifestyle. The role of gender was perhaps insufficiently considered in the development of the intervention programmes, and future interventions and research must include greater consideration of the impacts of gender on adolescents.

In hierarchical societies with a collective ethos, such as Tonga, interventions that target social groupings, e.g. families, schools and churches, may be more effective than those that rely on individual behaviour changes. Furthermore, a multi-pronged approach that engages government leaders to implement policies, community leaders to role model and influence practices, and families to influence adolescents' behaviours may be more effective in societies that are hierarchically organized than only targeting a specific age group, e.g. adolescence, that has relatively little status or autonomy. It seems that the community capacity-building approach taken in MYP was probably necessary but not sufficient for slowing the rapid weight gain in this population. Other innovative strategies will need to be tested to address the powerful socio-cultural barriers to healthy weight in the Tongan community.

One of the measures of quality of life showed a smaller increase in the adolescents from Tongatapu (intervention group) compared to the less urbanized, outer island of Vava'u (comparison group). The interpretation of this finding is uncertain and a potential explanation may be that adolescents on the main island are exposed to more pressure in terms of achieving high examination results and obtaining employment or overseas tertiary education places. Further analyses or studies will be needed to explore this issue.

The strengths of MYP were that it was the first major, solutions-based approach to an enormous health problem in Tonga and that it used a whole of community approach targeting the critical period of adolescence. Solid conclusions which advance knowledge can usually be made from such solutions-oriented research (39)– in this case, that the intervention did not have sufficient duration, intensity or integration of socio-cultural factors to bring about the desired changes in reducing overweight and obesity among Tongan adolescents. The major caveat to the strength of this conclusion is that the quasi-experimental approach of having three districts on a more urbanized island for the intervention populations and a more rural island for the comparison population intrinsically runs the risks of false positive conclusion (type 1 error) or false negative conclusion (type 2 error). In this instance, we have made a negative conclusion about the impact of MYP so a type 2 error is possible. For example, the trajectory of weight gain of adolescents on Vava'u may naturally be lower because they lead a more traditional lifestyle compared to those on Tongatapu who watched significantly more television and consumed more canned corned mutton and soft drinks than adolescents on Vava'u and Ha'apai (3,13).Therefore, it is possible that MYP was successful in reducing unhealthy weight gain by moving the trajectory of weight gain in the adolescents on Tongatapu down to the same trajectory as the Vava'u adolescents, thus showing up as no difference in trajectories. However, there is no evidence from the behavioural data that the adolescents on Tongatapu made significant progress towards healthier behaviours and the null result remains the most likely true conclusion. Other designs, such as cluster randomized trials, were considered for MYP but rejected because of other design considerations such as achieving sufficient clusters and the high risk of the interventions contaminating the comparison populations in a small island nation like Tonga. Cluster effects associated with the practicalities of data collection, such as differences in seasonality of measurements (e.g. clashing with school ‘seasons’ for examinations or inter-school sports) or measuring each arm of the study in blocks, can potentially add bias to the measurements of both behaviours and quality of life. In addition, the lower numbers of completed behavioural questionnaires completed at follow-up may have introduced some bias to those results, and are difficult to interpret.

One of the major lessons learned during the MYP project was around the nature of the interventions needed for populations where socio-cultural factors are strong determinants of eating, physical activity and body size perceptions. Qualitative studies on these socio-cultural factors were conducted in Tonga prior to the start of the intervention and these informed the development of the MYP action plan as well as the intervention. Subsequent quantitative and qualitative studies were conducted which helped to inform the interventions. However, the extremely strong impact of socio-cultural factors on behaviours and body size in this population and the inevitable delays in collection and analysis and feedback of the socio-cultural data were underestimated.

The school food policy was the only policy to be implemented in intervention schools between 2006 and 2008 but this policy was neither audited nor enforced. The implementation of more policies relating to food and physical activity environments would have provided high-level support for the MYP intervention. For example, the National Non-Communicable Diseases Committee has highlighted the need for a compulsory physical education policy for schools (40). Policies to encourage students to be physically active at school, either at lunch time or during physical education periods, would have supported the MYP action plan. National-level policies and regulations that promote physical activity would increase opportunities to be active outside school hours, e.g. more control of dogs and more footpaths. More emphasis on the wider food environment is also needed, through the use of policy to influence pricing and accessibility of healthy foods (13).


Ma'alahi Youth Project was the first community-based obesity prevention intervention to be conducted with adolescents in Tonga, a Pacific nation with a youthful population, a high prevalence of obesity, a focus on community and family, and other strong socio-cultural factors underpinning food and physical activity patterns as well as body size perception. The decrease in percentage body fat was a promising shift in the desired direction; however, the lack of change in BMI and BMI-z and most behaviours suggests that the intervention was insufficient to slow the rapid increase in unhealthy weight gain that this population experiences during adolescence and early adulthood. Community-based interventions in countries with a high prevalence of obesity may require a stronger intervention dose than was achieved in the MYP. Alternatively, different approaches may be necessary for populations with high obesity prevalence. The experience of MYP is that a strong emphasis is needed on addressing socio-cultural barriers to healthy behaviours and implementing more policy, as well as regulatory approaches at the national and local levels.

Conflicts of Interest Statement

L. Millar, M. Malakellis, H. Mavoa, M. Moodie, B. A. Swinburn, P. Kremer and M. P. McCabe's institutions have received grants from National Health and Medical Research Council. Support was provided to cover the costs of travel to New Zealand and to Investigator meetings. The authors were employed by Deakin University.

J. T. Schultz's institution has received grants, and support to cover the costs of travel to Investigator meetings in the Pacific, from Wellcome Trust Grant. The author was employed by Fiji School of Medicine.

W. Snowdon's institution has received grants from Wellcome Trust to fund the entire OPIC project. The author also received consulting fee or honorarium, for the concept paper on food policy options for the region, and support for travel to attend meetings to discuss actions to control NCDs in the region, both funded by World Health Organization.

J. Utter's institution has received a grant from the Health Research Council of New Zealand.

K. F. Fotu, P. Vivili and G. Roberts declared no conflict of interests.


The authors would like to thank the many people involved in the Pacific OPIC project including co-investigators, other staff and postgraduate students, partner organizations, and especially the schools, students, parents and communities. The funding for the project was from the Wellcome Trust (UK), the National Health and Medical Research Council (Australia) and the Health Research Council (New Zealand) through their innovative International Collaborative Research Grant Scheme. We also acknowledge the support of the following: FAO, SPC, AusAID, NZAID and other local donors; Tongan Ministries of Health, Education; Agriculture (DSAP); Training, Employment, Youth and Sports; and Labour and Commerce. Also the Tonga National Youth Congress, Tonga Secondary Schools, districts and villages involved, the participants in the studies, staff of the Ma'alahi Youth Project (Tilema Cama, ‘Inoke Taufa, Litiola Kava, Sifa Pomana, Hekisou Fifita, Laukau ‘Aholova, Sisilia Moala, Siaosi Polovili, Moana Kioa) and the OPIC team at Deakin University (especially Katherine Scalzo) and the University of Auckland.