Article Text

The effectiveness of nutrition interventions on dietary outcomes by relative social disadvantage: a systematic review
1. J Oldroyd1,
2. C Burns2,
3. P Lucas3,
4. A Haikerwal4,
5. E Waters5
1. 1
Cabrini Institute Department of Epidemiology and Preventative Medicine, Monash University, Malvern, Victoria, Australia
2. 2
School of Exercise and Nutrition Sciences, Deakin University, Burwood, Victoria, Australia
3. 3
School for Policy Studies, University of Bristol, Bristol, UK
4. 4
Research and Public Health Unit, Centre for Community Child Health, University of Melbourne, Royal Children’s Hospital, Melbourne, Victoria, Australia
5. 5
The McCaughey Centre, School of Population Health, University of Melbourne, Carlton, Victoria, Australia
1. Dr J Oldroyd, Cabrini Institute, Department of Epidemiology and Preventative Medicine, Monash University, 183 Wattletree Rd, Malvern, Victoria, 3144, Australia; john_oldroyd{at}yahoo.com.au

Abstract

Objective: To determine whether nutrition interventions widen dietary inequalities across socioeconomic status groups.

Design: Systematic review of interventions that aim to promote healthy eating.

Data sources: CINAHL and MEDLINE were searched between 1990 and 2007.

Review methods: Studies were included if they were randomised controlled trials or concurrent controlled trials of interventions to promote healthy eating delivered at a group level to low socioeconomic status groups or studies where it was possible to disaggregate data by socioeconomic status.

Results: Six studies met the inclusion criteria. Four were set in educational setting (three elementary schools, one vocational training). The first found greater increases in fruit and vegetable consumption in children from high-income families after 1 year (mean difference 2.4 portions per day, p<0.0001) than in children in low-income families (mean difference 1.3 portions per day, p<0.0003). The second did not report effect sizes but reported the nutrition intervention to be less effective in disadvantaged areas (p<0.01). The third found that 24-h fruit juice and vegetable consumption increased more in children born outside the Netherlands (“non-native”) after a nutrition intervention (beta coefficient  = 1.30, p<0.01) than in “native” children (beta coefficient  = 0.24, p<0.05). The vocational training study found that the group with better educated participants achieved 34% of dietary goals compared with the group who had more non-US born and non-English speakers, which achieved 60% of dietary goals. Two studies were conducted in primary care settings. The first found that, as a result of the intervention, the difference in consumption of added fat between the intervention and the control group was −8.9 g/day for blacks and –12.0 g/day for whites (p<0.05). In the second study, there was greater attrition among the ethnic minority participants than among the white participants (p<0.04).

Conclusions: Nutrition interventions have differential effects by socioeconomic status, although in this review we found only limited evidence that nutrition interventions widen dietary inequalities. Due to small numbers of included studies, the possibility that nutrition interventions widen inequalities cannot be excluded. This needs to be considered when formulating public health policy.

Statistics from Altmetric.com

There is growing evidence that inequalities in population health are increasing.1 Even in countries where overall health outcomes are improving, health inequalities persist across socioeconomic position, geographical area and by ethnicity, age and gender.2 Importantly, health inequalities persist intergenerationally, affecting the health outcomes of future generations.3

Policy directed to reducing these inequalities needs to address the evidence that health promotion interventions may be widening health inequalities. For example, there is evidence from health promotion studies that tobacco control interventions, which aim to reduce inequalities, paradoxically further widen health inequalities.4 5 With respect to nutrition, there is evidence that people of lower socioeconomic status (SES) tend to eat less fruit and vegetables and fewer foods rich in dietary fibre compared with people in a higher socioeconomic position.6 In addition, there is a gradient in diet-related diseases by socioeconomic disadvantage.7 However, whether nutrition interventions increase inequalities has not been studied in detail. The current review aimed to determine whether nutrition interventions widen inequalities by affecting dietary outcomes differentially with respect to SES.

METHODS

We used the recommendations outlined in the Cochrane Collaboration Handbook8 and the Health Promotion and Public Health Field Guidelines.9

Search strategy

CINAHL and MEDLINE were searched for studies published between January 1990 and October 2007.

The keywords included: (diet or nutrition or meal or food) and (intervention and health) (Appendix 1).

The reviewers hand searched published conference and symposia proceedings, existing review articles and bibliographies of identified trials. The reviewers also contacted experts in the field to enquire of further published or unpublished studies.

Selection of studies

The titles and abstracts of potentially relevant publications identified were reviewed independently by two reviewers (AH and EW) for eligibility using the following criteria: nutrition interventions (defined as any strategy from screening to health policy aimed at increasing the consumption of a healthy diet)10 delivered to healthy populations at a group level; randomised controlled trials (RCT) or concurrent controlled trials (CCT); and interventions delivered to low socioeconomic groups or studies where it was possible to disaggregate data by SES. Low socioeconomic groups were defined as ethnic minorities and those of low income or educational level. Animal studies or those published in languages other than English were excluded. Full copies of all those deemed eligible by one of the reviewers were retrieved for closer examination. Reasons for the non-inclusion of the studies were noted.

Data were extracted independently by two reviewers (AH and EW). Information on both the intervention and the control groups was described: country and study site, methodology including modes of delivery of the interventions, types of interventions, intervention settings, duration of intervention and follow-up, study design, recruitment, exclusion/inclusion criteria of the participants, sample size; participants’ demographics including age, gender, ethnicity and SES; process indicators (the extent to which the intervention was delivered as intended), outcome indicators, potential bias during selection, intervention and follow-up, loss to follow-up; cost data, analysis used, quality assessment, size and direction of effects reported.

The reviewers (AH and EW) independently rated the quality of each study according to the study design, sample size and power, comparability of intervention and control group, rates of attrition, reliability of method assessing outcome, blinding of outcome assessment and treatment of potential bias and potential confounding factors.

Data analysis

Owing to the heterogeneity of study characteristics, it was decided that it would not be appropriate to carry out a meta-analysis. Instead, a narrative synthesis was undertaken, accounting for target population, type and reported impact of intervention and the effect size.

Reported outcomes

All the reported outcomes were taken directly from the studies. Outcomes included frequency and portions of food consumed, fruit and vegetable consumption, fat intake, fat-related dietary habits as well as dietary knowledge, behaviours and preferences for healthy foods. Where possible, data were presented as mean change with standard deviations and 95% confidence intervals or p values. The direction of the effects was also noted. Results with p<0.05 were considered significant.

RESULTS

The electronic database searches identified 3344 references. A flow diagram of included studies is shown in figure 1. Four RCTs and two CCTs met the inclusion criteria and were included in the review (table 1).

Figure 1 Flow chart of study selection
Table 1 Description of the six studies

Methodological quality

As shown in table 2, all the six studies had some methodological weakness, using standard measurement criteria.9 None of the six studies provided power calculations. All studies collected the outcome data using self-reported food intake over the past 24 hours (eg food frequency questionnaire). Of the six studies, the studies by Calfas et al15 and Winkelby et al14 were assessed to have the lowest risk of bias (table 1).

Table 2 Quality criteria that were used to assess the six studies

Context

Description of the studies

The six included studies were: Reynolds et al 2000, Friel et al 1999, Reinaerts et al 2007, Winkelby et al 1999, Calfas et al 2002 and Kristal et al 1999. As shown in table 1, four of the six studies were conducted in the USA, one was conducted in Ireland and another in the Netherlands. Study participants ranged in age from 4 to 79 years. All six studies included ethnically diverse participants (eg European-Americans, African-Americans, Caucasians, Hispanics, Asians or Irish and Dutch).

Three studies were conducted in an elementary school-based setting, three in primary care and one in an adult vocational training setting.

Differential effects by socioeconomic status

Elementary school-based setting

Reynolds et al11 conducted a randomised controlled trial to evaluate the effectiveness of a school-based nutrition intervention. Here, 28 schools with mixed populations were randomised (1698 fourth grade children, mean age 8.7 years) to either intervention or control conditions. Some 83% were from European-American backgrounds and 16% were African-Americans. The intervention aimed to increase fruit and vegetable consumption. Intervention students consumed more fruit and vegetables after 1 year (mean, 95% CI: 3.96 servings (3.51 to 4.44)) than control subjects (2.28 (1.92 to 2.66), p<0.0001) and after 2 years follow-up intervention (3.20 (2.89 to 3.52) versus control 2.21 (1.94 to 2.49), p<0.0001). The pre-test knowledge score for cafeteria workers rose from 3.6 (maximum score of 5) to 4.6 at post-test. There was low family involvement (participation level 24%), decreasing the intervention delivered to the families.

Stratified analysis found greater increases in fruit and vegetable consumption in higher income (mean difference after 1 year: 2.4 portions per day, p<0.0001) than in lower income children (mean difference: 1.3 portions per day, p<0.0003).11 The time spent per lesson was 42.6 minutes for African-American children, 36.3 minutes for European-American children (p<0.0001), and for participants from low SES was 40.3 minutes and from high SES was 35.2 minutes (p<0.0001). In addition, there was a lower attendance rate in schools with a larger enrolment of African-American children from lower SES (attendance rate 94% and homework return rates 41%) than in schools with European-American children from higher SES (attendance rate 95% and homework return rates of 48%, p<0.001).

Friel et al12 conducted a comparative quasi-experimental study to evaluate a nutrition and physical activity education programme consisting of 20 sessions over 10 weeks delivered to Irish children aged 8–10 years. A total of 821 children were recruited from eight schools mainly from economically disadvantaged areas. The intervention consisted of worksheets and homework assignments and aerobic training. Outcomes included knowledge, behaviour and preferences for health foods assessed by questionnaire. Overall, chip consumption (⩾four times in the last 5 days) increased by 5% in the intervention group and by 10% in the control group pre to post intervention. High sugar snack consumption (one per day) decreased in the intervention group by 7% but increased in controls by 10% (pre to post intervention). These findings were not significant. The authors state that the NEAPS programme was less effective in pupils in disadvantaged areas (p = 0.018); however, they do not provide more details.

Reinaerts et al13 conducted a concurrent controlled trial to evaluate (1) a free fruit and vegetable distribution intervention or (2) a multicomponent school-based programme that consisted of classroom activities and parental involvement, delivered to children aged 4–12 years. A total of 929 parents completed dietary questionnaire data on behalf of their children across 12 schools (six intervention and six control). Outcomes were fruit consumption (portions per day), vegetable consumption (grams per day), vegetable snacks consumed (times per day) or 24-h fruit juice or vegetables consumed (times per day). Vegetable consumption increased more in children whose parents were born outside the Netherlands (“non-natives”) (beta coefficient  = 1.10, p<0.01) compared with “native” children (beta coefficient  = 0.23, not significant). 24-h fruit juice and vegetable consumption also increased more in “non-native” children (beta coefficient  = 1.30, p<0.01) compared with “native” children (beta coefficient  = 0.24, p<0.05). Parents of “non-native” children dropped out more often than parents of “native” children (52% versus 42%, p<0.001).

Vocational training setting

Winkelby et al14 conducted a cluster randomised trial to compare two curriculum interventions over 6 months of follow-up. The study randomised 24 classes matched in pairs on type of class (eg vocational training, basic skills training), and one of each pair was randomly assigned to receive either (1) a Stanford Nutrition Action Program (SNAP) intervention (using a food pyramid model) or (2) the general nutrition (GN) intervention. Both interventions consisted of six 60-minute classes, taught once a week. The SNAP curriculum also provided six maintenance telephone contacts during the 12-week period following the intervention. They found that 23% of participants who had a high baseline dietary fat intake successfully reduced their dietary fat to <30% of daily calories after the intervention, irrespective of which intervention they received. Only 34% of participants in group 2 (GN) successfully reduced their dietary fat. These participants had higher education levels than other groups. Some 60% of participants in group 1 (SNAP) were successful in reducing dietary fat. These participants were more likely not to speak English at home and to have been born outside the US.

Primary care settings

Calfas et al15 conducted a randomised controlled trial to evaluate a multicomponent programme for nutrition and physical activity change. In total, 173 adult participants were recruited from four primary health care centres with ethnically diverse and varied socioeconomic backgrounds. They were randomised to a control group or one of three interventions: (1) mail follow-up only (every 2 weeks ×8); (2) infrequent telephone (three 10-minute counseling calls) and mail follow-up (every 6 weeks ×2); (3) frequent telephone contact (weekly for 8 weeks) and mail follow-up (every 2 weeks ×8). There were no significant differences between intervention groups; however, participants who set dietary goals improved (pre–post intervention) significantly more than those who did not set goals for dietary fat consumption (−1.05 versus −0.07 servings per day, p<0.002), fruit and vegetable consumption (+1.66 versus +0.34 servings per day, p<0.001) and reduced overeating habits (−4.06 versus +3.42 habits per day, p<0.002).

There was greater attrition among the ethnic minority participants than among Caucasian participants (p<0.04), which the authors attributed to language barriers and lack of experience with computers.

Kristal et al16 undertook a randomised controlled trial to examine the effects of a nutrition education intervention programme to reduce dietary fat intake in blacks (not Hispanics), Hispanics (Hispanic black or Hispanic white) and white women (white or Caucasians not Hispanics). The study recruited 1702 post-menopausal women from three clinical centres in Atlanta, Birmingham and Miami. The intervention was delivered by registered dietitians and aimed to reduce dietary fat intake to 20% or less of total energy, increase servings of fruit and vegetables and reduce saturated fat intake. It is not stated how the control group was managed.

The intervention resulted in a larger decrease in added fats (eg butter and salad dressing) for whites (mean decrease −12 g/day) than for blacks (mean decrease −8.9 g/day, p<0.05). Conversely, the intervention effect for adopting low-fat meat purchasing and preparation methods was larger for blacks (−0.50 g/day) than for whites (−0.39 g/day, not significant). The intervention effect for replacing high-fat foods with fruit and vegetables was larger for Hispanics (−0. 43 g/day) than for whites (−0.30 g/day, not significant).

DISCUSSION

We searched for evidence that nutrition interventions widen dietary inequalities by SES. We found six studies that met the inclusion criteria. The studies showed that nutrition interventions have differential effects by SES, but they provided only limited evidence for widening of inequalities. For example, one relatively unbiased study found greater increases in fruit and vegetable consumption in school children from high-income families after 1 year (mean difference between baseline and follow-up 2.4 portions per day, p<0.0001) than in children in low-income families (mean difference 1.3 portions per day, p<0.0003) as a result of a nutrition intervention.11 A second study found that the difference in consumption of added fat between the intervention and control groups was greater in white than in black participants (−8.9 g/day for blacks and –12.0 g/day for whites, p<0.05).16 Kristjansson17 defines an effective intervention for reducing health inequalities as one that is more effective in people of low SES or at least equally effective across groups (ie does not widen inequalities). In most of the studies we found, the effects were smaller in low SES groups (eg increases in fruit and vegetable consumption in low-income children) suggesting a widening of inequalities. However, there were still some benefits in the low SES groups. We cannot exclude the possibility that different interventions (eg in context, intensity or uptake) would not result in greater benefits in lower SES groups. Owing to the small numbers of available studies and the beneficial effects observed in some disadvantaged groups, we think that this constitutes limited evidence that nutrition interventions widen dietary inequalities.

As noted, only six studies included a statistical measure summarising the magnitude of health differences between individuals in different SES groups. This was despite a broad search spanning 17 years. Many studies did not report SES or contained mixed SES groups in which the results could not be disaggregated by SES grouping. This highlights the fact that research that specifically looks at the impact of interventions on inequalities is a relatively new area. For example, Cochrane reviews rarely present information of outcomes stratified by SES, mainly because the primary studies have also not reported SES subgroup effects. Given the emerging evidence of the social gradients in health, there is a moral imperative to analyse differential effects of interventions by SES groups.

We had some difficulty defining low SES. There are many factors that may contribute to disadvantage including place of residence, ethnicity, occupation, income, gender, religion, education and social capital. We agreed on ethnic minority, low income and low education level as pertinent indicators of disadvantage for nutritional interventions. However, other indicators may have been informative and increased the number of included studies.

This review, in which only a small number of studies met the inclusion criteria, underlines the need for more investigations.18 19 Future research must involve large numbers of participants from disadvantaged backgrounds. For example, it is not clear from the included studies in this review how successful the recruitment of people from disadvantaged groups was. Larger studies will be able to control for differential effects due to differences in the recruitment of low SES groups. Interventions also need to be of longer duration and intensity. Based on our sense of the literature and on the requirements of a rigorous review process, the intervention elements that we believe are important to include are information needed to assess generalisability (eg the recruitment method), characteristics of intervention intensity (eg delivery methods, duration of intervention periods, uptake of the interventions), training of the individuals involved with intervention delivery, specific theoretical model used and how they are applied to the intervention and the ongoing maintenance of the intervention. We included two non-randomised studies in this review, although many studies were excluded because they had no concurrent comparison group. This was done to include only the highest quality evidence and maximise the confidence that we had in our conclusions. However, future systematic reviews with respect to equity will need to include qualitative and non-experimental data to capture the effects of interventions on all dimensions of health inequalities.19

A limitation of this review is that, although the included studies were randomised controlled trials or concurrent controlled trials, they differed markedly in setting, target population and outcome measures, making it difficult to compare the results across studies. Another limitation of the included studies was the reliability of dietary outcome measurements (eg self-reported consumption of fruit and vegetables), which was questionable. All the six studies collected the outcome data using self-reported food intake over the past 24 hours (eg food frequency questionnaire). Dietary measurement recommendations suggest that dietary data should ideally be collected over 3 days, including one weekend day to provide the best estimate of the food consumed.20 Additionally, our review did not measure whether the changes in food and nutrient consumption that we identified were of clinical, not just statistical, significance.

CONCLUSION

Nutrition interventions have differential effects by SES although in this review we found limited evidence that nutrition interventions widen dietary inequalities. This may be an absence of evidence rather than evidence of an absence of an effect. Owing to the small number of studies in this review, the possibility that nutrition interventions widen inequalities cannot be excluded. This needs to be considered when formulating public health policy and planning future nutrition interventions.

What is already known on this subject

• People in lower socioeconomic groups have poorer quality diets than people of higher socioeconomic position.

• Health promotion interventions (eg tobacco control) have been shown to widen health inequalities even while improving health for some.

• Nutrition interventions have differential effects by SES.

• Nutrition interventions can be effective in disadvantaged groups.

• The evidence that we found about nutrition interventions widening inequalities was limited (small effects in very few randomised controlled trials).

• Despite the absence of strong evidence that nutrition interventions increase dietary inequalities, caution should be used when formulating nutrition-related public health policy.

Appendix 1

Search strategy

CINAHL 1990–2007

1. ((diet$or nutrition$) adj3 (intervention$or program$ or education$)).tw. 2. (health$ adj3 (diet$or eat$ or meal$or food$)).tw.

3. randomi$.mp. [mp = title, cinahl subject headings, abstract, instrumentation] 4. clin$.mp. [mp = title, cinahl subject headings, abstract, instrumentation]

5. trial$.mp. [mp = title, cinahl subject headings, abstract, instrumentation] 6. (clin$ adj3 trial$).mp. [mp = title, cinahl subject headings, abstract, instrumentation] 7. singl$.mp. [mp = title, cinahl subject headings, abstract, instrumentation]

8. doubl$.mp. [mp = title, cinahl subject headings, abstract, instrumentation] 9. tripl$.mp. [mp = title, cinahl subject headings, abstract, instrumentation]

10. trebl$.mp. [mp = title, cinahl subject headings, abstract, instrumentation] 11. mask$.mp. [mp = title, cinahl subject headings, abstract, instrumentation]

12. blind$.mp. [mp = title, cinahl subject headings, abstract, instrumentation] 13. (7 or 8 or 9 or 10) and (11 or 12) 14. crossover.mp. [mp = title, cinahl subject headings, abstract, instrumentation] 15. random$.mp. [mp = title, cinahl subject headings, abstract, instrumentation] 1

16. allocate$.mp. [mp = title, cinahl subject headings, abstract, instrumentation] 17. assign$.mp. [mp = title, cinahl subject headings, abstract, instrumentation]

18. (random$adj3 (allocate$ or assign$)).mp. 19. Random Assignment 20. Clinical Trials 21. Meta Analysis 22. 18 or 14 or 6 or 3 or 19 or 20 or 21 23. 1 or 2 24. 22 and 23 Medline 1990–2007 1. ((diet$ or nutrition$) adj3 (intervention$ or program$or education$)).ti.

2. (health$) adj3 (diet$ or eat$or meal$ or food$)).ti. 3. randomized controlled trial.pt. 4. controlled clinical trial.pt. 5. randomized controlled trials.sh. 6. random allocation.sh. 7. double blind method.sh. 8. single-blind method.sh. 9. 3 or 4 or 5 or 6 or 7 or 8 10. clinical trial.pt. 11. clinical trials 12. (clin$ adj25 trial$).ti,ab. 13. ((singl$ or doubl$or trebl$ or tripl$) adj25 (blind$ or mask$)).ti,ab. 14. Placebos.sh. 15. placebo$.ti,ab.

16. random$.ti,ab. 17. research design.sh. 18. 12 or 13 or 14 or 15 or 16 or 17 19. comparative study.sh. 20. evaluation studies/ 21. follow up studies.sh. 22. prospective studies.sh. 23. (control$ or prospectiv$or volunteer$).ti,ab.

24. 19 or 20 or 21 or 22

25. 1 or 2

26. (25 and 9) and (25 and 18) and (25 and 24)

View Abstract

Footnotes

• Contributors: JO led in revising drafts of the paper, extracted data and critically reviewed it before submission. CB extracted data and contributed to writing the paper. PL participated in revising drafts of the paper and critically reviewed it before submission. AH participated in the design of the search strategy, conducted early literature searches, extracted data and wrote a first draft of the paper. EW identified the research question, participated in the design of the search strategy, extracted data and contributed to writing the paper.

• Funding: None.

• Competing interests: None.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.