Health promotion interventions are an effective way of disease prevention and improving overall health1,2. Well-established approaches focus amongst others on healthy eating, exercising, avoiding excessive use of alcohol, quitting smoking, and sexual health behaviors, such as condom use3,4. Promoting these health behaviors on a large scale by using health promotion interventions might be a promising mean for reducing the burden of disease5,6,7. However, these intervention offers do not always reach or fit populations equally between and within countries8.

The reach of these interventions could be expanded globally via the internet9 using the Internet- and mobile-based interventions (IMI)10,11,12,13,14. IMI offers time and place flexibility, potential cost-effectiveness, and scalability without necessarily losing effectiveness15. Thereby, IMI could reach a diverse group of people including minorities and people of a cultural background that is not yet well covered by the established health care systems16,17,18,19. However, IMI also come with some substantial limitations, most prominently risk of low adherence and uptake20,21. In order to tackle these issues, population-related factors such as the needs and expectations of the users should be taken into account15. Tailoring the intervention content and delivery method to the target group’s culture can thus be a means for increasing engagement and effectiveness22.

Researchers in the field of health promotion are encouraged by WHO to approach their practice and research considering diverse human experience and intersectionality of various factors, including culture, gender, immigration status, and ethnicity, which may be related to lower physical and mental health23,24,25. In order to offer an acceptable and relevant health service to people of a certain cultural background, health promotion interventions could be developed from scratch with cultural sensitivity or a less resource-consuming way can, for example, be tailoring an already existing intervention for specific cultural groups26. This process is defined as cultural adaptation27. Cultural adaptation could be adopted on surface structure modifications (pairing materials and messages to apparent features of the target population such as language) or deep structure modifications (concerning intersecting effects of social, cultural, and historical variables on the target behavior)28,29. Culturally adapted face-to-face interventions are shown to be effective in smoking cessation30,31, health education, and healthy eating32. Moreover, IMI developed for ethnic minorities and underserved populations are also shown to be accepted and effective in the promotion of various health behaviors such as physical activity33 and healthy eating16.

Previous reviews included IMI for ethnic minority and historically underserved populations in developed countries22,34,35,36. However, none of them specifically examined culturally adapted IMI for health promotion. The number of studies exploring the development and dissemination of culturally adapted health promotion IMI is increasing and thus a systematic review and meta-analysis seems timely. Hence, this review aims to systematically identify culturally adapted IMI on health promotion and explore their effectiveness among populations that is different from the original intervention’s target group.


All the predefined characteristics of included articles are presented in Table 1.

Table 1 Characteristics of included articles.

Study selection

We identified a total of 20,012 records. After screening titles and abstracts and full-texts, 13 studies were included in the quantitative analyses. The main characteristics of the studies are outlined in Table 1. During the full-text screening, 38 studies were excluded due to lack of cultural adaptation, 29 studies were concerning culturally sensitive interventions, 13 studies did not report relevant outcome data, eight studies were no original articles, four studies were not RCTs, five were excluded due to other reasons (i.e. novel interventions, study protocols, non-English full-text) and one due to a differing health promotion topic.

Study characteristics

The 13 RCTs included in this review comprise a total of N = 10,747 randomized participants, mostly adult populations (N = 9710). Trials were conducted between 2012 and 202037,38,39,40,41,42,43,44,45,46,47,48,49. The mean age of the participants varied from 14 to 57 (see Table 1). Primary studies focused on smokers (k = 3, N = 2546), individuals diagnosed with Type 2 diabetes (k = 2, N = 171), and HIV + individuals (k = 1, N = 433).

Four studies dealt with smoking cessation37,39,41,49, two with both healthy eating and physical activity38,42, three with physical activity only45,46,47, two with healthy eating only40,43, one with sexual health behavior44, and one with alcohol consumption48.

Four studies provided follow-up data on short-term effectiveness (one to five months follow-up)39,41,42,46, three studies on long-term effectiveness (six to 12-months follow-up)40,45,47, six studies provided follow-up data on both assessment points.

Seven studies were conducted in the USA38,39,43,44,45,47,48, two in China37,42, two in Brazil40,41, one in Hong Kong46, and one in Australia49.

The cultural adaptation was based on a theory or a guideline in three studies38,39,40. Eight studies based their cultural adaptation on a formative research/pilot study or expert review37,41,42,44,45,47,48,49. Two studies did not provide information regarding the basis of cultural adaptation43,46. In terms of alterations of the intervention content, four studies incorporated both surface and deep structure changes38,40,43,45, while nine studies used surface structure changes only28. Details of the culturally adapted and original interventions are presented in Table 2.

Table 2 Summary of culturally adapted and original IMI.

Risk of bias of in included studies

The risk of bias assessment of the included studies is presented in Figs. 1, 2. The interrater reliability suggested substantial agreement between the raters, κ = 0.79. Five studies were assessed to have a low risk of bias, five studies had some risk of bias and three studies were rated to have a high risk of bias. Four studies were assessed to have some risk of bias due to deviations from the intended intervention40,43,46,48, three studies were assessed to have some risk of bias in the measurement of the outcome domain41,44,46. Two studies37,38 were assessed to have some risk of bias due to missing outcome data, one study43 had a high risk of bias in this domain. One study was assessed to have some risk of bias arising from the randomization process40. This study had an unequal number of clusters, which resulted in significant baseline differences in primary outcomes.

Fig. 1: Risk of bias summary.
figure 1

Reviewers’ judgments about each risk of bias item for each included study.

Fig. 2: Risk of bias graph.
figure 2

Reviewers' judgments about each risk of bias item presented as percentages across all included studies.


Nine studies that provided data to calculate standardized mean difference effect sizes were pooled. Individual data points are presented in Supplementary Table 1. Four studies concerning smoking cessation did only provide dichotomous outcomes. We conducted separate meta-analyses for the different areas of health promotion (physical activity, smoking cessation), in the case of at least three studies reported on the same outcome. Since we included only nine studies in the analysis, we refrained from exploring the publication bias via a funnel plot50. Following, we report on pooled effectiveness across health promotion domains.

The meta-analysis of six studies38,43,44,45,47,48 examining any long-term health promotion interventions revealed that culturally adapted IMI of health promotion were not superior to active control conditions in the long-term. In addition, four studies38,43,44,48 that provided short-term follow-up data were not superior to active control conditions in improving health behavior outcomes. Due to substantial heterogeneity, the results are reported descriptively, see Figs. 3, 4.

Fig. 3: Summary of culturally adapted IMI of health promotion vs. active controls in the long-term.
figure 3

Due to substantial heterogeneity among the culturally adapted IMI of health promotion vs. active controls in long-term meta-analytical pooling did not perform.

Fig. 4: Summary of culturally adapted IMI of health promotion vs. active controls in the short-term.
figure 4

A summary plot of effect sizes of four studies of culturally adapted IMI of health promotion vs. active controls in short-term are presented.

Comparisons with passive control groups could not be pooled given the low number of studies (k = 4), see Fig. 5. One study39 reported only Odds ratios ([OR], 1.13; 95% CI 0.18 to 7.04).

Fig. 5: Summary of culturally adapted IMI of health promotion vs. passive controls.
figure 5

Due to few numbers of studies (two studies reported data in the long-term, two in the short-term, while one study reported dichotomous outcome) comparing culturally adapted IMI to a passive control group, meta-analytic pooling did not perform.

Due to a small number of studies (k = 9), we did not perform the predefined subgroup and sensitivity analyses.

Effectiveness of culturally adapted IMI of physical activity

Five studies reported physical activity outcomes, four of which provided data via accelerometers/pedometers38,45,47 and one via a self-reported questionnaire46. Pooling the three studies with active control conditions resulted in a small significant long-term effect favoring culturally adapted IMI (N = 296; g = 0.48; 95% CI 0.25 to 0.71; I2 = 41%; fixed effect), see Fig. 6.

Fig. 6: Fixed effects meta-analysis of culturally adapted IMI for physical activity vs. active control conditions.
figure 6

Forest plot presenting fixed effects meta-analysis of culturally adapted IMI for physical activity vs. active controls.

Effectiveness of culturally adapted IMI of smoking cessation

Three RCTs (N = 8,112) reported smoking cessation outcomes measuring short-term abstinence at the end of the intervention versus active controls. The meta-analysis findings of these studies were not significant (Odds Ratio [OR], 1.75; 95% CI 0.51 to 6.05, I2 = 56%), see Fig. 7. The number of included studies was small (k = 3) and the effect size was mainly based on one large-scale study with N = 8000 participants37.

Fig. 7: Summary of culturally adapted IMI for smoking cessation vs. active controls in short-term.
figure 7

Three studies reported smoking cessation outcomes measuring short-term abstinence at the end of the intervention vs. active controls are presented on the forest plot.

Effectiveness of culturally adapted IMI of healthy eating, sexual health behavior, and alcohol consumption

Fewer than three studies reported on sexual health behavior44, alcohol consumption48, and healthy eating40,43. Only one generic outcome, health-related quality of life, was reported in one trial42. These studies were not eligible for pooling due to the number of trials in the respective outcome but reported in Figs. 35.


To our knowledge, this is the first systematic review and meta-analysis that assessed the effectiveness of culturally adapted IMI on health promotion among populations which the intervention was originally not developed for. Our results suggested that culturally adapted IMI are not more effective in promoting health behaviors than active conditions in short- (g = 0.10) and long-term (g = 0.20), respectively. When regarding health promotion topics separately, health promotion of physical activity resulted in a small to moderate effect favoring culturally adapted IMI over active control conditions in the long-term (g = 0.48). This is in line with a previous umbrella review of health promotion IMI for minority and historically underserved populations, which, however, not exclusively included culturally adapted IMI22. Similarly, another meta-analysis of IMI concerning physical activity did highlight the superiority of IMI over a control group or no-treatment condition, without a specific focus on culturally adapted interventions14. No other significant effect was revealed for the other addressed health promotion topics.

Subgroup analyses aiming to detangle the substantial between trial heterogeneity were not feasible. Heterogeneity among the included studies was moderate to substantial, I2 ranged from 0.36 to 0.66. Prior research points to differential effects on culturally adapted health promotion interventions in terms of different populations (age and ethnicity51), different intervention features (professional vs. non-professional provider)32, intervention duration and follow-up times in culturally adapted face-to-face interventions52, interventions focusing on general population groups53, different methodological decisions (RCT methodology use, different control groups, e.g. tailored website, no-treatment controls53) and cultural adaptation contents (inclusion of social support and/or family members51, integrating cultural beliefs and values54). There might be further possible explanations for between-study heterogeneity, and future research needs to provide a better understanding of the impact such factors have on the effectiveness of culturally adapted IMI for health promotion. The present findings are inconsistent with previous meta-analyses on IMI in several ways. Among western populations, these meta-analyses yielded positive effects favoring IMI compared to a waitlist and/or active controls (e.g. other internet-based or face-to-face interventions): for smoking cessation and abstinence outcome10,55, which is maintained at 12-month follow-up56 and higher effects achieved with the use of tailored messages10,13; healthy eating57,58,59 and weight loss60; sexual health behavior promotion61 and regarding HIV prevention and condom use62. Another meta-analysis found significant positive effects of tailored (based on personal relevance) web-based interventions on health behaviors compared to non-tailored web-based interventions53 and a different meta-analysis of SMS-based interventions on various health behavior outcomes suggested that targeted and tailored (based on demographic and psychosocial factors) SMS yielded larger effect sizes, especially for physical activity interventions (g = .51), which yielded a similar effect size to our results (g = 0.48)10. Most surprisingly in this context is our null-finding regarding the effectiveness of culturally adapted IMI compared to waitlist control conditions, which are known to provide a rather upper benchmark of the benefit of interventions63, usually associated with significantly larger between-group effect sizes in IMI for health promotion as well10,13,64. Although some of the meta-analyses mentioned above concern tailored intervention contents, none of the above-mentioned meta-analyses were specifically examining culturally adapted IMI. Therefore, our results cannot be easily compared with prior meta-analyses. However, if culturally adapted IMI for health promotion are not effective at all, even when compared to waitlist controls, we might need to challenge the idea of providing culturally adapted IMI to populations for which the intervention was originally not developed for at large and examine whether IMI developed with cultural sensitivity are effective in the same target groups. Hence, explanations for this surprisingly limited effectiveness seem warranted.

One possible explanation of our results might be related to the quality of cultural adaptation of the interventions. The cultural adaptation processes were rarely well defined in the included studies. Therefore, it was not clear whether aspects of cultural adaptation were appropriate. In addition, the high dropout among included studies could be an indicator of cultural adaptation not working as intended by the researcher. Moreover, only three studies based their cultural adaptation process on a theory, which might contribute to its quality. A tested theory of cultural adaptation of IMI is missing. However, there are guidelines developed for culturally adapting face-to-face interventions28,65 and researchers could implement these guidelines when adapting an IMI66. A recent taxonomy of cultural adaptation of IMI for mental disorders serves as a basis for future cultural adaptations of IMI67. Adopting a theoretical basis in intervention development is suggested to result in higher effects, as was shown in a meta-analysis14. However, this could not be shown in our results: only one out of the three IMI that utilized a theory of cultural adaptation resulted in an improvement in physical activity outcome38. In the future, culturally adapted intervention studies should consider supporting cultural adaption with an established theory and report the adaptation process in more detail to lead prospective cultural adaptations and replications. In the process of cultural adaptation, some of the included studies sought expert reviews, focus group feedback, and conducted a pilot study, and at least altered one aspect of the intervention. However, the majority of the changes were regarded as taking place only at the surface structure28, which might be one reason for the limited impact of culturally adapted IMI shown in the present review. Surface structure changes aim at improving feasibility while deep structure changes target program’s effect for the participants28. Implementing deep structure changes involves core cultural values of a certain population, such as beliefs towards a health issue, or performing a behavior as a member of gender identity. A meta-analysis of face-to-face culturally adapted health interventions showed that incorporating family members and religious values, which are considered as deep structure changes, in the intervention was related to improvements31. Future research should consider exploring surface versus deep structure alterations on the effectiveness of IMI e.g. in the framework of dismantling and additive clinical trial designs aiming at detangling active components and mechanisms of change of the respective interventions68,69,70,71.

Moreover, culturally adapting interventions is not free from criticism. It should be taken into account that the majority of the interventions developed in the fields of psychology and behavioral medicine are for a rather homogeneous group (white, educated, middle to high socioeconomic status) but not representative for the majority19,72. In the cultural adaptation process, the same intervention is often altered to cater to the needs of a different group of people that are non-white, occasionally less educated, and/or bilingual. This process could be seen as a form of assimilation for the target group because even the topic of the intervention might also be representative of western, white, educated humans1. Therefore, in order to avoid these issues, the first step of cultural adaption might include approaching the problem and defining it with the cultural sensitivity of its target group73. In addition, people of a certain cultural background are not homogenous within themselves, each member’s experience is affected by intersecting factors74. Therefore, it might be more complicated than often expressed to adapt an intervention for a cultural group75. One possible solution might be to invest in adapting interventions to cultural specifics of the users, e.g. based on user needs assessments76 or community leaders’ input77. Another solution, especially for migrant/immigrant populations could be developing interventions based on the target groups’ acculturation levels, i.e. a process on a spectrum of either orientation to the host culture or maintaining the native culture1,76. It seems also worthwhile to pay attention to intersecting factors that might influence a member of a cultural group, namely gender and literacy. However, we first need to establish whether to culturally adapt health promotion IMI at all. The present findings at least suggest—except for physical activity IMI—a non-favorable cost-benefit ratio, a result that still is in need of stronger evidence.

Another topic of relevance to our findings is the reach, uptake, and intervention adherence of culturally adapted IMI. Even if such IMI would be effective, they still need to be used in order to exploit their full potential. Although internet technologies are widely used globally78, there are still barriers to utilizing these technologies, which cause inequalities in accessing the internet and mobile technologies and comprehending health information16,79,80. Moreover, pure mobile-based interventions are seemingly less effective than internet-based or combined interventions81,82,83, which might also affect for instance minority populations, where they are more likely to use a smartphone to access the internet than non-minority populations where multiple device (e.g. tablets, desktops) ownership is common84. To increase adherence, multimodal content and guidance (direct contact with the provider)85 might be useful via diminishing issues of health literacy, motivational and volitional aspects, and the digital divide71,86,87. These aspects might be particularly important in people of certain cultural backgrounds living in a high-income country, people living in low-income countries, and/or vulnerable populations, such as immigrants, given the limited representation of many of these populations in the health research, and high rates of drop-out86,88,89,90,91.

This review has some limitations. First, we included interventions concerning only five prominent areas of health promotion thus results might not be generalizable to other health promotion domains. Second, the definition of cultural adaptation varies, and our operationalization of culturally adapted interventions resulted in the exclusion of studies that investigated IMI that were developed newly in a culturally sensitive way. Comparison of culturally adapted vs. culturally sensitive interventions is an interesting further research topic. Moreover, comparing culturally adapted versus culturally sensitive IMI might present insights into whether it is worthwhile to develop a novel IMI for a group or adapt an already existing one. Third, we were able to pool data from only 13 studies, which further limit the generalizability of our results. Due to this limited evidence base, analyses were restricted to the main research questions while subgroup analyses were not feasible yet. Future updates might allow for exploring the between-study heterogeneity highlighted in the present review, while the findings reported here could guide researchers in what to examine next. In this context, we suggest adding generic outcomes such as health-related quality of life, daily functioning, or self-efficacy to domain/disease-specific outcomes to allow for cross-trial cross health promotion domain comparisons in the future. Fourth, although the studies included in this meta-analysis were culturally adapted, the adaptation process was rarely well defined. Fifth, only three out of 13 studies used a theory to adapt the intervention. And last, none of the studies were comparing culturally adapted IMI to non-adapted IMI. This creates a difficulty to draw any firm conclusions about the differential effectiveness of culturally adapted interventions. Despite these limitations, this meta-analysis had some strengths. To our knowledge, this is the first meta-analysis concerning culturally adapted health promotion IMI. Moreover, this a-prior registered systematic review included mostly scientifically sound RCTs from a broad sample and health behavior topics. A summary of recommendations for future research is represented in Box 1.


Based on the present findings, culturally adapted IMI might not be superior compared to control conditions in the short- and long-term, except for physical activity. Although they might exhibit a more attractive health offer to their target group, their usefulness is questionable or at least need further examination. Thereby, it might be worthwhile to take into consideration intersecting aspects of experiences of people of certain cultural groups regarding health behaviors to assure acceptability and effectiveness when designing interventions and contribute to diminishing health inequalities.


Protocol and registration

This systematic review and meta-analysis has been registered at PROSPERO (Registration number: CRD 42020152939) and follows the format of the PRISMA guideline92. Review protocol93 described the aim, methodology, and data analysis plan in advance. Changes to study protocol are listed in the supplementary notes.

Eligibility criteria

Studies were included if they (1) were RCTs, (2) had no treatment, treatment as usual (TAU), placebo, waitlist, or active control conditions, (3) were delivered via the internet, (4) were culturally adapted for a population that differed from the original intervention’s target group, (5) examined a health promotion intervention on healthy eating, physical activity, alcohol consumption, sexual health behavior and/or smoking cessation (6) reported one of the respective health promotion-specific outcomes: body mass index (BMI), time spent exercising, change in condom use, level of smoking, level of alcohol consumption, or one of the following generic outcomes: health-related quality of life and self-efficacy.

Information sources and study selection

The initial search was conducted in the following databases on 26. August 2019: Cochrane Central Register of Controlled Trials (CENTRAL), EbscoHost/MEDLINE, Ovid/Embase, EbscoHost/PsychINFO, and Web of Science. A combination of keywords (including MeSH terms) indicating culturally adapted IMI for health promotion has been used. The search terms are published in this review’s protocol93 and cover comprehensively both the topic of the present review as well as one of a parallel systematic review on culturally adapted IMI for mental health conditions94,67. There were no restrictions on the publication date. A search update was conducted on 15 October 2020.

All search results were merged into Covidence95 and duplicates were automatically removed. Two reviewers screened titles and abstracts of the identified articles against the inclusion criteria and selected potentially relevant articles for the full-text screening. Full-text screening has been performed by two reviewers independently, disagreements have been solved by consensus or a third reviewer where needed. The study selection is illustrated in the PRISMA flow diagram (see Fig. 8).

Fig. 8: Prisma Flow chart92.
figure 8

Study identification, selection, and inclusion represented on the diagram. An asterisk symbol represents a parallel review conducted regarding the culturally adapted internet- and mobile-based interventions concerning mental health.

Data extraction

Data extraction was conducted by two independent reviewers and then extracted data was then checked by a third reviewer. The following data were extracted from the included studies: publication details, study participants (demographics and cultural background, baseline characteristics), study design, study setting, characteristics of the original and culturally adapted intervention, health behavior-specific and generic outcome measures, information regarding cultural adaptation (content, utilization of theoretical or evidence-based components). Behavioral outcomes are defined as: physical activity measured via physical activity minutes per week with accelerometers or self-report questionnaires; healthy eating measured via BMI; alcohol consumption assessed via the level of alcohol consumption; smoking cessation assessed via the level of smoking or the abstinence percentage; sexual health behavior assessed with condom use. Generic outcomes were defined as health-related quality of life and self-efficacy, assessed by means of validated self-report questionnaires. Control conditions were categorized into active (placebo, other health promotion interventions & TAU) and passive controls (waitlist & No treatment). When related information could not be extracted, corresponding authors of the articles were contacted to obtain information. The extracted data was tabulated.

Risk of bias

Two independent reviewers performed quality assessments with Cochrane Collaboration’s Risk of Bias Tool 2.096. The RoB tool 2.0 has five domains including bias arising from the randomization process, bias due to deviations from intended interventions, bias due to missing outcome data, bias in the measurement of the outcome, bias in the selection of the reported result. A third reviewer solved disagreements following a discussion between the reviewers. The Kappa statistic was used to calculate interrater reliability97.


For each study, a standardized mean difference (SMD) and 95% confidence intervals (CI) were calculated with mean scores of intervention and control groups. In order to decrease the bias of small samples, Hedges’ g was calculated98. Effect sizes were recoded when higher scores of an outcome assessment indicated worsening results (e.g. BMI and level of alcohol consumption). For continuous outcomes, Hedges’ g and 95% CIs were reported; for dichotomous outcomes, odds ratios and CIs were reported. Random effects model was chosen for analyses due to an expected diversity among IMI of health promotion, sample size, and duration of intervention among studies99. Data were pooled to calculate a standardized mean effect size for each outcome and a forest plot with 95% CIs, in the case of at least three studies reporting the respective outcome. Otherwise, results were presented descriptively. Sensitivity analysis was planned to assess the impact of studies with a high risk of bias. Analyses were performed in R package meta and metafor, and Review Manager 5100,101,102,103. Continuous effect sizes were categorized along with Cohen’s rule of thumb with 0.20 considered a small effect, 0.50 medium effect, and 0.80 large effects104. In order to assess publication bias, we planned to conduct funnel plots.

Statistical heterogeneity among studies was analyzed with the I2 statistics97,105. Statistical heterogeneity refers to the variability among effect sizes in a meta-analysis106. However, the veracity of measures of heterogeneity is arguable; therefore, their interpretation should be made with caution97. Statistical heterogeneity was calculated with the I2 test for each outcome domain. Heterogeneity I2 ≥ 60% was regarded as substantial heterogeneity, in which case no pooled effect sizes are reported. Subgroup analyses were planned to explore possible sources of heterogeneity on population, duration of the intervention, delivery of the intervention, and presence of guidance.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.