Effectiveness of Cultural Adaptations of Interventions Aimed at Smoking Cessation, Diet, and/or Physical Activity in Ethnic Minorities. A Systematic Review

Background The importance of cultural adaptations in behavioral interventions targeting ethnic minorities in high-income societies is widely recognized. Little is known, however, about the effectiveness of specific cultural adaptations in such interventions. Aim To systematically review the effectiveness of specific cultural adaptations in interventions that target smoking cessation, diet, and/or physical activity and to explore features of such adaptations that may account for their effectiveness. Methods Systematic review using MEDLINE, PsycINFO, Embase, and the Cochrane Central Register of Controlled Trials registers (1997–2009). Inclusion criteria: a) effectiveness study of a lifestyle intervention targeted to ethnic minority populations living in a high income society; b) interventions included cultural adaptations and a control group that was exposed to the intervention without the cultural adaptation under study; c) primary outcome measures included smoking cessation, diet, or physical activity. Results Out of 44904 hits, we identified 17 studies, all conducted in the United States. In five studies, specific cultural adaptations had a statistically significant effect on primary outcomes. The remaining studies showed no significant effects on primary outcomes, but some presented trends favorable for cultural adaptations. We observed that interventions incorporating a package of cultural adaptations, cultural adaptations that implied higher intensity and those incorporating family values were more likely to report statistically significant effects. Adaptations in smoking cessation interventions seem to be more effective than adaptations in interventions aimed at diet and physical activity. Conclusion This review indicates that culturally targeted behavioral interventions may be more effective if cultural adaptations are implemented as a package of adaptations, the adaptation includes family level, and where the adaptation results in a higher intensity of the intervention. More systematic experiments are needed in which the aim is to gain insight in the best mix of cultural adaptations among diverse populations in various settings, particularly outside the US.


Introduction
There is a high prevalence of chronic diseases among ethnic minorities in high-income societies. For example, hypertension and diabetes are highly prevalent among populations of African and South-Asian origin in the United States and Europe, particularly compared with populations of European origin in those regions [1,2]. Important preventable risk factors for these diseases are health-related behaviors that include smoking, diet and physical activity (PA). Hence, interventions to promote healthier behaviors are crucial to a reduction in these diseases. Often, however, interventions targeted to the general population do not reach ethnic minorities. In addition, there are indications that such interventions have limited effects on health behavior [3][4][5].
Increasing the cultural sensitivity of lifestyle interventions is generally expected to enhance their appropriateness and effectiveness [3]. Cultural sensitivity can be described as the extent to which a target population's ethnic and/or cultural characteristics, experiences, norms, values, behavioral patterns, and beliefs as well as relevant historical, environmental, and social forces are incorporated into the design, delivery, and evaluation of targeted health promotion materials and programs [6]. This can be done through cultural adaptations such as matching materials to group characteristics or targeting cultural values of the population [6].
Increasing numbers of studies and reviews have examined the effectiveness of a broad range of culturally adapted interventions targeted at a variety of diseases (i.e. diabetes, asthma, or stroke) or health-related behaviors such as physical activity (PA) [7][8][9][10][11]. However, none of them make clear whether the reported effectiveness can be attributed to particular cultural adaptations. More importantly, in most of the studies included in previous reviews, different types of interventions were used for the control and intervention groups. For example, a culturally adapted 12-week group intervention compared with a control group that received a leaflet only. The differences in study arms of such studies do not allow us to draw firm conclusions about the effectiveness of the cultural adaptations.
Instead, a study design in which the only variation between the intervention and control group is the specific cultural adaptation that will be studied on effectiveness, is precisely what is needed to gain more insight into the effectiveness of specific cultural adaptations, and to determine which adaptations are really necessary and which are not.
This paper aims to fill this gap by systematically reviewing randomized and non-randomized studies that evaluate the effectiveness of health interventions targeted to ethnic minorities and focusing on smoking cessation, diet and/or PA. We were interested in evaluating studies in which the intervention without a particular cultural adaptation is compared with the same intervention with a particular cultural adaptation. Our specific objectives were: 1) to systematically review the effectiveness of cultural adaptations in interventions that targeted smoking cessation, diet and/or PA among ethnic minorities in high income societies, in terms of changes in the abovementioned behaviors. 2) to explore those features of cultural adaptations that may account for their effectiveness.

Methods
We conducted a systematic review of the effectiveness of culturally sensitive interventions that targeted smoking cessation, diet and/or PA among adults from ethnic minorities in high income societies. We based our review methodology on the Guidelines for Systematic Reviews of Health Promotion and Public Health Interventions [12]. The protocol has not been published or registered but is detailed below.
The protocol included the aim, the search strategy (Table 1) and the selection procedure. For the selection procedure, we developed a registration sheet in which the studies were registered and in which the reason for selection or non-selection was registered.
We developed, tested, adjusted, and retested the search strategy to ensure it was as sensitive as possible. We searched for controlled studies with no language restrictions in MEDLINE, PsycINFO, Embase (all from January 1997-September 2009), and the Cochrane Clinical Trial Database (January 1997-April 2010). We choose to include papers that were written in the context of increased refinement of culturally sensitive interventions as defined by Resnicow in his landmark paper in 1999 [6]. The year 1997 was chosen as the starting point to minimize the chance for missing important studies.
To ensure the consistency of the search strategy across the three different domains (smoking, diet and PA), the strategy consisted of two parts: a basic search strategy aimed at finding intervention studies among ethnic minority populations, and a behaviorspecific supplement aimed at the three behaviors. Both MeSH terms and free-text words were used (Table 1).
Subsequently, the publications from each database were crosschecked manually and duplicates were removed. At least two reviewers per lifestyle domain (smoking cessation, diet and PA) screened the titles and abstracts of the identified publications and excluded all publications that clearly did not meet our inclusion criteria. For all the other articles, full texts were obtained and two reviewers independently assessed eligibility. If there was uncertainty about inclusion, consensus was achieved by discussion.
Criteria for inclusion in our review: a) The study described the effect evaluation of a lifestyle intervention culturally adapted to a specific adult, ethnic minority population living in a high-income society. b) The study included a control group (randomized or nonrandomized) that was exposed to the basic intervention without one or more cultural adaptations that were evaluated on their effectiveness in this study (i.e. the adaptation under study). c) The outcome measure of interest was behavior change regarding smoking cessation, diet, or PA.
Our inclusion criteria allowed studies in which the control intervention made use of cultural adaptations but in which the intervention group received one or more additional cultural adaptations (i.e., the specific cultural adaptation under study). Studies were excluded if the control and intervention groups seemed to have received different interventions (e.g., a leaflet versus a culturally adapted group intervention), and the authors did not describe the difference between the intervention and control groups as including a difference in use of cultural adaptations.
Two reviewers per lifestyle domain used a pretested form to independently summarize the data, including study characteristics, intervention characteristics, details about the intervention and cultural adaptations, study design, and intervention effects of the included articles. To define 'cultural adaptations' we used the definition of Resnicow et al. [6] for surface-structure and deepstructure adaptations. We considered surface-structure adaptations as those adaptations that match materials or messages to observable, superficial characteristics of the target population. For example, interventions that used pictures of the target population. Deep-structure adaptations are those adaptations that address core cultural values or those ethnic, cultural, historical, social or environmental factors that may influence specific health behaviors. For example, interventions address the role of food in hospitality in the culture of the target population.
For the included articles, the Effective Public Health Practice Project (EPHPP) quality assessment tool for quantitative studies was used to assess the individual studies' risk of bias [13]. This tool rates design, selection bias, allocation bias, confounding, blinding, data-collection methods, withdrawals and dropouts (attrition bias), with ''strong'', ''moderate,'' or ''weak''. Based on the number of strong, moderate, or weak scores, the study was given an overall score. Studies were rated as strong if there were no weak ratings, as moderate if there was one weak rating, and as weak if there were more than two weak ratings. Each of the two reviewers rated the quality independently. If there was disagreement, the results were discussed and a final decision was made. We e-mailed all authors of the included studies to retrieve missing information, and completed the quality assessment using the information they provided. We assessed whether we could find an association between studies with a statistically significant effect or no effect and the quality of the studies or whether publication bias could affect the results. We found a large variety in outcome measures, which made a detailed analysis of effect sizes impossible. Therefore, a narrative synthesis was conducted describing the study design, the intervention and cultural adaptations of the effective and non-effective studies, target population, setting, attrition rates, and reported effectiveness of the intervention (based on statistical significance).
To provide detailed information about the results of the studies, we present the effect sizes in the tables. Effect sizes were described in terms of differences in mean and percentage, depending on the outcome measures of the studies. To answer research question two, we compared the characteristics of the adaptations for effective with those of non-effective interventions.

Study selection
The search across the four databases yielded 19222 publications for smoking cessation, 10294 publications for diet, and 15388 publications for PA ( Figure 1). Full texts were obtained for 32 studies on smoking, 72 on diet, and 51 on PA. Of these studies, 132 (27 on smoking, 62 on diet, and 43 on PA) were excluded because the intervention and control groups received different types of interventions. Within the 132 excluded studies was 1 study where the control group spontaneously took over the cultural adaptations, so that it no longer differed from intervention group [14]. We decided to include studies that investigated the effectiveness of additional cultural adaptations to an intervention that already included one or more cultural adaptations. Six studies were found in both the search strategy for diet and the search strategy for PA. One study that was included via the search for smoking cessation also targeted diet and PA. After removal of overlap because of multiple behaviors addressed, this left 17 studies that met all the inclusion criteria: 4 on smoking cessation [15][16][17][18], 4 on diet [19][20][21][22], 2 on PA [23,24], 6 on PA and diet combined [25][26][27][28][29][30], and 1 on all three behaviors [31,32].
Outcome measures in all of the smoking cessation studies were self-reported abstinence. Three studies validated these outcomes with CO levels [15,16,18]. Follow-up time varied between four weeks [15,17] and five years [31,32]. Primary outcomes varied widely in studies targeting diet and/or PA. For example, Babamoto et al. [26] measured if participants were physically active at least three times a week, while Keyserling et al. [29] measured the PA increase to 30 minutes per day and Fitzgibbon et al. [28] measured energy expenditure as indicator of PA. Outcome measures of studies targeting diet included fruit and vegetable intake and/or fat intake and in one study percent calories from fat and grams of fiber intake [20]. Each study used a different follow-up time. In the one study that targeted all three health behaviors, outcome measures for diet and PA were BMI, LDL cholesterol, energy expenditure and metabolic equivalent [31,32].

Quality assessment
Of the 17 included studies 10 authors responded to our questions on clarifications regarding quality assessment. Subsequently, 5 studies were assessed as strong, 11 as moderate, and 1 as weak ( Table 2). Moderate and weak studies mostly had problems with attrition rates or lacked information about response or attrition rates. This implies that there might be a chance of selection bias in these studies, namely the possibility that the intervention had a different effect in that part of the study population that was lost to follow-up. Other problems identified were insufficient blinding, not reporting information about reliability of data collection tools or not controlling for possible confounders. The strong and moderate studies reported mixed outcomes, i.e. some found that cultural adaptation(s) were effective, while others did not find statistically significant effects. Hence, we could not find a clear association between quality of the studies and effectiveness.

Overall effectiveness of interventions per behavior
Five out of 17 studies that tested one or more cultural adaptations found statistically significant results on the effectiveness of cultural adaptations. Regarding smoking cessation, three studies (of four) reported statistically significant decrease in  [19][20][21][22][23][24]28,33]. The study that was aimed at all three behaviors reported effectiveness on smoking cessation and energy expenditure [31,32]. The interventions and the cultural adaptations that were tested (with or without effectiveness) are described below.
Cultural adaptations that were reported to be effective Table 3 shows the content of the five intervention studies with cultural adaptations that reported statistically significant effective-ness on primary outcomes. Three studies targeted smoking cessation [16][17][18] and one targeted diet and PA [26]. The study by Cene et al. [31,32] which targeted smoking, diet and PA, reported effectiveness on smoking behavior and energy expenditure [31,32]. All studies [16][17][18]26,32] incorporated a package of adaptations.
Smoking cessation. The content of the package of adaptations varied greatly. The adaptations were all added to standard telephone counseling that had been adapted for language and culture. Wetter et al. [17] provided proactive telephone calls in the adapted intervention. These focused on practical counseling and  social support as well as Hispanic values like 'culture of respect' and 'pleasant and agreeable family' (deep-structure adaptations). The adapted intervention resulted in statistically significant higher abstinence rates after 12 weeks (27.4% vs. 20.5%) after controlling for demographic and tobacco related variables. Orleans et al. [16] adapted standard telephone counseling and a self help guide. Telephone counseling was tailored to motives and barriers specific for African Americans (deep-structure adaptations). The self help guide used African-American models and provided information about smoking among African Americans (surface-structure adaptations) as well as incorporating African-American values (deep-structure adaptations). At 12 months follow-up abstinence in intervention group was statistically significantly higher (25% vs. 15.4%). Woodruff et al. [18] replaced three of the six standard telephone consultations by four home visits. In the sessions the communication style was adapted and social cognitive principles were congruent with Latino smokers. In addition they focused on family concerns (deep-and surface-structure adaptations). At three months, seven day abstinence rates were statistically significantly higher in the intervention group (20.5% vs. 8.7%).
Three behaviors. In the study of Cene [32] the basic intervention consisted of consultations with a healthcare provider in a clinical setting and free pharmacotherapy. In the adapted intervention a community health worker (CHW) provided consultations in a non-clinical community setting (surface-structured adaptations). At one year follow-up smoking prevalence decreased significantly more in the intervention group (decrease of 6% vs. 3%). Energy expenditure increase was statistically significantly higher in the intervention group. Regarding the outcomes BMI, LDL cholesterol and metabolic equivalent [31,32] no significant effects were found at one year follow-up.
Diet and PA. The intervention of Babamoto [26] had three intervention arms: 1) counseling by a healthcare provider, using Spanish-language materials culturally adapted to the local population; 2) counseling by healthcare provider plus a professional case manager (who used standardized clinical protocols, was trained to be culturally sensitive, and was able to communicate in Spanish) and written materials; and 3) an enhanced counseling intervention which consisted of the healthcare provider plus cultural tailoring by a culturally competent trained CHW, family educational sessions and the written materials (deep-structure adaptations). At six months follow-up statistically significantly more respondents reported increasing their fruit and vegetable intake in the CHW and the case manager group compared with the standard care group. Fat intake was statistically significantly lower in the CHW group only (13% decrease vs. 2% and 9% decrease respectively).
Smoking cessation. One study evaluated the cultural adaptation of a self help guide and videotaped education sessions designed for the general population [15]. Cultural adaptations consisted of surface-structure adaptations such as the use of African-American role models and deep-structure adaptations by incorporating socio-cultural values into the self help guide and   video. Smoking cessation rates, although lower in the intervention group, did not differ significantly between groups. Diet and PA combined. One study [29] compared standard clinic-based counseling which already included some culturally specific written materials with an intervention condition that employed group sessions and monthly telephone contact with a lay health advisor as additional adaptations. Both the group sessions and contact with the lay health advisor were adapted to the sociocultural values of the target group (deep-and surface-structure adaptations). This study was not sufficiently powered to demonstrate differences between the two conditions.
Four studies investigated a single adaptation. One study [25] investigated whether group composition (ethnically mixed versus homogenous groups) in a group intervention influenced weight loss among African Americans (surface-structure adaptation) and found no difference between the groups. Fitzgibbon et al. [28] evaluated incorporation of faith in a weight loss intervention. This deep-structure cultural adaptation showed no significant effectiveness. In the study of Staten et al. [30] the cultural adaptation consisted of extra contacts with a CHW who communicated with patients, provided information, organized group walks and encouraged participants to find walking partners (deep-and surface-structure adaptations). There were no statistically significant differences between groups at follow-up although the intervention group had a slightly higher fruit intake. Finally, Campbell et al. [27] tested the effectiveness of employing a lay health advisor (LHA) (surface-structure adaptation) in addition to tailored printed materials (TPV): in the intervention group the    n/a n/a Intervention group had slightly higher fruit/vegetable intake at follow-up.
Group received the same as the PC + HE group but also communicated on a regular basis -semi weekly to monthly with community health workers. The CHW provided information and support and organized bi-monthly walks.
In addition, this group was offered 2 health education classes and monthly newsletter for 12 months.
Participants received brochures and nurse practitioner discussed benefits of and barriers to increasing PA and Fruit and Vegetable consumption this advice was tailored to individual.

Adaptations tested
No differences between all groups in intake of % calories from fat and in recreational activity (MET hrs/week): Surface-structure adaptations  In the tailored condition 9.7% returned all activity inserts; 41.9% returned none. No significant differences were detected among groups for the primary outcomes % calories from fat and total dietary fiber.
-Tailored print newsletters and activity insert as in the tailored print condition.
In addition: weekly home visits or telephone calls from sequentially assigned promotoras over a 12-week period.
12 weeks of tailored print newsletters (based on baseline data) and activity inserts mailed to their homes Percent calories from fat decreased: Promotora: 2.2% vs. tailored print: 0,6% vs. usual care: 1.5% Secundary outcomes:

Adaptations tested
The promotora group was significantly or marginally lower than the control and tailored groups for energy, total fat, total saturated fat, and total carbohydrates (p.,.05-.10).

Surface-structure adaptation
The promotora group was significantly lower than the tailored group for glucose and fructose (p,.05).
-Use of Promtora's For every outcome, the group differences seen were no longer significantly different at 6-month followup and 12-months follow-up. Instead, there were group-by-time interactions; effects achieved by the promotoras dissipated over the 12month follow-up period while the effects of the tailored group concurrently improved.
Campbell et al. 1999 [19] The spiritually oriented bulletin   LHA provided the same information to respondents as they received at home, in the form of TPV. In this study the 'TPV only' condition had a statistically significant effect on fruit intake and percentage that meet the recommendations of five servings a day; implying that the cultural adaptation was not effective.
Diet. The studies of interventions aimed to change diet all assessed one type of cultural adaptation instead of a package of different cultural adaptations. Kreuter et al. [34] studied the effect of cultural tailoring and behavioral tailoring using magazines (deep-structure adaptations). At six months follow-up African-American women receiving a combination of both behaviorally tailored as well as culturally tailored magazines had greater increases in daily fruit and vegetable consumption although this was not statistically significantly different from the group receiving behaviorally tailored magazines only. Elder et al. [20,33] assessed the effect of weekly home visits and/or telephone calls from LHA in addition to tailored print materials. LHA provided the same information as the print materials (surface-structured adaptation). No effects on primary outcomes were found at 12 weeks and effects on other outcomes disappeared at six-month follow-up. The study by Campbell et al. [19] tested whether a behaviorally tailored bulletin incorporating health messages with a spiritual orientation (deep-structure adaptations) were more effective when the source was the pastor versus an expert (surface-structure adaptation). No differences on fruit and vegetable consumption were found between the two conditions. Finally, Rescinow et al. [22] tested the effectiveness of targeting messages to the ethnic identity of respondents by using untailored, ethnically neutral graphics versus newsletters with information targeted at ethnic identity in a study population of African Americans. A significant increase in fruit and vegetable intake was only found for a subpopulation of participants with Afro-centric identity but not for the total group.
PA. The two culturally adapted interventions aimed at increasing PA without treatment effect tested a package of cultural adaptations. Chiang et al. [23] investigated the incorporating of cultural values in the messages and the involvement of family members in a standard eight-week walking program with translated written materials for Chinese Americans and weekly telephone monitoring. At follow-up walking duration was not significantly different between the two study arms. In the study of Newton et al. [24], deep-and surface-structure adaptations were made to an intervention consisting of ten group sessions. In the adapted intervention, all group members and the counselor were African American, the location was preferred by the target population and socio-cultural values regarding exercising were incorporated. There were no differences in cardio respiratory fitness and exercise frequency between the adapted and basic intervention group.
Are there features of the adaptations that may explain their effectiveness?
We could distinguish five broad categories of adaptations: level of adaptation, i.e. surface-vs. deep-structure; cultural values vs. interventions involving community health workers or lay health advisors; incorporating family vs. religious values; interventions employing intensive vs. non-intensive strategies; and use of a package of adaptations vs. one type of adaptation (Table 5).
Surface-vs. deep-structure adaptations. We found no indication that the level of adaptations influenced effectiveness. Studies that demonstrated statistically significant effects (n = 4) as well as several studies that did not (n = 9) included deep-structure adaptations. The same applies to surface-structure adaptations.
Adaptations based mainly on cultural values vs. involvement of lay health advisors/community health workers. We also observed no pattern of effectiveness when we distinguished the studies on the basis of adaptations that involved community health workers or lay health advisors versus adaptations mainly based on incorporating cultural values into intervention materials or in the counseling conducted by professionals.
Distinguishing incorporating family versus religious values. We found that we could distinguish topics used in the cultural adaptations, i.e. religious values, family values and/or family involvement and other cultural values which were not further specified. Statistically significant effects on primary outcomes were found by three interventions (of four) that incorporated family values and/or involved family members and by none interventions (of five) that incorporated religious values.
Intensity of the adaptation. In 9 of the 17 studies, the cultural adaptation implied an increase of the intervention's intensity (e.g. extra sessions with a lay health advisor). This was the case in all studies that reported statistically significant effects on primary outcomes.
Number of adaptations tested. Five studies (of nine) that incorporated a package of adaptations, e.g.; additional proactive calls together with tailoring to cultural values [17] reported statistically significant effects. Studies using one type of adaptation, e.g. use of homogeneous groups, [25] didn't show statistically significant effects (n = eight).

Discussion
To our knowledge, this is the first review that investigated the effectiveness of specific cultural adaptations in interventions for ethnic minorities targeted at smoking cessation, diet, and/or increasing PA. We identified 17 studies that evaluated the effectiveness of one or more cultural adaptations. The adaptations tested ranged from incorporating socio-cultural values of the target population to involving community health workers or lay health advisors; from employing a single adaptation to a package of adaptations; and from using deep-structure adaptations to surfacestructure adaptations. In five studies the adapted intervention had a positive statistically significant effect on the primary outcomes. These were mainly interventions that targeted smoking cessation. Twelve studies showed no statistically significant effects on primary outcomes, although some studies presented trends favorable for cultural adaptations. We observed that interventions incorporating a package of cultural adaptations, cultural adaptations that implied a higher intensity and those incorporating family values were more likely to report statistically significant effects.

Limitations
Before interpreting the results, the limitations of our review need to be considered. Like in all systematic reviews, publication bias (e.g. because studies were small or non-effective) may have limited the number of studies to be found. However, as our search did yield several small non-effective studies, we do not expect that this would have changed our results.
Second, the studies show heterogeneity in design and quality. Outcome measures varied too much to conduct a detailed analysis of effect sizes. In addition, follow-up time varied between studies. It may be expected that studies with a short follow-up interval would be more likely to report effects; however we did not find such a pattern. A similar argumentation can be followed for the quality of the studies; studies that have a strong design are less likely to be prone to bias compared with those studies with a moderate or weak quality score. For example, it may be that studies with a moderate or weak design have hidden effects due to a more motivated control group, or, in contrast, that effects were found just because of differences between the experimental and control group. In our analysis we found both moderate and strong designs in studies either with or without effect. Based on these results we conclude that we found no association with quality of the design and expect that the conclusions of our review are not likely to be affected by differences in design quality. If we consider the isssue of randomization as a separate study design issue, we see that two of the studies were not an RCT. Both of these studies showed no effect. It may be that some of the potential effects could be hidden due to bias in these studies. However, the fact that 10 out of the 15 RCT's also found no effect, means that this is unlikely to affect our conclusions.

Interpretation of results
We found that studies that reported effects had tested a whole package of adaptations. Furthermore, the adaptations were more likely to be infused through the intervention as a whole and at several levels. For example, alongside the individual, the level of the family was included. Thus, in line with evidence about behavioral interventions in general, it seems that interventions targeting multiple levels result in larger effect than interventions focused at a single level [35].
The finding that including cultural adaptations is likely to result in a more intense intervention may reflect that enhancement of the intervention is an important aspect of the cultural adaptations. Most study populations in studies incorporating cultural adaptations have lower educational levels and are living in disadvantaged circumstances (e.g. [17,21,29]). It's possible that more time and attention is needed to make the information understandable to these groups [36] and that the effects of the included studies are also attributable to the adaptation to lower health literacy. This implies that adaptations to characteristics other than norms and values should also be taken into account when developing interventions.
One striking result of our review is that the effectiveness of specific cultural adaptations was seen mainly in studies on smoking cessation and less often in studies on diet and PA. Regarding PA, this is in line with an other review [37]. An additional factor which may play a role is the relationship between certain behaviors (like diet) and deeply rooted cultural values and personal identity. These behaviors may therefore be less changeable, as suggested by several studies [38][39][40]. As a result, changing these habitsincluding what, when, and how food is eaten -may need more attention when adapting an intervention.

Implications for future research
The current review indicates that the current knowledge about the effectiveness of cultural adaptations is limited. To date most studies have been conducted in the United States among African Americans and Hispanics, which may not be generalizable to other population groups or to other countries. Therefore, future research needs to be conducted among other ethnic minority groups with different migration histories or local circumstances.
In addition the results indicate that the questions ''what is a cultural adaptation?'' and ''how is it employed?'' are important issues to consider at the inception of culturally sensitive interventions and their subsequent reporting. We started with the assumption that we would find adaptations such as the use of prevalence rates of diseases from the target population or the use of cultural values in the health promoting message. But, we found that factors as the intensity and of combination of adaptations appear to play an additional, important role. Our analysis was limited by its reliance on the authors' varying descriptions of the adaptations they applied. It seems that in order to move this field forward, more explicit description and discussion of these issues is needed.
The results of this review provide useful input for further research. We recommend, first, that the design of future studies should include a standard care arm with regular care as provided by general population without any extra interventions. Examples can be seen in Babamoto et al., Keyserling et al.,and Newton et al. [22,24,29]. With this design we are better able to assess the effect of the basic intervention and the adaptations tested.
Second, although a broad mix of cultural components appears most effective, it remains important to identify the specific cultural adaptations that are most effective within this mix. This calls for more 'experimental' designs in which well-defined adaptations or combinations of adaptations are tested. Examples of studies that tested one type of adaptation are studies done by Resnicow et al. [22] on assessing the incorporation of ethnic identity and by Ard et al. [25] on assessing the effect of group composition. However, in these studies the basic intervention was already culturally adapted which may have limited the potential effect of the single adaptations. To assess the potential effect of single adaptations, studies can be designed similar to pharmacological trials. Adaptations that are considered for use could be tested in firstor second-phase studies solely and combined [41]. If effective, they could subsequently be included in larger controlled trials to provide insight into the package of adaptations that is most effective.
Finally, there are other characteristics of cultural adaptations that might be worth studying in more detail. The process of developing the cultural adaptation (e.g. was the target population involved in this development?), the way the adaptation was executed (e.g. related to the competence of the executers), the types of determinants addressed by the adaptations (e.g. enabling or cognitive factors) and the environmental levels addressed by the intervention (e.g. individual versus family level) are all examples of factors that may affect the effectiveness of cultural adaptations. More insight into these characteristics of cultural adaptations might help to select whether or not a cultural adaptation will contribute to the effectiveness of an intervention.

Conclusion
The results of our review indicate that: 1) Culturally targeted interventions may be more effective if cultural adaptations are implemented as a package of adaptations, the adaptation addresses family influences, and where the adaptation implies a higher intensity of the intervention; 2) Adaptations in smoking cessation interventions seem to be more likely to be effective than adaptations in interventions aimed at diet and PA; 3) More systematic experiments are needed in which the aim is to gain insight in the best mix of cultural adaptations among diverse populations in various settings, particularly outside the US.