A two-armed randomized controlled trial was conducted to examine the effectiveness of AOC-TR, comparing the experimental group with a waiting list control group. The experimental group obtained direct access to the intervention and the waiting list control group received access after 4 months. Ethical approval was granted by an independent medical ethics committee (METc VUmc registration no: 2010/82). The trial is registered in the Dutch Trial Register (NTR2303). A detailed description of the trial design has been published earlier [23].

Participants aged 18 years or older with depressive symptoms as measured by a Center for Epidemiologic Studies Depression Scale (CES-D) score ≥16 [24] and a Turkish background (participant or at least 1 parent was born in Turkey) were included in the trial. To be included, participants also needed to have access to a computer with Internet, have an email address, and have returned a signed informed consent form.

Exclusion took place if the participant was suicidal, which was assessed in 2 steps as part of the online screening. First, the suicide item on the Beck Depression Inventory II (BDI-II) was presented [25,26]. The BDI-II is validated among Dutch [27,28] and Turkish populations [29,30]. Second, if the response was affirmative, the suicide risk was measured with the suicidality section of the Mini-international Neuropsychiatric Interview (MINI) [31,32] in Dutch [33] or Turkish [34]. Participants with a relatively high risk were advised to contact their general practitioner or were referred to the online portal for suicide prevention [35].

Recruitment took place from June 16, 2010 to March 15, 2012. Participants were recruited among the adult Turkish migrant population via several recruitment strategies. The following strategies were applied: advertisements in Dutch and Turkish national newspapers, magazines, and community websites; banners on health-related websites for migrants; and through social media. Information brochures were distributed at Turkish associations in the Netherlands, mental health care organizations, and sociocultural organizations. The recruitment took place in 2 languages, Dutch and Turkish. Facebook was the most effective recruitment strategy during the trial.

Recruitment on Facebook took place between January 1, 2011 and March 15, 2012. A personal profile and a fan page about the trial were created on Facebook, where pictures from the research website, information about the project, and status updates were shared. Facebook groups related to Turkish migrants and Turkish groups focusing on (general) health and psychology were joined. Next, random people from these groups were invited to join the fan page and friend requests were sent. A total of 584 friend requests were accepted by these invited people. Afterwards, friends of our friends list and people from the Facebook groups began to add our Facebook profile, which resulted in 3308 friends on the research profile by the end of the trial. Friends from our friends list and from the joined Facebook groups sent us messages or, if the researcher was online, chat conversations took place with them. We received or had chat conversations (about diverse topics, including application to the trial) with 348 people.

The advertisements contained a link to our research website with detailed information about the trial [36,37]. Interested parties could apply by sending an email to the researcher, who then returned a digital information brochure about the study, the informed consent form, and a unique Web link for an online screening questionnaire.

The original version of the self-guided, problem-solving intervention (AOC [20]) was adapted to a culturally sensitive intervention (AOC-TR) in collaboration with the Trimbos Institute (Netherlands Institute of Mental Health and Addiction). First, the intervention was translated from Dutch into Turkish, and then both versions were culturally adapted. Although there are multiple descriptions of cultural adaptation of psychotherapy to specific populations, it has been defined as the modification of intervention protocols according to the clients’ values, contexts, and worldviews [38]. Culture-specific adaptations in our intervention included several components: (1) the participants’ preferred language, (2) describing psychological problems in terms of idioms of distress (eg, using symptoms of depression instead of the term depression), (3) explicitly discussing migration and culture by using culture-specific cases and problems that are recognizable for the target group concerned, and (4) including recognizable examples of persons with similar problems (eg, a young woman who migrated 2 years ago and can’t find her way in the Netherlands). After adapting the intervention from the original Dutch version, 2 native Turkish persons evaluated the interventions both for language- and cultural-specific items in close collaboration with the first author who is a Turkish person herself. Finally, recommendations from these reviewers in terms of culture and language were incorporated in the interventions. Screenshots of the interventions are shown in Multimedia Appendices 1 and 2.

The AOC-TR consists of 5 sessions over 5 weeks. During the intervention, participants indicate what they think is important in their lives, they make a list of their problems and worries, and they categorize their problems into 3 groups: (1) unimportant problems, which are not related to what they think is important in their lives, (2) important and solvable problems, which are approached by a systematic problem-solving approach consisting of 6 steps, and (3) important but unsolvable problems, such as having lost someone through death or having a chronic general medical disease and making a plan for how to live with it. The core of the intervention is the 6-step problem-solving procedure, which teaches to use this technique during the course for several of their important and solvable problems. The idea is that by mastering this technique people will regain mastery of their problems and ultimately their lives.

The participants received feedback on their homework assignments in brief weekly emails in either Turkish or Dutch from the researcher (BÜI).

The control condition was a waiting list comparator; participants in this condition did not receive access to the intervention after randomization. However, they were provided with access to the intervention 4 months after the baseline measures.

The sample size was calculated on an expected difference of d=0.45 between the experimental and control groups. This expected difference was based on effect sizes derived from previous effect studies on Internet-guided problem-solving therapy for depression [21]. To achieve a power of 0.80 and an alpha of .05, we needed 78 participants at baseline in each condition (N=156). In keeping with our hypothesis, the primary and secondary outcomes were analyzed with a 1-tailed t test as in the study of Warmerdam and colleagues [21].

Participants were randomly assigned to the experimental or the control group after baseline assessment. The allocation schedule was generated by an independent researcher using a computerized system.

Figure 1 shows the flow of participants through the trial. A total of 287 individuals applied for participation. However, 66 of them did not complete the screening. The screening questionnaire was filled in by 221 individuals, of whom 125 were excluded primarily because of suicidal ideations (64/125, 51.2%). A total of 96 individuals met all inclusion criteria and were randomized to 1 of the 2 conditions.

Table 1 provides the baseline characteristics of study participants. The mean age of the participants was 35.2 years (SD 9.3) and 62% (59/96) were women. Most participants were born in Turkey (91%, 87/96) and preferred the Turkish language for study participation (89%, 85/96). More than three-quarters of the participants (78%, 75/96) were recruited through the Internet. The most important reason for choosing an Internet intervention was flexibility of use (62%, 59/96), followed by privacy and anonymity (23%, 22/96).

The mean score at baseline for all the participants on the CES-D was 29.9 (SD 9.6, range 11-52). There were no statistically significant differences between the experimental and control group at baseline on any of the demographic and secondary outcomes.

Of the 96 original participants, a total of 40 (42%) participants did not complete the posttest (6 weeks), and 59 participants (62%) did not complete the follow-up assessment at 4 months. Reasons for the high attrition rates are not known. There were no significant differences in attrition rates between the experimental (47%, 23/49) and the control group (36%, 17/47) at posttest (P=.29). However, at follow-up, the experimental group (74%, 36/49) had a higher attrition rate than the control group (49%, 23/47; P=.01).

Because of high attrition, we conducted completers-only analysis for follow-up results at 4 months after the start of the intervention. Results are shown in Tables 3-5. As is shown, the experimental group did significantly better on the primary and secondary outcomes analyses (clinically significant change, per-protocol analysis, and completers-only analysis).

Our results show no significant difference in improvement of depressive complaints in the experimental group compared to the control group on intention-to-treat analysis at posttest. This may be because our study was underpowered [57] (see also the limitations section). However, we found an effect size for the primary outcome (depression) of d=1.68 at posttest and d=1.13 at follow-up, which is an indication that the intervention could be effective with a sample size indicative of sufficient power. We did not find any differences on secondary outcomes. Recovery occurred significantly more often in the experimental group than in the control group at posttest (P=.01) but clinically significant change was not (P=.09). Completers-only analyses showed that the results of the analysis for the primary outcome differed from the imputation data, suggesting that the analysis was influenced by data imputation.

Our results did not support the effectiveness of the Internet-based, guided, self-help intervention, in contrast to the original version (AOC) [20]. As mentioned before, AOC was previously shown to be clinically effective in the general Dutch population in the reduction of depressive symptoms with a moderate effect size (Cohen’s d=0.50). However, the effect size at posttest in the current study was high, which might be an indicator of the possible effectiveness of the intervention for Turkish migrants when assessed in a larger sample.

The inclusion of ethnic minorities in clinical research has been a challenge for many years. Ethnic minorities are generally underrepresented in scientific and clinical analyses and are known to be a hard-to-reach population for research purposes [8,58]. Although previous research shows that ethnic minorities may have participated in Internet interventions (eg, [59]), randomized controlled trials on the effectiveness of psychotherapy for common mental disorders, such as depression, are still sparse. Ethnic minorities are underrepresented in mental health research and literature about ethnic differences in this field is very small [8]. In our recent meta-analysis, we found only 56 randomized controlled trials on the psychological treatment of depression in adults reporting the proportion of participants from ethnic minorities, of which none of them made distinct comparisons between ethnic populations [15].

It is generally believed that ethnic minorities are less willing to participate in clinical research; however, very small differences between ethnic minorities and majorities are found in the willingness to participate in health research [60]. Other factors, such as higher costs associated with the recruitment, the exclusion criterion of insufficient ability to speak the second (native) language, the shortage of ethnic minority coordinators in trials, and stereotypes and myths, are considered to be important barriers for their participation [61].

Our study shows that recruitment of ethnic minorities is possible when appropriate recruitment strategies are applied. For example, almost 80% of our participants were recruited through the Internet (primarily on Facebook). Traditional media, such as advertisements in newspapers or banners on websites, appear not to be successful recruitment strategies for this target group, although many studies have applied this strategy successfully for recruiting participants for randomized controlled trials and studies in routine practice among the general population [21,62]. Flexibility and privacy of the Internet were the main reasons for respondents to agree to participate. The use of social media in research is a relatively new development, and may potentially prove more effective for recruiting ethnic minorities in research trials. The contact through social media and the visibility of the researcher seemed to lower the threshold for participation in research and for help seeking. Although we did not find significant results from the intervention, the current trial shows that the Internet (1) is an effective way to reach hard-to-reach populations, (2) lowers the threshold to get in contact with a professional, (3) can be an effective recruitment strategy for clinical trials, and (4) is potentially an effective way to deliver cognitive behavioral therapy for ethnic minorities.

Furthermore, our participants consisted primarily of first-generation migrants who had a preference for the Turkish (native) language. Offering the intervention and assessment measures in 2 languages may have been another successful strategy to lower the threshold for study participation. Generally, participants are only included in intervention studies when they can read and speak the language of the country they live in [61].

Another argument for the low-access threshold of our intervention might be found in the large number of applicants with suicidal ideation (30%, 64/221). We had to exclude these applicants (51.2% of excluded group) because they are a high-risk group not suited to our guided self-help intervention. In keeping with the protocol, we referred these individuals to their primary care physician or to the online portal for suicide prevention [35]. These applicants were primarily women (59.4%) with a mean age of 33.5 years (range 18-53).This is a rather high number when compared with those excluded because of suicidal ideation among primarily native population studies [63,64]. The number also appears high when compared to prevalence rates at the population level. For example, in the Netherlands, 8.3% of the Dutch population have ever had suicidal ideation in their lifetime and 2.2% have attempted suicide [65]. As mentioned in the introduction, young Turkish women are at increased risk of committing suicide [7]. It seems that our study has reached a large number of this group. Future Internet research could focus more specifically on this high-risk group. These individuals experience a high burden of disease and unmet needs, but appear to be reachable by Internet. Thus, the delivery of psychotherapy through the Internet appears to be a promising way to target hard-to-reach ethnic minority groups.

This study has several limitations. First, the attrition ratio was high at posttest. We compensated for this high attrition by means of multiple imputation. Attrition was even higher at follow-up; therefore, we decided not to apply multiple imputation for this time point. Instead, we conducted completers-only analyses. High attrition remains a common problem in Internet interventions, with rates of up to 50% [66,67]. Analyses showed no differences in study attrition rates at posttest between the experimental group and control group. However, at follow-up, the experimental group had a higher attrition rate than the control group. Reminders in the form of emails were sent (maximum 5 times per assessment), but this did not result in a low attrition rate for study dropout. Reasons for this high attrition rate are not known; as a result, we can only guess why this happened. It is possible that participants in the experimental group stopped with the trial after finishing the intervention. Filling in the posttreatment questionnaires might not have been regarded as an obligatory part of the trial. Furthermore, the control group was waiting to receive access to the intervention, they could have perceived filling in the questionnaires as an obligation to partake in the intervention.

Second, although we reached a relatively large number of Turkish migrants, recruitment and inclusion were challenging and complex. One of the challenges was to find an appropriate recruitment strategy. In the end, we did find one (ie, Facebook), but we had only limited time left for recruitment because of the overall period available for this study. This could be one of the main reasons for not having been able to obtain the required sample size. Another problem was that we had to exclude most of the eligible participants for reasons such as high suicidality risk. Thus, we did not achieve the target sample size (N=200) during the study period, which may have resulted in an underpowered study. In turn, this may have been the reason that significant effects were not detected and it limited the generalizability of our results.

Third, our target population focused on the online population because participants were required to have access to the Internet and an email address to be included in our trial. Moreover, it is known that almost 80% of the Turkish population in the Netherlands has Internet access [68]. However, our population may have differed from most Turkish people in another way. When we look at the demographic characteristics of our sample, we notice that younger women (mean 35.2 years) with a middle to higher educational level (70%) took part. This is a higher proportion than the Turkish population in the Netherlands, of which 30.1% had at least a middle educational level [69]. Our sample conforms to the sample characteristics of nonmigrant populations in similar trials, in which women (aged 35-55 years) with higher educational levels have taken part [20,21].

Furthermore, although participants were required to have access to a computer with Internet and have an email address, we did not assess the reading and Internet comfort level of the participants. Given that most of our respondents were recruited through Facebook, we assume that respondents at least were able to understand our intervention and questionnaires. Next, respondents could choose the language in which they wanted to follow the intervention and answer the questionnaires (Turkish or Dutch).

Finally, we used only self-report assessments to measure the severity of depressive symptoms in participants. We used self-report on purpose because we wanted to keep the access barrier for study participation as low as possible. Diagnostic interviews are an extra burden for participants and it is not yet possible to conduct them through the Internet. Therefore, our study lacks a diagnosis of depression in the study participants. Research has shown that online self-report questionnaires have good validity (eg, [41]) and yield scores equivalent to paper-and-pencil questionnaires (eg, [70]).

The results of this study have promising implications for the clinical field. Our study is one of the first to assess the clinical effectiveness of guided self-help interventions by Internet for Turkish migrants with depressive complaints. In addition, the guided self-help intervention for Turkish migrants in the Netherlands could also be suitable for Turkish populations in other EU countries or in Turkey itself, where guided self-help is not yet common practice. It may be a welcome intervention both for clinicians and for minorities because there is a lack of evidence-based culturally sensitive psychotherapy for ethnic minorities and there is a high threshold to these services.

Future research should replicate our findings with adequately powered samples for posttest and follow-up measurement to assess the clinical effectiveness in a robust manner. Future research should focus on monitoring participants who drop out prematurely from the study at follow-ups to evaluate the reasons for withdrawal. It is also important to evaluate the impact of culturally sensitive components in Internet interventions for ethnic minority populations with depression.

The results of this study did not show a significant effect on the reduction of depressive symptoms. However, the effect size at posttest was high, which might be an indicator of the possible effectiveness of the intervention when assessed in a larger sample and robust trial. Future research should replicate our study with adequately powered samples.