Asian American and Pacific Islander (AAPI) persons are disproportionately affected by diabetes, with a higher prevalence of type 2 diabetes compared to White persons, even when individuals have similar BMI or waist circumference [1]. The high burden of disease becomes even more apparent when examining disaggregated diabetes prevalence across AAPI ethnic groups [1]. For example, in the United States, Filipino men have one of the highest type 2 diabetes rates among all ethnic groups (15.8%), and Korean women are 5 times more likely to be affected by type 2 diabetes compared to non-Hispanic White women [2].

As recommended by the American Diabetes Association, lifestyle and educational interventions are key components of diabetes care and empower patients to make informed health decisions in coping with their disease [3]. Such interventions have been shown to be highly effective in promoting diabetes knowledge and self-care behaviors. Diabetes self-management education, for example, has been shown to reduce hemoglobin A_1c (HbA_1c) in racial or ethnic minority populations [4]. However, only 3 reviews of diabetes interventions have focused on AAPI populations, and they have been limited in terms of geography, heterogeneity, and AAPI subpopulations included [5-7]. This systematic review and meta-analysis focused on AAPI persons with type 2 diabetes in the United States and aimed to broadly investigate the effect of nonpharmacological interventions on HbA_1c outcomes.

This review was registered with the International PROSPERO (CRD42019122625) database and has been reported in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines (Table S1 in Multimedia Appendix 1).

Database searches returned 111,289 results. After removing duplicate results and screening titles and abstracts, 1835 articles remained for full-text review. Ultimately, 9 RCTs met all inclusion criteria (Figure 1) [10-18]. These enrolled a total of 1492 patients who were primarily aged between 40 and 60 years (Table 1); 5 RCTs included all AAPI patients. Specific races or ethnicities included Filipino, Samoan, Hawaiian, Micronesian, Bangladeshi, and Korean. Two RCTs only assessed whether patients identified as Asian or Pacific Islander.

Most RCTs (n=7) included a component of diabetes self-management education (DSME) in the intervention arm. Other strategies consisted of bilingual counseling (n=4), glucose or weight self-monitoring (n=3), motivational interviewing (n=1), and financial support (n=1). Approximately all (n=8) studies modified some element of their intervention toward the specific cultures of their study populations. Per the National Institute on Minority Health and Health Disparities Research Framework, studies were conducted at the individual (n=9) and interpersonal levels of influence (n=5), targeting primarily behavioral (n=9) and sociocultural domains (n=8). For controls, 6 RCTs used waitlist patients, 2 used a minimal intensity intervention, and 1 used usual care.

Overall, individuals were followed for a median of 6 (range 3‐12) months. The interventions on average resulted in a −0.39% (95% CI −0.64 to −0.14%; I²=65%; Figure 2; Table 2) decrease in HbA_1c. A similar pattern was observed with subgroup analyses for Asian individuals (n=5), with an average reduction in HbA_1c by −0.43% (95% CI −0.66 to −0.20%; I²=43%) and for Pacific Islander individuals (n=3), with a decrease of −0.62 % (95% CI −0.94 to −0.29%; I²=0%). For individuals who preferred non-English language (n=5), these interventions on average decreased HbA_1c levels by −0.47% (95% CI −0.70% to −0.23%; I²=45%). The most effective interventions specifically described using culturally tailored and/or language-tailored modalities of DSME.

Using the Cochrane Collaboration risk of bias for randomized trials tool, 7 of the 9 studies were determined to be at low risk of bias (Table S4 in Multimedia Appendix 1). One study [12] had some concerns in the domain of deviations from the intended intervention, and another study [17] was at high risk for bias in the domain of missing outcome data. Publication bias was evaluated visually by plotting changes in HbA_1c by study on a funnel plot, where the y-axis represents study size and the x-axis the size of effect found (Figure S1 in Multimedia Appendix 1). Of the 9 RCTs, 8 fell within the expected 95%, and approximate symmetry was achieved with the exception of the single study, whose intervention relied solely on financially incentivizing diabetes self-management. Additionally, 1 study with a small sample size (n=47) and relatively large HbA_1c SD at baseline affected the vertical distribution. The Egger test indicated no evidence of publication bias (P=.22). Quality of evidence was high by the GRADE (Grading of Recommendations, Assessment, Development, and Evaluation) approach (Table S5 in Multimedia Appendix 1).

We investigated the efficacy of nonpharmacologic interventions for type 2 diabetes in lowering HbA_1c in Asian American and Pacific Islander individuals. Our systematic review and meta-analysis identified a substantive reduction of HbA_1c, with an average change of −0.39% over 6 months. Most interventions incorporated components of DSME and social engagement such as counseling or other interpersonal interactions. Furthermore, nearly all of the interventions tailored their approach based on the languages and cultures of their target population, often collaborating with community health workers or community-based organizations.

Identifying evidence-based ways to enhance diabetes management is essential for populations at greater risk for type 2 diabetes development and progression, including AAPI persons. AAPI populations have experienced alarming rises in diabetes; for instance, one study found a 68% increase in incidence between 1993 and 2001, the highest among all racial or ethnic groups [19]. Consequently, the American Diabetes Association has incorporated race or ethnicity into adult screening guidelines, and people with AAPI race or ethnicity are considered a “high-risk” population [20]. Furthermore, while AAPI race or ethnicity is comprised of many heterogeneous populations, current practices typically aggregate them as “Asian” or “Other,” when in reality over 40 ethnic subgroups fall under this categorization [21]. Encouragingly, in this review, few studies reported AAPI as a single group, and a broad range of AAPI subpopulations were represented (Filipino, Samoan, Hawaiian, Micronesian, Bangladeshi, and Korean). Heterogeneity was moderately high for overall results but lower in subgroup analyses. Data aggregation across AAPI subpopulations can mask higher-risk subpopulations and negatively impact screening, diagnosis, and treatment. For example, the DISTANCE study, which delineated AAPI subgroups when estimating diabetes rates, found that Pacific Islander, South Asian, and Filipino individuals had the highest prevalence of diabetes (18.3%, 15.9%, and 16.1%, respectively) across all racial or ethnic categories [22]. They also detected substantial differences in the rates of diabetes-related complications across AAPI subgroups [22].

AAPI persons constitute 7.7% of the US population and are one of the fastest growing racial or ethnic groups. Nonetheless, research funding for studies focusing on AAPI populations has not progressed over the past several decades. From 1986 to 2000, AAPI was represented in 0.2% of health-related grants across 7 federal agencies [23]. Between 1992 and 2018, 0.2% of the National Institutes of Health budget was allocated to 529 AAPI-related clinical research projects [24]. In the same time frame, inclusion of Native Hawaiians and Pacific Islanders in extramural National Institutes of Health–defined phase 3 clinical trials decreased from 0.3% (n=1011) to 0.2% (n=271) [24]. Given the small and shrinking number of RCTs among people with AAPI race or ethnicity, additional RCTs are needed to further our understanding of optimal strategies to manage type 2 diabetes in people with AAPI race or ethnicity.

Nonpharmacological interventions can meaningfully improve prevention and management of diabetes and have been found to be cost-effective; however, underrepresentation of AAPI persons in trials reduces generalizability [25-28]. For instance, the landmark Look Action for Health in Diabetes trial showed that intensive lifestyle interventions improved glycemic control and cardiovascular outcomes [29]. Although the Look Action for Health in Diabetes trial aimed to recruit at least a third of participants from racial and ethnic minority groups, there were only 50 (<2%) AAPI participants [30]. A meta-analysis of 16 randomized lifestyle intervention trials in general populations with type 2 diabetes found an average change of −0.37% in HbA_1c [31]. While this meta-analysis did not report stratified results or primarily enroll AAPI populations, their reduction in HbA_1c is very similar to our findings (−0.39%), providing some reassurance in the generalizability of larger studies.

Given the tremendous diversity of genetics, culture, diet, and lifestyle within the AAPI category, the effectiveness of specific interventions may vary by population. With our limited sample size, we were unable to directly compare types of interventions. However, we identified DSME as the most common shared component in these studies, which is one form of intervention that is strongly supported in the literature. In a systematic review of 118 randomized trials for adults with type 2 diabetes, DSME resulted in a −0.74% reduction in HbA_1c [32]. Furthermore, the study found that combining individual and group engagement resulted in greater reduction in HbA_1c. However, studies that aggregate data at this level are unable to delineate racial or ethnic differences. Recently, different meta-analyses have demonstrated that nonpharmacologic interventions such as DSME may differ in effectiveness based on the population in focus. For example, in 2 meta-analyses, DSME interventions were shown to reduce HbA_1c (−0.24% and −0.25%) in Latino individuals with type 2 diabetes [33,34]. However, in another meta-analysis, DSME did not reduce HbA_1c in African American persons [35]. Mixed results have been reported in other meta-analyses, in which lifestyle interventions were effective in lowering HbA_1c for White individuals but not Black or African individuals or Hispanic populations [36]. Another systematic review found that lifestyle intervention reduced the incidence of type 2 diabetes more in the “(predominantly) White” group compared to the Asian group [37]. Moreover, beyond HbA_1c, significant differences may exist in the impact on 2-hour glucose, BMI, and waist circumference depending on the racial or ethnic subgroup [38]. Taken together, these studies suggest that a uniform approach for nonpharmacologic management of type 2 diabetes would not lead to optimal health outcomes or reduction in health disparities.

In a report by the National Heart, Blood, and Lung Institute on improving cardiometabolic risk factors in Asian American individuals, the committee recommended more research to understand the social and cultural contexts of lifestyle interventions, perceptions, and barriers to addressing such risk factors [39]. This notion has taken root in the United States—for example, several studies focusing on South Asian individuals, such as the South Asian Healthy Lifestyle Intervention, use culturally tailored strategies to reduce cardiovascular risk factors [40]. Reviewing lifestyle interventions in South Asian individuals found that tailoring interventions to linguistic and cultural norms such as modifying traditional Indian cooking and incorporating physical activities from South Asia were effective in lowering HbA_1c [41].

In our study, 8 [10,11,13-18] of the 9 studies culturally tailored their intervention prior to implementation—for instance, Ing et al [13] in Hawaii created their intervention by partnering with local organizations and based their educational content on prior focus groups created in the community, highlighting relevant images, foods, and physical activities in Hawaii. In other studies, community health workers, who often are trusted members of the community, were employed to effectively carry out interventions in the study population. Overall, these studies highlight the potential benefits of actively partnering within and beyond traditional public health institutions, including community organizations and services, to bring culturally and linguistically appropriate care. Given the financial and logistical barriers to providing these more customized interventions to patients, telemedicine may be a promising approach. In the studies we reviewed, telephone counseling and remote blood glucose monitoring were strategies that proved effective in reducing HbA_1c. Results of recent meta-analyses also support the effectiveness and potential cost-saving of telehealth and digital interventions for improving glycemic control [42-47]. Therefore, culturally adapting interventions, collaborating with community organizations, and using telemedicine and digital health tools are all possible strategies that health care systems can incorporate to improve healthy behaviors and outcomes for type 2 diabetes in AAPI communities.

This study has several limitations. First, a relatively small number of trials met all inclusion criteria for this meta-analysis. In part, this reflects a greater need for clinical trials to increase enrollment across underrepresented races or ethnicities, including those of AAPI backgrounds. While not feasible to conduct comparable studies by individual subgroup, stratifying AAPI populations by a limited number of diabetes and cardiovascular risk groups may prove useful to guide screening and management. Additionally, as with any meta-analysis, the dissimilarity of interventions makes differences in effect on HbA_1c difficult to compare. This is particularly true with nonpharmacologic interventions, given that each study developed a unique intervention for their populations. A larger sample size and further exploration of heterogeneity related to intervention approaches and study populations would be useful for understanding the generalizability of findings. Finally, it is possible that there are intervention studies published in the gray literature or indexed in other databases that were not included in this analysis.

In summary, this meta-analysis examined 9 nonpharmacologic interventions for type 2 diabetes for primarily AAPI populations. Random-effects meta-analysis found that such interventions substantively lowered HbA_1c, offering support to leverage these interventions in clinical practice and community efforts. Successful interventions were culturally tailored and often used diabetes self-management education. More RCTs are needed to add to the small body of literature of effective nonpharmacologic interventions for AAPI populations with type 2 diabetes. Additionally, future studies are required to evaluate how other markers of cardiovascular risk and systemic health can be improved in these populations. Developing evidence-based approaches to managing type 2 diabetes in AAPI populations will further reduce the burden of diabetes in the United States and enhance quality of care.