PLoS Med plos plosmed PLOS Medicine 1549-1277 1549-1676 Public Library of Science San Francisco, CA USA 10.1371/journal.pmed.1003434 PMEDICINE-D-20-01702 Research Article Social sciencesEconomicsEconomic geographyLow and middle income countries Earth sciencesGeographyEconomic geographyLow and middle income countries Research and analysis methodsMathematical and statistical techniquesStatistical methodsMetaanalysis Physical sciencesMathematicsStatisticsStatistical methodsMetaanalysis Medicine and health sciencesEndocrinologyEndocrine disordersDiabetes mellitus Medicine and health sciencesMedical conditionsMetabolic disordersDiabetes mellitus Medicine and health sciencesDiagnostic medicineDiabetes diagnosis and managementHbA1c Biology and life sciencesBiochemistryProteinsHemoglobinHbA1c Medicine and health sciencesEndocrinologyEndocrine disordersDiabetes mellitusType 2 diabetes Medicine and health sciencesMedical conditionsMetabolic disordersDiabetes mellitusType 2 diabetes Medicine and health sciencesClinical medicineClinical trialsRandomized controlled trials Medicine and health sciencesPharmacologyDrug research and developmentClinical trialsRandomized controlled trials Research and analysis methodsClinical trialsRandomized controlled trials Medicine and health sciencesHealth careQuality of life Medicine and health sciencesEndocrinologyEndocrine disordersDiabetes mellitusType 2 diabetesType 2 diabetes risk Medicine and health sciencesMedical conditionsMetabolic disordersDiabetes mellitusType 2 diabetesType 2 diabetes risk Health system interventions for adults with type 2 diabetes in low- and middle-income countries: A systematic review and meta-analysis Systematic review of health system interventions for type 2 diabetes in low- and middle-income countries https://orcid.org/0000-0002-4372-7387 Flood David Conceptualization Data curation Formal analysis Investigation Methodology Project administration Resources Software Supervision Visualization Writing – original draft Writing – review & editing 1 2 * https://orcid.org/0000-0002-1957-6128 Hane Jessica Data curation Formal analysis Methodology Project administration Writing – review & editing 3 Dunn Matthew Data curation Formal analysis Project administration Visualization Writing – review & editing 4 https://orcid.org/0000-0001-7699-4417 Brown Sarah Jane Data curation Methodology Resources Software Writing – review & editing 5 https://orcid.org/0000-0002-3351-7175 Wagenaar Bradley H. Conceptualization Formal analysis Methodology Writing – review & editing 6 7 https://orcid.org/0000-0002-4819-7178 Rogers Elizabeth A. Conceptualization Formal analysis Methodology Writing – review & editing 8 9 Heisler Michele Formal analysis Investigation Methodology Writing – review & editing 10 11 12 https://orcid.org/0000-0001-7274-8315 Rohloff Peter Conceptualization Formal analysis Investigation Methodology Supervision Writing – original draft Writing – review & editing 1 https://orcid.org/0000-0001-8670-9376 Chopra Vineet Data curation Formal analysis Investigation Methodology Project administration Supervision Writing – original draft Writing – review & editing 12 13 Center for Research in Indigenous Health, Wuqu’ Kawoq, Tecpán, Guatemala Division of Hospital Medicine, Department of Internal Medicine, National Clinician Scholars Program, University of Michigan, Ann Arbor, Michigan, United States of America Medicine-Pediatrics Residency Program, University of Minnesota, Minneapolis, Minnesota, United States of America School of Public Health, University of Michigan, Ann Arbor, Michigan, United States of America Health Sciences Libraries, University of Minnesota, Minneapolis, Minnesota, United States of America Department of Global Health, University of Washington, Seattle, Washington, United States of America Department of Epidemiology, University of Washington, Seattle, Washington, United States of America Division of General Internal Medicine, Department of Medicine, University of Minnesota, Minneapolis, Minnesota, United States of America Department of Pediatrics, University of Minnesota, Minneapolis, Minnesota, United States of America Department of Internal Medicine, University of Michigan, Ann Arbor, Michigan United States of America Institute for Healthcare Policy and Innovation, University of Michigan, Ann Arbor, Michigan United States of America Center for Clinical Management Research, Veterans Affairs Ann Arbor Healthcare System, Ann Arbor, Michigan United States of America Division of Hospital Medicine, Department of Medicine, University of Michigan, Ann Arbor, Michigan United States of America Kengne Andre P. Academic Editor South African Medical Research Council, SOUTH AFRICA

The authors have declared that no competing interests exist.

 * E-mail: david@wuqukawoq.org 12 11 2020 11 2020 17 11 e1003434 28 4 2020 19 10 2020 This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication. Background

Effective health system interventions may help address the disproportionate burden of diabetes in low- and middle-income countries (LMICs). We assessed the impact of health system interventions to improve outcomes for adults with type 2 diabetes in LMICs.

 Methods and findings

We searched Ovid MEDLINE, Cochrane Library, EMBASE, African Index Medicus, LILACS, and Global Index Medicus from inception of each database through February 24, 2020. We included randomized controlled trials (RCTs) of health system interventions targeting adults with type 2 diabetes in LMICs. Eligible studies reported at least 1 of the following outcomes: glycemic change, mortality, quality of life, or cost-effectiveness. We conducted a meta-analysis for the glycemic outcome of hemoglobin A1c (HbA1c). GRADE and Cochrane Effective Practice and Organisation of Care methods were used to assess risk of bias for the glycemic outcome and to prepare a summary of findings table. Of the 12,921 references identified in searches, we included 39 studies in the narrative review of which 19 were cluster RCTs and 20 were individual RCTs. The greatest number of studies were conducted in the East Asia and Pacific region (n = 20) followed by South Asia (n = 7). There were 21,080 total participants enrolled across included studies and 10,060 total participants in the meta-analysis of HbA1c when accounting for the design effect of cluster RCTs. Non-glycemic outcomes of mortality, health-related quality of life, and cost-effectiveness had sparse data availability that precluded quantitative pooling. In the meta-analysis of HbA1c from 35 of the included studies, the mean difference was −0.46% (95% CI −0.60% to −0.31%, I2 87.8%, p < 0.001) overall, −0.37% (95% CI −0.64% to −0.10%, I2 60.0%, n = 7, p = 0.020) in multicomponent clinic-based interventions, −0.87% (−1.20% to −0.53%, I2 91.0%, n = 13, p < 0.001) in pharmacist task-sharing studies, and −0.27% (−0.50% to −0.04%, I2 64.1%, n = 7, p = 0.010) in trials of diabetes education or support alone. Other types of interventions had few included studies. Eight studies were at low risk of bias for the summary assessment of glycemic control, 15 studies were at unclear risk, and 16 studies were at high risk. The certainty of evidence for glycemic control by subgroup was moderate for multicomponent clinic-based interventions but was low or very low for other intervention types. Limitations include the lack of consensus definitions for health system interventions, differences in the quality of underlying studies, and sparse data availability for non-glycemic outcomes.

 Conclusions

In this meta-analysis, we found that health system interventions for type 2 diabetes may be effective in improving glycemic control in LMICs, but few studies are available from rural areas or low- or lower-middle-income countries. Multicomponent clinic-based interventions had the strongest evidence for glycemic benefit among intervention types. Further research is needed to assess non-glycemic outcomes and to study implementation in rural and low-income settings.

In this meta-analysis of published studies, David Flood and colleagues assess the impact of health system interventions to improve outcomes for type 2 diabetes patients.

 Author summary Why was this study done?

Approximately 80% of the 463 million adults with type 2 diabetes worldwide live in low- and middle-income countries (LMICs).

Evidence-based treatments for diabetes exist, but health systems in LMICs have difficulty meeting diabetes patients’ needs.

Health system interventions can help address this gap by improving the delivery of diabetes care within health systems.

 What did the researchers do and find?

We conducted a systematic review and meta-analysis of 39 health system interventions aiming to improve outcomes of glycemic (i.e., blood glucose) control, mortality, quality of life, or cost-effectiveness for people with type 2 diabetes in LMICs.

We found that health system interventions for type 2 diabetes may be effective in improving glycemic control in LMICs, but few studies were available from rural areas or low- or lower-middle-income countries.

Among intervention types, multicomponent clinic-based interventions had the strongest evidence for improving glycemic control.

 What do these findings mean?

Our findings support the scaling up of diabetes health system interventions to improve patients’ glycemic control in LMICs.

Further research is needed to assess other outcomes beyond glycemic control, especially in rural areas and in low- or lower-middle-income countries.

 http://dx.doi.org/10.13039/100000062 National Institute of Diabetes and Digestive and Kidney Diseases K23DK118207 https://orcid.org/0000-0002-4819-7178 Rogers Elizabeth A. DF is supported by the National Clinician Scholars Program at the University of Michigan Institute for Healthcare Policy & Innovation. BHW is supported by grant number K01MH110599 from the National Institute of Mental Health. EAR is supported by the National Institute of Diabetes and Digestive and Kidney Diseases of the National Institutes of Health under award number K23DK118207. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data Availability The study’s dataset and statistical code are available through Dataverse at: https://doi.org/10.7910/DVN/NIESKT. Introduction

Type 2 diabetes disproportionately affects people in low- and middle-income countries (LMICs). Of the estimated 463 million adults worldwide with type 2 diabetes, approximately 80% reside in LMICs [1]. The absolute number of adults and percentage of the population with diabetes have increased more quickly in LMICs than in high-income countries (HICs) [2]. Despite the existence of cost-effective and evidence-based clinical treatments for type 2 diabetes [3], health systems in LMICs have difficulty meeting the rising need for quality care [4]. Improving and scaling up care in LMICs is an urgent global health priority.

Health system interventions can help address this priority. In contrast to clinical therapies for individual patients, health system interventions emphasize the behavior of health workers and the way healthcare is practiced and delivered [5]. Examples of health system interventions include quality and safety programs, health information systems, health worker incentives, and changes in scope of practice [5]. Effective health system interventions are needed to implement type 2 diabetes care in settings with different resources, cultures, and population risk factors [6].

While health system interventions improve type 2 diabetes outcomes in HICs [7–9], the evidence from LMICs is limited. A 2012 meta-analysis of 142 randomized trials primarily conducted in HICs found that interventions targeting the health system rather than healthcare providers or patients alone were most effective [9]. However, health system interventions designed and tested in HICs may not be generalizable to LMICs [10]. In LMICs, prior reviews draw from diverse study designs and together suggest a modest yet increasing number of studies on the implementation of evidence-based type 2 diabetes care into health systems in LMICs [11–13]. To our knowledge, no review has systematically assessed evidence from randomized controlled trials (RCTs) or conducted a meta-analysis.

Therefore, we conducted a systematic review and meta-analysis to examine the impact of health system interventions that aimed to improve outcomes of glycemic (i.e., blood glucose) change, mortality, health-related quality of life, or cost-effectiveness for adults with type 2 diabetes in LMICs.

 Methods

This systematic review and meta-analysis was conducted based on guidance from Cochrane Effective Practice and Organisation of Care (EPOC), a group focusing on reviews of the delivery of health services [14]. We registered the review in PROSPERO (CRD42018106765; S1 Appendix) and followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines (S2 Appendix) [15]. Ethical approval was not required as the research used publicly available data.

Search strategy and selection criteria

We performed systematic searches in several bibliographic databases. The search strategy was built and tested for sensitivity in Ovid MEDLINE (S3 Appendix) and translated to 5 other bibliographic databases: Cochrane Library, EMBASE, African Index Medicus, LILACS, and Global Index Medicus. Databases were chosen to be inclusive of international and interdisciplinary literature. The search strategy was built in English, and no language filters were applied. We also hand-searched the references of included studies, related systematic reviews, and the websites of major international diabetes organizations. To ensure high search quality, a second reference librarian peer-reviewed the search terms. The search dates were from database inception through February 24, 2020.

We included RCTs of health system interventions targeting non-pregnant, ambulatory adults with type 2 diabetes in LMICs. We defined LMICs using the 2019 World Bank income groups. Included studies reported at least 1 of the following outcomes: glycemic change, mortality, health-related quality of life, or cost-effectiveness. Given our interest in durable health system interventions, we prespecified that studies enroll 100 or more participants, with follow-up of at least 24 weeks. No date or language restrictions were applied.

We used the EPOC review group’s definition of health system interventions as those designed to “improve the professional practice and the delivery of effective health services” through changes in healthcare delivery, financing, governance, and implementation [5,14]. Consistent with EPOC, we excluded studies of patient behavior change alone if the intervention did not primarily target healthcare professionals [14]. For example, an intervention training healthcare professionals on diabetes education was included; however, an intervention aiming to improve outcomes solely through individualized diabetes education was excluded [14]. We defined “healthcare professional” broadly to encompass physicians, nurses, pharmacists, and other allied health workers.

 Data analysis

A medical librarian (SJB) downloaded all records, removed duplicates, and imported records to the review management tool Covidence. Two authors (DF and JH) independently screened studies by title and abstract and, subsequently, by full-text review. Disagreements were resolved first by consensus and, if needed, in consultation with another author (PR). We used language proficiency among the members of our review team and Google Translate to abstract data from non-English trials [16]. Multiple reports from the same study were identified by reviewing the country setting, intervention details, and authorship list. When multiple reports were identified, we linked the reports together for extraction and analysis. We used the TIDieR checklist and EPOC template to structure extraction [17]. We extracted study elements including the 4 outcomes, country, setting, duration and follow-up, number of participants enrolled, intervention description, and comparator. One author (DF) extracted summary data into a customized electronic spreadsheet, and 2 other authors (JH and MD) independently verified the extracted data. We classified each study by EPOC domain (S4 Appendix) [5] and then grouped interventions into similar types. Our main unit of analysis was at the level of intervention type. If outcomes were missing or not reported, we contacted authors twice to obtain data. We used GRADE and EPOC guidance to assess risk of bias for the glycemic outcome and to prepare a summary of findings table [18–21].

 Statistical analysis

As quantitative data were reliably reported for only 1 of our 4 included outcomes, we limited our meta-analysis to the glycemic outcome of hemoglobin A1c (HbA1c) change. The meta-analysis was performed with random effects using the DerSimonian–Laird method for mean between-group HbA1c difference. Prespecified subgroup analyses were done by intervention type. Sample sizes for cluster RCTs were adjusted to account for the design effect using the intracluster correlation coefficient (ICC) [22]. We inferred an ICC from the literature if one was not reported in the study or its trial protocol [23]. We followed the methodology recommended in the Cochrane handbook to calculate within-group mean and standard deviation when this information was not directly reported in the study or made available by authors [22]. To provide a range of the effects of individual studies, we calculated an overall prediction interval [24].

We conducted 2 sensitivity analyses. First, we excluded studies with high risk of bias. Second, we assessed the influence of individual studies by using the leave-one-out method to recalculate estimates omitting 1 study at a time [25]. Heterogeneity was explored by calculating I2 and T2, and we report 95% confidence intervals for I2 if 3 or more studies are pooled [26]. Publication bias was assessed by visual inspection of funnel plots and the Egger test. The trim-and-fill method was also applied to impute the number of studies potentially missing from the meta-analysis and to re-estimate an overall effect size accounting for publication bias [27]. We analyzed data in Stata (version 16.0).

 Results Overview of results

Our search strategy identified 12,921 references (Fig 1). After removing 1,093 duplicates, we screened 11,828 references by title and abstract and assessed 322 full-text articles for eligibility. Of the 283 articles excluded after full-text review, 103 articles were excluded due to the type of intervention, and 94 articles were excluded due to incomplete data. We included 39 trials in the narrative review and 35 trials in the meta-analysis of glycemic change.

10.1371/journal.pmed.1003434.g001 PRISMA study flow diagram.

HbA1c, hemoglobin A1c.

Of the 39 studies included in the narrative review, 19 were cluster RCTs and 20 were individual RCTs (Table 1; S5 Appendix). There were 21,080 total participants enrolled across included studies. The greatest number of studies were conducted in the East Asia and Pacific region, followed by South Asia. Twenty-nine studies were conducted in upper-middle-income countries as defined by the World Bank, and only 1 trial included a site in a low-income country. All but 1 of the studies were published in the year 2010 or after [28]. The study setting was primarily urban in 27 trials and primarily rural in 5 trials. The median study duration was 10 months (interquartile range 6 to 12). Most interventions involved the EPOC domains of delivery arrangements and implementation strategies. Only 1 intervention incorporated a change in governance [29], and no study tested changes in financial arrangements. The comparator group in most studies was usual care as defined by the local program or setting of care. Two studies described the comparator group as enhanced usual care, where the enhancement consisted of clinical training for health professionals [30,31], and in 1 study the medical fees were waived in the comparator arm [32].

10.1371/journal.pmed.1003434.t001 Characteristics of the 39 studies included in this review.
Characteristic Number of studies (n = 39) References
EPOC domains
    Delivery arrangements 27 [28,32–57]
    Delivery arrangements and implementation strategies 9 [30,31,58–64]
    Implementation strategies 2 [65,66]
    Delivery arrangements, governance arrangements, implementation strategies 1 [29]
Intervention type
    Multicomponent clinic-based 8 [30,31,58–63]
    Pharmacist task sharing 14 [28,33–35,37,38,41,42,47,48,50,51,55,57]
    Diabetes education or support alone 9 [32,39,40,43,46,49,53,54,56]
    Case management by nurses 2 [36,52]
    Physician clinical training alone 2 [65,66]
    Nurse task sharing 1 [29]
    mHealth screening and quality improvement 1 [64]
    Internet-based glucose telemonitoring alone 2 [44,45]
Study design
    Individual RCT 20 [28,33–35,37–39,41,42,44,47,48,50–53,55,58,59,61]
    Cluster RCT 19 [29–32,36,40,43,45,46,49,54,56,57,60,62–66]
World Bank region*
    East Asia and Pacific 20 [28,32,34–37,39,40,44,45,49,52–54,57,61–63,66]
    South Asia 7 [30,31,38,42,43,51,58]
    Latin America and Caribbean 4 [47,48,59,60]
    Sub-Saharan Africa 4 [29,33,46,53]
    Middle East and North Africa 4 [41,50,55,64]
    Europe and Central Asia 1 [65]
World Bank income group*
    Low 1 [53]
    Lower middle 11 [30,31,33,38,42,43,49,51,53,58,66]
    Upper middle 29 [28,29,32,34–37,39–41,44–48,50,52,54–57,59–66]
Setting
    Mostly rural 5 [30,31,36,49,64]
    Mostly urban 27 [29,32,33,35,37–42,44,46,48,50,52,54–62,65,66]
    Mixed 4 [43,45,53,63]
    Not reported 3 [28,34,47]
Outcome reported
    Mortality 19 [29–31,34,36,39,42,44,45,49,51–53,57–59,61,63,66]
    Health-related quality of life 11 [29,32,33,40,45,46,51,52,58,59,62]
    Cost-effectiveness 5 [33,36,46,58,59]
    Change in glycemic control 39 All included studies

*The studies by Van Olmen et al. [53] and Reutens et al. [66] are counted twice, as they were conducted in multiple countries of different World Bank regions and income groups.

EPOC, Effective Practice and Organisation of Care; RCT, randomized controlled trial.

 Narrative description of interventions Multicomponent clinic-based interventions

Eight trials were classified as clinic-based multicomponent interventions, which we defined as studies involving multiple types of health workers implementing a bundle of quality improvement or health system strengthening interventions [30,31,58–63]. Most studies incorporated primary care doctors in a team-based intervention [30,31,59,60,62,63], and the study by Ali and colleagues incorporated endocrinologists [58]. Each intervention included self-management education or support delivered by peers [59], non-physician care coordinators [58], clinicians [30,59,60,62,63], or an automated short message service (SMS) text-messaging system [31]. Other components in the bundles included health record establishment [61], electronic decision support [31,58], physician education [30,59,60,62,63], care coordination or case management [58,62], clinical information systems [30,60,62,63], and clinical audit and feedback [62,63]. Three studies were based on the Chronic Care Model [60,62,63].

 Pharmacist task-sharing interventions

Fourteen studies were classified as pharmacist task-sharing interventions, which we defined as studies in which patients received activities performed by pharmacists such as care coordination, medication review and counseling, and prescription suggestions to physicians [28,33–35,37,38,41,42,47,48,50,51,55,57]. All pharmacist task-sharing interventions incorporated diabetes self-management education, but no trial included education alone. Seven interventions included counseling and reminders through telephone calls or text messages [34,35,38,39,41,55,57], and a study in Iran incorporated only telephone calls with no face-to-face encounters [50]. No intervention involved pharmacists independently prescribing or titrating medications. All 11 of the studies that described the study setting were conducted in an urban area [33,35,37,38,41,42,48,50,51,55,57]. The intensity of interventions was incompletely reported but ranged from 3 to 16 telephone calls or face-to-face visits.

 Diabetes education or support alone

Nine studies involved health workers primarily implementing diabetes education or support without additional services [32,39,40,43,46,49,53,54,56]. We defined these interventions as diabetes education or support alone. Six of the studies primarily involved in-person delivery [32,40,43,46,49,56], and 3 studies delivered the intervention in group format [46,49,56]. The health workers in these studies varied between and within studies and included peers [49,53,56], community health workers [43,46,53], nurses [32,39,40,53,54], psychologists [32], and physicians [32,39,54]. Face-to-face encounters were supplemented with telephone calls in 2 studies [32,54] and by computer-assisted instruction in another study [40]. Motivational interviewing techniques were incorporated in 2 studies [32,46]. The intensity of the in-person interventions ranged from 4 to 24 total face-to-face encounters.

 Other intervention types with fewer studies

Two studies involved nursing case management interventions. In these trials, a nurse [52] or nurse–community health worker team [36] facilitated patient support and care coordination. Both trials varied intervention intensity by a patient’s risk factors. DePue and colleagues conducted a cluster RCT in American Samoa that primarily used home visits and individual rather than group sessions [36]. In the trial conducted by Tutino and colleagues in China, both the intervention and comparator arms included implementation of a web-based clinical information portal, and the intervention arm received additional nurse-led care coordination [52].

Two health system interventions involved physician clinical training alone [65,66]. Akturan and colleagues trained physicians on a therapeutic interviewing technique [65]. Reutens and colleagues trained physicians in multiple countries on diabetes guidelines using 2 in-person sessions and reminders [66].

One study was classified as a nurse task-sharing intervention [29]. Conducted in South Africa, this intervention involved authorizing, training, and supporting nurses to independently prescribe a set of drugs for several noncommunicable diseases including diabetes using an algorithmic management tool [29].

One study was classified as an mHealth screening and quality improvement intervention [64]. This trial involved an intervention for diabetes and hypertension involving SMS educational messages and appointment reminders, community-based screening, and deployment of electronic clinical tools for physicians and nurses [64].

Two studies tested glucose telemonitoring interventions in which participants uploaded glucose data to an online system and then received feedback from health workers regarding treatment changes to improve glycemic control [44,45].

 Summary of outcomes

We describe outcomes of glycemic change, mortality, quality of life, and cost-effectiveness by study in S5 Appendix. Glycemic changes were reported based on HbA1c values in 36 studies and based on fasting glucose alone in 3 trials. Among studies reporting fasting glucose only, 2 trials of multicomponent clinic-based interventions found glycemic improvement [56,61], while there was no improvement in a trial of diabetes education or support alone [43]. The primary outcome involved change in HbA1c or the proportion of participants meeting HbA1c goals in 23 studies.

Outcomes of mortality, health-related quality of life, and cost-effectiveness were reported in 19, 11, and 5 studies, respectively. Of the 19 studies reporting mortality, 14 studies had 10 or fewer deaths combined in the intervention and comparator groups (S6 Appendix). Studies with larger numbers of deaths appeared to have generally similar mortality between trial arms though a formal meta-analysis was not conducted due to sparseness of data [29,31,53,58]. No study’s primary outcome was mortality.

Of the 11 studies reporting quality of life, 6 studies reported no significant differences between the intervention and comparator arms [29,32,45,46,52,59], and 5 studies showed improved quality of life in the intervention arm [33,40,51,58,62]. Seven different scales were used to assess quality of life, and only the EuroQol EQ-5D was used in more than 1 study [29,45,46,52,58]. Only 1 study reported quality of life as a primary outcome [51].

Cost-effectiveness was reported as an incremental cost-effectiveness ratio (ICER) in 5 studies. An ICER of $1,121 per 1% decrease in HbA1c was reported in the trial by DePue and colleagues [36] and $1,850 in the study by Ali and colleagues [58]. The study by Mash et al. reported an ICER of $1,862 per quality-adjusted life year (QALY) based on improvements in blood pressure [46]. Two other trials calculated an ICER between trial arms [33,59]. No study reported cost-effectiveness as a primary outcome.

In the meta-analysis of HbA1c in 35 trials, there were 10,060 total participants when accounting for the design effect of cluster RCTs (5,240 in intervention arms and 4,820 in comparator arms). The overall between-arm HbA1c mean change was −0.46% (95% CI −0.60% to −0.31%, I2 87.8% [95% CI 84.0% to 90.6%]; Fig 2). Within subgroups of intervention type, mean HbA1c difference was −0.37% (95% CI −0.64% to −0.10%, I2 60.0% [95% CI 8.2% to 82.6%], n = 7) in multicomponent clinic-based interventions, −0.87% (95% CI −1.20% to −0.53%, I2 91.0% [95% CI 86.5% to 94.0%], n = 13) in pharmacist task-sharing studies, and −0.27% (95% CI −0.50% to −0.04%, I2 64.1% [95% CI 18.8% to 84.1%], n = 7) in trials of diabetes education or support alone. The effect sizes of other intervention types with 2 or fewer studies reporting HbA1c are summarized in Fig 2. The overall HbA1c prediction interval was −1.19% to 0.28%.

10.1371/journal.pmed.1003434.g002 Forest plot for meta-analysis of hemoglobin A1c (%) mean difference.

Studies are listed in the figure by first author [28–36,38–42, 44–55,57–60,62–66]. The intervention arms were combined in the study by Anzaldo-Campos and colleagues [59]. Only the health literacy intervention arm was included in the study by Wang and colleagues [54]. The proportion of participants from low- and middle-income countries was inferred to be 60% in the study by Reutens and colleagues [66]. Participant numbers in cluster RCTs are adjusted for design effect as described in the Methods. The prediction interval is depicted as the horizontal whiskers intersecting the overall effect diamond marker.

 Risk of bias and sensitivity analysis

Eight studies were at low risk of bias for the summary assessment of glycemic control, 15 studies were at unclear risk, and 16 studies were at high risk (S7 Appendix). The overall funnel plot and Egger test for the HbA1c meta-analysis suggested possible bias (Egger p < 0.001; S8 Appendix), but there was little evidence of bias within subgroups of intervention types (S9 Appendix). Using the trim-and-fill method, we estimated that there were 8 missing studies, and inclusion of these imputed studies resulted in an estimated overall HbA1c mean difference of −0.28% (95% CI −0.43% to −0.13%; S10 Appendix). In the sensitivity analysis of studies not at high risk of bias (n = 21 trials), the overall HbA1c mean difference was −0.20% (95% CI −0.32% to −0.08%, I2 71.8% [95% CI 56.2% to 81.8%]; S11–S13 Appendices). In the sensitivity analysis using the leave-one-out method, we found that exclusion of the study with the largest effect size [42] would result in a HbA1c mean difference of −0.39% (95% CI −0.52% to −0.26%, I2 84.5%; S14 Appendix).

The certainty of evidence using the GRADE/EPOC approach for glycemic control by subgroup was moderate for multicomponent clinic-based interventions but was low or very low for other intervention types (S15 Appendix). The most common reasons for downgrading the certainty of evidence for intervention types were concerns regarding risk of bias or inconsistency across studies (S16 Appendix). For example, in the case of pharmacist task-sharing interventions, 9 of the 14 studies were classified as being at high risk of bias, 5 were at unclear risk of bias, and none were at low risk of bias. The absence of high-quality trials resulted in a low certainty of evidence for pharmacist task-sharing studies despite their sizeable pooled HbA1c estimate in the meta-analysis. Conversely, in the case of multicomponent clinic-based interventions, only 1 of the 8 studies was deemed to be at high risk of bias, and 3 of the studies were at low risk of bias. The result was a moderate certainty of evidence for the glycemic outcome for these interventions despite a lower pooled HbA1c estimate than for the pharmacist-led studies.

 Discussion

We systematically reviewed the literature and identified 39 RCTs of health system interventions for adults with type 2 diabetes in LMICs that assessed glycemic control, mortality, health-related quality of life, or cost-effectiveness. Most included studies were conducted in upper-middle-income countries, and few studies were carried out in rural areas or low- or lower-middle-income countries. Mirroring global patterns [67], this research disparity is discordant with epidemiologic evidence showing a substantial diabetes burden in low-income countries and in rural areas of LMICs [68,69]. The EPOC domains of delivery arrangements and implementation strategies were most commonly involved in the included interventions. In the overall meta-analysis of HbA1c from 35 trials, we found that health system interventions modestly improved glycemic control on average. At the same time, the wide prediction interval overlapping 0 in the meta-analysis of HbA1c showed that there was a wide range of effectiveness across studies, and some health system interventions may not be effective in all settings. Non-glycemic outcomes of mortality, health-related quality of life, and cost-effectiveness were less frequently reported. There was considerable heterogeneity in the overall pooled analysis that was partially explained by intervention type and baseline HbA1c. Within intervention types, multicomponent clinic-based interventions had moderate evidence of glycemic benefit, but the certainty of evidence was low or very low for other intervention types.

Our review complements prior meta-analyses of studies primarily from HICs showing the benefit of systems-level quality improvement interventions on surrogate outcomes such as HbA1c, blood pressure, and cholesterol [7–9]. However, these prior reviews have included few trials outside of HICs, which limits generalizability to health systems in LMICs. In LMICs, published reviews of health system interventions for diabetes care have explored diabetes care models [13], integrated hypertension and diabetes care [12], and interventions with a lifestyle component [11]. Incorporating heterogeneous study designs, these previous reviews have surveyed the limited evidence and described various approaches that have been implemented in LMICs. Our review adds to the literature by focusing on clinical and patient-oriented outcomes from the increasing number of randomized trials conducted in these settings.

The most common intervention types we identified were multicomponent clinic-based interventions, pharmacist task-sharing interventions, and interventions of diabetes education or support alone. Multicomponent clinic-based interventions were modestly effective in improving glycemic control, with moderate certainty of evidence. At the same time, multiple well-conducted trials had null results [30,31,62]. These findings may reflect differences in participants, setting, or the implementation of different components in the bundle. Of note, the comparator arm in 2 of these well-conducted multicomponent clinic-based trials consisted of enhanced usual care [30,31], potentially causing an attenuation of effect size. In HICs, components with the largest effect sizes have been team change, patient education or patient self-management, electronic registries, and promotion of patient–provider communication [7].

Interventions focusing solely on implementing diabetes education or support within the health system also were effective in improving glycemic control, but the certainty of evidence was low. All 3 trials judged as low risk of bias had null results [32,43,46]. One potential explanation for inconsistent findings is the relatively low contact intensity of many studies. In HICs, a dose-dependent relationship has been observed between contact intensity and glycemic effectiveness, with interventions with 10 or fewer hours found to be ineffective [70]. Another consideration is that research trial infrastructure in resource-limited settings may catalyze the delivery of standard clinical care across trial arms. This revitalization of underlying care may make it difficult to detect modest differences attributable to education or support alone. A dramatic example of this effect was the Happy Life Club trial in China, in which both trial arms experienced 3.7% within-group HbA1c improvement over 18 months [32]. Both the intervention and comparator arm in this trial had out-of-pocket medical fees waived, which may have contributed to catalyzing participants to seek medical care. Importantly, we included diabetes education or support trials that primarily changed the behavior of health workers within the health system, and we excluded lifestyle trials focusing on patient behavior alone without systems-level change.

Task sharing was a common thread across intervention types. Distinct from task shifting, task sharing emphasizes the shared responsibility for a task between the health workers’ different levels and types of training [71]. Previous reviews of trials predominantly conducted in HICs have suggested task sharing with pharmacists as an effective strategy [8]. We found that pharmacist task-sharing interventions appeared to improve glycemic control in the pooled analysis, but the certainty of evidence was low for these types of interventions, primarily due to concerns about studies’ risk of bias.

Task sharing also was a fundamental component in the nurse-led intervention by Fairall and colleagues in South Africa [29], a nurse care coordination trial [52], and multicomponent clinic-based studies [31,58,59]. We observed differences across studies relating to task sharing such as type of health worker, training, and assigned tasks. Prior reviews of task shifting for chronic diseases in LMICs have identified few trials in type 2 diabetes [72,73]. A 2019 meta-analysis by Anand and colleagues concluded that task-sharing interventions were effective in improving blood pressure in LMICs [74]. Our review shows an increase in research incorporating task sharing into health system interventions for type 2 diabetes in these settings.

Our review should be considered in the context of the movement to strengthen health systems in LMICs [75]. Diabetes has been referred to as a “tracer condition” for assessing the strength of health systems [76], and inadequate diabetes care has been reported in nationally representative surveys in many LMICs [4]. RCTs are not the only form of evidence generation in the field of health policy and research [77], and diverse research strategies are needed in conditions like type 2 diabetes that have a strong clinical evidence base yet weak evidence on implementation [3,78]. Logistical challenges in conducting randomized studies within health systems likely explain why we identified few interventions testing financial or governance arrangements. An advantage of including only RCTs is that we are able to offer robust evidence of the impact of health system interventions on glycemic control and reveal the limited data on other outcomes. Further studies in LMICs are needed to assess non-glycemic outcomes and, given the wide prediction intervals, to determine the specific components and details of health system interventions most likely to promote effectiveness and limit potential harms. Excellent examples of implementation research include the portfolio of ongoing projects funded by the Global Alliance for Chronic Disease [79].

Our review has limitations. Defining a health system intervention is challenging, and there is no consensus definition. We justify our use of the EPOC definition as reasonable given its use in prior Cochrane EPOC reviews on health systems in LMICs. We excluded non-randomized study designs given the challenge in attributing causality for outcomes and inconsistent reporting of these designs in pilot searches. Randomized designs in health system research have limitations, including the possible attenuation of effect sizes [80]. There was statistical evidence for publication bias and substantial differences in the quality of underlying studies that limited the certainty of evidence of glycemic benefit for intervention types including pharmacist task-sharing interventions. We did not assess blood pressure outcomes given our primary interest in the evidence of interventions attempting to achieve glycemic control and prior meta-analyses supporting the effectiveness of health system interventions for blood pressure control [74,81]. The only outcome in our review for which a meta-analysis was conducted, HbA1c, is only a surrogate outcome, but it is commonly used in meta-analyses of systems-level interventions for diabetes [7,9]. Our review was restricted to studies with at least 6 months of follow-up and 100 enrolled participants. Multiple trials were included within some countries, but we did not formally account for a potential lack of independence among studies conducted within the same health system context. This aspect reflects a limitation of the evidence generated rather than one of the analysis itself. We also did not systematically assess important implementation science outcomes such as reach, fidelity, or acceptability. Finally, although there were substantial similarities within intervention types, individual studies varied by setting and population, limiting our ability to make conclusions with high degrees of certainty.

This review has notable strengths. We synthesized evidence of outcomes by focusing on RCTs and performing a meta-analysis of HbA1c. Our comprehensive search strategy facilitated this choice as we identified a larger number of trials in LMICs than previous reviews. Our review was supplemented with unpublished data received from multiple study authors, and we were able to pool HbA1c statistical estimates reported differently across studies.

In conclusion, we found that health system interventions for type 2 diabetes may be effective in improving glycemic control in LMICs, but few studies were available from rural areas or low- or lower-middle-income countries. Multicomponent clinic-based interventions had the strongest evidence for glycemic benefit among intervention types. Data were generally limited for non-glycemic outcomes such as mortality, quality of life, and cost-effectiveness. Our findings imply a need for implementation research to investigate the details of health system interventions that confer durable improvements in clinical and patient-centered outcomes in LMICs, especially in rural areas and in low- and lower-middle-income countries.

 Supporting information PROSPERO registration and final review protocol.

(PDF)

 PRISMA checklist.

(PDF)

 Search strategy.

(PDF)

 Main domains of the EPOC taxonomy of health systems.

(PDF)

 Characteristics of included studies.

(PDF)

 Forest plot of overall deaths by study.

(PDF)

 EPOC risk of bias assessment.

(PDF)

 Overall funnel plot (all studies).

(PDF)

 Funnel plot by intervention type (all studies).

(PDF)

 Funnel plot generated using trim-and-fill method.

(PDF)

 Forest plot for meta-analysis of HbA1c (%) mean difference excluding studies at high risk of bias.

(PDF)

 Overall funnel plot excluding studies at high risk of bias.

(PDF)

 Funnel plot by intervention type excluding studies at high risk of bias.

(PDF)

 Forest plot generated using leave-one-out method.

(PDF)

 Summary of findings table.

(PDF)

 Certainty assessment of evidence for the glycemic control outcome.

(PDF)

We thank the following authors of included studies for contributing supplementary information used in this review: María Cecilia Anzaldo-Campos, MD, MBA; Anna Chapman, PhD; Jeroen De Man; Shaun Wen Huey Lee, PhD; Aditya Khetan, MD; Professor Dr. Anis Safura Ramli; Professor Hong-Mei Wang, PhD; and Xuefeng Zhong, MD, MPH, PhD.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the funders.

 Abbreviations EPOC

Effective Practice and Organisation of Care

 HbA1c

hemoglobin A1c

 HICs

high-income countries

 LMICs

low- and middle-income countries

 RCT

randomized controlled trial

 SMS

short message service

 References International Diabetes Federation. IDF diabetes atlas. 9th edition. Brussels: International Diabetes Federation; 2019. NCD Risk Factor Collaboration. Worldwide trends in diabetes since 1980: a pooled analysis of 751 population-based studies with 4.4 million participants. Lancet. 2016;387(10027):1513–30. doi: 10.1016/S0140-6736(16)00618-8 27061677 Ali MK, Siegel KR, Chandrasekar E, Tandon R, Ascher Montoya P, Mbanya JC, et al. Diabetes: an update on the pandemic and potential solutions. In: Prabhakaran D, Anand S, Gaziano TA, Mbanya JC, Wu Y, Nugent R, editors. Disease control priorities. 3rd edition. Volume 5: cardiovascular, respiratory, and related disorders. Washington (DC): World Bank; 2017. Manne-Goehler J, Geldsetzer P, Agoudavi K, Andall-Brereton G, Aryal KK, Bicaba BW, et al. Health system performance for people with diabetes in 28 low- and middle-income countries: a cross-sectional study of nationally representative surveys. PLoS Med. 2019;16(3):e1002751. doi: 10.1371/journal.pmed.1002751 30822339 Effective Practice and Organisation of Care. EPOC taxonomy—topics list. Oxford: Effective Practice and Organisation of Care; 2015 [cited 2020 Mar 17]. Available from: https://epoc.cochrane.org/sites/epoc.cochrane.org/files/public/uploads/taxonomy/epoc_taxonomy.pdf. Atun R, Davies JI, Gale EAM, Barnighausen T, Beran D, Kengne AP, et al. Diabetes in sub-Saharan Africa: from clinical care to health policy. Lancet Diabetes Endocrinol. 2017;5(8):622–67. doi: 10.1016/S2213-8587(17)30181-X 28688818 Lim LL, Lau ESH, Kong APS, Davies MJ, Levitt NS, Eliasson B, et al. Aspects of multicomponent integrated care promote sustained improvement in surrogate clinical outcomes: a systematic review and meta-analysis. Diabetes Care. 2018;41(6):1312–20. doi: 10.2337/dc17-2010 29784698 Ong SE, Koh JJK, Toh SES, Chia KS, Balabanova D, McKee M, et al. Assessing the influence of health systems on type 2 diabetes mellitus awareness, treatment, adherence, and control: a systematic review. PLoS ONE. 2018;13(3):e0195086. doi: 10.1371/journal.pone.0195086 29596495 Tricco AC, Ivers NM, Grimshaw JM, Moher D, Turner L, Galipeau J, et al. Effectiveness of quality improvement strategies on the management of diabetes: a systematic review and meta-analysis. Lancet. 2012;379(9833):2252–61. doi: 10.1016/S0140-6736(12)60480-2 22683130 Miranda JJ, Zaman MJ. Exporting ‘failure’: why research from rich countries may not benefit the developing world. Rev Saude Publica. 2010;44(1):185–9. doi: 10.1590/s0034-89102010000100020 20140343 Afable A, Karingula NS. Evidence based review of type 2 diabetes prevention and management in low and middle income countries. World J Diabetes. 2016;7(10):209–29. doi: 10.4239/wjd.v7.i10.209 27226816 Correia JC, Lachat S, Lagger G, Chappuis F, Golay A, Beran D, et al. Interventions targeting hypertension and diabetes mellitus at community and primary healthcare level in low- and middle-income countries: a scoping review. BMC Public Health. 2019;19(1):1542. doi: 10.1186/s12889-019-7842-6 31752801 Esterson YB, Carey M, Piette JD, Thomas N, Hawkins M. A systematic review of innovative diabetes care models in low-and middle-income countries (LMICs). J Health Care Poor Underserved. 2014;25(1):72–93. doi: 10.1353/hpu.2014.0037 24509014 Effective Practice and Organisation of Care. Scope of our work. Oxford: Effective Practice and Organisation of Care; 2020 [cited 2020 Mar 17]. Available from: https://epoc.cochrane.org/scope-our-work. Moher D, Liberati A, Tetzlaff J, Altman DG, PRISMA Group. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med. 2009;6(7):e1000097. doi: 10.1371/journal.pmed.1000097 19621072 Jackson JL, Kuriyama A, Anton A, Choi A, Fournier JP, Geier AK, et al. The accuracy of Google Translate for abstracting data from non-English-language trials for systematic reviews. Ann Intern Med. 2019;171(9):677–9. doi: 10.7326/M19-0891 31357212 Hoffmann TC, Glasziou PP, Boutron I, Milne R, Perera R, Moher D, et al. Better reporting of interventions: template for intervention description and replication (TIDieR) checklist and guide. BMJ. 2014;348:g1687. doi: 10.1136/bmj.g1687 24609605 Effective Practice and Organisation of Care. Suggested risk of bias criteria for EPOC reviews. Oxford: Effective Practice and Organisation of Care; 2017 [cited 2020 Oct 22]. Available from: https://epoc.cochrane.org/sites/epoc.cochrane.org/files/public/uploads/Resources-for-authors2017/suggested_risk_of_bias_criteria_for_epoc_reviews.pdf. Effective Practice and Organisation of Care. Summary assessments of the risk of bias. Oxford: Effective Practice and Organisation of Care; 2017 [cited 2020 Oct 22]. Available from: http://epoc.cochrane.org/sites/epoc.cochrane.org/files/public/uploads/Resources-for-authors2017/summary_assessments_of_the_risk_of_bias.pdf. Effective Practice and Organisation of Care. EPOC resources for review authors: worksheets for preparing a Summary of Findings tables using GRADE. Oxford: Effective Practice and Organisation of Care; 2020 [cited 2020 Oct 22]. Available from: http://epoc.cochrane.org/resources/epoc-resources-review-authors. Guyatt GH, Oxman AD, Vist GE, Kunz R, Falck-Ytter Y, Alonso-Coello P, et al. GRADE: an emerging consensus on rating quality of evidence and strength of recommendations. BMJ. 2008;336(7650):924–6. doi: 10.1136/bmj.39489.470347.AD 18436948 Higgins J, Green S, editors. Cochrane handbook for systematic reviews of interventions. Version 5.1.0. London: Cochrane Collaboration; 2011. Littenberg B, MacLean CD. Intra-cluster correlation coefficients in adults with diabetes in primary care practices: the Vermont Diabetes Information System field survey. BMC Med Res Methodol. 2006;6(1):20. doi: 10.1186/1471-2288-6-20 16672056 Riley RD, Higgins JP, Deeks JJ. Interpretation of random effects meta-analyses. BMJ. 2011;342:d549. doi: 10.1136/bmj.d549 21310794 Wallace BC, Schmid CH, Lau J, Trikalinos TA. Meta-Analyst: software for meta-analysis of binary, continuous and diagnostic data. BMC Med Res Methodol. 2009;9:80. doi: 10.1186/1471-2288-9-80 19961608 Borenstein M, Hedges LV, Higgins JP, Rothstein HR. Introduction to meta-analysis. West Sussex (UK): John Wiley & Sons; 2009. Duval S, Tweedie R. Trim and fill: a simple funnel-plot-based method of testing and adjusting for publication bias in meta-analysis. Biometrics. 2000;56(2):455–63. doi: 10.1111/j.0006-341x.2000.00455.x 10877304 Phumipamorn S, Pongwecharak J, Soorapan S, Pattharachayakul S. Effects of the pharmacist’s input on glycaemic control and cardiovascular risks in Muslim diabetes. Prim Care Diabetes. 2008;2(1):31–7. doi: 10.1016/j.pcd.2007.12.001 18684418 Fairall LR, Folb N, Timmerman V, Lombard C, Steyn K, Bachmann MO, et al. Educational outreach with an integrated clinical tool for nurse-led non-communicable chronic disease management in primary care in South Africa: a pragmatic cluster randomised controlled trial. PLoS Med. 2016;13(11):e1002178. doi: 10.1371/journal.pmed.1002178 27875542 Khan MA, Walley JD, Khan N, Hicks J, Ahmed M, Khan SE, et al. Effectiveness of an integrated diabetes care package at primary healthcare facilities: a cluster randomised trial in Pakistan. BJGP Open. 2018;2(4):bjgpopen18X101618. doi: 10.3399/bjgpopen18X101618 30723803 Prabhakaran D, Jha D, Prieto-Merino D, Roy A, Singh K, Ajay VS, et al. Effectiveness of an mHealth-based electronic decision support system for integrated management of chronic conditions in primary care: the mWellcare cluster-randomized controlled trial. Circulation. 2018 Nov 10. doi: 10.1161/CIRCULATIONAHA.118.038192 30586732 Chapman A, Browning CJ, Enticott JC, Yang H, Liu S, Zhang T, et al. Effect of a health coach intervention for the management of individuals with type 2 diabetes mellitus in China: a pragmatic cluster randomized controlled trial. Front Public Health. 2018;6:252. doi: 10.3389/fpubh.2018.00252 30283767 Adibe MO, Obinna UP, Uchenna IN, Michael UC, Aguwa CN. Effects of an additional pharmaceutical care intervention versus usual care on clinical outcomes of type 2 diabetes patients in Nigeria: a comparative study. Sci Res Essays. 2014;9(12):548–56. Ayadurai S, Sunderland VB, Tee LBG, Md Said SN, Hattingh HL. Structured tool to improve clinical outcomes of type 2 diabetes mellitus patients: a randomized controlled trial. J Diabetes. 2018;10(12):965–76. doi: 10.1111/1753-0407.12799 29877609 Chung WW, Chua SS, Lai PS, Chan SP. Effects of a pharmaceutical care model on medication adherence and glycemic control of people with type 2 diabetes. Patient Prefer Adherence. 2014;8:1185–94. doi: 10.2147/PPA.S66619 25214772 DePue JD, Dunsiger S, Seiden AD, Blume J, Rosen RK, Goldstein MG, et al. Nurse-community health worker team improves diabetes care in American Samoa: results of a randomized controlled trial. Diabetes Care. 2013;36(7):1947–53. doi: 10.2337/dc12-1969 23393217 Gillani SW. Determining effective diabetic care; a multicentre—longitudinal interventional study. Curr Pharm Des. 2016;22(42):6469–76. doi: 10.2174/1381612822666160813235704 27526787 Goruntla N, Mallela V, Nayakanti D. Impact of pharmacist-directed counseling and message reminder services on medication adherence and clinical outcomes in type 2 diabetes mellitus. J Pharm Bioallied Sci. 2019;11(1):69–76. doi: 10.4103/jpbs.JPBS_211_18 30906142 Guo Z, Liu J, Zeng H, He G, Ren X, Guo J. Feasibility and efficacy of nurse-led team management intervention for improving the self-management of type 2 diabetes patients in a Chinese community: a randomized controlled trial. Patient Prefer Adherence. 2019;13:1353–62. doi: 10.2147/PPA.S213645 31616135 Jaipakdee J, Jiamjarasrangsi W, Lohsoonthorn V, Lertmaharit S. Effectiveness of a self-management support program for Thais with type 2 diabetes: evaluation according to the RE-AIM framework. Nurs Health Sci. 2015;17(3):362–9. doi: 10.1111/nhs.12198 25818472 Jarab AS, Alqudah SG, Mukattash TL, Shattat G, Al-Qirim T. Randomized controlled trial of clinical pharmacy management of patients with type 2 diabetes in an outpatient diabetes clinic in Jordan. J Manag Care Pharm. 2012;18(7):516–26. doi: 10.18553/jmcp.2012.18.7.516 22971205 Javaid Z, Imtiaz U, Khalid I, Saeed H, Khan RQ, Islam M, et al. A randomized control trial of primary care-based management of type 2 diabetes by a pharmacist in Pakistan. BMC Health Serv Res. 2019;19(1):409. doi: 10.1186/s12913-019-4274-z 31234840 Khetan A, Zullo M, Rani A, Gupta R, Purushothaman R, Bajaj NS, et al. Effect of a community health worker-based approach to integrated cardiovascular risk factor control in India: a cluster randomized controlled trial. Glob Heart. 2019;14(4):355–65. doi: 10.1016/j.gheart.2019.08.003 31523014 Kim HS, Sun C, Yang SJ, Sun L, Li F, Choi IY, et al. Randomized, open-label, parallel group study to evaluate the effect of internet-based glucose management system on subjects with diabetes in China. Telemed J E Health. 2016;22(8):666–74. doi: 10.1089/tmj.2015.0170 26938489 Lee JY, Chan CKY, Chua SS, Ng CJ, Paraidathathu T, Lee KKC, et al. Telemonitoring and team-based management of glycemic control on people with type 2 diabetes: a cluster-randomized controlled trial. J Gen Intern Med. 2020;35(1):87–94. doi: 10.1007/s11606-019-05316-9 31512187 Mash RJ, Rhode H, Zwarenstein M, Rollnick S, Lombard C, Steyn K, et al. Effectiveness of a group diabetes education programme in under-served communities in South Africa: a pragmatic cluster randomized controlled trial. Diabet Med. 2014;31(8):987–93. doi: 10.1111/dme.12475 24766179 Mourao AO, Ferreira WR, Martins MA, Reis AM, Carrillo MR, Guimaraes AG, et al. Pharmaceutical care program for type 2 diabetes patients in Brazil: a randomised controlled trial. Int J Clin Pharm. 2013;35(1):79–86. doi: 10.1007/s11096-012-9710-7 23161124 Neto PR, Marusic S, de Lyra Junior DP, Pilger D, Cruciol-Souza JM, Gaeti WP, et al. Effect of a 36-month pharmaceutical care program on the coronary heart disease risk in elderly diabetic and hypertensive patients. J Pharm Pharm Sci. 2011;14(2):249–63. doi: 10.18433/j3259q 21733413 Paz-Pacheco E, Sandoval MA, Ardena GJ, Paterno E, Juban N, Lantion-Ang FL, et al. Effectiveness of a community-based diabetes self-management education (DSME) program in a rural agricultural setting. Prim Health Care Res Dev. 2017;18(1):35–49. doi: 10.1017/S1463423616000335 27640303 Sarayani A, Mashayekhi M, Nosrati M, Jahangard-Rafsanjani Z, Javadi M, Saadat N, et al. Efficacy of a telephone-based intervention among patients with type-2 diabetes; a randomized controlled trial in pharmacy practice. Int J Clin Pharm. 2018;40(2):345–53. doi: 10.1007/s11096-018-0593-0 29435911 Sriram S, Chack LE, Ramasamy R, Ghasemi A, Ravi TK, Sabzghabaee AM. Impact of pharmaceutical care on quality of life in patients with type 2 diabetes mellitus. J Res Med Sci. 2011;16(Suppl 1):S412–8. 22247727 Tutino GE, Yang WY, Li X, Li WH, Zhang YY, Guo XH, et al. A multicentre demonstration project to evaluate the effectiveness and acceptability of the web-based Joint Asia Diabetes Evaluation (JADE) programme with or without nurse support in Chinese patients with type 2 diabetes. Diabet Med. 2017;34(3):440–50. doi: 10.1111/dme.13164 27278933 Van Olmen J, Kegels G, Korachais C, de Man J, Van Acker K, Kalobu JC, et al. The effect of text message support on diabetes self-management in developing countries—a randomised trial. J Clin Transl Endocrinol. 2017;7:33–41. doi: 10.1016/j.jcte.2016.12.005 29067248 Wang L, Fang H, Xia Q, Liu X, Chen Y, Zhou P, et al. Health literacy and exercise-focused interventions on clinical measurements in Chinese diabetes patients: a cluster randomized controlled trial. EClinicalMedicine. 2019;17:100211. doi: 10.1016/j.eclinm.2019.11.004 31891144 Wishah RA, Al-Khawaldeh OA, Albsoul AM. Impact of pharmaceutical care interventions on glycemic control and other health-related clinical outcomes in patients with type 2 diabetes: randomized controlled trial. Diabetes Metab Syndr. 2015;9(4):271–6. doi: 10.1016/j.dsx.2014.09.001 25301007 Zhong X, Wang Z, Fisher EB, Tanasugarn C. Peer support for diabetes management in primary care and community settings in Anhui Province, China. Ann Fam Med. 2015;13(Suppl 1):S50–8. doi: 10.1370/afm.1799 26304972 Shen M, Zhao H, Sun K, Lu Y, Wang Z. Influence of pharmaceutical care intervention on type 2 diabetes patients with multi-drug therapy in the community. Pharm Care Res. 2016;16(3):170–4. Ali MK, Singh K, Kondal D, Devarajan R, Patel SA, Shivashankar R, et al. Effectiveness of a multicomponent quality improvement strategy to improve achievement of diabetes care goals: a randomized, controlled trial. Ann Intern Med. 2016;165(6):399–408. doi: 10.7326/M15-2807 27398874 Anzaldo-Campos MC, Contreras S, Vargas-Ojeda A, Menchaca-Diaz R, Fortmann A, Philis-Tsimikas A. Dulce Wireless Tijuana: a randomized control trial evaluating the impact of Project Dulce and short-term mobile technology on glycemic control in a family medicine clinic in northern Mexico. Diabetes Technol Ther. 2016;18(4):240–51. doi: 10.1089/dia.2015.0283 26914371 Barcelo A, Cafiero E, de Boer M, Mesa AE, Lopez MG, Jimenez RA, et al. Using collaborative learning to improve diabetes care and outcomes: the VIDA project. Prim Care Diabetes. 2010;4(3):145–53. doi: 10.1016/j.pcd.2010.04.005 20478753 Chao J, Yang L, Xu H, Yu Q, Jiang L, Zong M. The effect of integrated health management model on the health of older adults with diabetes in a randomized controlled trial. Arch Gerontol Geriatr. 2015;60(1):82–8. doi: 10.1016/j.archger.2014.10.006 25456892 Kong JX, Zhu L, Wang HM, Li Y, Guo AY, Gao C, et al. Effectiveness of the chronic care model in type 2 diabetes management in a community health service center in China: a group randomized experimental study. J Diabetes Res. 2019;2019:6516581. doi: 10.1155/2019/6516581 30719455 Ramli AS, Selvarajah S, Daud MH, Haniff J, Abdul-Razak S, Tg-Abu-Bakar-Sidik TM, et al. Effectiveness of the EMPOWER-PAR intervention in improving clinical outcomes of type 2 diabetes mellitus in primary care: a pragmatic cluster randomised controlled trial. BMC Fam Pract. 2016;17(1):157. doi: 10.1186/s12875-016-0557-1 27842495 Saleh S, Farah A, Dimassi H, El Arnaout N, Constantin J, Osman M, et al. Using mobile health to enhance outcomes of noncommunicable diseases care in rural settings and refugee camps: randomized controlled trial. JMIR Mhealth Uhealth. 2018;6(7):e137. doi: 10.2196/mhealth.8146 30006326 Akturan S, Kaya CA, Unalan PC, Akman M. The effect of the BATHE interview technique on the empowerment of diabetic patients in primary care: a cluster randomised controlled study. Prim Care Diabetes. 2017;11(2):154–61. doi: 10.1016/j.pcd.2016.12.003 28122690 Reutens AT, Hutchinson R, Van Binh T, Cockram C, Deerochanawong C, Ho LT, et al. The GIANT study, a cluster-randomised controlled trial of efficacy of education of doctors about type 2 diabetes mellitus management guidelines in primary care practice. Diabetes Res Clin Pract. 2012;98(1):38–45. doi: 10.1016/j.diabres.2012.06.002 22784926 Viergever RF, Li K. Trends in global clinical trial registration: an analysis of numbers of registered clinical trials in different parts of the world from 2004 to 2013. BMJ Open. 2015;5(9):e008932. doi: 10.1136/bmjopen-2015-008932 26408831 Dagenais GR, Gerstein HC, Zhang X, McQueen M, Lear S, Lopez-Jaramillo P, et al. Variations in diabetes prevalence in low-, middle-, and high-income countries: results from the prospective urban and rural epidemiological study. Diabetes Care. 2016;39(5):780–7. doi: 10.2337/dc15-2338 26965719 Seiglie JA, Marcus ME, Ebert C, Prodromidis N, Geldsetzer P, Theilmann M, et al. Diabetes prevalence and its relationship with education, wealth, and BMI in 29 low- and middle-income countries. Diabetes Care. 2020;43(4):767–75. doi: 10.2337/dc19-1782 32051243 Pillay J, Armstrong MJ, Butalia S, Donovan LE, Sigal RJ, Vandermeer B, et al. Behavioral programs for type 2 diabetes mellitus: a systematic review and network meta-analysis. Ann Intern Med. 2015;163(11):848–60. doi: 10.7326/M15-1400 26414227 Meara JG, Leather AJM, Hagander L, Alkire BC, Alonso N, Ameh EA, et al. Global surgery 2030: evidence and solutions for achieving health, welfare, and economic development. Lancet. 2015;386(9993):569–624. doi: 10.1016/S0140-6736(15)60160-X 25924834 Joshi R, Alim M, Kengne AP, Jan S, Maulik PK, Peiris D, et al. Task shifting for non-communicable disease management in low and middle income countries—a systematic review. PLoS ONE. 2014;9(8):e103754. doi: 10.1371/journal.pone.0103754 25121789 Ogedegbe G, Gyamfi J, Plange-Rhule J, Surkis A, Rosenthal DM, Airhihenbuwa C, et al. Task shifting interventions for cardiovascular risk reduction in low-income and middle-income countries: a systematic review of randomised controlled trials. BMJ Open. 2014;4(10):e005983. doi: 10.1136/bmjopen-2014-005983 25324324 Anand TN, Joseph LM, Geetha AV, Prabhakaran D, Jeemon P. Task sharing with non-physician health-care workers for management of blood pressure in low-income and middle-income countries: a systematic review and meta-analysis. Lancet Glob Health. 2019;7(6):e761–71. doi: 10.1016/S2214-109X(19)30077-4 31097278 Hafner T, Shiffman J. The emergence of global attention to health systems strengthening. Health Policy Plan. 2013;28(1):41–50. doi: 10.1093/heapol/czs023 22407017 Nolte E, Bain C, McKee M. Diabetes as a tracer condition in international benchmarking of health systems. Diabetes Care. 2006;29(5):1007–11. doi: 10.2337/diacare.2951007 16644629 Gilson L, editor. Health policy and systems research: a methodology reader. Geneva: World Health Organization; 2012. Owolabi MO, Yaria JO, Daivadanam M, Makanjuola AI, Parker G, Oldenburg B, et al. Gaps in guidelines for the management of diabetes in low- and middle-income versus high-income countries—a systematic review. Diabetes Care. 2018;41(5):1097–105. doi: 10.2337/dc17-1795 29678866 Daivadanam M, Ingram M, Sidney Annerstedt K, Parker G, Bobrow K, Dolovich L, et al. The role of context in implementation research for non-communicable diseases: answering the ‘how-to’ dilemma. PLoS ONE. 2019;14(4):e0214454. doi: 10.1371/journal.pone.0214454 30958868 English M, Schellenberg J, Todd J. Assessing health system interventions: key points when considering the value of randomization. Bull World Health Organ. 2011;89(12):907–12. doi: 10.2471/BLT.11.089524 22271948 Uthman OA, Hartley L, Rees K, Taylor F, Ebrahim S, Clarke A. Multiple risk factor interventions for primary prevention of cardiovascular disease in low- and middle-income countries. Cochrane Database Syst Rev. 2015;(8):CD011163. doi: 10.1002/14651858.CD011163.pub2 26272648 10.1371/journal.pmed.1003434.r001 Decision Letter 0 Arikainen Artur A. Senior Editor 2020 Artur A. Arikainen This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Submission Version 0

4 May 2020

Dear Dr Flood,

Thank you for submitting your manuscript entitled "Health system interventions for adults with type 2 diabetes in low- and middle-income countries: A systematic review and meta-analysis" for consideration by PLOS Medicine.

Your manuscript has now been evaluated by the PLOS Medicine editorial staff and I am writing to let you know that we would like to send your submission out for external peer review.

However, before we can send your manuscript to reviewers, we need you to complete your submission by providing the metadata that is required for full assessment. To this end, please login to Editorial Manager where you will find the paper in the 'Submissions Needing Revisions' folder on your homepage. Please click 'Revise Submission' from the Action Links and complete all additional questions in the submission questionnaire.

Please re-submit your manuscript within two working days, i.e. by .

Login to Editorial Manager here: https://www.editorialmanager.com/pmedicine

Once your full submission is complete, your paper will undergo a series of checks in preparation for peer review. Once your manuscript has passed all checks it will be sent out for review.

Feel free to email us at plosmedicine@plos.org if you have any queries relating to your submission.

Kind regards,

Artur Arikainen,

Associate Editor

PLOS Medicine

 10.1371/journal.pmed.1003434.r002 Decision Letter 1 Veitch Emma Senior Editor 2020 Emma Veitch This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Submission Version 1

7 Jul 2020

Dear Dr. Flood,

Thank you very much for submitting your manuscript "Health system interventions for adults with type 2 diabetes in low- and middle-income countries: A systematic review and meta-analysis" (PMEDICINE-D-20-01702R1) for consideration at PLOS Medicine.

Your paper was evaluated by a senior editor and discussed among all the editors here. It was evaluated by four independent reviewers, including a statistical reviewer. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of these reviews, I am afraid that we will not be able to accept the manuscript for publication in the journal in its current form, but we would like to consider a revised version that addresses the reviewers' and editors' comments. Obviously we cannot make any decision about publication until we have seen the revised manuscript and your response, and we plan to seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the PACE digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at PLOSMedicine@plos.org.

We expect to receive your revised manuscript by Jul 28 2020 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see http://journals.plos.org/plosmedicine/s/submission-guidelines#loc-methods.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

We look forward to receiving your revised manuscript.

Sincerely,

Emma Veitch, PhD

PLOS Medicine

On behalf of Clare Stone, PhD, Acting Chief Editor,

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

*In the last sentence of the Abstract Methods and Findings section, please add a brief note summarising any key limitation(s) of the study's methodology.

*At this stage, we ask that you include a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract. Please see our author guidelines for more information: https://journals.plos.org/plosmedicine/s/revising-your-manuscript#loc-author-summary

-----------------------------------------------------------

Comments from the reviewers:

Reviewer #1: This is an important but challenging area to investigate. The authors have done an excellent job of bringing together a large number of trials. The challenge is in the interpretation of findings when the results of studies are grouped and meta-analysed. The large I-squared values indicate the statistical risks in pooling these studies. This is compounded by the obvious heterogeneity of study design. Even within the broad categories of study type, there are unlikely to be two studies whose description of their intervention was the same, never mind how and how successfully the protocol was implemented. Added to this is the variation in the setting and populations, which would likely impact on the potential for and size of any benefit. Thus, it is quite likely that within the studies in each group there are interventions that work and some that don't. Indeed, it is likely that the need for some studies was justified on the basis that they were attempting to improve on other similar interventions that failed. In such circumstances it is very hard to know what a pooled estimate means, or even how to interpret general comments about the benefits of, for example, multicomponent interventions. It seems that it would be more useful to try to understand what features characterized the studies with the larger effect sizes. Of course this is also inherently difficult, as conclusions may then be driven by very small numbers of studies.

Ultimately, it seems that with current methodologies, it is very hard to make robust conclusions in this area.

-----------------------------------------------------------

Reviewer #2: See attachment

Michael Dewey

-----------------------------------------------------------

Reviewer #3: This systematic review and meta-analysis achieves its intended aims and has been conducted with sufficient rigour.

The introduction provides a clear rationale of the need for this study. It may be of benefit to include example/s of some health service interventions within the introduction (paragraph 2). While the EPOC definition is stated within the methods section, some examples earlier in the introduction would assist readers to interpret the distinction between clinical interventions and health service interventions.

The methods section provides sufficient detail as per the PRISMA recommendations. Information is provided with regard to eligible interventions, however, eligible control conditions do not appear to be stated throughout the manuscript. Given the intention of a meta-analysis is to quantify the effects of a specific intervention in relation to a specified control condition, this is an important consideration to include throughout the entire manuscript. This would also assist the reader to determine if the meta-analysis is comparing similar studies. It is possible that the heterogeneity observed may have been due to studies using different control conditions.

The results section provides a excellent synthesis of both the narrative and quantitative components. When summarizing outcomes, I would find it of interest to state how many of the included studies had a primary outcome of HbA1c (line 195).

The discussion section solidly interprets the findings in relation to previous literature and acknowledges the strengths and limitations of the review.

Overall a well written, well conducted, interesting piece of work.

-----------------------------------------------------------

Reviewer #4: Dr Flood and coworkers have conducted a systematic review and meta-analysis of reports (trials) on health system intervention for diabetes control in LIMC, using change in HbA1c as main outcome. After extensive searchers they identified 38 eligible studies, of which 34 were included in meta-analysis. They found a significant effect of the intervention with a change in HbA1c of 0.46% overall, 0.37% in multi-component clinic-based interventions, 0.93% in pharmacist task-sharing studies, and -0.27% in trials of diabetes education and support alone. Other pre-specified outcomes were inconsistently reported. They concluded on a likely efficacy of health interventions and advocated for further studies to assess the impact on non-glycaemic outcomes.

The review addresses a question of global relevance, the manuscript is generally well written and easy to read. I have the following suggestions to the authors.

1) There seems to be suggestions that non-English language studies were excluded in the selection process with 20 such studies excluded in the second last stage. Should some of these studies be eligible, then it will be a bias and the authors should rather consider using translated to access the information in non-English language studies for possible inclusion in the review.

2) Did the investigators make effort to contact the primary investigators for missing data; there seems to be indications that soem studies were excluded for missing data.

1) The authors have identified possible publication bias from funnel plots and have tried to explore this further using subgroup analyses. However, because of the overall small number of studies included, subgroups analyses are likely to be under powered to reliably explore publication bias. I am advising the authors to consider using the 'trim and fill' methods to impute the missing studies, and see if this results in imputed studies with plausible effect size. Similarly the authors should consider running other sensitivity analyses to confirm the robustness of their findings. I am thinking here of the 'leave one out' approach for instance, to check if some particular studies had more effect on the overall estimates.

-----------------------------------------------------------

Any attachments provided with reviews can be seen via the following link:

[LINK]

 10.1371/journal.pmed.1003434.r003 Author response to Decision Letter 1 Submission Version 2

29 Jul 2020

Submitted filename: 2020_07_28 - Responses to reviewers - Flood review.docx

 10.1371/journal.pmed.1003434.r004 Decision Letter 2 Arikainen Artur A. Senior Editor 2020 Artur A. Arikainen This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Submission Version 2

25 Aug 2020

Dear Dr. Flood,

Thank you very much for submitting your manuscript "Health system interventions for adults with type 2 diabetes in low- and middle-income countries: A systematic review and meta-analysis" (PMEDICINE-D-20-01702R2) for consideration at PLOS Medicine.

Your revised paper was evaluated by a senior editor and discussed among all the editors here. It was also sent to three of the four original independent reviewers once more, including a statistical reviewer. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of these reviews, I am afraid that we will still not be able to accept the manuscript for publication in the journal in its current form, but we would like to consider another revised version that addresses the reviewers' and editors' comments. Obviously we cannot make any decision about publication until we have seen the revised manuscript and your response, and we plan to seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the PACE digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at PLOSMedicine@plos.org.

We expect to receive your revised manuscript by Sep 15 2020 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see http://journals.plos.org/plosmedicine/s/submission-guidelines#loc-methods.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

We look forward to receiving your revised manuscript.

Sincerely,

Artur Arikainen

Associate Editor

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

1. Please address all of the reviewers’ comments below. Notably, we request that you update your study to include non-English studies and the additional sensitivity analyses requested. We appreciate that the inclusion of non-English language studies may require additional expense and effort, but we agree with reviewer #4 that the review would be incomplete without these, particularly given the focus on LMICs.

2. Abstract:

a. Please report all of the elements required by PRISMA for abstracts: http://www.plosmedicine.org/article/info:doi/10.1371/journal.pmed.1001419 . Specifically, please include a description of how study quality/risk of bias was assessed, and what the overall quality/risk of bias was found to be.

b. Please give the total number of patients across all included studies.

c. Please clarify (here and in the main Methods) that the search dates were “from inception through February 24, 2020”, as appropriate.

d. Please include a breakdown of the number of each study type for the included studies (eg. “18 cluster RCTs, 20 individual RCTs”), and regions covered (as from table 1).

e. Please include line numbers in the margin throughout.

3. Page 25: Please remove the Competing Interests, Data Availability, Funding and Author Contribution statements – these should be filled in on the online submission form instead. Please keep the Acknowledgements here as they are.

4. When completing the PRISMA checklist, please use section and paragraph numbers, rather than page numbers. Your checklist also appears to be cut off after item 14 – please include the full checklist.

Comments from the reviewers:

Reviewer #1: I think that the reviewers have done an adequate job in responding to my initial concerns.

The authors conclude that the most robust evidence of benefit is for multi-component interventions. However, this seems at odds with the data in Fig 2. Pharmacist task sharing interventions include more trials (albeit with fewer participants), a larger point estimate for the effect size, and 8/12 trials with individually statistically significant benefits. It appears that the downgrading of the Pharmacist trials is based on the GRADE score in Appendix S12, which is rated as 2/4 for the pharmacy trials compared to 3/4 for the multi-component trials. No discussion is provided on what the actual difference is between the quality scores, though this is central to the main study conclusion. The relevance of this is not only in regard to whether or not the evidence supports pharmacist task sharing interventions, but whether or not single-component interventions in general might be effective. Several other studies in this setting have concluded that multi-component interventions are required to be effective, and the current presentation of this meta-analysis supports this. However, that conclusion turns out to be entirely dependent on a difference in GRADE scores that is not explained or discussed.

Reviewer #2: The authors have addressed my points but I still have a couple of things to raise.

Even is the authors do not wish to do a formal analysis of mortality and the other outcomes I think expecting the reader to work all through S5 is unnecessary when they could be presented by outcome. I would put them in a forest plot but if the authors wish they can suppress the summary estimates.

I see there is some divergence of opinion between the referees about the heterogeneity and possible small study effects. For the record I have no problem with doing a summary in the presence of extreme statistical heterogeneity as measured by I^2. Whether the clinical heterogeneity precludes it is not for me to say. One thing I should have suggested and I apologise for not having done this at the beginning is that prediction intervals in addition to confidence intervals would be useful. They give a picture of where the next study would be likely to fall which I would imagine is more useful to someone working in public health in a different country than knowing the properties of the distribution of the summary statistic. Riley et al have argued for the use of prediction intervals https://doi.org/10.1136/bmj.d549 and it is not super-difficult to calculate them even if the authors' software does not do it. In the presence of the heterogeneity I do not think it is worth spending much time on possible small study effects and trim and fill has rather fallen out of favour recently.

I think I failed to explain adequately my concern about lack of independence of studies within a country. It is not just a concern about the same research team being involved which I am sure the authors have checked but also the fact that studies in a single country must share the same background health system and so will be more similar than those from a different country. I do not think country needs inclusion in a formal analysis but I would mention this as a possible limitation.

Michael Dewey

Reviewer #4: Thanks for the opportunity of reading the revised manuscript.

I note however that the authors have basically declined answering my two major comments:

1) They seem to be unable to access appropriate translation service to enable them assessing all potentially eligible studies; and therefore have limited inclusion to only studies published in English. In doing so, they have excluded at least 20 studies from further assessment due to language issue. Not only this deliberately introduces publication bias, but I am not sure if it would be acceptable for robust systematic review. The authors should consider the option of accounting at least for few other major international languages.

2) I proposed that they use sensitivity analyses (trim and fill) to explore the effect of missing studies imputation on the publication bias. I am not convinced by the rebuttals of the authors. Should the sensitivity analyses show that imputed studies have plausible effect sizes, then the reader needs to know, since accounting for missing studies could completely change the overall estimates from the analyses conducted by the authors. I also advised the authors to run other sensitivity analyses (leave one out) to confirm the robustness of their findings, what they haven't done.

Any attachments provided with reviews can be seen via the following link:

[LINK]

 10.1371/journal.pmed.1003434.r005 Author response to Decision Letter 2 Submission Version 3

1 Sep 2020

Submitted filename: 2020_09_01 - Responses to reviewers - Flood review.docx

 10.1371/journal.pmed.1003434.r006 Decision Letter 3 Arikainen Artur A. Senior Editor 2020 Artur A. Arikainen This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Submission Version 3

2 Oct 2020

Dear Dr. Flood,

Thank you very much for re-submitting your manuscript "Health system interventions for adults with type 2 diabetes in low- and middle-income countries: A systematic review and meta-analysis" (PMEDICINE-D-20-01702R3) for review by PLOS Medicine.

I have discussed the paper with my colleagues and the academic editor and it was also seen again by three reviewers. I am pleased to say that provided the remaining editorial and production issues are dealt with we are planning to accept the paper for publication in the journal.

The remaining issues that need to be addressed are listed at the end of this email. Any accompanying reviewer attachments can be seen via the link below. Please take these into account before resubmitting your manuscript:

[LINK]

Our publications team (plosmedicine@plos.org) will be in touch shortly about the production requirements for your paper, and the link and deadline for resubmission. DO NOT RESUBMIT BEFORE YOU'VE RECEIVED THE PRODUCTION REQUIREMENTS.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

In revising the manuscript for further consideration here, please ensure you address the specific points made by each reviewer and the editors. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments and the changes you have made in the manuscript. Please submit a clean version of the paper as the main article file. A version with changes marked must also be uploaded as a marked up manuscript file.

Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. If you haven't already, we ask that you provide a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract.

We expect to receive your revised manuscript within 1 week. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

We ask every co-author listed on the manuscript to fill in a contributing author statement. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

If you have any questions in the meantime, please contact me or the journal staff on plosmedicine@plos.org.

We look forward to receiving the revised manuscript by Oct 09 2020 11:59PM.

Sincerely,

Artur Arikainen,

Associate Editor

PLOS Medicine

plosmedicine.org

------------------------------------------------------------

Requests from Editors:

1. Short title: Please amend to: “Systematic review of health system interventions for adults with type 2 diabetes in low- and middle-income countries”

2. Abstract:

a. Please include a summary breakdown of included studies by region.

b. Lines 65-68: Please include p values for each result.

c. Around line 69: Please include this sentence from lines 362-363: “Eight studies were at low risk of bias for the summary assessment of glycemic control, 15 studies were at unclear risk, and 16 studies were at high risk.”

d. Line 76: Please begin with: “In this meta-analysis, we found that…”

3. Author summary: Line 94: Please define “glycemic control” for a lay reader, particularly as you mention “blood glucose control” later.

4. Line 213: Please remove this section on Funding. Please include this sentence elsewhere in the methods: “Ethical approval was not required as the research used publicly available data.”

5. Please remove spaces from within citation callouts throughout, eg “…professionals [30,31], and in…” (line 241).

6. You state that the multicomponent clinic-based interventions had the "most supporting evidence" (lines 78, 99-100, and 509-510). Yet there were apparently more pharmacist task sharing intervention trials included, and in these studies there was a stronger effect on the glycemic endpoint. It looks like the clinic-based studies were larger and higher quality, though. We would ask you to consider adapting the phrase "most supporting" to explicitly refer to strength of evidence, eg “…had the strongest evidence for…”.

Comments from Reviewers:

Reviewer #1: I am satisfied with the responses.

Reviewer #2: The authors have addressed all my points.

Michael Dewey

Reviewer #4: The authors have addressed all my comments.

Any attachments provided with reviews can be seen via the following link:

[LINK]

 10.1371/journal.pmed.1003434.r007 Decision Letter 4 Arikainen Artur A. Senior Editor 2020 Artur A. Arikainen This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Submission Version 4

19 Oct 2020

Dear Dr. Flood,

On behalf of my colleagues and the academic editor, Dr. Andre P Kengne, I am delighted to inform you that your manuscript entitled "Health system interventions for adults with type 2 diabetes in low- and middle-income countries: A systematic review and meta-analysis" (PMEDICINE-D-20-01702R4) has been accepted for publication in PLOS Medicine.

PRODUCTION PROCESS

Before publication you will see the copyedited word document (within 5 business days) and a PDF proof shortly after that. The copyeditor will be in touch shortly before sending you the copyedited Word document. We will make some revisions at copyediting stage to conform to our general style, and for clarification. When you receive this version you should check and revise it very carefully, including figures, tables, references, and supporting information, because corrections at the next stage (proofs) will be strictly limited to (1) errors in author names or affiliations, (2) errors of scientific fact that would cause misunderstandings to readers, and (3) printer's (introduced) errors. Please return the copyedited file within 2 business days in order to ensure timely delivery of the PDF proof.

If you are likely to be away when either this document or the proof is sent, please ensure we have contact information of a second person, as we will need you to respond quickly at each point. Given the disruptions resulting from the ongoing COVID-19 pandemic, there may be delays in the production process. We apologise in advance for any inconvenience caused and will do our best to minimize impact as far as possible.

PRESS

A selection of our articles each week are press released by the journal. You will be contacted nearer the time if we are press releasing your article in order to approve the content and check the contact information for journalists is correct. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact.

PROFILE INFORMATION

Now that your manuscript has been accepted, please log into EM and update your profile. Go to https://www.editorialmanager.com/pmedicine, log in, and click on the "Update My Information" link at the top of the page. Please update your user information to ensure an efficient production and billing process.

Thank you again for submitting the manuscript to PLOS Medicine. We look forward to publishing it.

Best wishes,

Artur Arikainen,

Associate Editor

PLOS Medicine

plosmedicine.org