Type 2 Diabetes Remission: A Systematic Review and Meta-analysis of Nonsurgical Randomized Controlled Trials

Abstract

BACKGROUND

Evidence that type 2 diabetes can be reversed has been limited by the understanding and implementation of these interventions.

PURPOSE

We assessed the effect of nonsurgical randomized controlled trials (RCTs) on type 2 diabetes remission and characterized core components.

DATA SOURCES

We reviewed articles from MEDLINE and Embase (inception to April 2025).

STUDY SELECTION

RCTs of multimodal pharmacological or nonpharmacological type 2 diabetes remission interventions for adults with type 2 diabetes were included.

DATA EXTRACTION

Study characteristics and outcomes for clinical/population health, patient-reported, and adverse event were extracted.

DATA SYNTHESIS

We performed a random-effects multilevel meta-analysis of studies, grouped based on type of intervention and by length of follow-up. A total of 18 studies were included in this review from 11 different countries. There was a higher likelihood of achieving type 2 diabetes remission through multimodal interventions (risk ratio [RR] 1.75 [95% CI 1.49–2.04]) and for nonpharmacological interventions (RR 5.80 [95% CI 4.28–7.87]), compared with the control group. Other significant outcomes for intervention groups compared with control groups included change in A1C, weight loss, and quality of life and improvements in adverse events of hypoglycemia.

LIMITATIONS

There was heterogeneity in our small pool of included studies (diversity of nonpharmacological components), stringent intervention protocols, narrow participant selection criteria, and lack of consistent diabetes remission definitions.

CONCLUSIONS

With specific protocols, a variety of tailored approaches can induce type 2 diabetes remission for patients with newly diagnosed type 2 diabetes who are able to subscribe to strict protocols. Consideration of long-term sustainability and effectiveness is needed in future research, along with patient preferences.

Graphical Abstract

Introduction

Type 2 diabetes is a global health problem, affecting more than 537 million adults worldwide as of 2021, with the number expected to rise to 783 million by 2045 (1). Currently, type 2 diabetes management involves health-behavior changes (2) and glucose-lowering medications in response to rising glucose levels and failing therapeutic regimens (3). Evidence that type 2 diabetes can be completely or partially reversed through combination of dietary, physical activity, intensive metabolic, and pharmacotherapy strategies (4–6) is now challenging the traditional management paradigm (4,7).

Type 2 diabetes remission (defined broadly as a return to glycemic levels below the threshold of diagnosis without the ongoing need for glucose-lowering medications [5]) is a rapidly evolving area of research. Various approaches to achieve remission have been studied, including bariatric surgery and pharmacological/multimodal interventions with use of intensive insulin therapy (8). For both types of interventions, multimodal pharmacological or nonpharmacological, the pathophysiological effect on β-cell capacity is more pronounced in the early stages of type 2 diabetes diagnosis (9). There is also a strong dose-response relationship between body weight loss and diabetes remission; every 1% decrease in body weight increases the probability of reaching remission by 2% (10). Loss of >10% of baseline body weight in the first year after diagnosis of type 2 diabetes is associated with a 70% higher chance of remission at 5 years (11). Thus, interventions with the aim of achieving type 2 diabetes remission are targeted at individuals who have been recently diagnosed with type 2 diabetes, although the definition of recently diagnosed varies across studies (11).

Long-term evidence demonstrates that maintenance of weight loss and remission decreases over time (12), particularly in the case of nonpharmacological interventions, where sustained behavior change is difficult to maintain outside of the resource-rich intervention environment. In addition, the implementations of these interventions have been limited (13). Type 2 diabetes remission implementation with diverse populations is poorly understood, with conversations that can involve shame and stigma (14), and little is known about supportive and practical, person-centered strategies. Current type 2 diabetes remission research has been focused mainly on demonstrating effectiveness, but no other review has included meta-analysis of multimodal pharmacotherapy interventions for type 2 diabetes remission.

In this article we assess the effectiveness of pharmacological and nonpharmacological type 2 diabetes remission interventions in this evolving field of research. Specifically, our objectives are to 1) assess the effect of pharmacological and nonpharmacological randomized controlled trial (RCT) interventions on type 2 diabetes remission and 2) characterize the components of these type 2 diabetes remission interventions.

Research Design and Methods

Data Sources and Searches

We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines from a registered protocol (International prospective register of systematic reviews [PROSPERO], reg. no. CRD42024511339), and the Cochrane Handbook for Systematic Reviews of Interventions, Version 6, 2019. The search terms, databases, and strategy were developed with a research librarian (Supplementary Appendix 1). We searched MEDLINE and Embase from inception to April 2025.

Study Selection

Studies were included if published in a peer-reviewed journal, in English, and according to the following criteria: 1) RCT (patient or cluster level), 2) adults ≥18 years of age with type 2 diabetes (diagnosed on the basis of included study criteria), 3) any pharmaceutical/drug and/or nonpharmacological (behavioral/lifestyle) intervention for type 2 diabetes remission, and 4) a control group (usual care, standard care, or minimal contact) without intervention components. For inclusion, type 2 diabetes remission was defined by the authors of the studies screened; surgical interventions were not included in this review, but information on surgical interventions related to this review has previously been published (15).

We identified and sorted available outcomes (see Supplementary Appendix 2) based on the Quadruple Aim framework (16) or safety/adverse event outcomes. This framework was developed to optimize health system performance and includes improved patient experience (patient-reported outcomes), better clinical/population health outcomes, lower costs (cost of care outcomes), and improved clinician experience (health care provider experience). Next, the team ranked outcomes (primary, secondary, or exploratory) based on Grading of Recommendations Assessment, Development and Evaluation (GRADE) methodology. Outcomes that were ranked as critical (for meta-analyses) were as follows: clinical/population health outcomes (type 2 diabetes remission, body weight/weight loss, type 2 diabetes relapse, glycated hemoglobin [A1C]), patient-reported outcome measures (adherence/program retention and acceptability, changes in medications, quality of life), and safety/adverse event outcomes (hypoglycemia and serious adverse events).

Data Extraction and Quality Assessment

A team conducted the screening and data extraction (D.T.S., M.E.R., M.K.G., P.E.A., Ruth Lewis, Milos Jovkovic, and H.C.G.) independently and in duplicate; any inconsistencies were discussed and resolved among the two reviewers and verified by a third reviewer. We used the DistillerSR AI tool to check for screening errors (17) against reviewed references to identify any that might have been erroneously excluded.

For each study, we used standardized forms to extract study characteristics, risk of bias assessment, the Template for Intervention Description and Replication (TIDieR) checklist and guide (18), and outcome data. Study characteristics included study details (e.g., country, setting, year), participant characteristics (e.g., age, sex/gender, ethnicity, socioeconomic status, comorbidities), intervention and comparison characteristics (e.g., inclusion/exclusion criteria, duration), and outcomes (e.g., definition and measurement). To elicit implementation information, we used the TIDieR tool. TIDieR is used to elucidate details of the core and adaptable components of an intervention of a study. In cases where studies had multiple measures for the same outcome, primary or direct measures were used before secondary outcomes or subgroup analysis data. Two team members independently verified all extracted data, and disagreements were resolved through discussion and/or third-party consultation.

Data Synthesis and Analysis

All data analyses were planned a priori with use of the published data from included studies. Analyses were conducted for our main outcome, type 2 diabetes remission, for all available time points and for secondary outcomes postintervention/posttreatment and the same time point when type 2 diabetes remission was first measured, to ensure these outcomes were considered for after/at the same time that remission may first occur. This was important for multimodal pharmacological interventions, as a period with no medications is considered indicative of type 2 diabetes remission.

For dichotomous outcomes of interest, we used both raw data (number of events, prevalence) and study-reported adjusted measures of effect (i.e., odds ratio, RR [RR], or hazard ratio [HR]). For consistency in quantitatively synthesizing results across studies, we used RRs or HRs as equivalent, as a summary measure of effect for dichotomous outcomes with use of a random-effects model (DerSimonian and Laird method).

For continuous outcomes, we used both raw change from baseline to posttreatment or follow-up and immediate posttreatment data (i.e., mean score along with its measure of variance) and study-reported adjusted measures of effect (i.e., mean differences or standardized mean differences), as well as summarized measures of effect (mean differences or standardized mean differences).

The analysis was based on the type of remission intervention (grouped as either multimodal pharmacological or nonpharmacological) and by the length of follow-up (such as immediate posttreatment, 6 to 12 months, longer than 12 months). The cut points of 6 and 12 months were determined based on the evidence from supporting literature and relevance for best practice of diabetes management.

We used a three-level random-effects meta-analysis (where applicable) instead of conventional two-level random-effects model to account for dependency across effect sizes (i.e., the correlation between effect estimates due to multiple comparison arms or multiple measures of the outcome within the same study) and to avoid possible unit of analysis error. In such cases, possibly correlated effect estimates were nested at the within-study level first. The variance in observed effect estimates was further decomposed into sampling variance, within-study variance, and between-study variance to account for intracluster (or intraclass) correlation in the true effects. Overall heterogeneity statistic I² for each summary effect estimate was estimated to represent the heterogeneity not attributable to sample error, as the sum of within-cluster (i.e., across effect sizes or multiple arms from the same study) and between-cluster (i.e., effect sizes across unique studies) heterogeneity. The Cochran Q test (α = 0.05) was used to detect statistical heterogeneity and I² statistic to quantify the magnitude of statistical heterogeneity between studies (I² > 50% represents moderate and I² > 75% represents substantial heterogeneity). Data syntheses were conducted in R software (metafor and dmetar packages) and the DataParty statistical platform.

For further exploration of heterogeneity, statistical stability, and robustness of our pooled results, further sensitivity and meta-regression analyses were done based on various study-level factors and definitions of type 2 diabetes remission (Diabetes Canada vs. American Diabetes Association) (5,19). Where meta-analysis was not possible, a narrative synthesis was reported (TIDieR tool).

Ethics Considerations

As this study was solely literature based and did not involve any research participants or subjects, no formal ethics approval from the McMaster Research Ethics Board (Hamilton, Ontario, Canada) was required.

Data and Resource Availability

Information on the the original contributions presented in the study can be found in the article/appendices. Further inquiries can be directed to the corresponding author.

Results

Our search yielded 6,470 unique citations; 53 full-text citations were reviewed and 18 RCTs were included in this review (Fig. 1) published between 2008 and 2025 from 11 countries. Most studies were published between 2020 and 2025 (n = 11). Eight studies were multimodal pharmacological interventions (7,20–26), and 10 were nonpharmacological interventions (12,27–35). In total, our review included 6,433 total participants for nonpharmacological studies and 1,488 total participants for multimodal pharmacological studies in both the intervention and control groups combined at baseline. Most studies included individuals diagnosed with type 2 diabetes, with diabetes duration spanning from as short as 14.6 months since diagnosis to 6.7 years and total mean A1C at baseline 6.6%–10.4% (49 mmol/mol to 90 mmol/mol). Participants were generally insulin naïve (n = 10 studies), but five studies included participants on insulin (ranging from 1% to 96% of participants at baseline) and for three studies there was lack of clarity regarding the insulin use of participants. Study characteristics are shown in Table 1 and Supplementary Appendix 3. Risk of bias analysis found mostly low risk of bias for the included studies (n = 12), for both multimodal pharmacological and nonpharmacological intervention studies (Supplementary Appendix 4). There were no trends related to publication date or risk of bias assessment.

Figure 1.

PRISMA 2020 flow diagram for new systematic reviews, including searches of databases, registers, and other sources. Adapted from Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021;372:n71. For more information, visit https://www.prisma-statement.org/.

Table 1.

Characteristics of included studies by intervention type

First author, year, country (reference no.)	Study design and duration	T2D remission definitiona	Sample size: total; I; C	T2D duration: total; I; C	Baseline BMI, kg/m2: total; I; C	Baseline A1C, %: total; I; C	Intervention category
Gregg, 2012, U.S. (27)	RCT, 12 months*	Full normalization of glucose (FPG <100 mg/dL and A1C <5.7%) with no antihyperglycemic medication	5, 145; 2,570; 2,575	5 years; 5 (8) years; 5 (7) years**	35.8 (NR); 35.8 (5.9); 35.9 (5.7)	NR; 7.3 (1.1); 7.4 (1.2)	Physical activity, nutrition, psychological
Hanick, 2025, Republic of Marshall Islands (35)	RCT, 24 weeks	A1C <48 mmol/mol (<6.5%) after not using glucose-lowering medications for at least 3 months	169; 79; 90	NR; NR; NR	29.8 (4.9); 30.5 (4.8); 29.2 (4.9)	10.4 (2.1); 10.2 (2.1); 10.5 (2.1)	Physical activity, nutrition
Joshi, 2023, India (28)	RCT, 12 months	A1C <6.5% without any glucose-lowering pharmacotherapy for at least 3 months	336; 250; 86	NR; 3.8 (2.6) years; 4.6 (3.3) years	NR; 27.2 (4.4); 28.3 (4.3)	NR; 9.0 (1.9); 8.5 (1.9)	Physical activity, nutrition, psychological, technology
Lean, 2018, England (12)	RCT (cluster), 12 months*	A1C <6.5% (<48 mmol/mol) after at least 2 months off all glucose-lowering medications	306; 150; 149	NR; 3.0 (1.7) years; 3.0 (1.9) years	NR; 35.1 (4.5); 34.2 (4.3)	NR; 7.7 (1.25); 7.5 (1.05)	Physical activity, nutrition (TDR)
Ried-Larsen, 2019, Denmark (29)	RCT, 12 months*	All three of the following criteria: fasting glucose ≤6.9 mmol/L, A1C <48 mmol/mol (6.5%), no glucose-lowering medications at the outcome assessments	98; 64; 34	5.1 (2.9) years; 4.7 (2.8) years; 5.6 (3.2) years	NR; 31.5 (3.9); 32.3 (4.4)	NR; 6.6 (3.0); 6.7 (3.0)b	Physical activity, nutrition
Romadlon, 2025, Indonesia (34)	RCT, 12 weeks*	A1C levels <6.5% for 3 months	84; 28, 28; 28	NR; 6.2 (3.0) years, 6.5 (2.8) years; 6.0 (2.3) years	NR; 26.3 (2.5), 25.0 (2.6); 25.2 (2.3)	NR; NR, NR; NR	Physical activity, nutrition, technology
Sattar, 2023, U.K. (30)	RCT, 3–5 months	A1C <48 mmol/L off diabetes treatment for at least 3 months	25; 13; 12	1.9 years; NR; NR	32.1 (3.8); 32.6 (4.4); 31.6 (3.3)	7.7 (3.2); 7.3 (2.7); 8.0 (3.4)b	Nutrition (TDR)
Taheri, 2020, Qatar (31)	RCT, 12 months*	A1C values <6.5% (<48 mmol/mol) and no pharmacological therapy for diabetes for at least 3 months	158; 79; 79	21.2 (12.3) months; 21.9 (11.5) months; 20.5 (13.0) months	34.9 (5.5); 35.0 (5.2); 34.8 (5.8)	7.0 (1.3); 6.95 (1.4); 6.95 (1.2)	Physical activity, nutrition (TDR), technology
Umphonsathien, 2022, Thailand (32)	RCT, 20 weeks	FPG <126 mg/dL and A1C level <6.5% in the absence of pharmacological therapy for diabetes	40; 14, 14; 12	4.9 (3.1) years; 5.5 (3.0) years, 3.1 (2.8) years; 5.2 (3.2) years	30.9 (5.9); 29.9 (1.6), 31.0 (1.6); 29.1 (1.7)	7.4 (1.1); 7.5 (0.3), 7.7 (0.3); 6.9 (0.3)	Nutrition
Yang, 2023, China (33)	RCT, 3 months*	Stable A1C levels <6.5% (48 mmol/mol) after discontinuation of antidiabetes medication for at least 3 months	72; 36; 36	NR; 6.7 (3.3) years; 6.5 (3.2) years	NR; 24.2 (2.6); 23.9 (2.5)	NR; 7.6 (1.8); 7.5 (1.3)	Nutrition
Hirukawa, 2018, Japan (20)	RCT, 18 months*	A1C <6.2% during the treatment phase and maintained during off-treatment phase	290; 101, 93; 84	3.8 (0.2) years; 3.6 (0.3), 4.3 (0.3) years; 3.6 (0.3) years	24.4 (0.2); 24.4 (0.4), 24.7 (0.4); 24.2 (0.4)	6.94 (0.03); 6.98 (0.05), 6.91 (0.06); 6.93 (0.05)	Thiazolidinediones, sulfonylureas
McInnes, 2017, Canada (21)	RCT (pilot), 28 weeks	A1C <6.0% (42 mmol/mol) and no chronic diabetes drugs	83; 28, 27; 28	14.6 (10.6) months; 12.6 (10.1), 15.3 (10.2) months; 16.0 (11.4) months	33.2 (5.8); 34.7 (7.0), 33.3 (5.5); 31.6 (4.4)	6.6 (0.6); 6.5 (0.4), 6.6 (0.1); 6.6 (0.6)	Metformin, insulin, other drug (acarbose), physical activity, nutrition
McInnes, 2020, Canada (24)	RCT, 12 weeks*	A1C <6% (<42 mmol/mol) without the need for glucose-lowering medications for at least 12 weeks	154; 77; 77	36.8 (27.2) months; 37.7 (26.7) months; 35.8 (27.8) months	32.9 (5.9); 33.3 (5.8); 32.5 (6.0)	6.7 (0.6); 6.7 (0.6); 6.7 (0.7)	Metformin, insulin, SGLT2 inhibitors, physical activity, nutrition
McInnes, 2022, Canada (22)	RCT, 12 weeks*	Absence of relapse after the 12-week intervention period defined as any of the following: ≥50% of capillary glucose values in any week >10 mmol/L in absence of illness, A1C ≥6.5%, use of diabetes drugs, FPG ≥7.0 mmol/L, 2-h postprandial plasma glucose ≥11.1 mmol/L on an OGTT	102; 50; 52	24 (16.9) months; 26.4 (17.3) months; 21.7 (16.3) months	31.6 (4.7); 30.9 (4.7); 32.2 (4.8)	6.6 (0.7); 6.6 (0.6); 6.6 (0.8)	Metformin, insulin, DPP-4 inhibitor, physical activity, nutrition
McInnes, 2023, Canada (23)	RCT, 12 weeks*	A1C <6.0% (42 mmol/mol); off diabetes medications for 12 weeks	160; 79; 81	24.7 (16.3) months; 25.9 (15.9) months; 23.6 (16.8) months	NR; 31.9 (5.7); 34.0 (8.4)	6.8 (0.7); 6.8 (0.7); 6.9 (0.7)	Metformin, insulin, SGLT2 inhibitor, physical activity, nutrition
Punthakee, 2024, Canada (25)	RCT, 16 weeks*	A1C <48 mmol/mol (6.5%) at least 12 weeks after stopping and remaining off glucose-lowering drugs and a median weekly self-monitored glucose ≤10.0 mmol/L (180 mg/dL)	159; 79; 80	2.6 (1.5) years; 2.5 (1.5) years; 2.6 (1.5) years	33.5 (6.5); 33.0 (5.9); 33.9 (7.1)	7.0 (0.5); 7.0 (0.5); 7.0 (0.5)	Metformin, insulin, GLP-1 receptor agonists, physical activity, nutrition, psychological, technology, other lifestyle (health coaching)
Shi, 2017, China (26)	RCT, 12 months*	FPG ≤7.0 mmol/L or 2-h PPG ≤10.0 mmol/L during on-treatment phase and maintained during off-treatment phase	130; 67; 63	NR	25.0 (3.2); 25.2 (0.8); 24.7 (0.8)	10.3 (2.8); 9.9 (0.6); 10.2 (0.6)	Insulin, GLP-1 receptor agonists, physical activity, nutrition
Weng, 2008, China (7)	RCT, 12 months	FPG ≤7.0 mmol/L or 2-h PPG ≤10.0 mmol/L during on-treatment phase and maintained during off-treatment phase	410; 137, 124; 121	NR	25.0 (3.0); 25.1 (3.0), 24.4 (2.7); 25.1 (3.3)	NR; 9.8 (2.3), 9.7 (2.3); 9.5 (2.5)	Metformin, insulin, sulfonylureas, physical activity, nutrition

Data are means (SD) unless otherwise indicated. A1C, glycated hemoglobin; C, control; FPG, fasting plasma glucose; I, intervention; NR, not reported; OGTT, oral glucose tolerance test; PPG, postprandial glucose; TDR, total diet replacement provided.

^aDiabetes remission definition is for complete remission if both partial and complete remission outcomes were provided.

^bConverted mmol/mol to % for A1C.

*With follow-up time points from postintervention.

**Data were reported as median (IQR).

Type 2 Diabetes Remission Intervention Characteristics

We described the type 2 diabetes remission interventions using the TIDieR checklist, which consists of 12 domains (Supplementary Appendix 5). For all of our included studies the publications reported a name for the intervention, along with a goal/aim for the study and methods or procedures of the study. Study aims were focused on the effectiveness of interventions; however, one pilot trial was focused on the feasibility and safety of the protocol to induce type 2 diabetes remission (21) and in one study investigators looked at whether the intervention was acceptable to a South Asian population (30). Few studies included reporting on modifications to the study protocol (n = 3) or explicit mention of assessing intervention delivery and fidelity, either planned (n = 3) or actual (n = 2). In two studies the impact of the pandemic was noted along with modifications needed due to restrictions (30), such as changing more than one-third of coaching visits from in-person to telephone contact (25). Three nonpharmacological studies (12,31,34) included stated plans to measure how well the intervention was delivered through ongoing observations, check-in calls, or interviews with participants and health care professionals, and there was only one study (34) where it was reported that all participants completed the peer support sessions.

Multimodal Pharmacological Interventions

In the included studies a variety of diabetes drugs and other components were used to deliver the type 2 diabetes remission intervention. In one study only pharmaceutical management was used (20), while seven studies included use of multimodal combinations of diabetes drugs with physical activity, nutrition/diet components, and health coaching. Five included studies were part of the same research intervention, with investigation of different combinations of diabetes drugs to induce remission in the populations (21–25). In these five studies metformin and insulin were used to “rest” the pancreas, while in each study a different pharmaceutical was investigated along with behavioral components. Intervention characteristics can be found in Table 1 and Supplementary Appendix 5.

All pharmacological studies included regulated health care professionals (e.g., nurses, kinesiologists, and registered dietitians) to deliver the interventions in diabetes centers or hospitals, likely due to medication prescribing, and deprescribing protocols that require medical directives. The interventions were delivered in person or via telephone calls. For one study (21) there was mention of a group component to the intervention for group exercise classes. Publications for only five multimodal pharmacological studies stated what materials, if any, were used during the intervention (beyond the pharmaceutical agents) such as pedometers to track activity and a continuous glucose monitor (25). While tailoring was mentioned in four studies, this was mostly for the tailoring of physical activity or nutrition components of the study.

Nonpharmacological Interventions

Overall, more components of the TIDieR checklist were reported on in studies of nonpharmacological interventions. In three included studies nutrition-only interventions were used (30,32,33): in one, a total diet replacement with a calorie intake of 825–853 kcal/day (30) and in two, low-calorie/very-low-calorie diets (600–800 kcal/day) during cycled periods of fasting, with the option for some meal replacement shakes for one of these studies (32). The other seven studies in our review involved nutrition intervention components combined with physical activity. Three studies also included total diet replacement (12,27,31). In three studies technology components were added to the diet and physical activity intervention (28,31). One study included wrist-worn accelerometers and smartphone apps to monitor food intake and activity (31); another study was a technology- and AI-focused intervention (28) that included wearables, AI data processing, and AI-enabled nudges along with human health coaching; and in the third study personalized text messages were provided to participants (34). Lastly, in one study psychological components were added to the intervention through a lifestyle counselor who provided support with individual barriers and needs. Details on the intervention characteristics can be found in Table 1 and Supplementary Appendix 5.

Nonpharmacological studies also included regulated health care professionals such as nurses, physicians, and registered dietitians or trained peer supporters. Of the nine studies with specification of how the intervention was delivered, nine included face-to-face visits and eight included telephone and/or email. In three studies the use of group sessions, to deliver the physical activity and/or nutrition intervention components, was mentioned. The interventions were delivered in the contexts of primary care offices, hospitals, community health/diabetes centers, free-living/home, and recreational/fitness centers. All the nonpharmacological intervention studies included reporting on what materials were provided to the participants (n = 10), meal replacement products/shakes, prepared meals, fitness trackers such as smartwatches or pedometers, fitness center memberships, body weight scales, and/or continuous glucose monitors. Tailoring of the intervention and/or study components was reported in seven nonpharmacological intervention studies and included individualized dietary plans or physical activity plans, cultural/ethnic considerations, and participant circumstances, and in one study AI was used to select behavioral nudges that were specific to the participant (28).

Benefits of Treatment

We extracted and categorized outcomes based on the Quadruple Aim framework (16) and were able to meta-analyze outcomes for clinical population health outcomes, patient-reported outcomes, and safety outcomes/adverse events. See Supplementary Appendix 6 for patient-reported outcomes and a summary table of all meta-analyses.

Clinical/Population Health Outcomes

In using the definitions provided by the authors of the included studies (shown in Table 1), there was a higher likelihood of achieving type 2 diabetes remission in the intervention groups than in the control groups both for multimodal pharmacological studies (RR 1.75 [95% CI 1.49–2.04]) and for nonpharmacological studies (RR 5.80 [95% CI 4.28–7.87]) at any point after the intervention (Fig. 2A and B). RRs for achieving type 2 diabetes remission did not vary at different time points for multimodal pharmacological interventions; however, the likelihood of achieving type 2 diabetes remission was lower at longer follow-up time points for nonpharmacological interventions. Participants in four multimodal pharmacological studies were 64% less likely to experience type 2 diabetes relapse compared with control participants (HR 0.64 [95% CI 0.53–0.76]). Similarly, participants in one nonpharmacological study also showed a reduced risk of type 2 diabetes relapse compared with control participants (RR 0.62 [95% CI 0.45–0.85]) (data not shown).

Figure 2.

The effect of multimodal pharmacological (A) and nonpharmacological (B) interventions on type 2 diabetes remission (any definition). Intervention and control columns denote the number of remission events per total group participants. Immediate post, immediately postintervention.

Data from five multimodal pharmacological studies on body weight at postintervention showed a significant decrease of large magnitude in percent body weight loss of 1.84 (95% CI 1.1–2.59) in comparison with control individuals (Fig. 3A). Data from eight nonpharmacological studies on body weight postintervention showed a significant improvement of large magnitude in the change in weight (in kilograms) of −5.05 (95% CI −7.10 to −3.01) in comparison with control (Fig. 3B). All data presented are mean difference with 95% CIs.

Figure 3.

A and B: The effect of multimodal pharmacological and nonpharmacological interventions on weight loss as percent body weight loss (A) or change from baseline (kilograms) (B). Weight change in kilograms was not presented for pharmacological interventions, and weight loss as percent body weight was only reported from two studies with nonpharmacological interventions. C and D: The effect of nonpharmacological interventions on A1C, postintervention (C), and as change from baseline (D). A1C results from pharmacological interventions were nonsignificant. MD, mean difference.

Data from nine nonpharmacological studies showed a significant improvement of large magnitude in A1C (%) postintervention (−0.98 [95% CI −1.41 to −0.55]) and A1C change from baseline to postintervention (−1.16 [95% CI −1.64 to −0.68]) in comparison with control (Fig. 3C and D). For multimodal pharmacological interventions, data from seven studies showed a nonsignificant improvement in A1C (%) in comparison with control measured postintervention (−0.13 [95% CI −0.27 to 0.01]) and A1C change from baseline to postintervention (−0.13 [95% CI −0.27 to 0.00]) (data not shown). All data presented are mean difference with 95% CIs.

Safety Outcomes/Adverse Events

Data from seven multimodal pharmacological studies showed that participants in the intervention group were more likely to experience mild-to-moderate hypoglycemia events in comparison with the control group participants (RR 2.95 [95% CI 1.95–4.47]), but there were no differences between groups for severe hypoglycemia events, as there were not any severe hypoglycemic events in the studies (data not shown). For both multimodal pharmacological studies and nonpharmacological studies, there were no significant differences between intervention and control groups concerning study-reported serious adverse events (data not shown).

Cost of Care

Data from one study showed that participants in the intervention group required fewer medications at a lower cost per month (33). In comparison with baseline, medication costs decreased by 77.2% for intervention group participants and by only 3.0% for the control group.

Conclusions

A total of 18 RCTs, comprising multimodal pharmacological and nonpharmacological type 2 diabetes remission interventions, contributed to this review. Across all studies, there was a higher likelihood of achieving type 2 diabetes remission through multimodal pharmacological interventions (RR 1.75 [95% CI 1.49–2.04]) and through nonpharmacological interventions (RR 5.87 [95% CI 4.96–6.94]) at any point, in comparison with the control group. Although achieving type 2 diabetes remission did not vary greatly for multimodal pharmacotherapy interventions across different time points, the likelihood of maintaining type 2 diabetes remission with nonpharmacotherapy type 2 diabetes remission interventions appeared to wane. There was a paucity of evidence, with reporting from only one study of any outcomes related to cost and no studies including consideration of provider satisfaction.

In our review of the literature we also examined type 2 diabetes remission approaches and the core components of interventions. Nonpharmacological interventions mainly comprised total or major meal replacement strategies to induce caloric reduction, total intake ranging from 600 to 853 kcal/day; using periods of fasting was only reported in one study. Five studies included physical activity, with two of the studies including movement trackers and/or technology-enhanced reminders to encourage movement. The pharmacological intervention studies were complex, with use of pharmaceutical agents and physical activity and nutrition components. Generally, greater detail was provided for nonpharmacological interventions in terms of the intervention components. Overall, the reporting of intervention components, with use of the TIDieR tool, aligns with findings of other type 2 diabetes reviews that have also included assessments of how type 2 diabetes studies are being reported and implemented (36).

Our work included a comprehensive and systematic review of both nonpharmacological and pharmacological type 2 diabetes remission trials. No other reviews have included meta-analysis of multimodal pharmacotherapy interventions for type 2 diabetes remission; however, others have summarized the potential effectiveness of such studies on an individual level (37). Highly intensive behavioral interventions can be more effective for managing type 2 diabetes, as research indicates that structured and intensive lifestyle intervention protocols, which include frequent follow-ups, can lead to significant improvements in clinical indicators such as glycemic control and weight management (6). This aligns with our meta-analysis, with a higher RR for nonpharmacological interventions than for multimodal pharmacological interventions. It is important to note that for many of the nonpharmacological studies, some level of adherence or a percent weight loss was required of participants for them to be allowed to participate or continue in the study (28,31,32). For some studies investigators also set drop-out rates of 25% in their power calculations, which may be indicative of the expected nonacceptability of such multicomponent interventions (12). The bias in the case of participants who are allowed to continue in the study leads to questions about whether these intense interventions can be sustained long-term, despite the relatively low drop-out rates.

Health benefits from weight management, which is related to type 2 diabetes remission, depend largely on long-term control of body weight (10). Thus, the sustainability of such interventions and their effects is important, and our results showed a small waning of effectiveness across time for nonpharmacotherapy studies. While in other work there has been findings that nonpharmacotherapy interventions, such as a total diet replacement protocol (formula diets provided in studies) with behavioral support, may be more cost-effective than standard care (38), consideration of individual needs, preferences, cultures, and adherence is needed as well as consideration of sustained costs to individuals. Food is important for personal and social well-being, and some of these intervention diets can be psychologically demanding (39). Overall, these type 2 diabetes remission studies are likely effective due to the unique combination of patient characteristics and strong commitments to the intervention components, which may not happen in real-world situations. Yet, what we are seeing is a rise in private health clinics offering these interventions without full research or consideration of implementation. For long-term success of these interventions, shared decision-making is needed between health care professionals and patients that includes consideration of patient preferences, culture, context, and lifestyle (37).

Definitions of type 2 diabetes remission were not consistent across studies, which was previously highlighted as a limitation of this field of research (37,40) and further complicates the interpretation of these interventions. Findings of a recent systematic review of type 2 diabetes remission literature showed nearly 100 different definitions of remission in type 2 diabetes (40). While groups have worked to develop consensus on type 2 diabetes remission definitions, discordance remains even among international expert groups (5,19). In most studies in our review an A1C threshold of <6.5% was used for type 2 diabetes remission (7,12,25,26,28–33); <6.0% was used for type 2 diabetes remission in only three studies (21,23,24). Based on the literature included in our review, questions remain about the clinical and patient significance of having A1C <6.0% compared with <6.5% and whether A1C <6.0% is achievable in real-world settings, considering that the criteria for type 2 diabetes remission used in most of these highly controlled RCTs included an A1C of <6.5%.

Our review was comprehensive, with searching of multiple databases and leveraging the expertise of researchers, patient partners, and health care professionals. Although there is no universally accepted definition of type 2 diabetes remission, we chose to accept author-defined criteria of type 2 diabetes remission and conducted sensitivity and meta-regression analyses in consideration of whether studies aligning with the Diabetes Canada definition resulted in contradictory results—it did not. We were unable to conduct tests for risk of bias due to the small number of studies in this emerging field that met the criteria for our review. Across studies in our review, there were risk of bias concerns related to the blinding of participants. Ideally, type 2 diabetes remission would be a primary end point because in these studies investigators are looking at efficacious and effective practice, and this was the case for 13 of our included studies. While type 2 diabetes remission has many definitions, they all overlap with specific components (e.g., glucose control may be a primary outcome in some studies and remission is a secondary outcome, and a component of the definition of remission is being assessed as the primary outcome), which may increase the power of remission as a secondary outcome. Lastly, we have noted that there is heterogeneity in our included studies, mostly with the diversity of nonpharmacological intervention components, but we could not perform any meta-regression analysis based on study-level factors as there were too few studies for such analysis.

Conclusion

Our review and meta-analysis confirmed the effectiveness of multicomponent and complex type 2 diabetes remission interventions. The evidence suggests that a variety of approaches, including pharmacotherapy, diet, physical activity, health coaching, and psychological components, may induce type 2 diabetes remission for patients with newly diagnosed type 2 diabetes when applied in optimal conditions and with participants ideally subscribing to strict protocols. Future research should include consideration of long-term sustainability and effectiveness of these interventions, along with patient preferences and real-world aspects, to improve generalizability of these protocols and reach consensus on a type 2 diabetes remission definition.

This article contains supplementary material online at https://doi.org/10.2337/figshare.29822177.

Appendix

Integrated Knowledge Translation (iKT) Type 2 Diabetes Remission Advisory Team

Members include Diana Sherifali (principal investigator), Monika Kastner (knowledge mobilization lead), Muhammad Usman Ali (collaborator), Paige Alliston (PhD trainee), Jennifer Brown (knowledge user), Laura Chiavaroli (co-investigator), Kathleen Chouinor (knowledge user), Kaberi Dasgupta (co-investigator), Ryan Emond (collaborator), Mark Ewer (lived experience partner), Donna Fitzpatrick-Lewis (lived experience partner), John-Michael Gamble (co-investigator), Hertzel Gerstein (co-investigator), Kristin Honshort (knowledge user), James Kim (knowledge user), Krista Lamb (knowledge user), Lorraine Lipscombe (co-investigator), Julie Makarski (collaborator), Natalia McInnes (co-investigator), Sean McKelvey (collaborator), Brian McKenna (knowledge user), Tracy McQuire (knowledge user), Kara Nerenberg (co-investigator), Zubin Punthakee (co-investigator), Doreen Rabi (collaborator), Megan Racey (collaborator), and Julia Roglich (knowledge user).

Funding Statement

This research was supported by the McMaster Evidence Review and Synthesis Team and the Canadian Institutes of Health Research Knowledge Mobilization Grant (KM2-196997). In addition, D.T.S. holds the Heather M. Arthur Population Health Research Institute/Hamilton Health Sciences Chair in Interprofessional Health Research, which also provided support for M.E.R.