Defining remission of type 2 diabetes in research studies: A systematic scoping review

Background Remission has been identified as a top priority by people with type 2 diabetes. Remission is commonly used as an outcome in research studies; however, a widely accepted definition of remission of type 2 diabetes is lacking. A report on defining remission was published (but not formally endorsed) in Diabetes Care, an American Diabetes Association (ADA) journal. This Diabetes Care report remains widely used. It was the first to suggest 3 components necessary to define the presence of remission: (1) absence of glucose-lowering therapy (GLT); (2) normoglycaemia; and (3) for duration ≥1 year. Our aim is to systematically review how remission of type 2 diabetes has been defined by observational and interventional studies since publication of the 2009 report. Methods and findings Four databases (MEDLINE, EMBASE, Cochrane Library, and CINAHL) were searched for studies published from 1 September 2009 to 18 July 2020 involving at least 100 participants with type 2 diabetes in their remission analysis, which examined an outcome of type 2 diabetes remission in adults ≥18 years and which had been published in English since 2009. Remission definitions were extracted and categorised by glucose-lowering therapy, glycaemic thresholds, and duration. A total of 8,966 titles/abstracts were screened, and 178 studies (165 observational and 13 interventional) from 33 countries were included. These contributed 266 definitions, of which 96 were unique. The 2009 report was referenced in 121 (45%) definitions. In total, 247 (93%) definitions required the absence of GLT, and 232 (87%) definitions specified numeric glycaemic thresholds. The most frequently used threshold was HbA1c<42 mmol/mol (6.0%) in 47 (20%) definitions. Time was frequently omitted. In this study, a total of 104 (39%) definitions defined time as a duration. The main limitations of this systematic review lie in the restriction to published studies written in English with sample sizes of over 100. Grey literature was not included in the search. Conclusions We found that there is substantial heterogeneity in the definition of type 2 diabetes remission in research studies published since 2009, at least partly reflecting ambiguity in the 2009 report. This complicates interpretation of previous research on remission of type 2 diabetes and the implications for people with type 2 diabetes. Any new consensus definition of remission should include unambiguous glycaemic thresholds and emphasise duration. Until an international consensus is reached, studies describing remission should clearly define all 3 components of remission. Systematic review registration PROSPERO CRD42019144619


Non-standard Abbreviations
2hr PG: two hour plasma glucose, ADA: American Diabetes Association, DG: diagnosis guidelines, FPG: Fasting Plasma Glucose, GLT: glucose-lowering therapy, PCOS: polycystic ovarian syndrome, PICOS: Patient, problem or population, Intervention, Comparison, control or comparator, Outcome, Setting, PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analyses, UK: United Kingdom, WHO: World Health Organisation,

ABSTRACT Background
Remission has been identified as a top priority by people with type 2 diabetes. Remission is commonly used as an outcome in research studies, however, a widely accepted definition of remission of type 2 diabetes is lacking. A report on defining remission was published (but not formally endorsed) in Diabetes care, an American Diabetes Association (ADA) journal. This Diabetes care report remains widely used. It was the first to suggest three components necessary to define the presence of remission: (1) absence of glucose-lowering treatment (2) normoglycaemia (3) for duration ≥1 year. Our aim is to systematically review how remission of type 2 diabetes has been defined by observational and interventional studies since publication of the 2009 report.  What did the researchers do and find?

Methods and Findings
 We systematically reviewed 178 research studies with an outcome of type 2 diabetes remission that had been published between 2009-2020 to evaluate how remission of type 2 diabetes had been defined.
 There were 96 unique definitions of type 2 diabetes remission, with substantial heterogeneity.
 93% of type 2 diabetes remission definitions stated that all glucose lowering therapy should be stopped. 87% included a numeric glycaemic threshold, but the exact threshold and combination of tests to measure normal glycaemic levels varied considerably. 61% did not include duration of stopping therapy or normal glycaemic levels as part of the definition of type 2 diabetes remission.
What do these findings mean?
 The heterogeneity of type 2 diabetes remission definitions in the recent research literature highlights a fundamental problem for remission research.
 Synthesising the findings of studies of remission will be limited by variation in definition. Future meta-analyses should ideally use individual participant data.
 Categorising people into remission and non-remission groups in a way that is consistent with the current type 2 diabetes diagnosis guidelines is challenging and should be carefully considered before integrating remission with clinical type 2 diabetes management pathways  Researchers should always clearly and unambiguously state their definition of remission. There is a clear need for an internationally accepted and routinely implemented definition of remission.

INTRODUCTION
Type 2 diabetes is a global health priority [1]. By 2045 an estimated 629 million people will be affected by diabetes of whom 90-95% will have type 2 diabetes [1]. Drivers for the twin epidemics of obesity and diabetes lie in complex interactions between obesogenic environments, a biological tendency for weight gain, and an ageing population [2,3]. The rapid spread of obesity and diabetes worldwide has considerable health implications for the individual and major financial consequences for health services [3]. In 2015, the global economic burden of diabetes was estimated to be US $1.3 trillion [4] and total costs are estimated to rise to $2.2-2.5 trillion by 2030 [4]. Type 2 diabetes has conventionally been considered a lifelong, progressive disease [5]. The concept of reversing type 2 diabetes by metabolic surgery was introduced in the early nineties [6]. Buchwald et al strengthened this concept by defining and demonstrating the resolution of clinical and laboratory manifestations of type 2 diabetes in their 2004 and 2009 systematic reviews and metaanalyses [7,8]. Whilst such research groups did not publish any guidance on defining resolution of diabetes, their definitions of resolution developed for use in review and metaanalysis were widely adopted [9]. In 2009, a multi-disciplinary expert group published a consensus report (2009 report) in Diabetes Care, an American Diabetes Association (ADA) journal to provide guidance on defining this concept [10], although this 2009 report does not represent the official ADA position [11]. The 2009 report recommends that the term remission should be applied to a chronic disease such as diabetes, rather than terms like resolution. It broadly defines remission as "achieving glycaemia below the diabetic range in the absence of active pharmacologic or surgical therapy" [10] (page 2134). Three specific types of remission are then explicitly defined: Partial, Complete and Prolonged Remission (cure) ( Table 1). Each remission type is composed of three components: (1) the absence of glucose lowering therapy (GLT); (2) the achievement of a glycaemic threshold; and (3) a duration during which the other two components have to be sustained for remission to have occurred. The 2009 report is not to be confused with the annually updated ADA Standards of Care, which contains the official ADA position on the diagnosis of diabetes and prediabetes [12], as yet, there is no official ADA position or internationally agreed consensus on the definition of remission [12]. Normal glycaemic measures d At least 5 years duration a. Partial remission is specifically defined as "Sub-diabetic hyperglycaemia (HbA1c not diagnostic of diabetes <48mmol/mol (<6.5%), fasting glucose 5.6-6.9 mmol/l (100-125 mg/dL)) of at least 1 year's duration in the absence of active pharmacologic therapy or ongoing procedures." [10] (p2134) b. Complete remission is specifically defined as "A return to "normal" measures of glucose metabolism (A1C in the normal range, fasting glucose <5.6 mmol/l (<100 mg/dl)) of at least 1 year's duration in the absence of active pharmacologic therapy or ongoing procedures." [10] (p2134) c. Prolonged remission is specifically defined as "Complete remission that lasts for more than 5 years and might operationally be considered a cure." [10] (p2134) d. There is ambiguity in terms of whether remission requires HbA1c<threshold AND FPG<threshold, or whether the logic is HbA1c<threshold OR FPG<threshold (see footnotes a-c where the join between HbA1c and FPG is a comma rather than a logical operator).
Remission of type 2 diabetes is of increasing interest to professionals and patients [12].

Search strategy and selection criteria
We adapted The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement for scoping reviews to report our systematic review of observational and interventional studies [19] (S1 Table). We included observational and interventional studies involving at least 100 participants with type 2 diabetes in their remission analysis, which examined an outcome of type 2 diabetes remission in adults >18 years old, and which had been published in English since 2009 (when the 2009 report was published) [10]. These restrictions were decided a priori and were expected to provide a sufficient number of papers specifically focussing on the effect of the 2009 report in reasonably large research studies within our time and resource constraints (S2 Table). Web-content or materials produced outside the traditional academic publishing were not searched as we were interested in how remission was defined in manuscripts that had been peer-reviewed in academic publication and distribution channels. We excluded case reports, systematic reviews, protocols or reviews. Studies focusing on: prediabetes; impaired glucose tolerance; impaired fasting glucose; gestational diabetes; maturity onset diabetes of the young; steroid induced diabetes; or type 1 diabetes were excluded as beyond the scope of this review (S2 Table).
Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research. The Diabetes UK-James Lind Alliance Priority Setting Partnership has recently identified remission as the top shared priority amongst people living with diabetes and their carers, healthcare professionals and black and minority ethnic groups [15].
We searched MEDLINE, EMBASE, CINAHL and the Cochrane library from 1 st January 2009 until 18 th July 2020 (S2 Table). The PICOS search strategy [20] was developed with a senior librarian and combined terms for: "remission" AND "type 2 diabetes" AND "weight loss strategies" and limits specified (human subjects, English language) (S3 Table). Two reviewers (MC and one of BG, SW and RP) independently screened all title and abstracts and all full texts against our inclusion criteria using Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia). MC and RP independently extracted data in duplicate to a piloted Microsoft Excel extraction form. In all cases, disagreement was discussed and a third reviewer involved to resolve these if needed. The systematic review protocol was pre-specified and prospectively registered (PROSPERO registration: CRD42019144619).

Outcomes
The primary outcome was the definition of remission used in each study irrespective of the term(s) used to name the outcome (partial, complete, or prolonged remission, and synonyms such as resolution, cure or reversal). Since an individual study could measure remission in more than one way, we extracted data for all distinct definitions used in each paper.

Data Analysis
For each definition of remission given by an underlying study, we sought to specify how each outcomes. Assessing whether poor reporting of remission definitions was linked to high risk of bias was outwith the scope of this preliminary exploration. Therefore, we did not undertake risk of bias assessment of the underlying studies in relation to the analysis they did.

RESULTS
The search identified 8966 citations, and 6772 title and abstracts underwent screening after removal of duplicates. After screening of 381 full texts against inclusion criteria, 178 studies from 33 countries met the inclusion criteria (Fig 1), comprising 164 cohort studies, eleven randomised-controlled trials, two cluster-randomised trials and one cost-effectiveness model (S4 Table).  Table). 81 (46%) of studies did not directly reference an existing guideline or paper.  Table). We combined definitions named as remission, remission of hyperglycaemia and resolution into a single category (Unspecified Remission) and definitions named as prolonged remission and cure into a single category (Prolonged Remission) to create four groups. There were 266 distinct definitions extracted, since 61 (34%) studies examined more than one remission type (for example, separately defining Partial Remission and Complete Remission). Almost half of definitions were categorised as Unspecified Remission (124 definitions (47% of all definitions)), compared to the more specific terms Partial Remission (57 definitions (21%)), Complete Remission (64 definitions (24%)) and Prolonged Remission (21 definitions (8%)) ( Table 2).  Fig 2) [23-29]. One study specified that metformin prescribed for a non-type 2 diabetes indication e.g. polycystic ovarian syndrome (PCOS) need not be stopped for remission to be achieved [29]. Remission, and nine in Prolonged Remission (Fig 2, Table 3). HbA1c was used more often than FPG and two-hour plasma glucose (2-hr PG). Of the 232 definitions that explicitly specified a numeric glycaemic threshold, 105 (45%) were based on HbA1c alone and HbA1c<42mmol/mol (6.0%) was used most frequently in 47 (20%) definitions. HbA1c <48mmol/mol (6.5%) with or without normoglycaemic interpretations of FPG (103 definitions) and HbA1c<42mmol/mol (6.0%) with or without normoglycaemic interpretations of FPG (89 definitions) were the most frequently used thresholds (S5 Table). It was not clear whether both or either of FPG and HbA1c needed to be attained in 23 definitions (S5 table).
There were 27 unique ways to categorise normoglycaemia (when all categories of remission type were merged, regardless of the term used by the authors) due to variation in the glycaemic thresholds for each of HbA1c, FPG and 2 hr PG, and also the combination of glycaemic tests specified (S5 Table).  Fig 2). When specified, the time periods defining duration of normoglycaemia ranged from 30 days to five years (Fig 2), with 17% of definitions defining time cross-sectionally (e.g. assessing remission at one moment in time such as at one year follow-up after an intervention) rather than longitudinally (e.g. absence of GLT and glycaemic threshold definitions met for a duration of one year) ( Table 2). Eleven studies (13 definitions) specified a time period or follow-up of less than one year in their definitions [13,23,25,30-37] (Fig 2). We found 13 unique definitions for the time component (when all categories of remission type were merged, regardless of the term used by the authors) ( Figure 2, Table 3). This was due to variation in the period of time specified, definition of time cross-sectionally or longitudinally and applying a different definition of time to the other two remission components (Fig 2).   Fig 2).

Unique definitions
For each remission category, heterogeneity in defining glycaemic thresholds was the most frequent driver of total heterogeneity (Table 3,   Strengths of the study include systematic searching, and that all screening and extraction was carried out independently by two reviewers. We used the PRISMA statement for scoping reviews to ensure clarity of reporting [19]. We had to adapt traditional PRISMA recommendations for a review assessing and evaluating a particular aspect of methodology (how remission was defined) rather than a traditional outcome focused systematic review. defined as a term in its own right. This precluded a meaningful analysis of fidelity (S6 Table).
Some studies responded to this ambiguity by discussing their interpretation and  [62]. This review highlighted a tendency or preference in the research literature to use the word "remission" without qualification rather than "partial", "complete" or "prolonged remission." The complexity in defining remission and creating partial and complete subcategories mirrors the complexity in defining diabetes and prediabetes [12]/intermediate hyperglycaemia [62]. Complete remission mirrors the diagnosis of prediabetes which is in itself controversial and is not internationally consistent.
We therefore suggest the alternative terms Remission of type 2 diabetes (instead of partial remission) and Remission of prediabetes (instead of complete remission) to make this clear.
Creating and defining a remission of prediabetes category will be highly challenging (Table   4) and it may be worth focussing on a consensus for Remission of type 2 diabetes in the first instance. There is an argument that combining multiple patient characteristics in a risk prediction model may be an alternative approach which sidesteps the issue of creating and defining a remission state and focusses instead on minimising future vascular complications rather than attaining remission [61]. However, until such tools are created, clinicians rely upon naming and classifying disease states which match diagnostic classification and coding schemas [63]. Future systematic reviews and meta-analyses would ideally use meta-analysis of individual participant data to apply a consistent definition of remission to make best use of existing data.
The first step in translating research to clinical practice is to provide clinicians with guidance on diagnosing type 2 diabetes remission. Given the fundamental difficulties in categorising type 2 diabetes, thought must be given to how remission guidelines maintain consistency with diabetes diagnostic guidelines. Type 2 diabetes is diagnosed by HbA1c, or FPG, or 2-hr PG.
These continuous measures of glycaemia are categorised using the following thresholds: HbA1c >48mmol/mol (6.5%); FPG >7.0mmol/l (126mg/dL); 2-hr PG >11.1mmol/l (200 mg/dL). As a consequence of diabetes diagnostic guidelines, the logical operator between HbA1c, FPG and 2-hr PG would ideally be AND. This would avoid the situation of a person diagnosed in remission from their type 2 diabetes based solely on an HbA1c<48mmol/mol (6.5%) (one of the most commonly used definitions of remission) whilst simultaneously meeting criteria for diabetes diagnosis if their FPG was over 7.0mmol/l (126mg/dL) [12,64].
However, achieving remission in all three glycaemic measures has substantial time and cost implications. This definition would be impractical to implement and would not be clinically appropriate in all patients (e.g. HbA1c is an inaccurate measure of glycaemia in certain patients and not easily available in all countries [64]) (Table 4).Therefore, using either HbA1c or FPG or 2-hr PG approach may be the best compromise to define remission of type 2 diabetes. Potential inconsistencies between diabetes diagnosis and remission diagnosis should, however, be specifically addressed in guidelines to avoid causing confusion amongst clinicians and patients.
People with type 2 diabetes have prioritised remission of type 2 diabetes and want more guidance about how to achieve remission and its implications [15]. This systematic review focuses on how remission has been defined in the recent literature. In doing so, we have demonstrated diverse multiple definitions of remission. If the concept of type 2 diabetes remission is to be pursued and implemented in clinical practice, then proceeding without a widely adopted consensus remission definition will be confusing for patients and clinicians alike. Heterogeneity also impedes research reproducibility, building knowledge and the provision of clear guidance to people with type 2 diabetes. This review supports the need for an international consensus definition of remission to guide both research and clinical practice.
In the meantime, any research study using 'remission' as an outcome should unambiguously report and justify their definition in terms of absence of GLT, glycaemic thresholds, and time.
20  [12,62] Consensus required on thresholds and whether all available glycaemic tests must be normal, or whether remission can be diagnosed based on the single most appropriate test for the patient 3. Absence of GLT and glycaemic thresholds both met for a duration of time Consensus is required on the duration required e.g. 6 months or 12 months. * Prediabetes itself is a disputed concept and a consensus definition for a remission of prediabetes/intermediate hyperglycaemia is likely to be very challenging. WHO does not use the term prediabetes and suggests Intermediate Glycaemia which can include: (1) Impaired Fasting Glucose; FPG 6.1-6.9mmol/l (110-124mg/dL) and (if measured) 2-hour plasma glucose <7.8mmol/l (140mg/dL) (2) Impaired Glucose Tolerance; FPG<7.0mmol/l (126mg/dL) and 2-hour plasma glucose>7.8 (140mg.dL) and <11.1mmol/l (200mg/dL)) [