Metformin in adults with type 1 diabetes: Design and methods of REducing with MetfOrmin Vascular Adverse Lesions (REMOVAL): An international multicentre trial

Aims Cardiovascular (CV) disease is a major cause of reduced life expectancy in type 1 diabetes (T1D). Intensive insulin therapy prevents CV complications but is constrained by hypoglycaemia and weight gain. Adjunct metformin reduces insulin dose requirement and stabilizes weight but there are no data on its cardiovascular effects. We have therefore initiated an international double‐blind, randomized, placebo‐controlled trial (REMOVAL: REducing with MetfOrmin Vascular Adverse Lesions in type 1 diabetes) to examine whether metformin reduces progression of atherosclerosis in adults with T1D. Individuals ≥40 years of age with T1D for ≥5 years are eligible if they have ≥3 of 10 specified CV risk factors. The enrolment target is 500 participants in 17 international centres. Materials and methods After 12 weeks of single‐blind placebo‐controlled run‐in, participants with ≥ 70% adherence are randomized to metformin or matching placebo for 3 years with insulin titrated towards HbA1c 7.0% (53 mmol/mol). The primary endpoint is progression of averaged mean far wall common carotid intima‐media thickness (cIMT) measured by ultrasonography at baseline, 12, 24 and 36 months. This design provides 90% power to detect a mean difference of 0.0167 mm in cIMT progression between treatment arms (α = 0.05), assuming that up to 20% withdraw or discontinue treatment. Other endpoints include HbA1c, weight, LDL cholesterol, insulin requirement, progression of retinopathy, endothelial function and frequency of hypoglycaemia. Conclusion REMOVAL is the largest clinical trial of adjunct metformin therapy in T1D to date and will provide clinically meaningful information on its potential to impact CV disease and other complications.


| INTRODUCTION
Period life expectancy in people with type 1 diabetes (T1D) is reduced by 11 to 13 years 1 ; rates of CV events are at least double those in the general population and account for approximately 45% of deaths. 2 Long-term post-randomization data from the Diabetes Control and Complications Trial (DCCT) participants followed up in the Epidemiology of Diabetes Interventions and Complications (EDIC) study demonstrate that intensive blood glucose control reduces both microvascular and CV complications in T1D. 3 However, in population-based data only approximately 30% of individuals with T1D are near target HbA1c (<7.5%/59 mmol/mol) and at least 30% have poor control (HbA1c ≥ 9.0%/75 mmol/mol). 4 A key barrier to optimizing glycaemia is hypoglycaemia; in the DCCT, rates of severe hypoglycaemia were 3-fold higher in those randomized to intensive therapy and with HbA1c at or below target. 5 Another long-term issue is insulin-induced weight gain, which may be accompanied by escalating insulin dose requirements, increased LDL cholesterol and/or raised blood pressure (BP). [6][7][8] Adjunct therapy with metformin reduces insulin dosage in T1D and may attenuate weight gain [9][10][11][12][13] ; some clinicians already use it in this context. As metformin reduces CV disease in type 2 diabetes (T2D), [14][15][16] and is recommended first-line therapy in most international guidelines, 17 we hypothesized that it might also provide CV protection in T1D.
In this largest and longest trial of metformin in T1D to date, we aim to gather data on cardiovascular and metabolic endpoints as well as key aspects of long-term safety (eg, vitamin B12 status, lactic acidosis). Progression of carotid artery intima-media thickness (cIMT) measured by ultrasonography is the primary endpoint as it was reduced by intensive glucose control in DCCT-EDIC 18 ; this was later validated by reduced CV events. 3

| Objectives and endpoints
The primary objective of the REducing with MetfOrmin Vascular Adverse Lesions in type 1 diabetes (REMOVAL: clinical trials.gov NCT01483560) is to test in adults with T1D whether metformin added to insulin therapy (titrated towards target HbA1c 7.0%/53 mmol/mol) reduces progression of atherosclerosis in the common carotid artery (CCA), defined as within-person change in bilateral averaged mean far wall carotid intima-media thickness (cIMT) measured annually over 3 years.
For measurement of cIMT, the same ultrasound system and preset image parameter settings (eg, depth, gain, persistence, dynamic range, post processing) are to be maintained at each site throughout the study. The reading centre (University College London [UK]) trains each site sonographer who then submits 5 accreditation scans. Right and left carotid arteries are interrogated in B mode with a 7.0 MHz or higher broadband linear array transducer with concurrent recording of a three lead ECG. A plaque screen (defined as focal thickening ≥1.5 mm or 50% greater than surrounding IMT) of the near and far walls of the CCA, bulb and internal carotid artery segments is performed also. Longitudinal images of the CCA are obtained at anterior, lateral and posterior angles using Meijer's arc during at least 5 cardiac cycles. Additional cIMT measurements are performed on a panel of 6 participants at each site annually to monitor reproducibility. At the reading centre, triplicate measurements are taken from the distal centimetre of the CCA (ie, immediately proximal to the bulb) by a single trained assessor using a validated semi-automated program. 19 The assessor undergoes repeated "masked" QC cycles to assess repeatability.
For the assessment of retinal disease, 2 colour 45 field photographs (field 1, optic disc; field 2, macula) are taken in each eye at randomization and at 36 months. In the UK these are acquired directly from national retinal screening systems. Images are graded using custom-designed software at the University of Wisconsin Ocular Epidemiology Reading Center (OERC) using the modified Airlie House classification scheme and the Early Treatment Diabetic Retinopathy Study severity scale as previously described. 20  Other secondary and tertiary endpoints are shown in Table 1.
Each of the secondary outcome measures will be analysed separately and the individual results will be reported. The protocol has a predefined composite interpretation of the secondary outcomes where results will be considered clinically meaningful with the potential to influence clinical practice in the event that a statistically significant improvement in 2 or more of the following individual outcomes is observed on metformin: (1) HbA1c (by DCCT-standardized local assays); (2) LDL-cholesterol (centrally-measured); (3) albuminuria, based on at least 2 separate urine specimens and routinely available assays (File S1, Appendix 3; (4) 2 or more step progression on the 11step modified concatenated retinopathy severity scale; (5) weight, by calibrated scales; (6) insulin dose; and (7) endothelial function (RHI). follow-up is permitted, we seek consent from participants for the local team to remain in contact at trial end.

| Screening, eligibility, enrolment and run-in period
Individuals aged ≥40 with ≥5 years T1D and at least 3 of 10 specified CV risk factors are eligible ( Table 2). T1D is defined as diagnosis of diabetes before age 35 years and insulin use within 1 year. Potential

Composite interpretation of all secondary endpoints
Improvement in 2 or more of these secondary endpoints will be considered clinically meaningful with the potential to influence clinical practice.   3 As judged by the site principal investigator based on at least 2 urine samples assayed locally and interpreted according to site reference ranges (File S1,

| Hypoglycaemia
Participants are asked to record all symptomatic or biochemicallyproven hypoglycaemic episodes (<2.8 mmol/L; 50 mg/dL) in the study diary. This information is used by site nurses at follow-up visits as a basis for completing the Steno Hypoglycaemia Questionnaire in which events are categorized as: minor (self-treated, resolved with short acting glucose and longer acting carbohydrate); major (requiring assistance from 1 or more other persons); or major with unconsciousness (self-reported). In addition to serious adverse event reporting, specific gastrointestinal, neurological, metabolic, renal and cardiovascular adverse events of medical interest are also recorded, as well as new diabetesrelated complications, operations or procedures.

| Safety and pharmacovigilance
Following each meeting of the IDMC (see above), a recommendation is made to the co-sponsors regarding the appropriateness of continuing the trial, from a safety and efficacy perspective. In addition to these arrangements, a Glycaemia Committee led by Dr Irene Hramiak (Ontario, Canada) sends detailed blinded reports of the participants' HbA1c status and rates of hypoglycaemia to each site every 6 months, along with "benchmarking" data from other sites in their region. The Committee can contact and support centres in which average HbA1c is higher than that in other comparable centres.

| Statistical considerations
All analyses will be conducted blinded to treatment allocation. The principal analysis will be on a modified intention-to-treat analysis set, ie, including all subjects from the intention-to-treat population (all randomized participants, regardless of subsequent participation in the study) with data available (without imputation). The target sample size is based on analysis of cIMT primary endpoint data using  repeated measures regression analysis, assuming a linear progression in the control arm of mean 0.044 mm and SD 0.050 mm over 3 years. 23 Regression model effect estimates with 95% confidence intervals and associated P values will be calculated. In order to minimize the residual SD, cIMT data will be adjusted for baseline cIMT as well as for age, sex and baseline levels of cardiovascular risk factors predictive of cIMT progression (specified in the Statistical Analysis Plan). To account for differences in ultrasound machines used at sites, a sensitivity analysis adjusting for ultrasound probe frequency is also specified along with a separate per protocol analysis.
A final sample size of 200 participants per treatment arm provides 90% power to detect an average mean cIMT difference of at least 0.0167 mm (one third of an SD) between treatment arms (α = 0.05); we therefore aim to recruit 500 patients allowing for 20% treatment withdrawal and/or treatment discontinuation. This sample size will provide 90% power to detect differences of approximately 0.3 SD in secondary endpoints including lipid, metabolic and endothelial function (α = 0.05). The retinopathy secondary endpoint is exploratory; if 3-year 2-step progression in the ETDRS category is estimated at 13.7%, treatment with metformin will necessarily be associated with a 60% reduction in risk for 80% power to declare significance at P < 0.05. No interim analyses are planned or pre-specified.

| RESULTS AND DISCUSSION
REMOVAL is the first adequately-powered long-term trial addressing the impact of metformin on a valid CV surrogate outcome (cIMT) in T1D. It will also collect data on metabolic endpoints (insulin dose, weight, HbA1c, hypoglycaemia, LDL cholesterol) as well as other vascular outcomes (endothelial function, retinal disease).
Metformin is a biguanide that undergoes active transport via cationic transporters and accumulates in intestinal cells. 24 During steady-state oral therapy, plasma glucose is reduced mainly by inhibition of hepatic glucose production. 25 Glucose-lowering is key in reducing microvascular complications in both T1D and T2D, but its effect on cardiovascular (macrovascular) complications is more complex, as other risk factors impact to a greater or lesser extent.
The differing molecular mechanisms of action of the various available classes of glucose-lowering "antidiabetic" agents are critical to their overall therapeutic profile as candidates for adjunct therapy. For metformin, mechanisms relevant to glucose-lowering include activation of AMP-activated protein kinase (AMPK), 26 inhibition of mitochondrial glycerophosphate dehydrogenase, 27 and release of gut hormones (including glucagon-like peptide-1). 28 However, other downstream effects of AMPK have been postulated to mediate vascular actions of metformin, 29 including modulation of proinflammatory pathways in perivascular adipose tissue 30 and inhibition of STAT3 (and thereby monocyte to macrophage differentiation) in vascular tissue. 31 Moreover, metformin can inhibit advanced glycosylation end product (AGE) formation by binding and inactivating methyglyoxal via an AMPK-independent pathway. 32 The primary focus of REMOVAL is to assess metformin's effects on the cardiovascular system in adults with T1D who are at high risk of CV disease, rather than its ability to lower glucose levels.
Accordingly, we adopted a double-blind, placebo-controlled, "treat to target HbA1c" design. When adjunct agents are prescribed in T1D, insulin doses are often down-titrated to avoid hypoglycaemia, such that an overall effect on glycaemia (measured by HbA1c) is not sustained. 10,11,13 Thus, although HbA1c is 1 of 7 pre-specified secondary endpoints, it is unlikely, by design, that a sustained separation in glycaemia between active and placebo arms will be observed. Instead, the trial is powered to detect whether 3 years of treatment with metformin reduces atherosclerosis progression as measured by cIMT. In addition to measuring vascular structure, we are assessing endothelial function (RHI) in 80% of participants, to provide an index of vascular function.
cIMT can be considered a validated surrogate endpoint for atherosclerotic disease in T1D on the basis of DCCT-EDIC. 3,18 However, despite the variety of pathways by which metformin has been hypothesized to exert potentially beneficial effects on the cardiovascular system, [29][30][31] there is conflicting evidence regarding its effects on cIMT. We recently reported that metformin had no impact on cIMT over 18 months in non-diabetic patients with established coronary heart disease 33 ; similarly, no reduction in cIMT progression was detected in insulin-treated people with T2D in the recent (underpowered) Copenhagen IMT trial. 34 However, metformin has been reported to reduce cIMT progression in metabolic syndrome 35 and also in T2D. 36 REMOVAL is the first cIMT progression trial concerning T1D; in this context it is important to note that mechanisms of accelerated atherosclerosis in T1D and T2D differ in a number of aspects. 37,38 Despite a paucity of evidence concerning T1D, metformin (embonate) already holds a product license for use in T1D in France 39 ; moreover, the UK National Institute for Clinical Excellence (NICE) recently recommended metformin for adults with T1D and BMI ≥ 25 kg/m 2 who "want to improve glucose control while minimising their effective insulin dose." 40 Currently, more than 50% of people with T1D are now obese or overweight. 8 Given that REMOVAL is planned to be 3 times longer and larger than any previous T1D metformin trial, the secondary endpoint data will be of considerable clinical utility in addressing longer term metabolic effects (eg, those on weight and insulin dose).
As metformin is structurally related to phenformin, which was withdrawn in the 1970s because of cases of lactic acidosis, concerns have been expressed regarding its use in ketoacidosis-prone T1D patients. 41 Metformin is commonly associated with gastrointestinal adverse effects, and long-term use in T2D is associated with vitamin B12 deficiency. 42,43 Rather than simply extrapolating its adverse effect profile and overall tolerability from T2D, REMOVAL will gather important specific safety data on metformin in T1D.
A key limitation of the study is use of a surrogate primary endpoint rather than clinical cardiovascular events. 44 Although several large T2D CV outcome trials have been reported recently and many more are in progress, 45 not a single randomized trial of any intervention with CVD as the primary outcome has been performed in T1D to date, despite the undoubted impact of CV disease in this condition. 1,2 Much of the current evidence base for CV preventive strategies in T1D (including that for statins) is extrapolated from T2D or from meta-analysis of T1D subgroups. 46,47 Another limitation is that focusing on atherosclerosis progression in the carotid artery may obviate detection of any beneficial cardiovascular effects mediated by other mechanisms. Data on which to base estimates of the degree of cIMT disease and the rate of change in our population were limited (mainly from DCCT); thus, there remains a degree of uncertainty in the power of a 3-year intervention study in this population. Finally, the retinal endpoint can be regarded only as exploratory; it was included to acquire a point estimate for any likely effect size to guide future research, given the relatively low marginal cost of acquiring images from routine screening (at least at UK sites).
Two different glucose-lowering agents used in T2D have recently been demonstrated to improve cardiovascular outcomes: a GLP-1 agonist 48 and an SGLT2 inhibitor. 49 We anticipate that our international effort in REMOVAL, the largest and longest clinical trial of adjunct metformin therapy concerning T1D to date, will illustrate the feasibility of conducting large collaborative multi-centre cardiovascular trials of adjunct therapy in T1D. Whether the data for metformin in the REMOVAL trial are positive or negative, we hope they will provide a stimulus to funding agencies and the wider diabetes community to support timely trials of other adjunct therapy candidates, with the twin aims of improving metabolic control and CV outcomes. Agents that are able to reduce rates of CV disease are urgently needed in T1D.