Improved mobility with metformin in patients with myotonic dystrophy type 1: a randomized controlled trial

Abstract

Metformin, the well-known anti-diabetic drug, has been shown recently to improve the grip test performance of the DMSXL mouse model of myotonic dystrophy type 1. The drug may have positively affected muscle function via several molecular mechanisms, on RNA splicing, autophagia, insulin sensitivity or glycogen synthesis. Myotonic dystrophy remains essentially an unmet medical need. Since metformin benefits from a good toxicity profile, we investigated its potential for improving mobility in patients. Forty ambulatory adult patients were recruited consecutively at the neuromuscular reference centre of Henri-Mondor Hospital. Participants and investigators were all blinded to treatment until the end of the trial. Oral metformin or placebo was provided three times daily, with a dose-escalation period over 4 weeks up to 3 g/day, followed by 48 weeks at maximum dose. The primary outcome was the change in the distance walked during the 6-minute walk test, from baseline to the end of the study. Concomitant changes in muscle strength and effect on myotonia, gait variables, biological parameters and quality of life were explored. Patients randomized into two arms eventually revealed similar results in all physical measures and in the mean 6-minute walk test at baseline. For the 23/40 patients who fully completed the 1-year study, differences between the groups were statistically significant, with the treated group (n = 9) gaining a distance of 32.9 ± 32.7 m, while the placebo group (n = 14) gained 3.7 ± 32.4 m (P < 0.05). This improvement in mobility was associated with an increase in total mechanical power (P = 0.01), due to a concomitant increase in the cranial and antero-posterior directions suggesting an effect of the treatment on gait. Subanalysis revealed positive effects of metformin treatment on the 6-minute walk test at the first intermediate evaluation (after 16 weeks of treatment), quantitatively similar to those recorded at 1 year. In contrast, except for the expected limited weight loss associated to metformin treatment, there was no change in any of the other secondary endpoints, including myotonia and muscle strength. Patients in the treated group had a higher incidence of mild-to-moderate adverse effects, mostly gastrointestinal dysfunctions that required symptomatic treatment. Although results were statistically significant only for the per protocol population of patients and not in the intent-to-treat analysis, metformin at the maximal tolerated dose provided a promising effect on the mobility and gait abilities of myotonic patients. These encouraging results obtained in a small-scale monocentric phase II study call for replication in a well-powered multicentre phase III trial.

myotonic dystrophy type 1, metformin, clinical trial, 6-minute walk test, gait

Introduction

Myotonic dystrophy type 1 (DM1; OMIM #160900), an autosomal dominant monogenic disease caused by an unstable CTG expansion, is the most common form of muscular dystrophy in adults. DM1 phenotype is characterized by a multi-systemic array of symptoms, among which is altered mobility secondary to major muscle weakness, wasting and myotonia, which represents the main disease burden (De Antonio et al., 2016). Beside recent advances in symptomatic treatments for sleepiness (Puymirat et al., 2012) or myotonia (Logigian et al., 2010), DM1 has remained essentially an unmet medical need.

Over recent years, we have assessed pathological mechanisms associated with DM1 using cells differentiated from human pluripotent stem cell lines derived from DM1 mutated gene-carrying embryos and from affected adult DM1 patients (Marteyn et al., 2011; Gauthier et al., 2013). During this research programme, we have identified that the compound metformin—the most widely used anti-diabetic drug—can induce changes toward normalization in ratios of protein isoforms, the alteration of which is associated to various DM1 clinical symptoms (Laustriat et al., 2015). Subsequently, we showed that similar metformin-induced biological effects on alternate RNA splicing changes occurred in vivo in diabetic patients exposed to the drug. These effects are, therefore, not specific to DM1. While non-DM1 specific, metformin counteracts alternative splicing deregulation of a subset of genes known to be affected in DM1, similar to other compounds reported by others [e.g. manumycin (Oana et al., 2013), chromomycin (Ketley et al., 2014), pentamidine (Warf et al., 2009), furamidine (Siboni et al., 2015), resveratrol (Takarada et al., 2015)]. These beneficial effects were observed at relatively high concentrations of the drug. A complementary study using the DMSXL mouse model of the disease revealed a statistically significant functional benefit on motor testing after metformin treatment. However, the link with a biological effect on alternate splicing of specific genes was not established and metformin may have induced that behavioural benefit via other molecular mechanisms, e.g. autophagia (Brockhoff et al., 2017), glycogen synthesis (Scalzo et al., 2017) or insulin sensitivity (for review see Dial et al., 2018; Kjobsted et al., 2018).

Altogether, despite this uncertainty on the mode of action, these positive experimental results and the very good safety profile of metformin led us to undertake the present clinical trial designed to explore the effects of the drug on mobility in non-diabetic adult DM1 patients. The 6-minute walk test (6MWT), a robust functional test to assess mobility, was selected as the primary endpoint of functional capacity outcome measures, in agreement with the recommendations of the Outcome Measures in Myotonic Dystrophy type 1 (OMMYD-2) expert consensus (Gagnon et al., 2015). A series of secondary outcomes explored concomitant changes in muscle strength, myotonia, gait variables, biological parameters and quality of life.

Materials and methods

Study design

MYOMET was designed as a 52-week monocentric, double-blind, placebo-controlled phase II randomized study that aimed at assessing the efficacy of metformin on ambulation in patients with myotonic dystrophy type I. Secondary objectives were to evaluate safety of metformin and its potential efficacy on myotonia, muscle strength, gait variables, quality of life and splice defects. Patients were recruited from the cohort assessed in the Paris area neuromuscular reference centre. The entire clinical trial was conducted at the Henri-Mondor Hospital (Créteil, France), in the clinical research centre and the neuromuscular reference centre. Study protocol was approved by an appropriate ethics committee (CPP IDF VI, Paris, France, approval date 4 September 2013) and by the French regulatory authority (ANSM; Number 130862-31, approval date 13 August 2013). All participants were provided with participant information and responded with written informed consent to participate. This clinical trial was conducted in compliance with the Declaration of Helsinki and according to the Good Clinical Practice (GCP) and was registered in the European Union Clinical Trials Register (EudraCT Number 2013-001732-21; https://www.clinicaltrialsregister.eu/ctr-search/trial/2013-001732-21/FR) before the first inclusion of a patient.

Participants

Included patients had a diagnosis of DM1 confirmed by genetic mutation, were aged between 18 and 60 years inclusive, and covered by a national health insurance scheme affiliated to the French healthcare system. They had a Muscular Impairment Rating Scale (MIRS) (Mathieu et al., 2001) score equal to 2 or 3, were ambulatory and able to perform the 6MWT. Female participants of child-bearing potential were not included in case of pregnancy evidenced by a urinary test at the screening visit and were asked to use one effective method of birth control during the conduct of the study. Exclusion criteria included: evidence of serious concomitant or past history of medical disorders, including, but not limited to renal dysfunction, liver impairment, uncontrolled cardiac diseases, or any other medical conditions that might significantly impact ambulation. Patients under legal protection were excluded as well as those with current or past history of psychiatric conditions, including, but not limited to psychosis, suicidal ideations, or major depression. Known contra-indications to metformin were considered and treatment by medications intended for the treatment of DM1 were not allowed, including glucocorticoids, anabolic steroids, testosterone, growth hormone, or insulin-like growth factor 1 (IGF1) within 1 year of entry into the study and throughout the study participation. At the end of the 52-week study period, all patients having completed the randomized period were invited to enter an open-label extension phase during which they received the active treatment according to the same dose escalation performed at the study entry, and without any therapy restriction for the treatment of DM1. The open-label phase was still ongoing at the time of manuscript submission.

Randomization and masking

Enrolled patients were randomly assigned (1:1; 20 patients per group) to either metformin therapy or a placebo, using a centralized randomization procedure. Upon satisfaction of all inclusion and exclusion criteria for patients who consented, the investigator sent the enrolment form to the Clinical Research Organization (CRO, Axonal-Biostatem, Nanterre, France) and in return received the assigned treatment arm. The computer-generated random allocation sequence was generated by the CRO prior to the onset of the study and concealed from participants and investigators. Randomization was stratified according to age (> or ≤ 50 years), gender and distance walked over 6 min (> or ≤500 m) at the screening visit. Patients were not informed of their treatment regimen and a matching placebo was manufactured (Bertin Pharma) to be identical to the metformin tablets in order to ensure the double-blind methodology. All study staff involved in the assessment of patient outcomes, including performance of the 6MWT, were blinded to the treatment group allocation. The blinding was maintained throughout the study period until both data entry and data processing were completed.

Procedures

To prevent tolerance issues, participants received escalating daily doses of treatment over a 4-week period, up to the total study dose of 3000 mg/d. This dose was maintained for 48 weeks until the end of the study, cumulating a 52-week treatment duration. Metformin/placebo was taken orally by the patients, three times a day during or after a meal. After the enrolment of the first five patients of each arm, a blinded assessment of the tolerance was performed in order to check the patients’ compliance to the study treatment and to take appropriate measures if required. The patients who did not enter in the open-label extension phase were followed for safety up to 4 weeks after the end of the treatment or after having discontinued all study medication.

A screening visit was conducted to check for eligibility criteria, including performance on the 6MWT. After patients were enrolled at the inclusion visit [Day 1 (D1)], follow-up visits were scheduled at Weeks 2, 4, 16, 28, 40 and 52 (W2, W4, W16, W28, W40 and W52). Details concerning follow-up of patients are available in Supplementary Table 1. Data collection included results from clinical evaluation (medical history, height, weight, body temperature, heart rate, respiratory rate and blood pressure), ECG, blood laboratory evaluation [including full blood count, liver function tests, renal function tests, electrolytes, glycaemia, lipase, amylase, calcium, creatine kinase (CK), INSR and ATP2A1 mRNAs], ambulation evaluation (6MWT), the activity and social participation DM1-Active scale (Hermans et al., 2015) and the health-related quality of life InQoL questionnaire. The InQoL questionnaire was scored according to published recommendations covering the following 12 subdomains: weakness, muscle locking, pain, fatigue, activities, independence, social relationships, emotions, body image, perceived and expected treatment effects, and overall neuromuscular disease-related quality of life.

Quantitative muscle testing was performed using a handheld dynamometer (Mark-10) for knee extension, finger extension and elbow flexion and using the MyoAnkle dynamometer for ankle flexion and extension (Moraux et al., 2013). Grip strength and myotonia were assessed using a grip dynamometer (MIE) during the MyoTone test (Hogrel, 2009). Accelerometric data were recorded along the three axes of space using a specific device (Locometrix©) during the 6MWT. The Locometrix© includes three accelerometers in a small (20 × 40 × 80 mm) and light (50 g) box and a data logger. The apparatus is incorporated into a semi-elastic belt, which is fastened around the subject’s waist, close to the centre of gravity. Signals were recorded at a sampling frequency of 100 Hz. The recorded signals were transferred to a computer and analysed using dedicated software. Several gait variables were then computed: walking velocity, stride frequency and length, symmetry and regularity indexes, mechanical power. The power was computed in each space direction as the integral of the Fourier transform of the three acceleration signals. This variable measures the mean mechanical power generated in each direction of the device near the centre of gravity of the body and is expressed in W/kg. Its value depends on both the amplitude and the speed of the movements. It can be taken clinically as reflecting body abilities to move. The computation of all variables is further detailed in previous articles (Barthelemy et al., 2009; Mignardot et al., 2014).

Quantification of splice defects on blood samples were performed for identifying potential changes due to the treatment in the isoforms ratios of two genes known to exhibit splice defects in DM1 patients, namely INSR ± exon 11 and ATP2A1 ± exon 22. Conventional reverse transcription polymerase chain reaction (RT-PCR) and quantitative RT-PCR were used. Conventional RT-PCR techniques used were as previously described (Laustriat et al., 2015). Quantitative PCR performed in 384-well plates used a QuantStudio™ 12 K Flex Real-Time PCR System (Applied Biosystems, ThermoFisher Scientific) with Luminaris Probe qPCR Master Mix (Invitrogen) and TaqMan® gene expression assays: INSR − exon 11: Hs00965956_m1; INSR+exon 11: Hs00169631_m1; and 18 S : Hs99999901_s1 (Invitrogen). For ATP2A1 ± exon 22 transcript analysis, the following primers and MGB probe were used: ATP2A1 − exon 22: 5′-CGGAACTACCTAGAGGATCCAG-3′, 5′-CACAGCTCTGCCTGAAGATG-3′ and FAM-GGAAGTGAGCATCCTTTTGC-MGB NFQ; ATP2A1 + exon 22: 5′-GGGGGAACAGTTATCCCTCT-3′, 5′- ACCTCACCCAGTGGCTCAT-3′ and FAM-TTCGTTGCTCGGAACTACCT-MGB NFQ. A mesenchymal stem cell sample reference was included in all 384-well plates for amplification efficiency determination and as a calibrator for normalization. The method described by Pfaffl (2001) was used to determine the relative expression level of each gene.

Outcomes

The primary outcome was the mean difference in the distance walked as measured by the 6MWT between the baseline and Week 52 between the study groups. The test was performed according to the American Thoracic Society guidelines (ATS Committee on Proficiency Standards for Clinical Pulmonary Function Laboratories, 2002).

Secondary outcome measures were: changes from baseline to Week 16 and Week 28 in the distance walked at the 6MWT, changes from baseline in the distance walked at the 6MWT expressed as Z-scores according to available references in general population (Casanova et al., 2011), splice defects, muscle function and strength, gait variables, social participation and quality of life. The safety of metformin was evaluated by the overall incidence of adverse events and serious adverse events as well as ECG and laboratory assessments.

Adverse events

Metformin has a long track record for treatment of non-insulin-dependent diabetes mellitus not responding to dietary modification. Metformin improves glycaemic control by improving insulin sensitivity and decreasing intestinal absorption of glucose. Metformin is associated with a very low incidence of lactic acidosis. The product resume of metformin was used as a reference of suspected adverse events, and intensity of an event was evaluated by WHO definitions of serious adverse event (SAE) and suspected unexpected serious adverse reaction (SUSAR). Participants were informed about the risk of side effects before inclusion. During the trial, inconvenience experienced by the participants was registered in the diary and participants were interviewed about adverse events at evaluations. Trial continuation was stopped in individuals in whom a suspected serious adverse event of metformin was encountered.

Statistical analysis and sample size calculation

Available data are scarce regarding the expected impact of treatments on ambulation as assessed by the 6MWT in patients with myotonic dystrophy type I. In a randomized controlled trial assessing the impact of a physical exercise programme on ambulation in patients with MIRS scores of 2 to 5, Kierkegaard et al. (2011) reported a 2% increase in the intervention group [+9 m (standard deviation, SD 38)] whereas no increase was observed in the control group [−2 m (SD 40)]. Data from another previous report indicate that MIRS 2–3 patients are expected to have a mean walking distance of 550 m (SD 100) at baseline 6MWT (Kierkegaard and Tollback, 2007). Based on the previous figures and at two-sided type 1 error of 5%, a sample size of 30 patients (15 per group) was required to provide an 80% power to detect a minimally clinically important difference of +35 m (SD 35) in favour of the experimental group, representing a relative increase of +6% in distance walked between baseline (D1) and final visit (Week 52). Enrolment of a total of 40 patients (20 patients in each group) was thus planned, assuming a 25% dropout rate, as observed in previous studies with DM1 patients (Kierkegaard et al., 2011).

Descriptive results are presented as means with SD or median [interquartile range (IQR)] for continuous variables, and as numbers with percentages for categorical variables. Unadjusted comparisons of continuous variables (including the primary endpoint) were performed using t-tests or Mann-Whitney tests, depending on the normality of the variable distributions as assessed by the use of the Shapiro-Wilk test for normality along with the assessment of kurtosis and skewness indicators. Comparisons of categorical variables were performed using the χ² or Fisher’s exact test, as appropriate. Comparisons between baseline and final assessments within groups were conducted using paired t-test or Wilcoxon signed ranks test, as appropriate.

For the primary outcome of the comparison of Week 0 with Week 52 for the 6MWT, patients were assessed according to randomized treatment group using an intent-to-treat (ITT) analysis and missing data imputation by the chained equations multiple imputation method. Supportive sensitivity analyses were performed to test the robustness of the results: (i) where the data were the complete set without imputing missing data; (ii) in ‘per protocol’ (PP) analysis after excluding patients with deviations to protocol; (iii) using linear multivariate regression models adjusted for variables used for stratification at randomization; and (iv) where the analysis was based on longitudinal data analysis using multiple linear mixed models to account for the correlation between repeated measures over time.

All tests were two-tailed and P-values < 0.05 were considered statistically significant. Analyses were pre-specified in the trial protocol. The analysis plan was performed using Stata v14.2 statistical software (StataCorp, College Station, TX, USA).

Data availability

Researchers wishing to access the data collected in the MYOMET study are welcome. They are requested to contact Dr Marc Peschanski ([email protected]) and sign a data access agreement.

Results

Forty-eight patients were identified by the mean of the DM-scope registry when applying study inclusion and non-incluion criteria. Eight patients were declared non-eligible due to test results: prolonged PR interval (>200 ms) at ECG in two, and elevated liver enzyme in serum (>1.5 N upper value) in five. Forty patients were thus included and randomized between February 2014 and May 2015, of whom 38 received at least one dose of treatment and were considered in the ITT population. Twenty-nine completed the 1-year follow-up (ITT complete cases), among whom 23 had no major deviation to protocol [per protocol (PP) population; Fig. 1]. Of the 15/38 patients excluded from the original ITT population, 10 were in the treated group and five had received placebo. More precisely, two patients were excluded because of dispensing errors of drugs; five patients in the treated and four patients in the placebo groups were lost because of deviation to protocol or lack of compliance. Four additional patients in the treated group withdrew between the pretreatment and the first post-treatment evaluations because of metformin-induced gastrointestinal adverse events that were not well controlled by usual anti- diarrhoeic treatment. Baseline characteristics of patients did not substantially differ between the two groups in either ITT (Table 1) or per protocol populations (Supplementary Table 2). All patients in the ITT group reached the maximal 3 mg/day dosage of metformin.

Table 1

General characteristics of the study population at baseline: ITT population

	Placebo group	Metformin group
	n = 19	n = 19
Demographics and physical examination
Age, years	38.31 (±10.08)	42.11 (±11.49)
Gender, female	11 (57.89%)	11 (57.89%)
MIRS 2	7 (36.84%)	7 (36.84%)
MIRS 3	12 (63.16%)	12 (63.16%)
Weight, kg	66.95 (±11.99)	70.91 (±18.55)
BMI, kg/m2	22.88 (±3.71)	25.12 (±6.53)
Vital capacity - total lung capacity, %	84.58 (±13.50)	85.16 (±14.31)
Systolic blood pressure, mmHg	111.79 (±12.33)	115.26 (±18.82)
Diastolic blood pressure, mmHg	62.37 (±9.23)	63.53 (±11.64)
Biological results
(CTG)n expansion mutation	600.00 (±254.80)	584.67 (±334.93)
Haemoglobin, g/dl	14.5 (13.6–15.1)	14.6 (13.0–15.4)
Haematocrit, %	44 (41–45)	44 (39–45)
Leucocytes, 109/l	6.4 (4.9–7.6)	5.4 (4.3–7.6)
Platelets, 109/l	195 (167–240)	191 (162–217)
ASAT, U/l	39 (28–46)	29 (27–35)
ALAT, U/l	45 (30–57)	35 (27–40)
GGT, U/l	45 (18–150)	36 (20–126)
Glycaemia, mmol/l	4.9 (4.6–5.2)	4.8 (4.2–5.1)

Results are means ± SD, median (IQR) or n (%).

ALAT = alanine amino-transferase; ASAT = asparine amino-transferase; BMI = body mass index; GGT = gamma-glutamyltransferase; MIRS = Muscle Impairment Rating Scale.

Figure 1

Flowchart of the study populations. PP = per protocol population.

Primary outcome

Results concerning the primary outcome are shown in Table 2. Altogether, the placebo group showed very stable values between baseline and the Week 52 examination. In contrast, the metformin-treated group systematically showed a trend to increase in performance, by whichever analysis carried out. Within-group analyses identified a statistically significant progression in the metformin group in the per protocol analysis (n = 9, P = 0.017). The difference between the treated and placebo groups in mean change in the total distance for 6MWT was statistically significant when considering patients who entirely fulfilled the protocol [per protocol, n = 23; control +3.7 m (±32.4); metformin +32.9 m (±32.7); difference in change +29.2 m, P = 0.048]. Modified-ITT analysis of the whole sample did not reach significance level despite systematically higher change estimates in the metformin group, whether after multiple imputation of missing values [n = 38; Control mean change +2.6 m (±8.2); metformin +20.5 m (±13.1); difference in change +17.9 m, P = 0.255], or considering only patients with complete information for the 6MWT [n = 29; Control +3.1 m (±31.2); metformin +14.6 m (±47.6); difference in change +11.5 m, P = 0.441].

Table 2

Primary outcome results: 6MWT total distance (m)

		Placebo group		Metformin group		Difference in changes	P-value
ITTa, n = 38		n	Mean (SE)	n	Mean (SE)
Raw values	Baseline	19	457.46 (±26.25)	19	442.89 (±26.28)	-	0.697
	Change from baseline, W16 –W0	19	0.12 (±6.36)	19	21.76 (±9.81)	+21.6	0.076
	Change from baseline, W28–W0	19	8.28 (±8.93)	19	18.18 (±13.53)	+9.9	0.549
	Change from baseline, W52–W0	19	2.57 (±8.19)	19	20.48 (±13.11)	+17.9	0.255
ITT (complete cases), n = 29		n	Mean (±SD)	n	Mean (±SD)
Raw values	Baseline	17	456.41 (±111.64)	12	475.92 (±101.81)	-	0.635
	Change from baseline, W16–W0	17	2.65 (±25.96)	12	20.58 (±38.55)	+17.9	0.145
	Change from baseline, W28–W0	17	11.06 (±38.48)	12	13.25 (±51.59)	+2.2	0.897
	Change from baseline, W52–W0	17	3.06 (±31.96)	12	14.58 (±47.56)	+11.5	0.441
Per protocol, n = 23		n	Mean (±SD)	n	Mean (±SD)
Raw values	Baseline	14	455.86 (±115.94)	9	457.67 (±93.36)	-	0.969
	Change from baseline, W16–W0	14	3.64 (±17.65)	9	32.56 (±36.94)	+28.9	0.019
	Change from baseline, W28–W0	14	11.57 (±36.10)	9	28.44 (±51.12)	+16.9	0.363
	Change from baseline, W52–W0	14	3.71 (±32.43)	9	32.89 (±32.75)	+29.2	0.048

^aITT analysis performed on all patients included except the two who withdrew between inclusion and first dosing. W = week; SE = standard error.

Intermediate visits for the primary outcome at 16 and 28 weeks follow-up yielded similar trends, with a statistically significant improvement in the metformin group in per protocol analysis as early as the Week 16 visit [Control +3.6 m (±17.6); metformin +32.6 m (±36.9); difference in change +28.9 m, P = 0.019; Fig. 2 and Supplementary Fig. 1].

Figure 2

6MWT evolution across follow-up visits: changes from baseline in per protocol (n = 23) and ITT (n = 38) populations. Results are shown as boxplots, with each box representing the IQR (first to third quartile), the line within the box indicating the mean, and the whiskers extending to 1.5 times the IQR above and below the box; the dots represent individual values for each patient.

Similar results were found after multiple adjustments on stratification variables, namely age, gender and baseline 6MWT [ITT: (metformin − control) difference in change+10.9 m, P = 0.491; per protocol: (metformin − control) difference in change +29.0 m, P = 0.055 and after using mixed models for longitudinal analysis of 6WMT values over time [interaction time × metformin group: ITT +0.03 m/day (95% confidence interval −0.04 to 0.10) P = 0.365; per protocol+0.07 m/day (−0.01 to 0.14), P = 0.077]. Analysis of age-sex standardized scores revealed decreased 6MWT Z-scores values compared to the general population [baseline (control) −1.84 SD; (metformin) −1.91 SD], with similar trends found for changes from baseline to those based on raw 6MWT values expressed in metres (Supplementary Table 3).

Secondary outcomes

Results for the other secondary outcomes in ITT and per protocol populations are shown in Supplementary Tables 4 and 5. Almost all clinical parameters evolved similarly over time, whether patients were treated or not with metformin. This included, in particular, myotonia indices, muscle function and strength, quality of life and ECG. Biological samples were also not different between the two groups. As concerns splicing defects associated with DM1, namely INSR ± exon 11 and ATP2A1 ± exon 22, results of conventional RT-PCR were not interpretable because the minority isoforms were below detection levels (Supplementary Fig. 2A). Quantitative PCR assays with greater sensitivity allowed us to detect the minority transcript INSR + exon 11 but at a barely visible level of detection. Variability of the measures precluded the determination of any statistically significant variation due to treatment (Supplementary Fig. 2B).

Some gait parameters showed a statistically significant difference between the two groups (Table 3). The total mechanical power during gait was significantly increased in the treated group compared to the placebo group (P = 0.011). This was due to an increase in mechanical power in both the cranio-caudal (P = 0.011) and antero-posterior (P = 0.028) directions. In contrast, the other gait variables showed no change in either group during the study.

Table 3

Secondary outcomes results: gait analysis

	Placebo group			Metformin group			Difference in changes	P-value
Gait analysis values	n	Baseline (complete W0–W52)	Change from baseline (W52–W0)	n	Baseline (complete W0–W52)	Change from baseline (W52–W0)
ITT population
Frequency	17	1.0 (0.9–1.0)	0.01 (−0.02–0.04)	15	1.0 (0.9–1.1)	0.03 (−0.01–0.05)	+0.0	0.692
Regularity	17	227 (202–276)	−9.0 (−26.9–6.1)	15	236.2 (198–263)	−16.9 (−26.1–13.3)	−7.9	0.895
Symmetry	16	103 (96–107)	−2.2 (−5.0–5.1)	15	106 (94–115)	0.3 (−6.6–6.4)	+2.5	0.937
pAP	17	2.4 (1.5–2.8)	0.13 (−0.07–0.50)	15	3.0 (1.9–4.0)	0.26 (−0.07–1.26)	+0.1	0.308
pML	17	2.1 (1.4–3.3)	0.34 (0.18–0.68)	15	2.2 (1.2–3.9)	0.20 (−0.25–0.61)	−0.1	0.282
pCC	17	4.8 (2.6–6.7)	0.59 (0.01–1.73)	15	5.2 (2.6–8.2)	0.86 (−0.12–2.41)	+0.3	0.777
PT	17	9.7 (5.4–12.1)	1.13 (0.29–2.96)	15	10.8 (5.3–16.9)	0.55 (−0.46–4.82)	−0.6	0.985
Per protocol population
Frequency	12	1.0 (0.9–1.0)	0.01 (−0.02–0.05)	7	1.0 (0.9–1.0)	0.05 (0.03–0.15)	+0.0	0.237
Regularity	12	223.9 (196–253)	−15.8 (−34.6–8.5)	7	236 (229–245)	1.6 (−26.1–26.7)	+17.4	0.499
Symmetry	11	106 (99–109)	−2.7 (−9.6–5.1)	7	114 (87–117)	0.3 (−7.9–8.0)	+3.0	0.751
pAP	12	2.5 (1.7–2.8)	0.25 (−0.02–0.91)	7	3.8 (3.0–4.0)	1.26 (1.16–2.55)	+1.0	0.028
pML	12	2.1 (1.4–3.3)	0.46 (0.27–0.92)	7	2.5 (2.0–4.1)	0.61 (0.25–1.79)	+0.2	0.735
pCC	12	4.6 (2.6–6.7)	0.61 (−0.46–1.65)	7	5.6 (3.3–8.2)	2.41 (1.83–3.73)	+1.8	0.011
PT	12	9.9 (5.7–12.0)	1.05 (0.54–3.66)	7	12.5 (8.3–16.9)	4.82 (3.49–5.90)	+3.8	0.011

All results are medians (IQR).

pAP = power in the antero-posterior direction; pML = power in the medio-lateral direction; pCC = power in the cranio-caudal direction; PT = total power.

Safety

Overall, 280 adverse events were reported in 38 (100%) participants, including 163 in the metformin group and 117 in the control group, of which 68% were not considered to be related to the study treatment (metformin group 53%; control group 88%). Five serious adverse events were reported in four participants (10.5%), including three in the metformin group, and one in the control group (Supplementary Table 6). Most frequent adverse events were gastrointestinal disorders including diarrhoea (n = 15 in the treated group versus n = 5 in the placebo group), abdominal pain (n = 10 versus n = 6) and dyspepsia (n = 4 versus n = 0). Diarrhoea could be controlled to the level of patient’s satisfaction using symptomatic treatments, except in four patients who withdrew from the study over the first trimester after inclusion. Asthenia occurred in eight participants in the metformin group and in three in the control group. Benign infections were more frequently observed in the control (n = 16) than in the metformin group (n = 9). Weight loss occurred in 11 participants, all in the metformin group, with a mean loss of −1.49 kg (±2.65) at Week 16 and −0.51 kg (±4.08) at Week 52 in the ITT population, −2.42 kg (±2.34) and −1.98 kg (±2.97) in the per protocol population. However, weight did not statistically differ between the two groups throughout the study. No effect of metformin was recorded on either ECG parameters or laboratory assessments, except for an intermittent increase of lipase in one participant. There was no case of hypoglycaemia or lactic acidosis in participants in the metformin group.

Discussion

This randomized placebo-controlled double-blind phase II study failed, possibly marred by the large number of patients who were lost to follow-up, to demonstrate a significant improvement after metformin in the ITT population for the walking distance over 6 min, which was the primary outcome of the study. However, a statistically significant increase in 6MWT distance was identified in the per protocol population of patients who fully completed the study, suggesting promising effects of metformin on the mobility of patients with myotonic dystrophy type 1. The more precise analysis of the various secondary outcomes that may have contributed to this functional effect points to positive changes in gait whereas muscle strength and myotonia did not improve in a statistically significant manner. Metformin may thus have an original target mechanism and be a complement to current candidate therapies. However, our study has limitations, in particular because its results are considered encouraging based only on the analysis of a reduced per protocol population of patients. This calls for confirmation in a high-powered multicentre phase III trial.

The primary endpoint of this study was the 6MWT, according to the recommendation of the international consensus workshop OMMYD-2 (Gagnon et al., 2015). There is a general agreement on the difficulty in finding a reliable and feasible test that would be fully informative on the progressive decline of patients with myotonic dystrophy type 1. This is especially due to the multi-systemic nature of the disease where the different symptoms develop according to different time schedules, as well as to their slow and non-uniform rate of progression. Within that general framework, the 6MWT was selected by the international committee in a large part because it has been very widely used over the years and facilitates a standard comparison of results. The test has also been shown to be reliable from one session to another when repeated in the same DM1 patients and feasible for most of them (Kierkegaard and Tollback, 2007). Nevertheless, one main challenge for our study was the reported lack of a statistically significant decline of the DM1 patients’ performance over 1 year for that test. This was indeed clearly the case in the control group in the present study. This did not allow us to reveal a potential preservation of performance under treatment, and only the potentially more challenging detection of a statistically significant improvement in mobility was to be pertinent. A meaningful change had been calculated by previous authors to 6% of baseline value, i.e. over 27 m in the present study (Kierkegaard and Tollback, 2007). The performance of the treated group indeed outperformed that theoretical success threshold at all time points in the per protocol population. It is worth mentioning that a statistically significant improvement of mobility was already observed at 16 weeks in the same per protocol population. A limited weight loss also evolved in a statistically significant manner in parallel to the 6MWT in the treated group but the two features have been shown to evolve independently in previous studies dealing with non-DM1 patients affected by heart failure (Wong et al., 2012). Interestingly, the difference observed in mobility performance between the two studies suggests that the effects of metformin recorded in the present study may bear some specificity for the DM1 disease process.

Among the main symptoms associated with DM1 that may impact mobility, metformin did not appear to have visible effects either on myotonia or on muscle strength. In contrast, there was a statistically significant improvement in some gait parameters, as revealed functionally by an increased mechanical power. It has been shown that the distance performed at the 6MWT is correlated with muscle strength, particularly ankle flexion and extension and knee extension but also in other muscle groups not involved in locomotion (Bachasson et al., 2016). In the absence of a gain in muscle strength, the increase in 6MWT distance may be due to an improvement in motor strategy or coordination. Accelerometry was used to get more information on the gait quality in addition to the gait quantity during the 6MWT. The mechanical power can be deduced from Newton’s second law, where it is said that the acceleration of an object equals the sum of the forces applied to its centre of gravity divided by its mass. The body mass being constant, we can speculate that the increased power reflects increased force acting upon the device. The increase of the power in the cranio-caudal and antero-posterior directions thus suggests a better efficiency in gait biomechanics possibly due to a slightly relative decrease of the rolling and waddling components. Gait disorders are one of the main functional impairments in myotonic dystrophy. During the progression of muscle atrophy, patients first maintain an autonomous gait for some time. However, gait defects will eventually couple with limited walking perimeter, and gait speed, as well as falls, contribute to the burden of the disease. As previously described by our group (Bachasson et al., 2016), lower trunk accelerometry analysis during walking at a self-selected pace reveals impairments in all spatiotemporal gait variables (i.e. slower walking speed, lower stride frequency, and smaller stride length) in patients with DM1 compared to controls. Patients with DM1 are unstable during standing and walking, which contributes to a restriction of their walking distance. Significant correlations have been observed between postural instability and muscle weakness. Therefore, metformin may have promoted an increased walking distance by improving gait parameters.

Metformin has not been shown in preclinical studies to affect specific DM1 molecular targets as do other drugs that were specifically designed for either decreasing expression of DMPK (Pandey et al., 2015), or else releasing MBNL1 from the expanded CUG repeats (Rzuczek et al., 2015). However, a deregulation of the AMPK/mTORC1 has additionally been recently identified in DM1 cells, and metformin is a well-known activator of AMPK (Brockhoff et al., 2017). One may rather consider more general metabolic effects, such as the well-known effects of metformin in partially stimulating autophagy (Jagannathan et al., 2015) and inhibiting the UPR transcription programme (Saito et al., 2009; Theriault et al., 2011), increasing glycogen synthesis (Scalzo et al., 2017) or improving insulin sensitivity in skeletal muscles (for review see Dial et al., 2018; Kjobsted et al., 2018). It is not possible to determine whether those hypothetical mechanisms may play a role in the observed clinical effects, in the absence of biological samples of skeletal muscles from patients. However, one conclusion can be drawn from this list of mechanisms, namely that they may well be very different from those of drugs that are currently tested in DM1 patients and, therefore, may point to metformin as an interesting add on treatment, whatever drug is used for targeting molecular mechanisms more directly related to DM1 (Konieczny et al., 2017).

Our phase II clinical study has, however, obvious limitations that need to be addressed before metformin may be widely prescribed to patients with DM1.The most important one is the general setting of our study, which was limited both in sample size and in the number of clinical centres involved. In addition, there was a substantial dropout rate in both groups, but more pronounced in the group receiving metformin. Mild to moderate drug side effects and imperfect compliance have obviously played a significant role in our observations. As a result, a potential loss in statistical power cannot be ruled out, as well as a possible, albeit limited, selection bias due to differential attrition. This large loss of patients to follow-up during the study may have participated to the lack of statistical significance between groups in the ITT analysis, despite an apparent systematic difference in values. Of note, statistically significant trends were nonetheless identified in the per protocol population in favour of the experimental group, confirming the appropriateness of the sample size calculations made before the conduct of the study. The results obtained in this monocentric study with just a handful of patients seem to us sufficiently encouraging to stimulate the planning of a larger-scale multicentre multi-national well-powered phase III clinical trial to provide definite proof of efficacy. In such a study, specific analysis of gait and balance parameters should definitely be added, although it is our contention that the 6MWT should be confirmed as the primary endpoint, in the absence of another test that would receive a similar consensus. In addition, the results on the mechanical power must only be considered as potential clues for the interpretation of the 6MWT improvement. As such, they should be assumed as hypotheses rather than definitive causal explanations.

Organizing a costly phase III clinical trial may, however, be a concern because a sufficient return on investment is difficult to foresee with such an old and cheap drug, accessible to patients as a generic. Another limitation of our study that calls for attention is the fact that a number of patients treated with metformin exhibited gastrointestinal adverse effects. In four cases, they were strong enough to cause the patients to withdraw. These gastrointestinal adverse effects were better tolerated under appropriate symptomatic treatment, but remained uncomfortable to many. Such adverse effects are well known with metformin treatment (Bonnet and Scheen, 2017). They require, however, particular attention and care in DM1 patients who suffer from disease-related gastrointestinal manifestations.

Abbreviations

Abbreviations
 
• 6MWT

  6-minute walk test

 
• DM1

  myotonic dystrophy type 1

 
• ITT

  intent-to-treat

Acknowledgements

We thank Dr Amit Chandra for linguistic revision of our paper, Raymond Zakhia and Jackie Gide of CECS/AFM (Corbeil-Essonnes), Dr Géraldine Honnet, Céline Labetoule and Florence Ganne of the Institute of Biotherapies (Evry), Isabelle Ledoux of Institute of Myology (Paris) and our colleagues of the Department of Physiotherapy of Henri-Mondor Hospital (Créteil) for valuable technical and management help in various parts of the trial.

Funding

The MYOMET clinical trial has been sponsored and fully funded by the Centre d'Etude des Cellules Souches, a research entity funded by the AFM-Telethon charity.

Competing interests

The authors report no competing interests.