Mechanistic Phase II Clinical Trial of Metformin in Pulmonary Arterial Hypertension.

Background Metabolic dysfunction is highly prevalent in pulmonary arterial hypertension (PAH) and likely contributes to both pulmonary vascular disease and right ventricular (RV) failure in part because of increased oxidant stress. Currently, there is no cure for PAH and human studies of metabolic interventions, generally well tolerated in other diseases, are limited in PAH. Metformin is a commonly used oral antidiabetic that decreases gluconeogenesis, increases fatty acid oxidation, and reduces oxidant stress and thus may be relevant to PAH. Methods and Results We performed a single-center, open-label 8-week phase II trial of up to 2 g/day of metformin in patients with idiopathic or heritable PAH with the co-primary end points of safety, including development of lactic acidosis and study withdrawal, and plasma oxidant stress markers. Exploratory end points included RV function via echocardiography, plasma metabolomic analysis performed before and after metformin therapy, and RV triglyceride content by magnetic resonance spectroscopy in a subset of 9 patients. We enrolled 20 patients; 19/20 reached the target dose and all completed the study protocol. There was no clinically significant lactic acidosis or change in oxidant stress markers. Metformin did not change 6-minute walk distance but did significantly improve RV fractional area change (23±8% to 26±6%, P=0.02), though other echocardiographic parameters were unchanged. RV triglyceride content decreased in 8/9 patients (3.2±1.8% to 1.6±1.4%, P=0.015). In an exploratory metabolomic analysis, plasma metabolomic correlates of ≥50% reduction in RV lipid included dihydroxybutyrate, acetylputrescine, hydroxystearate, and glucuronate (P<0.05 for all). In the entire cohort, lipid metabolites were among the most changed by metformin. Conclusions Metformin therapy was safe and well tolerated in patients with PAH in this single-arm, open-label phase II study. Exploratory analyses suggest that metformin may be associated with improved RV fractional area change and, in a subset of patients, reduced RV triglyceride content that correlated with altered lipid and glucose metabolism markers. Registration URL: http://www.clinicaltrials.gov; Unique identifier: NCT01884051.

fractional area change

false discovery rate

pulmonary arterial hypertension

Clinical Perspective

What Is New?

Although metabolism and insulin resistance have been recognized in pulmonary arterial hypertension, there have been no trials of therapies targeting these alterations in this deadly disease.

In this pilot trial, we found that metformin is safe in patients with pulmonary arterial hypertension and that, in secondary analyses, metformin may improve right ventricular function and noninvasive markers of right ventricular lipotoxicity that tracked with changes in the plasma metabolome.

What Are the Clinical Implications?

This trial provides a rationale for further study of metformin in pulmonary arterial hypertension and suggest that the plasma metabolome may be useful in identifying patients with improvements in right ventricular metabolism.

Metabolic dysfunction resembling features of insulin resistance is prevalent in pulmonary arterial hypertension (PAH) and likely contributes to both pulmonary vascular disease and right ventricular (RV) failure, the primary cause of death in PAH. Animal models of PAH demonstrate increased pulmonary vascular and right ventricular (RV) glycolysis, , , , , , impaired fatty acid oxidation, , and insulin resistance that is associated with more advanced pulmonary vascular disease, , accumulation of myocardial lipid, and impaired RV function. Observational studies in humans with PAH corroborate the clinical relevance of these findings: patients with PAH have an elevated prevalence of insulin resistance and impaired fatty acid metabolism compared with matched controls. , , , , Noninvasive imaging studies demonstrate increased glycolysis in the human PAH RV, and we have reported evidence of lipotoxic RV triglyceride accumulation in PAH RV. , , There is currently no cure for PAH and the availability of Food and Drug Administration‐approved, safe, and well‐tolerated metabolism‐targeted therapies coupled with the high clinical relevance of metabolic disease in PAH makes study of metabolic interventions in this disease potentially relevant.

There are no approved therapies for right heart failure in patients with PAH despite its high mortality. Multiple lines of evidence demonstrate that impaired glucose and lipid homeostasis affect the RV in PAH; , , , , , , however, whether correction of these alterations in human PAH is beneficial or may specifically affec the RV is largely unexplored. Metformin therapy in a rodent model of PAH reduces myocardial lipid and improves some indices of RV function. The mechanisms through which metformin may work are multiple, including improved insulin sensitivity, enhanced fatty acid oxidation, and reduced oxidant stress. Metformin is generally well tolerated with an excellent safety profile in humans including those without manifest diabetes mellitus, making it a potentially effective metabolic therapy in human PAH that may reduce oxidant stress. Further, left ventricular myocardial triglyceride is known to be increased in type 2 diabetes mellitus , , and can be modified by caloric restriction, suggesting that right ventricular lipid may also be modifiable.

We sought to test the hypothesis that metformin therapy is safe in human PAH and that it may have beneficial effects on the RV via exploratory analyses. We performed an open‐label, single‐arm phase II study to examine the safety and potential efficacy of metformin in patients with PAH. Our secondary exploratory end points were selected to assess potential mechanisms through which metformin may affect RV function and lipid content using proton magnetic resonance spectroscopy. Further, we explored the metabolomic consequences of metformin therapy in PAH to define potentially important pathways altered by this therapy and identify patients most likely to respond to therapy.

Methods

The authors will make the data, analytic methods, and study materials reported here upon request in accordance with our institutional review board and Vanderbilt University Medical Center policies. We performed a single‐center, open‐label, single‐arm study of the safety and efficacy of metformin in patients with PAH. The study was approved by Vanderbilt's institutional review board and all subjects gave written informed consent. The study was registered on clinicaltrials.gov before the first subject was enrolled (NCT01884051). Full details of the protocol can be found in Data S1. The primary safety outcomes were development of clinically significant lactic acidosis and study withdrawal attributed to metformin. The primary efficacy outcome was change in plasma markers of oxidant stress (isoprostanes, isofurans). Secondary outcomes included change in measures of glucose and lipid homeostasis, RV function on echocardiography, 6‐minute walk distance, myocardial lipid content (in subjects without a contraindication to cardiac magnetic resonance imaging), brain natriuretic peptide levels, and World Health Organization functional class. In exploratory analyses, we examined changes in metabolomic profiles before and after exposure to metformin. Details of testing procedures and biochemical assays are provided in Data S2. Safety was monitored by an independent data safety monitor (TR).

Inclusion and Exclusion Criteria

Full inclusion and exclusion criteria are presented in Data S1. Briefly, we included subjects who were ≥18 years of age diagnosed with heritable or idiopathic PAH. Relevant exclusion critieria included known type 1 diabetes mellitus, prednisone use, and evidence of chronic kidney disease or cirrhosis.

Study Procedures

After providing informed consent, subjects underwent testing at baseline and again after 8 weeks. Baseline testing included phlebotomy, 6‐minute walk distance, transthoracic echocardiography, and proton magnetic resonance spectroscopy to quantify myocardial triglyceride in a subset. Details of study procedures are described in Table S1. Interim visits at weeks 2, 4, and 6 for safety labs were conducted for the first 12 subjects. After no safety concerns emerged, the interim visits during weeks 2 and 6 were replaced with phone calls for the subsequent 8 subjects. All subjects received phone calls at 2, 4, and 6 weeks. Metformin was given in escalating doses as follows: 500 mg daily for 1 week, 500 mg twice daily for 1 week, 500 mg three times daily for 1 week, and 1 g twice daily for 5 weeks.

Statistical Analysis

We used standard graphing and screening techniques to detect outliers and to ensure data accuracy. We reported summary statistics for both numerical (mean±SD) and categorical (proportion and count) baseline demographics. For a comprehensive set of safety end points, insulin‐related end points, echocardiogram data, and 6‐minute walk distance, we reported summary statistics (mean±SD) at baseline and at the end of the study (week 8). Their change from baseline was evaluated using paired t test. The data analysis was conducted using the open‐source statistical package R (R Core Team, 2017) .

For the responder analysis, metabolomic data were scaled to set the median level of each metabolite to 1 across all samples. The ratio of postmetformin levels and premetformin levels was obtained for each patient and metabolite and compared to the difference between postmetformin and premetformin triglyceride content by magnetic resonance imaging. Patients were defined as high responders to metformin if their triglyceride content was decreased by 50% or greater after treatment, whereas low responders all had a <50% decrease in content. The ratio of post‐ and premetformin levels of each metabolite were then compared between responders and nonresponders by Mann‐Whitney nonparametric test, and P<0.05 was considered significant because the small sample size and exploratory nature of the analysis. Treating metabolite measure as count data, R package “DESeq2” was used to assess posttreatment metabolomic changes from baseline. In this analysis each metabolite was modeled using a Negative Binomial distribution. The posttreatment metabolite count was analyzed using generalized linear model with baseline in the model and the posttreatment log2 fold change was tested using the Wald test. Separate analyses were conducted to generate results with and without adjusting for fractional area change. False discovery rate (FDR) adjusted P values were reported. A cluster analysis was performed using metabolites with an FDR‐adjusted P<0.1. Data were first scaled and then log‐transformed to improve the distribution. Results were graphically presented with heatmaps. The linear relationship between pre‐ versus posttreatment change in metabolite level and the fractional area change (FAC) was assessed using Spearman's correlation coefficients. Linear mixed‐effect models were fitted to estimate the association between metabolite level and FAC. The fixed effect factors are treatment and time point (pre and post). Subjects were treated as a random effect factor to take into account the correlation between pre‐ and posttreatment measurements.

Metabolomics pathway analysis was performed using the R package “FELLA,” a tool for building hierarchical network representation of the chosen organism using the KEGG database. Metabolites satisfy the following conditions were used in the pathway analysis: (1) FDR‐adjusted P<0.1, (2) <25% missing data, (3) not in the Xenobiotics pathway, and (4) with a KEGG ID. Thirty‐three and 40 metabolites were used as input for the pathway analysis, respectively for analysis adjusted and not adjusted for FAC. Diffusion P scores based on the heat diffusion model were generated to compare a node to its null distribution under input permutation. The idea behind the heat diffusion is the usage of the finite difference formulation of the heat equation to propagate labels from the metabolites to the rest of the graph. The top scoring nodes contain not only relevant pathways, but also the intermediate entities that build a plausible explanation on how the input metabolites translate into reported pathways. For visualization purposes, top 100 nodes prioritized by P score were displayed on the graphs. The hypergeometric test was used to assess whether a biological pathway contains more hits within the input list than expected from chance given its size. Pathways were ranked according to their P value after multiple testing correction. The multiple testing was adjusted using the Benjamine‐Hochberg FDR criteria. Such adjusted P values were then compared against a prespecified FDR of 0.05 as they are reported as adjusted P<0.05. The same adjustment was applied to the metabolics analysis. For other analyses, the nominal P value was reported.

Results

Study Population

We screened 68 patients (see Figure 1) and enrolled 20 subjects (85% female) with an age of 42±10 years and body mass index of 30±6 kg/m2 (Figure S1). All consented subjects completed the study. Subject demographics and baseline clinical features are shown in Table 1. Most subjects had idiopathic PAH and 8/20 (40%) were receiving intravenous prostacyclin therapy. Mean 6‐minute walk distance in our cohort was 434±75 m. Features of the metabolic syndrome were common: 3 or more features of the metabolic syndrome were present in 5 patients.

Figure 1

Flow diagram.

DM indicates diabetes mellitus; MRS, magnetic resonance spectroscopy; PAH, pulmonary arterial hypertension; and PFT, pulmonary function test.

Table 1

Baseline Characteristics

Variable	Value
Age, y	42±10
Sex, n (%) female	17 (85)
Race, n (%)
White	17 (85)
Black	3 (15)
Body mass index, kg/m2	30±6
PAH etiology (heritable PAH/idiopathic PAH)	4/16
PAH medications, n (%)
Phosphodiesterase type 5 inhibitor	13 (65)
Endothelin receptor antagonist	5 (25)
Intravenous/subcutaneous epoprostenol or treprostinil	10 (50)
Inhaled prostacyclin analog	2 (10)
>1 PAH therapy	12 (60)
6‐minute walk distance, m	434±75
World Health Organization functional class
I	8
II	12
Features of metabolic syndrome*, n (%)
Elevated glucose	3 (15)
HDL cholesterol	6 (30)
Triglycerides	9 (45)
Obesity	10 (50)
Systemic hypertension	3 (15)
Number of MS features, n (%)
0	4
1	7
2	4
3	4
4	1
Right heart catheterization, most recent
Right atrial pressure, mm Hg	8.1±2.5
Mean pulmonary artery pressure, mm Hg	45.5±14.0
Pulmonary artery wedge pressure, mm Hg	13.0±5.1
Cardiac index, L/min per m2	2.4±0.7
Pulmonary vascular resistance, Wood units	7.3±4.1

Data are presented as mean±SD. HDL‐C indicates high‐density lipoprotein cholesterol; PAH, pulmonary arterial hypertension; MS, metabolic syndrome.

National Cholesterol Education Program Adult Treatment Panel III criteria used : glucose > or =100 mg/dL, HDL cholesterol <40 mg/dL men or <50 mg/dL women or drug treatment for low HDL‐C, triglycerides > or =150 mg/dL or drug treatment for elevated triglycerides, obesity waist > or =102 cm men, > or =88 cm women, hypertension > or =130/85 or drug treatment for hypertension.

Primary End Point: Safety

A total of 19/20 (95%) subjects reached the target metformin dose, and 1 subject was able to tolerate only a maximal dose of 1500 mg per day. The most common side effects were diarrhea, which occurred in 9/20 (45%) subjects, followed by heartburn (20%) and abdominal pain (15%) (Table 2). There were no serious adverse events and all subjects completed the 8‐week protocol. No subjects experienced clinical signs or symptoms of lactic acidosis. No patient required adjustment of PAH therapies or diuretics during treatment with metformin for worsening symptoms. There was no significant change in functional class (P=0.99). We observed a 1.9 kg reduction in body weight, consistent with known effects of metformin.

Table 2

Subject Reported New Side Effects and Adverse Events

Symptom	Frequency
Nausea	2 (10%)
Diarrhea	9 (45%)
Vomiting	1 (5%)
Abdominal pain	3 (15%)
Blue lips or fingers	1 (5%)
Chest pain	1 (5%)
Dyspnea	3 (15%)
Edema	3 (15%)
Fatigue	2 (10%)
Syncope	0
Heartburn	4 (20%)
Dizziness	0

Data are presented as N (%).

Primary End Point: Biochemical Outcomes

Although it was not significant, we observed a trend to reduction in the primary efficacy end point of plasma isoprostane (0.041±0.017 ng/mL pre versus 0.035±0.019 ng/mL post, P=0.18, Figure S2A). Similarly, the ratio of isoprostane/isofurans did not change significantly (8.5±5.7 pre versus 8.3±5.4 post, P=0.98, Figure S2B), though plasma isofurans were commonly undetectable. There was a trend to reduction in glycosylated hemoglobin from 5.33±0.42% to 5.21±0.27% (P=0.07) and there was no significant change in fasting insulin, glucose, or Homeostatic Model Assessment of Insulin Resistance (pre 1.9±1.7 versus post 1.7±1.0, P=0.57, Table S2). There was no effect on plasma lipid profile (Table S2).

Exploratory End Points: Clinical, Echocardiography, and RV Lipid Content

Metformin had no effect on 6‐minute walk distance in our study (pre 434±75 versus post 424±68 m, P=0.26) or brain natriuretic peptide level (66±77 pg/mL versus 80±159 pg/mL, P=0.60).

Full echocardiographic data are shown in Table 3 before and after metformin exposure. We observed a 3% absolute increase in FAC, a global measure of RV systolic function (pre 23±8% versus post 26±6%, P=0.02, Table 3, Figure S3). There were no changes in measures of RV longitudinal function (tricuspid annular plane systolic excursion, S'; Table 3). There was no change in tricuspid regurgitant velocity or estimated pulmonary artery systolic pressure.

Table 3

Echocardiographic Measurements Before and After Metformin

Variable	Baseline	8 wk	P Value
Left ventricular ejection fraction, %	70±5	70±5	0.67
Left atrial volume index, mL/m2	21±8	20±6	0.51
Ratio of early mitral inflow velocity/mitral annular early diastolic velocity	10.4±3.5	10.8±2.7	0.60
Right ventricle basal dimension, cm	4.4±0.6	4.4±0.6	0.90
Tricuspid regurgitant peak velocity, m/s	3.3±0.7	3.4±0.7	0.90
Right ventricle systolic pressure, mm Hg	51±21	52±21	0.85
Tricuspid annular plane systolic excursion, cm	1.9±0.4	2.0±0.3	0.37
Tricuspid S', mm/s	11.8±3.5	11.5±2.1	0.23
Fractional area change, %	23±8	26±6	0.02

Data are presented as mean±SD.

Eight subjects had no contraindication to cardiac magnetic resonance imaging and were able to complete the imaging protocol. Myocardial triglyceride was markedly elevated at baseline , and decreased significantly with metformin (3.2±1.8% to 1.6±1.4%, P=0.015; Figure 2). Triglyceride content declined in 8/9 subjects.

Figure 2

Effect of metformin on right ventricular triglyceride content and metabolomic correlates of improvement.

A, Change in myocardial triglyceride content by cardiac magnetic resonance spectroscopy (n=8). Published reference ranges for obesity and diabetes mellitus are shown. , B, Plasma metabolomic pattern was compared pre‐ and postmetformin therapy and segregated by ≥50% reduction in right ventricular (RV) triglyceride (Responders) or <50% reduction (non‐responders). Significantly different ratios were identified in several plasma metabolites of relevance to PAH and lipid and glucose metabolism. *P<0.05.

Exploratory End Point: Metabolomics

In an exploratory analysis, we sought to define peripheral blood correlates with significant improvement in RV triglyceride content, defining high responders as a ≥50% reduction in RV triglyceride content between initial and follow‐up magnetic resonance spectroscopy. Using nontargeted metabolomics (Metabolon, Research Triangle Park, NC), we found significant differences in metabolites of potential relevance to pulmonary hypertension and right heart failure including dihyroxybutyrate, acetylputriscene, hydroxystearate, and glucuronate (Figure 2B, all adjusted P<0.05 versus baseline, see Table S3 for full list of significant metabolites).

We further sought to determine which pathways metformin might be affecting including all enrollees. In total, 143 biochemicals were significantly different between baseline and end point with 84 increased and 58 decreasing in concentration. Using a principal clustering analysis, there was not a significant shift between the 2 groups in aggregate; however, when analyzing individual patients, there were noted significant alterations as a function of metformin therapy as shown in Figure S4 and demonstrated in the hierarchical clustering analysis in Figure 3A. A random Forest analysis demonstrated biomarkers that may separate the 2 groups (baseline and after therapy). A predictive accuracy of 80% was identified, which is improved from 50%, which would be expected by random chance. Separation was driven most strongly by lipid metabolites, which made up the top 5 most important to group separation (Figure 3B). Markers of lipid beta‐oxidation such as the medium and long‐chain acylcarnitine, whose synthesis is an entry point to mitochondrial beta oxidation, were increased after metformin therapy (hydroxybutyrylcarnitine, hydroxypalmitoylcarnitine, and hydroxyoleoylcarnitine all adjusted P<0.05 versus baseline).

Figure 3

Effects of metformin on plasma metabolome in pulmonary arterial hypertension.

A, Heatmap demonstrating significantly altered pathways after metformin therapy. B, Random Forest analysis of baseline vs end point demonstrating significantly altered metabolites and their pathway.

In addition to the changes in lipid metabolites with metformin therapy, there were other pathways that were significantly affected. There was a small, nonsignificant decrease in glucose, whereas levels of lactate and pyruvate were increased significantly at end point compared with baseline (1.20‐ and 1.28‐fold increase respectively, adjusted P<0.05). Although there was no clinically significant lactic acidosis, these changes likely reflect an increase in aerobic glycolysis rates with metformin therapy. Similarly reflecting reduced glucose availability, glucose‐related metabolism markers, for example, pentose metabolites, were significantly decreased at end point compared with baseline (adjusted P<0.05). Metformin therapy was associated with significant reductions in metabolites in the urea cycle, tryptophan metabolites, and other markers of inflammation such as DiHOMEs produced by active leukocytes and markers of oxidative stress (adjusted P<0.05 for all versus baseline, Figures S5 through S7).

We next sought to understand the changed metabolites using a pathway analysis. Metabolites were included in the analysis if they had (1) FDR‐adjusted P<0.1, (2) <25% undetectable data, (3) not in the Xenobiotics pathway, and (4) an assigned KEGG ID. For visualization purposes, the top 100 nodes prioritized by P score are displayed in Figure 4. We identified the following key nodes in descending order of strength of association: arginine, malate, 4‐methyl‐2‐oxopentanoate, mannose, citrulline. The sex steroids dhydroepiandrosterone and androstenediol monosulfate were also identified as significantly changed by metformin (1.32‐ and 1.24‐fold higher respectively, adjusted P<0.05).

Figure 4

Pathway analysis to identify key nodes in changed metabolites.

 

Discussion

We sought to determine the safety and a potential signal for efficacy for metformin in PAH. Because our prior work in rodent models suggested that metformin improves RV function and lipid deposition in the PAH RV and prior literature had noted improved oxidant stress with metformin in human studies and some, but not all, rodent models, , , , we focused on these potential exploratory efficacy metrics in our trial. First, our study demonstrated that metformin is likely well tolerated in PAH. Although there was no significant change in plasma markers of oxidant stress, the trend was toward reduction in these markers and not worsening. Our exploratory analyses further suggested that metformin may improve RV function and reduce RV lipid deposition in PAH.

Most important, metformin appeared to be safe in patients with PAH without advanced heart failure symptoms. We enrolled a high proportion of patients with parenteral prostacyclin therapy (50%) and combination PAH medical therapy (60%), suggesting a patient population with a significant burden of pulmonary vascular disease and currently using drugs with well‐described gastrointestinal side effects including diarrhea. , Despite this, only 1/20 patients did not reach the target dose owing to side effects. Further, there are reports of metformin causing lactic acidosis in the literature; however, the incidence is so rare that metformin is now considered safe in patients with heart failure. , We thus thought it was safe to proceed with a trial in a disease categorized by right heart failure. Most enrolled patients had evidence of reduced RV FAC and, yet, tolerated the drug well. Although there was a nonclinically significant increase in plasma lactate by metabolomics, this biochemical alteration may reflect the metabolic effect of metformin on lactate metabolism, , rather than a clinically important lactic acidosis as no patient suffered an adverse event.

Our prior work had demonstrated in an animal model of RV failure in PAH that metformin improves RV function and may improve histologic evidence of lipid deposition. Further, we have previously published that patients with PAH had increased RV triglyceride content. Based on these data, we tested the hypothesis that metformin therapy improves RV lipid content and function in human patients with PAH. We tested this hypothesis using echocardiography, because of its clinical relevance. In addition, we recruited a subcohort of patients to undergo cardiac magnetic resonance spectroscopy. Despite relatively few enrollees, we observed a significant decrease in percentage of cardiac triglyceride, signifying reduced RV lipid content. We did not measure invasive hemodynamics; however, there was no change in RV systolic pressure estimates, suggesting that if there are pulmonary vascular effects of metformin therapy, these are less pronounced than the effects on the RV. Although metformin did not normalize RV FAC, there was a significant improvement even in our short‐term study of RV function with an absolute difference of 3% from baseline to study end, which may be clinically significant. Taken together, these data suggest that metformin could have a direct effect on the failing RV and that metformin's RV effects should be tested in a more robust fashion in future studies.

We further hypothesized that peripheral blood metabolomic changes may reflect RV metabolism. We thus we sought metabolomic correlates in the peripheral blood of subjects who achieved a >50% reduction in RV lipid content in an exploratory analysis. We identified several compounds that may be of relevance to RV and PAH metabolism. Dihyroxybutyrate, a ketone synthesized from fatty acids during ketogenesis, was reduced, which may suggest enhanced lipid beta oxidation. Elevation of this metabolite has been reported in children with persistent pulmonary hypertension of the newborn indicated potential relevance to pulmonary vascular disease. Acetylputrescine was elevated and associated with mortality in a large multinational cohort of patients with PAH and was reduced in patients with significant reduction in RV triglyceride after metformin in our cohort. Alterations in hydroxysterate and glucuronate similarly suggest potential changes in lipid and glucose metabolism of relevance to PAH. , Safai et al recently demonstrated the effect of metformin on branched chain amino acid metabolites in a clinical trial; however, the larger, particularly extended lipid metabolite, effects of metformin in humans are largely unknown. Our pathway analysis suggests a major effect of metformin on urea cycle metabolism, sex hormones, and glucose and fat metabolism in patients with PAH. Although our data do not show that RV lipid deposition accelerates RV failure, or that these biomarkers reflect RV composition changes specifically, they do suggest there may potentially be peripheral blood markers of drug responses specific to metformin and to the RV that could be studied prospectively.

Our work has limitations. This was a single‐center, open‐label study focused on safety and potential early markers of efficacy and thus there was no control group and we were underpowered to detect markers of clinical efficacy. Our RV findings, in particular, require confirmation in a larger, multicenter trial that is now enrolling (NCT 03617458). We were unable to measure RV ejection fraction by magnetic resonance imaging, a more accurate measure of RV function than echocardiographic metrics, due to the duration of the spectroscopy protocol and we recognize that the low number of participants in this procedure may bias to error, however the results were nonetheless compelling and the body mass index was similar in this substudy (30±4.8) to the rest of the trial participants. We did not identify a significant change in plasma markers of oxidant stress (isoprostanes, isofurans); however, we also did not find any concerning worsening of these markers either that would preclude further study. Although approximately half of our patients were obese, they are similar in body mass index to other US‐based studies of PAH. , The positive exploratory analyses are likely not attributable to simply improving obesity as the final mean body mass index was 29.7±5.9. Additionally, although all plasma samples were taken in the fasting state and subjects were not instructed to change their diet, we cannot rule out a change in diet that is reflected in the plasma metabolome.

In conclusion, metformin therapy was safe and well tolerated in patients with PAH in this single‐arm, open‐label mechanistic phase II study. Exploratory analyses suggest that metformin may be associated with improved RV FAC and, in a subset of patients, reduced RV triglyceride content that correlated with altered lipid and glucose metabolism markers. Further study of metformin in PAH is warranted.

Sources of Funding

This work was supported by NIH 1 R01 HL 122417‐01A1 (Hemnes), NIH 5 P01 HL 108800‐07 (Loyd & Hemnes, Co‐PI), American Heart Association 13FTF16070002 (Brittain), This publication was supported by CTSA award No. UL1 TR002243 from the National Center for Advancing Translational Sciences. Its contents are solely the responsibility of the authors and do not necessarily represent official views of the National Center for Advancing Translational Sciences or the National Institutes of Health.

Disclosures

ARH has served as a consultant to Actelion, Bayer, Complexa, United Therapeutics. She has received grant funding from the Cardiovascular Medical Research and Education Fund and National Institutes of Health and holds stock options with PHPrecisionMed. IMR has received research support from Complexa and Bayer. The remaining authors have no disclosures to report.

Data S1–S2

Tables S1–S3

Figures S1–S7

References , , ,

Click here for additional data file.