Effect of Metformin on Microvascular Endothelial Function in Polycystic Ovary Syndrome

Abstract

Objective:

To investigate the factors that are associated with the effect of metformin on endothelial dysfunction in polycystic ovary syndrome (PCOS).

Patients and Methods:

From March 24, 2014, to November 18, 2016, 48 women with PCOS were randomly assigned to 1500 mg/d of metformin (N=29) or no treatment (N=13) for 3 months; 42 patients (29 in the initial treatment group and 13 in the no treatment group) completed the study. Study variables were measured at baseline and after 3 months. Participants who did not receive metformin initially were then treated with metformin for another 3 months, and study variables were measured again. Endothelial function was measured as reactive hyperemia–peripheral arterial tonometry (RH-PAT) from the index finger.

Results:

The age and baseline endothelial function (mean ± SD) of the participants were 32.7±6.9 years and 1.8±0.5, respectively. No notable change was observed in endothelial function after 3 months with metformin compared with no treatment. However, after stratifying participants who received metformin based on baseline endothelial function, there was a significant improvement following metformin treatment in participants with abnormal baseline endothelial function (1.3±0.3 vs 1.7±0.3; P<.001) but not in those with normal baseline endothelial function (2.1±0.4 vs 2.0±0.5; P=.11).

Conclusion:

Metformin improves endothelial function in women with PCOS and endothelial dysfunction independent of changes in glucose metabolism, dyslipidemia, or presence of prediabetes. Metformin has a direct effect on endothelial function in PCOS, and measurement of endothelial function can stratify and follow response to metformin treatment in PCOS.

Trial Registration:

clinicaltrials.gov Identifier: NCT02086526.

Polycystic ovary syndrome (PCOS) is a highly prevalent endocrine disorder affecting up to 19.9% of women of reproductive age.¹ The combination of increased androgens and insulin resistance in PCOS results in an increased risk for the cardiometabolic syndrome.^2,3 Impaired glucose tolerance (IGT), type 2 diabetes mellitus, obesity, dyslipidemia, metabolic syndrome, and nonalcoholic fatty liver disease have been found to be more prevalent in PCOS than in healthy controls.^4–6 Subsequently, PCOS has also been associated with biomarkers of subclinical atherosclerosis such as coronary and aortic calcification, greater carotid intima-media thickness,^7–9 and endothelial dysfunction compared with healthy controls.^7,8

Metformin treatment improves ovulatory frequency and can improve both insulin resistance and hyperandrogenism.⁹ Metformin treatment is also associated with weight loss and improved plasma lipid profile in PCOS,^9,10 although this is not universally observed in all women with PCOS.¹¹ There is also mixed evidence regarding metformin’s effect on endothelial function.^7,12,13 Differences seen between studies could be related to greater effect seen specifically with macrovascular vs microvascular endothelial function or due to the heterogeneous phenotypes of PCOS and specific factors related to these phenotypes.¹¹

We sought to determine whether specific metabolic, hormonal, or phenotypic features of PCOS might be associated with metformin’s effect in PCOS on peripheral microvascular endothelial function or if metformin can improve endothelial function in women with PCOS and endothelial dysfunction. The use of digital measurement of microvascular endothelial function could be more clinically relevant in the treatment of PCOS because it is less dependent on expertise required for the measurement of macrovascular flow-mediated dilation and interpretation of the data.¹⁴

PATIENTS AND METHODS

Study Design and Treatment Allocation

We conducted an open-label study of women with PCOS randomized to metformin vs no treatment at Mayo Clinic in Rochester, Minnesota, from March 24, 2014, to November 18, 2016 (clinicaltrials.gov Identifier: NCT02086526). The study protocol was approved by the Institutional Review Board of Mayo Clinic. Written informed consent was obtained from each participant before study enrollment. Participants could not be taking any other medications for the treatment of PCOS for at least 3 months before screening and enrollment.

We randomly assigned 48 women with the diagnosis of PCOS to metformin (1500 mg daily, extended-release) or no treatment (delayed start, Figure 1A). Metformin dosage was titrated to 1500 mg weekly during the first 3 weeks of the trial. Participants randomized to the no treatment arm had the option to continue the study for an additional 3 months, during which they received metformin and completed an additional set of study measurements after metformin treatment (Figure 1B).

FIGURE 1.

Participants’ assignments to immediate or delayed metformin treatment. A, Number of participants in each step and treatment received. B, Timing of follow-up visits and workups.

Definitions and Procedures

Premenopausal women aged 18 to 50 years with a body mass index (BMI) of 25 or greater who had a diagnosis of PCOS were enrolled in this study. The diagnosis of PCOS was made by an endocrinologist at Mayo Clinic using Rotterdam criteria.¹⁵ Criteria of both oligoanovulation and androgen excess were required for inclusion in the study to increase the likelihood of insulin resistance. Oligoanovulation was defined as less than 9 menses per year or the absence of a progesterone increase above 10 ng/mL (to convert to nmol/L, multiply by 3.18) in the luteal phase (days 20–22 of menstrual cycle) in women with monthly menses. Androgen excess was defined as elevated testosterone or dehydroepiandrosterone sulfate, severe acne, androgenic alopecia, or clinical hirsutism (Ferriman-Gallwey score >8). Exclusion criteria included elevated prolactin level, untreated hypothyroidism or hyperthyroidism, Cushing syndrome, congenital adrenal hyperplasia, diabetes, creatinine level greater than 1.5 mg/dL (to convert to μmol/L, multiply by 88.4), pregnancy, breastfeeding, smoking, taking oral contraceptive pills or other medications that would affect androgen levels, insulin sensitivity, or endothelial function.

Body mass index was calculated as weight in kilograms divided by height in meters squared. Homeostasis model assessment of insulin resistance (HOMA-IR) was calculated by multiplying fasting plasma glucose (mg/dL) by fasting plasma insulin (U/L) divided by 405.¹⁶

Endothelial function values of more than 1.6 were considered normal, and those less than or equal to 1.6 were considered abnormal (endothelial dysfunction).¹⁷

Impaired fasting glucose was defined as fasting glucose levels of 100 to 125 mg/dL (5.6 to 6.9 mmol/L), and IGT was defined as 2-hour glucose levels of 140 to 199 mg/dL (7.8 to 11.0 mmol/L) following a 75-g oral glucose tolerance test (OGTT). Impaired fasting glucose and IGT are prediabetic states of hyperglycemia, associated with insulin resistance and increased risk of cardiovascular pathology, and thus patients with impaired fasting glucose and IGT are an important target group for primary prevention.¹⁸

Participants’ height and weight were measured with light clothing without shoes. Waist circumference was measured in the midpoint between the iliac crest and the lower rib after normal expiration. Hip circumference was measured as the widest diameter of the hips. A standard mercury sphygmomanometer was used to measure systolic and diastolic blood pressure after 10 minutes of resting in a sitting position.

Before placement of intravenous access, an EndoPAT 2000 device (Itamar Medical Ltd) was used to assess endothelial function. As previously described, it is a noninvasive method for the assessment of the peripheral endothelial function, correlates well with coronary endothelial function, and has good reproducibility^17,19–22 Finger reactive hyperemia (RH)–peripheral arterial tonometry (PAT), which measures flow-mediated dilation, was used to assess endothelial function. The measurement methods have been published previously.^17,20,23 Briefly, the EndoPAT 2000 device uses finger probes that record and analyze the volume of blood. Two finger probes were placed on 2 index fingers of each participant to continuously record the blood volume changes accompanying pulse waves. A standard blood pressure cuff was placed on one arm. At first, 5 minutes of PAT recording was done for equilibration. Next, while the PAT recording was still continuing, the cuff was inflated to higher than systolic pressure and held for 5 minutes. After 5 minutes, the cuff was deflated, and the PAT recording continued for another 5 minutes. The recordings from the other index finger were used as a control for each measurement. The recorded RH-PAT data were analyzed by computer software through an operator-independent process. The RH-PAT index was obtained as a measure of RH. It was calculated as the average PAT signal amplitude during 1 minute starting from 1.5 minutes after cuff deflation divided by the average PAT signal amplitude during the 2.5 minutes before cuff inflation.

Venous blood samples were collected after 12 hours of overnight fasting for measurements of fasting plasma glucose, hemoglobin A_1c, C-peptide, fasting plasma insulin, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, triglycerides, C-reactive protein, leptin, adiponectin, estradiol, dehydroepiandrosterone sulfate, and total testosterone. After baseline samples were obtained, a 75-g OGTT was performed. After the participant ingested 75 g of glucose, blood samples were collected every 10 minutes for the following 3 hours to measure plasma glucose and insulin levels. The OGTT samples were collected via a retrograde venous catheter with the hand being placed within a heating box and warmed to approximately 60°C to collect arterialized blood.

Study variables were measured twice for those initially assigned to metformin, once before and once after 3 months of treatment. For control participants who were in the no treatment group and received metformin treatment later on (delayed start group), variables were measured 3 times: 2 times before metformin therapy (day 1 and day 90) and once after metformin therapy (after 6 months) (Figure 1B).

Statistical Analyses

The study participants were first grouped based on the randomized treatment (metformin vs no treatment). Next, all participants treated with metformin (including those in the metformin group and those in the delayed start group) were divided into 2 groups according to their baseline endothelial function—normal or endothelial dysfunction as defined previously. Baseline characteristics of the participants are presented as mean ± SD for normally distributed continuous variables, median and interquartile range for nonnormally distributed continuous variables, and number (percentage) for dichotomous variables. The Kolmogorov-Smirnov test was used to assess normality in the study variables. The independent sample t test was used to compare means for normally distributed variables and the Mann-Whitney U test for nonnormally distributed variables at baseline. The χ² test was used to compare the dichotomous variables between study groups at baseline. A paired t test was used to compare means for normally distributed variables and the Wilcoxon signed rank test for nonnormally distributed variables within each study group before and after treatment. The McNemar test was used to compare the dichotomous variables within each study group before and after treatment. Linear regression analysis was used to assess predictors of change in RH-PAT index following metformin treatment. With the sample size of 42, our study had the power of 80% to detect the mean difference of 0.3 in RH index before and after treatment with metformin. SPSS statistical software, version 22 (IBM Corp) was used for data analysis. Two-sided P<.05 was considered statistically significant.

RESULTS

In this study, 48 women with PCOS were randomly assigned to metformin (33 patients) or no treatment (15 patients). Four participants in the metformin treatment group (12.1%) and 2 participants in the no treatment group (13.3%) did not complete the trial (Figure 1A). Of the participants in the no treatment group who completed the trial, 13 continued the study to take metformin for 3 months (delayed start group). Among those, 12 participants completed therapy.

Table 1 shows the characteristics of the metformin and no treatment groups at baseline and after 3 months. The mean ± SD age of the participants was 32.7±6.9 years. The participants were obese (BMI, 36.8±9.4 kg/m²). Mean baseline endothelial function was 1.8±0.5 for the study population. Table 1 also shows the characteristics of the delayed start and metformin groups at baseline and after 3 months. There was no notable difference between baseline characteristics in the two groups. There was no notable change in study variables in the delayed start group after 3 months. In metformin-treated participants, there was a significant decrease in body weight (P<.05), fasting plasma glucose (P<.01), fasting plasma insulin (P<.05), HOMA-IR (P<.05), total plasma cholesterol (P<.05), and testosterone (P<.001) after 3 months of treatment. There was no notable change in endothelial function in the 2 groups during this period.

TABLE 1.

Characteristics of the Study Population at Baseline and After 3 Months^a,b,c,d

Variable	Metformin group (N=29)		Delayed treatment group (N=13)
	Baseline	Month 3	Baseline	Month 3
Age (y)	32.4±7.5	NA	33.1±5.9	NA
Height (cm)	166.1±6.9	166.2±7	165.4±5.5	165.7±6.7
Weight (kg)	103.4±30.8	102.3±31.6e	104.7±20.9	100.9±19.7
BMI (kg/m2)	37.1±9.1	36.2±10.3	38.4±8.3	37.7±8.1
Waist circumference (cm)	107.6±18.8	106.8±18.4	111.4±15.7	113.1±21.5
Hip circumference (cm)	123±18.4	120.2±14	124.4±14.4	123.9±15.1
Waist to hip circumference	0.9±0.1	0.9±0.1	0.9±0.1	0.9±0.1
SBP (mm Hg)	120.7±14.5	117.9±12.1	120.2±17.9	118.8±16.4
DBP (mm Hg)	72.8±10.2	69.7±9.1	73.5±14.1	71.3±12.2
Hypertension	4 (13.8)	2 (6.9)	2 (15.4)	2 (15.4)
FPG (mg/dL)	94.4±11.1	87.5±9.4f	89.8±8.1	91.3±9.2
IFG	5 (17.2)	2 (6.7)	1 (7.7)	1 (7.7)
2-h Glucose OGTT (mg/dL)	143.5±23.5	142.3±28.3	143±23	136.4±31.8
IGT	13 (44.8)	11 (37.9)	6 (46.2)	7 (53.8)
HbA1c (%)	5.3±0.3	NA	5.3±0.3	NA
C-peptide (ng/mL)	1.1 (0.74–1.3)	0.93 (0.59–1.2)	0.93 (0.59–1.2)	0.86 (0.54–1)
Insulin (μIU/mL)	13.7 (9.8–24.1)	10.2 (5.9–16)e	10 (7.75–22.75)	11.2 (5.4–17.9)
HOMA-IR	3.2 (2.4–5.5)	2.1 (1.1–3.5)f	2.1 (1.6–5.6)	3.6 (1.4–6.5)
IR (%)	63.3	46.7	58.3	58.3
Total cholesterol (mg/dL)	177.7±30.3	169.4±26.2e	170±27.4	170.8±24.3
HDL-C (mg/dL)	45.6±13.7	45.7±12.1	47.8±13.4	51±21.7
Adiponectin (mg/dL)	6046.7±2786.4	5946.9±2928.5	6122.2±2478.1	6413.7±2499.4
Leptin (μg/L)	47.4±23.3	48.4±23.2	45.8±15.1	43.2±18.3
LDL-C (mg/dL)	104.1±31.6	101.8±19.8	100.8±25.5	100.6±20.2
TG (mg/dL)	123.1±75.7	109.7±47.9	106.6±43.8	95.6±30.3
Estradiol (pg/mL)	49.6 (35–98.7)	42.4 (33.4–56.5)	61.6 (40–121.5)	58.9 (46–71)
DHEAS (μg/dL)	147 (73.4–257.5)	123 (75.1–246)	150 (99.3–164)	126 (98.5–150)
Total testosterone (ng/dL)	33 (23.5–37)	24 (15.5–35.5)g	31 (24–61)	27.5 (21.3–44)
CRP (mg/L)	2.3 (1.2–5.4)	3.1 (1.4–6.2)	5.1 (1.4–11.3)	7.2 (4.3–11.1)
RHI	1.9±0.6	1.9±0.5	1.7±0.4	1.8±0.5

^aBMI = body mass index; CRP = C-reactive protein; DBP = diastolic blood pressure; DHEAS = dehydroepiandrosterone sulfate; FPG = fasting plasma glucose; HDL-C = high-density lipoprotein cholesterol; HOMA-IR = homeostasis model assessment of insulin resistance; IFG = impaired fasting glucose; IGT = impaired glucose tolerance; IR = insulin resistance; LDL-C = low-density lipoprotein cholesterol; NA = not applicable; OGTT = oral glucose tolerance test; RHI = reactive hyperemia index; SBP = systolic blood pressure; TG = triglycerides.

^bData are presented as mean ± SD for normally distributed continuous variables, median (interquartile range) for nonnormally distributed continuous variables, and No. (percentage) for dichotomous variables.

^cSI conversion factors: To convert glucose values to mmol/L, multiply by 0.0555; to convert HbA_1c values to proportion of total hemoglobin, multiply by 0.01; to convert C-peptide values to nmol/L, multiply by 0.331; to convert insulin values to pmol/L, multiply by 6.945; to convert cholesterol values to mmol/L, multiply by 0.0259; to convert triglyceride values to mmol/L, multiply by 0.0113; to convert estradiol values to pmol/L, multiply by 3.671; to convert DHEAS values to μmol/L, multiply by 0.027; to convert testosterone values to nmol/L, multiply by 0.0347; to convert CRP values to nmol/L, multiply by 9.524.

^dThere was no significant difference between the 2 groups at baseline.

^eP<.05, comparing values at baseline and after 3 months.

^fP<.01, comparing values at baseline and after 3 months.

^gP<.001, comparing values at baseline and after 3 months.

To further explore the effect of metformin therapy on endothelial function, all study participants who received metformin were stratified according to baseline endothelial function (normal vs endothelial dysfunction) (Figure 2). Table 2 compares the characteristics of the study participants stratified by baseline endothelial function before and after 3 months of treatment with metformin. As shown in Table 2, the group with endothelial dysfunction did not have significantly different concentrations of fasting glucose (P<.01), insulin (P<.05), and C-peptide (P<.05) compared with the normal group. There was no significant change in endothelial function in participants with baseline normal endothelial function following 3 months of metformin treatment (2.1±0.4 vs 2.0±0.5; P=.11) (Table 2). In contrast, there was a significant improvement in participants with abnormal baseline endothelial function (1.3±0.3 vs 1.7±0.3; P<.001).

FIGURE 2.

Change in endothelial function following treatment with metformin. Error bars represent 95% CIs of the mean.

TABLE 2.

Characteristics of the Study Population at Baseline and After 3 Months of Metformin Therapy, Stratified by Baseline Endothelial Function Status^a,b,c

	Normal baseline endothelial function (N=26)		Endothelial dysfunction (N=15)
Variable	Baseline	Month 3	Baseline	Month 3
Age (y)	33.8±6.9	NA	30.6±6.7	NA
Height (cm)	165.3±5	165.8±6.4	166.9±8.6	166.6±8.7
Weight (kg)	104.4±28.8	102.8±31.4d	102.7±27.2	100.7±25.9
BMI (kg/m2)	37.9±9.2	37.5±9.9	36.7±8.4	35.0±9.7
Waist circumference (cm)	107.6±15.4	111.8±24.3d	110.9±21.9	109.9±19.9
Hip circumference (cm)	119.4±28	122.6±14.7	117.8±13.1	117.3±12.4
Waist to hip circumference	0.8±0.2	0.9±0.1	0.9±0.1	0.9±0.1
SBP (mm Hg)	121.2±12.5	119.4±12.5	119.4±16.4	116.1±14.0
DBP (mm Hg)	73.3±11.9	70.7±8.9	72.5±10.5	69.1±11.2
Hypertension	4 (15.4)	2 (7.7)	2 (13.3)	2 (13.3)
FPG (mg/dL)	93.7±10.4	85.6±7.6e	91.7±10.7	91.3±9.2
IFG (%)	3 (11.5)	1 (3.8)	3 (20)	1 (6.7)
2-h OGTT glucose (mg/dL)	141.5±25.3	138.7±30.6	146.5±19	145.2±25.7
IGT	11 (42.3)	10 (38.5)	8 (53.3)	6 (40.0)
HbA1c (%)	5.3±0.3	NA	5.3±0.3	NA
C-peptide (ng/mL)	0.99 (0.76–1.2)	0.83 (0.75–1.15)d	0.99 (0.57–1.4)	0.82 (0.51–1.3)
Insulin (μlU/mL)	11.6 (9–20.3)	11.6 (8.1–16.1)d	11.9 (7.5–24.6)	6.9 (3.3–16.2)
HOMA-IR	2.9 (2.2–4.9)	2.7 (1.6–3.6)e	2.7 (1.5–6.0)	1.7 (0.6–4.1)d
Total cholesterol (mg/dL)	175.4±27.5	168.3±23.9	175.1±33.3	172.5±29.5
HDL-C (mg/dL)	44.1±12.5	44.1±10.5	50.1±14.6	51.4±18.4
Adiponectin (μg/dL)	6084.7±2382.6	5836.1±2095.2	6044±3202	6924.1±3615.4
Leptin (μg/L)	46.6±20.2	47.2±21.6	47.6±23.4	46.8±23.3
LDL-C (mg/dL)	102.6±31.9	101.7±20.2	103.8±25.8	101.4±18.9
TG (mg/dL)	124.9±76	111.9±52.5	105.7±47.8	98.3±26.3
Estradiol (pg/mL)	47 (33.1–96.3)	46.6 (34.3–55.1)	51 (42–149)	46.1 (38.8–145)
DHEAS (μg/dL)	95.4 (72.8–164)	121 (74.7–217)	164 (149–308)f	173.5 (114.1–307.7)
Total testosterone (ng/dL)	29 (22–42)	22 (15.5–30.5)g	35 (28–50)	30 (21–43)
CRP (mg/L)	3.4 (1.1–5.1)	2.2 (1.2–5.4)	5.3 (1.4–11.2)	5.3 (1.1–9.2)
RHI	2.1±0.4	2.0±0.5	1.3±0.3h	1.7±0.3g

^aBMI = body mass index; CRP = C-reactive protein; DBP = diastolic blood pressure; DHEAS = dehydroepiandrosterone sulfate; FPG = fasting plasma glucose; HDL-C = high-density lipoprotein cholesterol; HOMA-IR = homeostasis model assessment of insulin resistance; IGT = impaired glucose tolerance; IFG = impaired fasting glucose; LDL-C = low-density lipoprotein cholesterol; NA = not applicable; OGTT = oral glucose tolerance test; RHI = reactive hyperemia index; SBP = systolic blood pressure; TG = triglycerides.

^bData are presented as mean ± SD for normally distributed continuous variables, median (interquartile range) for nonnormally distributed continuous variables, and No. (percentage) for dichotomous variables.

^cSI conversion factors: To convert glucose values to mmol/L, multiply by 0.0555; to convert HbA_1c values to proportion of total hemoglobin, multiply by 0.01; to convert C-peptide values to nmol/L, multiply by 0.331; to convert insulin values to pmol/L, multiply by 6.945; to convert cholesterol values to mmol/L, multiply by 0.0259; to convert triglyceride values to mmol/L, multiply by 0.0113; to convert estradiol values to pmol/L, multiply by 3.671; to convert DHEAS values to μmol/L, multiply by 0.027; to convert testosterone values to nmol/L, multiply by 0.0347; to convert CRP values to nmol/L, multiply by 9.524.

^dP<.05, comparing values at baseline and after 3 months inside each group.

^eP<.01, comparing values at baseline and after 3 months inside each group.

^fP<.05, comparing values between two groups at baseline.

^gP<.001, comparing values at baseline and after 3 months inside each group.

^hP<.001, comparing values between two groups at baseline.

Next, we assessed what factors can predict the improvement of endothelial function following metformin treatment when adjusted for other variables. Table 3 shows the univariable and multivariable regression analysis considering the individual change in endothelial function as the dependent variable. Impaired baseline endothelial function was the only variable that was significantly associated with improved endothelial function following treatment with metformin in the univariable analysis (β=.519; P=.001). Because we had improvements of endothelial function in 23 participants, we could include only 2 variables in the multivariable model. Since we observed consistent significant changes in HOMA-IR following metformin treatment (Table 1 [P<.01] and Table 2 [P<.01]) and it has been reported to be significantly associated with impaired endothelial function,²⁴ we included HOMA-IR and baseline endothelial function in the final multivariable model (Table 3). As shown, baseline endothelial dysfunction remained the single significant predictor of change in endothelial function following adjustment for HOMA-IR (β=.518; P=.002).

TABLE 3.

Linear Regression Analysis Considering Endothelial Function Change After Metformin Administration Dependent Variable

	Univariable		Multivariable
Variable	β	P value	β	P value
Baseline endothelial dysfunction	.519	.001	.518	.002
Age	−.004	.76	NA	NA
BMI	−.002	.81	NA	NA
Hypertension	.161	.61	NA	NA
FPG	−.003	.71	NA	NA
IGT	.114	.49	NA	NA
Insulin	.001	.87	NA	NA
HOMA-IR	.007	.833	.001	.92
Estradiol	.002	.15	.002	.15
Total testosterone	.004	.46	NA	NA
CRP	.127	.51	NA	NA
DHEAS	.001	.15	NA	NA
Total cholesterol	−.001	.61	NA	NA
HDL-C	.010	.09	NA	NA
Adiponectin	.001	.25	NA	NA
Leptin	.001	.8	NA	NA
LDL-C	−.001	.7	NA	NA
TG	−.001	.27	NA	NA

BMI = body mass index; CRP = C-reactive protein; DHEAS = dehydroepiandrosterone sulfate; FPG = fasting plasma glucose; HDL-C = high-density lipoprotein cholesterol; HOMA-IR = homeostasis model assessment of insulin resistance; IGT = impaired glucose tolerance; LDL-C = low-density lipoprotein cholesterol; NA = not applicable; TG = triglycerides.

DISCUSSION

The current study documents that treatment with metformin can improve peripheral endothelial function, but specifically in the subset of women with PCOS and endothelial dysfunction. This improvement in endothelial function was not mediated through changes in androgens, glucose metabolism, or insulin resistance. In addition, endothelial dysfunction was highly prevalent in our study population—15 of 42 women with PCOS (35.7%) had endothelial dysfunction (Table 2).

To our knowledge, only one previously published study has assessed the effect of metformin on peripheral microvascular endothelial function in PCOS.⁷ That study found that although endothelial function was decreased in women with PCOS compared with healthy controls, 3 months of metformin treatment did not notably change endothelial function in the women with PCOS.⁷ The 2 factors that distinguish our study is that (1) Lowenstein et al⁷ had a lean sample of women with PCOS whereas our patients were obese and (2) they did not have any data about glucose metabolism or insulin resistance. Their study concluded that the shorter duration of treatment—3 months—was not long enough to produce an observable effect and overlooked the additional possibility that their sample was also different because they were not overweight or obese compared with patients in previous studies. In our study, over one-third of the women with PCOS had endothelial dysfunction, and similar to the aforementioned study, we did not see a notable improvement in endothelial function after 3 months of metformin therapy in the whole study population. However, stratified by baseline endothelial function, we found that those with abnormal baseline endothelial dysfunction had notable improvement whereas those with normal baseline endothelial function did not. Our study results support the majority of studies in PCOS that reported improvement of macrovascular flow-mediated dilation following metformin administration.^12,13,25 We know that both flow-mediated dilation and RH-PAT measure endothelial function and are associated with unfavorable cardiovascular outcomes. However, the magnitude of their correlation regarding the prediction of future cardiovascular events is not yet elucidated and they are discussed as separate concepts in the current literature.¹⁴ Therefore, this study establishes that a notable effect of metformin on peripheral microvascular endothelial function can be observed even in obese women with PCOS as early as 3 months after treatment.

The improvement of endothelial function in our study was independent of the effect of metformin on other cardiometabolic factors, including glucose metabolism, insulin resistance, and dyslipidemia. This result is similar to that from a previous study documenting that metformin improved macrovascular endothelial dysfunction in women with PCOS independent of changes in insulin sensitivity.¹² Insulin resistance and glucose intolerance are hallmarks of PCOS. Metformin is an antidiabetic agent with insulin-sensitizing properties that has become increasingly accepted in the treatment of PCOS to improve ovulatory frequency and prevent the development of type 2 diabetes. However, its mechanism of action is not fully understood, and studies reveal a plethora of pleiotropic effects that are not always related.⁹ Metformin might improve endothelial function directly by multiple potential mechanisms. Although metformin is a glucose-lowering agent, it is currently known as an antiaging medication that targets several pathways of aging. It has been found to be effective in increasing the life span in animal models and delays age-related diseases as well. The effect of metformin on aging includes multiple mechanisms such as increasing adenosine monophosphate-activated protein kinase, which results in increased nitric oxide production in the endothelium and subsequently better vascular relaxation. Metformin decreases mammalian target of rapamycin, which can subsequently increase nitric oxide. It decreases inflammation and oxidative stress, which may subsequently result in better endothelial function and vascular dilation.^26,27 It has a protective effect against inflammation and can reduce endothelial cell damage and apoptosis caused by mitochondrial membrane damage.

It might reduce vascular remodeling and fibrosis. Also, by inhibiting the renin-angiotensin system, it can cause vascular relaxation as well as a decrease in systolic blood pressure and skin hyperemia.²⁸ In addition, metformin can increase production of nitric oxide in the endothelial cells and improve its effect on vasodilation.²⁹ Although the current study did not identify a specific mechanism of action or pathway through which metformin improves endothelial function, the aforementioned clinical and laboratory studies provide possible explanations through which metformin exerts a direct protective effect on vascular endothelium.^28,29

In this sample of obese women with PCOS studied for 3 months, metformin had no major effect on body weight, waist circumference, cholesterol and triglyceride concentrations, or markers of chronic subclinical inflammation, including C-reactive protein and the adipose tissue hormones adiponectin and leptin. Similarly, a previous study¹² in overweight women with PCOS did not find notable changes in glucose metabolism and insulin response during an OGTT after 6 months of metformin treatment. It is also possible that in obese women it is more difficult to see the effect with 3 months of treatment. It might be that the criterion standard hyperinsulinemiceuglycemic clamp is needed to formally measure changes in insulin resistance that have been demonstrated previously with metformin. It has also been found that the metabolic abnormalities in PCOS do not necessarily change together. As an instance, differences in serum low-density lipoprotein cholesterol levels in these patients are at least partially independent of BMI.⁹ Interestingly, a recent study has reported that endothelial function in PCOS is not associated with cardiometabolic factors of obesity and insulin resistance.⁸ Lambert et al⁸ found that women with PCOS have lower endothelial function and higher sympathetic activation compared with matched controls with a similar metabolic profile. Both observed findings were associated with PCOS but not cardiometabolic variables such as BMI, glucose level, insulin concentration, or lipid profile.⁸ This finding suggests that endothelial dysfunction is at least partially independent of prevalent cardiometabolic risk factors in PCOS.⁸ Thus, improvement of endothelial function in PCOS following metformin administration can be mediated through mechanisms other than weight reduction, change in plasma glucose and insulin levels, or lipid profile as documented in previous studies.^28,29 Endothelial dysfunction could thus be an independent abnormality or an early factor in the course of the metabolic dysregulation, leading to increased cardiovascular risk in women with PCOS. These data suggest that screening for endothelial dysfunction can be done independently and in addition to the metabolic risk profile in women with PCOS to help decision making regarding possible treatment strategies.

Polycystic ovary syndrome is a heterogeneous disorder with various presentations and ongoing challenges in treatment strategies.^2,9 Metformin can have variable effects in women with PCOS, at least partially because different subtypes of PCOS have different cardiometabolic profiles and should be considered as separate subpopulations for therapeutic approaches.^9,30 This issue highlights the need for individualized treatment, especially in patients with PCOS. We have previously proposed that measuring endothelial function can be a useful tool for individualized medicine.^31,32 Assessing endothelial function provides information about the atherosclerosis itself rather than quantifying risk factors that may have variable effects from person to person. Hence, it can be used as a tool for individualized risk assessment and provides an opportunity to control atherosclerosis at earlier stages.^31,32 In the current study, we found that women with PCOS and endothelial dysfunction can benefit from taking metformin, whereas metformin did not have any notable effect on those with normal baseline endothelial function. Our study suggests a potential beneficial role for the measurement of endothelial function to individualize the treatment approaches in women with PCOS.

In addition to its prognostic value, endothelial function can be targeted for treatment to reduce future cardiovascular events. As an example, a study by Kitta et al³³ found that improvement of endothelial dysfunction can reduce future adverse outcomes in patients with coronary artery disease. Another study reported that improved endothelial function following antihypertensive treatment in postmenopausal women is associated with more favorable cardiovascular outcomes.³⁴ These findings further highlight the value of targeting endothelial function for prevention of coronary heart disease. Although a higher long-term risk of cardiovascular events in PCOS remains to be further clarified,⁹ several studies have established the higher prevalence of cardiometabolic risk factors in PCOS that raises concerns about their future risk of adverse cardiovascular events.^{2–6,9,35–37} Therefore, screening for cardiometabolic risk factors is currently recommended in all women with PCOS.^3,9 Heart disease is still the leading cause of death for women, with the development of type 2 diabetes notably increasing a woman’s risk for heart disease to the equivalent of a prior cardiovascular event. Our study results suggest that screening for endothelial dysfunction might be a promising method to screen for higher-risk individuals who could receive the greatest benefit from metformin or other therapies independent of their glucose tolerance and cardiovascular risk factor profile. Future studies are needed to further validate this strategy.

The strengths of the current study are the use of a control group to assess endothelial function in a randomized manner and the inclusion of detailed dynamic measurements of glucose metabolism and insulin secretion using the 2-hour OGTT. The null correlations of metformin’s effect with changes in these metabolic parameters are important indicators that metformin could have a direct effect on the endothelium and reinforces the dialogue to pursue other mechanistic routes of metformin in future studies. Besides the novel findings, our study has some limitations. Similar to most studies of the effect of metformin in women with PCOS,³⁰ we do not have a long follow-up to establish metformin’s effect on hard cardiovascular end points of morbidity and mortality. Studies with long-term follow-up would be needed to assess whether this improvement in endothelial function can actually reduce the incidence of cardiovascular events in addition to confirming maintenance of endothelial function improvement long term. The final comparison between patients with endothelial dysfunction and normal endothelial function did not have a placebo control group. Because this was a secondary analysis, we could not confirm this result against a placebo control group. A placebo control group would not be expected to improve endothelial function differently by baseline endothelial function, and this could be confirmed with future studies. In addition, whether the endothelial dysfunction is associated with specific PCOS phenotype or genotype that can benefit from metformin treatment remains to be further elucidated.

CONCLUSION

Metformin improves peripheral endothelial function in women with PCOS and endothelial dysfunction independent of changes in glucose metabolism, dyslipidemia, or presence of prediabetes. The study findings suggest that metformin has a direct effect on endothelial function in PCOS and that measurement of peripheral microvascular endothelial function can stratify and follow response to metformin treatment in PCOS.

Abbreviations and Acronyms:

BMI

body mass index

HOMA-IR

homeostasis model assessment of insulin resistance

IGT

impaired glucose tolerance

OGTT

oral glucose tolerance test

PCOS

polycystic ovary syndrome

PAT

peripheral arterial tonometry

RH

reactive hyperemia