A total of 48 women with obesity and affected by PCOS from the infertility clinic of Al-Zahra hospital and Sheykholrais Polyclinic in Tabriz, Iran were recruited. The patients’ diagnoses were made by a gynecologist based on the Rotterdam criteria [2]. Inclusion criteria were: women aged 20–45 years, with BMI ranged 30–40 kg/m2, and taking oral contraceptive pills. Exclusion criteria were: women experiencing menopausal transition, pregnancy (or having tendency to become pregnant), or lactation, women with an endocrine disorder interfering with the function of the reproductive system (e.g., diabetes, thyroid, liver, adrenocortical dysfunction, hyperprolactinemia, Cushing’s syndrome, or adrenal hyperplasia), those who were smokers or passive smokers, women using ovulation induction agents or taking blood pressure regulating drugs, and those receiving fertility therapy or willing to receive it. In addition, women taking insulin-sensitizing medications, insulin infusion drugs, vitamin and mineral supplements, antioxidants, or any supplements affecting their body weight, as well as those following certain diets during two months prior to study were excluded from the present study. All participants in the present study, as mentioned earlier, were on oral contraceptive pills prescribed by a gynecologist for their medical treatment, and they were asked to keep their regular medication during the intervention period of the present study. The sample size was calculated using the following formula: \(\text=\frac}_}\right)}+ }_\right)}\right)}^(}_^+}_^)}}_-}_)}^}\) where N, number of individuals, Z, the value from the table of probabilities of the standard normal distribution for the desired confidence level; β, power of the test; α, statistical significance level; S, standard deviation and µ, variable mean.

This formula was used by taking into account the results of mean ± SD for MDA and the previous result of oxidative stress in PCOS [24] as well as by assuming an α error of 0.05 and the power of 80%. Thus, the total sample size was determined to be 44, which was increased to 48 by predicting a 10% dropout rate.

Thylakoid membranes were extracted from the fresh baby spinach (i.e., Spinacia oleracea) leaves. The leaves were then homogenized, filtered, and diluted with distilled water for ten times. The pH was adjusted to 4.7, which allowed the maximum precipitation. The green precipitate was collected, purified, washed using repeated centrifugation technique, and dried. The green thylakoid powder was manufactured by Darook Pharmaceutical Co., Tehran, Iran. Placebo sachets containing corn starch were colored with edible green color and flavored with kiwi fruit flavor (similar to the flavor added to thylakoid powder) so that the two powders could not be distinguishable in terms of appearance and taste. The packages containing the sachets were coded and distributed monthly by a third party who had no other role in performing the study. The participants’ adherence to taking the sachets on a monthly basis was assessed by counting the sachets returned by them. Consuming more than 80% of the sachets was considered as an acceptable adherence effort.

The current study was a double-blind, randomized, placebo-controlled, and clinical trial. The present randomized controlled trial was conducted in accordance with the Consolidated Standards of Reporting Trials (CONSORT) 2010 statement (Supplementary Material) [25]. The study protocol was registered on the Iranian Registry of Clinical Trials system under code number IRCT20140907019082N9, and was also approved by the ethics committee of the research vice-chancellor of Tabriz University of Medical Sciences, Tabriz, Iran with a code of ethics (IR.TBZMED.REC.1401.088). A written, informed consent was obtained from all study participants. Eligible women were randomly assigned to two treatment conditions in a ratio of 1:1 with varying block sizes stratified by age and BMI adopting the block randomization method, and sequential allocation was performed using the random allocation software (RAS). Sequentially numbered opaque sealed envelopes are used to perform allocation concealment. The envelopes are opened sequentially in the order of entrance at the baseline of the study and the grouping of participants is decided by the envelope carrying the allocation plan inside it. Until the final analysis, the researchers and patients were kept concealed regarding the randomization and distribution. For both control and thylakoid groups, a hypocaloric diet with 500 calories less than the total energy expenditure (TEE) was designed by a nutrition consultant taking into account the individual characteristics of the participants and using the Mifflin formula. All participants were provided with thylakoid-rich spinach extract or raw corn starch as placebo in sachets, and were instructed to dissolve one sachet in a glass of water and consume it on a daily basis 30 min before the lunch for 12 weeks while following the prescribed low-calorie diet.

The primary study outcome was to measure and compare the value LPS, neurotrophic factors such as BDNF and S100B levels and also oxidative stress markers like MDA, TAC and CAT among PCOS individuals at the baseline and end of the intervention period of thylakoid supplementation along with low-calorie diet in two groups of intervention and placebo. Secondary outcome measures were evaluation and comparison of glycemic markers, including FBG, insulin and HOMA-IR and hormonal profiles, including the FSH/LH ratio, FTI and DHEA-S in two groups of the study at the baseline and after 12 weeks of intervention.

The height, weight, waist and hip circumference and body composition of the participants were measured at the beginning, sixth week after the study, and at the end of the study by a trained person to lessen the individual error. BMI was calculated as the body weight (kg) divided by the square of the height (m). Body composition was measured performing bioelectrical impedance analysis and using Tanita MC-780 SMA (Tanita Corporation, Tokyo, Japan). The International Physical Activity Questionnaire-Short Form (IPAQ-SF) was used to measure the level of physical activity at baseline, 6th week, and end of the intervention. This validated questionnaire records self-reported activity at four levels of intensity, namely high-intensity activity (e.g., aerobics), moderate-intensity activity (e.g., leisure cycling), walking, and sitting [26].

At the baseline and end of the study, a 10 mL venous blood sample was taken from all participants after 12 h of fasting. Blood sampling was performed between 8:00 AM and 9:00 AM during the early follicular phase (days 2 to 5) of a spontaneous menstrual cycle or P-induced menstrual cycle. Blood samples were immediately centrifuged at a speed of 3500 rpm for ten minutes, and serum samples were separated from whole blood and directly frozen at -80 °C until examination time. Oxidant and antioxidant parameters, including CAT (Teb Pazhouhan Razi Co., Catalog Number: TPR-CAT-96T, Tehran, Iran. Intra and inter-assay coefficient of variation (CV) 4.1% and 9.9% respectively), TAC (Teb Pazhouhan Razi Co., Catalog Number: TPR-TAC-96T, Tehran, Iran. Intra and inter-assay CVs 5.7% and 3.7% respectively), and MDA (Teb Pazhouhan Razi Co., Catalog Number: TPR-MDA-96T, Tehran, Iran. Intra and inter-assay CVs 6.7% and 7.2% respectively), were measured using enzyme-linked immunosorbent assay (ELISA) kit and based on the manufacturer’s instructions. The serum fasting blood sugar (FBS) was measured adopting enzymatic methods and colorimetric technique and using commercial kits (Pars Azmoon, Catalog Number:1,500,017, Karaj, Iran. Intra- and inter-assay CVs were 1.63 and 2.2 respectively) with an auto-analyzer (Hitachi-917, Tokyo, Japan). Furthermore, serum insulin level (Monobind, Inc.,Catalog Number: 2,425,300, Lake Forest, CA, USA. Intra and inter-assay CVs < 5.6% both) was determined by performing chemiluminescence (IMMULITE 2000, SIEMENS), and the homeostatic model of assessment for insulin resistance (HOMA-IR) was calculated using the suggested formula [27]. Hormonal profiles, including the follicle stimulating hormone, luteinizing hormone, total testosterone, dehydroepiandrosterone sulfate, and sex hormone binding globulin, were measured using ELISA kits (Bioassay Technology Laboratory, Shanghai Korean Biotech, Shanghai City, China. Catalog Numbers: E1001Hu, E1037Hu, E1036H, E1057Hu, E1011Hu respectively, Intra and inter-assay CVs < 5.0% for these measurements). Free testosterone index (FTI) was calculated using the formula: FTI = 100 x serum testosterone (nmol/L)/sex hormone binding globulin (SHBG, nmol/L). The serum BDNF (Padginteb Co., Catalog Number: CSB-E04501h, Iran. Intra-assay CV < 8% and CV < 10%), S100B (Biovendor, Catalog Number: RD192090100R, Heidelberg, Germany. Intra- and inter-assay CVs were 2.0 and 5.9 respectively.) and LPS (Bioassay Technology Laboratory, Shanghai Korean Biotech. Catalog Number: E1791H, Shanghai City, China. The intra and inter-assay CVs were 5.6% and 6.9% respectively) levels were determined by adopting sandwich ELISA method and following the manufacturer’s recommendation for the commercial kit.

Data in the text and tables were presented as means ± standard deviation and were analyzed using SPSS version 23 (SPSS Inc., Chicago, IL). After testing the normal distribution using the Kolmogorov–Smirnov test, independent sample t-tests or Mann–Whitney or chi squared tests were performed to determine the between-group differences at baseline. The analysis of covariance (ANCOVA) was performed to compare the two groups after 12 weeks of intervention after adjusting for baseline values of each variable and potential confounders such as age, changes in weight and body mass index during 12 weeks, and physical activity as well as reporting the crude P-value at the end of the trial. Meanwhile, one-way ANCOVA was performed to estimate the effect size by taking the difference in the means of two study groups and dividing by the pooled standard deviation. Independent t-test was used for reporting mean difference (MD) with a 95% confidence interval (CI) for between-group differences. P-values less than 0.05 were considered statistically significant.