Polycystic ovary syndrome (PCOS), with an incidencerate of 6-10% (1), is one of themost commonly encountered endocrinopathy in women of reproductive age, withmetabolic and reproductive consequences, including anovulation, infertility andincreased risk of type 2 diabetes mellitus, hypertension, impaired glucose tolerance (IGT),coronary artery disease, endometrial and perhaps breast cancer(2).
Current evidence indicate an association between PCOSand lifestyle patterns (3,4). It is suggested that there is apossible association between diet and PCOS risk (5), whereas the associationbetween nutrient intakes and development of PCOS has not been thoroughlyexplored (6). Current evidence recommenda low glycemic index diet with low content of saturated fatty acids and high infiber as an optimal diet for PCOS patients (7-12).Although the association between dietary intakes and PCOS risk has been shown previously, only fewstudies have evaluated the effect of isolated nutrients on PCOS characteristics(2,9, 13-15), and the impact of overall nutrientintakes in the etiology of this syndrome has not yet been investigated. Nutrientpattern assessment provides us a snapshot of the interactions among differentdietary nutrients (16-18), which might be different from theeffects of each one intake separately. Therefore, this study was designed to determineif there is a relationship between the dietary nutrient patterns and PCOS risk.SUBJECTS AND METHODSStudy population This case-control study was conducted on the consecutiveincident PCOS cases (n=288) and healthy controls (n=491), aged 20-35 years, admittedto the endocrinology clinics of hospitals from February 2012 to March 2014 inTehran, Iran.
Sample size wascalculated based on a pilot study.Newly diagnosed PCOScases (identified within 3 month of diagnose) were diagnosed by at least two ofthe three symptoms including oligo- or chronic anovulation, clinical and/orbiochemical signs of hyperandrogenism and polycystic ovaries by ultrasonography (19). Exclusion criteriaconsisted: Any history suggestive of other potential causes ofhyperandrogenism/oligo/amenorrhea congenital adrenal hyperplasia, androgensecreting tumor, hypothyroidism, cushing’s syndrome, hyperprolactinemia, otherpituitary/adrenal disorders, other insulin resistance conditions (acromegaly),history of any drug intake and pregnancy.Controls were selected among patients referred to thesame hospitals for a wide spectrum of diseases such as orthopedic problems,ear/nose/throat diseases or elective surgeries, did not have any history of specialdiet, not on any treatment and had regularmenstrual cycles (25-33 day cycles) (20). Neither group sufferedfrom any nutrition-related conditions such as diabetes, cardiovasculardiseases, cancers, osteoporosis or renal disease and was on any medicationwhich might affect hormone metabolism.Socio-demographic, anthropometrics andphysical activityUsing pre-structured pre-tested questionnaires, informedand consenting patients in case and control groups were interviewed face-to-faceby experienced interviewers.
The questionnaire details have been describedpreviously (15). Participants’ weight and height weremeasured, while minimally clothed and standing on digital scales (Soehnle,Berlin, Germany) without shoes. Body mass index (BMI was calculated by dividingweight in kg by the square root of height in meter. Waist circumference (WC)was measured at just above the hip bone, below the rib cage. .Allanthropometric data were collected by the same person to reduce the randomobserver error.
Physical activity was measured through interviews usinga previously validated questionnaire (21,22). Levels of serum follicle stimulating hormone (FSH),luteinizing hormone (LH), prolactin (PRL) and testosterone were determinedusing electrochemiluminescence immuno assays (ECLIA) using Roche Elecsys 1010(Roche Diagnostics, Mannheim, Germany).Dietary assessmentA validated semi-quantitative food frequencyquestionnaire (FFQ) was used to determine the usual dietary intake ofparticipants one year before diagnosis in cases and one year before interviewfor controls (23). The averageintake of energy and nutrients per day was calculated using the USDA foodcomposition table (FCT) and Iranian FCT.Totally, 7 cases and 19 controls were excluded from the analysis becausetheir log scales of total energy intake was either >3 or <3 SD from themean, indicating under/over-reporting. Totally, 281 PCOS cases and 472 controls were included inthe analyses. Participation rates were 97.5% among cases and 96.
3% amongcontrols.All participants signed the informed consent form. Thestudy protocol was approved by our local ethics committee (94/44287).
Statistical analysisInthe present study, “principal components method for factor analysis” of theStatistical Package for the Social Sciences software version 20 (SPSS Inc.,Chicago, IL, USA) was employed to derive potential nutrient patterns on thebasis of 32 nutrients. This statistical procedure aggregates specific nutrientsinto nutrient patterns on the basis of the degree to which nutrients in thedata set are correlated with one another (24).Thecorrelation matrix among the 32 nutrients was visually and statistically evaluatedto justify undertaking factor analysis.
The Chi-square for Bartlett’s test ofsphericity was statistically significant at P<0.001, and theKaiser–Meyer–Olkin measure of sampling adequacy resulted in a score of 0.76,showing that the correlation among the nutrients was adequately strong for afactor analysis. Using the correlation matrix, the orthogonal varimax rotationmethod was applied to achieve a simpler structure facilitating interpretation.The number of meaningful components to retain from the total number ofextracted patterns depended mainly on the assessment of scree plots andcomponents' interpretability. Factor score for each pattern was then calculatedfor each participant by summing the intake frequency of nutrients with weightsthat were proportional to their component (factor) loadings. Factor loadingsare correlation coefficients between nutrient and nutrient patterns; a positiveloading in a factor indicates a direct association with the factor, whereas anegative loading indicates that the food is inversely associated with the factor.Scores for two nutrient patterns identified in this study (factor1 and 2) werethen divided into three categories based on the tertile.
Chi-squaretest was used to evaluate the differences in distribution of categoricalvariables, and independent t-test was used to check the differences indistribution of continuous variables. Unconditional logistic regression wasused to estimate odds ratio (OR) with 95% confidence interval (CI).Models weremutually adjusted for age (years), BMI (kg/m2), physical activity (METhours/day), WC (cm), age atmenarche(years), familial historyof PCOS (yes/no) and total energy intake (kcal/day) as potential confounders. Ap-value < 0.05 was used as the statistical evaluation tool.RESULTSThe distribution of general characteristics of studyparticipants among cases (n = 281) and controls (n = 472) is shown in Table 1.
Factor-loadingmatrix for the two retained factors is shown in Table 2. These factorsexplained 58.3% of the total variance in the original dataset. Factor 1explained29.6% of the total variance and had high loadings for vitamin B1, B2,B3, B5, B6, B9, B12, C,D, E, K, magnesium, total fiber, selenium, phosphorus, manganese,monounsaturated fatty acids, polyunsaturated fatty acids, potassium andvegetable protein. Factor 2 displayedhigh loadings for sodium, cholesterol, saturated fatty acid, fat, biotin, carbohydrate,iron, fluoride, zinc, copper, calcium and animal protein.Table 3 shows the crude and multivariate adjusted odds ratio of the PCOS bytertiles of scores for each nutrient pattern.
After adjusting for potentialconfounding factors, the highest tertile of the first pattern was associatedwith a lower risk of PCOS (OR: 0.48, 95% CI: 0.21–0.82, P for trend = 0.002).
Conversely, in the fully adjusted model, those in the highest tertile of thesecond pattern had a 2.38 times higher odds of PCOS (95% CI: 1.69–3.21; P fortrend = 0.
012).DISCUSSIONTo our knowledge, the present study, based on a validand detailed FFQ, is the first case-control study to examine the relationshipbetween nutrient pattern and PCOS risk. Our findings suggest that the first pattern (abundant inriboflavin, niacin, pyridoxine, thiamin, magnesium, pantothenic acid, cobalamin,vitamin C, folate, vitamin D, total fiber, selenium, phosphorus, vitamin E,manganese, vitamin K, monounsaturated fatty acids, polyunsaturated fatty acids,potassium and vegetable protein) had a significant negative relationship withPCOS risk among a sample of Iranianwomen. In contrast, the second pattern (abundant in sodium, cholesterol,saturated fatty acid, fat, biotin, carbohydrate, iron, fluoride, zinc, copper,calcium and animal protein) was positively associated with the risk of PCOS.Limited studies have evaluated the relationshipbetween dietary factors and PCOS risk, and none of them assessed the dietary nutrients,as a pattern. Douglas et al (5) have found that womenwith PCOS consume a greater amount of specific foods with a high glycemic indexin comparison to healthy controls. Ahmadi et al reported that women with PCOS had a diet with higher totalenergy and fat, saturated fat and poly-unsaturated fat(25). The beneficial effects of omega-3 fattyacids (13) (26).
, combination of high-protein and low-glycemic-loadfoods in a modified diet (14), vitamin D without (27), and with calcium (28) on PCOS comorbidities have been shownpreviously. However the results of somestudies evaluating the effects of nutritional factors on PCOS are inconclusive (27). Altieriet al (29)did not find major differences between PCOSand normoandrogenic control women in regards of energy and macronutrient exceptfor a lower percentage of energy from lipids and a higher intake of fibers byPCOS women in comparison to control healthy ones. None of previous studies on the role ofnutrients in PCOS risk have focused on the impact of dietary nutrients patterns,whereas when an association with overall intake of a nutrient is observed,investigators may identify significant cumulative effects that may be too smallto detect with individual nutrients(30).The nutrient pattern approach iscomplementary to analyses using individual nutrients, which are limited bybiologic interactions and colinearity among nutrients.
The logic behind thenutrient pattern approach is that nutrients are not eaten separately but areeaten in the form of specified dietary patterns. In this study, the components of the firstpattern could be found in a diet high in vegetarian foods, while the secondpattern’s components are usually found in an animal based diet, so it seemsthat the interaction of nutrients in a diet high in vegetables can beprotective for PCOS, while this interaction in an animal based diet proneswomen to PCOS, which might be explained by the high volume and low energydensity of vegetable foods.This study has several strengths; being thefirst study to evaluate the relationship between dietary nutrient patterns andrisk of PCOS in a developing countrywith enough dietary variations due to various economic and cultural situations (31); high participation rate, and inclusion ofnewly diagnosed patients made this study more reliable. Selection of controlsfrom the same socioeconomic population, and using the factor analysis are otheradvantages of this study (24).There were somelimitations in this study including measurement errors, and selection bias. Furthermore,we did not match cases and controls for BMI to avoid overmatching problem whichcould lead to loss of efficiency, since the matching effect could narrow theexposure range.CONCLUSIONIn conclusion, thefirst nutrient pattern comprising mainly vegetable dietary sources showed aninverse association and the second nutrient pattern comprising mainly animaldietary sources showed a direct association with PCOS among Iranian women.
Itseems that substitution of animal sources of proteins with vegetarian sourcescan be beneficial in management of PCOS.Due to case-control nature of the study, we can notdemonstrate causation relationship. Thus, the results should be confirmed infuture prospective studies.
Theauthors’ responsibilities were as follows G.E. conceptualizedthe study, collected the data, analyzed the data and wrote the manuscript. A.H.
conceptualized the study and analyzed the data and wrote the manuscript. Noneof the authors had any personal or financial conflicts of interest.Funding/SupportThis project is financially supportedby the Students’ Research Committee of Shahid Beheshti University of MedicalScience, Tehran, Iran.