{"paper_id":"4c012afc-1bab-4ed3-b7bc-d733bc537f4e","body_text":"Protein intake and the risk of premenstrual syndrome\nSerena C Houghton 1,*, JoAnn E Manson 2,3,4, Brian W Whitcomb 1, Susan E Hankinson 1,2,\nLisa M Troy 5, Carol Bigelow 1 and Elizabeth R Bertone-Johnson 1\n1Department of Biostatistics and Epidemiology, University of Massachusetts, 715 North Pleasant Street, Arnold\nHouse 412, Amherst, MA 01003, USA: 2Channing Division of Network Medicine, Department of Medicine,\nBrigham and Women ’s Hospital and Harvard Medical School, Boston, MA, USA: 3Division of Preventive Medicine,\nDepartment of Medicine, Brigham and Women ’s Hospital and Harvard Medical School, Boston, MA, USA:\n4Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, MA, USA: 5Department of\nNutrition, University of Massachusetts, Amherst, MA, USA\nSubmitted 25 March 2018: Final revision received 12 October 2018: Accepted 10 December 2018: First published online 18 February 2019\nAbstract\nObjective: To examine the relationship between protein intake and the risk of\nincident premenstrual syndrome (PMS).\nDesign: Nested case –control study. FFQ were completed every 4 years during\nfollow-up. Our main analysis assessed protein intake 2 –4 years before PMS\ndiagnosis (for cases) or reference year (for controls). Baseline (1991) protein\nintake was also assessed.\nSetting: Nurses’ Health Study II (NHS2), a large prospective cohort study of\nregistered female nurses in the USA.\nParticipants: Participants were premenopausal women between the ages of 27\nand 44 years (mean: 34 years), without diagnosis of PMS at baseline, without a\nhistory of cancer, endometriosis, infertility, irregular menstrual cycles or\nhysterectomy. Incident cases of PMS ( n 1234) were identi fied by self-reported\ndiagnosis during 14 years of follow-up and validated by questionnaire. Controls\n(n 2426) were women who did not report a diagnosis of PMS during follow-up\nand con firmed experiencing minimal premenstrual symptoms.\nResults: In logistic regression models adjusting for smoking, BMI, B-vitamins and\nother factors, total protein intake was not associated with PMS development. For\nexample, the OR for women with the highest intake of total protein 2 –4 years\nbefore their reference year (median: 103 ·6 g/d) v. those with the lowest (median:\n66·6 g/d) was 0 ·94 (95 % CI 0 ·70, 1 ·27). Additionally, intakes of speci fic protein\nsources and amino acids were not associated with PMS. Furthermore, results\nsubstituting carbohydrates and fats for protein were also null.\nConclusions: Overall, protein consumption was not associated with risk of\ndeveloping PMS.\nKeywords\nPremenstrual syndrome\nDiet\nProtein\nNurses’ Health Study II\nEpidemiology\nUp to 20 % of reproductive-aged women meet clinical\ndiagnostic criteria for premenstrual syndrome (PMS) (1,2),a\ncyclical disorder characterized by physical and emotional\nsymptoms occurring during the late luteal phase of the\nmenstrual cycle and abating within a few days following\nthe onset of menses. While the aetiology of PMS is still\nlargely unknown, an interaction between hormonal,\nneural, genetic, psychosocial and dietary factors likely\ncontributes\n(3).\nWe hypothesize that protein intake may be related to\nPMS through several potential physiological mechanisms,\nincluding actions of sex steroid hormones and neuro-\ntransmitters, and/or the renin –angiotensin–aldosterone\nsystem\n(4). Protein intake may alter sex hormone levels, as\n17β-oestradiol and progesterone levels are found to\ndecrease with increasing soya protein intake (5). Higher\nanimal protein intake has been associated with higher total\nand free oestradiol levels and lower sex hormone-binding\nglobulin level, potentially due to the increase in exogen-\nous hormones\n(6). Additionally, high protein intake and\nintake of speci fic amino acids may plausibly lower PMS\nrisk, as tryptophan, glutamate and other amino acids are\nprecursors to neurotransmitters implicated in PMS aetiol-\nogy\n(7,8). Lastly, protein intake is reported to increase levels\nof renin, aldosterone and vasopressin (9), vasoactive hor-\nmones of the renin –angiotensin–aldosterone system,\nPublic Health Nutrition: 22(10), 1762 –1769 doi:10.1017/S1368980018004019\n*Corresponding author: Email shoughto@schoolph.umass.edu © The Authors 2019\nhttps://doi.org/10.1017/S1368980018004019 Published online by Cambridge University Press\n\ndysfunction of which has been suggested to contribute to\nPMS(10,11).\nWomen with PMS consumed higher intakes of protein\nin the premenstrual phase (luteal) compared with the\npostmenstrual phase (follicular), with no change in intake\namong controls, in one study examining energy intake\nover the menstrual cycle\n(12). The small number of retro-\nspective studies of the relationship between premenstrual\nsymptoms and consumption of protein have reported\ninconsistent findings\n(13–15). Additionally, due to the retro-\nspective study design, it is uncertain whether increased\nprotein or amino acid intake precedes the development of\nPMS or whether intake is affected by symptom occurrence.\nTo our knowledge, no previous study has prospectively\nevaluated whether protein intake is associated with risk of\ndeveloping PMS.\nTherefore, we evaluated the relationship between pro-\ntein intake and the development of PMS in the Nurses ’\nHealth Study II (NHS2) PMS Sub-Study, a case –control\nstudy nested within the prospective NHS2.\nMethods\nStudy population\nThe NHS2 is an ongoing prospective cohort study that has\nfollowed 116 429 US female nurses, aged 25 –42 years in\n1989, since the first mailed questionnaire. Information on\nhealth-related behaviours and medical history has been\nupdated biennially and diet quadrennially for over 25\nyears\n(16). Response rates have been at least 89 % for all\nquestionnaire cycles.\nClassification of premenstrual syndrome cases and\ncontrols\nThe NHS2 PMS Sub-Study, described previously\n(16,17),\nincludes a subset of premenopausal women who did not\nreport that they had or ever had PMS on the 1989 or 1991\nquestionnaire. Over 14 years of follow-up (1993 –2007\nquestionnaires), 4108 partici pants reported new clinician-\nmade diagnoses of PMS. For these women we assigned\ndiagnosis year as their reference year. Women who had\nnever reported a diagnosis of PMS by a clinician were\nrandomly assigned a reference year between 1991 and\n2005, of whom 3248 were frequency-matched to cases\nb a s e do na g ea n dr e f e r e nce year. Among both groups,\nwomen with a history of cancer other than non-\nmelanoma skin cancer, endometriosis, extremely irre-\ngular menstrual cycles, infertility and hysterectomy prior\nto their reference year were excluded to limit the possi-\nbility that PMS-like symptoms were due to another con-\ndition. Additionally, because of our interest in diet, those\nwith implausible energy intakes (i.e. those below 2092 kJ\n(500 kcal) and above 14 644 kJ (3500 kcal)) were also\nexcluded. Potential cases and controls were then mailed\nam o d ified version of the Calendar of Premenstrual\nExperiences (COPE) questionnaire\n(17,18) assessing\noccurrence, timing and impact on several domains of\ndaily functioning of twenty-si x premenstrual symptoms in\nthe speci fied 2-year period before their individual refer-\nence year, to con firm case and control status\n(16).T h e\nresponse rates were 86 % for potential cases and 79 % for\npotential controls.\nPMS cases included women who met case criteria for\nPMS de fined by Mortola et al .\n(18). Speci fically, case cri-\nteria included: (i) ≥1 physical and ≥1a f f e c t i v em e n s t r u a l\nsymptoms; (ii) overall symptom severity of ‘moderate’ or\n‘severe’ OR ‘moderate’ or ‘severe’ effect of symptoms on\nat least one life activity or relationship domain; (iii)\nsymptoms begin ≤14 d prior to start of menses; (iv)\nsymptoms end ≤4 d after start of menses; and (v) symp-\ntoms not present in the week after menses ended (16).\nAmong self-reported cases who responded, 14 % did not\nmeet the first criterion, 52 % for the second criterion, 6 %\nfor the third criterion, 12 % for the fourth criterion and\n17 % did not meet the fifth criterion (percentages not\nmutually exclusive). Controls included women who had\nno or minimal symptoms that did not impact daily func-\ntion domains. Control criteri a included: (i) no PMS diag-\nnosis; (ii) either no menstrual symptoms OR an overall\nsymptom severity of ‘minimal’ or ‘mild’; and (iii) either\n‘no effect ’ or ‘mild’ effect of symptoms on the life activity\nand relationship domains. Among those who had not\nreported a PMS diagnosis and responded, 6 % did not\nmeet the first criterion, 12 % for the second criterion and\n11 % did not meet the third criterion. To minimize the\nlikelihood for misclassi fication of the outcome, women\nwho did not meet either case or control criteria ( n 2946)\nwere excluded from further analysis. This resulted in\n1257 validated PMS cases and 2463 validated controls\nthat met criteria.\nAssessment of protein intake and other factors\nIntakes of protein-containing foods were assessed via a\nsemi-quantitative 131-item FFQ beginning in 1991 and\nsubsequently every 4 years thereafter. We assessed the\nintake of total protein, sources of protein (i.e. animal,\nvegetable, dairy), the ratio of animal to vegetable protein\nand the intake of speci fic amino acids (i.e. tryptophan,\ntyrosine, glutamate). To calculate each woman ’s total\nintake of protein and amino acids, the portion size of a\nsingle serving of each food or supplement was multiplied\nby the reported intake frequency. The total amount of\neach food consumed was then multiplied by the protein or\namino acid nutrient content of the food item, and con-\ntributions from all food items were summed. Protein intake\nwas then adjusted for total energy intake using the residual\nmethod\n(19).\nThe validity of similar FFQ for measuring total protein\nintake has been demonstrated previously (19).I na n\nProtein and PMS 1763\nhttps://doi.org/10.1017/S1368980018004019 Published online by Cambridge University Press\n\nanalysis of ninety-two women, the energy-adjusted cor-\nrelation between intake reported by the FFQ and the mean\nof intake measured with two 1-week diet records was 0 ·42\nfor total protein intake\n(19).\nFor each participant, we evaluated protein intake at\nboth baseline (1991) and 2 –4 years before her individual\nreference year (the most recent, but still prospective\nFFQ), to assess longer-term and recent protein intake,\nrespectively. For analyses, dietary information was\navailable for 3660 NHS2 PMS Sub-Study participants at\nbaseline (cases, n 1234; controls, n 2426) and 3638\nwomen 2 –4 years prior to their reference year (cases, n\n1222; controls, n 2416).\nInformation on other factors potentially associated with\nPMS and diet were collected on the biennial ques-\ntionnaires, including age, smoking status, weight, preg-\nnancy history and oral contraceptive use. Height and\nmenstrual cycle characteristics were assessed on the 1989\nquestionnaire. History of depression and antidepressant\nuse were assessed on the menstrual cycle questionnaire.\nChildhood trauma was assessed in 2001 on a separate\nquestionnaire\n(20). Lastly, macronutrients and micro-\nnutrients including vitamin D, B-vitamins, Ca and other\nminerals were assessed by FFQ.\nStatistical analysis\nAge-adjusted means and\nSD for continuous variables and\nfrequencies for categorical variables were calculated using\ngeneralized linear modelling to compare distributions of\ndemographic, behavioural and lifestyle characteristics\nbetween cases and controls.\nWe used unconditional logistic regression to estimate\nOR and 95 % CI of PMS for women across quintiles of\nprotein and amino acid intake. Covariates were selected as\neither being important a priori or producing a 10 % change\nin estimates. Multivariable logistic regression was con-\nducted to assess the relationship between protein intake\nand PMS risk, controlling for age, reference year, age at\nmenarche, BMI (kg/m\n2; weight/height 2), physical activity,\never use of oral contraceptives, parity (pregnancies lasting\n≥6 months), smoking status and quantity (pack-years),\never use of antidepressants, signi ficant childhood trauma,\nvitamin D from dietary sources and total intake of vitamin\nB\n6, thiamin, Fe and Zn.\nAdditionally, we mutually adjusted vegetable, animal\nand dairy protein for one another to control for potential\nconfounding by variation in protein source, where\npotential associations could be due to increases or\ndecreases in the other protein sources. For example,\nvegetable protein was adjusted for intake of dairy and\nanimal protein. Linear trend across quintiles was\nassessed using the Mantel extension test for trend,\nwhere the median value of each protein category was\nentered into the regression model as a continuous\nvariable.\nWe further assessed whether a relationship between\nprotein and amino acid intake and PMS varied by age at the\nreference year (<40 v. ≥40 years) and smoking status (past/\nnever v. current) via strati fied analyses, as the aetiology of\nPMS may vary between younger and older premenopausal\nwomen, and between smokers and non-smokers. The\nmultiplicative interaction terms were evaluated using like-\nlihood ratio tests, where the interaction terms were calcu-\nlated as the products of a binary strati fication factor and\nindicators of macronutrient quintile.\nTo assess the possibility that associations between\nhigher protein intake and risk of PMS could be due to\nlower intake of fats or carbohydrates, we conducted sub-\nstitution analyses. For example, we compared associations\nwhen protein was substituted for fat by including terms in\nthe model for percentage of energy from protein, per-\ncentage of energy from carbohydrates, percentage of\nenergy from alcohol and total energy, excluding percen-\ntage of energy from fat. Additional substitution models\nwere also conducted looking at substitutions for carbo-\nhydrates and fats.\nAnalyses were conducted using the statistical software\npackage SAS version 9 ·3. Two-sided P values <0·05 were\nconsidered statistically signi ficant.\nResults\nCharacteristics of cases and controls 2 –4 years prior to the\nreference year are shown in Table 1. Compared with\ncontrols, cases were younger and had a higher mean BMI\nboth at 2 –4 years prior to the reference year and age 18\nyears. Cases were more likely to have used oral contra-\nceptives, smoked, have been diagnosed with depression,\nused antidepressants and have had signi ficant childhood\ntrauma. Additionally, cases had lower intakes of vitamin D\nfrom food sources and higher intakes of B-vitamins at 2 –4\nyears prior to the reference year.\nTotal protein intake 2 –4 years prior to the reference year\nwas not associated with development of PMS (Table 2).\nOverall, sources of protein were not associated with the\ndevelopment of PMS. While higher intakes of dairy protein\nwere associated with lower risk of PMS in the age-adjusted\nmodel, the results were no longer signi ficant after adjust-\nments for vitamin D, B-vitamins and other covariates.\nHigher vegetable protein intake was non-signi ficantly\nassociated with increased risk of PMS in multivariable-\nadjusted models ( P\ntrend = 0·08; OR quintile 5 v. quintile\n2 = 1·26; 95 % CI 0 ·97, 1 ·65). Results for vegetable, animal\nand dairy protein intake were similar for mutually adjusted\nmodels. Lastly, intakes of tryptophan, tyrosine and gluta-\nmate were not associated with the development of PMS\n(Table 3).\nAnalyses evaluating protein and amino acid intake at\nbaseline in 1991 were similar to results presented for the\nreference year (results not shown). For example, the OR\n1764 SC Houghton et al.\nhttps://doi.org/10.1017/S1368980018004019 Published online by Cambridge University Press\n\nfor total protein comparing the highest quintile with the\nlowest quintile was 0 ·98 (95 % CI 0 ·72, 1 ·34). As BMI may\npotentially lie within the causal path between protein\nintake and PMS, the analyses were repeated without BMI\nand estimates were unchanged. Analyses strati fied by\nsmoking status did not suggest effect measure modi fica-\ntion and there were no signi ficant interactions found.\nHowever, the association between protein and risk of PMS\ndid differ by age at the reference year (Table 4). For total\nprotein, women younger than 40 years at the reference\nyear had non-signi ficant lower risk of PMS development\nwith increasing protein intake. Additionally, interactions\nwere signi ficant for animal protein and vegetable protein\nsources (both P\ninteraction < 0·01). Among women who were\nyounger than 40 years at the reference year, the OR for\nanimal protein and vegetable protein with PMS comparing\nthe highest quintile of intake with the lowest quintile were\n0·58 (95 % CI 0 ·35, 0 ·96) and 1 ·70 (95 % CI 1 ·10, 2 ·62),\nrespectively.\nTable 5 presents the results of substitution models,\nwhere we assessed the effect of substituting equivalent\nenergy from different macronutrients for others. This\nlooks at the effect of the compensatory changes in other\nmacronutrients while holding total energy intake con-\nstant. In age-adjusted models , substitution of protein or\nfat for carbohydrate energy appeared to increase the\nrisk of developing PMS. Substitution of protein for car-\nbohydrate energy was associated with a 13 % increase\nin PMS risk (95 % CI 1 ·01, 1 ·26). However, after\nadjustment for micronutrient intake and other covari-\nates, substitution of protein for carbohydrate\nenergy was not associated with PMS (MV2: OR = 1·00;\n95 % CI 0 ·85, 1 ·17). Similarly, substitution of fat for\ncarbohydrate energy was not associated with PMS after\nadjusting for micronutrients and other covariates (MV2:\nOR = 1·00; 95 % CI 0 ·92, 1 ·07). Additional substitutions\nfor fat or carbohydrates were not associated with PMS\nrisk.\nTable 1 Age-standardized characteristics of premenstrual syndrome (PMS) cases and controls at 2 –4 years\nprior to the reference year ( n 3638); Nurses ’ Health Study II PMS Sub-Study, 1991 –2005\nCases ( n 1222) Controls ( n 2416)\nCharacteristic* Mean SD Mean SD P value†\nAge (years) 37 ·24 ·33 8 ·64 ·4 < 0·001\nBMI (kg/m 2)\nAt 2 –4 years prior to reference year 25 ·35 ·62 4 ·65 ·2 < 0·001\nAt age 18 years 21 ·43 ·32 1 ·13 ·10 ·02\nAge at menarche (years) 12 ·41 ·41 2 ·51 ·40 ·04\nAge at first birth (years) ‡ 26·44 ·22 6 ·53 ·90 ·74\nNumber of full-term pregnancies ( ≥6 months) 1 ·91 ·21 ·91 ·20 ·40\nPhysical activity (MET/week) 28 ·59 5 ·82 4 ·66 5 ·30 ·06\nPack-years of cigarette smoking 8 ·66 5 ·25 ·05 0 ·40 ·07\nAlcohol intake (g/d) 3 ·36 ·03 ·76 ·80 ·31\nT otal energy intake (kJ/d) 7594 2247 7640 2167 0 ·50\nVitamin D intake from food sources ( µg/d)§ 5 ·92 ·86 ·13 ·10 ·02\nT otal vitamin B6 intake (mg/d)§ 9 ·92 8 ·76 ·81 9 ·1 < 0·001\nT otal vitamin B12 intake (mg/d)§ 11 ·21 8 ·91 0 ·91 9 ·90 ·19\nT otal thiamin intake (mg/d)§ 4 ·41 0 ·63 ·99 ·50 ·04\nT otal riboflavin intake (mg/d)§ 4 ·91 0 ·64 ·39 ·20 ·02\nT otal Fe intake (mg/d)§ 23 ·92 3 ·42 4 ·02 2 ·50 ·45\nT otal Zn intake (mg/d) 16 ·31 0 ·51 6 ·31 2 ·10 ·87\nT otal K intake (mg/d) 2996 542 2991 548 0 ·22\nT otal Ca intake (mg/d)§ 1054 449 1084 447 0 ·26\n%%\nHistory of tubal ligation 20 21 0 ·89\nOral contraceptive use\nEver 86 79 < 0·001\nCurrent 8 6 0 ·47\nDuration > 4 years 61 57 0 ·004\nSmoking status\nCurrent 12 6 < 0·001\nPast 28 18 < 0·001\nPreviously diagnosed with depression 18 8 < 0·001\nPreviously used antidepressant medication 15 7 < 0·001\nHistory of childhood trauma 17 9 < 0·001\nMET , metabolic equivalent of task.\n*All characteristics, except age, are standardized to the age distribution of participants at 2 –4 years prior to the reference year.\n†Calculated using generalized linear model.\n‡Limited to parous women.\n§Energy-adjusted value.\nProtein and PMS 1765\nhttps://doi.org/10.1017/S1368980018004019 Published online by Cambridge University Press\n\nDiscussion\nTo our knowledge, the present study is one of the first to\nevaluate prospectively if protein and amino acid intakes\nare associated with the development of PMS. Overall, we\nfound little evidence that protein intake relates to PMS.\nResults from previous studies of protein intake and\npremenstrual symptoms have been inconsistent. Nagata\net al. evaluated the relationship of total protein intake and\npremenstrual symptoms among Japanese women aged\n19–34 years ( n 189)\n(13). Total protein (mean protein\nintake = 76·9( SD 35·3) g/d) was not correlated with change\nin total menstrual distress scores in the premenstrual\nphase. Barnard et al . conducted a crossover study among\nthirty-three women comparing a low-fat vegetarian diet\nwith a normal diet with B-vitamin supplements and found\nthat the low-fat vegetarian diet decreased the duration of\npremenstrual symptoms\n(14). Intakes of protein and fat\nwere significantly different between the normal diet (mean\nprotein intake = 59·8( SD 17·7) g/d) and low-fat vegetarian\ndiet (mean protein intake = 43·5( SD 11·5) g/d). However, it\nis unclear whether this is due to the vegetarian diet, lower\nprotein intakes, B-vitamin supplements and/or the low-fat\ndiet. Lastly, Steinberg et al . conducted a clinical trial\nassessing supplementation of tryptophan (6 g) in women\nwith premenstrual dysphoric disorder for 17 d, where\nsupplementation with tryptophan ( n 37) was more effec-\ntive than placebo ( n 34) in reducing mood symptom\nseverity among women with premenstrual dysphoric dis-\norder\n(21). Our study found no association with tryptophan\nand risk of developing PMS; however, our mean intake of\ntryptophan was less than 1 g (mean = 0·98 (\nSD 0·17) g/d).\nSubstitution of protein for energy from either fat or car-\nbohydrates was not associated with risk of developing PMS\nafter adjusting for potential confounders. This is consistent\nwith our previous findings that fat\n(22) and carbohydrates(23)\nTable 2 Age-adjusted and multivariate OR and 95 % CI for dietary protein intakes 2 –4 years prior to the reference year\nand risk of premenstrual syndrome (PMS; n 3638); Nurses ’ Health Study II PMS Sub-Study , 1991 –2005\nQuintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 Ptrend\nT otal protein\nRange (g/d) < 73·47 3 ·4–80·68 0 ·7–87·28 7 ·3–95·0 > 95·0\nCases:controls ( n) 236:419 233:556 269:542 272:516 212:383\nOR\nAge-adjusted Ref. 0 ·73 0 ·88 0 ·93 0 ·98 0 ·49\nModel 1* Ref. 0 ·73 0 ·91 0 ·90 0 ·94 0 ·83\n95 % CI†– 0·57, 0 ·94 0 ·70, 1 ·18 0 ·68, 1 ·19 0 ·70, 1 ·27\nAnimal protein\nRange (g/d) < 48·64 8 ·6–56·95 7 ·0–64·06 4 ·1–72·8 > 72·8\nCases:controls ( n) 236:455 245:521 284:549 250:472 207:419\nOR\nAge-adjusted Ref. 0 ·90 0 ·99 1 ·02 0 ·94 0 ·99\nModel 1* Ref. 0 ·86 0 ·96 0 ·94 0 ·82 0 ·33\n95 % CI†– 0·67, 1 ·11 0 ·75, 1 ·24 0 ·71, 1 ·24 0 ·61, 1 ·11\nVegetable protein\nRange (g/d) < 19·41 9 ·4–21·82 1 ·9–24·32 4 ·4–27·5 > 27·5\nCases:controls ( n) 226:415 244:488 243:492 237:512 272:509\nOR\nAge-adjusted Ref. 0 ·94 0 ·92 0 ·87 1 ·01 0 ·95\nModel 1* Ref. 1 ·01 1 ·03 0 ·99 1 ·26 0 ·08\n95 % CI†– 0·79, 1 ·29 0 ·80, 1 ·32 0 ·76, 1 ·28 0 ·97, 1 ·65\nDairy protein\nRange (g/d) < 11·31 1 ·3–15·31 5 ·4–19·71 9 ·8–26·3 > 26·3\nCases:controls ( n) 201:379 249:429 251:495 257:516 264:597\nOR\nAge-adjusted Ref. 1 ·08 0 ·93 0 ·90 0 ·81 0 ·01\nModel 1* Ref. 1 ·18 1 ·10 0 ·98 0 ·91 0 ·25\n95 % CI†– 0·91, 1 ·54 0 ·83, 1 ·45 0 ·73, 1 ·31 0 ·65, 1 ·26\nAnimal:vegetable protein\nRange (g/d) < 1·91 ·9–2·42 ·5–2·93 ·0–3·6 > 3·6\nCases:controls ( n) 239:525 265:481 257:519 255:472 206:419\nOR\nAge-adjusted Ref. 1 ·21 1 ·08 1 ·18 1 ·06 0 ·72\nModel 1* Ref. 1 ·09 0 ·97 1 ·03 0 ·86 0 ·25\n95 % CI†– 0·86, 1 ·38 0 ·76, 1 ·23 0 ·79, 1 ·33 0 ·65, 1 ·14\nRef., reference category; MET , metabolic equivalent of task.\n*Adjusted for age (continuous), reference year (1991 –1992, 1993, 1994 –1996, 1997 –1998, 1999 –2000, 2001 –2002, 2003 –2004), age\nat menarche (continuous), BMI ( ≤19·9, 20·0–22·9, 22·5–24·9, 25·0–27·4, 27·5–29·9, ≥30 kg/m2), physical activity ( <3, 3–8, 9–17, 18–26,\n27–41, ≥42 MET/week), oral contraceptive use (none, 1 –23, 24 –71, 72 –119, ≥120 months), parity (nulliparous, 1 –2, 3 –4, ≥5 preg-\nnancies ≥6 months), smoking status (never, past 1 –14, past 15 –34, past ≥35, current 1 –14, current 15 –34, current ≥35 cigarettes/d),\never use of antidepressants (never, ever), childhood trauma score (5, 6 –10, 11 –15, 16 –20, 21 –25), vitamin D from dietary sources\n(quintile) and quintile of total intake for vitamin B 6, thiamin, Fe and Zn at 2 –4 years prior to the reference year.\n†95 % CI is for multivariable model.\n1766 SC Houghton et al.\nhttps://doi.org/10.1017/S1368980018004019 Published online by Cambridge University Press\n\nwere not associated with PMS risk. This further suggests that\nmacronutrient intake is not associated with PMS risk\nafter controlling for intake of micronutrients (e.g. Ca\n(16),\nB-vitamins(24)) and other factors (e.g. smoking (25), BMI(26))\nthat are potentially correlated with macronutrient intake\nand have been signi ficantly associated with PMS risk.\nDifferences in our results compared with previous study\nfindings could potentially be due to confounding by\nmicronutrients. Nagata et al . did not adjust for micro-\nnutrients such as vitamin D or B-vitamins (13). However,\nwhen we controlled for several micronutrients, we still\nfound no association. The reduction in premenstrual\nsymptom severity for the crossover study by Barnard et al.\nmay have been due to additional differences other than fat\nintake and source of protein, including differences in\nmicronutrient intakes\n(14).\nOne potential reason why the previous studies found\nassociations with tryptophan whereas we found no asso-\nciations is study design. The previous studies were treat-\nment trials for premenstrual symptoms, while our study\nassessed risk of developing PMS. Factors that are asso-\nciated with treatment of existing PMS may not be similarly\nTable 3 Age-adjusted and multivariate OR and 95 % CI for amino acid intakes 2 –4 years prior to the reference year and\nrisk of premenstrual syndrome (PMS; n 3638); Nurses ’ Health Study II PMS Sub-Study , 1991 –2005\nQuintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 Ptrend\nT ryptophan\nRange (g/d) < 0·80 ·8–0·90 ·9–1·01 ·0–1·1 > 1·1\nCases:controls ( n) 229:417 232:523 284:580 270:486 207:410\nOR\nAge-adjusted Ref. 0 ·80 0 ·89 1 ·01 0 ·92 0 ·85\nModel 1* Ref. 0 ·83 1 ·00 1 ·07 0 ·91 0 ·94\n95 % CI†– 0·65, 1 ·07 0 ·77, 1 ·30 0 ·81, 1 ·41 0 ·67, 1 ·24\nT yrosine\nRange (g/d) < 2·62 ·6–2·82 ·9–3·13 ·2–3·4 > 3·4\nCases:controls ( n) 226:407 244:552 272:529 259:491 221:437\nOR\nAge-adjusted Ref. 0 ·79 0 ·92 0 ·95 0 ·91 0 ·92\nModel 1* Ref. 0 ·82 1 ·02 0 ·92 0 ·90 0 ·76\n95 % CI†– 0·64, 1 ·06 0 ·78, 1 ·33 0 ·69, 1 ·23 0 ·66, 1 ·23\nGlutamate\nRange (g/d) < 14·01 4 ·0–15·21 5 ·3–16·41 6 ·5–17·6 > 17·6\nCases:controls ( n) 217:379 237:544 265:525 276:526 227:442\nOR\nAge-adjusted Ref. 0 ·75 0 ·88 0 ·90 0 ·89 0 ·86\nModel 1* Ref. 0 ·84 0 ·98 1 ·00 1 ·01 0 ·58\n95 % CI†– 0·65, 1 ·08 0 ·76, 1 ·28 0 ·76, 1 ·32 0 ·75, 1 ·36\nRef., reference category.\n*Adjusted for age, reference year, age at menarche, BMI, physical activity, oral contraceptive use, parity, smoking status, ever use of\nantidepressants, childhood trauma, vitamin D from dietary sources and total intake of vitamin B 6, thiamin, Fe and Zn at 2 –4 years prior to\nthe reference year.\n†95 % CI is for multivariable model.\nTable 4 Multivariate OR and 95 % CI for dietary protein intakes (g/d) 2 –4 years prior to the reference year and risk of premenstrual syndrome\n(PMS; n 3638), stratified by age at the reference year; Nurses ’ Health Study II PMS Sub-Study , 1991 –2005\nQuintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5\nOR 95 % CI OR 95 % CI OR 95 % CI OR 95 % CI OR 95 % CI Pinteraction\nT otal protein\n< 40 years old 1 ·00 Ref. 0 ·71 0 ·48, 1 ·07 0 ·65 0 ·42, 1 ·02 0 ·68 0 ·43, 1 ·06 0 ·80 0 ·48, 1 ·31 < 0·0001\n≥ 40 years old 1 ·00 Ref. 0 ·66 0 ·47, 0 ·92 1 ·05 0 ·75, 1 ·47 1 ·01 0 ·70, 1 ·45 1 ·02 0 ·68, 1 ·51\nAnimal protein\n< 40 years old 1 ·00 Ref. 0 ·74 0 ·49, 1 ·12 0 ·73 0 ·48, 1 ·12 0 ·63 0 ·40, 1 ·00 0 ·58 0 ·35, 0 ·96 < 0·0001\n≥ 40 years old 1 ·00 Ref. 0 ·87 0 ·63, 1 ·20 1 ·06 0 ·76, 1 ·48 1 ·11 0 ·77, 1 ·58 0 ·95 0 ·64, 1 ·40\nVegetable protein\n< 40 years old 1 ·00 Ref. 0 ·84 0 ·56, 1 ·26 1 ·38 0 ·92, 2 ·06 1 ·55 1 ·01, 2 ·38 1 ·70 1 ·10, 2 ·62 < 0·0001\n≥ 40 years old 1 ·00 Ref. 1 ·04 0 ·75, 1 ·45 0 ·81 0 ·57, 1 ·14 0 ·72 0 ·50, 1 ·03 1 ·06 0 ·74, 1 ·52\nDairy protein\n< 40 years old 1 ·00 Ref. 1 ·16 0 ·74, 1 ·82 1 ·10 0 ·68, 1 ·79 1 ·00 0 ·61, 1 ·65 0 ·74 0 ·42, 1 ·30 0 ·0001\n≥40 years old 1 ·00 Ref. 1 ·12 0 ·80, 1 ·57 1 ·02 0 ·72, 1 ·45 0 ·91 0 ·63, 1 ·33 0 ·95 0 ·62, 1 ·46\nRef., reference category.\n*Models are adjusted for age, reference year, age at menarche, BMI, physical activity, oral contraceptive use, parity, smoking status, ever use of an ti-\ndepressants, childhood trauma, vitamin D from dietary sources and total intake of vitamin B 6, thiamin, Fe and Zn at 2 –4 years prior to the reference year.\nProtein and PMS 1767\nhttps://doi.org/10.1017/S1368980018004019 Published online by Cambridge University Press\n\nrelated to risk of developing PMS. Additionally, the sup-\nplementation dose in the treatment trials was substantially\nhigher than the average dietary intake of tryptophan in our\nstudy; potential bene fits of tryptophan are perhaps only\nachievable with higher intakes than observable in our\nstudy or supplementation. Furthermore, in studies of\nprevalent PMS observing associations with protein intake,\nit is unclear whether women may have altered their pro-\ntein intake in response to symptoms of PMS as a method of\nmanaging them, or whether protein or amino acids intake\ncontributes to PMS development.\nIn stratified analyses, among younger women (<40 years\nold), higher intake of protein from animal sources was\ninversely associated with PMS risk, whereas higher intake of\nprotein from vegetable sources was positively associated\nwith PMS risk. These findings suggest that risk factors\nmay differ for PMS diagnoses at younger v. older ages.\nHowever, as these findings were unexpected and the\nmechanism by which this could occur is unclear, future\nstudies are needed.\nSimilar to other epidemiological studies that use FFQ to\nassess diet, protein intakes may be misclassi fied due to\nissues in the accuracy of food composition tables to assign\na mean protein value for each food and women accurately\nreporting diet history. As exposure was assessed before\nthe diagnosis of PMS, this misclassi fication is likely not\ndifferent between women with PMS and women without\nPMS, and estimates would be biased towards the null.\nHowever, misclassi fication is minimized through use of a\nvalidated FFQ, exclusion of those with implausible energy\nintakes, adjustment for total energy, and ranked compar-\nisons of high intake v. low intake using quintiles. Lastly,\nprevious studies within the NHS2 cohort, using the same\nFFQ, have detected associations of meat and protein\nintake with other chronic illnesses\n(27–29). Additionally, as\nthe aetiology of PMS is unknown, it is unclear which\ndietary exposure period would be most relevant to the\ndevelopment of PMS. While we assessed both longer-term\n(baseline) and more recent protein intakes (2 –4 years prior\nto diagnosis), we cannot exclude the possibility of asso-\nciations with intakes even closer to diagnosis ( <2–4 years\nprior) or further from diagnosis (e.g. childhood and ado-\nlescence). However, prior studies in this cohort have\nadditionally detected associations with dietary intakes 2 –4\nyears prior to PMS\n(16,24,30), indicating that it is a potentially\nrelevant aetiological period. Additionally, with the expo-\nsure being assessed prior to diagnosis of PMS, we exclude\nthe potential for recall bias and reverse causation.\nDue to the large prospective cohort study design, pro-\nspective charting was not feasible; however, mis-\nclassification of the outcome is minimized by comparing\nthe two ends of the symptom spectrum and excluding\nthose in the middle who met criteria for neither cases nor\ncontrols. Symptom recall is likely to be accurate for those\nwho regularly experience severe symptoms that impair\ndaily functioning and for those who regularly experience\nfew, if any symptoms, and is unlikely to be misclassi fied\nbetween these two groups\n(16). Second, participants had\nprospectively reported incident PMS diagnoses by a clin-\nician, which were then con firmed by validated retro-\nspective questionnaire. We previously observed that\nwomen meeting our criteria for PMS were comparable to\nthose who also reported prospective charting\n(17).\nConclusion\nIn conclusion, we did not observe evidence that protein or\namino acid intake was associated with PMS risk. Further-\nmore, macronutrient intake overall was not associated\nwith PMS after adjusting for micronutrients. As the present\nstudy is the first to examine protein intake and develop-\nment of PMS, con firmation from additional prospective\nstudies that there does not appear to be an important\nassociation is needed. Additionally, future studies should\nexamine micronutrients as potential risk factors for PMS\ndevelopment.\nAcknowledgements\nAcknowledgements: The authors thank the participants\nand staff of the NHS2 for their valuable contributions.\nTable 5 Age-adjusted and multivariate OR and 95 % CI for macronutrient (5 % of energy) substitution models 2 –4 years prior to the\nreference year and risk of premenstrual syndrome (PMS; n 3638); Nurses ’ Health Study II PMS Sub-Study , 1991 –2005\nAge-adjusted MV1* MV2 †\nSubstitution OR 95 % CI OR 95 % CI OR 95 % CI\nProtein for fat 1 ·10 0 ·97, 1 ·25 1 ·12 0 ·95, 1 ·32 1 ·01 0 ·84, 1 ·23\nProtein for carbohydrate 1 ·13 1 ·01, 1 ·26 1 ·04 0 ·91, 1 ·19 1 ·00 0 ·85, 1 ·17\nFat for carbohydrate 1 ·06 1 ·00, 1 ·13 0 ·98 0 ·92, 1 ·05 1 ·00 0 ·92, 1 ·07\nFat for protein 1 ·05 0 ·95, 1 ·17 1 ·06 0 ·94, 1 ·20 1 ·01 0 ·87, 1 ·18\nCarbohydrate for fat 0 ·97 0 ·92, 1 ·02 1 ·05 0 ·99, 1 ·12 1 ·01 0 ·94, 1 ·09\nCarbohydrate for protein 0 ·99 0 ·91, 1 ·07 1 ·06 0 ·97, 1 ·17 1 ·02 0 ·90, 1 ·15\nMV , multivariable-adjusted.\n*MV1 = age, reference year, age at menarche, BMI, physical activity, oral contraceptive use, parity, smoking status, ever use of antidepressants, childho od\ntrauma, vitamin D from dietary sources and total intake of vitamin B 6, thiamin and Fe.\n†MV2 = MV1 + history of depression and total intake of Ca, vitamin B 12, riboflavin, folate, Zn and K.\n1768 SC Houghton et al.\nhttps://doi.org/10.1017/S1368980018004019 Published online by Cambridge University Press\n\nFinancial support: This work was supported by the\nNational Institutes of Health (grant number\nUM1CA176726), (E.R.B.-J., grant number MH076274); a cy\npres distribution from Rexall/Cellasene settlement litiga-\ntion; and a grant from GlaxoSmithKline Consumer\nHealthcare. The funders had no role in the design, analysis\nor writing of this article. Conflict of interest: None.\nAuthorship: J.E.M., S.E.H. and E.R.B.-J. designed the\nresearch; S.C.H. and E.R.B.-J. conducted the research;\nJ.E.M. and S.E.H. provided essential materials; S.C.H. and\nE.R.B.-J. performed the statistical analysis; S.C.H. and\nE.R.B.-J. wrote the paper; B.W.W., L.M.T. and C.B. inter-\npreted study results, reviewed the manuscript for important\nintellectual content and contributed knowledge of under-\nlying biological mechanisms; S.C.H., J.E.M. and E.R.B.-J.\nhad primary responsibility for the final content. All authors\nread and approved the final manuscript. Ethics of human\nsubject participation:This study was conducted according to\nthe guidelines laid down in the Declaration of Helsinki and\nall procedures involving human subjects were approved by\nthe Institutional Review Board at Brigham and Women ’s\nHospital in Boston, MA; return of mailed questionnaires was\nconsidered to be informed consent.\nReferences\n1. Halbreich U, Borenstein J, Pearlstein T et al . (2003) The\nprevalence, impairment, impact, and burden of pre-\nmenstrual dysphoric disorder (PMS/PMDD). Psychoneuro-\nendocrinology 28, Suppl. 3, 1 –23.\n2. Johnson SR (2006) The epidemiology and social impact of\npremenstrual symptoms. Clin Obstet Gynecol 30, 367 –376.\n3. Matsumoto T, Asakura H & Hayashi T (2013) Biopsycho-\nsocial aspects of premenstrual syndrome and premenstrual\ndysphoric disorder. Gynecol Endocrinol 29,6 7 –73.\n4. Houghton SC & Bertone-Johnson ER (2015) Macronutrients\nand premenstrual syndrome. In Advances in Medicine and\nBiology, vol. 87, pp. 137 –156 [LV Berhardt, editor]. Haup-\npauge, NY: NOVA Science Publishers, Inc.\n5. Liu S, Willett WC, Stampfer MJ et al . (2000) A prospective\nstudy of dietary glycemic load, carbohydrate intake, and risk\nof coronary heart disease in US women. Am J Clin Nutr 71,\n1455–1461.\n6. Brinkman MT, Baglietto L, Krishnan K et al . (2010) Con-\nsumption of animal products, their nutrient components and\npostmenopausal circulating steroid hormone concentra-\ntions. Eur J Clin Nutr 64, 176 –183.\n7. Backstrom T, Andreen L, Birzniece V et al . (2003) The role\nof hormones and hormonal treatments in premenstrual\nsyndrome. CNS Drugs 17, 325 –342.\n8. Association of Reproductive Health Professionals (2008) A\nQuick Reference Guide for Clinicians: Managing Pre-\nmenstrual Symptoms . Washington, DC: ARHP.\n9. Daniels BS & Hostetter TH (1990) Effects of dietary\nprotein intake on vasoactive hormones. Am J Physiol 258,\n1095–100.\n10. Janowsky DS, Berens SC & Davis JM (1973) Correlations\nbetween mood, weight, and electrolytes during the\nmenstrual cycle: a renin –angiotensin–aldosterone\nhypothesis of premenstrual tension. Psychosom Med 35,\n143–154.\n11. Rosenfeld R, Livne D, Nevo O et al . (2008) Hormonal and\nvolume dysregulation in women with premenstrual syn-\ndrome. Hypertension 51, 1225 –1230.\n12. Cross GB, Marley J, Miles H et al. (2001) Changes in nutrient\nintake during the menstrual cycle of overweight women\nwith premenstrual syndrome. Br J Nutr 85\n, 475 –482.\n13. Nagata C, Hirokawa K, Shimizu N et al . (2004) Soy, fat and\nother dietary factors in relation to premenstrual symptoms\nin Japanese women. BJOG 111, 594 –599.\n14. Barnard ND, Scialli AR, Hurlock D et al . (2000) Diet and\nsex-hormone binding globulin, dysmenorrhea, and pre-\nmenstrual symptoms. Obstet Gynecol 95, 245 –250.\n15. Menkes DB, Coates DC & Fawcett JP (1994) Acute trypto-\nphan depletion aggravates premenstrual syndrome. J Affect\nDisord 32,3 7 –44.\n16. Bertone-Johnson ER, Hankinson SE, Bendich A et al. (2005)\nCalcium and vitamin D intake and risk of incident pre-\nmenstrual syndrome. Arch Intern Med 165, 1246 –1252.\n17. Bertone-Johnson ER, Hankinson SE, Johnson SR et al .\n(2007) A simple method of assessing premenstrual syn-\ndrome in large prospective studies. J Reprod Med 52,\n779–786.\n18. Mortola JF, Girton L, Beck L et al . (1990) Diagnosis of pre-\nmenstrual syndrome by a simple, prospective, and reliable\ninstrument: the calendar of premenstrual experiences.\nObstet Gynecol 76, 302 –307.\n19. Willett WC (2012) Nutritional Epidemiology , 3rd ed. New\nYork: Oxford University Press.\n20. Bertone-Johnson ER, Whitcomb BW, Hankinson SE et al .\n(2014) Early life emotional, physical, and sexual abuse and\nthe development of premenstrual syndrome: a\nlongitudinal study. J Womens Health 23, 729 –739.\n21. Steinberg S, Annable L, Young SN et al . (1999) A placebo-\ncontrolled clinical trial of\nL-tryptophan in premenstrual\ndysphoria. Biol Psychiatry 45, 313 –320.\n22. Houghton SC, Manson JE, Whitcomb BW et al . (2017)\nIntake of dietary fat and fat subtypes and risk of pre-\nmenstrual syndrome in the Nurses ’ Health Study II. Br J Nutr\n118, 849 –857.\n23. Houghton SC, Manson JE, Whitcomb BW et al . (2018)\nCarbohydrate and fiber intake and the risk of premenstrual\nsyndrome. Eur J Clin Nutr 72, 861 –870.\n24. Chocano-Bedoya PO, Manson JE, Hankinson SE et al .\n(2011) Dietary B vitamin intake and incident premenstrual\nsyndrome. Am J Clin Nutr 93, 1080 –1086.\n25. Bertone-Johnson ER, Hankinson SE, Johnson SR et al .\n(2008) Cigarette smoking and the development of pre-\nmenstrual syndrome. Am J Epidemiol 168, 938 –945.\n26. Bertone-Johnson ER, Hankinson SE, Willett WC et al. (2010)\nAdiposity and the development of premenstrual syndrome.\nJ Womens Health (Larchmt) 19, 1955 –1962.\n27. Malik VS, Li Y, Tobias DK et al . (2016) Dietary protein\nintake and risk of type 2 diabetes in US men and women.\nAm J Epidemiol 183, 715 –728.\n28. Boutot ME, Purdue-Smithe A, Whitcomb BW et al . (2018)\nDietary protein intake and early menopause in the Nurses ’\nHealth Study II. Am J Epidemiol 187, 270 –277.\n29. Farvid MS, Cho E, Chen WY et al . (2014) Dietary protein\nsources in early adulthood and breast cancer incidence:\nprospective cohort study. BMJ 348, g3437.\n30. Chocano-Bedoya PO, Manson JE, Hankinson SE et al .\n(2013) Intake of selected minerals and risk of premenstrual\nsyndrome. Am J Epidemiol 177, 1118 –1127.\nProtein and PMS 1769\nhttps://doi.org/10.1017/S1368980018004019 Published online by Cambridge University Press","source_license":"public-domain-us","license_restricted":false}