Results Over 18.1 years of follow-up, 2820 cases of ischemic heart disease and 1072 cases of entire throw ( 519 ischemic stroke and 300 hemorrhagic stroke ) were recorded. After adjusting for sociodemographic and life style confounders, fish eaters and vegetarians had 13 % ( guess proportion 0.87, 95 % confidence interval 0.77 to 0.99 ) and 22 % ( 0.78, 0.70 to 0.87 ) lower rates of ischemic heart disease than kernel eaters, respectively ( P < 0.001 for heterogeneity ). This deviation was equivalent to 10 fewer cases of ischemic heart disease ( 95 % assurance interval 6.7 to 13.1 fewer ) in vegetarians than in meat eaters per 1000 population over 10 years. The associations for ischemic heart disease were partially attenuated after alteration for self reported high blood cholesterol, high rake pressure, diabetes, and body mass index ( gamble proportion 0.90, 95 % confidence interval 0.81 to 1.00 in vegetarians with all adjustments ). By contrast, vegetarians had 20 % higher rates of entire stroke ( hazard proportion 1.20, 95 % assurance interval 1.02 to 1.40 ) than kernel eaters, equivalent to three more cases of full throw ( 95 % confidence interval 0.8 to 5.4 more ) per 1000 population over 10 years, by and large due to a higher rate of hemorrhagic stroke. The associations for throw did not attenuate after far allowance of disease gamble factors. Participants 48 188 participants with no history of ischemic affection disease, solidus, or angina ( or cardiovascular disease ) were classified into three distinct diet groups : meat eaters ( participants who consumed meat, careless of whether they consumed fish, dairy, or eggs ; n=24 428 ), pisces eaters ( consume pisces but no kernel ; n=7506 ), and vegetarians including vegans ( n=16 254 ), based on dietary information collected at baseline, and subsequently around 2010 ( n=28 364 ). For stroke, two previous reports, 5 6 including one that included EPIC-Oxford data, 6 found no significant differences in risk of total stroke deaths between vegetarians and non-vegetarians. however, no previous studies have examined the incidence of stroke in sexual intercourse to vegetarian diets, or have examined the main stroke types. For ischemic heart disease, some but not all former studies reported importantly lower risks of mortality from ischemic heart disease in vegetarians than in non-vegetarians. 5 6 7 In terms of incidence, the only former study ( the European Prospective Investigation into Cancer ( EPIC ) -Oxford ) reported that vegetarians had a lower risk of ischemic center disease than non-vegetarians, 8 but at the clock of publication the sketch had an insufficient duration of follow-up to individually examine the risks in other diet groups ( fish eaters and vegans ).

vegetarian and vegan diets have become increasingly popular in recent years, partially due to perceived health benefits, ampere well as concerns about the environment and animal benefit. 1 In the United Kingdom, both the representative National Diet and Nutrition Survey 2008-12 and a 2016 Ipsos MORI survey estimated about 1.7 million vegetarians and vegans living in the country. 2 3 testify suggests that vegetarians might have different disease risks compared with non-vegetarians, 4 but data from large scale prospective studies are limited, because few studies have recruited sufficient numbers of vegetarian participants .

Methods

Study population and design

EPIC-Oxford is a prospective cohort cogitation of about 65 000 men and women who were recruited across the UK between 1993 and 2001. Details of the recruitment process have been described previously. 9 Individuals were recruited from either general practices or by postal questionnaire. The general drill recruitment method acting recruited 7421 men and women aged 35 to 59 who were registered with participating general practices, all of whom completed a full moon questionnaire on their diet, life style, health characteristics, and checkup history. The postal recruitment preferentially target vegetarians, vegans, and other people interested in diet and health, and recruited 57 990 participants aged 20 or older. A full questionnaire was mailed to all members of the vegetarian Society and all surviving participants of the Oxford Vegetarian Study, 10 and respondents were invited to provide names and addresses of relatives and friends who were besides matter to in receiving a questionnaire. A short questionnaire was besides distributed to all members of the Vegan Society, enclosed in vegetarian and health food magazines, and displayed in health food shops ; and a full questionnaire was subsequently mailed to all those who returned the short questionnaire. Despite the target recruitment of the postal method acting, about 80 % of meat eaters in the cohort were recruited by post. subsequently, a follow-up questionnaire was sent to participants in 2010, which asked similar questions on their diet and life style, and participants returned the questionnaires between 2010 and 2013. A player flow chart of the recruitment procedure and inclusion into this study is shown as auxiliary figure 1. The discipline protocol was approved by a multicentre research ethics committee ( Scotland A Research Ethics Committee ) and all participants provided written inform consent .

Assessment of diet group and diet

The full baseline questionnaire collected responses to four questions about consumption of meat, fish, dairy products, and eggs, in the shape of “ Do you eat any kernel ( including bacon, overact, poultry, game, kernel pies, sausages ) ? ” or like for the other three food groups. These four questions were used to classify participants into kernel eaters ( participants who reported eating kernel, regardless of whether they ate pisces, dairy, or eggs ), pisces eaters ( participants who did not eat meat but did eat fish ), vegetarians ( participants who did not eat kernel or fish, but did eat one or both of dairy products and eggs ), and vegans ( participants who did not eat kernel, fish, dairy products, or eggs ). The follow-up questionnaire station in 2010 included identical questions on consumption of kernel, fish, dairy products, and eggs ( yes/no ). consequently, at both baseline and follow-up, participants were classified into one of four diet groups : kernel eaters, fish eaters, vegetarians, and vegans. Owing to the minor phone number of vegans, vegetarians and vegans were combined as one diet group in the chief analyses, but the two groups were examined individually for each result in secondary analyses. The service line questionnaire besides included a semiquantitative food frequency section containing 130 items, which asked about dietary inhalation over the past year, and which was previously validated using 16 days ( in four sets of four days ) of weighed dietary records and selected convalescence and concentration biomarkers. 11 12 13 For calculation of food and nutrient intakes, the frequency of pulmonary tuberculosis of each food or beverage was multiplied by a standard assign size ( largely based on data from the UK Ministry of Agriculture, Fisheries, and Food ) 14 and alimentary contentedness of each food or beverage ( based on McCance and Widdowson ’ s food musical composition tables ). 15 Because our prespecified analysis design was to examine disease risks associated with clear-cut dietary groups, the associations of person foods and nutrients with risks were not assessed in this study, but information on intakes of foods and nutrients were used in descriptive and secondary coil analyses .

Assessment of other characteristics

In addition to diet, the baseline questionnaire besides asked questions on sociodemographic characteristics, life style, and checkup history, including questions on education level, smoke, physical activeness, consumption of dietary supplements, and habit of oral contraceptives or hormone substitute therapy in women. socioeconomic condition was categorised by use of the Townsend loss index, 16 based on the participants ’ postcodes. For physical action, based on their responses to questions asked about their occupation and their prison term spent participating in activities including walk, bicycle, and other physical exercises, participants were categorised by a validate physical action index with four levels. 17 Alcohol consumption was determined from responses to five items on the food frequency questionnaire. Questions relating to smoke and alcohol consumption were besides asked on the follow-up questionnaire in 2010.

For biological measurements, torso mass index was calculated from participants ’ self reported acme and weight at recruitment, which was previously found to be accurate compared with measured altitude and system of weights in a establishment study of about 4800 participants. 18 All participants were besides asked at recruitment whether they were will to have their blood imperativeness measured at their general practice and to provide a rake sample. Details of the procedures for rake coerce measurement and blood sample collection, which were conducted in subsets of the cohort, have been previously reported. 8 19 20

Outcome ascertainment

Participants were followed up via record linkage to records from the UK ’ s health service up to 31 March 2016. Outcomes of concern were ischemic heart disease ( code 410-414 from ICD-9 ( international classification of diseases, 9th revision ) or codes I20-I25 from ICD-10 ), including acute myocardial infarct ( ICD-9 410 or ICD-10 I21 ) ; and total stroke ( ICD-9 430-431, 433-434, 436 ; or ICD-10 I60-I61, I63-I64 ), including ischemic stroke ( ICD-9 433-434 or ICD-10 I63 ) and hemorrhagic stroke ( ICD-9 430-431 or ICD-10 I60-I61 ). Details of events, using the relevant ICD-9 or ICD-10 codes, were obtained from hospital records or death certificates .

Exclusion criteria

Participants who were not resident in England, Wales, or Scotland ( n=945 ) were excluded, as were those with no Hospital Episode Statistics data or NHS number ( n=20 ). We besides excluded participants who completed the short questionnaire only ( n=7619 ) ; were younger than 20 ( n=1 ) or older than 90 at recruitment ( n=58 ) ; had no follow-up ( were censored at or before the date of recruitment ( eg, if they were living overseas ), n=364 ) ; could not be traced by the NHS ( n=14 ) ; had an nameless diet group ( if they did not answer the relevant questions to be classified, n=132 ) ; had unreliable alimentary data ( ≥20 % of food frequencies missing, or casual energy intakes < 500 kcal or > 3500 kcal for women or < 800 kcal or > 4000 kcal for men ( 1 kcal=4.18 kJ=0.00418 MJ ), n=1219 ) ; had a self reported history of acute myocardial infarct, stroke, or angina pectoris at recruitment ( n=6837 ) ; or had a date of diagnosis that preceded or equalled the date of recruitment ( n=14 ) .

Statistical analyses

Baseline characteristics and food and alimentary intakes of the EPIC-Oxford participants were summarised by diet group. For self reported body mass index, and measures of blood pressure ( systolic and diastolic rake blackmail ) and blood lipids ( total cholesterol, high concentration lipoprotein cholesterol ( HDL-C ), non-HDL-C ), the means and 95 % assurance intervals are presented, after adaptation for sex and historic period at submission ( in 5-year old age groups ), alcohol consumption ( < 1, 1-7, 8-15, ≥16 g/day ), and physical natural process ( inactive, low natural process, moderately active, very active, obscure ). 17 Cox proportional hazards regression models were used to estimate the hazard ratios and 95 % confidence intervals for the associations between diet group ( meat eaters, fish eaters, vegetarians including vegans ) and each result of interest, with kernel eaters as the reference group. For participants who completed both the baseline and follow-up questionnaire, diet group and relevant time varying covariates ( smoking and alcohol pulmonary tuberculosis ) were updated at follow-up. The implicit in fourth dimension variable was the long time at recruitment to the age at diagnosis, death, or administrative censor, whichever occurred first. For acute accent myocardial infarct or ischemic affection disease, events were censored on the respective outcomes of interest. For total stroke, ischemic stroke, and hemorrhagic accident, events were censored on any stroke. All analyses were stratified by sexual activity, method of recruitment ( general exercise or postal ), and area ( seven regions across the UK ), and adjusted for class of recruitment ( per year from 1994 to 1999 ), education ( no qualifications, basic junior-grade ( eg, O flat ), higher secondary ( eg, A flat ), degree, unknown ), Townsend privation index ( quarters, unknown ), 16 smoke ( never, former, light, heavy, obscure ), alcohol consumption ( < 1, 1-7, 8-15, ≥16 g/day ), forcible activeness ( inactive, low natural process, reasonably active, very active, unknown ), dietary supplement use ( no, yes, strange ), and oral contraceptive use ( no, yes always, stranger ) and hormone surrogate therapy practice ( no, yes ever, nameless ) in women. We used Wald tests to test for heterogeneity of gamble between diet groups. The proportional hazards assumption was assessed on the basis of Schoenfeld residuals, and was not violated for the variables of sake in the adjust model for either ischemic heart disease or stroke ( P > 0.05 for all categories ). Self reported history of high blood pressure ( no, yes, unknown ), high blood cholesterol ( no, yes, unknown ), diabetes ( no, yes, strange ), and body multitude index ( < 20, 20-22.5, 22.5-25, 25-27.5, ≥27.5, unknown ) were assessed as electric potential physiological mediators, since these factors were known to be associated with vegetarian diets, 19 20 21 22 23 a good as being established cardiovascular hazard factors. 24 total fruit and vegetable consumption, total dietary fiber, and sum department of energy inhalation ( each continuous ) were assessed as possible relevant dietary factors. We assessed the effects of likely physiologic mediators and possible relevant dietary factors by adding each variable one at a time to the previous mannequin. An extra model was besides fitted including all electric potential physiologic mediators.

To estimate the population affect of vegetarian diets on cardiovascular health, we assessed the absolute hazard difference for each result between kernel eaters, fish eaters, and vegetarians. Predicted incidence and absolute risk differences were presented as per 1000 population over 10 years, and were estimated by manipulation of luck ratios and 95 % assurance intervals expressed as floating absolute risks, 25 26 which do not alter the value of the hazard ratios but assign an appropriate 95 % assurance interval to all groups, including the character group ( thereby allowing an estimate of the uncertainty in the effect size in the address group ). In kernel eaters, predicted incidence over this time menstruation of each consequence was calculated as ( 1−Sr ) ×1000, where Sr= ( 1−observed incidence in kernel eaters ) 10, and represents the bode 10 year survival ( that is, non-incidence ) in meat eaters. By subtracting this estimate of survival from 1, and multiply by 1000, the resulting appraisal represents incidence per 1000 population over 10 years. For all early diet groups, predicted incidence was calculated as ( 1−SrHR ) ×1000, where HR represents the guess proportion or confidence intervals for each consequence in that diet group. By applying HR or assurance interval estimates in each diet group as an exponential to survival in the reference group, SrHR represents the predict 10 year survival pace in the each of the other diet groups. Absolute risk differences were then calculated as the crude differences between the predict incidence per 1000 population over 10 years between each diet group and the meat eaters. extra sensitivity analyses included analyses using service line diet group only, excluding participants with less than five years of follow-up, including participants recruited via the postal method lone, censoring at age 70 or setting submission time at historic period 70 to evaluate possible differences by old age at event, and performing multiple imputation ( with 10 imputations ) for missing covariates. The percentages of missing values in the covariates were 12.7 % for the Townsend privation exponent, 10.9 % for physical natural process, 6.3 % for education flush, and less than 2 % for each of the other covariates. We assessed heterogeneity in the associations between diet group and hazard of ischemic heart disease or stroke by sexual activity, old age at recruitment ( < 60 or ≥60 years ), smoking condition ( never, former, or current ), torso aggregate index ( < 25 or ≥25 ), bearing of risk factors ( one or more of self reported history of high lineage atmospheric pressure, high blood cholesterol, or diabetes ), and any long term treatment for any illness or condition ( no, yes ) by adding allow interaction terms to the Cox models and testing for statistical significance of interaction across strata using likelihood proportion tests. All analyses were performed with Stata adaptation 14.1 or 15.1 ( Stata Corp, TX, United States ) and P values less than 0.05 were considered significant .

Patient and public involvement

No members of the community or patients were involved in setting the research wonder or the consequence measures, nor were they involved in developing plans for recruitment, design, or execution of the study. They were not asked to advise on interpretation or writing up of results. We are appreciative of our participants who, although not partners, were engaged in the advance of EPIC-Oxford through follow-up questionnaires. The results are disseminated to study participants through newsletters and the study web site ( www.epic-oxford.org/ ) .