In this prospective analysis of a previous cancer-screening cohort of middle-aged and older adults in the US, over 99% of the participants self-identified as non-vegetarian and the rest followed lacto- and/or ovo-vegetarian, pesco-vegetarian or vegan diets. Pesco-vegetarians had a lower mortality risk than omnivores and vegans, which lost statistical significance after further adjusting for lifestyle and sociodemographic factors and comorbidity. We observed no difference in risk of all-cause mortality between vegetarian and non-vegetarian diets.

Previously, observational studies reported that vegetarian diets were associated with a lower risk of chronic diseases (e.g., CVD and type 2 diabetes) and their risk factors (e.g., obesity, blood lipid profiles and glucose control), compared with omnivorous diets; however, findings were less clear in terms of all-cause mortality [3, 4]. Comparing with omnivores, no difference in risk of mortality among vegans and lacto- and/or ovo-vegetarians were reported in all three previous studies [8, 11, 12]. For pesco-vegetarians, comparing with omnivores, the Adventist Health Study 2 reported a statistically significant inverse association (0.81 [0.69–0.94]) [12], but not the other study (0.79 [0.59–1.06]) [11]. Although the comparison was with regular meat eaters, instead of omnivores as in our and other previous studies, in the Oxford Vegetarian/Oxford- European Prospective Investigation into Cancer and Nutrition (EPIC) Study, the risk of mortality did not differ from vegans, lacto- and/or ovo-vegetarians or pesco-vegetarians [8]. Furthermore, previous six studies compared risk of mortality between vegetarians and non-vegetarians. Three studies of Seventh-Day Adventists—the Adventist Mortality Study and Adventist Health Study 1 and 2—reported an inverse association [12, 16], but other three previous studies reported no significant association [9,10,11]. Altogether, our null finding is consistent with the previous studies of vegetarianism and all-cause mortality, except for cohort studies of Adventists.

Differences in associations may be due to the fact that self-identified vegans in our study might have differed in chronic diseases and their risk factors from those of previous studies targeting vegetarians such as Adventist studies. The high prevalence of comorbidities and BMI in the vegan group compared with the Adventist and other previous studies, in which vegans had the lowest average BMI among the diet groups [12, 17], may indicate that our self-identified vegans recently switched to a vegan diet because of health concerns. In our questionnaire, vegetarianism questions only pertained to the past 12 months, lacking the information on the specific duration of following vegetarian/vegan diets. In the previous pooled analysis of five prospective studies, participants who have followed a vegetarian diet for > 5 years had a lower risk of all-cause mortality than omnivores (0.93 [0.79–1.09]), but those followed ≤ 5 years had an increased risk (1.20 [1.04–1.38]) [16]. In addition, smoking and alcohol drinking were far more common in our study than Adventist Health Study 2 (9.3% vs. 1.1% for current smokers and 72.0% vs 10.1% for alcohol drinkers, respectively). Given that the PLCO study was conducted in a general, non-vegetarian population and same association patterns persisted after excluding the first two years of the follow-up period, the potentially shorter duration of following vegetarian/vegan diets might not have been sufficiently long enough for our study participants to have the typical health benefits associated with vegetarian/vegan diets.

Strength of this study is its prospective collection of diet and other important confounders with a relatively long follow-up (18 years) of death and vital status. In addition, our study is the first study conducted in a general, non-vegetarian population in the US to investigate associations between vegetarianism and mortality. A potential limitation, however, is low proportions of vegans/vegetarians in our study population (0.7%), which was lower than a previous report (1.9%) [18], but higher than another (0.4%) [19]. In addition, the self-identified vegetarian diet status was based on two questions regarding foods they excluded from their diet. Because of this, a possibility of misclassifications for each diet group cannot be fully ruled out, as reported previously on discrepancies between self-identified vegetarianism and their food and beverage consumption [20]. Classifying vegetarian diet status based on food and beverage consumption is beyond the scope of this study and warrants further investigation. Furthermore, we did not have information on the duration of vegetarian diets, which was reported to modify the direction of associations with mortality [16]. Due to the very small number of deaths in vegans and vegetarians, we were unable to conduct cause-specific nor subgroup analyses. Lastly, physical activity information at baseline was not collected and we were unable to adjust in our analyses.

In conclusion, we found no difference in risk of all-cause mortality by self-identified vegetarian diet status. Similar to previous studies, our vegetarians attained more education, were less likely to smoke cigarettes and drink alcohol, and had a lower BMI; however, vegans had higher comorbidities and similar BMI to omnivorous participants, suggesting that they likely recently switched to vegans because of health concerns. Future studies of vegetarian diets and mortality need to be conducted in general populations and consider the duration of following vegetarian diets.