It should be stressed that the meat intake in the non-vegetarian group in most of these studies was significantly lower than that of the general population. As
described in a different review, the selection criteria for the non-vegetarians in these studies generally included being affiliated with vegetarians in some way or another, likely explaining their relatively low meat intakes. Furthermore, evidence suggests that many of the self-proclaimed vegetarians in these studies actually consumed meat on a regular basis, resulting in minimal differences in meat intake between the groups.
10Considering the lack of difference in meat intake between the vegetarian and non-vegetarian groups, it would only be expected that these studies would not have the statistical power to demonstrate a significant benefit of a vegetarian diet. Two studies which may be considered as especially having limited statistical power due to such limitations were the Heidelberg Study and Health Food Shoppers Study. In the Heidelberg Study, the non-vegetarian group were predominantly semi-vegetarians, while in the Health Food Shoppers Study, a validity assessment of the survey used to classify the participants vegetarian status suggested that 34% of the participants classified as vegetarians actually consumed meat. In fact, a slightly greater percentage of participants classified as vegetarians in the Health Food Shoppers Study were found to consume meat 3 or more times a week than the non-vegetarians in the Heidelberg Study (7.6% and 6.9%, respectively).
14 18 Taking this into consideration, I performed a sensitivity analysis excluding either the Heidelberg Study or Health Food Shoppers Study from the meta-analysis. Excluding the Health Food Shopper Study alone reduced heterogeneity and strengthened the association between vegetarian status and a reduced risk of mortality from all-causes (RR 0.91 [95% CI, 0.87-0.94], and 0.92 [95% CI, 0.86-0.98] using the fixed effects and random effects model, respectively).
Most of these studies did not provide separate data for length of adherence to a vegetarian diet. As already described earlier in this review, evidence from several of these studies suggest a stronger effect on mortality would have been observed if the analysis was restricted to long-term vegetarians.
10 Another important limitation was that most studies also did not verify changes to vegetarian status of the participants throughout the follow-up, which may in part explain why the association between a vegetarian diet and a reduced risk of mortality weakened over time in several of the studies.
9 10Another important potential limitation of these studies described in detail in the next section, is that some of the participants likely adopted a vegetarian diet in order to improve poor health, such as symptoms of an undiagnosed or developing illness that would ultimately became fatal. Each of these limitations described are expected to have either biased these findings towards null, or even in favor of the non-vegetarian group, suggesting that the findings of this meta-analysis may have significantly underestimated the benefits of an appropriately planned vegetarian diet.
In order to determine which factors may have contributed to the observed reduced risk of death in vegetarians, I also performed separate meta-analyses for the major causes of death. This included mortality from coronary heart disease and cardiovascular disease, and the incidence of all cancers combined. For mortality from coronary heart disease, based on 7 cohorts, the vegetarian group had a statistically significant 25% and 24% reduced risk, using the fixed effects and random effects models, respectively (Fig. 3).
13 14 15 16 17 For mortality from cardiovascular disease, based on 7 cohorts, the vegetarian group had a statistically significant 17% and 14% reduced risk, using the fixed effects and random effects models, respectively (Fig. 4).
13 14 15 16 17 For the Adventist Mortality Study and Adventist Health Study, mortality from cardiovascular disease was derived from pooling the relative risk for mortality from coronary heart disease and stroke.
17 Excluding the Adventist Mortality Study and the Adventist Health Study, the vegetarian group had a statistically significant 8% reduced risk of mortality from cardiovascular disease, using both the fixed effects or random effects model (0.92 [95% CI, 0.85-0.99]).
 |
Figure 3. Coronary heart disease mortality for vegetarians compared to health conscious non-vegetarians in a meta-analysis using the fixed effects model. *Indicates that participants classified as semi-vegetarians were included in the vegetarian group |
 |
Figure 4. Cardiovascular disease mortality for vegetarians compared to health conscious non-vegetarians in a meta-analysis using the fixed effects model. *Indicates that participants classified as semi-vegetarians were included in the vegetation group |
As
described in a previous review, the degree of reduction in risk of mortality from coronary heart disease observed in vegetarians in these cohort studies was generally in proportion to the expected reduced risk based on the differences in levels of total and non-HDL cholesterol, and blood pressure. There is a plethora of evidence, not only from epidemiological studies, but also clinical trials that plant-based diets and nutrients have favorable effects on
total and LDL cholesterol,
blood pressure, among
several other factors which are established risk factors for cardiovascular and all-cause mortality.
19 20 21 22In the Oxford Vegetarian Study, high compared to low intake of saturated animal fat was associated with a nearly 3-fold increased risk of coronary heart disease mortality.
23 Similarly, in a meta-analysis of 11 cohort studies, high compared to low intake of saturated fat was associated with a 32% increased risk of coronary heart disease mortality, despite the inclusion of over-adjustments for dietary and serum lipids.
24 It was also found in the Oxford Vegetarian Study that high compared to low intake of total animal fat and dietary cholesterol was associated with a greater than 3-fold increased risk of coronary heart disease mortality.
23 Furthermore, evidence from thousands of experiments carried out over the last century have shown that the feeding of dietary cholesterol and saturated fat has accelerated the development of atherosclerosis in
virtually every animal species in which researchers were able to find a method to sufficiently elevate cholesterol concentrations. This includes herbivores, omnivores and carnivores from mammalian, avian and fish species, and over one dozen different species of nonhuman primates.
In both the Oxford Vegetarian Study and the Adventist Mortality Study, high compared to low intake of eggs was associated with an increased risk of coronary heart disease mortality.
23 25 However, for ill-defined reasons, these studies were excluded from
several recent meta-analyses. In the Adventist Mortality Study and Heidelberg Study, high compared to no intake of meat was associated with a 50% and almost 5-fold increased risk of coronary heart disease mortality, respectively.
14 25 Similarly, in the Adventist Health Study, high compared to no intake of beef was associated with a greater than 2-fold increased risk for men.
26 In addition, recent meta-analyses of prospective cohorts found that an increment of 1 mg/day of
heme iron, found only in animal tissue, is associated with a 16% and 27% increased risk increased risk of type II diabetes and coronary heart disease, respectively.
27 28 Therefore, the totality of evidence strongly suggests that the observed greater longevity of vegetarians can be explained, at least in part, by the reduced risk of cardiovascular disease as the result of the replacement of animal foods with minimally processed plant foods.
For incidence of all cancers combined, based on 5 cohorts, the vegetarian group had a statistically significant 8%, and borderline significant 6% reduced risk, using the fixed effects and random effects model, respectively (Fig. 5).
13 14 29 30 Excluding the Health Food Shoppers Study removed evidence of heterogeneity and strengthened these findings (RR 0.90 [95% CI, 0.85-0.99] using both the fixed effects and random effects models).
 |
Figure 5. Cancer incidence for vegetarians compared to health conscious non-vegetarians in a meta-analysis using the fixed effects model. *Indicates that participants classified as semi-vegetarians were included in the vegetation group |
The finding of a decreased risk of cancer in vegetarians may also be explained, in part, by a diet devoid in heme iron.
Controlled feeding trials have established that NOCs (N-nitroso compounds) arising from heme iron in meat forms potentially cancerous DNA adducts in the human digestive tract, likely in part, explaining the significant association between heme iron and an increased risk of colorectal cancer in recent meta-analyses of prospective cohort studies.
31 32 33 Heme iron has also been associated with
numerous other cancers. These lines of evidence also provide confidence in the validity of the findings of greater longevity in vegetarians.
Why Some People Choose to Become Health Conscious
 |
Which came first,
vegetarianism or ill health? |
In health research, the reasons why some people chose to become health conscious is critically important when interpreting data from observational studies. This is because it is possible that it may not have been the health conscious lifestyle that caused the examined outcome, but rather the outcome that caused the health conscious lifestyle, ie. reverse causality. As previously described in a different review, reverse causality occurs when the studied effect precedes the cause. An example in health research is the frequent paradoxical observation that former smokers have worse health outcomes than current smokers. These unfavorable outcomes are not explained as being caused by smoking cessation, but rather that those who quit smoking tend to have done so because they were showing symptoms of illness, illnesses that ultimately resulted in the observed unfavorable health outcomes. A similar phenomenon has been observed in nutritional research where sick people tend to adopt a more plant-based diet, suggesting that this would bias observational studies towards showing an unfavorable effect of plant-based diets and nutrients, and therefore a favorable effect of animal based diets and nutrients.34
Although Denise Minger suggests that the favorable health outcomes for vegetarians observed in many studies maybe unrelated to dietary factors, but rather explained by other healthy habits associated with vegetarianism, she fails to consider the possibility that these vegetarians may have become health conscious, including adopting a vegetarian diet, in order to improve poor health. The results of a recent study from the Netherlands illustrates the critical importance of considering reverse causality in research on plant-based diets. The researchers found that 75% of the vegetarian participants with cancer adopted a vegetarian diet after diagnosis, consistent with previous research which found that cancer survivors are highly motivated to adopt a more plant-based diet with the intention of improving poor health.35 36Although health researchers often attempt to partially control for reverse causality by excluding participants who were diagnosed with cancer, cardiovascular disease and other life threatening conditions prior to baseline of a study, it is nearly impossible to fully control for reverse causality, as participants may make dietary changes due to symptoms or unfavorable risk factors that act as markers of an undiagnosed or developing disease. For example, it is known that in studies carried out as far back as the late 1950s, participants with unfavorable blood cholesterol levels tended to reduce the intake of dietary cholesterol and saturated fat (ie. in part, by abstaining from animal foods). This resulted in paradoxical findings where participants who consumed more dietary cholesterol and saturated fat actually had lower serum cholesterol levels. These paradoxical findings were produced, in part, because those participants who continued to consume a diet rich in these lipids were choosing to do so because they were able to maintain lower cholesterol levels despite consuming such a diet (ie. due to favorable genetics).37 Many individuals who attempt to downplay the harmful effects of these lipids, such as Minger has done in her book have chosen to ignore this critical factor when reviewing research on the diet-heart hypothesis.24
In the chapter Herbivore’s Dilemma, Minger reviews several prospective cohort studies that compared the mortality rates of vegetarians to health conscious non-vegetarians, carefully selecting only those studies in which vegetarians were not found to live longer. These studies were the Oxford Vegetarian Study, the Health Food Shoppers Study and the Heidelberg Study, all included in my meta-analysis.
Minger emphasizes the lack of reduced risk of mortality in the vegetarian groups, while neglecting to mention that the difference in intake of meat between the groups was relatively small. Minger especially focuses on the Heidelberg Study, happening to be the smallest study, which found a non-significant increased risk of all-cause mortality in the vegetarian group. Minger points out that in this particular cohort, the vegetarians had greater levels of physical activity, consumed less alcohol and smoked less. Minger then states “Mortality and disease rates, in this case, might be expected to turn up in favor of the vegetarian crowd, even though the opposite ended up happening”[p.270]. This statement is outright misleading as all these factors were controlled for in the analysis. One could also correctly point out that the vegetarians tended to be older than the meat eaters, but suggesting that this would be expected to turn up in favor of the meat eaters would also be misleading, as age was also controlled for. Either way, the differences in mortality between the groups did not even come close to being statistically significant.
There are several important findings that Minger neglected to mention that cast doubt on the suggestion that a vegetarian diet had a harmful effect on the participants in the Heidelberg Study. For example, it was shown in
an earlier follow-up of this study that, similar to the Adventist studies, the participants who adhered to a very low meat diet for at least 20 years had a 29% lower risk of all-cause mortality compared to those who adhered to such a diet for fewer years.10 Excluding the first 5 years of follow-up significantly strengthened this association, resulting in a 45% lower risk of all-cause mortality, suggesting that reverse causation may have attenuated the results for the entire follow-up.
38 This difference in mortality was considerably stronger than the mortality difference between vegetarians and non-vegetarians. Unfortunately the researchers appear to not have provided any data comparing mortality in long-term vegetarians and non-vegetarians.
Indeed, in other cohorts it has been observed that compared to non-vegetarians, short-term vegetarians had a higher rate, whereas long-term vegetarians had a lower rate of all-cause mortality.34 These findings suggest that those who adopted a vegetarian diet more recently had done so due to deteriorating health, biasing the results in favor of the non-vegetarians.
Also as already briefly mentioned, in this study it was also found that intake of meat 3 or more time per week was associated with an almost 5-fold increased risk of mortality from ischemic heart disease, whereas there was a greater than 2-fold risk increase for those who consumed fish more than once per month. One of the reasons that this did not translate into an increased risk of all-cause mortality in the non-vegetarian group may have been due to the very low intake of meat, with less than 7% of the non-vegetarians consuming meat 3 or more times a week.
14 Another important finding was that there was a trend towards a favorable effect of a vegetarian diet on mortality in the early years of follow-up, which declined over time (Fig. 6).
14 This suggests the likelihood of undocumented changes in vegetarian status in a portion of the participants throughout the follow-up, which would be expected to have attenuated a possible favorable effect of a vegetarian diet. A similar phenomenon was also observed in several other studies included in my meta-analysis.
9 |
Figure 6. Standard Mortality Rates (SMRs) and their 95% CIs for all-cause mortality, dietary group, and 5-year follow-up period, separately for males and females. (○, female vegetarians; •, female nonvegetarians; □, male vegetarians; ▪, male nonvegetarians). |
Minger also points out that there was a trend towards a higher rate of mortality among the vegans in the Heidelberg Study, which included only 60 vegans, while neglecting to mention that in the Adventist Health Study 2, which included almost 100 times many vegans found that compared to the non-vegetarians, the vegans, especially male vegans experienced a lower rate of mortality.
16 In the Heidelberg Study, while current smokers had the same mortality rate from cancer as never smokers, past smokers suffered from a 70% greater risk of cancer mortality than current smokers. This strongly suggests that many of the past smokers in this cohort quit smoking because they had deteriorating health.
14 This raises the question as to whether the participants in this cohort also had a tendency to adopt a vegetarian diet after becoming ill, possibly explaining the unfavorable mortality rates in the vegetarian group. Perhaps the reason why the vegetarians in this group tended to be older, was not because they were becoming more ethical with old age, but because they were becoming more health conscious, which included adopting a vegetarian diet due to deteriorating health with old age.
As these lines of evidence described suggest, the trends towards elevated mortality in the vegetarian groups in the Heidelberg Study may be explained by reverse causation. It is more than possible that it was not the vegetarian diet that caused these unfavorable outcomes, but deteriorating health, which would ultimately resulted in these unfavorable outcomes that caused a portion of the participants to adopt a vegetarian diet.
