Wheat flour versus rice consumption and vascular diseases: Evidence from the China Study II data

Why does wheat flour consumption appear to be significantly associated with vascular diseases? To answer this question, we analyzed data on rice consumption, wheat flour consumption, total calorie consumption, and mortality from vascular diseases obtained from the China Study II dataset. This dataset covers the years of 1983, 1989 and 1993 and includes data related to biochemistry, diet, lifestyle, and mortality from various diseases in 69 counties in China. Our analyses point to a counterintuitive conclusion: it may not be wheat flour consumption that is the problem, but the culture associated with it, which is characterized by decreased levels of physical activity, decreased exposure to sunlight, increased consumption of processed foods, and increased social isolation. Wheat flour consumption may act as a proxy for the extent to which this culture is expressed in a population. The more this culture is expressed, the greater is the prevalence of vascular diseases.


Introduction
Recent years have seen the vilification of wheat flour as a food ingredient, particularly in health books aimed at non-technical readers, with its consumption being presented as a major cause of a variety of modern diseases, including vascular diseases (Davis, 2011;Taubes, 2008).The term "vascular diseases" is used here to refer to diseases that affect the heart and blood vessels, including ischemic heart disease, stroke, hypertensive heart disease, aortic aneurysms, cardiomyopathy, atrial fibrillation, and peripheral artery disease (Mendis et al., 2011).
A key hypothesis underlying this indictment of wheat flour is that foods made with it induce acute insulin elevations, thus promoting excess body fat gain when consumed regularly over a period of several years.Insulin is a hormone that tends to rise sharply in response to the intake of foods that, like those made with wheat flour, have high glycemic indices and, particularly, high glycemic loads (Barclay et al., 2008;Cordain et al., 2003;Ludwig, 2002).Excess body fat gain then induces insulin resistance, which, in turn, increases the likelihood of vascular diseases (Apovian & Gokce, 2012;Poirier & Eckel, 2002;Rexrode et al., 1996;Sowers, 1998;Van Gaal et al., 2006).
However, one of the key problems with the above hypothesis comes from epidemiological evidence related to rice consumption and its health effects.Boiled rice also scores relatively high in terms of glycemic index and glycemic load, and its regular consumption is generally associated with a low prevalence of vascular diseases in China and Japan (Liu & Li, 2000;Matsuzaki, 1992).This dilemma leads to an important research question: Why does wheat flour consumption appear to be significantly associated with vascular diseases?
The importance of this research question comes from the fact that wheat flour products are widely consumed around the world, and that their broad availability and relatively low cost help mitigate global hunger (Singh, 2009).To answer this research question, we analyzed data on rice consumption, wheat flour consumption, total calorie consumption, and mortality from vascular diseases obtained from the China Study II dataset (Junshi et al., 2006).This is an extensive dataset that covers the years of 1983, 1989 and 1993 and includes data is related to biochemistry, diet, lifestyle, and mortality from various diseases in 69 counties in China.
Our investigation suggests that wheat flour consumption in China, and perhaps in many other countries, is accompanied by culturally-induced factors including decreased levels of physical activity, decreased exposure to sunlight, increased consumption of processed foods, and increased social isolation.We conclude that consumption of wheat flour per se may not be the problem and that evidence points to these culturally-induced factors correlated with wheat flour consumption playing a role in cardiovascular disease.Therefore, wheat flour consumption may be a proxy for the extent to which these culturally-induced factors are expressed in a population.The more they are expressed, the greater the prevalence of vascular diseases.

Research Background
There is strong evidence linking consumption of modern processed foods based on wheat flour and vascular diseases (Davis, 2011).The evidence refers primarily to modern foods made with refined wheat flour and not to more traditional wholewheat products (Harris & Kris-Etherton, 2010;Smith & Tucker, 2011).Examples of foods based on refined wheat flour include white flour, white bread, hominy, and most types of pasta.Foods based on refined wheat flour are widely available in modern cities and widely consumed by urban populations.
Although the link between consumption of wheat flour and vascular diseases has a hypothetical connection to elevated insulin, boiled rice also scores relatively high in terms of glycemic index and glycemic load, but its consumption is generally associated with a low prevalence of vascular diseases in China and Japan (Liu & Li, 2000;Matsuzaki, 1992).However, obesity is a clear, major cause of vascular diseases, as determined by innumerable studies conducted over different periods of time and with a variety of participants (Apovian & Gokce, 2012;Poirier & Eckel, 2002;Rexrode et al., 1996;Sowers, 1998;Van Gaal et al., 2006).Total calorie consumption has been found to be associated with obesity in several studies (Rosenheck, 2008;Waxman & Stunkard, 1980), whereas other studies found no link between these two variables (Braitman et al., 1985).Wheat flour consumption may significantly influence total calorie consumption, thus causing obesity (Galal, 2002).Rice, however, appears to contribute significantly to total calorie consumption and yet not to be a major cause of obesity in countries such as China and Japan (Liu & Li, 2000;Matsuzaki, 1992).
The discussion above indicates the necessity of a path analysis (Kock, 2015;Kock & Gaskins, 2014;Wright, 1934) examining the relationships among rice consumption (Rice), wheat flour consumption (Wheat), total calorie consumption (Calorie), and mortality from vascular diseases (MVasc; Figure 1).Our goal is to try to understand why consumption of wheat flour, but not rice, appears to be significantly associated with vascular diseases.
The path analysis includes a possible correlation between rice consumption (Rice) and wheat flour consumption (Wheat) and allows us to explore the extent to which they are consumed together or separately.If they are consumed together, a positive correlation will be observed, and if they are consumed separately, we will see a negative correlation.In countries such as China and Japan, rice consumption is associated with adherence to traditional cultural practices in rural contexts, whereas consumption of products based on refined wheat flour is associated with urbanization and adoption of modern lifestyles (Liu & Li, 2000;Matsuzaki, 1992).This cultural divide would typically be reflected in a strong negative correlation between rice consumption (Rice) and wheat flour consumption (Wheat) in these countries.
Next, the path analysis assesses rice (Rice) and wheat flour consumption (Wheat) in the context of total calorie consumption (Calorie).
Finally, the analysis allows us to assess the extent to which those hypothesized predictors influence mortality from disease (MVasc).Given that there is evidence that total calorie consumption is not associated with obesity (Braitman et al., 1985), and thus may not be associated with vascular diseases, it would be interesting if rice and wheat flour were nevertheless associated, positively or negatively, with vascular diseases.Counterintuitive results such as these would arguably help in our understanding of why wheat flour consumption appears to be significantly associated with vascular diseases by enabling us to rule out some possible reasons and look for alternative explanations.

Research Method
Data on rice consumption (Rice), wheat flour consumption (Wheat), total calorie consumption (Calorie), and mortality from vascular diseases (MVasc) was obtained from the China Study II dataset (Junshi et al., 2006).This dataset, which is described in more detail in Appendix A, contains data for the years of 1983, 1989 and 1993.The data is related to biochemistry, diet, lifestyle, and mortality from various diseases.It covers 69 counties in China.
The data was analyzed through a nonlinear robust path analysis (Kock, 2015;Kock & Gaskins, 2014) and implemented with the software WarpPLS 5.0 (Kock, 2015).This software was chosen due to some of its advanced features that were needed in our study, such as the ability to conduct nonlinear robust path analyses, generation of outputs for model fit and quality assessment, as well as support for multicollinearity and predictive validity tests.The main software settings used are provided in Appendix B.
In a nonlinear robust path analysis, nonlinearity among linked variables is taken into consideration in the calculation of coefficients of association.This is an important feature in our analyses, because food-mortality associations tend to be nonlinear (Johansson, 1994;O'Sullivan et al., 2013).Since this is a robust type of analysis, restrictive parametric conditions such as multivariate normality do not have to be met.This is noteworthy because we conducted a normality assessment that suggested that none of the four variables in our model was normally distributed.This assessment employed the classic Jarque-Bera test (Bera & Jarque, 1981;Jarque & Bera, 1980) and Gel and Gastwirth's (2008) robust modification of this test.
Because data from multiple years was used, we controlled for a variable storing the year in which the data were collected and for cluster mean variables that stored the average rice consumption (Rice), wheat flour consumption (Wheat), and total calorie consumption (Calorie), for each of the years in which the data was collected.We also replicated our analyses using data for only a single year, namely 1989, obtaining results nearly identical to those based on the whole sample.The year of 1989 was chosen for this replication because it was the intermediate year between the extremes represented by the two other years in the dataset, and because it was the year for which the most complete data subset was available.Additionally, we controlled for the ratio of females-to-males in each county to account for differences in diet and lifestyle between the sexes and related differences in mortality (He et al., 2005).
We conducted several model and data validation tests, summarized in Appendix B, to ensure that the results were not biased and that there was a good fit between our model and the data.Those tests suggest that the model is free of multicollinearity (Kock & Lynn, 2012), that acceptable predictive validity values were achieved in connection with all variables (Geisser, 1974;Stone, 1974;Kock 2015), and that all model fit and quality indices reached suitable levels (Kock, 2011;2015).

Results and Discussion
Coefficients of association were assigned to each of the hypothesized relationships in our model (Figure 2).Rice consumption (Rice) and wheat flour consumption (Wheat) were significantly and negatively correlated (r = -0.803,P < 0.001) with each other.Both rice consumption (Rice) and wheat flour consumption (Wheat) were significantly and positively associated with total calorie consumption (Calorie), with the association involving rice consumption (Rice) (β = -0.318,P < 0.001) being much stronger than the association involving wheat flour consumption (Wheat) (β = -0.174,P < 0.001).
These results suggest a major divide in terms of rice and wheat flour consumption.The very strong and negative correlation between rice consumption (Rice) and wheat flour consumption (Wheat) suggests that the more rice is consumed in a county, the less wheat flour is consumed, and vice-versa.
As noted, in terms of total calorie consumption (Calorie), the association involving rice consumption (Rice) is much stronger than the association involving wheat flour consumption (Wheat).This suggests that rice tends to make up a bigger percentage of total calorie consumption than wheat flour.In other words, a diet primarily of rice is one in which other components do not contribute as many calories as in a diet primarily of wheat flour.Modern processed foods tend to have high calorie contents and be generally less healthy than foods that are not processed or are minimally processed (Cutler et al., 2003).Therefore, in our dataset, rice consumption (Rice) may be a marker of a healthier diet, with fewer processed foods in it.This seems to have historically been the case in China and Japan (Liu & Li, 2000;Matsuzaki, 1992).
Overall, the relationship between rice consumption (Rice) and mortality from vascular diseases (MVasc) is negative (Figure 3), which means that the linear equivalent of the relationship would have a negative slope.The nonlinear relationship depicted, which presents a better fit with the data than the equivalent linear relationship, suggests that at around 400 grams of rice consumed per day per person, the relationship turns from negative to positive.
The nonlinear pattern of the association between rice consumption (Rice) and mortality from vascular diseases (MVasc) is what one would expect in connection with consumption of almost any healthy food or liquid, including water (Johansson, 1994;Sonnenblick et al., 1993).Intake up to a point promotes health, thus decreasing mortality but beyond that point becomes health-detrimental.The left side of the curve shows a pattern of increased mortality from vascular diseases (MVasc) with decreased rice consumption (Rice), which may be due to malnutrition negatively affecting resistance against infections and thus promoting vascular diseases (Leinonen & Saikku, 2002).Here, rice consumption is a marker of nutrition status.
The right side of the curve shows a pattern of increased mortality from vascular diseases (MVasc) with increased rice consumption (Rice), likely due to excess rice consumption.Even though the detrimental effect of excessive rice consumption in terms of mortality from vascular diseases is not mediated by total calorie consumption, it may be mediated by excess body fat gain in response to a combination of excess calories and high insulin levels (Cordain et al., 2003;Taubes, 2008).The excess body fat gain promotes insulin resistance, which is a known precursor of vascular diseases (DeFronzo & Ferrannini, 1991).Also, excessive rice consumption may crowd out other foods that bring in important nutrients.
Overall, the relationship between wheat flour consumption (Wheat) and mortality from vascular diseases (MVasc) is positive, and, although nonlinear, no change in sign is observed (Figure 4).That is, increases in wheat flour consumption (Wheat) are associated with increases in mortality from vascular diseases (MVasc) throughout the dataset when we control for rice consumption (Rice) and total calorie consumption (Calorie).
The pattern of the association between wheat flour consumption (Wheat) and mortality from vascular diseases (MVasc) is what one would expect in connection with consumption of a food that is detrimental to health.This is puzzling, given that wheat flour is similar to rice in terms of carbohydrate content, glycemic index and load, and total calorie content.In fact, wheat flour contains greater contents of protein and various vitamins than rice (USDA, 2015), being in this respect more nutritious.The human body has minimal daily protein and vitamin intake requirements (Campbell et al., 1994;Selhub et al., 1993;Young et al., 1973).While both wheat flour and rice are poor sources of many required nutrients, such as vitamins A and C, generally speaking, the minimal daily nutrient requirements appear to be met more easily with wheat flour than rice (USDA, 2015).The primary protein in wheat flour is gluten, which has been hypothesized to be detrimental to the health of individuals who suffer from celiac disease (Mora et al., 1993).However, the incidence of celiac disease in Japan and China appears to be extremely small compared with other countries (Cummins et al., 2009), and there is very limited evidence of gluten-related health problems in individuals who do not suffer from celiac disease (Biesiekierski et al., 2011;Biesiekierski et al., 2013).Therefore, it is difficult to explain the pattern of the association between wheat flour consumption (Wheat) and mortality from vascular diseases (MVasc) based on the gluten content of wheat flour.
Rice consumption (Rice) ranges from zero to about 785 grams of rice consumed per day per person.Wheat flour consumption (Wheat) ranges from zero to about 751 grams of wheat flour consumed per day per person.These similar ranges lead to different ranges of variation in mortality from vascular diseases (MVasc); with a range from 2.48 to 2.96 associated with rice consumption (Rice), and a range from 2.44 to 3.85 associated with wheat flour consumption (Wheat).These different ranges are reflected in the strength of the associations with mortality from vascular diseases (MVasc); wheat flour consumption (Wheat) was more strongly associated with mortality from vascular diseases (MVasc) than rice consumption (Rice).

Wheat and Culture
The results of our analyses strongly suggest that wheat flour consumption appears to be significantly associated with vascular diseases, but so far the answer to our original research question has been elusive: Why does wheat flour consumption appear to be significantly associated with vascular diseases?Biochemical explanations, such as acute insulin responses, have proven to be less than satisfactory.In this section, we put forth the notion that wheat consumption may be a marker of cultural elements that increase the prevalence of vascular diseases.
In Asian countries, the move from rice to wheat flour consumption has historically paralleled a move from more traditional and rural to more modern and urban lifestyles (Liu & Li, 2000;Matsuzaki, 1992;Pingali, 2007).This is essentially a cultural move, with impacts on lifestyle variables such as physical activity, exposure to sunlight, and consumption of processed foods.Compared with modern urban life, traditional rural life generally incorporates greater levels of physical activity, greater exposure to sunlight, and lower consumption of processed foods.
Greater levels of physical activity are strongly associated with decreased mortality from vascular diseases (Blair et al., 1996;Thompson et al., 2003).Greater exposure to sunlight leads to greater levels of circulating vitamin D, which is associated with lower prevalence of vascular diseases (Holick, 2004;Wang et al., 2008).Lower consumption of processed foods, and their replacement with unprocessed or minimally processed foods such as fruits and vegetables, is also associated with lower prevalence of vascular diseases (Bazzano et al., 2002).
When we make the reasonable assumptions that rice consumption (Rice) and wheat flour consumption (Wheat) are respectively correlated with rice and wheat farming, another cultural change paralleling the move from rice to wheat flour consumption is increased social isolation (Talhelm et al., 2014).There is conclusive evidence that social isolation is a significant cause of vascular diseases (Rozanski et al., 1999).The link between vascular diseases and social isolation seems to be mediated by chronic mental stress (Mallon et al., 2002).Talhelm et al. (2014) studied 1162 Han Chinese participants in six sites.Their study suggests that a history of farming rice or wheat leads to differences in the levels of social isolation found in different cultures.The authors found evidence that farming rice makes cultures more interdependent, leading to lower levels of social isolation; whereas farming wheat makes cultures more independent, promoting greater levels of social isolation.The study, consistently with reviews of it (Henrich, 2014), concludes that these agricultural legacies extend to other parts of the modern world.
The above discussion provides the basis on which we can provide a counterintuitive answer our original research question: Why does wheat flour consumption appear to be significantly associated with vascular diseases?In China, the reason may be what could be called a "wheat culture," characterized by increased consumption of wheat flour and a set of correlated culturally-induced factors including decreased levels of physical activity, decreased exposure to sunlight, increased consumption of processed foods, and increased social isolation.We do not believe that there is enough evidence to unequivocally point to consumption of wheat flour per se as the problem.The problem appears to lie in the set of correlated culturally-induced factors, which occur together with wheat flour consumption and which appear to be health-detrimental.

Limitations and Alternative Explanations
Like any empirical research study, ours may be biased by omitted confounders.That is, it is possible that the relationships we uncovered among rice consumption (Rice), wheat flour consumption (Wheat), total calorie consumption (Calorie), and mortality from vascular diseases (MVasc) are biased by correlated variables that have not been explicitly included in the model.While we have conducted a number of tests to rule out biases, the problem of omitted confounders is pervasive in epidemiological research.In this type of research, investigators try to create and validate models based on incomplete information.
The possibility of the existence of omitted confounders calls for the consideration of alternative explanations.For example, there is evidence that vascular diseases may be caused, at least in part, by infectious diseases (Fong, 2000;Leinonen & Saikku, 2002).This is partly supported by our analysis of the nonlinear relationship, with a parabolic relationship between rice consumption and vascular diseases.The left side of the parabola shows a pattern of increased mortality from vascular diseases with decreased rice consumption, which may be due to malnutrition negatively affecting resistance against infections and thus promoting vascular diseases.
However, our analyses do not rule out other possible factors.In fact, the existence of a complex interplay of factors would be consistent with past research on the multifaceted and intricate nature of vascular diseases (Fong, 2000).We speculate here that the emergence of a wheat culture in China may be a fundamental causative factor behind an increasing incidence of vascular diseases.Still, one could argue that such a wheat culture would be more urbanized, with better access to treatments against infectious diseases and, thus, should be associated with less vascular disease.This underscores the need for more research on the topic, including a possible confounder-incidence of infectious diseases.
Another argument that could be raised against our wheat culture hypothesis is related to wheat flour consumption and gut health.Indeed, as we noted earlier, the primary protein in wheat flour is gluten, which can be detrimental to the health of individuals who suffer from celiac disease (Mora et al., 1993).Although celiac disease appears to be very rare in China compared with other countries (Cummins et al., 2009), and there is scant evidence of gluten-related health problems in individuals who do not suffer from celiac disease (Biesiekierski et al., 2011;Biesiekierski et al., 2013), our findings underscore the need for more research on the topic.
We recommend further research using the general methodological approach we employed here but with more complex models that include additional variables that could have confounded the results of our analyses.Such variables include level of urbanization, level of poverty, incidence of infectious diseases, access to treatments against infectious diseases, and incidence of gut health problems.We also recommend future research addressing the wheat culture hypothesis we put forth in other parts of the world because our hypothesis builds primarily on data from China.

Conclusion
We set out to answer a research question: Why does wheat flour consumption appear to be significantly associated with vascular diseases?In order to do so, we analyzed data on rice consumption, wheat flour consumption, total calorie consumption, and mortality from vascular diseases obtained from the China Study II dataset (Junshi et al., 2006).This dataset covers the years of 1983, 1989 and 1993; the data is related to biochemistry, diet, lifestyle, and mortality from various diseases in 69 counties in China.
Our search for an answer to the research question above has led us to investigate biochemical explanations, related to carbohydrate content, glycemic index and load, and total calorie content.We also explored the role of gluten to explain the apparently health-detrimental nature of wheat flour.None of these explanations were found to be fully satisfactory.This was in part due to the similarities in biochemical responses elicited by both wheat flour and rice, and the fact that rice consumption seems to be generally associated with good health in Asian countries.
Wheat flour consumption in China, and perhaps in many other countries, may be accompanied by culturally-induced factors characteristic of a rural-to-urban transition, including: decreased levels of physical activity, decreased exposure to sunlight, increased consumption of processed foods, and increased social isolation.This health-detrimental "wheat culture" is perfectly compatible with the pattern of the association between wheat flour consumption and mortality from vascular diseases that our analyses uncovered.The pattern suggests that any wheat flour consumption is detrimental in terms of vascular diseases, which cannot be reconciled with biochemical evidence.
Our analyses point at a counterintuitive conclusion: it may not be wheat flour consumption that is the problem, but the wheat culture associated with it.In our analyses, wheat flour consumption appears to act as a proxy for the extent to which this wheat culture is expressed in a population.The more this culture is expressed, the greater is the prevalence of vascular diseases.

Figure 2 .
Figure 2. Model with results in terms of overall effects.The correlation between Rice and Wheat is a bivariate coefficient of association.The path coefficients next to straight arrows are multivariate coefficients of association; they assume hypothesized directions of causality and control for competing effects.The existence of competing effects is represented by two or more arrows pointing at the same variable.Notes: * P<.001; NS non-significant.

Figure 3 .
Figure 3. Mortality from vascular diseases and rice consumption.Notes: Horizontal (X) axis values are grams of rice consumed per day per person, averaged for each county; vertical (Y) axis values are deaths from all vascular diseases within ages 35-69 per 1,000 people, averaged for each county.

Figure 4 :
Figure 4: Mortality from vascular diseases and wheat flour consumption.Notes: Horizontal (X) axis values are grams of rice consumed per day per person, averaged for each county; vertical (Y) axis values are deaths from all vascular diseases within ages 35-69 per 1,000 people, averaged for each county.