# Revisiting Eggs and Dietary Cholesterol

Buckle up, folks. This is going to be a long one. In this post, I want to discuss three main things:

Let’s get right into it.

## A Summary of the JAMA Study

Last week, a nutritional epidemiological study (from now on nutri epi) published in JAMA posited that egg consumption was associated with an increased risk of cardiovascular disease incidence.1 The results gained significant coverage from the media.

With that coverage came the usual criticism of “correlation not being causation,” and how nutri epi studies produce findings that are often contradicted by randomized trials, etc. Unlike many of these surface-level criticisms, I want to dive a little deeper into the findings of this study and the utility and limitations of studies like this. But first, a quick summary of the study.

## Who’s Being Studied?

This study included data from six prospective cohorts and had a total sample size of 29,615 adults. That’s a lot of people. Those who follow nutrition studies should be familiar with many of the cohorts included such as:

• the Atherosclerosis Risk in Communities (ARIC) Study
• the Coronary Artery Risk Development in Young Adults (CARDIA) Study
• the Framingham Heart Study (FHS)
• the Framingham Offspring Study (FOS)
• the Jackson Heart Study (JHS)
• the Multi-Ethnic Study of Atherosclerosis (MESA)

Data collection started in early 1985 and halted in the summer of 2016 (nearly 31 years of data). The project was involved in answering the question of whether dietary cholesterol was associated with cardiovascular disease incidence.

## What and How Were the Data Collected?

So how were dietary data collected? The authors give a very brief description in the methods,

Using a standardized protocol and data dictionary, diet data were harmonized cohort by cohort. Briefly, consumption frequencies were converted into estimated number per day using the middle value (eg, 3-4 times per week = 0.5 times per day).

However, a look at the supplementary files gives us more insight into how the data were collected:

So they mainly used various food frequency questionnaires which differed depending on the cohort.

The primary outcomes were…

The primary outcomes were incident CVD and all-cause mortality. Incident CVD was a composite end point of fatal and nonfatal coronary heart disease (0), stroke, heart failure, and CVD death from other causes.

Here are some of the covariates they collected data for

Standardized questionnaires and laboratory protocols were used to collect information on the following variables: age, sex, race/ethnicity, education, lifestyle factors (including smoking, alcohol intake, and physical activity), body mass index (BMI, calculated as weight in kilograms divided by height in meters squared), blood pressure, lipid profile, medication use, and medical conditions.

and what they adjusted for in their three statistical models. Let’s focus mainly on model 2 since they use that as a reference.

Models were sequentially adjusted for age, sex, race/ethnicity (white, black, Hispanic, Chinese), education (<high school, high school, some college or more) (model 1); plus total energy, smoking status (current, former, never), smoking pack-years (0, 0.1-4.9, 5.0-9.9, 10-19.9, 20-29.9, 30-39.9, ≥40), cohort-specific physical activity z score, alcohol intake (gram), use of hormone therapy (y/n) (model 2);

## What Were the Results?

The authors graphed the hazard ratios by the amount of dietary cholesterol consumed per day (mg) for CVD incidence…

So what were the main results that the authors reported?

Based on model 2, each additional 300 mg of dietary cholesterol consumed per day was significantly associated with higher risk of incident CVD (adjusted HR, 1.17 [95% CI, 1.09-1.26]; adjusted ARD, 3.24% [95% CI, 1.39%-5.08%]; Figure 2A) and all-cause mortality (adjusted HR, 1.18 [95% CI, 1.10-1.26]; adjusted ARD, 4.43% [95% CI, 2.51%-6.36%]; Figure 2B).

So a 17% increase in the risk of incident CVD where a 9% increase all the way up to a 26% increase were reasonably compatible with the data and its assumptions. The results are practically the same for all-cause mortality.

The authors also adjusted for individual foods or combinations of foods to see how the effect deviated with model 2 being used as the reference.

• For incident CVD (shown above), the effect estimate increased from 17% to 20% when adjusted for eggs alone, and decreased to 13% when adjusted for eggs, unprocessed red meat, and processed meat.

• For all-cause mortality (graph not shown), the effect estimate dropped from a 17% increase to a 13% increase when adjusted for eggs alone and dropped to a 5% increase when adjusted for eggs, unprocessed red meat, and processed meat.

Here’s what the authors concluded:

The current study found that the significant associations of dietary cholesterol consumption with incident CVD and all-cause mortality were independent of fat amount and quality of the diet. Further, the significant association between egg consumption and incident CVD was fully accounted for by the cholesterol content in eggs

Mechanistically, eggs and processed or unprocessed red meat are rich in other nutrients such as choline, iron, carnitine, and added sodium (for processed meat) that have been implicated in CVD risk via different pathways.

And then they discuss the limitations of their study:

First, appropriate interpretation of the study findings requires consideration of measurement error for self-reported diet data. Further, this study relied on single measurement of egg and dietary cholesterol consumption. Exposure misclassification may be of concern, but results were similar when censoring participants at different time points…

…Fifth, generalizing our results to non-US populations requires caution due to different nutrition and food environments and chronic disease epidemiology. Sixth, the study findings are observational and cannot establish causality.

Interestingly enough, right after saying the results cannot establish causality, they write this,

Among US adults, higher consumption of dietary cholesterol or eggs was significantly associated with higher risk of incident CVD and all-cause mortality in a dose-response manner. These results should be considered in the development of dietary guidelines and updates.

Well, I have a lot to say, but let’s first take a look at how the media interpreted and reported these results. For that, we shall consult the Google oracle.

## How Did the Media Respond?

This is one of my favorite ones.

Okay, now that we’ve done that, we can discuss the measurements and analyses.

## How Reliable Are the Measurements and Analyses?

As the authors mentioned in the methods, and supplementary files, most of the dietary data were collected via a food frequency questionnaire (FFQ). This is important for a couple of reasons, and I’m going to dive a little deeper than the usual “FFQ data are useless”.

FFQ data are known to suffer from serious measurement error, and there is a giant body of literature on this. For example, such data can be prone to classical measurement error ($$W_{ij}=X_{i}+ \epsilon_{ij}$$), systematic error, heteroscedastic error, and differential error.

Methods like regression calibration, moment reconstruction, multiple imputation, and graphical methods can be used to help address many of these errors, along with sensitivity analyses. For a more detailed read of this, see Keogh & White, 2014.2

Unfortunately, the authors of the JAMA study hardly bothered with this, and merely pointed out in their limitations section that their study was limited by measurement error.

This study has several limitations. First, appropriate interpretation of the study findings requires consideration of measurement error for self-reported diet data. Further, this study relied on single measurement of egg and dietary cholesterol consumption. Exposure misclassification may be of concern, but results were similar when censoring participants at different time points…

This isn’t something to take this lightly given how serious of a problem it is. There have been extensive criticisms published of FFQs with one of the most prominent of reviews stating the following six arguments as to why memory-based questionnaires should not be used.3

1. The use of memory-based methods is founded upon two inter-related logical fallacies: a category error and reification
2. Human memory and recall are not valid instruments for scientific data collection
3. The measurement errors of memory-based dietary assessment methods are neither quantifiable nor falsifiable; this renders these methods and data pseudoscientific
4. The post hoc process of pseudoquantification is impermissible and invalid
5. Memory-based dietary data were repeatedly demonstrated to be physiologically implausible (i.e., meaningless numbers)
6. The failure to cite or acknowledge contrary evidence and empirical refutations contributed to a fictional discourse on diet-disease relations

The authors don’t even believe that the measurement error from memory-based questionnaires can even be addressed! They conclude the following:

Our conclusion is that nutrition epidemiology is a degenerating research paradigm in which the use of expedient but nonfalsifiable anecdotal evidence impeded scientific progress and engendered a fictional discourse on diet-health relations. The continued funding and use of memory-based methods such as Food Frequency Questionnaires and 24-hour dietary interviews is anathema to evidence-based public policy.

Thus, our recommendation is simply that the use of memory-based methods must stop and all previous articles presenting memory-based methods data and conclusions must be corrected to include contrary evidence and address the consequences of a half century of pseudoscientific claims.

Ouch. Those are quite polarizing statements. Some researchers have argued that Archer et al have overstated their case against FFQs, but agree that FFQs are far from perfect.4 This discussion isn’t new though. FFQs have been criticized for so long that they’ve spurred parody videos like the one below:

Those are some of the issues with the measurements, now onto the analyses (I’ll be using many of the arguments from my nutri epi blog post). Unfortunately, the analytical plans of studies like this are rarely registered in advance, especially because data collection for many of these studies began decades ago, and most of the analyses are often viewed as being “exploratory.” This means that there is a great deal of analytical flexibility in what covariates to include in a model.

Some of the biggest critics of this have been statistician Stan Young and meta-researcher John Ioannidis. Young wrote the following in the journal Significance.5

For example, consider the use of linear regression to adjust the risk levels of two treatments to the same background level of risk. There can be many covariates, and each set of covariates can be in or out of the model. With ten covariates, there are over 1000 possible models. Consider a maze as a metaphor for modelling (Figure 3).

The red line traces the correct path out of the maze. The path through the maze looks simple, once it is known. Returning to a linear regression model, terms can be put into and taken out of a regression model. Once you get a p‐value smaller than 0.05, the model can be frozen and the model selection justified after the fact. It is easy to justify each turn.

Patel, Burford, and Ioannidis call this flexibility and its impact on the effect size, the “vibration of effects.”6 In their study (which I’ve mentioned several times in previous blog posts), they downloaded several variables from the NHANES dataset that were linked to all-cause mortality and were able to substantially change the effect size by including certain covariates and excluding others,

Ioannidis and colleagues downloaded 13 variables from the NHANES dataset that were linked to all-cause mortality, and that had a substantial number of participants associated with each variable (at least 1000 participants and 100 deaths). From those 13 variables, they were able to produce 8,192 different statistical models that all resulted in different hazard ratios.

Their recommendations were to report all possible statistical models from all possible combinations of covariates and to report the median of all these models, rather than selectively reporting a few models. And to note whether there was something called a “Janus effect”, where the effect size would go in both directions.

In the last pattern, as exemplified by α-tocopherol, the estimated HRs can be both greater and less than the null value (HR > 1 and HR ≤ 1) depending on what adjustments were made. We call this the Janus effect after the two-headed representation of the ancient Roman god.

For α-tocopherol, most of the HR and p-values were concentrated around 1 and non-significance, respectively… The Janus effect is common: 131 (31%) of the 417 variables had their 99th percentile HR>1 and their 1st percentile HR<1.

However, statistician/epidemiologist, Sander Greenland has been a critic of meta-research and epidemiological studies that lack biochemical sophistication and lump several compounds together as if they were the same and behaved the same in the body.

“For nutrition, the lack of biochemistry sophistication among the trial designers leads to a lot of dubious and noncomparable studies, while the meta-analysts and reviewers do a lot of distortive lumping, e.g., talking about “vitamin E” as if that were a single entity.

A recent review by prominent authors didn’t even notice that almost all trials used the racemic synthetic mixture, dl-alpha-tocopheryl, (which they misidentify with “alpha tocopherol”) with vastly diferent and unjustifed dosages, and which hardly resembles the eight or so natural d-tocopherols or d-tocotrienols that account for dietary intake."6

Ioannidis and Young are also prominent critics of the fact that nutri epi studies and randomized trials have poor concordance. For example, in Young’s 2011 paper published in Significance (which Ioannidis cites often), he (Young) states the following,

We ourselves carried out an informal but comprehensive accounting of 12 randomised clinical trials that tested observational claims – see Table 1. The 12 clinical trials tested 52 observational claims. They all confirmed no claims in the direction of the observational claims. We repeat that figure: 0 out of 52. To put it another way, 100% of the observational claims failed to replicate.

In fact, five claims (9.6%) are statistically significant in the clinical trials in the opposite direction to the observational claim. To us, a false discovery rate of over 80% is potent evidence that the observational study process is not in control. The problem, which has been recognised at least since 1988, is systemic.

This seems pretty bad. I can agree with Young and Ioannidis that nutri epi studies can often produce much noise given the analytical flexibility and measurement error. However, I’m not particularly convinced by the poor concordance data for two reasons.

1. Most RCTs have designs that are likely to be underpowered for an effect size of interest. Contrast this with nutri epi studies which are often enormous and span decades. So fixating on statistical significance with this in mind makes little sense to me.

2. Statistical significance has far more utility in studies that have some random mechanism, whether that be random assignment or random sampling. Otherwise, they make little sense to fixate on, especially in nutri epi studies. Greenland, 1990 summarizes this well below.7

Randomization provides the key link between inferential statistics and causal parameters. Inferential statistics, such as P values, confidence intervals, and likelihood ratios, have very limited meaning in causal analysis when the mechanism of exposure assignment is largely unknown or is known to be nonrandom. It is my impression that such statistics are often given a weight of authority appropriate only in randomized studies

I’m reasonably confident that Greenland has changed his mind since then on the appropriate weight of authority such statistics should be given, even in randomized studies. Especially given the recent Nature commentary he published on the problems with statistical significance and dichotomous thinking.8

In the same 1990 paper, Greenland also points out the problems with several probabilistic arguments that are used to justify classical statistics in areas where the exposure is unknown or unlikely to be random. He proposes some possible solutions:

In causal analysis of observational data, valid use of inferential statistics as measures of compatibility, conflict, or support depends crucially on randomization assumptions about which we are at best agnostic and more usually doubtful. Among the possible remedies are:

• Restrain our interpretation of classical statistics by explicating and criticizing any randomization assumptions that are necessary for probabilistic interpretations;
• train our students and retrain ourselves to focus on nonprobabilistic interpretations of inferential statistics;
• deemphasize inferential statistics in favor of pure data descriptors, such as graphs and tables;
• expand our analytic repertoire to include more elaborate techniques that depend on assumptions in the “agnostic” rather than the “doubtful” realm, and subject the results of these techniques to influence and sensitivity analysis.

These are neither mutually exclusive nor exhaustive possibilities, but I think any one of them would constitute an improvement over much of what we have done in the past.

Now, with all of these nuances in mind, how convincing is this observational study that has produced a hazard ratio of 1.17 with a 95% CI ranging from 1.09 - 1.26? I don’t think it’s that convincing. I’m not saying that dietary cholesterol or eggs don’t impact CVD incidence and serum cholesterol, but just that I don’t find this study’s data to be very convincing.

Perry Wilson also has an interesting argument where he utilized graphical methods to assess the plausibility of the results,

## Does Dietary Cholesterol Impact Serum Cholesterol?

This has been a very controversial subject ever since dietary cholesterol was first implicated in cardiovascular disease in the 1900s with one of the most famous epidemiological studies being the Seven Countries Study.9 I’m not going to touch on that study since I wouldn’t be doing it justice, but there’s plenty of reading material out there.

But here’s a very brief timeline of dietary cholesterol:

## Brief Timeline of Dietary Cholesterol

• 1950s | Dietary cholesterol is implicated in heart disease.10
• Late 1900s | Seven Countries Study is published.9
• Late 1900s | Metabolic ward experiments find that dietary cholesterol impacts serum cholesterol.11
• Late 1900s | Other trials fail to find an effect.12
• Late 1900s - Early 2000s | Government agencies and medical organization recommend limiting dietary cholesterol.13
• 2013 | American Heart Association states that there’s insufficient evidence whether removing cholesterol is beneficial.14
• 2015 | Dietary guidelines drop arbitrary recommendations on limiting dietary cholesterol, but still recommend little intake.

Wow, that can be pretty confusing and frustrating… Wait till you see the next study I talk about.

## What Do Reviews of Randomized Trials Say?

A few months ago, a meta-regression of only randomized controlled trials that were at least two weeks long (and a million other inclusion/exclusion criteria) was published in The American Journal of Clinical Nutrition.15 This study was interested in assessing the relationship between dietary cholesterol and serum cholesterol. Here are some characteristics of the study:

• Included 55 randomized trials (2,652 subjects), with 44 being crossover trials.
• Used Bayesian methods with uninformed priors.
• Used the Mensink equation to subtract the impact of fatty acids on cholesterol to isolate the impact of dietary cholesterol better.
• Explored both linear and nonlinear models.

So what were the results?

The nonlinear models better fit the data than the linear model and at 100 mg/d of dietary cholesterol change, there was an associated 4.58 mg/dL change in LDL cholesterol. The linear model estimated this change to be 1.90 mg/dL, which is similar to estimates found by the meta-analysis of metabolic ward studies.11 At 200-400 mg/d of dietary cholesterol change, the nonlinear and linear models had substantial differences in their estimations (see the table below and then the graph).

Wow. So dietary cholesterol does indeed impact serum cholesterol (LDL)! Here’s what the authors write in their conclusion…

The meta-regression analyses presented herein indicate that, after providing a theoretical control for intakes of SFAs, MUFAs, PUFAs, and when possible, TFAs, there is a positive, dose-related trend between changes in cholesterol intake and changes in circulating LDL cholesterol, a finding that is consistent with the published literature.

The nonlinear dose-response shapes (MM and Hill) best fit the data. For a 100-mg/d increase in dietary cholesterol intake, circulating LDL-cholesterol concentrations are predicted to increase by 1.90 mg/dL (linear model), 4.46 mg/dL (MM model), or 4.58 mg/dL (Hill model).

Then they discuss the practical significance of this.

The results from these meta-analyses have potential implications for dietary recommendations. For comparison, the equation given by Mensink et al. predicts a change of 1.23 mg/dL for each 1% increase in SFAs in exchange for carbohydrate.

Therefore, based on the nonlinear models, increasing dietary cholesterol by 100 mg/d would be predicted to have an effect comparable to that of increasing dietary SFAs by 3.7%, and increasing dietary cholesterol by 200 mg/d is predicted to be comparable to increasing dietary SFAs by 5.5%.

A large egg contains ∼185 mg cholesterol and would therefore be expected to increase the LDL-cholesterol concentration by 6.5 mg/dL based on the nonlinear models. However, the majority of dietary cholesterol is not attributable to egg intake in the United States, except for individuals in the highest quartile of TC intake.

According to a recent publication reporting dietary sources of cholesterol in US adults aged ≥20 y based on NHANES 2013–2014 data, mean dietary cholesterol intake was 293 mg/d and the primary dietary cholesterol source in the overall population was meat (defined as poultry, mixed dishes, red meat, processed meat, and seafood), which accounted for 42% of dietary cholesterol.

## Wait, so Should We Be Worried About Eggs?

I’m going to argue that we probably shouldn’t. The reason is that just because a food contains dietary cholesterol doesn’t mean it will be absorbed efficiently in the gastrointestinal tract. Foods are not homogenous and do not contain the same nutrients, anti-nutrients, and minerals. Consuming eggs in a sitting is not the same as consuming pure dietary cholesterol nor is it the same as consuming other foods high in cholesterol (poultry, red meat, and processed meat).

This also seems to be supported by a few randomized crossover trials that have shown that the cholesterol in eggs doesn’t seem to be well absorbed.16 Several other trials have also failed to show an effect on serum cholesterol.12,17,18 Is it possible that these trials were just underpowered in the first place? Very possible.

This all needs to be taken into account with the fact that in addition to the randomized trials included in the meta-regression mentioned above,15 there exists a plethora of metabolic ward studies suggesting an effect of dietary cholesterol on serum cholesterol.11 However, many of the metabolic ward trials included in the prominent meta-analysis by Clarke et al11 were also nonrandomized,

Solid food diets were assessed in 72 of these reports among 129 groups of subjects in 395 experiments with various designs (109 randomised crossover, 57 randomised or matched parallel, 77 non-randomised Latin square, and 152 non-randomised sequential).

So while there is plenty of evidence suggesting that a diet high in dietary cholesterol impacts serum cholesterol, the evidence to suggest that eggs in particular have this effect seems far less concrete. There also may be a number of nutrients found in eggs that inhibit cholesterol absorption.19,20 Here are some excerpts from Kuang et al20

Yang et al. reported that dietary PUFAs, PC, and SM significantly inhibit the uptake of cholesterol in Caco-2 monolayer, which may have potential therapeutic effect on reducing cholesterol absorption as functional food ingredients [44, 45]….

Ezetimibe reduces the absorption of cholesterol in the small intestine by inhibiting the activity of NPC1L1 protein [52].

Therefore, consuming eggs may inhibit the intestinal absorption of cholesterol via the phospholipids in eggs. A fundamental question is how this biological effect is regulated.

How convincing is this evidence? I’m still not sure. Much of it seems to be very preliminary, but I’m leaning towards the idea that most people probably don’t have to worry about eggs, especially since Americans don’t even get most of their dietary cholesterol from eggs.15 I don’t find the existing data for eggs (including those from this study) to be convincing. More better research is needed.

We need less research, better research, and research done for the right reasons.21 - Doug Altman

See my other articles on nutritional epidemiology:

### References

1. Zhong VW, Horn LV, Cornelis MC, et al. Associations of Dietary Cholesterol or Egg Consumption With Incident Cardiovascular Disease and Mortality. JAMA. 2019;321(11):1081-1095. doi:10.1001/jama.2019.1572

2. Keogh RH, White IR. A toolkit for measurement error correction, with a focus on nutritional epidemiology. Statistics in Medicine. 2014;33(12):2137-2155. doi:10.1002/sim.6095

3. Archer E, Marlow ML, Lavie CJ. Controversy and debate: Memory-Based Methods Paper 1: The fatal flaws of food frequency questionnaires and other memory-based dietary assessment methods. Journal of Clinical Epidemiology. 2018;104:113-124. doi:10.1016/j.jclinepi.2018.08.003

4. Subar AF, Freedman LS, Tooze JA, et al. Addressing Current Criticism Regarding the Value of Self-Report Dietary Data12. The Journal of Nutrition. 2015;145(12):2639-2645. doi:10.3945/jn.115.219634

5. Young SS, Karr A. Deming, data and observational studies. Significance. 2011;8(3):116-120. doi:10.1111/j.1740-9713.2011.00506.x

6. Patel CJ, Burford B, Ioannidis JPA. Assessment of vibration of effects due to model specification can demonstrate the instability of observational associations. J Clin Epidemiol. 2015;68(9):1046-1058. doi:10.1016/j.jclinepi.2015.05.029

7. Greenland S. Randomization, statistics, and causal inference. Epidemiology. 1990;1(6):421-429.

8. Amrhein V, Greenland S, McShane B. Scientists rise up against statistical significance. Nature. 2019;567(7748):305. doi:10.1038/d41586-019-00857-9

9. Feinleib M. Seven Countries: A Multivariate Analysis of Death and Coronary Heart Disease. JAMA. 1981;245(5):511-512. doi:10.1001/jama.1981.03310300063026

10. Duff GL, McMILLAN GC. Pathology of atherosclerosis. The American Journal of Medicine. 1951;11(1):92-108.

11. Clarke R, Frost C, Collins R, Appleby P, Peto R. Dietary lipids and blood cholesterol: Quantitative meta-analysis of metabolic ward studies. BMJ (Clinical research ed). 1997;314(7074):112-117.

12. Bowman MP, Van Doren J, Taper LJ, Thye FW, Ritchey SJ. Effect of dietary fat and cholesterol on plasma lipids and lipoprotein fractions in normolipidemic men. The Journal of Nutrition. 1988;118(5):555-560. doi:10.1093/jn/118.5.555

13. Brownawell AM, Falk MC. Cholesterol: Where science and public health policy intersect. Nutrition Reviews. 2010;68(6):355-364. doi:10.1111/j.1753-4887.2010.00294.x

14. Eckel RH, Jakicic JM, Ard JD, et al. 2013 AHA/ACC guideline on lifestyle management to reduce cardiovascular risk: A report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines. Journal of the American College of Cardiology. 2014;63(25 Pt B):2960-2984. doi:10.1016/j.jacc.2013.11.003

15. Vincent MJ, Allen B, Palacios OM, Haber LT, Maki KC. Meta-regression analysis of the effects of dietary cholesterol intake on LDL and HDL cholesterol. Am J Clin Nutr. 2019;109(1):7-16. doi:10.1093/ajcn/nqy273

16. Kim JE, Campbell WW. Dietary Cholesterol Contained in Whole Eggs Is Not Well Absorbed and Does Not Acutely Affect Plasma Total Cholesterol Concentration in Men and Women: Results from 2 Randomized Controlled Crossover Studies. Nutrients. 2018;10(9). doi:10.3390/nu10091272

17. Chenoweth W, Ullmann M, Simpson R, Leveille G. Influence of dietary cholesterol and fat on serum lipids in men. The Journal of Nutrition. 1981;111(12):2069-2080. doi:10.1093/jn/111.12.2069

18. Vorster HH, Benad’e AJ, Barnard HC, et al. Egg intake does not change plasma lipoprotein and coagulation profiles. The American Journal of Clinical Nutrition. 1992;55(2):400-410. doi:10.1093/ajcn/55.2.400

19. Jesch ED, Carr TP. Food Ingredients That Inhibit Cholesterol Absorption. Preventive Nutrition and Food Science. 2017;22(2):67-80. doi:10.3746/pnf.2017.22.2.67

20. Kuang H, Yang F, Zhang Y, Wang T, Chen G. The Impact of Egg Nutrient Composition and Its Consumption on Cholesterol Homeostasis. Cholesterol. 2018. doi:10.1155/2018/6303810

21. Altman DG. The scandal of poor medical research. BMJ. 1994;308(6924):283-284. doi:10.1136/bmj.308.6924.283

• Cite this blog post