We have all fallen victim to it. The latest and greatest epidemiologic study reveals that coffee prevents cancer only to be replaced by a similar study days later showing that coffee causes cancer. When it supports our coffee habit, we email the news release to our friends, family members, and coworkers. When a newer study reveals that coffee causes cancer, we simply hope it’s not true. In medical school, I used to keep track of which substances may harm or hurt us based on these studies. I quickly abandoned that practice when I realized that nearly all foods could kill us or cure us of all ailments depending on the study. Diet studies and cancer were all over the place, and the issues seem to be worsening with each passing moment. Expansion of internet reach, coupled with the media’s love of promoting “latest and greatest” studies only worsened things.
So how do we deal with the endless ups and downs of these studies? Below are the lessons I have learned while dealing with the explosion of diet studies and cancer over the past decade.
Diet Studies and Cancer Lesson 1: Dealing with Epidemiologic Studies
Epidemiologic studies are important. They analyze patterns of disease in large groups of people and expose potential connections. They show correlations between certain behaviors and the occurrence of cancer, leaving us with testable hypotheses to ascertain which lifestyle characteristics can help reduce the risk of cancer. A famous example is the handful of studies linking cigarette smoke and lung cancer.1 One of the most famous of these population studies by Doll and Hill revealed that those who smoked 35 or more cigarettes per day experienced a 40 times higher risk of dying from lung cancer (that is 40 times higher, not 40% higher). At that point, it was considered “beyond a reasonable doubt” that smoking caused lung cancer.
However, correlation does not equal causation. We have all heard this comment dozens of times, but it is a disclaimer that should be repeated before most studies are reported. We all get caught in the trap of overstating the importance of correlation studies, especially when they support our views.
There are a couple of issues here. First and foremost, when these studies tell us our risk of cancer based on the studied activity, we must be careful. For instance, while our risk of dying from lung cancer is 40 times higher if we smoke, our absolute risk of getting lung cancer if we smoke is still about 16% (16 out of 100 smokers are diagnosed). However, because the rate of lung cancer in the average nonsmoker is so low, a change to 16% is massive, and as much as a 3,000% increase for heavy smokers. Do not misunderstand me – a 16% risk of the most common fatal malignancy is still very worrisome and certainly one of the million reasons to never smoke.
However, the link between smoking and lung cancer is one of the strongest we have ever seen in epidemiologic studies. Fast forwarding a half century, nearly all the recent epidemiologic studies published reveal links that are miniscule compared to the tobacco studies. For instance, a recent analysis of several studies assessing meat consumption showed that while red meat was not associated with pancreatic cancer, processed meat was associated with the deadly malignancy.2 Yet, the data was almost statistically insignificant as it revealed a 19% increase in risk of pancreatic cancer in those that ate over 50g per day.
That 19% sounds large and is bigger than the 16% of smokers that get lung cancer. But upon looking closer, that 19% increase is in relative risk of pancreatic cancer, not overall risk. The lifetime risk of pancreatic cancer is 1.5% (1 in 67 men). In other words, if this correlation were in fact true, pancreatic risk would go from 1.5% to 1.79%, a tiny fraction of the increased risk seen with lung cancer and smoking. This increased risk is so small, that studies like this provide little, if any, strong takeaways. Yet, the results of this study were reported on nearly every nightly news show, newspaper, and online source with headings like “Red Meat is as Bad as Smoking.” The actual findings may have been weak, but the reach and scare tactics were certainly not.
Correlation does not equal causation, and really, really small correlations more often than not leave us with little more than annoying media coverage and news stories.
Diet Studies and Cancer Lesson 2: Avoiding Jumping to Conclusions
In the movie Office Space, Tom reveals his brilliant idea of manufacturing a “jump to conclusions mat” to his coworkers. This idea, which he thought would make him millions, would be labeled with different conclusions that the user could jump to. The other office characters respond that this is the worst idea they have ever heard.
Jumping ahead several decades, with each new population study published, it appears we are continually jumping to conclusions. Again, these studies only reveal associations, stimulate hypotheses, and yield no certain truths, and jumping to conclusions, in this instance, is just as bad as Tom’s idea.
Gary Taubes, author of the famous Good Calories, Bad Calories, has used the following example many times throughout his writing, but it serves as a valuable lesson of the benefits and dangers of jumping to conclusions from epidemiologic studies. In 1985, a massive epidemiologic study of 122,000 women was published out of Harvard revealing that those who took hormone replacement therapy (HRT) experienced a decrease in coronary artery disease. Specifically, those women who were currently taking HRT had a 70% decreased risk, while current and former users experienced half the risk.3 Tobacco, diabetes, hypertension, cholesterol levels, parental heart attacks, oral contraceptives, and obesity were adjusted for by the statisticians and the association held. It was final, HRT cut the risk of coronary artery disease.
The results of this study were met with open arms. HRT helped women feel substantially better during and after menopause, and this newfound benefit only strengthened their desire for their physicians to prescribe them these life-saving meds. After decades of prescribing hormones for postmenopausal women, HRT was eventually tested against a placebo in healthy women – these results painted a drastically different picture.
After 6.8 years of the study taking place at 20 US clinical centers, our worst nightmare had come true. Not only did HRT not lower the risk of coronary artery disease, it increased it. In fact, it increased the risk of many illnesses, including heart disease by 30%, breast cancer by 26%, stroke by 40%, blood clots in the lung by 113%, and total cardiovascular disease by 22%.4
So, what happened?
While the study statisticians used all the available data to equate the groups of women who did and did not use HRT, they could not account for some missing information, a fatal flaw of these types of studies. The woman who were taking HRT were generally healthier – they smoked less, exercised more, ate better, were wealthier, were more educated, and were more proactive. The last may have been one of the key issues, as this proactive nature led them to frequent their doctor’s office more often, asking to be prescribed HRT. The massive initial study simply found that healthier women used HRT more often, not the other way around. Their health was not from their HRT, but their HRT was a product of their good health. What had happened is known as the healthy user bias, or in other words the healthier women were using the medication – HRT in this case – and it had nothing to do with their health.
Diet Studies and Cancer Lesson 3: Isolating One Aspect of Food in a Pill in Mega Doses Rarely Works
Fruits and vegetables are associated with many health benefits, including a decreased risk of cancer.5 So what is it about fruits and vegetables that make them so healthy? The potential benefits are plenty, including their bowel bacteria-feeding fiber, vitamins, minerals, and chemicals that stimulate our immune system. What distinguishes plants from well-sourced animals (which contain hefty amounts of vitamins and minerals) is their fiber and chemicals that help in the battle against free radicals. As excess free radicals can cause cancer, the topic of antioxidants dominates most discussions on diet studies and cancer.
For instance, some fruits and vegetables contain a significant amount of beta-carotene. When scientists attempt to calculate how much beta-carotene a group of people is consuming by handing out surveys and questionnaires (yes, there many issues with this method), they find that it reduces the risk of breast cancer by 18%.5 Based on the information above, we know that this decrease may not be that significant for each woman eating vegetables (breast cancer rates are 12.5%, thus this would lower the overall risk to about 10%), but can be large based on an entire population. Studies like this may suggest that fruits and vegetables are healthy, but can it narrow down this benefit to beta-carotene or any one nutrient?
Dating back several decades, a handful of other studies revealed similar findings – dietary beta-carotene appears to reduce the risk of several different types of cancer.6 Studies even showed that increases in retinol in the blood (from the conversion of beta-carotene) may reduce the risk of cancer.
The next step seemed like a slam dunk: isolate beta-carotene, put mega doses of it in a pill, and test it in a clinical trial. Before we dig too deep into this topic, let’s take a step back and discuss some of the benefits of eating fruits and vegetables that have traditionally been associated with several beneficial antioxidant substances found in our foods.
Diet Studies and Cancer – The Problem with Fruits, Vegetables, and Antioxidants
The interwoven relationship of typical antioxidants is confusing, and even a basic understanding of them helps establish the issues with studies that test their benefits outside of food. Some of the different substances in our plant sources, their exact benefits, and even the terminology to describe them is less than clear. As described above, people who eat fruits and vegetables are generally healthy, but what is it about the plants that fuel our health?
The most commonly described healthy substances in fruits and vegetables include:
- Phytonutrients – A broad term for plant-based chemicals that serve as toxins to threats like insects, fungus, and bacteria. There are tens of thousands of different phytonutrients (or phytochemicals) found in different plants, fruits, and vegetables and they are often responsible for their bright color. They include ellagic acid, carotenoids like beta-carotene, flavonoids, glucosinolates, and the recently famous resveratrol. Besides the vitamins that some of them provide (for instance, carotenes are converted to vitamin A), they stimulate our cells to fight disease.
- Flavonols – Phytonutrients that are flavonols, like anthocyanins, quercetin, catechins, and epicatechins are known as tannins, and they give tea and wine their dark color and bitter taste (wine aficionados call this complexity). They are also found in spices and berries and upon consumption, they upregulate our innate antioxidant systems like glutathione and stimulate our immune system. Studies are mixed, but they have been shown to fight inflammation, heart disease, and cancer.
- Beta-Carotene – Another substance that gives food its color, beta-carotene is the precursor to vitamin A. It causes carrots to look orange and is used as food coloring. If we eat too much carotene-rich foods, our skin can develop an orange hue. Beta-carotene is considered a potent antioxidant and has been studied extensively.
- Sulfur – My affinity for sulfur and those cruciferous vegetables where it is found is well known, as I have written about it plenty. Cruciferous vegetables, like broccoli and Brussels sprouts, contain significant amounts of sulfur-containing glucosides, which can stop cancer. The bacteria within our gastrointestinal tract also break down glucosinolates to isocythianates and sulforaphane, which help defend the body from cancer. The sulfur stresses our cells and turns on several anticancer pathways.
Many of these antioxidant benefits are the same benefits seen is spices, which is no coincidence as they often provide distinct flavors and colors of the spices.
The key with many of these compounds is that they stress our cells. In the past, many of these benefits would be referred to as antioxidant value or ORAC, while in actuality these substances act like potential toxins to our cells and stimulate them to mount a defensive front, much like a military drill prepares a unit for battle without any actual fighting. This is where some of the confusion begins when we try to extract the antioxidant benefits from fruits, vegetables, and spices, replicate them in a lab, and stuff them into a convenient pill that can be swallowed with water. I dedicated an entire chapter to just some of these issues in Misguided Medicine, but this could be an entire book by itself.
Due to the potential benefits in epidemiologic studies of the “antioxidants” from fruits and vegetables listed above, multiple clinical trials have assessed them in pill form. Some of the most famous of these studies include:
- Beta-Carotene – As discussed above, the potential for beta-carotene in reducing cancer risk was immense. The outcome, however, left us with a lot to be desired. In a massive study from Finland with over 29,000 male smokers, those that were randomized to 20mg of beta carotene were found to have higher rates of lung cancer and heart attacks, and their risk of dying since taking the supplement increased by 8%. An analysis of all studies reveals absolutely no reduction in overall cancer risk with beta-carotene supplementation, and an increase in lung and gastric cancer risk.7
- Vitamin E – Much like beta-carotene, the vitamin E trials revealed more harm than benefit. Nine of 11 trials assessing high-dose vitamin E showed that it increased an individual’s risk of death.8 The massive SELECT trial included over 35,000 men and revealed that supplementation with Vitamin E increased prostate cancer risk in those men that were taking 400 IU.9 Many claim the scientists used the wrong type of vitamin E (alpha-tocopherol, which some consider a more unnatural version), while others claim that the oil in these supplements – much like the inherent issues with vegetable oils and polyunsaturated fats – was rancid and contained more free radicals than antioxidants. Perhaps vitamin E in its natural food form holds potential that cannot be captured in a pill, and this was what produced a benefit in the epidemiologic studies.
- Antioxidant Supplements – A large analysis of 22 randomized trials consisting of 161,000 participants revealed no benefit of antioxidant supplements including vitamins A, C, and E, and beta-carotene and selenium in lowering the risk of prostate cancer.10 Some data even reveal that supplementing with antioxidants after exercise blocks many of the benefits of a workout.11
These studies were well-intentioned, and the point here is not that they failed – basing studies on epidemiologic findings are complicated and more likely to fail than succeed (there are infinite wrong answers). The point is that testable hypotheses that were once promoted as truth often fail when tested. When we prematurely jump to conclusions from these original epidemiologic studies before testing them, we will often find ourselves promoting an incorrect message, especially if it is from a “latest and greatest” study finding that defies all logic.
The bad news is simple: taking one compound naturally found amid several and trying to isolate its health benefit in pill form is flawed from the start and has failed us dozens if not hundreds of times. The issues with this approach are many, and go broadly beyond the obvious inability to isolate and amplify benefits that we have trouble even defining.12 The good news, however, yields a more powerful message. Fruits and vegetables have a plethora of potential health benefits and we need not try to isolate one in pill form; we can simply eat a variety of berries and green leafy vegetables to capture all their combined benefits.
Diet Studies and Cancer Lesson 4: All Foods Cause and Prevent Cancer
When all else fails, keep in mind that all foods cause and prevent cancer. For instance, Drs. Jonathan Schoenfeld and John Ioaniddis reviewed the top 50 ingredients from random recipes included in The Boston Cooking-School Cook Book.13 They found that depending on the study, we could basically find support for any food or ingredient to cause or cure cancer. Ioaniddis, who has been critical of the scientific process, exposed the ridiculousness of many of these epidemiological publications by beating them at their own tricks.
They also found that many of the “positive” studies revealed a weak or nominally significant increase or decrease in cancer based on the food. In total, over 70% of the studies described results that were insignificant or weak. These same weakly significant studies are often the ones the media jumps on to inform us of the next food that has been killing us.
He also found that studies often compare extremes to find their conclusions. For instance, drinking five cups of coffee per day is associated with a small risk when compared to less than one cup, but the study results are extrapolated and we are warned that any amount of coffee may lead to cancer.
And this assessment only came from published studies. It is foreseeable that plenty of similar studies revealed insignificant results, thus were never accepted for publication (although there is a Journal of Negative Results, negative studies most often go unpublished as they create no buzz for journals). Along these lines, the study authors found that basically the higher “quality” the studies were, the higher chance of showing no relationship between a food and cancer.
The result? Plenty of poorly done studies exist to support promotion of demonization of a food when it comes to cancer, so we can take our pick. However, a more prudent response is to question the study, and perhaps more so, the source promoting its findings.
Each data point represents the results of a study. Any point to the left of the vertical line at “1” represents a study the reduces risk, while anything to the right increases the risk of cancer. The top section represents each cancer type, while the bottom represents specific foods. From Jonathan D. Schoenfeld and John P.A. Ioannidis Am J Clin Nutr 2012.13
Diet Studies and Cancer – Where Do We Go from Here?
Looking at the picture of the raspberry or spices from above, picking any one of the many potential benefits and isolating them to find a large benefit is difficult, if not impossible. Looking at the picture of the healthy women, isolating her HRT usage in this instance and extrapolating to an entire population led to significant harm. Even the best of statisticians does not have the capability to account for the inherent issues with nearly all the datasets used for these epidemiologic studies. Findings can be interesting and stimulate further discussion and studies, but when we count them as good as gold, we get in trouble.
Even those flawless fruits and vegetables have been shown to be unsuccessful at lowering cancer risk in a massive study out of Harvard assessing over 100,000 people.14 So what is the moral of the story? With the large amount of studies these days with findings all over the map, support for nearly anything can be found. For instance, did you know that being a male surgeon is associated with significantly better looks than being a male physician?15 I wish I had known that before choosing oncology as, according to these data at least, a career in surgery would have made me better looking.
Hormone replacement therapy, beta-carotene, and antioxidants are only some of the many studies revealing opposite findings when rigorously tested in randomized trials. Yet, moving forward, we must go with all available information to make informed decisions. I personally use studies (yes, even epidemiologic studies), human history, evolutionary medicine,16 common sense, and applicability for my decisions and health recommendations. For instance, if advice is impossible to be followed, regardless of how good it is, it is useless.
Simply put, I ask myself three questions:
- Does this information fit with our past?
- Is there biologic/scientific/physiologic and/or randomized study support for the findings?
- Can we do it?
It is not foolproof, but it works for me – and it works much better than my local news source.