Surviving epidemiological whiplash: Fruits and vegetables and colorectal cancer

Do fruits and vegetables protect from colorectal cancer? Last year, we heard that fruits and vegetables were not linked to colon cancer and that risks were similar between men and women. A Harvard study had been unable to find any tenable correlations among data of 756,217 men and women. This month, we heard that fruits and vegetables may protect men from colorectal cancer, but not women.

What we have here is another case of epidemiological whiplash.

This study, published in the current issue of the American Journal of Clinical Nutrition was led by Dr. Abraham M.Y. Nomura, M.D., Ph.D., from the Epidemiology Program at the University of Hawaii in Honolulu. These researchers looked for correlations between fruit, vegetable and grain intakes and cases of colorectal cancer among the nearly 200,000 Multi-ethnic Cohort Study of Diet and Cancer participants.

Our first clue that this study was done in a computer and unable to provide us with any meaningful clinical information was the large Trojan number of nearly 200,000 participants. The second was the fact that the authors were looking for associations, which are irrelevant to us until actual clinical studies test any hypotheses generated. As we know, correlations, no matter how significant they might seem, don’t necessarily mean anything. When culling through large databanks, the chances of finding unusual correlations become highly probable. Based on the principles of probability and statistics, they’re best thought of as coincidences. Even remarkable and huge ones aren’t at all uncommon or extraordinary.

Overview of methodology and data

The Multi-ethnic Cohort Study databank used in this study had been created between 1993 and 1996, when more than 215,000 men and women of mostly African-American, Japanese, Latino, Native Hawaiian and Caucasian ethnicity living in Hawaii and Los Angeles had completed a 26-page mail-in questionnaire about their diet history, demographics, medical history and lifestyles.

For this analysis, the researchers used data from 191,011 of the cohort, restricting it to those five major ethnicities and excluding data on another 8,265 participants because of implausible dietary information. Over the next seven years, through 2001, 2,100 cases of colon and rectal cancers were identified via the Hawaii Tumor Registry, the Cancer Surveillance Program for Los Angeles County, and the California State Registry, along with the death certificates on file in Hawaii and California.

According to the raw demographic data, there were no statistical differences in BMIs, current cigarette smoking, red meat intake, vigorous physical activity per week, or aspirin use among either men or women and cases of colorectal cancer. The most significant differences between those who were later diagnosed with cancer and those who weren’t were age and family history of colorectal cancer. Compared to the remaining cohort, the cancer patients had been about 5 years older at the start of the study and had been more than half again as likely to have had a family member with colorectal cancer.

The dietary database for the Multi-ethnic Cohort Study had been created from self-administered food frequency questionnaires which had asked each participant how often they consumed foods in 8-9 categories and to select one of three portion sizes for each food item. The authors used food composition tables and a large recipe database from the Cancer Research Center of Hawaii to calculate intakes for specific food groups. To calibrate their estimates of weights and calorie counts, the authors had randomly selected 1,606 study participants to complete three 24-hour dietary recalls administered over the phone during a 3-month period and additional food frequency questionnaires.

For this analysis, the authors divided estimated food intakes for fruits, vegetables and grains into quintiles, with Q1 the lowest fifth of consumption and Q5 the highest. The authors then let Cox proportional hazards computer models calculate the relative risks between the different quintiles of each food group and the cases of colon and rectal cancers. The relative risks were adjusted in their multivariate model for age, ethnicity, family history, history of colorectal polyp, cigarette smoking, BMI, physical activity, aspirin use, vitamin use, hormone replacement therapy, caloric intake, alcohol, red meat, folate, vitamin D and calcium intakes.

Findings

The authors stated: “In women, none of the associations with vegetables, fruit, or vegetables and fruit combined were significant. Grain intake was not associated with colorectal cancer for either men or women.”

For men, they reported a 26% lower relative risk of colon and rectal cancers between the highest quintile and lowest quintile for vegetables and fruits combined, and a 20% lower relative risk associated with fruit intake. A 15% lower relative risk associated with vegetables alone was of borderline statistical significance. Looking at rectal and colon cancers separately, none of the correlations with rectal cancer were significant.

In other words, these results were null. None of the relative risks were even tenable for computer-derived correlations. Relative risks under at least 200% - 300% are considered no better than what could come up a random error or chance, and are most likely explained by co-factors.

“What we have here is a failure to communicate”

But, did you catch what wasn’t reported?

Here’s the trick: Reporting just the relative risks between the top and bottom quintiles of fruit and vegetable consumptions leaves out all of the other consumption levels in between.

The reported inverse relationship to colorectal cancer between eating the highest amounts of fruits and vegetables and lowest was not linear. The correlations among the various quintiles for each type of produce looked like a see-saw, up and down at each amount consumed. In other words, there was no dose-response. Relative risks for cancer didn’t steadily go down as the amounts of fruits and/or vegetables went up.

So, for example, as the consumption of vegetables went up with each quintile among the men, the relative risks for colorectal cancer went up to 7% higher, then down to a 14% lower relative risk, then up to 9% lower risk, and then down to a 15% lower risk. The relative risks all hugged either side of 1 (null).

Looking just at colon cancer among the men, fruits and grains didn’t correlate at all, as the authors reported. And the correlations with vegetable intakes did the roller coaster starting at no intake: going up to 3% higher relative risk, to 23% lower risk, up to 14% lower risk and down to 20% lower risk.

In other words, men consuming the highest amounts of vegetables could “lower” their risks for colon cancer by cutting their vegetable consumption in half! That’s not quite the results being reported in the news.

“You want fries with that?”

To help bring perspective to any lingering concerns that these correlations are clinically meaningful, look at the actual amounts of vegetables consumed at the lowest and highest quintiles. Among men, the lowest veggie consumption group ate about 2.5 ounces/1,000 kcal and those in the highest group ate about 8.3 ounces/1,000 kcal. The range among women was 3 ounces to 10 ounces/1,000 kcal.

Total average daily intakes* were about 10 ounces each for men and women, with a 3.4-fold difference in vegetable intake between the highest and lowest groups, according to the authors. In other words, the men eating the fewest veggies ate about 4.5 ounces and the most about 15.3 ounces per day — the difference of a potato and a few slices of tomato!**

Of course, we didn’t hear headlines claiming an order of fries could lower men’s risk for colorectal cancer! Why, that would have made no sense. :-)

The range of fruit intake among this cohort was wider, with a 9.8-fold difference: men’s intake varied from 1 to 10.4 ounces and women from 1.6 to 13.4 ounces per day. Yet, the authors still found no tenable correlations with colorectal cancers.

Some want to believe that there are special cancer-fighting types of fruits or vegetables. To investigate this possibility, the authors next looked at correlations between colorectal cancer and types of fruits and vegetables at each quintile of consumption. They were unable to find any statistical differences in risks associated with any particular type of veggie or fruit among the men or women at any level of intake.

Computer games. To illustrate the often nonsensical correlations that these types of studies can find and how easy it is to pull out individual correlations and try to suggest they mean something, here are a few examples of other untenable correlations to colorectal cancer this study found — all of which contradict the belief that more produce is better:

● a 10% higher relative risk associated with men eating 20% more light green vegetables or broccoli compared to the least

● an 18% higher relative risk associated with men eating the highest amounts of citrus fruit compared to those eating 1/5 less

● a 4% higher relative risk associated with women eating the highest amounts of citrus and light green vegetables compared to those eating none at all — but intakes cut by a fifth were associated with relative risks that were 8% higher for citrus and 21% lower for light green vegetables

Despite popular beliefs surrounding wholegrains, the authors reported: “The consumption of grains, which included both refined and whole grains, was not related to colorectal cancer in our cohort. Previously, we found no association with fiber from grains and therefore did not examine whole grains separately in the current analyses… Grain intake was not associated with colorectal cancer for either men or women.”

Yet, even pulling out one of those correlations with grains could be made to sound really scary to consumers who don’t understand tenable correlations. For example, there was a 45% higher relative risk associated with rectal cancer among women eating 20% more grains than none at all (of course, then the relative risks went down, then up, then down again).

In the scheme of things

The bottom line, this National Cancer Institute-funded study found no tenable correlation between fruits, vegetables or grains eaten by these Americans and colorectal cancer. As the study authors noted, null findings of an association concur with the bulk of the epidemiological research — including the Health Professionals’ Follow-up Study, the Nurses’ Health Study, the Women’s Study, and the Japan Health Center study — all of which found no association between vegetable or fruit consumption with colorectal cancer in men and women.

Past cohort studies also haven’t supported an inverse relationship between fruit intake and colorectal cancer, they noted. “There was no association between fruit intake and colorectal cancer among men in six studies… Six other cohort studies also reported no relationship between fruit and colorectal cancer among men…Six other cohort studies also reported no relation between fruit and colorectal cancer among women.” The only contrary studies, said the authors, were two other epidemiological studies (the Swedish Mammography Screening Cohort Study and the Cancer Prevention Study II). Both of those, however, had reported untenable relative risks also comparing the highest quintile of fruit intake with the lowest.

Enjoying a varied diet continues to be our best defense to help prevent nutritional deficiencies and make eating loads more fun. But, popular claims in media — even in medical continuing education materials offered by Medscape — of special benefits of produce to protect against colorectal cancers continue to be unsupported in the medical literature.

Only you can protect yourself from epidemiological whiplash.

© 2008 Sandy Szwarc

* “The median daily intake of vegetables and fruit was 293 grams [10.3 ounces] and 259 grams [9.1 ounces], respectively, for men; and 286 grams [10 ounces] and 298 grams [10.5 ounces], respectively, for women. In addition, the range of intake of these foods was wide. There was a 3.4-fold difference in vegetable intake to a 9.8-fold difference in fruit intake between the lowest and highest quintile groups.”

** A medium tomato or zucchini weighs about 5 ounces, a carrot about 3 ounces, and medium sweet potato or potato weighs around 8 ounces.