The nut study -- the good, the bad, and the ugly

We take rare pleasure in praising an Associated Press piece about a paper just out in the New England Journal of Medicine ("Association of Nut Consumption with Total and Cause-Specific Mortality", Bao et al., NEJM 2013 369: 2001-2011).  The paper reports a study showing that risk of death from all causes is reduced among study participants who frequently eat nuts and author Marilynn Marchione does a nice job pointing out the caveats (well, apart from the suggestion that nut eaters are less likely to die).

Why might nuts be protective of your health?  Boa et al. note that "Nuts are nutrient-dense foods that are rich in unsaturated fatty acids, fiber, vitamins, minerals, and many other bioactive substances, such as phenolic antioxidants and phytosterols."

Boa et al. used data from the longstanding Nurses Health Study and the Health Professionals Follow-up Study, looking at the dietary intake and disease history of a total of 119,000 people. Dietary intake was assessed with food-frequency questionnaires sent to study participants every 2 to 4 years.  And the primary endpoint was death from any cause. 

In two large prospective U.S. cohorts, we found a significant, dose-dependent inverse association between nut consumption and total mortality, after adjusting for potential confounders. As compared with participants who did not eat nuts, those who consumed nuts seven or more times per week had a 20% lower death rate. Inverse associations were observed for most major causes of death, including heart disease, cancer, and respiratory diseases. Results were similar for peanuts and tree nuts, and the inverse association persisted across all subgroups.

Other studies have reported similar findings.  Bao et al. suggest several alternative explanations for these findings, other than the apparent protective effects of nuts.  The findings could reflect "confounding by unmeasured or poorly measured variables", although the authors believe this is unlikely because so much data have been collected on these study participants that they feel they could control for most potential confounders.  Unknown confounders can't be controlled for, as they point out.

Reverse causality is another possible explanation for their findings -- unhealthy people might stop eating nuts, for example, rather than that people who don't eat nuts become unhealthy.  But the researchers excluded people from the study once they reported illness, in order to limit this potential problem.

Bao et al. note that because their study is observational, they haven't demonstrated cause and effect, but they also note that their results are consistent with many other reports of the health benefits of eating nuts.  Indeed, the authors are, to an usual but praiseworthy extent, responsibly circumspect in their reporting of their results.

So, the ground was well-laid for Marchione to write a responsible story in turn, and she did. Among other issues, she points out, for example,

Researchers don't know why nuts may boost health. It could be that their unsaturated fatty acids, minerals and other nutrients lower cholesterol and inflammation and reduce other problems, as earlier studies seemed to show.

Observational studies like this one can't prove cause and effect, only suggest a connection. Research on diets is especially tough, because it can be difficult to single out the effects of any one food.

People who eat more nuts may eat them on salads, for example, and some of the benefit may come from the leafy greens, said Dr. Robert Eckel, a University of Colorado cardiologist and former president of the American Heart Association.

Read more here: http://www.centredaily.com/2013/11/20/3900498/eating-nuts-is-tied-to-lower-risk.html#storylink=cpy

.So, kudos to Marchione for pointing out that there are limitations to this kind of study.  The evidence seems to be building in favor of the protective role of nuts, and this study adds to the evidence, but it is not definitive.

Funder bias
Marchione notes rather in passing that one of the funders of the study was the International Tree Nut Council Nutrition Research and Education Foundation, but she says that they had no role in designing or reporting the results.  This, of course, raises the question of whether they were involved in analysis, but let's assume they were not.  Marchione was correct to mention this bit of information because studies funded by industry are much more likely to report findings in the funder's favor than are independently funded studies -- this is common enough that there's a Wikipedia entry for "funding bias".

Knowing that the Tree Nut people weren't involved in study design or reporting is not entirely reassuring about there being no conflict of interest.  A study published in PLoS Medicine in 2007 ("Relationship between Funding Source and Conclusion among Nutrition-Related Scientific Articles", Lesser et al.) reported the results of an examination of the relationship between sponsorship and conclusions of studies of the health effects of non-alcoholic beverages. 

206 articles were included in the study, of which 111 declared financial sponsorship. Of these, 22% had all industry funding, 47% had no industry funding, and 32% had mixed funding. Funding source was significantly related to conclusions when considering all article types (p = 0.037). For interventional studies, the proportion with unfavorable conclusions was 0% for all industry funding versus 37% for no industry funding (p = 0.009). The odds ratio of a favorable versus unfavorable conclusion was 7.61 (95% confidence interval 1.27 to 45.73), comparing articles with all industry funding to no industry funding.

The authors conclude that this has "potentially significant implications for public health."

Why this bias exists isn't clear, and there are probably multiple causes. Some of it is perhaps malfeasance or bad science, industry buying results, but it can't all be that. It's possible that when results are negative they just aren't published, and, ok, it can be argued that that's a clear sort of malfeasance. Or Lesser et al. suggest that perhaps industry funds only studies that they believe will favor their product.  But then the prior knowledge should be reported accurately and built into the testing (e.g., in a 'Bayesian' way), rather than pretending that one has a true 'null' hypothesis.  Or investigators, knowingly or not, propose studies that are likely to show that a sponsor's product is beneficial.

Lesser et al. also suggest that all investigators are biased somehow -- e.g., they want to support their favorite hypothesis -- but that financial conflict of interest is a separate animal.  "We contend that financial conflict of interest is qualitatively different, producing selective bias that acts consistently in one direction over time." This is certainly true in clinical trials funded by pharmaceutical companies.  Cosgrove and Wheeler, for example, go so far as to call the involvement of pharmaceuticals in drug testing corruption in a paper published this year in the Journal of Law, Medicine and Ethics.

The problem here is that the nut study might well have been devised, conducted and analyzed in exactly the same way without industry funding.   But it might well not have, and we have no way of knowing.  Again, Marchione was right to bring this up, but stating conflict of interest doesn't automatically prevent it.