The Nattering Nabobs of Nutrition

Professional Discipline

All professional disciplines have some standards of performance, but how those standards are developed vary tremendously.  Some professions, like engineering, have obvious external measures of success.  A bridge must work like a bridge. Under those circumstances, the credentials an engineer must have to practice their trade are themselves designed to insure, for example, that the bridge will stand and cross the river.  Yet, even when there are objective of performance, there are examples where the credentialing process has proven to be inadequate.  For example, stock market crashes often come about because standard accounting practices show one or more failing businesses to be entirely solvent.

A profession with absolutely no objective measures of performance isn’t providing much of a service.  So, there aren’t very many of those.  But, there are plenty of professions where the definition of successful performance is not at all obvious even when there is a clear purpose.  Examples can be found in medicine that cover the entire range.  Some treatments are highly effective; and under those circumstances medicine is essentally biological engineering.  If the treatment doesn’t work, then it is likely that the doctor is at fault.  But there are also many other medical treatments where success is not so straightforwardly defined.  The treatment may not always work, or it may only partially mitigate or cure a disease.   In that case, the standard of performance cannot be entirely dictated by the success of an outcome.  Instead a doctor is likely to be judged as to whether or not they followed the norms of the profession.

Academic Fiefdoms

Many academic disciplines correspond to professions, so standards of academic practice will generally correspond to those of the profession.  However, because performance in academia is less often tied directly to objective measures of professional success, normative standards of behavior are more often what determines acceptable practice.  In other words, peer review.  One might presume that the journal reviewers, editors and study section members that essentially regulate a scientific discipline all have some objective measures of success in mind, but maybe they don’t.  There is no written guarantee.

The other thing about academia is that there are many areas of academic study that don’t correspond to a profession at all.  For example, if you get a degree in literature or art history, about the only thing you can do with it is teach literature or art history.  Since there are no objective measures of performance, peer review is the only game in town.  If the reviewers say the paper is good, it is.  If they say it’s bad, then it is.  Why?  "Because we say so".

In the realm of public health, there are plenty of tweeners.  Ostensibly, public health is concerned with the health of the entire population, rather than the health of individuals.  You might think that would entail the same sense of objective performance that is found in medicine, but the reality is that the quantitative analyses that define success can be so nebulous that only a "properly qualified" expert can judge it. Sure, there are standards of practice, but do those standards actually work?  Is “statistical significance” indicative of actual significance? Maybe not.  Do environmental epidemiologists think it is their responsibility to sort out when an association can be considered to be causal?  Often, they don’t. 

The other thing about a public health is that you can’t just take your degree and hang out your shingle.  Outside of academia, the main employers are governments.  Consequently, the norms of public health tend to be very political.  For example, if you are a toxicologist, the quality of a scientific argument may be judged by whether or not it advocates proper public policy.   It also explains why, when push comes to shove, the safety assessment paradigm is preferred to the risk assessment paradigm: It gives the profession of toxicology more control over public policy.  It can also add the authority of the government to “because we say so”.

The Nutrition Paradigm

Some nutritionists work at a personal level, and they often have truly objective measures of performance (e.g. weight loss, energy).  But, in public health the main game is setting standards.  Like the standards in toxicology, there are a number of acronyms, including the older Recommended Dietary Allowance (RDA), which has been replaced by the RDI (Reference Daily Intake or Recommended Daily Intake, which lets decide whether you prefer to sound like the FDA or the EPA).   The RDI is defined as

The daily intake level of a nutrient that is considered to be sufficient to meet the requirements of 97–98% of healthy individuals in every demographic in the United States.

But unlike the standards of toxicology, there is no regimented procedure for the creation of RDAs.  On the hand, there is a "framework" (IOM, 2015).  NOT having a prescriptive procedure allows every nutrient to be treated as a unique issue, so that can be a good thing.  But here's the problem: when nutrients are also toxic, the standards of practice of toxicology and nutrition result in a cultural clash (Greger, 1998; Oilin, 1998).   In particular, all of the following issues arise:

• Safety factors.  This is the most obvious difference.  In setting standards, toxicologists routinely apply safety factors, while nutritionists do not.  While there are many nutrients where perhaps an extra factor of 10 could be applied “just to be safe”, there are many others where 10 times the RDA is toxic (e.g. iron, vitamin A).  It is also not so clear that toxicologists really should be completely enamored with safety factors either, so let’s score this one in favor of the nutritionists.
• Burdens of Proof.  Toxicologists tend to want to err on the side of safety, which often means using a worst-case analysis as the basis for setting standards.  For example, the Benchmark Dose has come to be preferred to the No Observed Adverse Effect Level (NOAEL) because it reverses the burdens of proof.  Although nutritionists tend to use human experiments (i.e. Randomized Clinical Trials - RCTs) to set standards, the generally use a standard of certainty (i.e. staitsically significant) akin to the NOAEL as the burden of proof.  If you want to balance nutritional requirement vs toxicity, then neither one of these standard are really appropriate.  Common sense dictates that what is really needed is a middle of the road as-likely-as-not standard (i.e. “preponderance of the evidence”).  Let’s call this one even.
• Dose-Response.  Toxicology has a long history of emphasizing the relationship between dose and response.  The dose makes the poison.   But, you just never hear “the dose makes the nutrient”.  Obviously, there are underlying dose response relationships for nutritional effects.  In general, it is presumed that while a certain amount is required, more is not necessary (Olin, 1988).  (But then again, there is the Vitamin C debate).  But somehow that fact doesn’t always get figured into the design of RCTs.  For example, Hurtado et al (2015) reported the results of an RCT concerned with omega-3 supplementation during pregnancy.  Yet the members of the study were all advised to consume fish 2-3 times, which follows guidelines that are partly justified by the fact that “fish and seafood also are sources of other important nutrients, including the long-chain polyunsaturated fatty acids (PUFAs) eicosapentaenoic acid/docosahexaenoic acid (EPA/DHA)” (IOM, 2015).   Yet, the authors concluded that “Omega-3 LCPUFA supplementation had no neurodevelopment effects”.   What’s wrong with that?  Well, let’s suppose you were going to design a toxicology experiment.  Would your control group be a population that had already been poisoned?   And then when they all died, would you conclude that the chemical being studied has no effect?  No, that would very silly.  What the Hurtado et al (2015) really needed for the study were pregnant women who eat no fish at all.  But that would run into the same ethical problems that prevent toxicologists from conducting experiments with humans.   Score this one for Toxicology.

That leaves our nattering nabobs both tied with a record of 1-1-1.  Based on their superior alliterative qualities, the Nutritionists get the nod for the post title.

References

Greger JL (1998).  Dietary Standards for Manganese: Overlap between Nutritional and Toxicological Studies.  J. Nutrition 128:368S-371S.

Hurtado JA, Iznaola C, Peña M, Ruíz J, Peña –Quintana L, Kajarabille N, Rodriguez-Santana Y, Sanjurjo P, Aldámiz-Echevarría L, Ochoa J, and Lara-Villoslada F (2015).  Effects of Maternal Ω-3 Supplementation on Fatty Acids and on Visual and Cognitive Development.  J Pediatr Gastroenterol Nutr. 61:472-80

Institute of Medicine (2015).  A Framework for Assessing Effects of the Food System.  National Academy Press, Washington, DC.

Olin, SS (1998). Between a Rock and a Hard Place: Methods for Setting Dietary Allowances and Exposure Limits for Essential Minerals.  J. Nutrition 128:364S-367S.

Official Post Soundtrack

Talking Heads (1977).  Don’t Worry About the Government.  In: Talking Heads 77, Track 8.

Post Notes

Thesis Post #58.  Most related to An Ethical Science and my next post which be about a nutritional dose-response for fish consumption.   My last post from Oxford.

Methylmercury in Your Child’s Brain

Politics

Under other circumstances, I might try to provide a discussion of the nuances of modeling the dose-response relationship between a mother’s exposure to methylmercury and neurobehavioral development of the child.  But, the fact of the matter is I probably wouldn’t do much better than I and others have done in the past (e.g. FDA, 2014), so I won’t bother.  Besides, a discussion of the vying political interests associated with having a dose-response model at all just might be far more interesting anyway.

With methylmercury, the political story revolves around an arbitrary number called the Reference Dose (RfD).  The meaninglessness of the RfD is entirely intentional – it was specifically designed to be devoid of any and all information value.  While the RfD and similar values do have some legitimate regulatory uses, the RfD is also often used as a form of political rhetoric.  The fact that it doesn’t mean a damn thing makes it perfect for that. 

The Environmental Working Group (EWG) discussion of the toxic effects of methylmercury provides a nice concise example of the use of the RfD as political propaganda.  Specifically, they make the following statement concerning the RfD:

In 2001 the EPA concluded that a pregnant woman could consume 0.1 micrograms of mercury per kilogram of bodyweight daily without ill effects to her fetus and that this amount of mercury would also be safe for children and adults (EPA 2001).

Actually, that’s wrong.  The EPA description of the Reference Dose is “likely to be without an appreciable risk of deleterious effects”, not “without ill effects”.  Maybe that sounds picky, but the difference is that the second sentence implies that scientists actually have some way of identifying a dose that has absolutely no “ill effects”.  No, we really don’t.   They then go on to say;

government and independent scientists have not reached a consensus on a safe level of mercury exposure

Yeah, well maybe that’s because there is no such thing.  Scientists have never agreed on how many angels can dance on the head of a pin either.  But that doesn’t stop EWG from taking one “safe” level and morphing it into another:

EWG recommends that pregnant women and children consume no more than 75 percent of EPA’s safety level.  By doing so, they are likely to build an extra margin of safety from lifelong damage mercury can inflict on the developing brain.

They’re just making stuff up – can you tell? I’m sure EWG thinks their political adversaries are dastardly money-grubbing fishmongers.  Maybe some of them are.  But, I’m all about defending the environment from ignorant political pablum.  That means I’m in favor of both science and democracy.  If that sounds trite, well trust me it isn’t. I want to give decision makers the best information possible – even if they happen to be Democrats, Republicans, or citizens of some other ilk.  Furthermore, I want scientists to be able to say what they know about methylmercury without passing moral judgment at the same time.  Is all that too much to ask?  No, it is not:  A nice scientifically-defensible dose-response function will send all those 'safe’ levels back to the netherworld from whence they came.

Science

Rant over; the FDA risk-benefit analysis used three different dose-response analyses, all of which have been published.  They are all derived from populations with either very high level fish consumption (New Zealand, Seychelles, and the Faroe Islands) or from an epidemic where there was accidental exposure to methylmercury.  There are also many studies at lower exposure levels, but they aren’t very useful for establishing a dose-response relationship.  The reason for that is straightforward:  Since many many factors can influence the behavioral development of a child, any statistical association observed in a low dose study can only be reasonably interpreted to be causal if it is consistent with what has been observed at higher doses (i.e. if there is a toxic effect a at a low dose, a bigger effect is to be expected at higher doses).  So, we may just as well stick with the higher dose studies:

• Delayed Walking and Talking.  This is an analysis I did while I was at the FDA (Carrington and Bolger, 2000).  It is based on a pooled data set from both Iraq and the Seychelles where the age at which each child began to walk or talk was recorded.  The exposure of mothers to methylmercury during pregnancy was measured by taking hair samples.  The main advantage this data set has over the others is that you don’t really have to worry about the whether or not the associations are really causal – the effects that occurred in Iraq were big and obvious (for example, there were two children who could not walk or talk at age 5).  In addition, it is also possible to tell something about how variable the effects are – not all the children exposed to very high levels of methylmercury responded in exactly the same way.
• Intelligence Quotient (IQ). This analysis (Axelrad et al, 2007) was developed by the EPA for doing cost-benefit analyses.  It was used by the Office of Air to support the Mercury Rule enacted in 2011.  Because the agency already had developed a methodology for calculating a dollar value for impacts on IQ, it was especially important to quantify the impact of methlymercury on this standardized measure of intelligence.  Sinc the Faroe island study only completed four of the 10 tests that are part of measuring IQ, (both the New Zealand and the Seychelles studies did the full set), this isn’t exactly IQ – but it’s as close as anyone is going to get.  [The Office of Water tried to block the use of this analysis because they correctly surmised that it would make the Reference Dose irrelevant].
• Domains.  The New Zealand, Faroes Island, and Seychelles studies all conducted a wide variety of different tests on each child in the study.  A group of scientists at Harvard (Cohen et al, 2005) went through the results from all three studies and sorted them into different functional “domains”: Motor, attention, visuospatial/visuomotor, language, memory, and intelligence.  By doing so, they hoped to determine whether or not methylmercury impacted specific areas of the brain.  But, as it turned out, there were no clear cut differences between the different categories of performance, which suggests that methylmercury has effects throughout the brain.  Nonetheless, this analysis does a good job of showing how big the effects of methylmercury may be across a wide variety of behavioral tests.

Software

Besides providing an opportunity to peruse the different dose-response models in isolation, the VBA macros provided here can be used as part of a larger model:

MeHg DR_Functions.xls

MeHg_DR_Function_Reference

References

Axelrad D.A., Bellinger, D.C., Ryan L.M., and Woodruff T.J. (2007), Dose-Response Relationship of Prenatal Mercury Exposure and IQ:  An Integrative Analysis of Epidemiological Data.  Environmental Health Perspectives 115:609-615.

Carrington, CD and Bolger, PM (2000).  A Pooled Analysis of the Iraqi and Seychelles Methylmercury Studies.  Human Ecological Risk Assessment 6:323-340.   Also, in Appendix E here.

Cohen, J.T., Bellinger, D.C., and Shaywitz, B.A. (2005). A Quantitative Analysis of Prenatal Methyl Mercury Exposure and Cognitive Development.  American Journal of Preventive Medicine 29:353-365.

Official Post Soundtrack

Rush (2007).  Workin’ Them Angels.  In: Snakes & Arrows, Track 3.

Post Notes

Thesis Post #57.  Part of the Individual Risk/Methylmercury series.

There’s Always a Bigger Fish

A Rapidly Changing Fish Market

Like poultry, pork, and beef, seafood is a source of protein.  But, unlike the other three categories of animal flesh that are consumed as food, there is far more variety among fish than chickens, pigs, or cattle.  This is largely because many fish are still wild-caught from the oceans.  But that’s not as true as it used to be; the majority of fish on the market are now produced by aquaculture (i.e. farm-raised), and the percentage of aquaculture seafood on the market is expected to increase as declining stocks of wild-caught fish are replaced with those that are farm-raised.  Major species raised by aquaculture include catfish, crawfish, mussels, salmon, scallops, shrimp, tilapia, and trout (FAO, 2016).   But, even though there are many varieties of seafood raised as aquaculture, there still are many more different varieties that are wild caught.

For the purposes of a discussion of the pluses and minuses of consuming fish while pregnant, there are two main characteristics of fish that are of interest.  The first is methylmercury.  Even though some methylmercury can be found in just about all seafood, the amount varies tremendously.  Since it bioaccumulates, higher methylmercury levels tend to occur in predatory fish that are higher up in the food chain.   Among different individual animals of the same species, larger older fish tend to have higher methylmercury levels than those that are younger.  However, because wild-caught fish supplies are dwindling, the supply of high mercury fish is too.  Since the aquaculture fish that are replacing them are lower down in the food chain and are harvested when they are young, they tend to have uniformly low methylmercury levels.  However, they can have higher levels of other contaminants like antibiotics and dioxins.

In addition to mercury, fish can be an important source of many nutrients.  Besides protein, the most noted of these are omega-3 fatty acids.  There have many studies on potential nutritional effects of isolated omega-3 fatty acids (i.e. fish-oil capsules), and most of the results are negative.  Perhaps the best evidence comes from epidemiological studies of fish consumptions where modest levels of fish consumption appear to slightly reduce the incidence of stroke, cardiovascular disease, and perhaps other outcomes like neurobehavioral development (FDA, 2014).  So, omega-3 fatty acids, or possibly some other constituent in fish, probably do not correct a common nutritional deficiency, but it does appear that they may address  a nutritional deficiency that is only found in people who consume little or no fish.

A Survey of Fish Species Sold in the United States

The FDA has been collecting samples from the U.S. market since the 1990.  While early effects focused only on shark and swordfish, the current data base has over 60 categories of fish.  The table below shows estimated average methylmercury for 51 categories, which as of 2012 included over 99% of all species sold in the United States.  However, the FDA database hasn’t been updated since 2010, and the fish on the market are constantly changing, so this table is not exactly current.  But, it will have to do for now.

Also included in the table are omega-3 concentrations.  Most of these are taken from the USDA nutrient database.  While the FDA methylmercury concentrations are reported with ranges (and raw data if you like), the USDA only reports average values.  In order to emphasize the nutrient value of each category, they are sorted by the ratio omega-3 concentrations to methylmercury concentrations.  If you consume fish on a regular basis anyway, then the omega-3s are probably not really an issue, so you may be more interested in just methylmercury concentrations.  They are sorted that way on the FDA web summary and in the FDA risk-benefit report (FDA, 2014).

Fish Category	Hg (µg per g)	Total Ω-3 g/100g	µg Hg per g Ω-3
Sardines	0.02	1.19	1.7
Salmon	0.02	1.18	2.0
Oysters and Mussels	0.02	0.70	2.1
Anchovies, Herring, and Shad	0.05	2.02	2.5
Shrimp	0.01	0.35	3.1
Trout, Freshwater	0.03	0.93	3.4
Scallops	0.01	0.19	3.7
Mackerel, Atlantic and Atka	0.05	1.20	4.1
Mackerel, Chub	0.09	1.25	7.0
Pollock	0.04	0.53	7.0
Smelt	0.07	0.89	7.5
Catfish	0.02	0.22	7.6
Butterfish	0.06	0.73	8.0
Whitefish	0.10	0.91	11.0
Clams	0.02	0.20	11.6
Tilefish, Atlantic	0.11	0.91	12.2
Squid	0.07	0.54	12.9
Tilapia	0.01	0.09	14.3
Crabs	0.06	0.38	16.6
Sablefish	0.37	1.81	20.4
Crawfish	0.03	0.16	20.7
Lobsters, Spiny	0.11	0.48	22.9
Flatfish	0.08	0.30	25.3
Bass, Saltwater	0.25	0.97	25.9
Mackerel, Spanish	0.37	1.25	29.7
Halibut	0.22	0.71	31.4
Bluefish	0.35	0.99	35.4
Carp and Buffalofish	0.17	0.45	37.7
Croaker, Atlantic	0.08	0.20	38.6
Tuna, Albacore Canned, Water	0.35	0.86	40.6
Haddock, Hake, and Monkfish	0.07	0.16	41.9
Bass, Freshwater	0.32	0.76	41.9
Trout, Saltwater	0.26	0.62	42.0
Tuna, Light Canned, Water	0.12	0.27	44.4
Skate	0.14	0.30	45.7
Perch, Ocean and Mullet	0.16	0.32	49.4
Perch, Freshwater	0.15	0.29	50.9
Pike	0.14	0.27	52.3
Cod	0.09	0.16	55.6
Lobsters, American	0.11	0.20	56.4
Tuna, Fresh	0.39	0.65	59.9
Snapper, Porgy, and Sheepshead	0.16	0.26	62.6
Marlin	0.49	0.50	98.0
Croaker, Pacific	0.30	0.30	100.0
Lingcod and Scorpionfish	0.29	0.26	108.7
Swordfish	1.00	0.90	111.2
Shark	0.98	0.69	142.2
Tilefish, Gulf	1.45	0.80	181.3
Mackerel, King	0.73	0.40	182.0
Grouper	0.46	0.25	185.5
Orange Roughy	0.57	0.03	1838.7

Software

One of the problems with public health advice is that it often doesn’t take into account other things that individual consumers want.  Maybe you would rather eat more fish, or less.  Maybe you prefer tuna to salmon.  Maybe fish is too expensive.  So, instead of handing out advice I’m going to concentrate on giving out information.  Then you can figure out whether or not it’s worth changing your fish eating habits for yourself.  The first step is figuring out how your fish consumption choices affect how much methylmercury and omega-3 are in your diet.

Personal_Seafood_and_MeHg_Intake_Estimator.xls

References

Food and Agriculture Organization (2016).  Aquaculture Fact Sheets.

U.S. Food and Drug Administration (2014). Quantitative Assessment of the Net Effects on Fetal Neurodevelopment from Eating Commercial Fish (As Measured by IQ and also by Early Age Verbal Development in Children).  

Official Post Soundtrack

Murphy, Peter (1989).  Deep Ocean Vast Sea.  In: Deep, Track 1.

Post  Notes

Thesis Post #56.  The is the first of what I think will be a five part series that describes a individualized version of the risk-benefit model I created for the FDA [part two was the last post, so part three will be next].  I'm putting it out in parts largely to provide documentation of the model -- the code is all open.  In the end, I will combine them all into a single program.