HOME
FISH CONSUMPTION AND MORTALITY:
RELATIVE RISK — WORSE THAN ANY MOTHER-IN-LAW
We are often told that eating fish will protect us from a variety of diseases, of which the most important is coronary heart disease (CHD) otherwise called ischaemic heart disease (IHD).
But just how strong is the evidence for this recommendation?
In the article below, Professor Joel Kauffman, Emeritus, University of the Sciences, Philadelphia, shows that it really isn't very strong at all; that it is the use of 'relative risk'* (see below), which is attention-grabbing, rather than 'absolute risk', which is more meaningful, which has artificially inflated the apparent benefits from eating fish.
The claim
Laura J. Ninger, ELS, in Life Extension magazine 2009;15(2):22 wrote that the "latest research" showed that a threshold level of omega-3 fatty acids from fish (EPA and DHA) protects against heart disease ... and should be endorsed by authorities". She continued: "At an intake of 566 mg/day of EPA plus DHA, the mortality reduction was 37%." But that was relative risk (RR) and cardiac mortality only, neither specified.
The only citation Ninger gave was to William S. Harris, and colleagues. Intakes of Long-Chain Omega-3 Fatty Acid Associated with Reduced Risk for Death from Coronary Heart Disease in Healthy Adults, Curr Atheroscler Rep 2008;10(6, Dec): 503-9. This "atest" research was a comparative review of six studies published in 1992, 1995, 1998, 2002 (two) and 2003.
One of the most useful things in this paper is confirmation that the benefits of eating fish, slight though they may be, are due to the content of the omega-3 fatty acids, EPA & DHA, not to some other component of fish. There is a useful table of the EPA & DHA content in 37 common kinds of fish. Orange roughy and tilapia have the least, and are useless, while farmed Atlantic salmon had the most.
But this review suffers from two common failings of medical and nutritional reporting: Firstly, relative risk (RR*) is used exclusively with no idea of absolute risk*, so most of the study results were of very low clinical significance; and secondly, cardiovascular deaths were used exclusively with no presentation of all-cause death, so no action should have been recommended in the absence of mortality data, since so many interventions that delay heart deaths increase cancer and other causes of death.
Four of the studies were examined in detail to demonstrate the problems with using Relative Risk in the absence of absolute risk or mortality.
The studies
(1). Mozaffarian D et al., Cardiac Benefits of Fish Consumption May
Depend on the Type of Fish Meal Consumed. Circulation
2003;107:1372
This study had no total mortality figures, but there were numbers of
subjects (3910, ≥65 years, both sexes at baseline) and events after
9.3 years. Adjustments gave only slightly different results, so in the
Cardiovascular Health Study:
| Fish Intake | < 1/mo | 1-3/mo | 1/wk | 2/wk | > 2/wk |
| Subjects | 381 | 917 | 800 | 610 | 1202 |
| Total IHD Deaths | 39 | 75 | 58 | 36 | 39 |
| Relative Risk | 1.00 | 0.78 | 0.65 | 0.53 | 0.28 |
| % Dead | 10 | 8 | 7 | 6 | 3 |
| % No IHD Death | 90 | 92 | 93 | 94 | 97 |
After 9.3 years, the clinical significance is moderate from least to most fish, around 0.7% per year with no IHD (heart attack) death. Comparing the mean fish intake of 1 per week with more than 2 per week, 93% vs. 97% with no IHD death is not much different, only 4% absolute per year with no IHD death, NNT** = 25. Without total mortality no recommendation should be made. This is the study with the largest difference in absolute risk, probably because the subjects were oldest, and the only study to find no benefit in fried fish or fish sandwiches compared with broiled or baked fish.
(2). Albert CM, et al., Blood Levels of Long-Chain n–3 Fatty Acids
and the Risk of Sudden Death. NEJM 2002;346:1113
This second study reported for the
Health Professionals Study a "significant" reduction in mortalityy
from sudden cardiac death (SCD) by eating more fish, but in the absence of
the number of subjects in each quartile, or the number of fatal events
in each quartile in the paper, the absolute risk could not be
ascertained. In this study there were quartiles set up, not of fish
intake, but of EPA/DHA levels in serum by actual assay. Adjusted for
age and smoking, the RRs were 1.00, 0.47 (Non Significant), 0.37 and
0.31. These authors did specifically write that the only change in
mortality related to EPA/DHA levels was sudden cardiac death, with no
other cause of death being affected.
But among 22,071 male physicians at baseline, there were just 201 sudden cardiac deaths in 17 years. That's 0.91% deaths, or 0.05% per year. Whatever was found in RR, the results are of small clinical significance for lifestyle change consideration.
(3). Hu FB, et al., Fish and Omega-3 Fatty Acid Intake and Risk of
Coronary Heart Disease in Women. JAMA 2002;287:1815-1821.
Here
84,688 women in the Nurses Health Study were followed for 16 years to
find 484 fatal heart attack (MI) (0.57%) and 1029 non-fatal MI (1.2%)
and 4121 total deaths. Here most deaths (88%) were not from MI.
Essentially all the benefit was obtained by eating fish 1-3 times per
month. With the total fatal MI so low in absolute terms, not much could
be expected from more than 1-3 fish meals per month.
| Fish Intake | < 1/mo | 1-3/mo | 1/wk | 2-4/wk | ≥ 5/wk |
| Person-Years | 67537 | 337393 | 690479 | 157711 | 54525 |
| Subjects PY/16 | 4221 | 21087 | 43195 | 9857 | 3408 |
| Fatal MI, n | 41 | 126 | 231 | 61 | 25 |
| Relative Risk | 1.00 | 0.58 | 0.46 | 0.52 | 0.59 |
| % Fatal MI | 0.97 | 0.60 | 0.53 | 0.62 | 0.59 |
| % NO Fatal MI | 99.0 | 99.4 | 99.5 | 99.4 | 99.3 |
Hu et al. wrote that: "These findings lend further support to current dietary guidelines recommending fish consumption twice weekly for the prevention of CHD." You can see that they do not. From less than 1 meal per month of fish to 2-4 meals per week the improvement in fatal MI was 0.4% absolute which corresponds with an NNT of 250! Worse, we do not know how fish consumption was related to the 3637 non-MI deaths. No recommendation was justified.
(4). Albert CM, et al., Fish Consumption and Risk of Sudden Cardiac
Death. JAMA 1998;279:23-28.
Here 20,551 male physicians were
followed for 11 years to find 133 sudden deaths (0.65%). There was a
multivariate adjusted RR = 0.48 for 1 - <2 fish meals per week vs.
less than 1 per month, an impressive drop in RR. But there was no
relation found between fish intake and all MI, contrary to studies (1)
and (3) above. The detailed findings on sudden cardiac death (SCD) and
fish intake was:
| Fish Intake | < 1/mo | 1-3/mo | 1 - <2/wk | 2 - <5/wk | ≥ 5/wk |
| Person-Years | 7715 | 15465 | 79561 | 123693 | 27343 |
| Subjects PY/11 | 701 | 1406 | 7233 | 11245 | 2486 |
| SCD, n | 9 | 12 | 38 | 64 | 10 |
| RR of SCD* | 1.00 | 0.58 | 0.46 | 0.52 | 0.59 |
| % Fatal SCD | 1.3 | 0.85 | 0.53 | 0.57 | 0.40 |
| % NO Fatal SCD | 98.7 | 99.2 | 99.5 | 99.4 | 99.6 |
From less than 1 meal per month of fish to 1 - <2 per week the improvement in fatal SCD was 0.8% absolute which corresponds with an NNT of 125! But the conclusions in the body of the paper were: "…our data add to the existing totality of evidence that consuming at least 1 meal of fish per week may substantially reduce the risk of sudden cardiac death."
A RR = 0.71 for all-cause mortality (95% CI 0.55-0.89) was reported for 1 - <2 fish meals per week vs. less than 1 per month. This is a less impressive figure, which was even less impressive with RR = 0.73 at 5 or more meals per week. Numbers of all-cause deaths vs. fish intake were not provided
More fish intake tends to go with more education, higher wealth and other lifestyle changes such as better diet overall and supplement use, most of which were not adjusted for in most of the studies.
Fish and Mercury
Harris et al. (2008) worried about mercury in fish, citing the FDA’s limits. But: "The Chicago Western Electric Study followed the effects of fish consumption in 2,107 men aged 40-55, and followed for 30 years. Those who ate an average of ≥35 g daily (about 1 big fish dinner every 5 days) had only 9/10 of the all-cause mortality rate of men who ate no fish. [This was a drop from 56% deaths to 53% deaths (0.1% per year). A big drop in MI was indeed accompanied by a lesser increase in non-cardiac deaths.]
"Pregnant women have been cautioned to restrict their intake of fish (http://www.cbc.ca/storyview/CBC/2002/10/21/Consumers/mercuryfish_021021) despite evidence that children receive most of their mercury from vaccines. Hepatitis b vaccine carries 12.5 micrograms per dose; influenza and other common vaccines carry 25 micrograms per shot. That is over 830 times the amount in a can of tuna. It has been reported that vaccines said not to contain the mercury compound, thimerosal, still might have it. The long duration of the diet studies makes it very clear that the mercury content of fish, in general, is not shortening life." From: Joel M. Kauffman, Malignant Medical Myths, Infinity Publ., West Conshohocken, PA, 2006, p11.
Clearly, the mercury content in common fish eaten by Americans is not shortening lives. Thus there is reason to recommend fish for minor benefits in longevity, but very little reason to recommend more than 1 per week. On the other hand, fish at up to 5 week is not harmful.
Researchers are less than honest in making RR and single-cause mortality the basis of press releases. Too bad that referees, editors, reporters, non-ICIM physicians and most of us do not notice.
-----Joel M. Kauffman, 28 Jan 09
* Relative Risk versus Absolute Risk
If one person in every 100,000 of a population suffers an ailment, then the absolute risk is 1 per 100,000, or 0.001%. If the number of people suffering the disease doubles, there has been an apparently spectacular increase of 100% in the relative risk, but the absolute risk, at 0.002%, or 1 in 50,000, is still extremely small.
When the pharmaceutical industry and the media want headlines, it is the more eye-catching relative risks that are quoted; the unspectacular absolute risks are rarely mentioned. Yet the absolute risks are the ones that matter: they are the ones that predict the likelihood of any one person suffering a complaint.
** NNT
Number Needed to Treat (NNT) is the number of patients who have to be treated to prevent one event. The greater the NNT, the less effective is the treatment, and the less significance it has for a single person. For example, if NNT is 1,000, that means that 1,000 people have to be treated for a single person to benefit - and there is only a one in a thousand chance that person will be you. When many treatments have adverse side effects, a high NNT means that most patients will suffer a great many adverse side effects yet get no benefit.
Related Articles
