Intended for healthcare professionals


Risks and benefits of omega 3 fats for mortality, cardiovascular disease, and cancer: systematic review

BMJ 2006; 332 doi: (Published 30 March 2006) Cite this as: BMJ 2006;332:752

Having your fish and eating it

I wish to add additional comment on the paper by Lee Hooper et al
which undertook a systematic review of randomised controlled trials and
prospective cohort studies of the effects of long and short chain omega-3
fats on cardiovascular disease, cancer and bleeding events. As far as
possible I have not repeated comments made by other contributors but focus
on reasons why this review has drawn different conclusions from a previous
meta-analyses (1) and from a review of the Scientific Advisory Committee
on Nutrition (SACN) and the Committee on Toxicity (COT) of the benefits
and risks of consumption of fish (2).

The main reason for these
differences was the inclusion of the data from Burr et al. (3) in the
pooled analysis for risk of death in high risk subjects, in the report
from Lee Hooper and colleagues. In the methods section of their paper the
authors refer to use of two specific quality criteria (concealment of
study arms and masking of participants, providers and outcome assessors)
in judging studies at potential risk of bias. The authors do not appear
to have taken account of (or acknowledged) the fact that concealment of
study arms and masking of participants and providers are design
characteristic that cannot be applied in dietary intervention trials in
which individual advice on food intake forms part of the study protocol.

However many other aspects of study design are used to evaluate the
quality of evidence used to guide policy and public health advice in the
area of diet and these were not used in judging the quality of the papers
included in this latest meta-analysis. The risk assessment protocol
applied by SACN/COT, concluded that the study of Burr et al contained a
number of significant flaws that reduced the reliability of the data.

These have also been fully addressed in the previous correspondence. The
SACN/COT review group were provide with a copy of the Cochrane review
which formed the basis of the present meta-analysis and concluded that a
meta-analytical approach, without consideration of wider study design
issues, was inappropriate for use in their overall assessment of the
evidence. A critique of the use of Cochrane reviews in assessing the
reliability of knowledge about diet and chronic disease has recently been
published (4). It clearly illustrates why meta- analysis cannot be
considered a ‘gold standard’ for evaluating dietary intervention trials
and is in direct conflict with the authors view that their data ‘provides
high quality evidence to guide policy and practice’.

The explanations put forward by the authors to explain differences
between the findings of Burr et al and other large long chain omega-3
intervention trials, appear to be unlikely. They suggest the study to be
the longest follow up of all the long chain omega -3 trials and consider
the findings may reflect cumulative adverse effects of methylmercury in
fish and fish oils. However, the adverse impacts were largely confined to
the second phase which represented subjects whose follow up (36 months
maximum) was shorter than phase one (72 months) and shorter than the 42
month follow up in the GISSI trial. The possibility that fish or fish oils
have different effects in men with angina than in those who have had a
recent myocardial infarction (MI) is also an unlikely explanation given
that many of the post MI patients in the GISSI-P and the DART I trial were
likely to be suffering angina. Burr et al suggested a possible explanation
lay in the differential behaviour of men given fish oil capsules, as
opposed to dietary advice. If so, this illustrates precisely why design
factors other than concealment to study arms and masking, are essential in
controlled intervention trials involving diet.

Whilst the size of the study conducted by Burr et al. provides a
potentially important contribution to the literature in this area, the
authors themselves have been extremely transparent concerning the
limitations of their study and fully discuss these in their paper, and
also in communications with the SACN/COT review group. However, despite
acknowledging the important contribution made by the Burr et al paper to
the conclusion drawn from their meta-analysis, Lee Hooper and colleagues
have failed to give any consideration to these quality issues in the
discussion of their findings, and as a consequence, have exaggerated the
strength of the findings from Burr et al.

During the significant media coverage which followed the publication
of the review, one of the authors was quoted as supporting the
recommendations of the SACN/COT fish report. The SACN/COT review group
could not have made these general population recommendations had they
concluded their advice to be hazardous to a significant proportion of the
adult population (angina sufferers), since this is contrary to the basic
tenets of sound public health policy and practice. On this basis, it would
seem the authors wish to have their fish and eat it.

Because of the media coverage given to the findings of this study and
the damage done to the clarity of public health policy on diet and heart
disease, it is hoped that this response will be given equal coverage, and
by doing so, will elevate the level of debate concerning the evaluation of
science in the area of diet and health.

Competing interests:
None declared

Competing interests: No competing interests

27 April 2006
Christine M Williams
Professor of Human Nutrition
University of Reading,UK RG6 6AO