Intended for healthcare professionals

Research

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

BMJ 2006; 332 doi: https://doi.org/10.1136/bmj.38755.366331.2F (Published 30 March 2006) Cite this as: BMJ 2006;332:752

Authors' reply - omega 3s and health

Thank you all for your comments about our systematic review on the
effects of oily fish and fish oils on mortality, cardiovascular disease
and cancers.(1)

Our findings

We did not report that ‘long chain omega-3 does not offer any
protection from heart disease’, that ‘omega-3 fats have no effect on total
mortality, combined cardiovascular events, or cancer’ or that omega 3 fats
are of ‘no benefit’ - this is not what we found, or what we reported
(despite our being misquoted in much of the press).

The overall result from the meta-analysis was a relative risk
suggesting benefit of omega 3 intake on mortality, but a confidence
interval including unity (total mortality RR 0.87, 95% CI 0.73 to 1.03
with significant heterogeneity) so that we are not quite so sure. This
raises the possibility that the protective effect of omega 3s may have
been overestimated OR the heterogeneity may imply that omega 3s may be
helpful in some people and less so (or even harmful) in others (perhaps
men with angina or people with higher omega 6 intakes?). If the first is
the case, this is not a big issue, and general advice on therapeutic doses
of omega 3 fats should probably not alter. If the second is the case then
we owe it to our patients to watch the unfolding evidence very carefully
so we can react fast and alter our advice for specific groups if it proves
necessary.

Apples, pears and different outcomes

Many of you suggest that if we kept different types of studies
separate, or separated out specific outcomes, the results of our meta-
analysis would have been different. This is not the case, we did run most
of the subgroupings and individual outcomes that you suggest in the
original Cochrane review.(2) However, there was no space to present all
these results in the BMJ version of this review. The full Cochrane review
can be found on the Cochrane Library at:
http://www.mrw.interscience.wiley.com/cochrane/clsysrev/articles/CD00317....
(The Cochrane Library is free to those from the UK and some other
countries, for others it is by subscription.)

Analyses that some of you requested (see Cochrane review table 04 and
forest plots):
• Subgrouping by oily fish advice vs. supplementation with fish oil: RR of
total mortality in trials where dietary advice to eat more oily fish was
given 0.91 (95% CI 0.57 to 1.44, 3 trials, 4727 participants, 664 deaths –
highly heterogeneous), RR of total mortality in trials where fish oil
supplements, capsules or oil were provided 0.90 (95% CI 0.76 to 1.07, 39
trials, 32641 participants, 1569 deaths)
• Subgrouping by plant vs. fish source: see BMJ paper, subgrouping results
are reported within or below all of the forest plots, none are
statistically significant
• Sudden death or fatal MI rather than total mortality: RR of sudden death
in all trials with data 0.85 (95% CI 0.49 to 1.48, 6 trials, 416 deaths –
highly heterogeneous), RR of fatal MI in all trials with data 0.86 (95% CI
0.60 to 1.25, 8 trials, 390 deaths).
• Primary vs. secondary prevention: RR of total mortality in trials where
participants have existing CVD 0.84 (95% CI 0.70 to 1.02, 18 trials, 20002
participants, 1907 deaths), RR of total mortality in trials where
participants have strong risk factors but not diagnosed CVD 1.04 (95% CI
0.04 to 24.53, 9 trials, 1564 participants, 5 deaths), RR of total
mortality in trials where participants are at low initial risk of CVD 1.07
(95% CI 0.70 to 1.64, 17 trials, 14599 participants, 83 deaths).

Unfortunately there are virtually no data from the trials on genetic
factors so further analysis will need new trials. We did run a meta-
regression of fish-based omega 3 dose (combination of EPA, DHA and DPA
where known) vs. total mortality (see the Cochrane review) – we did not
find a statistically significant relationship. We did not break down the
results into individual fatty acids.

Including the Burr 2003 paper

We accept that the Burr 2003 trial(3) was not perfect, but neither
were the other two very large trials (and many of the smaller ones) that
contribute most of the data to the analysis. The DART trial(4) was
similar in methodology to the later Burr trial. The GISSI-P trial,(5) the
largest, was also imperfect as it did not provide a placebo. We find it
hard to understand how one could justify including DART and GISSI-P (and
these trials are included as star evidence within most reviews that cite
the effectiveness of omega 3s) without including Burr 2003 – you would
need to write your exclusion criteria to specifically exclude it. The
strength of the Cochrane Systematic Review process is that it aims to
remove any bias during the meta-analysis and precludes post-hoc
exclusions.

A couple of points to note about Burr 2003 – there is evidence from
weighed dietary records of 1202 subjects (38.6% of the total) that fish
advice increased the intake of EPA (from fish and capsules) and decreased
total fat and saturated fat intake. Also, around 50% of the men with
chronic stable angina had had a myocardial infarction in the past, so the
populations in Burr 2003 and DART may not have been very dissimilar. The
main difference is that DART recruited patients immediately post MI.

When we excluded studies at moderate or high risk of bias (in
sensitivity analysis) the relative risk of death was 0.98, 95% CI 0.70 to
1.36, including only 138 deaths (see figure 2 of the long BMJ version).
Because of the low number of events, the confidence interval does not
preclude a beneficial effect, nor does it exclude a detrimental effect,
but the closeness of the point estimate to unity lends weight to the idea
that we may have overestimated the protective effect of omega 3 fats.
Also, the evidence could change rapidly when data from further RCTs at low
risk of bias become available. It is interesting to note that the point
estimate of each of the three largest studies (DART, GISSI-P and Burr
2003) lies within the confidence intervals of the sensitivity analysis –
we could be seeing natural variability around a negligible effect size.

Inclusion criteria

As stated in both the BMJ paper and the Cochrane review, the Burr
2003 paper was included before publication (Andy Ness, one of our review
authors, was working with Michael Burr and we included a manuscript which
had been submitted for publication). We also wrote to all authors of
included studies asking whether they knew of unpublished studies, and
would have included more if they had been relevant. This was stated
clearly in the review protocol.

We also stated in our published protocol that we would exclude
studies of multiple risk factor interventions unless the effect of diet or
supplementation could be separated out from the other interventions, to
ensure that any effect could be more clearly attributed to omega 3. The
Lyon Heart study was therefore excluded as the effect of alpha-linolenic
acid could not be separated from changes in the consumption of many other
foods (for example, fruit and vegetable intake rose and saturated fat and
linoleic acid intake fell in the intervention arm).(6) However, factorial
trials were included as it is clearly possible to separate out the effects
of the different interventions. GISSI-P and DART were both factorial
trials, not just Burr 2003.

It might be possible to use these factorial designs to assess the
effects of omega 3 with regard to omega 6 intake – interactions were not
seen between the dietary fat arm (which advised and achieved an increase
in the P/S ratio) and the oily fish arm (increasing omega 3) of DART.
This tends not to support the idea that omega 3 fats work better when
omega 6 intakes are lower.

Inclusion of specific cohort studies

We did include prospective cohort studies that assessed either
dietary intake or biomarkers (and used biomarker analyses for preference
where both types of analysis of a single cohort were available). The
reason we included cohorts was that adverse effects might be observed in
observational but not in relatively short-term randomised controlled
trials. Of the 41 analyses of 26 cohort studies included in our review
intake of omega 3 (varying combinations of eicosapentaenoic acid,
docosahexaenoic acid, docosapentaenoic acid, along with  linolenic
acid, supplemental fish oils, or dietary oily fish) was assessed by
dietary and biochemical means in two cohorts, dietary means only in 18,
and biochemical means only in 10. The Albert(9) analysis of the
Physician’s Health Study (PHS) was indeed one of the 5 analyses of the PHS
cohort relevant to the review which we included. The Morris 1995 paper(10)
provided the most useful data on cardiovascular events and stroke for the
PHS study and is seen in these forest plots in the review (for details
please see the references to included studies provided on the BMJ
website).

We excluded cohort studies which assessed general fish intake (rather
than oily fish or fish oil intake, or biomarkers of omega 3 intake) as we
were interested in the effect of omega 3 fats, not fish, on health. We
agree that white fish consumption may have health effects other than those
we have studied, but this was not our interest. He has conducted this
analysis and there is not much point in our repeating it.(11)

Relying on cohort studies to inform us

The problem with reliance on cohort evidence for assessing the
effects on health of omega 3 fats is that in all cohorts we examined,
where there were data on socioeconomic and lifestyle characteristics,
participants consuming more omega 3 fats (or with higher body levels) were
better off, better educated, had better social support, were more
interested in health (as evidenced by them taking more vitamin supplements
of various types), ate less saturated fats, more fruit and vegetables,
more fibre, smoked less and took more physical activity than those taking
less omega 3. Adjusting for this level of confounding is difficult and
our analysis demonstrates that adjustments are often not adequate. We
should remember that claims based on observational studies about the
health benefits of beta-carotene, folate and vitamin E (for example) have
been shown to be incorrect following randomised controlled trials.(7, 8)
While observational data are interesting clues to the health effects of
foods, we need to rely on high quality trial evidence to fully understand
the health benefits of increasing our intake of oily fish or fish oil
supplements.

Overall

Like many of you, we had anticipated that our review would provide a
simple comprehensive conclusion that omega-3’s are protective against
cardiovascular disease. In fact this is not the case. Surely, however, it
is more important to have a balanced and objective analysis of existing
evidence so that we can be certain about what we do not know and be clear
about what future trials should be designed to answer?

Our review has not questioned UK healthy eating advice suggesting
that the public eat two portions of fish per week, one of which should be
oily. This seems very sensible given the essentiality of omega 3 fats and
evidence of wide ranging benefits of fish. It also suggests that low dose
fish oil supplements in those not eating oily fish (the sorts of doses
taken to help with arthritic symptoms for example) would cause no harm.

However we are less sure about the higher, therapeutic, doses of
omega 3 - three or more large portions of oily fish each week or 1g+ of
omega 3 fats daily. The evidence here is less clear. If high doses of
omega 3 are not helpful for some people, or are even harmful for some
groups of people, it is important that we recognise this and make sure
that only the right people get the advice to take more oily fish or fish
oil supplements. We need to keep watching the evidence as it emerges and
adjust our advice as appropriate, for the best health of people
everywhere.

1. Hooper L, Thompson RL, Harrison RA, Summerbell CD, Ness AR, Moore
HJ, et al. Risks and benefits of omega 3 fats for mortality,
cardiovascular disease, and cancer: systematic review. British Medical
Journal 2006;332:752-60.

2. Hooper L, Thompson RL, Harrison RA, Summerbell CD, Moore H,
Worthington HV, et al. Omega 3 fatty acids for prevention and treatment of
cardiovascular disease. Cochrane Database Syst Rev 2004;4:CD003177.

3. Burr ML, Ashfield-Watt PA, Dunstan FD, Fehily AM, Breay P, Ashton
T, et al. Lack of benefit of dietary advice to men with angina: results of
a controlled trial. European Journal of Clinical Nutrition 2003;57:193-
200.

4. Burr ML, Fehily AM, Gilbert JF, Rogers S, Holliday RM, Sweetnam
PM, et al. Effects of changes in fat, fish and fibre intakes on death and
myocardial reinfarction:diet and reinfarction trial (DART). Lancet
1989;ii:757-61.

5. Dietary supplementation with n-3 polyunsaturated fatty acids and
vitamin E after myocardial infarction: results of the GISSI- Prevenzione
trial. Lancet 1999;354:447-55.

6. de Lorgeril M, Renaud S, Mamelle N, Salen P, Martin JL, Monjaud I.
Mediterranean alpha-linolenic acid-rich diet in secondary prevention of
coronary heart disease. Lancet 1994;343(8911):1454-9.

7. MRC/BHF Heart Protection Study of antioxidant vitamin
supplementation in 20536 high-risk individuals: a randomised placebo-
controlled trial. Lancet 2002;360:7-22.

8. Davey Smith G, Ebrahim SBJ. Folate supplementation and
cardiovascular disease. Lancet 2005;366:1679-81.

9. Albert CM, Campos H, Stampfer MJ, Ridker PM, Manson JE, Willett
WC, et al. Blood levels of long-chain n-3 fatty acids and the risk of
sudden death. New England Journal of Medicine 2002;346(15):1113-8.

10. Morris MC, Manson JE, Rosner B, Buring JE, Willett WC, Stampfer
MJ. Fish consumption and cardiovascular disease in the physician's health
study: a prospective study. American Journal of Epidemiology
1995;142(2):166-175.

11. He K, Song Y, Daviglus ML, Liu K, Van Horn L, Dyer AR, et al.
Accumulated evidence on fish consumption and coronary heart disease
mortality: a meta-analysis of cohort studies. Circulation 2004;109:2705-
11.

Competing interests:
As shown in the published paper.

Competing interests: No competing interests

07 April 2006
Lee Hooper
Lecturer
Rudolph Riemersma, Paul Durrington, Carolyn Summerbell, Helen Moore, Rachel Thompson, Roger Harrison, Helen Worthington, Nigel Capps, Shah Ebrahim, and George Davey Smith
School of Medicine, Health Policy and Practice, University of East Anglia, Norwich NR4 7TJ, UK