The Hooper et al (2015) Meta-Analysis on Saturated Fat and Cardiovascular Disease

Hooper et al recently published a new meta-analysis, this one entitled “Reduction in saturated fat intake for cardiovascular disease”.  This meta-analysis focused on the effect of reducing saturated fat (SFA), as opposed to their previous meta-analysis that looked at the effect of reducing and/or modifying fat for CVD.

This new meta-analysis included many of the original trials from the previous meta-analysis (from all three categories), but had more selective inclusion criteria than the previous, such as an intention to reduce SFA and the intervention lasting at least 24 months.  This led to the exclusion of many of the trials that were included in the previous meta-analysis

	Trials in 2012	Included in 2015	Reason for Exclusion
Modified Fat Trials	RCOT	X
	LAVAT	X
	ODHS	X
	NDHS		Follow-up less than 24 months
	MRCT	X
	MCS		Although the study proceeded for over 4 years participants (patients) came and went and mean follow-up was only 1 year
	SDHS	X
	Houtsmuller	X
Reduced Fat Trials	Ball		Study aim was to assess effects of a low-fat diet and methods state that the "nature of the fat consumed was not altered". Saturated fat content of diet was not reported.
	WINS	X
	Black	X
	Polyp Prevention		Neither mortality nor cardiovascular morbidity data available (only decided after contact with at least 1 author)
	Ley	X
	Moy	X
	BRIDGES		Follow-up less than 24 months
	PREMIER		Follow-up less than 24 months
	DO IT		Intervention aim was for a "mediterranean diet" with total fat 27 - 30%E, protein 15 - 18%E, CHO 50 - 55%E, no specific aim to reduce saturated fat (though polyunsaturated margarine given to intervention group), and intervention group saturated fat was more than 80% of that in the control.
	WHI	X
	WHEL		Study aimed to reduce total fat, but saturated fat goals were not mentioned, and saturated fat intake in the intervention group was more than 80% of that in the control (81%)
Modified and Reduced Fat Trials	DART	X
	STARS	X
	Sondergaard		Follow up less than 24 months
	MeDiets		Follow-up less than 24 months
	NDHS		Follow-up less than 24 months

This meta-analysis included the same figures as the previous one.  This means they generally included minor events (angina, surgery, increased anti-anginal treatment) and the same error in total mortality in ODHS (48 vs. 65 should be 41 vs. 55)

This meta-analysis got a very similar result to the previous one, with a significant reduction in combined CVD events and no effect for CVD mortality or total mortality (the links take you to the Forest plots)

	CVD Events	CVD Mortality	Total Mortality
Hooper, et al (2012) (modified fat)	0.82 (0.66-1.02) P = 0.073	0.92 (0.73-1.15) P = 0.46	1.02 (0.88-1.18) P = 0.81
Hooper, et al (2012) (modified and reduced fat)	0.77 (0.57-1.03) P = 0.077	0.98 (0.76-1.27) P = 0.88	0.97 (0.76-1.23) P = 0.78
Hooper, et al (2015) (reduced SFA)	0.83 (0.72-0.96) P = 0.013	0.95 (0.80-1.12) P = 0.51	0.97 (0.90-1.05) P = 0.47

The lack of effect for CVD mortality and total mortality is predictable.  The inclusion of the Finnish Mental Hospital Study (FMHS) is the major determinant of whether a meta-analysis finds a significant reduction in CHD or CVD mortality as evidenced by Skeaff & Miller and Mozaffarian et al (blog).  If FMHS is not included, as is the case with this meta-analysis, it’s very hard to get a significant effect.  Similarly almost every meta-analysis has found no effect for total mortality.  The only exception is the meta-analysis by Skeaff & Miller because they incorrectly calculated the risk ratio for FMHS using group size rather than by using person years.  If FMHS is not included or its risk ratio is calculated using person years then you get a risk ratio very close to 1.0 (blog).

The significant effect for CVD events is similar to their previous result, significant this time probably due to the inclusion of more trials under one category.  This significant result is particularly interesting because the inclusion of the Women’s Health Initiative (WHI) would be expected to strongly draw the pooled RR close to 1.0 as it did previously in the pooled result of reduced fat trials.  There are some reasons that can explain this significant result:

1) In the analysis for total mortality the WHI had 2404/4376 (55%) of all deaths and 59.6% of the weighting.  However, in the analysis for CVD events the WHI had 3445/4377 (79%) of all CVD events, but only 32.7% of the weighting.

2) The trials that drove the significant result included Houtsmuller, MRC, ODHS, STARS and LAVAT.  Neither of these trials are a particularly accurate representation of reducing SFA.  The control group in LAVAT had a vitamin E intake that was 9.4-fold lower than the experimental group and only 16% of the RDI, ODHS and STARS used a multifactorial diet intervention, and the control group in Houtsmuller likely had a very high intake of TFA.  Even though I have categorised both MRC and DART as ‘adequately controlled trials’ they aren’t perfect either as the method to reduce SFA included forbidding “…most biscuits and cakes” (MRC) and limiting “..cakes, pastries, biscuits, meat pies and pasties, crisps, chocolates and toffees” (DART), which confounds the results due to the unhealthy and/or lack of healthy components of those foods.

3) Unlike their previous meta-analysis, none of the sensitivity analyses excluded studies that “were not free of dietary differences other than fat (or unclear)”, which excluded ODHS and STARS in their previous meta-analysis.  This criteria of bias assessment wasn’t even present in this meta-analysis

So I still think that the most accurate representation of the effect of reducing SFA or replacing it with PUFA is my analysis of ‘adequately controlled trials’ (with or without the Sydney Diet Heart Study) (blog)

Finally, I am waiting for the same people who criticised the meta-analyses by Ramsden et al and Harcombe et al to also criticise this meta-analysis because it is inconsistent with the results of several meta-analyses of observational studies that find no independent association between intake of SFA and CHD (which is close enough to CVD) [1] [2] [3] [4] [5] [6]