Saturday, February 14, 2015

Intro to the Diet Heart Trials

The lipid hypothesis led to the development of cholesterol lowering drugs and cholesterol lowering diets as a means to try and reduce the incidence of coronary heart disease (CHD).  The efficacy of cholesterol lowering diets (low SFA, high PUFA, low dietary cholesterol) were tested in several clinical trials, many of which took place approximately 50 years ago.  Five main meta-analyses have pooled the results of these clinical trials, but have come to different conclusions (see table below).

CHD Events
CHD Mortality
Total Mortality
0.83 (0.69-1.00)
P = 0.073
0.84 (0.62-1.12)
P = 0.335
0.88 (0.76-1.02)
P = 0.005
0.81 (0.70-0.95)
P = 0.008
0.80 (0.65-0.98)
P = ???
0.98 (0.89-1.08)
P = ???
1.13 (0.84-1.53)
P = 0.43
1.17 (0.82-1.68)
P = 0.38
1.16 (0.95-1.42)
P = 0.15
0.78 (0.65-0.93)
P = 0.005
0.81 (0.64-1.03)
P = 0.08
0.92 (0.80-1.06)
P = 0.25
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
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
0.86 (0.69-1.07)
P = ???
* Ramsden, et al categorised trials as replacing SFA with either just omega 6 PUFA (n-6) or both omega 6 and omega 3 PUFA (n-6+3) and performed a separate analysis for each category
** Hooper, et al categorised trials modified fat (replacing SFA with MUFA and/or PUFA) or both modified fat and reduced fat and performed a separate analysis for each category.  The primary outcome measures in Hooper, et al were cardiovascular events, cardiovascular mortality and total mortality (rather than CHD events and CHD mortality)
*** Chowdhury, et al did not analyse CHD mortality or total mortality

All the meta-analyses, except Ramsden, et al (n-6), found a significant (P < 0.05) or near significant (P < 0.10) reduction in CHD events in the order of approximately 20%.  Only Mozaffarian, et al and Ramsden, et al (n-6+3) found a significant or near significant reduction in CHD mortality.  And only Skeaff & Miller found a significant reduction in total mortality, with the hazard ratios in other meta-analyses being very close to 1.0 on average.

The different results between the meta-analyses is partially due to differences in the trials each of them included.  There are nine trials that have been included in at least two of the meta-analyses and only five of the nine trials were included by all of them.  In addition Ramsden, et al and Hooper, et al performed separate analyses based on their categorisation of the trials.  The trials, their inclusion in meta-analyses and how they are categorised in Ramsden, et al and Hooper, et al is shown in the table below.

* Categorised as an omega 6 (6) or omega 6+3 trial (6+3)
** Categorised as a fat modification (M) or fat modification and reduced fat trial (M,R)
*** Hooper, et al was the only meta-analyses that included other trials not listed in this table.  They included four small trials, a feasibility study (The National Diet Heart Study), two trials testing the effect of a Mediterranean diet (The MeDiet and one by Sondergaard, et al), and one which tested the effect of linoleic acid on complications of diabetes (Houtsmuller, et al)


  1. I notice that the Hooper analysis may have some errors:

    It seems that the figures for STARS should be 3/27 vs. 10/28. Do you know how they arrived at 8/27 vs. 20/28 for CV events?

    For MRC, heart attacks should be 40/199 vs. 39/194, not 39/199 vs. 40/194.

    The control group in that Houtsmuller trial seems to be not using butter:

    "The fiber content was similar in both groups as was the cholesterol content, being 88 rag/1000 Kcal in both groups except for 4 patients of group I who preferred butter over saturated margarines"

    Saturated margarines? Trans fat?

    1. I agree, Hooper has errors, and so do the others. I'll put up a post on this shortly.

      Their figures for STARS seems to come from CVD events (including surgery) + 'requiring increased anti-anginal treatment' (5 vs. 10; see page 567 Took me a while to figure that one out.

      Regarding Houtsmuller, that would be my guess too, and it would be consistent with the very large difference between the groups, which I don't think anyone would have expected from simply replacing SFA with LA

  2. Their figures for STARS seems to come from CVD events (including surgery) + 'requiring increased anti-anginal treatment' (5 vs. 10; see page 567 Took me a while to figure that one out."

    I would have never figured that out! I really don't understand how those are cardiovascular events.

    1. I suppose one could argue that 'requiring increased anti-angina treatment' is indicative of events of angina, which can be considered a CHD event, but not a major CHD event.

      I don't really like it either, and non-major events as well, because just like using surgery as a CHD event, it may be influenced by insufficient participant and/or researcher blinding.

      When I was trying to understand some of the odd figures the meta-analyses used, I mainly focused on what numbers (or combinations of numbers) made sense.