Sunday, March 1, 2015

The Diet Heart Trials

This blog post is a (rather long) summary of the main trials testing the diet heart hypothesis I mentioned earlier.  I've previously discussed each of these trials in a little more depth in separate blog posts I wrote about a year ago (see table for links)

The table below provides some basic information on the trials that may be of interest.  Most of the trials only included males with pre-existing CHD (secondary prevention).  Only FMHS and MCS included females and only LAVAT, FMHS, MCS included participants both with and without pre-existing CHD.  All the trials used a parallel design, except FMHS which used a crossover design.


Randomised
Blinding
Population*
N
Age on Entry
Follow Up
X
Single
FL
80
< 70
2.0
X
Double
DC
846
≥ 55
8.0
X
Single
FL
393
< 60
2.0-7.0
X
Single
FL
412
30-64
5.0
X
Single
FL
458
30-59
2.0-7.0

Single
MP
676/591
4178/6434**
34-64/
44-64
> 15
6.0
***
Double
MP
4393/4664
****
4.5
X
Single
FL
2033
30-69
2.0
X
Single
FL
90
< 66
3.25
* FL = free living; DC = domiciliary; MP = patient in a mental hospital
** Data for CHD events in FMHS comes from male and female patients aged 34-64 and 44-64 respectively and “whose initial electrocardiogram was free from coronary patterns”.  Whereas data for mortality comes from all patients aged > 15.  The study did not report the total number of participants and instead reported the number of participants at each hospital during each diet period.  The N used in this table came from the sum of these figures, but this figure is greater than the total number of participants as those participants who were present during both diet periods are counted twice
*** The participants in MCS were stratified rather than randomised
**** The age of participants on entry ranged from < 30 to > 70

Note: for the rest of the post unless stated otherwise figures will be compared as ‘experimental vs. control’ in that order (e.g. ‘experimental group had higher X (10 vs. 5)’).  Also, unlike other blog posts, the reference numbers will reset to 1 at each subheading.

The Rose Corn Oil Trial

Dietary Intervention

Participants were randomly allocated to a control group, an olive oil group or a corn oil group.  The control group was no advice on dietary fat.  The two oil groups were given an oil supplement of 80g/d of their respective oil, “were instructed to avoid fried foods, fatty meat, sausages, pastry, ice-cream, cheese, cakes (except plain sponge), etc” and “milk, eggs, and butter were restricted”.  Both oil groups consumed approximately 60g/d of their respective oil and had a similar calorie intake as the control group [1].

Results


Corn Oil
(n=28)
Olive Oil
(n=26)
Control
(n=26)
Relative
Risk*
Major CHD Events [1]
12
9
6
1.857
Total CHD Events [1]
15
11
11
1.266
Major CVD Events [1]
13
9
6
2.012
CHD Mortality [1]
5
3
1
4.643
Total Mortality [1]
5
3
1
4.643
* Comparing corn oil vs. control

No statistics were reported except regarding the difference in the percentage of participants who remained in the trial and were free of major cardiac events, which was lower in the corn oil group compared to the control group (52.0% vs. 75.0%, p = 0.05-0.10) (olive oil group was 57.1%) [1]

Comment

The instruction to avoid certain junk food in the oil groups would be expected to improve diet quality.  However, both oil groups had a higher incidence of major CHD events than the control group suggesting this was offset by the oil supplements.  This may suggest an adverse effect of high refined oil consumption, but also as this result is inconsistent with similar trials (MRC), it may be explained by other factors, such as poor diet quality – indicated by low protein intake – being more sensitive to nutrient dilution by high oil consumption

The Los Angeles Veterans Administration Trial

Dietary Intervention

Participants were randomly allocated to an experimental group or a control group.  The two diets were well matched for calories and macronutrient composition.  Both diets included ~40% of calories from fat, which came mostly from animal foods in the control diet and mostly from vegetable oils in the experimental diet, resulting in a higher intake of PUFA (15.6% vs. 4.9%) and a higher PUFA:SFA ratio (1.7 vs. 0.3) in the experimental diet [1] [2].

Results


Experimental
(n=424)
Control
(n=422)
Relative
Risk
Statistical Significance
Major CHD Events
[3] [4] [5] [6]
60
78
0.766
?
Participants with Major CHD Events
[3] [4] [5] [6]
52
65
0.796
P < 0.2
Total CHD Events [3] [6]
89
105
0.843
?
Major CVD Events
[3] [4] [5] [6]
85
119
0.711
?
Participants with Major CVD Events
[3] [4] [5] [6]
66
96
0.684
P = 0.01
CHD Mortality
[3] [4] [5] [6]
41
50
0.804
?
CVD Mortality
[3] [4] [5] [6]
48
70
0.682
P < 0.05
Total Mortality**
174
177
0.978
NS
Cancer Mortality [7]
31
17
1.815
P = 0.06
* The figure for total mortality is a bit ambiguous.  The original papers reported it as being 174 vs. 177 [3], 177 vs. 174 [5] and 174 vs. 178 [7].  Also, the graph in [4] suggests total mortality is higher in the experimental group, while the graph in [6] suggests it’s lower.  I’m using the figure from the monograph ([3]), which is the figure used most often in the meta-analyses

Comment

This trial has been criticised for there being more heavy smokers in the control group.  However, the differences between the groups in the incidence of CHD events persisted when stratified by cigarette use [8].

Micronutrient intake was not reported except for α-tocopherol (the main form of vitamin E), which was 9.4-fold higher in the experimental group (22.6mg vs. 2.4mg) [9] and only 16.0% of the RDA in the control group [10], indicating a very inadequate intake and a substantial confounding variable in favour of the experimental group.  While vegetable oils tend to be richer sources of vitamin E than animal fats, this is insufficient to explain the 9.4-fold difference between the groups and the very low intake of α-tocopherol in the control group.  This may be explained by the differences in food preparation, as butter for the control group was reused after cooking, whereas the researchers were careful not to do the same with the vegetable oils for the experimental group [11].

Another potential issue is that many more participants in the experimental group (117 vs. 58) withdrew from the study [3] [6], which skews the results in favour of the experimental group.  A rough adjustment for this would increase the relative risks by a factor of 1.147.

The Medical Research Council Trial

Dietary Intervention

Participants were randomly allocated to an experimental group or a control group.  The control group was instructed to eat their usual diet, but reduced intake of fried food following myocardial infarction.  The experimental group was instructed to take 85g/d of soybean oil (43g of that to be unheated and was often consumed with fruit juice) and to avoid “butter, other margarines, cooking-fat, other oils, fat meat, whole milk, cheese, egg yolk, and most biscuits and cakes”.  In addition some participants were placed on a reduced carbohydrate diet (73 vs. 90) or a very low calorie diet (5 vs. 6) to lose weight.  Compliance was high in the experimental group, who reported consuming 80g/d of soybean oil and a much higher PUFA:SFA ratio than the control group (1.8:1 vs. 1:6) [1] [2].

Results


Experimental
(n=199)
Control
(n=194)
Relative
Risk
Statistical Significance
Participants with Major CHD Events* [1]
45
51
0.860
NS
Participants with Any CHD Event** [1]
62
74
0.817
NS
CHD Mortality [1]
25
25
0.975
NS
CVD Mortality [1]
27
25
1.053
NS
Total Mortality [1]
28
31
0.881
NS
* The number of major CHD events wasn’t reported.  These figures come from the number of men who had a ‘major relapse’ during the trial
** The number of total CHD events wasn’t reported.  The figure that’s reported for this is the number of ‘first relapses’

Comment

It’s interesting that the experimental group had lower total CHD events, but not major CHD events or CHD mortality.  This result in consistent other similar studies (MCS and DART), though neither produced the same effect size for CHD events.  The authors suggested that: “One possibility is that the men in the control group themselves may have felt that they were having less active treatment, and might have been psychologically motivated both to complain more and more readily of cardiac symptoms. The other possibility, of course, is that the diet may have had a real effect in these cases of minor ischaemia.” [1].

The Oslo Diet Heart Study

Dietary Intervention

Participants were randomly allocated to an experimental group or a control group.  The control group consumed their regular diet, while the experimental group was instructed to consume a diet low in animal fats and cholesterol and high in vegetable oils.  Dietary information only came from 17 “especially conscientious” participants in the experimental group.  These participants reported that consuming 75g of soybean oil, which 28% of their calories.  Reported intake of SFA was 8.5%, PUFA was 20.7% and the PUFA:SFA ratio was 2.44 [1] [2]

Results


Experimental
(n=206)
Control
(n=206)
Relative
Risk
Statistical Significance
Major CHD Events [1]
70
91
0.769
?
Participants with Major CHD Events [1]
56
70
0.800
?
Total CHD Events [1]
80
120
0.667
?
Participants with any CHD Event [1]
64
90
0.711
P = 0.011
CHD Mortality [1] [2]
37
50
0.740
?
CVD Mortality [2]
38
52
0.731
P = 0.09
Total Mortality [2]
41
55
0.745
P = 0.13

Comment

While not presented in the papers that are readily accessible, the dietary intervention was highly multifactorial in favour of the experimental group: (1) the experimental group was instructed to replace meat and eggs with fish and was supplied sardines canned in cod liver oil, which greatly increased LCO3 and vitamin D intake; (2) the experimental group were encouraged to eat more nuts, fruits, and vegetables and to restrict their intake of refined grains and sugar; and (3) in the control group ~6.9% of total calories came from partially hydrogenated fish and vegetable oil margarines, that are associated with CHD, and were entirely restricted in the experimental group [3].  The Hooper meta-analysis also flagged ODHS as not being “free of other dietary differences besides fat” [4].

The Sydney Diet Heart Study

Dietary Intervention

Participants were randomly allocated to an experimental group or a control group.  The experimental group was advised to reduce SFA to 10% TC, cholesterol to 300mg or less and increase PUFA to at least 15% TC, and the participants achieved those targets.  The control group was given no dietary advice except to reduce calories if thought to be overweight, but made similar dietary changes as the experimental group, though to a much lesser extent as PUFA intake was higher in the experimental group (15.1% vs. 8.9%) and SFA intake was lower (9.8% vs. 13.5%) [1].

Results

Only the breakdown of total mortality by groups was initially published and Ramsden, et al later recovered data on the SDHS regarding the causes of death [2].  There is no data on CHD or CVD events


Experimental
(n=221)
Control
(n=237)
Relative
Risk
Statistical Significance
CHD Mortality [2]
36
24
1.609
P = 0.04
CVD Mortality [2]
38
26
1.567
P = 0.04
Total Mortality [1]
39
28
1.494
P = 0.051

Comment

Many of the participants made diet and lifestyle changes following infarct, which led the authors to conclude “that because of multiple changes in lifestyle men who have had myocardial infarction are not a good choice for testing the lipid hypothesis” [1].  While lifestyle changes can be a confounding variable, this is less of a concern in larger randomised trials.  Even though the prevalence of smoking and physical activity in each group was not reported, given that the groups were similar at baseline and lost a similar amount of weight between infarct and starting the trial (3.3kg vs. 2.7kg) [1], it can reasonably be assumed that the degree of lifestyle change was similar in each group.  There was also a large enough difference in SFA and PUFA intakes (PUFA:SFA ratio = 1.7 vs. 0.8)) to make this trial a meaningful test.

TFA intake was not measured and may have been a confounding variable in this trial, but whether TFA intake in the experimental group was higher, lower or the same as the control group is unknown and debatable.  It can be argued that TFA intake may have been higher in the experimental group as the experimental was provided with margarine that was high in TFA.  It can also be argued that TFA intake may have been lower in the experimental group as only the experimental group was advised to restrict intake of common margarines and shortenings, which are major sources of TFA [2].

The Finnish Mental Hospital Study

Dietary Intervention

This trial used a cross-over design.  During the first six year period participants in hospital N were served the experimental diet while participants in hospital K served the control diet, during the last six year period the diet was reversed such that participants in hospital N were served the control diet while participants in hospital K served the experimental diet.  The experimental diet was designed so that a large part of SFA was replaced with soybean oil [1].  Consequently the experimental group had a lower intake of SFA (27.3g vs. 54.7g) and higher intake of PUFA (40.5g vs. 13.6g).  There were other dietary differences though.  The experimental group had a lower intake of sugar (74g vs. 83g) and common margarine (1g vs. 12g), a major source of TFA [2].  Consequently, TFA intake has been estimated to be substantially lower in the experimental group (0.1% vs. 1.5%) [1]

Results

The closest thing to CHD events is the development of abnormal ECG patterns in participants with normal ECG patterns whose initial ECG pattern was not abnormal.  Major CHD events includes ‘major ECG change or coronary death’.  Total CHD events also includes ‘intermediate ECG change’ ‘CHD events’ and mortality rates are measured in 1000 person years and mortality rates are also age adjusted.

Men
Experimental
(1000 PY)
Control
(1000 PY)
Relative
Risk
Statistical Significance
Major CHD Events [2]
4.18
12.69
0.330
P = 0.001
Total CHD Events [2]
13.54
24.32
0.557
P = 0.008
CHD Mortality [3]
6.61
14.08
0.469
P < 0.002
CVD Mortality [3]
11.53
18.97
0.608
?
Total Mortality [3]
34.84
39.50
0.882
NS

Women
Experimental
(1000 PY)
Control
(1000 PY)
Relative
Risk
Statistical Significance
Major CHD Events [4]
1.97
5.00
0.393
P = 0.10
Total CHD Events [4]
25.04
39.41
0.635
P = 0.04
CHD Mortality [3]
5.21
7.90
0.659
P < 0.1
CVD Mortality [3]
10.58
12.32
0.859
?
Total Mortality [3]
30.87
29.01
1.064
NS

Comment

Besides the unintended dietary differences, the lack of adequate randomisation resulted in a number of confounding variables in the trial, though many of those were quite minor.  However, a notable confounding variable is that the use of a cardiotoxic medication (thioridazine) was lower in the experimental group in hospital N (0.82 vs. 1.79) and higher in the experimental group in hospital K (0.43 vs. 0.14), but was used less overall in the experimental group (0.63 vs. 0.97).  It could be argued that the substantially higher TFA consumption by the control group in hospital K (0.2% vs. 2.0%) and substantially higher use of cardiotoxic medication by the control group in hospital N can explain the differences in CHD events and CHD mortality, or at the very least explain how this study is an outlier.

The Minnesota Coronary Survey

Dietary Intervention

Participants were stratified to an experimental group and a control group.  The control diet was what was normally served at the institutions.  The experimental diet “represented a compromise between the B and C diets of the National Diet-Heart Study”, with targets of 45% fat, PUFA:SFA ratio of 2.5 and <150mg cholesterol.  The experimental diet ended up having a similar fat intake (37.8% vs. 39.1%), was lower in SFA (9.2% vs. 18.3%) and higher in PUFA (14.7% vs. 5.2%) [1].

Results

CHD events and mortality rates are measured in 1000 person years.  CHD and CVD mortality isn’t reported in this way, but can be calculated with the data that was published

Men
Experimental
(1000 PY)
Control
(1000 PY)
Relative
Risk
Statistical Significance
Major CHD Events [1]
28.1
31.4
0.895
?
CHD Mortality [1]
15.9
14.4
1.102
?
CVD Mortality [1]
37.5
39.9
0.940
?
Total Mortality [1]
64.4
64.9
0.992
?

Women
Experimental
(1000 PY)
Control
(1000 PY)
Relative
Risk
Statistical Significance
Major CHD Events [1]
26.2
19.9
1.317
?
CHD Mortality [1]
9.3
8.5
1.097
?
CVD Mortality [1]
27.5
26.7
1.029
?
Total Mortality [1]
46.9
40.3
1.164
?

Comment

Without access to the dietary information it is difficult to know what the participants actually ate and whether there were any confounding variables related to diet.  As the results for men are similar to MRC and DART, two trials with minor confounding variables favourable to the experimental group, it’s unlikely that there are any substantial confounding variables in MCS.  Similar to FMHS, an issue of this trial was that there was a high turnover rate of participants, such that the average time a participant was in the study was 384 days and only 17.3% of the participants were in the study for over 2 years.

The Diet and Reinfarction Trial

Dietary Intervention

Participants were randomly allocated to one of eight groups, each with different combination of dietary advice.  There were three dietary advices were to: (1) reduce fat intake to 30% of total calories and increase the PUFA:SFA ratio to 1 (fat advice); (2) include at least two portions of fatty week per week (fish advice); and (3) increase intake of cereal fibre to 18g/d (fibre advice) [1].  The fat advice group reported a lower intake of total fat (31% vs. 35%) and SFA (11.2% vs. 14.9%) and a higher PUFA:SFA ratio (0.85 vs. 0.45).  Intake of other macronutrients was similar, although, the fat advice group reported a slightly higher intake of white fish, fruit and vegetables and a slightly lower intake of cakes and biscuits [1].

Results
  

Experimental
(n=1018)
Control
(n=1015)
Relative
Risk
Statistical Significance
Major CHD Events [2]
132
144
0.914
NS
CHD Mortality [2]
97
97
0.997
NS
CVD Mortality [3]
101
100
1.007
NS
Total Mortality [2]
111
113
0.979
NS
* Adjusting for confounding variables had little effect on the results, only moving the RR for total mortality from 0.97 to 1.00 [2].

Total mortality was reported for each group, which enables the possibility of exploring nutrient interactions.  Some patterns emerged when comparing all eight groups.  Fat advice had lower mortality than no fat advice when combined with fish advice (8.45% vs. 10.10%), but had higher mortality when combined with no fish advice (13.35% vs. 12.20%).  Fat advice had lower mortality than no fat advice when combined with fibre advice (10.75% vs. 13.45%), but had higher mortality when combined with no fibre advice (11.05% vs. 8.85%) [2].

Comment

DART had a modestly multifactorial diet intervention and the results are very consistent with similar studies (MRC and MCS).  While “there were no significant interactions between the three dietary interventions”, the results are interesting and could be explored in future research.

The St. Thomas Atherosclerosis Regression Study

Dietary Intervention

Participants were randomly allocated to a usual care group, a diet group and a diet + cholestyramine group.  The dietary intervention was intended to be low in fat (27%), SFA (8-10%) and cholesterol (100mg/1000kcal), and high in PUFA (8%) and soluble fibre mainly from pectin [1].  However, the dietary intervention is also based on two other papers [1] which promoted a multifactorial diet such as also increasing whole plants foods [2] [3] and severely restricting commercial baked goods [2].  Participants in the diet group who were overweight (BMI>25) were also put on a low calorie diet (1,000-1,200kcal/d) to achieve a BMI of 25, which may explain how energy intake was lower in the diet group (8610kJ vs. 9135kJ) and was likely responsible for the weight loss in the diet group (-2.9kg vs. 0.3kg, p = 0.001) [4].  The diet group reported a lower intake of SFA (8.9% vs. 17.1%) and TFA (2.47g/d vs. 4.36g/d) and a higher intake of PUFA (7.3% vs. 4.7%) and long chain omega 3 (0.48g/d vs. 0.24g/d) and fibre (27.9g vs. 18.2g) [4] [5]

Results


Diet
(n=28)
Usual Care
(n=27)
Relative
Risk
Statistical Significance
Major CHD Events [1]
2
5
0.415
?
Total CHD Events* [1]
3
9
0.346
?
Total CVD Events* [1]
3
10
0.311
P < 0.05
CHD Mortality [1]
1
3
0.346
?
Total Mortality [1]
1
3
0.346
?
* Includes surgery, but not angina

Comment 

The diet intervention was multifactorial which makes it difficult to determine which dietary factors were responsible for the improvements.  As PUFA intake wasn’t substantially different between the groups (7.3% vs. 4.7%), it should certainly not be concluded that such a small increase in PUFA was responsible for the impressive result in STARS.