Monday, January 27, 2014

The Diet and Reinfarction Trial

Studies Associated with the Trial

The effect of dietary advice on nutrient intakes: evidence from the diet and reinfarction trial (DART) (1989) [1]
Diet and reinfarction trial (DART): design, recruitment, and compliance (1989) [2]
Effects of changes in fat, fish, and fibre intakes on death and myocardial reinfarction: diet and reinfarction trial (DART) (1989) [3]
Fish and the heart (1989) [4]
Haematological prognostic indices after myocardial infarction: evidence from the diet and reinfarction trial (DART) (1992) [5]
Diet and reinfarction (1994) [6]
The long-term effect of dietary advice in men with coronary disease: follow-up of the Diet and Reinfarction trial (DART) (2002) [7]
Is Fish Oil Good or Bad for Heart Disease? Two Trials with Apparently Conflicting Results (2005) [8] 

Participants and Diets 

2033 men, aged 30-69, who had previously had a heart attack, were randomised to one of 8 groups, each with a different combination of dietary advice [3].  The three dietary factors were: 

1.      “Fat advice, designed to reduce fat intake to 30% of total energy and to increase the polyunsaturated/saturated (P/S) ratio to 1.0”*
2.      “Fish advice, at least two weekly portions (200-400 g) of fatty fish (mackerel herring, kipper, pilchard, sardine, salmon, or trout)”
3.      “Fibre advice, increased intake of cereal fibre to 18 g daily” 

“Those randomized to the ‘no advice’ group were given a ‘sensible eating’ sheet which did not include advice on any of the intervention dietary components.” 

The dietary groups were similar at baseline, all smokers were advised to give up the habit and people who were overweight were put on calorie restricted diets [3] 

So there were 8 unique combinations, see the table below [3]:  

 
Overall, the fat advice group reduced fat and saturated fat and increased polyunsaturated fat , although there was some overlap [1].
 
 
The other macronutrients were similar between the groups [1].

 
However, there were some differences in food intake.  The fat advice group ended up eating more white fish, fruit and vegetables and less cakes and biscuits [1]
 
 
The fish advice and fibre advice groups had very similar diets besides fish or fibre intake (obviously) [3].  Although the fibre advice group had slightly more milk (245g vs. 226g) and spreading fats (22.3g vs. 21.3g) and less white bread and low fibre cereals (in total they had the same amount of bread and 11% more cereals) [1]

 
“Advice to alter one intervention component did not appear to interfere with intakes of other intervention components.  For example, the fat advice group had similar fibre and EPA intakes to the no fat advice group. The fibre or fatty fish advice groups had similar fat intakes to those not given this advice.” [1] 

* I like the design.  It’s efficient and interesting because it can test a few things at once and potentially find interactions between the different dietary factors 

Results 

Cholesterol levels in the fat advice vs. non-fat advice group didn’t change much (just a few %) [3]
 
 
At 2 years, the fat advice group had no difference in mortality, the fish advice group had a 29% decrease in mortality and the fibre advice group had a non-significant increase in mortality.  Interestingly, the fat advice group had slightly fewer non-fatal heart attacks and IHD events, even though deaths from IHD were basically the same, and the fish advice group had more non-fatal heart attacks even though the number of IHD events was slightly lower (not significant) [3]

 
Adjusting for some confounding variables didn’t change much [3]

 
* However, perhaps the most important thing someone can do following a heart attack is to stop smoking, as those who gave up smoking had a lower mortality than those who continued to smoke (4.1% vs. 7.9%) after 18 months (which appeared to be mediated by fibrinogen levels”) [5] 

Fat Advice 

Despite the fat advice group eating more fruits and vegetables and fewer cakes and biscuits, there was no difference in mortality between the groups, but non-significantly fewer IHD events in the fat advice group.  The researchers suggest that the lack of difference in mortality between the groups was due to there being no significant difference in cholesterol levels after 2 years [3]

Even though the no fat advice group made changes to their diet and the fat advice group didn’t have perfect adherence, I still would have expected a difference in cholesterol as that’s true of basically every trial and the experimental group in the trials I’ve looked at reduce their cholesterol  

Since the confounding variables were (yet again) in favour in the fat advice group and that there was no difference in mortality between the groups, this is actually an unfavourable result for the diet heart hypothesis

 
Fish Advice 

The fish advice group had a 29% decrease in mortality.  Cholesterol levels didn’t decrease in the fish advice group and actually slightly increased (not what you would expect).  The researchers discussed that the reduction in mortality therefore couldn’t have been due to cholesterol and instead may be due to fish oil preventing ventricular fibrillation during a heart attack, fish oil inhibiting blood clotting or something else present in lean fish [3].  However later studies found that the fish advice group had similar blood clotting [5] [6] 

* In a letter to the editor (not by the DART researchers) we get this gem of wisdom: Their data, however, do show that the "fish effect" more than doubles in patients who received additional advice on dietary fat, strongly suggesting that advice to reduce total fat intake and increase the ratio between polyunsaturated and saturated fat in the diet greatly enhanced the beneficial effect of advice on fish intake.”.  They then mention the traditional Eskimo diet as being low in SFA high in PUFA and LCO3s [4].  While the first part of the quote is accurate, what this person didn’t mention is that is only true in the context of fibre advice.  In the absence of fibre advice the no fat advice did better with or without fish.  The group that did the best was fish, no fat, no fibre (see below) 

** However, in DART 2 the researchers found that “taking fish oil capsules was associated with a higher risk of cardiac and sudden death. The adverse effects of fish or fish oil were restricted to men not taking b-blockers or dihydropyridine calcium-channel blockers, and were greater in those taking digoxin.” [8] 

 
Fibre Advice 

Contrary to observational studies, the fibre advice group had an almost significant increase in mortality (1.25 (0.99-1.65)) and IHD events (1.23 (0.97-1.57)) [3].  In the long-term follow up (below) there were more smokers in the fibre advice group, which may have been true during the trial period, but also may have been adjusted for.  Most of fibre was likely from wheat.  This website discusses the possibility of wheat promoting cardiovascular disease. 

 
Long Term Follow Up 

The trial period for DART was between 1983 and 1987, where participants had 2 years in the trial.  In February 2000 a study looked at how the groups travelled.  There were still some differences in diet between the groups from 13-15 years ago.  The fat advice group had more strokes (103%) and slightly less CHD (11%) (non-significant).  The fish advice group had an increase in total mortality at 2-5 years, but not after, and slightly less CHD (8%) and slightly more strokes (23%) (both are non-significant).  The fibre advice group had slightly more mortality (7%), CHD (11%) and strokes (34%) (all are non-significant), although did have more smokers (20.6% vs. 15.6%), which was almost significant (p=0.06) [7] 

Potential Nutrient Interactions 

If we compare the differences between the eight groups and look for interactions we get these results.  Fish advice is consistently beneficial, fibre advice is inconsistently harmful and fat advice is inconsistent (from table V, just added the other two tables) [3]: 

Mortality (%)
Fat
No Fat
Difference
(Fat - No Fat)
Fish, Fibre
8.0
12.4
-4.4
Fish, No Fibre
8.9
7.8
1.1
No Fish, Fibre
13.5
14.5
-1.0
No Fish, No Fibre
13.2
9.9
3.3

Mortality (%)
Fish
No Fish
Difference
(Fish - No Fish)
Fat, Fibre
8.0
13.5
-5.5
Fat, No Fibre
8.9
13.2
-4.3
No Fat, Fibre
12.4
14.5
-2.1
No Fat, No Fibre
7.8
9.9
-2.1

Mortality (%)
Fibre
No Fibre
Difference
(Fibre - No Fibre)
Fat, Fish
8.0
8.9
-0.9
Fat, No Fish
13.5
13.2
0.3
No Fat, Fish
12.4
7.8
4.6
No Fat, No Fish
14.5
9.9
4.6

And arranged in another way to compare groups (the right column is the average): 

Average Mortality (%)
Fat
No Fat
Fibre
8.0 (fish),
13.5
10.75
12.4 (fish),
14.5
13.45
No Fibre
8.9 (fish),
13.2
11.05
7.8 (fish),
9.9
8.85

Average Mortality (%)
Fish
No Fish
Fat
8.0 (fibre),
8.9
8.45
13.5 (fibre),
13.2
13.35
No Fat
12.4 (fibre),
7.8
10.10
14.5 (fibre),
9.9
12.20

Average Mortality (%)
Fibre
No Fibre
Fish
8.0 (fat),
12.4
10.20
8.9 (fat),
7.8
8.35
No Fish
13.5 (fat),
14.5
14.00
13.2 (fat),
9.9
11.55

From these results we can put together some observations that may suggest nutrient interactions (this is not definitive and can easily occur by chance):
 
  • No fat advice did better than fat advice without fibre advice and worse with fibre advice
  • Fibre advice was only harmful if paired with no fat advice.  If fat advice was given, then fibre advice had no effect
  • Compared to no fat advice, fat advice did better with fish advice and worse without fish advice.  However, the fish advice group had the lowest mortality of the 8 groups 

As for the mechanisms that explain the potential interactions, I’m not sure.  The fat and fish advice thing could be related to omega 3:6 ratios, but I have no good ideas on the fat and fibre relationship. 

* Chances are, being in this study was harmful to your health (no advice 9.9%, average advice 11.2%)

Sunday, January 19, 2014

The Minnesota Coronary Survey

Studies Associated with the Trial

The Minnesota Coronary Survey: composition of their diets, adherence, and serum lipid response (1975)
The Minnesota Coronary Survey: methodology and characteristics of the population (1975)
The Minnesota Coronary Survey: effect of diet on cardiovascular events and deaths (1975)
Test of effect of lipid lowering by diet on cardiovascular risk. The Minnesota Coronary Survey (1989) [1]

Participants and Diets

A total of 4393 male and 4664 female patients of several mental hospitals were put in two groups an experimental group or a control group.

“The original population was initially stratified into 512 cells on the basis of eight variables. These were: age, sex, length of stay in the hospital, weight, blood pressure, diabetes, cigarette smoking, and evidence by electrocardiogram of a previous myocardial infarction. When new subjects were admitted later, they were divided among four cells, based on only age and sex.”

The experimental diet was “a compromise between the B and C diets of the National Diet-Heart Study with target values of 45% of calories from fat, a polyunsaturated/ saturated fat (P/S) ratio of 2.5, and less than 150 mg of cholesterol daily”.  The control diet was what was normally served at the institutions.  Unfortunately I don’t have access to what the diets in the NDHS were or further dietary information most likely provided in the first paper

The replacing on SFA with PUFA in the experimental diet was achieved by “filled milk and ice cream, a whole egg substitute, soft margarine, whipped topping, filled cheese, low fat ground beef with added vegetable oil, and filled sausage products”

The only other dietary information they give is:

Experimental
Control
Fat (%)
SFA/PUFA (%)
P/S
37.8
9.2/14.7
1.6
39.1
18.3/5.2
0.3
Cholesterol
166
446

Results

Cholesterol levels decreased (by 14.5%), and were lower in the experimental group.  Adherence to the experimental diet resulted in greater reductions of cholesterol (fairly proportionally), but not much (decrease of 15.4 if < 5% meals were missed).  Triglycerides were lower in the experimental group, but increased after 3.5 years on the diet and were higher than the control group after 4.0 years on the diet (may not be significant)


Primary end-points for CHD were slightly lower in men consuming the experimental diet (10.5%), but higher in women consuming the experimental diet (31.7%).  The difference in men is probably not significant, whereas the difference in women may be, but this wasn't tested


Deaths from all-causes were similar for men, but slightly higher in women consuming the experimental diet (16.4%).


The causes of death are presented in table 7.  There doesn't seem to be a consistent pattern with large enough numbers to draw any conclusions, but we can use this to calculate CHD mortality ('atherosclerotic heart disease' + 'cardiac arrest, heart block').  We can also estimate the number of person years based on table 4 and table 5, then report CHD mortality in person years.  The results are in the table below


Men
Women
Exp
Con
Exp
Con
Atheriosclerotic heart disease
18
13
12
11
Cardiac arrest, heart block
21
21
10
9
CHD mortality
39
34
22
20
Estimated person years
2454
2357
2367
2360
CHD mortality per 1000 person years
15.89
14.42
9.30
8.48

The increase in primary end-points and deaths in women consuming the experimental diet led to a very modest (probably non-significant) total increase in primary end-points and deaths in both sexes, which you can see in the tables and these life tables.


An advantage of the trial was the large sample size, but a disadvantage was that “the mean duration of time on the diets was 384 days, with 1568 subjects consuming the diet for over 2 years”, which is only 17.3%.  The trial was not randomised, but the initial stratification and the large group size would have controlled for individual variation.  This leaves bias, particularly in the later stratification, as a potential factor.  However, potential bias would likely be in favour of the experimental group.