A re-analysis of six randomized controlled trials substituting polyunsaturated fat for saturated fat. Switch the effect measure and model, then include or exclude trials to probe how robust the pooled estimate is.
Effect measure
Model
Forest plot
Risk ratio · 95% CI · Random effects
Trial
RR (95% CI)
Weight
DART
0.91 (0.73–1.14)
27.7%
LA Veterans
0.74 (0.53–1.03)
14.5%
Minnesota CS
1.08 (0.84–1.37)
23.7%
MRC Soya
0.86 (0.61–1.22)
13.1%
Oslo Diet-Heart
0.75 (0.57–0.99)
20.1%
STARS
0.41 (0.09–1.96)
0.8%
Pooled (random)
0.87 (0.76–1.00)
100%
0.1
0.2
0.25
0.33
0.5
0.7
1
1.5
2
◀ Favours PUFA
Favours control ▶
Pooled estimate
0.87
95% CI 0.76 – 1.00 · p 0.052
6
trials
897
CHD events
12,796
participants
Heterogeneity
16%
I² · low
0.005
τ² between-trial var.
5.95
Cochran Q · df 5
0.310
Q p-value
Across 6 trials, replacing saturated with polyunsaturated fat is associated with a 13% reduction in coronary events (risk ratio 0.87, 95% CI 0.76–1.00). The interval crosses the null, and between-trial heterogeneity is low (I² = 16%).
Trial data
Incl.
Trial
PUFA arm
Control arm
RR
95% CI
Weight
DART
132 / 1018
144 / 1015
0.91
0.73 – 1.14
27.7%
LA Veterans
53 / 424
71 / 422
0.74
0.53 – 1.03
14.5%
Minnesota CS
131 / 4541
121 / 4516
1.08
0.84 – 1.37
23.7%
MRC Soya
45 / 199
51 / 194
0.86
0.61 – 1.22
13.1%
Oslo Diet-Heart
61 / 206
81 / 206
0.75
0.57 – 0.99
20.1%
STARS
2 / 27
5 / 28
0.41
0.09 – 1.96
0.8%
Effects pooled by inverse-variance weighting; random-effects model uses the DerSimonian–Laird estimator of between-trial variance. Ratio measures are pooled on the log scale. Risk difference is shown per 100 participants. Data: six randomized trials substituting polyunsaturated for saturated fat, coronary-heart-disease events.