There were 15 studies recruiting inpatients in the acute phase of myocardial infarction or unstable angina (n=10,198), and 19 studies for the analysis of outpatients with stable angina or antecedents of myocardial infarction (n=1,236).
Acute phase: the authors' reanalysis of Furberg's original meta-analysis, with correction of data errors, produced a statistically significant increase (p=0.04) in mortality of approximately 17% (95% CI for odds ratio 1.01, 1.36). A dose-response effect was observed, with significantly higher mortality rates associated with doses of 80 mg per day or more of short acting nifedipine.
Results of sensitivity analyses using Peto's odds ratios:
When an additional study assessing mortality at 6 months was included in the meta-analysis, the summary OR changed minimally (1.15) and confidence intervals crossed the central line. Results were altered for the 100 mg per day sub-group (OR 1.31: 95% CI: 0.656, 2.617), but no other dose-response effects were observed.
Exclusion of 2 studies with beta-blockers given to controls altered the summary estimate minimally. However, the CIs now cross the central line (OR 1.16: 95% CI: 1.000, 1.355).
Exclusion of a study recruiting patients who were not in the acute phase of disease reduced the global OR (1.14: 95% CI: 0.979, 1.328). The detrimental effect of the 80 mg per day dose became non-significant (95% CI 0.463, 4.474).
When all four of the above studies were excluded, and the six month results were taken into account, the summary OR for the 12 remaining trials was 1.112 (95% CI: 0.953, 1.297), and the dose-response effect disappeared.
Exclusion of five studies where data were not reported on an intention to treat basis changed the results minimally (OR 1.177 95% CI 1.010, 1.371).
Review of studies for stable angina or antecedents to MI.
Mortality data were only available for one out of 19 studies. In this study the mortality rate in the nifedipine group was 5.6% and in the placebo group 0.9% after 3 year follow-up (p=0.012, Fisher's exact test).