Review of the ISIS-4 Trial

A randomized factorial trial assessing early oral captopril, oral mononitrate, and intravenous magnesium sulphate in 58,050 patients with suspected AMI

Lancet 1995;345:669-685.

Background The premises for ISIS-4 were similar to GISSI-3 with the notable exception of adding an arm to test the effect of magnesium in acute myocardial infarction.

Up to this point magnesium was commonly used in patients with acute MI. It was suspected to limit infarct size and reduce arrhythmic events and death; however, it was backed only by animal experiments and small trials in humans that were underpowered to test for realistic differences in mortality. A meta-analysis of such trials was cited in the ISIS-4 manuscript, which reported a 50% relative reduction in mortality! You read that correctly.

ISIS-4 was undertaken to test 3 distinct hypotheses: that 1 month of Captopril , 1 month of isosorbide mononitrate, and 24 hours of intravenous magnesium reduced mortality in patients presenting with definite or suspected AMI.

Patients Patients were eligible if they were thought to be within 24 hours of the onset of symptoms of suspected or definite AMI (with or without ECG changes) and if they had no definite contraindications to any of the study treatments. Contraindications were not specified by protocol but by the responsible physician but standard contraindications were suggested like conditions associated with a high risk of adverse events like cardiogenic shock or persistent hypotension (SBP <90-100mmHg) or conditions associated with only a small likelihood of worthwhile benefit, such as a high-risk of death from other life-threatening conditions.

Baseline characteristics The majority of patients were men (74%) under the age of 70 years (72%). Eighty percent of patients had ST elevation on their presenting ECG and 40% were within 6 hours from symptom onset. Only 2% of patients had SBP <100 mmHg; however, 14% had clinical heart failure. This represents a departure from the early beta blocker trials which excluded patients with clinical heart failure at study entry.

Procedures The study used a 2x2x2 factorial design, resulting in 8 treatment groups:

1. Captopril, placebo nitrate, control magnesium (C alone)

2. Captopril, placebo nitrate, magnesium (C + Mg)

3. Captopril, nitrate, control magnesium (C + N)

4. Captopril, nitrate, magnesium (C + N + Mg)

5. Placebo control, placebo nitrate, control magnesium (nil)

6. Placebo control, placebo nitrate, magnesium (Mg alone)

7. Placebo control, nitrate, control magnesium (N alone)

8. Placebo control, nitrate, magnesium (N + Mg)

Study treatments were to be started within the first hour of lytic treatment, if lytics were indicated, and immediately otherwise.

Patients randomized to captopril received an initial dose of 6.25 mg, 12.5 mg 2 hours later, 25 mg 10-12 hours later and then 50 mg twice daily for 28 days.

Those randomized to isosorbide mononitrate received an initial dose of 30 mg, 30 mg 10-12 hours later and then 60 mg each morning for 28 days.

Patients randomized to magnesium received 24 hours of intravenous magnesium sulphate with an 8 mmol initial bolus injection over 15 minutes followed by 72 mmol in about 50 ml infused over 24 hours.

Placebo captopril and isosorbide mononitrate were utilized but open control was used for magnesium because flushing and other cutaneous signs and symptoms from the initial bolus were thought likely to “unblind” the active treatment.

Endpoints The primary endpoint of the trial was intended to be vascular mortality within the first 5 weeks (35 days) but the investigators ended up reporting all-cause mortality since non-vascular mortality was rare and divided evenly between groups and did not impact the main results.

A sample size of 40,000 patients was the original aim; however, sufficient treatment was made available for the randomization of up to 60,000.

The primary analyses were captopril vs placebo (half of patients in each group received nitrate and magnesium), isosorbide mononitrate vs placebo (half of patients in each group received captopril and magnesium), and magnesium vs control (half of patients in each group received captopril and nitrate).

Results 58,050 patients were randomized from 1086 hospitals in 31 countries. AMI was confirmed in 92% of all randomized patients. The use of non-study treatments was evenly balanced between groups with 94% receiving antiplatelet therapy, 70% fibrinolytic therapy and 9% intravenous beta-blockade. 

Fidelity with study treatment was high with 81% of patients allocated captopril discharged on the drug, 83% of patients allocated isosorbide mononitrate were discharged on the drug and 88% of patients allocated magnesium completed the full infusion.

Compared to placebo, captopril significantly reduced all-cause mortality (OR 0.93; 7.19% vs 7.69%; 95% CI 0.87-0.99). There was no significant treatment effect heterogeneity observed for any of the captopril treatment subgroups (e.g., in combination with the other study treatments or not). The event rates for the captopril alone vs nil groups were 6.9% vs 7.8%, respectively. As anticipated, captopril increased hypotension in hospital requiring termination of study drug.

Inspection of various high-risk subgroups showed preservation of captopril’s treatment effect. Captopril reduced mortality in patients with anterior STEMIs (8.5% vs 9.8%), heart failure (14.5% vs 16.0%), heart rate >100 bpm at entry (13.7% vs 14.8%) and SBP 100-104 mmHg (7.5% vs 9.4%); however, mortality was increased in patients with SBP <100 mmHg (14.2% vs 12.4%) but this was a small subgroup with a difference of only 13 events. It is also noteworthy that there was no benefit observed in patients >70 years of age (14.8% vs 14.7%). This was a large subgroup that accounted for 28% of patients and 55% of all deaths.

Nitrate use did not reduce all-cause mortality compared to placebo (OR 0.97; 7.34% vs 7.54%; 95% CI 0.91-1.03). There was no significant treatment effect heterogeneity observed for any of the nitrate treatment subgroups (e.g., in combination with the other study treatments or not). The event rates for the nitrate alone vs nil group were 7.6% vs 7.8%, respectively. Unlike GISSI-3, no synergistic effect was observed when combined with captopril. Similar to captopril, it was associated with significant increases in hypotension requiring termination of the study drug. Subgroup analyses did not yield any surprising results.

In a surprise finding, magnesium increased all-cause mortality compared to placebo (OR 1.06; 7.64% vs 7.24%; 95% CI 1.0-1.12) and the effect was consistent across the magnesium treatment subgroups. No significant differences were identified among patient subgroups.

Conclusions The ISIS-4 trial demonstrated that early initiation of captopril reduced death at 5 weeks in patients with definite or suspected AMI and was associated with an NNT of approximately 200 patients. The relative treatment effect was preserved in various high-risk subgroups resulting in lower NNTs, particularly patients with anterior STEMI’s, evidence of hemodynamic instability (elevated HR and reduced SBP) and congestive heart failure. However, caution is urged in translating these results to patients >70 years of age who did not benefit despite their significant numbers and contribution to events in the trial. These results are in contrast to GISSI-3, which prespecified a subgroup analysis in elderly patients and found a statistically significant reduction in the primary combined endpoint and reduction in the rates of death (no statistics provided).

In the discussion of ISIS-4 the authors provide results from a meta-analysis involving 15 trials of ACE inhibitors on short-term mortality when started early in AMI. This includes 11 small trials, and 4 larger trials; GISSI-3 and ISIS-4 are the largest and presented in this section; CONSENSUS-II and CCS-1 were smaller but still provide important data. All combined, early ACE inhibition reduced short-term mortality and the margin was very similar to that found in ISIS-4 and GISSI-3 (7.27% vs 7.73%; p=0.006).

The ISIS-4 trial did not demonstrate a benefit for isosorbide mononitrate. Again, in the discussion, the authors provide results from a meta-analysis involving 22 trials, including 20 small trials as well as GISSI-3 and ISIS-4. When all results are combined, a statistically significant reduction in death is found for nitrates; however, significant treatment effect heterogeneity is demonstrated between the small trials and GISSI-3 and ISIS-4, which eliminates any confidence in the overall result. This demonstrates the importance of basing clinical translation on large trials, appropriately powered to detect differences in the endpoints of interest; otherwise, there is a good chance of being fooled by false positive findings.

Finally, ISIS-4 conclusively shattered all delusions related to the efficacy of magnesium in AMI and effectively represented a reversal in standard practice. The authors present a meta-analysis involving 9 small trials, a larger but still underpowered trial called LIMIT-2, and ISIS-4. Extreme heterogeneity is identified between small and large studies and the treatment effect goes from a very large benefit in small studies to a likely small harm in the one appropriately powered trial.

The story of magnesium in AMI is an excellent demonstration of positive outcome bias in medical research. It highlights how the published literature, especially for small, underpowered studies is more likely to contain studies reporting positive treatment effects but these should not be trusted. Multiple mechanisms for positive outcome bias exist including reporting bias on the part of investigators (i.e., the “file drawer effect”) and biases on the part of journal reviewers and editors to accept and publish positive studies compared to those that report negative findings.