Pablo Lamelas –
Multiple small randomized trials evaluating the effect of complete revascularization in STEMI suggested that complete revascularization was associated with a marked reduction (50%) in cardiovascular death, as well as lower myocardial infarction and a massive reduction in repeat revascularization (summarized here). Now we finally have the COMPLETE trial published, so do we know the answer then?
In patients with ST-segment elevation myocardial infarction (STEMI) and multi-vessel disease, compared complete vs culprit-only revascularization in a large randomized trial (n=4,041) for two composite co-primary outcomes: cardiovascular death or myocardial infarction, and cardiovascular death or myocardial infarction or ischemia-driven repeat revascularization. At a median follow-up of 3 years, first co-primary 7.8% in the complete-revascularization group vs 10.5% in the culprit-lesion-only PCI group (HR 0.74; 95% CI 0.60 to 0.91; P=0.004). Second co-primary outcome 8.9% in the complete-revascularization group vs 16.7%) in the culprit-lesion-only PCI group (HR 0.51; 95% CI 0.43 to 0.61; P<0.001). Authors conclude complete revascularization was superior to culprit-lesion-only PCI in reducing the risk of cardiovascular death or myocardial infarction in post-STEMI patients.
The study was originally powered for death and myocardial infarction alone. Adding a second primary outcome may induce a positive result by chance. Therefore, authors updated the p-values for both co-primary outcomes based on the expected correlation between this outcomes and used two different p-values (both obviously lower than 5%), reducing the likelihood of a type I error. Regardless, both primary outcomes were highly-statistically significant.
Composite outcomes are usually needed, not only for increasing statistical power, but also to account for competing risks. However, we are at risk of making wrong inferences if not interpreted properly. When appraising the results of composite outcomes, Dr Guyatt and colleagues suggest the following:
- Are the components of similar importance to patients? Cardiovascular death and myocardial infarction are usually considered of similar importance to patients, based on prior literature. But I am personally unsure if this applies to current/newer definitions of myocardial infarction. The difference in myocardial infarction was mainly driven by non-ST-segment elevation myocardial infarction (STEMI events were 43 vs 53), which carries a 30-day mortality under 10% and usually requires a new PCI and then discharged home (few will require CABG since were eligible for complete revascularization through PCI after the STEMI).
- Did occur with similar frequency? The event rate in the control group was 7.9% for myocardial infarction and 3.2% for cardiovascular death. This difference generated reasonably precise 95% confidence intervals for myocardial infarction. On the other hand, the precision for cardiovascular death was not that great (0.65 – 1.32). This was expected, since was already known that myocardial infarction was supposed to happen more frequently than cardiovascular death.
- Do the components of the composite outcome share similar risk reductions? Here is where I think we need to focus. The person-years incidence of cardiovascular death was identical (1.0%), while myocardial infarction was reduced to a third (2.8% to 1.9%).
So, was correct to analyze this data as a composite of cardiovascular death or myocardial infarction? I think yes, to deal with competing risks in the primary analysis and enhance study power, when usually are both considered two patient-important outcomes of similar importance. Can we claim a win in cardiovascular death in this specific trial data? the most likely answer is no.
Repeat revascularization outcome
We debated in a prior article the limitations of such outcomes in these trials. If COMPLETE did not showed reduction in myocardial infarction and a drastic reduction in the second co-primary outcome (which included repeat revascularization) then I will be very disappointed of calling this a positive study! Is not that repeat revascularization is an outcome that does not matter, the problem is the context: if we assign people to get a PCI vs medical therapy only, is surprising that those not revascularized have more PCI in the future? of course not. Should this outcome alone guide clinical decision making? not in this context in my opinion. In DES vs BMS literature where all get the PCI to start and are blinded to stent type, then repeat revascularization is a less biased and reliable outcome, but not the setting of STEMI with residual disease.
Most of the subgroups resulted consistent, with some “interesting trends”, but one had a statistically significant p-value: the >80% stenosis visual or >60% through core-laboratory QCA suggested more benefit (interaction p-value = 0.03). This goes in the pre-specified direction, but the number of events is not balanced: the <80% subgroup had small number of events. Then, this subgroup analysis results suggests an interaction, but the imbalanced precision with few events in one of the subgroups reduces the confidence of a true interaction. Would be great to understand the rationale why 80% visual was correlated with 60% by QCA, but likely do not change much.
External validity: comparator group
Practice related to medical management of residual disease vary between centers, regions, countries, etc. This trial tested complete revascularization (plus medical treatment) vs medical treatment alone. Many centers in the world do systematic stress testing after discharge and decide revascularization based on the amount of ischemia, and the PCI may happen weeks or months after discharge. Then, the efficacy of complete revascularization vs a systematic stress testing guided revascularization may be different than a “try to stay on medical therapy alone” arm. The Kaplan-Meier within 1-year suggest a difference of new-revascularization less than 5% between study groups, so revascularization of non-culprits was uncommon. Future analysis of this trial related to this topic are needed to better understand this.
Peri-procedural myocardial infarction
When doing PCI for COMPLETE I wondered: doing a staged PCI one day after a STEMI, when troponins are peaking, may reduce the chance of detecting a peri-procedural myocardial infarction since troponins are already high. On the other hand, a planned ischemia-driven revascularization one year after elevating some troponin will be clearly visible. Given the small prognostic impact of these small peri-procedural myocardial infarctions, the impact of this likely small.
COMPLETE trial in context
Adding COMPLETE to a prior meta-analysis we published on the topic, these are the results:
What is clear is that before COMPLETE was added to the meta-analysis, cardiovascular death estimate was 50% relative risk reduction, that went to 25% (the pooled estimate diluted by a factor of 50%! that is a lot). If we exclude PRAMI trial as a sensitivity analysis, which was an early-stop, the p-value goes to 0.15.
Now comes the key question: with a p-value of 0.05 do we have the evidence to claim reduction in cardiovascular mortality? When the literature is mainly composed by small and underpowered (for cardiovascular death and myocardial infarction) trials and comes a new large trial, we maybe feeling like the “systematic IV magnesium era”, when tons of smaller prior randomized trials suggested benefit, until the very large ISIS-3 trial showed not even a hint of benefit, or even the “systematic STEMI thromboaspiration era” until TOTAL came to light (in terms of cardiovascular death). Looking for explanations for complete revascularization in STEMI, these are some that come to my mind:
- Methodological heterogeneity: since PRAMI was an early stop and CvLPRIT was originally planned as a pilot study, we may be facing a random-high finding in these trials that affect the observed estimates biased towards more benefit.
- Clinical heterogeneity: The prior trials observed higher incidence of events, suggesting a higher risk population vs COMPLETE. However, the subgroup analysis of COMPLETE (which of course have the limitation of inherently being underpowered for primary outcomes) does not confirm that higher risk characteristics is associated with higher treatment efficacy, some trends observed of course. Then, cannot rule out that higher risk populations may benefit more from this intervention based on the available data.
- Other methodological and clinical aspects, like systematic FFR use or the way of defining non-culprit lesions, may have a role, but no robust evidence to support it.
- And the last but not the least: since this intervention does reduce myocardial infarction with a reasonable safety profile, is not surprising that this could be translated in small but real lower cardiovascular death during follow-up. Then, we maybe facing a real signal, but given its mall size not clearly visible now. What I am pretty sure is that we are not facing a 50% relative reduction in cardiovascular death before COMPLETE was published.
Myocardial infarction estimates from COMPLETE are aligned with prior literature, with moderate heterogeneity among the studies.
No surprises here, as stated above.
Why do we needed a large randomized trial like COMPLETE? to simply understand the underlying truth! COMPLETE “excluded” the possibility of a 50% relative risk reduction of cardiovascular death suggested by all the prior evidence taken together. Now we are uncertain if this intervention is translated to clinically-important benefits in cardiovascular death (5 to 15% maybe?), which may be true or not, and we may have ore insight when newer trials come along.
What to do with my patient after COMPLETE? Tough question. For the repeat revascularization outcome alone, I would not change practice. The outcome driving future recommendations is myocardial infarction (NNT = 40). My first reflex would be to say yes: the body of evidence suggest a 30% relative risk reduction in myocardial infarction and suggests a small (and weak from the certainty point of view) reduction in cardiovascular death. Now we do not know if specific subgroups benefit or not, future individual-patient data analyses or longer term follow-up may bring more light to this issue. My guess is that future clinical guidelines will recommend complete revascularization after STEMI, unless a good reason for not doing so. This was already in practice in many centers around the world.
Another key implication in my opinion: Randomized trials in stable coronary disease favoured medical therapy over revascularization (exception the borderline significant long-term lower myocardial infarction of FAME II favouring PCI). COMPLETE and these other STEMI trials, which correspond to a more contemporary practice of PCI and medical therapy, favoured revascularization, setting a new precedent that PCI may have a prognostic role in “stable plaques” vs medical therapy. Few patients in clinical practice come with real chronic angina, while most of them have a form a recent-onset (a feature of plaque instability) angina. We will leave this to ISCHEMIA trial.