I agree that longer follow-ups are needed in order to properly compare the treatment modalities. If I am reading it correctly, the 3 year follow-up shows that the "statistical significance" for the difference in the primary end points disappears at 3 years. I don't see any indication in the original paper whether the patients were symptomatic or not. It is my understanding from prior papers that CABG had slightly better numbers with symptomatic disease but none with asymptomatic. So that is a serious omission from the table of characteristics. Another aspect worthy of mention is that the End Point Table shows that the largest differences are in the Safety End Points: arrhythmias and hospitalizations in the first 30 days. Both favor the PCI group and they show a far greater difference as well as larger incidence than the differences in the primary end points. Does this cancel out the rather tiny difference in the primary end points? The description of the statistical methods in the original paper is somewhat confusing and this always starts my BS meter flashing red. Since the incidence of events is rather small, I can see no objection to running a study with a true control group i.e. no intervention.
The primary outcome favored CABG at 1-year and the difference was even larger, in favor of CABG, at 3 years.
We agree that including more information on symptoms and extent of ischemia would be helpful. The authors specified in the supplement that patients had to have angina and/or evidence of myocardial ischemia to be included.
As seen with prior trials, some outcomes are worse with CABG at 30-days but long term outcomes were generally better with CABG (this applies to the populations studied and outcomes assessed). The Kaplan-Meier curves are helpful to examine to see if the curves continue to diverge on follow up.
Unless I am reading this incorrectly, the difference at 3 years is smaller--not larger.
Differences for mortality and MI are insignificant at both 1 & 3 years. The difference in the composite primary outcome is driven chiefly by the difference in revascularization procedures. This is certainly the softest end point in the composite. In my opinion, all the differences here are trivial and insignificant. Perhaps that is why the language used in the data analysis is so confusing---using p for noninferiority and words like "favoring" and "tending toward".
Great observation. The specified primary endpoint of the trial was death from any cause, myocardial infarction, stroke, or repeat revascularization. In the abstract of the 3-year follow up study, authors deviated from that primary endpoint and omitted the endpoint of repeat revascularization (the numbers you provided refer that new endpoint used).
If you look at table 3 of the 3-year follow up study, the original primary endpoint was 18.6% with PCI and 12.5% with CABG.
This highlights how follow up studies can sometimes make it more difficult to appraise trials.
I agree that longer follow-ups are needed in order to properly compare the treatment modalities. If I am reading it correctly, the 3 year follow-up shows that the "statistical significance" for the difference in the primary end points disappears at 3 years. I don't see any indication in the original paper whether the patients were symptomatic or not. It is my understanding from prior papers that CABG had slightly better numbers with symptomatic disease but none with asymptomatic. So that is a serious omission from the table of characteristics. Another aspect worthy of mention is that the End Point Table shows that the largest differences are in the Safety End Points: arrhythmias and hospitalizations in the first 30 days. Both favor the PCI group and they show a far greater difference as well as larger incidence than the differences in the primary end points. Does this cancel out the rather tiny difference in the primary end points? The description of the statistical methods in the original paper is somewhat confusing and this always starts my BS meter flashing red. Since the incidence of events is rather small, I can see no objection to running a study with a true control group i.e. no intervention.
The primary outcome favored CABG at 1-year and the difference was even larger, in favor of CABG, at 3 years.
We agree that including more information on symptoms and extent of ischemia would be helpful. The authors specified in the supplement that patients had to have angina and/or evidence of myocardial ischemia to be included.
As seen with prior trials, some outcomes are worse with CABG at 30-days but long term outcomes were generally better with CABG (this applies to the populations studied and outcomes assessed). The Kaplan-Meier curves are helpful to examine to see if the curves continue to diverge on follow up.
Primary Outcome at 1 year: PCI 10.6% CABG 6.9% Difference 3.7%
Primary Outcome at 3 Years: PCI 12% CABG 9.2% Difference 2.8%
Unless I am reading this incorrectly, the difference at 3 years is smaller--not larger.
Differences for mortality and MI are insignificant at both 1 & 3 years. The difference in the composite primary outcome is driven chiefly by the difference in revascularization procedures. This is certainly the softest end point in the composite. In my opinion, all the differences here are trivial and insignificant. Perhaps that is why the language used in the data analysis is so confusing---using p for noninferiority and words like "favoring" and "tending toward".
Great observation. The specified primary endpoint of the trial was death from any cause, myocardial infarction, stroke, or repeat revascularization. In the abstract of the 3-year follow up study, authors deviated from that primary endpoint and omitted the endpoint of repeat revascularization (the numbers you provided refer that new endpoint used).
If you look at table 3 of the 3-year follow up study, the original primary endpoint was 18.6% with PCI and 12.5% with CABG.
This highlights how follow up studies can sometimes make it more difficult to appraise trials.