There have been dueling schools of thought regarding atraumatic intracerebral hemorrhage: let the brain autoregulate its own blood supply and don’t artificially lower the blood pressure, or use intravenous agents to lower blood pressure because there’s evidence it decreases hematoma expansion. However, until now, there’d been no evidence that decreased hematoma size correlated with meaningful patient-oriented outcomes.
So, what are they saying about INTERACT2, the open-label, randomized trial of intensive BP control (SBP <140 mmHg) versus guideline-concordant BP control (<180 mmHg)?
@medwireNews INTERACT2 contradicts “longstanding dogma”, supports intensive BP reduction in ICH #eurostroke2013
@MedscapeNeuro INTERACT2: Intensive Blood Pressure Lowering Benefits ICH
@IctusClnic #ESCLondon2013 Surely INTERACT2 will have a great impact in blood pressure management after intracranial hemorrhage.
Pfffft.
The primary outcome was reduction in death or major disability (modified Rankin scale 3 to 6) at 90 days. Unadjusted outcome was statistically negative, 52.0% to 55.6% (OR 0.87, 95% CI 0.75 to 1.01), but favoring intensive BP control. Their secondary outcomes, which uses the conceptually messy tool of ordinal analysis, essentially magnified the effect of that 3.6% absolute difference in mRS outcomes and goes on to show that folks with less disability end up happier and more functional.
However, the baseline functional characteristics favored the intensive BP group, with median NIHSS score of 10 vs. 11. 68% of the cohort was from China – which has uncertain effects on external validity. Over seven different intravenous antihypertensives – including the most popular agent, urapidil – were used for BP lowering, further muddling precise treatment guidance. Most ICH was small volume hemorrhage, and BP treatment didn’t seem to have much different on hematoma expansion – so it’s hard to say why the intensive control group seemed to have a trend towards superiority.
And, finally, even though approximately half of the 1436 assigned to guideline-recommended treatment group had baseline systolic BP >180 mmHg, only 303 of them received an anti-hypertensive agent within 1 hour of study assignment. It might be more appropriate to describe this study as “intensive” vs. “poorly guideline-concordant” BP control – would outcomes have been more favorable if more of the guideline-concordant group actually had their systolic BP lowered below 180 mmHg?
In any event, to call this a practice-changing paradigm is a only a half-truth. It does appear safe, at least, to make a brisk and reasonable effort to lower BP in atraumatic, intracerebral hemorrhage. Whether “intensive” control is needed with a nicardipine infusion, such as in the upcoming Antihypertensive Treatment of Acute Cerebral Hemorrhage (ATACH) II trial, is still uncertain.
“Rapid Blood-Pressure Lowering in Patients with Acute Intracerebral Hemorrhage”
http://www.nejm.org/doi/full/10.1056/NEJMoa1214609
My impression is that ordinal analysis (which I still fail to understand as if no one could simply explain) is as you say a magnifier.
But it seems to blur our perception of the effect's size (when there is a significant difference).
And we need more than statistical significance, we need a clinically significant result i.e. with a decent effect size. But effect size and absolute risks are usually hard to find in contemporary studies, why is that ?
My impression is that ordinal analysis (which I still fail to understand as if no one could simply explain) is as you say a magnifier.
But it seems to blur our perception of the effect's size (when there is a significant difference).
And we need more than statistical significance, we need a clinically significant result i.e. with a decent effect size. But effect size and absolute risks are usually hard to find in contemporary studies, why is that ?
You're not the only one who has difficulty translating ordinal shift analysis into a clinically meaningful construct. Proponents of shift analysis feel it better describes the "lost" information from a dichotomous endpoint when there's an ordinal scale. The problem is, should the differences between outcomes at mRS 4-5 or mRS 5-6 be relevant? These are patients with dismal "alive but dead" outcomes, and the reason the mRS 0-2 outcome is selected is because it's the most clinically relevant.
The other problem with ordinal analysis is the kappa for mRS isn't fabulous, and there's actually a lot more variation in the data when you start looking at all those different ordinal points. The statistical outcome seems to indicate greater power to detect a difference, but, in reality, it ought to have less – it's more likely to describe results that occur by chance alone.
You're not the only one who has difficulty translating ordinal shift analysis into a clinically meaningful construct. Proponents of shift analysis feel it better describes the "lost" information from a dichotomous endpoint when there's an ordinal scale. The problem is, should the differences between outcomes at mRS 4-5 or mRS 5-6 be relevant? These are patients with dismal "alive but dead" outcomes, and the reason the mRS 0-2 outcome is selected is because it's the most clinically relevant.
The other problem with ordinal analysis is the kappa for mRS isn't fabulous, and there's actually a lot more variation in the data when you start looking at all those different ordinal points. The statistical outcome seems to indicate greater power to detect a difference, but, in reality, it ought to have less – it's more likely to describe results that occur by chance alone.