The TRICC trial was published in 1999, and I must admit until now I’ve never looked too closely at it, having only read the abstract.
TRICC was an equivalence study. The question was not whether a Hb of 7-9 was better than 10, but whether it was equivalent. All enrolled patients had a Hb of <9g/dl and were not bleeding, although to be classified as bleeding the patient needed to drop their Hb by 3g/dl or need 3u RBC in 12 hours so ‘ooze’ was allowed. The headline results were that there was no difference in mortality, but the authors did report a significantly significant reduced hospital mortality in the restrictive group (although the confidence interval for the absolute risk was a massive -0.3 to 11.7% - more on that later).
The study was powered to be able to detect a 4% difference in absolute mortality, the reasonable argument being that less than 4% difference would be clinically insignificant. 2300 patients were needed. An interim analysis showed that the patients had a higher than predicted mortality (23% rather than the expected 20%). To detect a 4% absolute difference would therefore require more patients, as a 4% absolute difference would be a smaller relative difference (so the study would need to be more sensitive to detect it). The authors decided therefore to keep the relative risk reduction of 20% the same, which with the higher than expected mortality made the absolute risk 5.5%. It’s easier to find a difference with higher baseline incidence of an outcome (in this case death), meaning the study does not need to be as sensitive, meaning less patients are needed. The end result of all that was that 1620 patients were required, rather than the original 2300. I mention this only because when you first read the paper it sounds like the authors are pulling a fast one.
What actually happened was that the study was terminated after 838 patients were enrolled, as recruitment had ground to a halt – the study had run out of steam. The overall mortality was 21%, closer to the predicted 20% than the 23% of the interim analysis. I therefore haven’t got a clue what the final power of the study was, however I suspect it was low and therefore unlikely to find a difference, which is clearly significant in an equivalence trial. You will notice that all the confidence intervals in the results section are pretty huge, and I think this is why.
Reading between the lines, there appears to have been little enthusiasm for the TRICC trial, 30% of eligible patients were not consented due to physician refusal and patients with cardiac disease were underrepresented (despite cardiac disease not being an exclusion criteria). The underrepresentation of patients with coronary artery disease in the TRICC trial has led to some clinicians not applying the results in this group.
The spin used in the conclusions are also worth pointing out. In the results, the authors report “33% of patients in the restrictive group didn’t receive blood vs. 0% in the liberal strategy group.” They even gave a p value of <0.01 to make it look all the more impressive. The reality is that all enrolled patients had a Hb of <9g/dl and the liberal group were given blood if the Hb was <9g/dl – of course they all got blood!! In the discussion is written “[the intervention]…decreased exposure to any red cells by 33%”. If you only read the discussion, you will be misled.
The applicability of the study has also been criticised in that the blood was pre the age of leukodepletion, but regardless practice changed and we have used a transfusion trigger of 7g/dl in critical care ever since.
The TRISS trial has just published, specifically considering patients with septic shock (TRICC looked at all critical care patients). This group used leukodepleted blood, and looked for equivalence between 7g/dl and 9g/dl. The design was very similar to the TRICC trial. Enthusiasm has clearly increased, with no physician refusal pre-intervention, and a massive consent rate of 97% in eligible patients. The authors set their sensitivity (power) for the same absolute risk reduction as in the TRICC trial (20%). The mortality of patients with septic shock is higher than the TRICC population, so the absolute risk reduction would be greater (9%) and less patients were needed (a very round 1000). Unlike the TRICC trial the recruitment target was achieved.
The TRISS trial was in my opinion an excellently designed and conducted trial, with the punchline being that there was no difference in mortality between the two groups.
I think one nagging doubt that clinicians may still have regards the patient with ischaemic heart disease. Patients who were actively having an MI or unstable angina were not included in either this or the TRICC trial. The transfusion strategy for patients with active ischaemia is not something for discussion now as the evidence is contradictory and I don’t think we have a clear answer yet. In the TRISS trial, a post-hoc analysis was undertaken of the patients with myocardial ischaemia (table S14 of the supplementary appendix). Whist 2.7% of patients in the restrictive group and only 1.2% in the liberal group had myocardial ischaemia, this did not reach statistical significance (small numbers with a low baseline incidence). The definition of myocardial ischaemia was STEMI / NSTEMI or unstable angina for which the patient received treatment, a pretty robust threshold. The two groups each contained similar numbers of patients with a previous MI (33 vs 30), angina being more common in the restrictive group (26 vs 7). Whilst this trial was not designed to look at patients with coronary artery disease I still think these results are relatively reassuring.
These are only my thoughts about these two studies, I’d be really interested to read yours. Please feel free to disagree and remember no comment is a stupid one. I’ll maybe try to link this discussion into later posts about ARISE and ProMISe, waiting to see if the Surviving Sepsis Campaign can finally bring themselves to say goodbye to a beautiful relationship with Rivers.