Stephanie and colleagues,
Please see a response from PI of FOCUS. I think this type of response should become a norm from all investigators around the world, and we should truly appreciate Dr. Carson's engagement in this open public discourse setting an example for new bar for transparency about clinical research ( which is what EBM is all about).
Ben Djulbegovic
----
1. The protocol is on the NEJM website. I can forward it to you if you wish, but I am on a train the internet connection is poor.
2. To the skeptic...check out the protocol to see that the outcomes were defined a-priori and sample size calculations were performed for whatever is worth which I don't think is much now that the trial is done. Evaluate power by the width of confidence intervals.
3. We apriori defined secondary and tertiary outcomes and state that we would need an alpha of 0.01 (99% confidence intervals) to consider significant. It was our way to address, at least in part, multiple comparisons.
4. We designated some outcomes as secondary and others tertiary outcomes. You will note that both in the paper and protocol we identified death, MI, and unstable angina as the most important secondary outcome. This was based on our view of that is extremely important clinical event. We did not make it out primary outcome because we did not collect troponin levels in our pilot and we did not know what the event rates were likely to be. We did not want our primary outcome to be under powered.
5. We have a large number of secondary and tertiary outcomes to account for the events that we thought were clinically important. There was a debate what was most important and I came to the conclusion that the answer to that question varied by doc and speciality. We observed that orthopedics and geriatrics really liked function. Cardiology really liked MI, death. Others really liked infections, chf, disposition, length of stay, etc. In the end we tried to collect as many of the clinically important outcomes we could and presented to the medical community a complete picture.
6.The composite outcomes were chosen a-priori. MI/unstable angina and death is a typical cardiology outocmes that combines ACS and death. MI, infection, clot etc were chosen to provide overall rate of bad outcomes. For example MI, pneumonia, death were combined because more blood was hypothesized to reduce MI but might increase infection
7. We also designated MI outcome and death as well several other outcomes as secondary outcomes because we could evaluate them blinded to treatment status...and what we had good power (at least in some outcomes). In contrast, our tertiary outcomes were mostly in-hospital events that were recorded from information in the chart and were classified based on data collected by research nurses. Those same nurses were not blinded to treatment group. Also many of these outcomes were expected to be uncommon and under powered. Also, part of the motivation of this classification was for grantsmanship purposes since we made it clear to all that there were some limitations of these outcomes. Many are uncommon events with low power.
Jeff
PS...stroke is certainly important but we did not have a compelling reason to think it was related to anemia or transfusion. In contrast, we were very concerned that heart has special vulnerability to anema...thus MI was thought to be more important.
On Feb 29, 2012, at 5:11 PM, "Stephanie Chan" <[log in to unmask]<mailto:[log in to unmask]>> wrote:
Good points, Michael & Ash. Interestingly, the tertiary endpoints were prespecified, but they did seem somewhat random - so looks like it violates your last rule.
Thanks,
Stephanie
On Wed, Feb 29, 2012 at 2:04 PM, Michael Power <[log in to unmask]<mailto:[log in to unmask]>> wrote:
Stephanie
Here is a skeptic’s take on 1ry. 2ry, 3ry, …, n-ry outcomes.
I wouldn’t bother about what they are called, it may just be camouflage.
I would want to know:
If they were defined in the study’s protocol – before the study was started
if power calculations were done before the study was started (can you see the protocol to compare what they said before with what they say now?)
if they are patient-important, or proxy/surrogate variables
if they are sensible composites, or composites designed to maximize the chance of a convenient p-value
Michael
< = >
From: Evidence based health (EBH) [mailto:[log in to unmask]<mailto:[log in to unmask]>] On Behalf Of Stephanie Chan
Sent: 29 February 2012 17:45
To: [log in to unmask]<mailto:[log in to unmask]>
Subject: Re: Secondary vs. tertiary outcomes?
Thanks, Jo. Those definitions make sense, but I don't think they apply to medical trials, or at least not in the one I'm reviewing. For example, in FOCUS:
Primary outcome: death or inability to walk at 60 days
Secondary outcome: in-hospital MI, unstable angina or death
Tertiary outcome: in-hospital pneumonias, wound infections, thromboembolisms, strokes, TIAs, MIs
The power analysis was based only on the primary outcome. The only difference I can perceive between secondary and tertiary may be the clinical importance of the endpoints (though you could argue that stroke is more important than unstable angina). Tertiary outcomes seem like sort of a wastebasket of "other things that would be interesting to look at."
On Wed, Feb 29, 2012 at 12:32 AM, jo kirkpatrick <[log in to unmask]<mailto:[log in to unmask]>> wrote:
It might be different in medicine, but in psychology secondary outcomes are those that occur as a result of the primary outcomes? So tertiary outcomes are those that result from secondary outcomes. These all have equal importance and you can't have one without the others. These terms are commonly used in epidemiology, where the primary outcome is diagnosis: identifying the existence of a disease and its causes; the secondary outcome is establishing an effective treatment and the tertiary outcome is cure and eradication of the disease. What these terms actually mean in medical research and clinical trials depends on the purposes of the research but this link gives examples:
http://www.mja.com.au/public/issues/176_10_200502/geb10185_fm.html
________________________________
From: Stephanie Chan <[log in to unmask]<mailto:[log in to unmask]>>
To: [log in to unmask]<mailto:[log in to unmask]>
Sent: Tuesday, 28 February 2012, 21:36
Subject: Secondary vs. tertiary outcomes?
Hi everybody,
Could someone explain to me the difference between secondary and tertiary outcomes in a trial? I know that primary outcomes are the ones that investigators use to do their power analysis, and that secondary ones are other outcomes of interest. But what is a tertiary outcome? One that's deemed less important? I'm preparing a talk on the FOCUS trial (liberal vs. restrictive transfusion for hip fracture surgery), and they list a few prespecified tertiary study outcomes.
Thanks for your help.
Stephanie
--
Stephanie Chan, M.D.
http://www.evidencebasedmommy.com/
--
Stephanie Chan, M.D.
http://www.evidencebasedmommy.com/
--
Stephanie Chan, M.D.
http://www.evidencebasedmommy.com/
