Hi Paul,
You bring up a few points that we should discuss further. First, let me clarify that my point is that it's not clear what they mean by 'sequentially'. Do they mean in an alternate fashion or using further coin tossing. If we make an educated guess that it's means the former, then it's an inadequate form of randomization. All I'm saying is that we have to be careful about making assumptions.
There are many ways to randomize and flipping coins is only one way. You can even use it to theoretically randomize patients in pairs, but this would be a very interesting study design. If you flip coins and include two patients into the trial at the same time, they will still be randomized because before the flip of the coin you have no idea where either the first will get the intervention or the control. Think of it like when a football team flips a coin to pick which direction they will be playing or who gets the ball first. Before the toss, neither team knows if they will get heads or tails. But once the flip takes place both teams are allocated to sides at the same time.
As for random sequence generation and allocation concealment, they do go hand in hand, but are two separate entities per se. You can have a perfectly well conducted randomization sequence without any form of allocation concealment and you can have allocation concealment without any randomization. For example, I can prepare a computer-generated randomization sequence with random block sizes, stratification, etc. and then hang the randomization sequence on the hospital ward wall so that people know how to assign treatments. In this case, the randomization sequence is sound, but there is absolutely no allocation concealment. The opposite can also be true (but unlikely). I can prepare dark, sealed, sequentially numbered envelopes and put intervention instruction in all the odd numbered envelopes and comparison instructions in all the even numbered envelopes. You can guess what the next allocation will be, but you won't be sure until you open the envelope.
Final point on 'known' and 'unknown' variables. While randomization helps to distribute the variables among the intervention and control arms, there is a minimal number of individuals needed to be randomized before you can confidently assume that the variables have been distributed evenly by 'chance'. For example, card counting became a great way to 'beat the house' in card games because you could calculate the probability that the variables (e.g. certain cards) still remain in the dealers hand compared to the ones already on the table. Today, they use several decks in the same randomization shuffle to make this less likely to work. In clinical research, we often tend to see smaller trials having a higher probability of unequal distributions by chance. I have personally come across this several time while conducting systematic reviews.
Ahmed
P.S. I'm not a gambler but I read an interesting article about the evolution of random shuffling of cards in casinos to prevent card counting.
-----Original Message-----
From: Evidence based health (EBH) [mailto:[log in to unmask]] On Behalf Of Paul Elias
Sent: Tuesday, July 02, 2013 6:29 AM
To: [log in to unmask]
Subject: Re: randomisation
There are several issues here I think (and only my thoughts):
1.) the provided information is unclear as to the sequence generation so we cannot really say
2.) if at any stage, the central tenet of randomization which is unpredictability of future allocation, is breached, then the sequence generation is inadequate
3.) the goal of randomization is to ensure balance of known and unknown prognostic factors so that the effect can be allotted to 'only' the treatment under consideration...the way it is described below, we cannot say prognostic factors are balanced (at least thats how I read it)
4.) it is critical to understand what the allocation concealment strategy was for if it were deficient, or there were none, then no matter how good the sequence generation strategy (randomization) was, the randomization has failed...they go hand in hand
5.) the way it is described thus far, it appears (unless one can read the methods) that this was a form of minimization (adaptive randomization), so a deterministic form of randomization where the first entrant is randomized adequately where there is an equal probability of being assigned to either group and there is no predictability, but the other subsequent entrants to the trial are assigned based on prior assignment to achieve some form of prognostic balance. The researcher is guiding entrants to certain groups. The draw back with this form of randomization as you know (from what I understand) is that there is a form of determinism and thus one is violating the central tenet of randomization this being unpredictability (increased selection bias risk). Once you guide or determine the next assignment (e.g it appears they did to the AB group) and breach the 'no predictability', then one cannot say this is suitably randomized. To say assigned 'sequentially' is not giving us enough information, and could mean one goes to group 1 the other to group 2 then the third to group 1 and the fourth to group 2 and so on..this is fraught with predictability. Thats how I read it.
this is a very interesting question and the responses are very informative..thanks for sharing.
Best,
Paul E. Alexander
--------------------------------------------
On Tue, 7/2/13, Dr. Ahmed M. Abou-Setta <[log in to unmask]> wrote:
Subject: Re: randomisation
To: [log in to unmask]
Received: Tuesday, July 2, 2013, 2:52 AM
I will have to disagree here... I'm
not saying that this is a correct form of randomization, but rather it is an unclear form of randomization until we know more.
The authors stated "We assigned participants to two groups by randomized sequential allocation. The first patient was randomly assigned to the LC group by tossing a coin, the second patient to the AB group, and the other patients were assigned to the two groups, sequentially." From this we can tell the following:
A) 1st patient was properly randomized.
B) 2nd patient was not quasi-randomized - unless they entered two patients in the study at the same time.
C) Every patient after that may or may not have been properly randomized depending on whether or not they kept flipping coins or whether they used alternate randomization.
This is the gray area where published protocols and contacting authors for additional information/ clarification comes in handy. In this case the protocol
(http://www.irct.ir/searchresult.php?id=1658&number=1)
doesn't help much. So the only remaining options are to either make a decision based on what is available in the published trial report or contact the author.
Ahmed
Date: Mon, 1 Jul 2013
20:57:35 -0400
From: [log in to unmask]
Subject: Re: randomisation
To: [log in to unmask]
Hi Mark,
I agree with Valerie, it also appears this study was not blinded. We did a post http://www.ithinkwell.org/why-is-randomization-needed-in-clinical-trials/
and there is an excellent paper on this embedded written by Jeremy Howick and a great 5 minute video supplied by Dr Terry Shanyfelt. In the comments Paul Lucas has made some astute observations and even uses an autologous example, I hope this could be helpful , The guidelines I like best for RCTs and systematic reviews come from a program called CASP they are 10-11 point checklists with full explanations and are a great way to see what you have in a simple and fair way, they are free to download here under checklists http://www.casp-uk.net/
`I always try to give pers a pass but sometimes you just cant! BestAmy
From: Valerie King <[log in to unmask]>
Reply-To: Valerie
King <[log in to unmask]>
Date: Monday, July
1, 2013 8:29 PM
To: <[log in to unmask]>
Subject: Re:
randomisation
In a word,
no.Could someone predict
which group the next subject would be allocated to? Yes.
Could that someone
therefore not enroll said subject, or consent for enrollment differently, or hold off on enrollment, or some other action in order to get the subject the care that the someone really wanted or the subject really wanted? Of course, that someone could do those things. Therefore it isn’t adequate randomisation. Valerie J. King, MD, MPHDirector of Evidence and PolicyCenter for Evidence-based PolicyOregon Health & Science UniversityMailcode
SN-4N3455 SW US Veterans
Hospital Rd.Portland, OR
97239Voice:
503-494-8694Fax:
[log in to unmask] From:
Evidence based health (EBH) [mailto:[log in to unmask]]
On Behalf Of Mark Ayson
Sent: Monday, July 01, 2013 5:05 PM
To: [log in to unmask]
Subject: randomisation Dear
group I was
wondering if you could help me. I am working on a report about autologous blood injections and have an RCT that describes their randomisation technique as follows: "We assigned participants to two groups by randomized sequential allocation. The first patient was randomly assigned to the LC group by tossing a coin, the second patient to the AB group, and the other patients were assigned to the two groups, sequentially."Is this techniques an adequate method of randomisation?I look forward to your comments. Kind regardsMark
Mark
Ayson, MBChB,
DPHResearch Advisor,
Governance, Policy
& Research, ACC
Tel 04-8166386 / Fax 04-8167562 / Ext
46386ACC / Research / Vogel House, 19 Aitken Street PO Box 242 / Wellington 6011 / New Zealand / www.acc.co.nz
ACC cares about the environment – please don’t print this email unless it is really necessary. Thank you.
Disclaimer:
"This message and any attachments may contain confidential and privileged information.
If you believe you have received this email in error, please advise us immediately by return email or telephone and then delete this email together with all attachments.
If you are not the intended recipient, you are not authorised to use or copy this message or any attachments or disclose the contents to any other
person. "
|