It is difficult to imagine how such a precedure should be done. legal aspects for trial performance like the need of dressing ethical committees and achieving patients informed consenter identical as is the data quality and thus documentation effort needed. However, I agree that high quality registries are of utmost importance and value for clarifying the role of novel drugs in unselected patients in a real world scenario.FDA and EMA have recently endorsed the value of such registries

No, not always. Every CT has unique SOP. There is very important role of such  registries.

I am having difficulty understanding the term "registry" in the context of a RCT.  Are you referring to a registry like the DES registry and taking pts from that group?  If so, does that not destroy the concept of randomization, since the pts in a registry would have to share the same criteria for inclusion in the registry.

As a member of the pharm industry, I conducted many RCTs and some were not randomized but utilized "matched pairs" assigned to a treatment group or control group.  In conditions like osteoporosis or arthritis, randomization could lead to a preponderance of severely afflicted pts in one group and mildly affected people in another.  When a clinical response ivolves MD global evals and pt evals (subjective criteria), something other than a strict randomization may be a better design. Yes, it does destroy the "statistical" eval of the data and preclude typcial T-test or ANOVA analyses, but you know what they say about statistics, there are liiars and damn liars.  Sometimes, non parametic analyses are preferable and permit a better distribution of pts according to severity characteristics than typical randomization.

I know that there will be statisticians and purists who will disagree with my comments, but experimental design is an art form and needs to be flexible enough to be adapted to each clinical trial.  Certainly, biological diversity in the pt population is a mahor concern to all researchers, but there must be some flexibility and creativity.  Pt and life cannot always be reduced to a statistical formula or algorithm ... people are not numbers, they are a multi-factorial group of heterogeneous entities which share some common feathres, but exhibit great diversity within the group.  To think it can all be ironed out with an algorithm or T test is naive.  Just my opinion.

Thank you Richard for your response. Can you please provide some references concerning the endorsement by FDA and EMA of quality registries?

Thank you Krishnat for your response.

Thank you David for your response.

The idea of Registry-based RCTs is to compare treatment A (new drug) with treatment B (reference drug) within the population of a clinical registry. Patients in arm B with a maximum of caractéristiques common to the arm A will be selected for a comparative analysis. The prospective recruitment of patients, and the logistics burden (cost) associated with classical RCTs is not necessary. The number of subjects enrolled, and the length of follow-up are optimal with this methodology. It will not replace the classical RCTs that is the gold standard and the only legal procedure to commercialize a new drug. But it could be a compementary control in real-life practice conditions of efficacy and long-terme results of drug, surgical technics or a combination of both.

I do believe  that  experimental design in each RCT  should have flexibility and creativity  to fit into each trial.  I also agree that prospective registry based RCT would be conducive to  cost effective and efficient clinical trials in many areas.

Thank you Pius for your response. 

…The randomized registry trial represents a disruptive technology, a technology that transforms existing standards, procedures, and cost structures. Will it be given serious consideration as a way to resolve the recognized limitations of current clinical-trial design? Theodore Roosevelt once said, “Do what you can, with what you have, where you are.” Today we can no longer afford to undertake randomized effectiveness trials that cost tens or hundreds of millions of dollars. But today we also have registries and other powerful digital platforms…

Michael S. Lauer, M.D., and Ralph B. D’Agostino, Sr., Ph.D. N Engl j Med 369;17 october 24, 2013.

I am very much in favor of comparing drug A with drug B and not doing just placebo comparisons.  I am pleased Dr. Karam agrees and I also am pleased that Dr. Maliakal shares my philosophy on flexibility and creativity in experinebtal design..

Wearable device technology opens up the potential to match large numbers of patients based on data obtained from remote monitoring of various disease states. Patients may need to be enrolled into a registry to enable analysis of de-identified data for trial reasons. We have been doing propensity score matching with multiple parameters obtained from implanted devices, and have been able to replicate RCT results using these methods, but with much greater granularity using populations over 100,000 and 7 years follow-up. There is the potential to use this type of an approach to carry out smaller RCT to confirm benefit, and then augment in some case the post market follow-up of patients to better understand the effects of interventions using larger populations, longer follow-up but with lower cost.

Thank you Arjun for your opinion. I totally share this idea provided that registry data quality are controlled. It can also be used to validate new techniques in case of surgical treatment.

Finally, it could be also used to study the efficacy of new formulations of a same molecule. Our registry (www.eltr.org) has performed recently this kind of study to compare long-term outcomes with prolonged-release tacrolimus versus tacrolimus BD in liver transplantation.

Improved survival in liver transplant recipients receiving prolonged-release tacrolimus in the European Liver Transplant Registry. Adam R, Karam V, Delvart V, et al. Am J Transplant. 2015 May;15(5):1267-82. 

It sounds like you are referring to something along the lines of a cohort multiple RCT (http://www.bmj.com/content/340/bmj.c1066).  You set up your cohort (in this case a registry) and then select a random group of eligible patients for a new intervention.  The remainder of eligible patients not selected remain on standard treatment and form the control group.  This design has the advantage that you only consent the randomised subset for the intervention - the control cohort have already consented to data collection within the registry.  Long term follow-up is much easier than in a standard comparative trial design, and multiple trials can run within the same registry at the same time.

The only caveat is that the data collected in the registry must include all of the important safety and efficacy outcomes that you want to measure, as collection of additional data outside of the register is difficult. This is where the concept of a Core Outcome Set for use in trials/registries can be extremely useful.

Thank you Simon for your comment.

To better explain the principle, I will use the example of our registry. The ELTR collects data on each liver transplantation (LT) performed in Europe since the first one in 1968. These data are prospectively and regularly updated and represent the cohort of LT in Europe, with a greater than 97% completeness. Suppose we wish to know if an anti-rejection treatment A (or A surgical technique) is still effective longtime after its release. We need just isolate all patients using the treatment A (or present surgical technique) and create a comparison group using the reference treatment B (reference surgical technique B) identical on the main variables by using the propensity score matching method. PSM puts us in the situation of randomization but afterwards. so we can compare A and B treatment with optimal statistical power (highest number of subjects included), with a longer length of patients follow-up, and in the situation of real-life practice (not only selected patients).