ASA Connect

 View Only
Expand all | Collapse all

Delayed Treatment

  • 1.  Delayed Treatment

    Posted 05-29-2016 10:46

    Consider a randomized clinical trial comparing surgery vs. medical therapy. Patients assigned to medical therapy start immediately. But, for the surgery patients, there may be a delay in scheduling and performing the surgery. Nevertheless, the primary outcome (e.g., relapse or death) may occur in the intervening period. Does ITT require us to count these events against surgery even though they occurred before it happened? Is there a literature supporting this view?

     

    Thanks,

    Jim R.

     

    - - - - - - - - - - - - - - - - - -

    James Rochon, PhD

    Principal Statistical Scientist

    Rho Federal Systems

    6330 Quadrangle Dr.

    Chapel Hill, NC 27517

    (919) 595-6857

     



  • 2.  RE: Delayed Treatment

    Posted 05-30-2016 02:29

    I am unaware of the ITT regulations, but - from an intent to treat point of view - I would count these events against treatment, since they reflect the trua outcomes in the real world

    ------------------------------
    Yiannis Bassiakos
    Associate Professor
    University Of Athens, Economics Department



  • 3.  RE: Delayed Treatment

    Posted 05-30-2016 02:55
    Yes, events while waiting are choice-related. ICH E9.
    --    Dr. Reinhard Vonthein     Wissenschaftlicher Mitarbeiter   UNIVERSIT??T ZU L??BECK     INSTITUT F??R MEDIZINISCHE BIOMETRIE UND STATISTIK     ZENTRUM F??R KLINISCHE STUDIEN      Ratzeburger Allee 160, Haus 24     23562 L??beck      Tel +49 451 500 2788     Fax +49 451 500 2999     vonthein@imbs.uni-luebeck.de      www.imbs-luebeck.de, zks-luebeck.de 





  • 4.  RE: Delayed Treatment

    Posted 05-30-2016 04:02

    Hi,

    When the delayed surgeries reflect the real world situation then one benefit of medical therapy is that is available immediately. Thus, the events before surgery definitely should be taken into account. 

    If this does not reflect the real world but is special to your trial then one could think of other approaches and excluding the events. But I doubt this is the case as in the trial you generally have all in place to speed up the process. In reality the surgery might be delayed even more. 

    In my eyes this relates to the discussion about estimands: What is the question you want to get answered? In the first place this might be "What is the patient's survival time after a decision between medical therapy or surgery?" and in the second situation the question would be more like "What is the patient's survival time after start of medical therapy or after surgery?"

    ------------------------------
    Martin Kappler
    Statalpha



  • 5.  RE: Delayed Treatment

    Posted 05-30-2016 09:24

    This is an interesting question, James.  If surgery is considered the standard of care (SOC) or the gold standard for treatment then the clock should start once a patient becomes eligible for surgery or the alternative therapy. That is the time of randomization. The fact the therapy may be able to start sooner  or delay the need for surgery could be the most important component of benefit if it is also effective.     

    Susan E. Spruill

    Susan E. Spruill, PStat®

    Statistical Consultant, President

    Applied Statistics and Consulting

    828-467-9184 (phone)

    Professional Statistician accredited by the American Statistical Association

    www.appstatsconsulting.com

     






  • 6.  RE: Delayed Treatment

    Posted 05-30-2016 10:18

    I would say, yes, to include in ITT.  The delay is part of the surgical intervention.  Perhaps a workaround would be to impose a delay after randomization to the medical group.

     

    Steve Crockett






  • 7.  RE: Delayed Treatment

    Posted 05-30-2016 17:52
    What precisely are we comparing? What is the research purpose of the study

    If we our goal is to assess the benefit of receiving surgery vs. medical therapy without surgery, then we would need to randomize prior to any surgery, all events following randomization should count.

    If our purpose is to assess the benefit of 2 treatment regimens following surgery, we randomize following surgery and pre-surgery events count.

    What is the research purpose? What is the role of surgery in the study? Is it a component of what is being studied, or simply a possible concomitant treatment? Can it occur in only one arm, or any arm? Why?

    Where possible, it's usually better to deal with delayed effects by being smart about when one randomized and what inferences one can draw, than by accounting for it in the sample aize.

    Jonathan Siegel
    Associate Director Clinical Statistics




  • 8.  RE: Delayed Treatment

    Posted 05-30-2016 18:01
    I don't know about a formalized guideline, but the ITT principle states that once a patient is randomized he or she is available for summarization regardless of whether the planned treatment is actually received. There are ways to "circumvent" (bad word, but can't think of another) this, such as defining a Safety population which typically only considers Safety events as due to assigned treatment only after the treatment has actually been received and identification of a Modified ITT population (mITT), which is often very similar in concept to the Safety Population, but extends to Efficacy as well. As for surgery, it may be wise to consult with surgeons working in the therapeutic area to find out if from their point of view a surgical procedure begins with pre-surgical testing and other preparations, or whether it is defined as occurring only when the scalpel or other instrument actually penetrates the skin.
     
    Nestor Rohowsky President and Principal Consultant IDCS Inc 5718 Laurel Ave La Grange, IL 60525 phone: (708) 205-4240 fax: (708) 784-1619 e-mail: idcsinc@yahoo.com





  • 9.  RE: Delayed Treatment

    Posted 05-30-2016 18:06
    I think the short answer to the query as stated is yes, events while waiting have to count.

    Imagine a treatment which is a perfect cure when given, but takes 10 years to prepare. If median survival is 1 year while waiting and 2 years on alternative treatment, waiting for the perfect cure is likely not the patient's best treatment choice. What are the actual comparative benefits of the complete alternatives regimens beginning now? ITT is designed to elicit that.

    Jonathan Siegel
    Associate Director, Clinical Statistics

    Sent from my iPhone




  • 10.  RE: Delayed Treatment

    Posted 05-31-2016 12:00
    In a randomized clinical trial comparing surgical to medical management, the intention-to-treat principle would count any outcome experienced in the "surgery" arm against that treatment method, even if the outcome occurred in the interval between randomization and the actual surgical procedure. In effect, one of the risks of surgery is that attendant on waiting for an opening in the surgical schedule, or the possibility that an intervening medical event would require postponing the procedure. As a result, the ITT comparison can be viewed as assessing the difference in two policies of treatment, i.e., a glimpse at how well patients like the ones eligible for the trial would fare if surgical treatment (with all of its attendant delays and risks and chances for changing one's mind prior to going under the knife) were adopted as the general rule, compared to how they would fare if medical treatment (with its attendant risks of adverse experiences, drug interactions, poor compliance) were adopted. The ITT comparison gets at the potential effectiveness of a mode of treatment. It does not directly address the efficacy of the treatment, that is, how well the treatment--when it is actually applied as intended--affects the disease process. Of course, if there are only a few patients in the ITT population who are not also in the "as treated" population, then the ITT analysis addresses both effectiveness and efficacy.


    ********************************************************************************
    This e-mail is intended only for the use of the individual or entity to which
    it is addressed and may contain information that is privileged and confidential.
    If the reader of this e-mail message is not the intended recipient, you are
    hereby notified that any dissemination, distribution or copying of this
    communication is prohibited. If you have received this e-mail in error, please
    notify the sender and destroy all copies of the transmittal.

    Thank you
    University of Chicago Medicine and Biological Sciences
    ********************************************************************************




  • 11.  RE: Delayed Treatment

    Posted 05-31-2016 14:23

    If delaying surgery is bad for the patient, why not concentrate on reducing the delay in the real world?

    ------------------------------
    Emil M Friedman, PhD
    emilfriedman@gmail.com
    http://www.statisticalconsulting.org



  • 12.  RE: Delayed Treatment

    Posted 06-05-2016 11:56

    Randomization partitions patients into stochastically equivalent groups. The primary analysis comparing group outcomes is not confounded (all randomized patients according to randomized group). It is good science because it is based on only one experimental factor. This primary analysis is often referred to as the Intent-to-Treat (ITT) analysis. The totality of the post randomization experience caused any group outcome difference. This is a valid inference of causality because of the simple one factor experimental design.

     

    The above does not mention the defined interventions! Suppose the interventions are simply defined, such as drug A versus drug B, or drug versus surgery, or drug versus placebo. The degree to which the primary analysis result can be claimed as evidence of causal association between the defined interventions is judgement. The judgement is most often (and should be) supported by waterboarding the data. Factors analyzed must include trial conduct factors, especially those associated with the integrity of implementation of the defined interventions.

     

    So, a causal claim for the defined interventions is an inferential leap based on analysis of data that includes confounding. It cannot be otherwise. Ideally, these experimental and analytic principles applicable to randomized clinical trials would be evident and structure reports of results.

    --
    Brent A. Blumenstein, PhD
    Trial Architecture Consulting
    2737 Devonshire Pl NW, Unit 318
    Washington, DC 20008-3462
    www.TriArcConsulting.com
    bab@TriArcConsulting.com
    mobile: +1 (206) 390-0081