Attitudes and Statistical Methods in Clinical Trials

1
Attitudes and Statistical Methods in Clinical
Trials

• Meiyi Chen
• April 25, 2005
• Department of Statistics
• University of Connecticut

2
Introduction

• Two attitudes in clinical trials
• Explanatory
• Pragmatic
• Statistical methods in clinical trials
• Frequentist approach for explanatory attitude
• Two approaches for pragmatic attitude
• Bayesian approach
• Intention-to-treat approach

3
Explanatory and Pragmatic Attitudes
Armitage(1998) gives the definitions according to
Schwartz and Lellouch (1967), The explanatory
attitude an attempt to understand precisely the
relative effects of alternative treatments when
administered under ideal conditions.It simulates
laboratory conditions, gives biological
information, studies principles of pharmacology
or physiology and assess efficacy. The Pragmatic
attitude a comparison of treatments made under
the less rigorous conditions likely to be found
in practice. It resembles the conditions of
normal clinical practice, focuses on the
question, "What is the better treatment in the
particular clinical circumstances of the patients
in the study?" and assess effectiveness The
following aspects in the design and analysis of a
clinical trial should be considered definition
of treatment regimens, assessment of results,
choice of subjects, method of comparison and
ethics.
4
Definition of Treatment Regimens

• In explanatory approach, the treatment regimens
  are strictly defined
• In pragmatic approach, the treatment regimens are
  loosely defined
• For example, consider the treatment of cancer in
  which radiotherapy alone is to be compared with
  radiotherapy preceded by the administration of a
  drug which has no effect by itself but which may
  sensitize the patient to the effects of radiation
  
• The explanatory approach requires a control group
  to have an initial 30-day period without drug
  treatment
• The pragmatic approach starts radiotherapy
  immediately in the control group

5
Assessment of Results

• The explanatory approach emphasizes on biological
  significance, such as tumour regression or
  biochemical changes.
• The pragmatic approach emphasizes practical
  measures, such as back to work, decrease in pain
  and duration of survival.

6
Choice of Subjects

• In the explanatory approach
• Subjects may be highly selected
• Tend to have "exclusive" eligibility criteria
• Only patients who take medication according to
  the study protocol should be included in the main
  analysis
• In the pragmatic approach
• A more heterogeneous patient population is
  acceptable
• Tend to have inclusive" eligibility criteria
• All randomized patients would include in analyses
  called "intention-to-treat" analyses

7
An Example for Choice of Subjects
Studying a new analgesic in patients with
visceral pain related to cancer1. Patients must
have documented radiological evidence of a tumor
impinging on a hollow viscus corresponding to an
area where they complain of poorly localized,
crampy pain. Patients are ineligible if there is
a lesion of somatic tissues such as ribs or
abdominal wall, or if their pain complaint
includes superficial and steady pain.  2.
Patients are eligible if they have a diagnosis of
metastatic cancer and their primary complaint is
of crampy abdominal pain in the absence of
obvious gastrointestinal disorders unrelated to
cancer.   Question Which one is a pragmatic
approach? Answer 1. Explanatory
2.Pragmatic Source http//symptomresearch.nih.gov
/chapter_1/sec5/cmms5pg2.htm
8
Ethical Problems

• Normally, explanatory approach must be done on
  animals, therapeutic trials on human subjects
  should be limited to pragmatic approach.
• Lellouch and Schwartz (1971) described the two
  ethics of clinical trials between explanatory and
  pragmatic approaches
• Individual ethics for explanatory approach
  interested in the individual patient.
• Collective ethics for pragmatic approach
  emphasize the benefit to society, or to
  population of patients exposed to which treatment
  is chosen.
• The modern clinical trial is a controlled
  experiment with efficient organization, careful
  plan and execution, and capability of
  interpretation by valid statistical analysis,
  such as stopping decision.
• Individual patients are safeguarded by the use of
  informed consent, the provision for withdraw and
  the data and safety monitoring.

9
Method of Comparison

• The classical testing of significance is
  appropriate for explanatory trials, concluding
  that a true difference exists only when the
  observed difference exceeds a threshold level
  which guarantees the required error probability
  ?.
• For pragmatic trials, a decision from some form
  of decision theory, such as Bayesian and
  intention-to treat may be needed to decide which
  treatment is better.

10
Frequentist Vs Bayesian Methods in Clinical
Trials
David J Spiegelhalter etc., An Introduction to
Bayesian Methods in Health Technology Assessment,
BMJ, 319, 508-511 (1999)
11
Why Aren't Bayesian Methods More Widely Used in
Clinical Trials Research?

• Bayesian analysis can be computer intensive and
  effective computational software are needed to
  perform Bayesian analysis.
• BUGS project began in 1989
• Few statisticians are proficient in the routine
  application of Bayesian analysis.

12
Bayesian Methods of Choice of Sample Size and
Allocation in Clinical Trials

• Adcock (1997) gives a review of both Bayesian and
  frequentist methods on determining the overall
  sample size for clinical trials
• Armitage (1989) suggests the use of a dynamic
  allocation method with a fixed number of patients
• Lindley (1997) discusses utility functions which
  balance the cost with the effectiveness of the
  design

13
Bayesian Methods of Choice of Sample Size and
Allocation in Clinical Trials (Continued)

• Tan and smith (1998) consider problems of choice
  of sample size and allocation using utility
  functions when utilities includes
• Benefits and cost (suited to the drug developer)
• Side-effects
• Trade-off between benefit and side-effects

14
Intention-to-treat in Pragmatic Clinical Trials

• The pragmatic attitude is often taken to be
  exemplified par excellence by the
  intention-to-treat approach to the analysis of
  results, whereby comparisons are made between the
  outcomes for the complete groups assigned to
  different treatment regimens, irrespective of the
  extent of departure from the treatment schedules
  laid down in the protocol.
• ---- Peter
  Armitage (1998)

15
Why Should Intention-to-treat Be Used ?

• Reasons including
• The benefit of randomization are maintained
  because compliance or noncompliance occurs after
  randomization.
• Excluding noncompliance can bias the treatment
  evaluation.
• It simulate routine practice.
• Possible counter-arguments
• The extent of non-compliance may be different
  from that in routine practice, so the simulation
  of routine practice by non-compliance is dubious.
• Using an explanatory approach is subject to
  selection bias, it might be better to accept such
  bias.
• An intention-to-treat analysis is likely to
  underestimate true differences, because the
  non-compliers will dilute the effect from the
  compliers.

16
Analyzing Non-compliance in clinical trials

• One very simple model stated in Sommer and Zeger
  (1991) supposed that the success proportions are
  known separately for compliers and non-compliers
  in the drug group.
• Efron and Feldman (1991) provided a regression of
  the outcome variable on the compliance measure.
• Sheiner and Rubin (1995) advocated the
  model-based analyses of efficacy, presenting
  intention-to treat limitation.

17
Analyzing Non-compliance in clinical trials
(Continued)

• Rubin (1998) suggests
• A test statistic is nominated a priori in a
  randomized experiment with non-compliance.
• A generally uncontroversial model is posited for
  non-compliance under the null hypothesis.
• Valid p-values can be obtained using
  randomization-based posterior-predictive
  distributions.

18
Conclusions

• Two attitudes in clinical trial
• Explanatory
• Pragmatic
• Methods to analyze clinical data
• Frequentist
• Bayesian
• Intention to treat

19
Discussion

• Which attitude is better?
• Schwartz and Lellouch tends the explanatory and
  pragmatic attitudes correspond to a clear choice
  between two sorts of trial.
• Armitage (1998) suggested that the two attitudes
  are likely to co-exist in any one trial.

20
Discussion(continued)

• Which method of analysis should be used?
• Considering noncompliance, the intention-to-treat
  is commonly used in analysis of randomized
  controlled trials, but it is inadequate.
• Alternative methods include
• Per protocol analysis compare outcomes only in
  those patients who are compliant.
• As-treated analysis compares observed outcomes
  according to treatment received.
• Classical methods used in randomized controlled
  trial include
• Power calculation at design stage
• Control type I error within sequential monitoring
• Calculate p value and confidence intervals at the
  final analysis
• Meta-analysis for pooling the results of multiple
  study
• Bayesian method
• More flexible and ethical
• Elegant ways to hand multiple sub-studying
• Provide conclusions in a suitable form for making
  decision
• Increasing interest is reflected in literature
  and pharmaceutical companies.

21
References

• Armitage, P. The search for optimality in
  clinical trials, International Statistical
  Review, 53, 15-24 (1985).
• Armitage, P. Attitudes in Clinical Trials,
  Statistics In Medicine, 17, 2675-2683 (1998).
• Schwartz, D. and Lellouch, J. Explanatory and
  pragmatic attitudes in therapeutic trials,
  Journal of Chronic Diseases, 20, 637- 648 (1967).
• Tan, S. B. and Smith, A. F. M. Exploratory
  Thought on Clinical Trials with Utilities,
  Statistics In Medicine, 17, 2771-2791 (1998).
• Sheiner, L. B. etc. Intention-to-treat analysis
  and the goals of clinical trials, Clinical
  Pharmacology and Therapeutics, 57, 6-15 (1995).
• Spiegelhalter D. J. Bayesian Methods in Health
  Technology Assessment a Review, Health
  Technology Assessment, Vol. 4 No. 38 (2000).