Overall and subgroup analysis

1
Overall and subgroup analysis

• If the OVERALL results show highly significant
  evidence of a worthwhile effect of treatment, but
  a few subgroups of the overview unexpectedly
  indicate no benefit (which could well happen by
  chance), then the appropriate question is whether
  there is good evidence that this life-saving
  treatment should be denied to these patients.
• REVERSAL of the usual demand that there should be
  proof of worthwhile benefit.

Courtesy of Dr. K. Wheatley
2
Meta-analysis vs. randomized controlled trials
internal validity vs. generalizibility

• Have complimentary roles
• RCT, large adequately powered
• If our desire is to assess the efficacy of
  treatment (i.e. understand a measure of benefit
  of the treatment under ideal conditions of a
  clinical trial using narrow defined eligibility
  criteria)
• Meta-analysis (of totality of evidence)
• If our goal is to obtain reliable estimate about
  the treatment effectiveness (i.e. understand the
  extent to which a given treatment can produce a
  beneficial effect under variety of circumstances
  and eligibility criteria)

3
Meta-analysis vs. randomized controlled trials
Small CTs
To study mechanisms
To generate hypotheses for more reliable RCTs
Meta-analyses of small RCTs
To obtain reliable overall answers under
specific conditions of a trial
Large RCts
To obtain a typical and unbiased and
generalizible estimate of treatment effect and
to explore interactions among subgroups
Meta-analyses of large RCTs
4
Literature-based vs. individual patient data
meta-analysis?

• IPD MA gold-standard
• LMA may be misleading
• Data extraction, patient exclusion, length of
  follow-up, method analysis may be less accurate
  in LMA

Lancet 1993341418-22 Stat Med 1998142057-2079
5
IV Ethical obligations to account of whats
already known

• To avoid unnecessary trials if reliable knowledge
  already exists
• Conversely, to determine if there is true
  uncertainty about relative values of competing
  treatment alternatives
• A new trial should be conducted if there is a
  substantial uncertainty which of the trial
  treatments would benefit the patient better
• Requirement that equipoise (uncertainty
  principle) is met

6
Ethical obligation of building systematically on
what is already known

• Clinical trials should be preceded by a
  systematic review and should be reported with a
  discussion of assessing the trials results in
  the context what is already known
• Ethical requirement for updating systematic
  reviews
• UK, Denmark, Holland now mandates search or
  conduct of SR before a new clinical trial is done

JAMA 1998280280-282Lancet 20013581648
7
V Knowledge resources
8
Archie Cochrane
It is surely a great criticism of our
profession that we have not organised a critical
summary, by specialty or subspecialty, adapted
periodically, of all relevant randomised
controlled trials.
9
Cochrane Database of Systematic Reviews -

• The Cochrane Collaboration - an international
  network of individuals and institutions committed
  to preparing, maintaining, and promoting the
  accessibility of systematic reviews of the
  effects of health care interventions.

Cochrane Systematic Reviews (2,796) (January
2003) Database of Abstracts of Reviews of
Effectiveness (3,875) Registry of Randomized
Controlled Trials (353,809)
10
How many systematic reviews are needed to cover
whole medicine?

• 10,000 systematic reviews to provide broad
  coverage of most health care topics

Clarke M, personal communication
11
Cochrane Centres
Canadian
Nordic
San Francisco
German
UK
Dutch
French
Italian
Iberoamerican
Chinese
San Antonio
New England
Brazilian
South African
Australasian
12
Cochrane Systematic reviews

• Cochrane reviews have been shown to be
  methodologically superior to non-Cochrane
  systematic reviews

BMJ 2000320537-40, JAMA 1998280278-80
13
The Cancer Library
Cochrane Cancer Network with Update Software Ltd
Courtesy of Dr. Chris Williams
14
Meta-analyses in radiation oncology

• 100 meta-analyses in the Cochrane Database of
  Systematic Reviews
• 22 Cochrane Reviews
• 78 DARE reviews
• MEDLINE (Clinical Queries) search
• 616 systematic reviews

15
Meta-analyses in radiation oncology an example
of reliable review with long-term (20 years)
follow-up

• Favourable and unfavourable effects on long-term
  survival of radiotherapy for early breast cancer
  an overview of the randomised trials

Early Breast Cancer Trialists' Collaborative
Group
Lancet 2000 355 175770 (20 May 2000 )
16
Proportional effects on all-cause mortality in 40
trials of radiotherapy
17
Lancet 2000 355 175770 (20 May 2000 )
18
(No Transcript)
19
Absolute effects of radiotherapy on
cause-specific survival
20
Absolute benefits and hazards
21
Part VIEvidence and decision-making
22
Clinical Decision Making
Patient circumstances
Evidence from research
Preferences, values and rights
Courtesy of Dr. G. Lyman
23
Reporting data on benefits and harms

• If evidence on benefits and harms are not
  reported or is of poor quality, one has to wonder
  how physicians make decisions and recommendations
  for their patients

Eddy D. JAMA 19902641737-39
24
Reporting data on benefits and harms RCTs in
myeloma

• Survival outcomes
• 111/136 (82)
• Survival beyond 5 years
• 15/111 (14)
• Treatment-related mortality
  33/136 (24)
• Non-fatal adverse events
• 91/136 (67)

Annals Oncol 2001121611-1617
25
Reporting harms in RTOG randomized trials
N 44
N 51
N 54
26
HOW TO INTEGRATE BENEFITS AND RISKS OF AVAILABLE
THERAPEUTIC OPTIONS

• Should we always use the option with the best
  benefit/risk ratio?
• Efficacy80 Toxicity10 E/R8
• Efficacy20 Toxicity1 E/R20

27
Decision-making at the bedside

• Minimal conditions for treatment benefit at which
  therapy is worth considering is met when
• Absolute benefitsgtabsolute harms (adjusted for
  the probability of bad event, e.g. relapse)
• Never administer treatment or order diagnostic
  test if treatment harm is greater than its
  efficacy

28
Integrating benefits and harms of radiation
therapy of breast cancer

• Threshold for administering radiation therapy
  (RT)
• probability of breast cancer recurrence (without
  RT)gt
• Deaths due to (RT) ()

Deaths due to breast cancer without RT- deaths
due to breast cancer on RT

4.3
89.6 (actual relapse30.1)
51.4-46.6 (4.8)