05 -10- 2020
Intention to
Treat
Analysis &
Drop-outs
Group b4
01 02
Introduction Analysis Methods
Case Scenario - Intention to treat
- Per protocol
Drop-outs Definition & - As treated
Impact - Comparison
03 04
How to handle the missing Conclusions
- data ?
Worst-case - Case Answer
scenario - Take Home
- Hot deck
Messages
imputation - References
- Sensitivity
analysis
01
Introducti
on
Case Scenario
Drop-outs Definition & Impact
Case Scenario
Dr Wilson returned from his lunch break
with renewed vigour. He had just read the
conclusion of a trial on the treatment of ear
infections. 500 children were given a new
antibiotic, zapitillin, and compared with 500
children who were given his usual choice,
amoxicillin. In the zapitillin arm, 240 out of
300 children who completed the study
improved (80%). In the amoxicillin arm,
300 out of 400 children who completed the
study improved (75%). He sent a copy of
the paper to his colleague, proposing that
zapitillin be the first-choice antibiotic in the
hospital formulary.
What is Randomized
Controlled Trial:
Simple Definition:
A study in which people are allocated at random
to receive one of several interventions
Simple but powerful research
tool
Dropouts:
A term of art for a subject in a clinical trial who for any reason fails to continue in
the trial until the last visit or observation, as required of him or her by the
study protocol
Examples of
Dropouts: Voluntary Death
withdrawal from
the trial
2 4
1 3
Loss to follow-up Non-compliance
individuals who
cannot be contacted
or who have moved
out of the study area
Analytical Methods
Called “On-Treatment Analysis” or “The Subjects analyzed according to
Explanatory Approach” treatment taken or not
Per-protocol Intention To Treat As Treated
Called “The Pragmatic Approach”,
“Once Randomized, Always
Analyzed”
Per-protocol Analysis
Data are collected Advantages
from only those who: Isolate &a identify the
- Received treatment
- complied with the trial biological effect of a
protocol treatment.
Drop-outs are
excluded!
If these subjects are not Disadvantages
accounted for in the analysis of Attrition or Exclusion
the results, the results and
conclusions can be misleading bias
and important effects of the
intervention (e.g. intolerable
side-effects) can be lost.
Does this mean explanatory analyses shouldn’t
Intention-to-treat Analysis
Considered
as The
Analysis of
Choice
ALL OR Why to Include A Non-
Advantages NOTHING! compliant Patient in The
Analysis?
• Tends to produce more • All study participants are • After randomization.
conservative results compared included whether they • May bias the treatment
to the Per-Protocol analysis. completed the study or not. evaluation.
• Success rates based on ITT tend • It keeps them in the original • In clinical practice, some
to mirror the results seen in real- groups for the purpose of patients are not
life practice. statistical analysis. fully compliant.
• Compliant subjects & better
outcomes
As treated protocol
Experimenta
l
1 2
Non compliance
from treatment in
• Subjects analyzed the experimental
according to group is
treatment taken or
Control
considered as a
not member of the
control group and
visa versa.
Missing Data:
O1
What are the missing data ?
Problems of missing data ?
How to analyze missing data?
3
Analyze for
patterns of
2
Analyze missing
mechanism data
1
of missing
How much
data
data is
missing ?
2. Analyze mechanism of missing
Data
Missing
Completely at 02 Missing Not at
Random (MNAR)
Random (MCAR)
01 Missing at 03
Random (MAR)
Analyze For Patterns Of Missing Data
Patterns
01 02
Monoto Arbitrar
ne y
How to deal with missing data ?
worst-case scenario
analysis Last observation
1 Carried forward LOCF
• Subjects who drop out are The last recorded results of
recorded as having the individuals who drop out are
worst outcome possible. carried forward to the end of
• The most cautious and the trial and incorporated into
pessimistic approach. 2 4 the final analysis of the results.
Hot-Deck Sensitivity
Imputation analysis
• Values from similar • Assumptions are made
subjects with complete 3 when the missing values are
records (the donors) put in.
can be used to fill in • Sensitivity analyses can also
missing values in the be carried out to include
drop-outs. different scenarios of
assumptions, such as the
worst-case and best-case
scenarios.
Conclusion
Each
Strategy Unequal
Can Perform drop-outs
Better and No Perfect on both
Worse Way To Reasons arms?
Compensat Drop-
e outs?
Degree of The chance
Adherence of there
Should be being drop-
Reported outs should
be
minimized
POSSIBLE
RESULTS CAUSES
Case Answer:
Dr Wilson’s colleague read the paper too, but came to a different
conclusion. He emailed Dr Wilson, ‘I disagree with the conclusions
the researchers have drawn. Five hundred patients were enrolled
into each arm of the study. I worked out that the results did not take
into account the drop-outs in both arms. Intention-to-treat analysis
shows that amoxicillin gave better results, with 300 out of 500
children improving (60%). In contrast, only 240 children improved
with zapitillin (48%). I won’t be prescribing zapitillin unless the child
has an allergy to amoxicillin, but thanks for drawing my attention to
the paper.’
Take Home Message
• ITT analyses try to preserve the randomized groups and address
pragmatic hypotheses about the clinical utility of treatment.
• Explanatory analyses address interesting hypotheses about the
biological effect of treatment, but are more prone to bias.
• ITT analyses should be the primary analysis for most clinical trials.
• Explanatory analyses should carefully consider the effect of
compliance.
• Clinical trial reports should document compliance and withdrawal
in detail.
References
● The CONSORT statement: revised
recommendations for improving the quality of
reports of parallel-group randomized
trials. The Lancet, 357(9263), 1191–1194. doi:
10.1016/s0140-6736(00)04337-3 By Moher,
D., Schulz, K. F., & Altman, D. G. (2001).
● 6 Different Ways to Compensate for Missing
Values In a Dataset (Data Imputation with
examples). Retrieved from
https://towardsdatascience.com/6-different-
ways-to-compensate-for-missing-values-data-
imputation-with-examples-6022d9ca0779.
● The Doctor's Guide to Critical Appraisal, Third
and Fourth Edition
THANKS 01 TAMIM SUBHI ALREFAAI 171110149
ALHASAN ALI ALAKHFASH
02 171110056
STUDENT MAHMOD TALAL ALNAJJAR
NAMES:
03 171110107
Any questions?
04 AHMED KHALIL AMASSI 171110118
Group B4
05 MUHAMMED HUSSAIN ALKAHLOT
171110106
