Supplementary Table 1 Risk of bias of each study in detail
Random
Sequence
Generation
(selection bias)
Allocation Concealment
(Selection Bias)
Cole et
al.
(2007)17
Computergenerated in
blocks of 6.
Treatment assignments were
concealed from participants and
study staff except for the
pharmacists. Placebo tablets were
identical in appearance to folic
acid tablets.
Jaszewski
et al.
(2008)19
Randomization
was performed
using a
"stratified
randomization
block". No
other
information
was provided.
Selected assignment was made in
advance and recorded in sealed
envelopes, numbered
consecutively.
Blinding of
Participants
and Personnel
(performance
bias)
See above.
Placebo
tablets were
identical in
appearance to
folic acid
tablets.
Blinding of outcome
assessment (detection bias)
Incomplete outcome data
(attrition bias)
Selective reporting
(reporting bias)
Other bias
Treatment study staff
except for the pharmacists.
No information on how this
was done.
Drop outs reported: Phase
1: died (n=34), lost to follow
up (n=9). Phase 2: died
(n=16), lost to follow up
(n=31).
Power
calculations
were
performed and
sufficient
sample
numbers
acquired.
Unspecified.
Unspecified.
No consort flow diagram.
139 eligible subjects
randomized. 94 completed
the 3 year analysis. 43
dropped out - unrelated
deaths (n=28), relocation
(n=15). Very high level of
deaths despite excluding for
co-morbidities. Analysis on
49 (intervention) and 45
(control). In drop-out
subjects, there was no
statistically significant
difference in key
demographics. Not clear
how many were in each
group at baseline as
numbers in chart incorrect.
After randomization,
there was limited
separate reporting on
the outcome of the six
sub-groups. The
reasons for these
omissions are thought
to be that the author
was focusing on the
placebo or folic acid
group (with our
without the two
different doses of
aspirin) as a whole.
Recurrent adenomas
reported as mean
number rather than
relative risk. Data
from paper cannot
therefore be used in
meta-analysis.
Inconsistencies
in
baseline
data provided.
No
power
calculations
were
performed.
Logan et
al.
(2008)18
Computergenerated
randomized list
with a block
size of 8, with
randomization
stratified by
centre.
The researchers and all clinical
staff involved with patient
recruitment were blind to this
treatment allocation schedule. All
trial medication was bottled,
labelled and supplied from the
central trial pharmacy and
dispensed to patients through
regional pharmacies at each
participating centre.
Wu et al.
(2009)20
Random
number
generator
No information provided.
Aspirin and
folic acid
placebo were
identical in
appearance to
active
treatments.
Study
participants,
clinical staff
involved with
patient
surveillance
and
investigators
at the
coordinating
centre were
all blind to
treatment.
Only trial
pharmacists
had
knowledge of
treatment.
No blinding of
participants
because the
control group
received no
placebo
supplement.
No further information
provided
Drop out data provided and
largely balanced across two
groups. Folate only
intervention - 19 drop outs
at stage 1 - died (n=4), lost
to follow up (n=5), poor
health (n=6), patients
refused (n=4). 54 drop outs
at stage 2 - withdrew from
both medication arms
(n=42), withdrew from
aspirin only (n=11),
withdrew from folic acid
only (n=1); Placebo
Intervention - 29 drop outs
at stage 1 - died (n=7), lost
to follow up (n=7), poor
health (n=7), patient refused
(n=8). 46 drop outs at stage
2 - withdrew from both
medication arms (n=25),
withdrew from aspirin only
(n=21).
Apparently free of
selective outcome
reporting - all stated
outcomes reported.
Power
calculations
provided but
required
sample size
not met.
Drop out figures were
provided and reasons for
drop out reported.
Difference in drop out
figures between groups:
n=41 discontinued folic acid
intervention v n=62 in
intervention group. Similar
numbers in final analysis
cohorts (n=237 in folic acid
intervention and n=238 in
cohort). Although number
of subjects excluded from
analysis were similar in
both groups it was not
explained how these
Apparently free of selective
outcome reporting - all
stated outcomes reported.
No inclusion of power calculations.
Bonelli et
al.
(2012)21
No further
information
provided than
randomization
lists stratified
by centre were
provided by
the
pharmaceutical
company.
Each centre received a random list
where the assigned treatment was
indicated by the label numbers of
the bottles containing the pills
with no reference to their
contents. The active compound
and the placebo had an identical
appearance and that the bottles
and pills contained no reference
to their contents.
Double
blinded but no
further
information
Limburg
et al.
(2011)23
"Dynamic
allocation
procedure".
No information provided.
Some
interventions
were
delivered as
tablets and
some as
powder. Also
different daily
dosages used.
related to drop out figures.
Baseline characteristics of
participants included in the
final analysis were similar to
all randomly assigned
participants. In the 194
participants who did not
receive an endoscopy,
characteristics such as age
or severity of adenoma at
baseline, that may have
affected the likelihood of a
follow up, did not differ
significantly at baseline.
States that the trial was
double blinded but no
further information about
blinding of personnel was
provided.
No information provided.
All drop outs reported:
intervention (n=36): refused
follow up colonoscopy
(n=29), died (n=6), cancer
(n=1); control (n=45):
refused follow up
colonoscopy (n=35), died
(n=9), cancer (n=1).
Apparently free of
selective outcome
reporting - all stated
outcomes reported.
Low drop-out rate and drop
out reasons provided: lost to
follow up (n=3), withdrew
consent (n=2), medical
contraindications (n=1),
other (n=2).
Apparently free of
selective outcome
reporting - all stated
outcomes reported.
Power
calculations
provided but
required
sample size
not met. No
collection of
information on
confounding
factors at
baseline.
Power
calculations
provided but
required
sample size
not met.
Principi
et al.
(2013)24
Computer
generated.
Unspecified.
States that
the patients
and
investigators
were blinded
to assignment
without
providing any
further
information.
States that both LI and
intensity staining were
calculated by two
independent observers in a
blinded fashion without
providing any further
information.
Shimizu
et al.
(2008)22
Not described.
Unspecified.
No blinding of
participants
because the
control group
received no
placebo
supplement.
States that endoscopists
were blinded to the study
group but no further
information provided.
Much higher percentage
withdrawal in intervention
than control group. 7 out of
30 patients withdrew from
the intervention group (4
withdrew consent before
T60 colonoscopy and 3 were
lost to T60 colonoscopy). 3
out of 30 patients withdrew
from the control group (3
withdrew consent before
T60 colonoscopy).
8 patients were lost to
follow up in intervention
arm only as they could not
be reached from the day
after the initiation of the
trial. No intention to treat
analysis included. Author
refers to it in the discussion
and states that the results
were insignificant.
Apparently free of
selective outcome
reporting - all stated
outcomes reported.
Power
calculations
provided but
required
sample size
not met.
Apparently free of
selective outcome
reporting - all stated
outcomes reported.
Power
calculations
not provided.