CRITICAL APPRAISAL

“Caffeine Therapy for Apnea of Prematurity”

Barbara Schmidt, M.D., Robin S. Roberts, M.Sc., Peter Davis, M.D., Lex W. Doyle, M.D.,
Keith J. Barrington, M.D., Arne Ohlsson, M.D., Alfonso Solimano, M.D., and Win Tin, M.D.

Critical Appraisal Worksheet

1 2 3
Can you find this Is the way this was done Does this problem threaten
information in the paper ? a problem? the validity of the study ?

1. What is the research 1. Is it concerned with 1. Does this problem

question and/or the impact on an threaten the validity
hypothesis? intervention, of the study ?
 To evaluate the casuality, or  No
function of determining the
Methylxanthines magnitude of a
which can reduce the health problem?
frequency of apnea of  Yes
prematurity and the
 As soon as possible
need for mechanical
after random
ventilation during the
assignment,
first seven days of
eligible infants
therapy
received an
intravenous
loading dose of
either 20 mg of
caffeine citrate per
kilogram of body
weight or an
equivalent volume
of normal saline.
This was followed
by a daily
maintenance dose
of 5 mg per
kilogram. If apneas
persisted, the daily
maintenance dose
could be increased
to a maximum of
10 mg of caffeine
citrate per
kilogram. The
maintenance doses
were adjusted
weekly for changes

1
 in body weight and
could be given
orally once an
infant tolerated full
enteral feedings.
The drug was
monitored
according to its
clinical effect only
 The potential for
harm exists
because
methylxanthines
are inhibitors of
adenosine
receptors.8
Adenosine
preserves brain
ATP levels and
protects brain cells
during
experimental
hypoxia and
ischemia in a
variety of animal
models.
2. What is the study type? 2. Is the study type 2. If not, how useful are the
 A randomized appropriate to the results produced by this type
placebo controlled research question? of study?
trial, not blinded.  Yes  Caffeine therapy for apnea
of prematurity reduces the
rate of bronchopulmonary
dysplasia in infants with
very low birth weight.
3. What is the 3. Is the sampling 3. Do these threaten the
reference frame internal validity of the
population? What appropriate for study?
are the sampling the reference  No
frame and sampling population? Is
method? there selection
bias?
 5292 Infants with a  Yes
birth weight of 500–  Yes, the study is
1250 g were not blinded so
candidates for maybe can made
methylxanthine some bias.
therapy during the
first 10 days of life.
977 Excluded. 242
Had dysmorphic

2
 features or congenital
abnormalities likely
to affect life
expectancy or
neurologic
development. 425
Were unlikely to be
available for long-
term follow-up. 277
Previously treated
with a
methylxanthine. 33
For unknown reasons.
4315 Infants eligible.
2309 Did not undergo
randomization. 1628
Because consent not
obtained. 681 Not
approached. Only
2006 underwent
randomization.
 A computer-generated
randomization
scheme was used to
assign the infants to
treatment groups in a
1:1 ratio.
Randomization was
stratified according to
the study center and
balanced in random
blocks of two or four
patients. A designated
pharmacist at each
center received a
binder containing the
prespecified sequence
of treatment-group
assignments from a
statistician at the
coordinating center
who was not
otherwise involved in
the trial. Access to the
binder was restricted
to selected pharmacy
personnel. The
pharmacy study logs
were retrieved after
the completion of

3
 recruitment to ensure
that all randomly
assigned infants were
included in the
analysis.
4. In an experimental 4. Has bias been 4.Does this threaten the
study, how were introduced? internal validity of the
subjects assigned to  Yes study?
group? In a  No
 They stratified the
longitudinal study
how many reached random
final follow up? assignment to
 They used the caffeine or
exclusion criteria : placebo according
977 Excluded. 242 to study center,
Had dysmorphic and They adjusted
features or congenital for center in the
abnormalities likely analysis to
to affect life eliminate the bias
expectancy or caused by the
neurologic variability of
development. 425 practice among
Were unlikely to be the clinical sites.
available for long-
term follow-up. 277
Previously treated
with a
methylxanthine. 33
For unknown reasons.
5. What are the study 5.Is there measurement 5.Is measurement error an
factors and how are error? important source of bias?
they measured?  No error  No
Determinan (illness) measurement in
in this study were this study
infant with apnea of
prematurity

Domain (cause):
Apnea of
prematurity
which was
happened in
infants who born
at less than 34
weeks of
gestation.

Assuming an
incidence of death or
neurodevelopmental

4
 disability of 20
percent, they needed
1000 infants in each
group for the study to
have a statistical
power of 80 percent
to detect a 25 percent
relative reduction in
the risk of the
primary outcome. The
current analyses of
secondary outcomes
were adjusted
according to study
center with the use of
a logistic-regression
model that included
terms for treatment
and center. The
quotient of the
estimated coefficient
and its standard error
was used as a z-test
statistic for the null
hypothesis of no
treatment effect.
Mean weight gain
and head
circumference were
compared between
the two groups at
weekly intervals with
the use of Student’s t-
test. Nonparametric
tests or Fisher’s exact
test were used as
appropriate to analyze
the use of the study
drug and
cointerventions. All P
values are twosided.
6. What are the 6. Are all relevant 6. How important are
outcome factors and outcome assessed? Is omitted outcomes?
how are they there measurement Is measurement error an
measured? error? important source of bias?
 The primary  Yes  There are no omitted
outcome of this outcomes
 No error
study is a composite  There are no error
measurement in
of death, cerebral meausrement
this study
palsy, cognitive

5
delay, deafness, or
blindness at a
corrected age of 18
to 21 months.
Ascertainment of
this outcome is
under way and is
expected to be
completed by
December 2006.
Bronchopulmonary
dysplasia was
defined by the need
for supplemental
oxygen at a
postmenstrual age
of 36 weeks.
Cranial
ultrasonography
was recommended
between the 14th
and the 28th days of
life, and between 34
and 36 weeks of
postmenstrual age if
the infant was still
hospitalized in the
study center at that
time. The following
lesions are
indicative of brain
injury and were
analyzed as a group:
intraparenchymal
echodense lesions,
cystic
periventricular
leukomalacia,
porencephalic cysts,
and
ventriculomegaly
with or without
intraventricular
hemorrhage.
Necrotizing
enterocolitis was
diagnosed during
surgery, at autopsy,
or by a finding of
pneumatosis

6
 intestinalis,
hepatobiliary gas, or
free intraperitoneal
air on radiography.
All stages of
retinopathy were
recorded according
to the international
classification.
Weight and head
circumference were
recorded weekly.

7. What important 7. Are potential 7. Is confounding an

potential confounders important source of bias?
confounders are controlled?  No
considered?  Yes, through
 There is no exclusion criteria
significant
difference in
characteristic
between two
groups, the data
considered as
homogen at the
beginning of the
study
 They stratified the
random assignment
to caffeine or
placebo according
to study center, and
They adjusted for
center in the
analysis to eliminate
the bias caused by
the variability of
practice among the
clinical sites.
8. Are statistical tests 8. Were the tests -
considered? appropriate for
 Yes the data? Are
confidence
 Assuming an intervals given?
incidence of death or Is the power
neurodevelopmental given?
disability of 20
 Yes,
percent, they needed
 Confidence
1000 infants in each
interval 95%.

7
 group for the study to
have a statistical
power of 80 percent
to detect a 25 percent
relative reduction in
the risk of the
primary outcome. The
current analyses of
secondary outcomes
were adjusted
according to study
center with the use of
a logistic-regression
model that included
terms for treatment
and center. The
regression coefficient
associated with
treatment in the fitted
model yielded a point
estimate and 95
percent confidence
interval for the
treatment effect
expressed as an odds
ratio. The quotient of
the estimated
coefficient and its
standard error was
used as a z-test
statistic for the null
hypothesis of no
treatment effect.
Mean weight gain
and head
circumference were
compared between
the two groups at
weekly intervals with
the use of Student’s t-
test. Nonparametric
tests or Fisher’s exact
test were used as
appropriate to analyze
the use of the study
drug and
cointerventions. All P
values are twosided.
9. Are the results 9. Was the sample 9. Is the study useful or
clinically/socially size sufficient to is the result

8
 important? detect a inconclusive?
 Yes clinically/socially  Yes, because
 Caffeine therapy for significant Caffeine therapy
apnea of prematurity result? for apnea of
reduces the rate of  Yes prematurity
bronchopulmonary reduces the rate of
dysplasia in infants bronchopulmonar
with very low birth y dysplasia in
weight. infants with very
low birth weight.
10. What conclusions 10. Do the results 10.Do you accept the results
did the authors apply to the of this study?
reach about the population in  Yes
research question? which you would
Did they generate be interested?
new hypotheses? Do  Yes
you agree with the
conclusions?
 Conclusion : caffeine
therapy for apnea of
prematurity reduces
the incidence of
bronchopulmonary
dysplasia. Except for a
temporary reduction in
weight gain, caffeine
has no apparent
shortterm risks. At the
doses used in this trial,
the incidence of drug-
induced toxicity is low
 No, they didn’t
generate a new
hypotheses
 Yes, I agree with the
conclusion