Caffeine for Apnea of Prematurity Trial: Benefits May Vary in Subgroups

Peter G. Davis, MD, Barbara Schmidt, MD, Robin S. Roberts, MSc, Lex W. Doyle, MD, Elizabeth Asztalos, MD,
Ross Haslam, MD, Sunil Sinha, PhD, and Win Tin, MD for the Caffeine for Apnea of Prematurity Trial Group*

Objective To determine whether the benefits of caffeine vary in three subgroups of 2006 participants in the
Caffeine for Apnea of Prematurity (CAP) trial.
Study design Post-hoc subgroup analyses were performed on the basis of: (1) indication for commencement of study
drug: treat apnea, prevent apnea, or facilitate extubation; (2) positive pressure ventilation (PPV) at randomization: endo-
tracheal tube (ETT), noninvasive ventilation, or none; and (3) timing of commencement of study drug: early or late (#3
versus >3 days). Outcomes assessed were those showing treatment effects in the original analyses. We investigated
the consistency of caffeine effects by using regression models that incorporated treatment/subgroup factor interactions.
Results There was little evidence of a differential treatment effect of caffeine in subgroups defined by the clinical
indication for starting study drug. The size and direction of the caffeine effect on death or disability differed depend-
ing on PPV at randomization (P = .03). Odds ratios (95% CI) were: no support, 1.32 (0.81-2.14); noninvasive support,
0.73 (0.52-1.03); and ETT, 0.73 (0.57-0.94). Adjustment for baseline factors strengthened this effect (P = .02). Start-
ing caffeine early resulted in larger reductions in days of respiratory support. Postmenstrual age at time of discon-
tinuing PPV was shorter with earlier treatment (P = .01). Mean differences (95% CI) were: early, 1.35 weeks
(0.90-1.81); and late 0.55 weeks (–0.11-0.99). Adjustment for baseline factors weakened this effect (P = .03).
Conclusions There is evidence of variable beneficial effects of caffeine. Infants receiving respiratory support
appeared to derive more neurodevelopmental benefits from caffeine than infants not receiving support. Earlier ini-
tiation of caffeine may be associated with a greater reduction in time on ventilation. (J Pediatr 2010;156:382-7).

T
he Caffeine for Apnea of Prematurity (CAP) trial compared short- and long-term outcomes of preterm infants treated
with caffeine with those of infants receiving a placebo. Infants were eligible for inclusion when their clinicians considered
them to be candidates for methylxanthine therapy during the first 10 days of life.1,2
Earlier trials and systematic reviews examined the safety and efficacy of methylxanthines in relation to the specific clinical
indication for commencement of treatment. Cochrane reviews evaluated methylxanthines for treatment of apnea (5 trials,
192 infants),3 prophylaxis for infants at risk of apnea (2 trials, 104 infants),4 and pre-extubation treatment (6 trials, 197 in-
fants).5 The small numbers of subjects included in the 3 pooled analyses limited the statistical power to evaluate the safety
and efficacy of methylxanthines.
The mechanisms of the short- and long-term beneficial effects of caffeine are not clearly understood. In human infants, meth-
ylxanthines reduce apneic events, increase minute volumes, improve respiratory system compliance,6 reduce diaphragmatic
fatigue,7 and act as a diuretic.8 However, methylxanthines reduce survival of weanling mice exposed to a period of anoxia.9
Therefore, the benefits and risks of caffeine may vary according to the respiratory status and postnatal age of the infant.
The broad inclusion criteria of the CAP trial meant that infants were randomized and commenced study drug at varying levels
of respiratory support. Moreover, some infants commenced the study drug immediately after birth, and others received their
first dose of study drug after they reached their second week of life.
We conducted these post hoc analyses of the CAP trial to determine the short- and long-term effects of caffeine according to
the clinical subgroups of: (1) indication for commencing study medication; (2) level of respiratory support at randomization;
and (3) age of starting treatment.

Methods From the Department of Obstetrics and Gynaecology,

University of Melbourne, Melbourne, Australia (P.D.,
L.D.); Departments of Clinical Epidemiology and
Biostatistics, McMaster University, Hamilton, Ontario,
The trial design and the short- and long-term outcomes of the CAP trial have Canada (B.S., R.R.); Department of Pediatrics, University
been reported previously.1,2 The research ethics boards of all participating cen- of Toronto, Toronto, Ontario, Canada (E.A.); Women’s
and Children’s Hospital, Adelaide, Australia (R.H.); and
ters approved the trial. Written parental consent was obtained prior to study en- Department of Pediatrics, James Cook University
Hospital, Middlesbrough, United Kingdom (S.S., W.T.)
*Additional members of the Caffeine for Apnea of Pre-
maturity Trial Group are available at www.jpeds.com
CAP Caffeine for Apnea of Prematurity (Appendix).
ETT Endotracheal tube Funded by the Canadian Institute of Health Research and
PDA Patent ductus arteriosus the National Health and Medical Research Council of
Australia. The authors declare no conflicts of interest.
PMA Postmenstrual age
PPV Positive pressure ventilation 0022-3476/$ - see front matter. Copyright ! 2010 Mosby Inc.
All rights reserved. 10.1016/j.jpeds.2009.09.069

382
 Vol. 156, No. 3 ! March 2010

try. In summary, infants with a birth weight between 500 and
1250 g were randomized to receive either caffeine citrate (n =
1006) or normal saline placebo (n = 1000). A computer-gen-
erated randomization scheme with variable block sizes was
used, and a designated pharmacist at each center prepared
vials of study solution. A loading dose of 20 mg of caffeine
citrate per kilogram of body weight was followed by a daily
maintenance dose of 5 mg/kg. Caregivers and those assessing
study outcomes were masked to the treatment group. The
dose of study drug could be increased to 10 mg/kg/day
when symptoms of apnea persisted or withheld or reduced
when there were clinical signs of toxicity. Caffeine levels
were not measured. Open label use of caffeine or other meth-
ylxanthines was strongly discouraged and occurred in 9.5%
of infants overall. The study drug was commenced at a me-
dian age of 3 days and discontinued before a median of 35
weeks postmenstrual age (PMA).
Infants receiving caffeine had reduced rates of broncho-
pulmonary dysplasia and spent approximately 1 fewer week
with an endotracheal tube (ETT), any positive pressure ven-
tilation (PPV), and supplemental oxygen.1 The primary out-
come of the trial was death before a corrected age of 18 to 21
months or survival with $1 of these conditions: cerebral
palsy, cognitive delay, hearing loss requiring amplification,
or bilateral blindness. Caffeine reduced the rate of the pri-
mary outcome and 2 of its components, cerebral palsy and
cognitive delay.2
Subgroup analyses
We documented why each infant was considered to be a can-
didate for methylxanthines as 1 of the following mutually ex-
clusive reasons: to prevent apnea, to treat apnea, or to
facilitate removal of an ETT. Analyses of a subset of outcomes
were conducted for each ‘‘indication’’ subgroup. In addition,
we performed subgroup analyses on the basis of the level of
respiratory support at randomization (no support, non-inva-
sive respiratory support, or ventilation via an ETT) and the
time of commencing study drug (< median age versus $
median age).
We assessed these outcomes: Death or major disability; ce-
rebral palsy; cognitive delay; broncopulmonary dysplasia,
any retinopathy of prematurity (ROP); severe ROP; drug
therapy for closure of patent ductus arteriosus (PDA); PDA
ligation; use of postnatal corticosteroids; PMA at last use of
oxygen treatment, last use of ETT, and last use of PPV. All
these outcomes showed an overall treatment effect in the
original analyses of the 2 randomly allocated CAP trial
groups. Definitions of outcomes are provided in full in earlier
publications.1,2 Bronchopulmonary dysplasia was defined as
the need for supplemental oxygen at a PMA of 36 weeks. Ce-
rebral palsy was diagnosed when the child had a non-progres-
sive motor impairment characterized by abnormal muscle
tone and decreased range or control of movements. Cogni-
tive delay was defined as a Mental Development Index score
<85 on the Bayley II Scales of Infant Development.10
For each dichotomous outcome, logistic regression models
were fitted separately to each subgroup. This yielded a point
estimate and a 95% CI for the subgroup treatment effect ex-
pressed as an odds ratio. The subgroup treatment effects for
continuous outcomes such as PMA of last respiratory sup-
port were expressed as mean differences between the treat-
ment groups with 95% CIs. For dichotomous outcomes,
we determined the statistical significance of the observed het-
erogeneity of treatment effect with logistic regression models
incorporating treatment/subgroup interactions. Multiple lin-
ear regression models were used for continuous outcomes.
Each analysis was then adjusted for prognostically important
variables—gestational age, sex, level of maternal education,
antenatal steroid treatment, and multiple births. A P value
for the test of heterogeneity (treatment x subgroup interac-
tion) <.05 was considered to represent statistically important
evidence of a subgroup effect.
Results
Adequate data for an analysis of the primary composite out-
come were available for 1869 (93.2%) of the infants who
were enrolled in the study. The characteristics of these 1869
children were similar in the 2 groups at birth (Table I; avail-
able at www.jpeds.com). There were no important differences
in these characteristics between the subgroups (Table II).
The number of infants in each subgroup with an adverse
outcome (n), the total number in each group (N), the odds
ratios, and their 95% CIs are shown in Figure 1, A,
Figure 2, A, and Figure 3, A (available at www.jpeds.
com). Continuous outcomes expressed as means (SD)
and the mean differences between caffeine and placebo

Table II. Baseline characteristics by subgroup

Respiratory support
Reason for caffeine at randomization Time of Randomization
Apnea Apnea PPV, PPV Early Late
Baseline characteristics treatment prophylaxis Extubation No PPV no ETT with ETT £3 days >3 days
Mother’s education college/university 54.2% 53.3% 47.5% 48.8% 53.2% 51.6% 48.9% 53.5%
Antenatal steroids 89.8% 86.3% 87.3% 89.4% 90.4% 86.3% 89.7% 86.9%
Gestational age (weeks), mean (SD) 27.8 27.3 26.8 28.8 27.8 26.7 27.4 27.3
(1.76) (1.84) (1.69) (1.54) (1.55) (1.69) (1.67) (1.91)
Singleton birth 69.9% 67.4% 76.3% 68.5% 68.7% 74.3% 71.9% 71.6%
Female 52.7% 45.2% 45.4% 54.6% 49.9% 45.5% 48.4% 48.3%

383
THE JOURNAL OF PEDIATRICS ! www.jpeds.com Vol. 156, No. 3

A Caffeine Placebo OR (fixed) OR (fixed)

n/N n/N 95% CI 95% CI
Death or major disability
Apnea treatment 141/400 153/367 0.76 [0.57, 1.02]
Apnea prophylaxis 94/219 88/204 0.99 [0.67, 1.46]
Pre-extubation 141/316 189/360 0.73 [0.54, 0.99]
0.80 [0.66, 0.96]
Test for heterogeneity: P = 0.44

Cognitive delay
Apnea treatment 110/374 117/341 0.80 [0.58, 1.09]
Apnea prophylaxis 78/207 66/189 1.13 [0.75, 1.70]
Pre-extubation 105/285 145/327 0.73 [0.53, 1.01]
0.84 [0.69, 1.02]
Test for heterogeneity: P = 0.25

Cerebral palsy
Apnea treatment 11/388 18/361 0.56 [0.26, 1.19]
Apnea prophylaxis 10/215 9/200 1.04 [0.41, 2.60]
Pre-extubation 19/305 39/339 0.51 [0.29, 0.91]
0.60 [0.40, 0.90]
Test for heterogeneity: P = 0.43

BPD
Apnea treatment 107/413 141/392 0.62 [0.46, 0.84]
Apnea prophylaxis 84/226 94/211 0.74 [0.50, 1.08]
Pre-extubation 158/322 212/350 0.63 [0.46, 0.85]
0.65 [0.54, 0.78]
Test for heterogeneity: P = 0.76

PDA ligation
Apnea treatment 16/427 40/400 0.35 [0.19, 0.64]
Apnea prophylaxis 10/233 23/220 0.38 [0.18, 0.83]
Pre-extubation 19/339 66/378 0.28 [0.16, 0.48]
0.32 [0.23, 0.46]
Test for heterogeneity: P = 0.77

0.1 0.2 0.5 1 2 5 10

Favors caffeine Favors placebo

B Caffeine Placebo MD (fixed) MD (fixed)

N Mean (SD) N Mean (SD) 95% CI 95% CI

PMA at last oxygen treatment

Apnea treatment 410 33.1(4.9) 387 34.0(4.3) -0.84 [-1.48, -0.19]
Apnea prophylaxis 223 34.4(5.8) 208 35.0(4.4) -0.61 [-1.66, 0.37]
Pre-extubation 319 35.3(4.5) 347 36.8(5.4) -1.53 [-2.29, -0.77]
-1.02 [-1.45, -0.58]
Test for heterogeneity: P = 0.24
PMA at last ETT
Apnea treatment 410 29.8(3.1) 387 30.4(3.1) -0.61 [-1.04, -0.19]
Apnea prophylaxis 224 30.0(3.3) 208 30.6(3.4) -0.64 [-1.28, -0.09]
Pre-extubation 319 30.3(3.3) 349 31.2(3.6) -0.88 [-1.41, -0.34]
-0.69 [-0.99, -0.40]
Test for heterogeneity: P = 0.72

PMA at last positive pressure ventilation

Apnea treatment 410 31.3(3.0) 387 32.2(3.1) -0.86 [-1.29, -0.43]
Apnea prophylaxis 224 31.5(3.2) 208 32.5(3.4) -1.02 [-1.63, -0.40]
Pre-extubation 319 31.9(3.5) 348 33.0(3.6) -1.03 [-1.57, -0.49]
Test for heterogeneity: P = 0.86 -0.98 [-1.28, -0.69]

-4 -2 0 2 4
Favors caffeine Favors placebo

Figure 1. A, Neonatal morbidities and long-term impairments by reason for treatment. B, PMA at last use of respiratory supports
by reason for treatment. Each black diamond represents the overall treatment effect for that outcome.

groups and their 95% CIs are shown in Figure 1, B, expressed as forest plots for each subgroup. The overall
Figure 2, B, and Figure 3, B. For dichotomous and contin- treatment effect of caffeine on each outcome is expressed
uous outcomes, the point estimates and their 95% CIs are as a diamond at the bottom of each graph.
384 Davis et al
March 2010 ORIGINAL ARTICLES

Caffeine Placebo OR (fixed) OR (fixed)

A n/N n/N 95% CI 95% CI
Death or major disability
No PPV 61/167 41/135 1.32 [0.81, 2.14]
Non-invasive vent 100/285 118/278 0.73 [0.52, 1.03]
Endotracheal tube 216/485 271/518 0.73 [0.57, 0.94]
0.80 [0.66, 0.96]
Test for heterogeneity: P = 0.03, adjusted 0.02

Cognitive delay
No PPV 55/162 33/129 1.49 [0.90, 2.49]
Non-invasive vent 77/263 91/260 0.77 [0.53, 1.11]
Endotracheal tube 161/442 204/468 0.74 [0.57, 0.97]
0.83 [0.69, 1.02]
Test for heterogeneity: P = 0.02, adjusted 0.01

Cerebral palsy
No PPV 3/168 4/138 0.61 [0.13, 2.77]
Non-invasive vent 8/273 13/274 0.61 [0.25, 1.49]
Endotracheal tube 29/468 49/488 0.59 [0.37, 0.95]
0.60 [0.40, 0.89]
Test for heterogeneity: P = 0.97

BPD
No PPV 41/182 32/151 1.08 [0.64, 1.82]
Non-invasive vent 73/290 107/293 0.58 [0.41, 0.83]
Endotracheal tube 236/491 308/509 0.60 [0.47, 0.78]
0.65 [0.54, 0.78]
Test for heterogeneity: P = 0.09

PDA ligation
No PPV 2/184 6/154 0.27 [0.05, 1.36]
Non-invasive vent 8/303 23/302 0.33 [0.14, 0.75]
Endotracheal tube 36/514 101/542 0.33 [0.22, 0.49]
0.33 [0.23, 0.46]
Test for heterogeneity: P = 0.94

0.1 0.2 0.5 1 2 5 10

Favors caffeine Favors placebo

B N
Caffeine
Mean (SD) N
Placebo
Mean (SD)
MD (fixed)
95% CI
MD (fixed)
95% CI

PMA at last oxygen treatment

No PPV 184 32.5(5.4) 154 32.59(3.82) -0.06 [-1.07, 0.96]
Non-invasive vent 303 33.4(5.6) 302 34.0(4.1) -0.65 [-1.43, 0.13]
Endotracheal tube 518 35.0(5.1) 541 36.3(5.5) -1.32 [-1.96, -0.67]
-0.85 [-1.29, -0.40]
Test for heterogeneity: P = 0.11
PMA at last ETT
No PPV 184 29.6(2.3) 154 30.1(2.6) -0.53 [-1.04, -0.01]
Non-invasive vent 303 29.5(3.0) 302 29.8(2.8) -0.38 [-0.83, 0.08]
Endotracheal tube 519 30.7(4.0) 543 31.4(3.9) -0.78 [-1.26, -0.31]
-0.55 [-0.83, -0.27]
Test for heterogeneity: P = 0.47
PMA at last positive pressure ventilation
No PPV 184 30.5(2.5) 154 31.4(3.0) -0.83 [-1.43, -0.24]
Non-invasive vent 303 31.3(3.6) 302 32.0(3.0) -0.68 [-1.21, -0.16]
Endotracheal tube 519 32.1(3.9) 542 33.1(3.9) -0.93 [-1.40, -0.46]
-0.88 [-1.18, -0.57]
Test for heterogeneity: P = 0.80

-4 -2 0 2 4
Favors caffeine Favors placebo

Figure 2. A, Neonatal morbidities and long-term impairments by level of respiratory support at randomization B, PMA at last use
of respiratory supports by level of respiratory support at randomization. Each black diamond represents the overall treatment
effect for that outcome.

Caffeine for Apnea of Prematurity Trial: Benefits May Vary in Subgroups 385
THE JOURNAL OF PEDIATRICS ! www.jpeds.com
Outcomes by reason for treatment with study drug
There was little evidence of heterogeneity of treatment effect
for any outcome assessed on either the unadjusted (Figure 1)
or adjusted analyses (data not shown).
Outcomes by level of respiratory support
There was evidence of heterogeneity of caffeine effect for the
outcomes death or major disability (P = .03) and cognitive
delay (P = .02; Figure 2, A). These effects remained statisti-
cally significant after adjustment for prognostically impor-
tant baseline variables. Infants receiving respiratory support
at the time of randomization appeared to derive greater
long-term neurological benefit from caffeine than infants
not receiving any respiratory support. There was
little evidence of heterogeneity for any other outcome
(Figure 2).
Outcomes by age at randomization
The median age at randomization was 3 days; the early
group comprised infants randomized at <3 days, and the
late group comprised infants randomized at $3 days. Un-
adjusted analysis showed evidence of significant heteroge-
neity for the outcomes of broncopulmonary dysplasia (P
= .02), PMA at last oxygen treatment (P = .03; Figure 3,
B), PMA at last endotracheal intubation (P = .02), and
PMA at last PPV (P = .01; Figure 3, B). After adjustment,
only PMA at last endotracheal intubation (P = .04) and
PMA at last PPV (P = .03) remained statistically significant.
Infants whose treatment commenced before 3 days of age
appeared to derive greater short-term respiratory benefit
from caffeine than infants commencing treatment at $3
days.
There was little evidence of any subgroup effects in the ad-
justed analyses for any of the following outcomes: any ROP;
severe ROP; drug therapy for closure of PDA, and use of
postnatal corticosteroids (data not shown).
Discussion
The CAP study showed that caffeine significantly improved
survival without neurodevelopmental disability at a corrected
age of 18 to 21 months.2 Caffeine also reduced the incidence
of bronchopulmonary dysplasia and shortened the duration
of supplemental oxygen therapy and assisted ventilation.1
The eligibility criteria were broad and pragmatic. Infants
with birth weights from 500 to 1250 g could be randomized
in the first 10 days of life when their caregivers considered
them to be candidates for methylxanthine therapy. Earlier
trials and meta-analyses ‘‘split’’ the question of efficacy of re-
spiratory stimulants according to the clinical indication for
commencing therapy. The numbers of patients enrolled
were small, and important clinical outcomes were not mea-
sured.
Mindful of the earlier trials, we recorded the clinical
indication for considering methylxanthine therapy. This
provided an opportunity to explore the potential for differ-
386
Vol. 156, No. 3
ential effectiveness of methyxanthines in these subgroups.
However, caution should be exercised in the interpretation
of these subanalyses. We did not stratify for clinical indica-
tion, time of study drug commencement, or level of respira-
tory support, and we did not undertake sample size
calculations for these subgroups.
Although effect sizes and directions of effect of caffeine
varied in the subgroups, we found no evidence of harm
reaching statistical significance. However, the power to detect
differences in the size of treatment effect in subgroups is rel-
atively low. Similarly, although not directly addressed here,
the power to make statements about the presence or absence
of a treatment effect within any specific subgroup is much
less than for the trial overall. Rather than focus on the point
estimates of treatment effects for each subgroup, we stress the
importance of looking for consistency of treatment effects
across the subgroups.
The stated clinical indication for the use of methylxan-
thines appeared to have little effect on the benefits of caf-
feine. However, the effectiveness of caffeine varied
significantly depending on the level of respiratory support
at time of study commencement. In relation to the 2 out-
comes, death or major disability and cognitive delay, it ap-
pears that infants not requiring any PPV derived less
benefit than infant receiving support via continuous posi-
tive airway pressure or an ETT. This is consistent with
our post hoc analysis of possible mechanisms of action of
the effect of caffeine on the rate of survival free of disabil-
ity.2 Earlier discontinuation of any positive airway pressure
was the most powerful of 6 mechanisms explored and ex-
plained 49% of the beneficial long-term drug effect. There
were some interesting trends suggesting that early com-
mencement of caffeine was associated with improved
short-term pulmonary outcomes. However, only PMA at
last endotracheal intubation and the last use of any positive
airway pressure remained statistically significant after ad-
justment for prognostically important variables. Caution
should be applied in interpreting this subanalysis because
infants who started the study drug early were likely to
have been seen by their clinicians as more ready to wean
from respiratory support than infants who were enrolled
later.
The CAP study provides a unique resource because of
its methodological strengths and large number of subjects
enrolled. Lagakos cautioned against over-interpretation of
the results of subgroup analyses, but suggested that when
well planned and executed, such analyses could make
a valuable contribution to the literature.11 One pitfall to
be avoided is that of multiple testing. We conducted 36
comparisons and would therefore expect that 2 nominally
significant interaction terms would occur by chance
alone.12
The message of the CAP trial for clinicians remains clear.
Caffeine improves neurodevelopmental outcomes to 2 years
of age. Follow-up of infants at 5 years is ongoing and may
further refine estimates of effect of caffeine on neurodeve-
lopmental outcome. Infants who start receiving caffeine
Davis et al
March 2010
while they are receiving respiratory support appear to expe-
rience the greatest improvements in neurodevelopmental
outcome. n
Submitted for publication Feb 17, 2009; last revision received Aug 20, 2009;
accepted Sep 25, 2009.
Reprint requests: Dr Peter G. Davis, The Royal Women’s Hospital, Locked Bag
300, Cnr Grattan St & Flemington Rd, Parkville, Victoria, 3052 Australia. E-mail:
pgd@unimelb.edu.au.
References
1. Schmidt B, Roberts RS, Davis P, Doyle LW, Barrington KJ, Ohlsson A, et al.
Caffeine therapy for apnea of prematurity. N Engl J Med 2006;354:2112-21.
2. Schmidt B, Roberts RS, Davis P, Doyle LW, Barrington KJ, Ohlsson A,
et al. Long-term effects of caffeine therapy for apnea of prematurity. N
Engl J Med 2007;357:1893-902.
3. Henderson-Smart DJ, Steer P. Methylxanthine treatment for apnea in
preterm infants. Cochrane Database Syst Rev 2001;CD000140.
4. Henderson-Smart DJ, Steer PA. Prophylactic methylxanthine for pre-
venting of apnea in preterm infants. Cochrane Database Syst Rev 2000;
CD000432.
Caffeine for Apnea of Prematurity Trial: Benefits May Vary in Subgroups
ORIGINAL ARTICLES
5. Henderson-Smart DJ, Davis PG. Prophylactic methylxanthines for
extubation in preterm infants. Cochrane Database Syst Rev 2003;
CD000139.
6. Davi MJ, Sankaran K, Simons KJ, Simons FE, Seshia MM, Rigatto H.
Physiologic changes induced by theophylline in the treatment of apnea
in preterm infants. J Pediatr 1978;92:91-5.
7. Heyman E, Ohlsson A, Heyman Z, Fong K. The effect of aminoph-
ylline on the excursions of the diaphragm in preterm neonates. A
randomized double-blind controlled study. Acta Paediatr Scand
1991;80:308-15.
8. Rieg T, Steigele H, Schnermann J, Richter K, Osswald H, Vallon V.
Requirement of intact adenosine A1 receptors for the diuretic and natri-
uretic action of the methylxanthines theophylline and caffeine. J Phar-
macol Exp Ther 2005;313:403-9.
9. Thurston JH, Hauhard RE, Dirgo JA. Aminophylline increases cerebral
metabolic rate and decreases anoxic survival in young mice. Science
1978;201:649-51.
10. Bayley N. Bayley Scales of Infant Development:Manual. San Antonio:
The Psychological Corporation, Harcourt Brace & Company; 1993.
11. Lagakos SW. The challenge of subgroup analyses—reporting without
distorting. N Engl J Med 2006;354:1667-9.
12. Wang R, Lagakos SW, Ware JH, Hunter DJ, Drazen JM. Statistics in
medicine—reporting of subgroup analyses in clinical trials. N Engl J
Med 2007;357:2189-94.
387
THE JOURNAL OF PEDIATRICS ! www.jpeds.com
Appendix
In addition to the authors, these investigators, research
nurses, and hospitals participated in the Caffeine for Apnea
of Prematurity Trial (study sites are listed according to the
number of infants they enrolled): McMaster University Med-
ical Center, Hamilton, Ontario, Canada—B. Schmidt, J.
D’Ilario, J. Cairnie, J. Dix, B. Adams; Royal Women’s Hospi-
tal, Melbourne, Australia—B. Faber, K. Callanan, N. Davis, J.
Duff, G. Ford; Sunnybrook Health Sciences Center, Toronto,
Canada—L. Golec, M. Lacy, D. Hohn; Women’s and Chil-
dren’s Hospital, Adelaide, Australia—C. Barnett, L. Good-
child, R. Lontis; Mercy Hospital for Women, Melbourne,
Australia—S. Fraser, J. Keng, K., Saunders, G. Opie, E. Kelly;
Centre Hospitalier Universitaire de Quebec, Quebec City,
Quebec Canada—A. Bairam, S. Ferland, L. Laperriere,
S. Bélanger, P. St. Amand; Ottawa Hospital, Ottawa, Ontario,
Canada—M. Blayney, D. Davis, J. Frank, B. Lemyre; British
Columbia Children’s Hospital, Vancouver, British Columbia,
Canada—A. Solimano, A. Singh, M. Chalmers, K. Ramsay,
A. Synnes, M. Whitfield, M. Rogers, J. Tomlinson; Academic
Medical Center, Amsterdam, The Netherlands—M. Offringa,
D. Nuytemans, E. Vermeulen, J. Kok, A. van Wassenaer; Meir
General Hospital, Kfar-Saba, Israel—S. Arnon, A. Chalaf, R.
Regev, I. Netter; Mount Sinai Hospital, Toronto, Ontario,
Canada—A. Ohlsson, K. Nesbitt, K. O’Brien, A.M. Hamilton;
Royal University Hospital, Saskatoon, Saskatchewan,
Canada—K. Sankaran, S. Morgan, P. Proctor; The Brooklyn
Hospital Center, Brooklyn, New York—M. LaCorte, P. LeB-
lanc , A. Braithwaite; Soroka University, Beer Sheva, Is-
rael—A. Golan, T. Barabi, E. Goldstein; The Canberra
Hospital, Canberra, Australia—G. Reynolds, B. Dromgool,
S. Meskell; Foothills Hospital, Calgary, Alberta, Canada—D.
McMillan,D. Schaab, L. Spellen, R. Sauve, H. Christianson,
D. Anseeuw-Deeks; St. Boniface, Winnipeg, Manitoba,
Canada—R. Alvaro, A. Chiu, C. Porter, G. Turner, D.
387.e1
Vol. 156, No. 3
Moddemann, N. Granke, K. Penner; University Hospital
Maastricht, Maastricht, The Netherlands—T. Mulder, A.
Ghys, M. van der Hoeven; Kingston General Hospital, King-
ston, Ontario, Canada—M. Clarke, J. Parfitt, H. MacLean;
Windsor Regional Hospital, Windsor, Ontario, Canada—C.
Nwaesei, L. Kuhn, H. Ryan, C. Saunders; Ludwig Maximilian
University, Munich, Germany—A. Schulze, P. Pudenz, M.
Muller; Astrid Lindgren Children’s Hospital, Stockholm,
Sweden—H. Lagercrantz, M. Bhiladvala, L. Legnevall, E. Her-
lenius; Victoria General Hospital, Victoria, British Columbia,
Canada—D. Matthew, W. Amos, S. Tulsiani, C. Tan-Dy, M.
Turner; Kaplan Medical Center, Rehovot, Israel—E. Shinwell,
R. Levine, A. Juster-Reicher; Royal Victoria Hospital, Mon-
treal, Quebec, Canada—K. Barrington, T. Kokkotis, M.
Khairy, P. Grier, J. Vachon; James Cook University Hospital,
Middlesbrough, United Kingdom—W. Tin, S. Fritz; Univer-
sity of Sherbrooke, Sherbrooke, Quebec, Canada—H. Walti,
D. Royer; Royal Maternity Hospital Belfast, Northern Ireland,
United Kingdom—H. Halliday, D. Millar, A. Berry, C. Mayes,
C. Cummings; Basel Children’s Hospital, Basel, Switzer-
land—H. Fahnenstich, K. Philipp, B. Tillmann, P. Weber;
Moncton Hospital, Moncton, New Brunswick, Canada—R.
Canning; Royal Victoria Infirmary, Newcastle upon Tyne,
United Kingdom—U. Wariyar, W. Tin, S. Fritz, N. Embleton;
University Hospital Zurich, Zurich, Switzerland—H-U.
Bucher, J-C. Fauchere; Northern Neonatal Initiatives, Mid-
dlesbrough, United Kingdom—W. Tin, S. Fritz; University
Hospitals of Geneva, Geneva, Switzerland—R. Pfister, V. Lau-
noy, P. Huppi; University of Tuebingen, Tuebingen, Ger-
many—C. Poets, P. Urschitz-Duprat. Steering Committee:
K. Barrington, P. Davis, L.W. Doyle, A. Ohlsson, A. Solimano,
W. Tin. External Safety Monitoring Committee: M. Gent
(Chair), W. Fraser, E. Hey, M. Perlman, K. Thorpe. Consul-
tant Pharmacist: S. Gray. Coordinating and Methods Center
in Hamilton, Ontario, Canada: R.S. Roberts, C. Chambers,
L. Costantini, E. McGean, L. Scapinello.
Davis et al
March 2010 ORIGINAL ARTICLES

Caffeine Placebo OR (fixed) OR (fixed)

A n/N n/N 95% CI 95% CI
Death or major disability
Early 152/375 206/408 0.67 [0.50, 0.89]
Late 225/562 225/524 0.89 [0.70, 1.13]
0.79 [0.66, 0.95]
Test for heterogeneity: P = 0.33

Cognitive delay
Early 123/348 151/372 0.80 [0.59, 1.08]
Late 170/519 178/486 0.84 [0.65, 1.09]
0.82 [0.68, 1.00]
Test for heterogeneity: P = 0.80

Cerebral palsy
Early 15/367 34/395 0.45 [0.24, 0.85]
Late 25/542 32/506 0.72 [0.42, 1.23]
0.58 [0.39, 0.88]
Test for heterogeneity: P = 0.27

BPD
Early 111/396 190/426 0.48 [0.36, 0.65]
Late 239/567 257/528 0.77 [0.61, 0.98]
0.64 [0.53, 0.77]
Test for heterogeneity: P = 0.02, adjusted 0.06

PDA ligation
Early 11/413 54/444 0.20 [0.10, 0.38]
Late 35/588 76/555 0.40 [0.26, 0.61]
0.32 [0.22, 0.45]
Test for heterogeneity: P = 0.08

0.1 0.2 0.5 1 2 5 10

Favors caffeine Favors placebo

Caffeine Placebo MD (fixed) WMD (fixed)

B N Mean (SD) N Mean (SD) 95% CI 95% CI

PMA at last oxygen treatment

Early 413 33.5(5.2) 442 35.1(4.6) -1.58 [-2.24, -0.92]
Late 592 34.4(5.5) 556 35.0(5.4) -0.57 [-1.21, 0.07]
-1.06 [-1.51, -0.60]
Test for heterogeneity: P = 0.03, adjusted 0.10

PMA at last ETT

Early 413 29.5(2.9) 444 30.6(3.5) -1.08 [-1.51, -0.65]
Late 593 30.5(3.8) 556 30.9(3.5) -0.37 [-0.79, 0.05]
-0.72 [-1.02, -0.42]
Test for heterogeneity: P = 0.02, adjusted 0.04
PMA at last positive pressure ventilation
Early 413 31.2(3.0) 443 32.6(3.7) -1.35 [-1.81, -0.90]
Late 593 31.8(4.0) 556 32.4(3.5) -0.55 [-0.99, -0.11]
Test for heterogeneity: P = 0.01, adjusted 0.03 -0.99 [-1.30, -0.67]

-4 -2 0 2 4
Favors caffeine Favors placebo

Figure 3. A, Neonatal morbidities and long-term impairments by age at commencement of study drug. B, PMA at last use of
respiratory supports by age at commencement of study drug. Each black diamond represents the overall treatment effect for that
outcome.

Caffeine for Apnea of Prematurity Trial: Benefits May Vary in Subgroups 387.e2
THE JOURNAL OF PEDIATRICS ! www.jpeds.com Vol. 156, No. 3

Table I. Baseline characteristics of the infants and their

mothers
Caffeine Placebo
(n = 937) (n = 932) P value
Baseline characteristics
Maternal college/university education 454 (52%) 438 (50%) .51
Antenatal steroids 829 (89%) 817 (88%) .62
Gestational age (weeks) 27.4 (1.8) 27.3 (1.8) .45
Singleton 672 (72%) 668 (72%) 1.00
Female 468 (50%) 435 (47%) .17

These data are for the 1869 children with adequate information for the ascertainment of the
composite primary outcome at a corrected age of 18 to 21 months.

387.e3 Davis et al