VIEWPOINTS

Challenges in the Design of Antibiotic Equivalency Studies:

The Multicenter Equivalency Study of Oral Amoxicillin
versus Injectable Penicillin in Children Aged 359 Months
with Severe Pneumonia
Patricia L. Hibberd,1 Archana Patel,2 and the Amoxicillin Penicillin Pneumonia International Study (APPIS) Groupa
1
Clinical Research Institute, TuftsNew England Medical Center, Boston, Massachusetts; and 2Clinical Epidemiology Unit, Indira Gandhi Medical College, Nagpur, India

The World Health Organization (WHO) recommends that children with severe pneumonia (characterized by cough or difficult
breathing, as well as lower chest wall indrawing) be hospitalized and treated with parenteral penicillin. Oral amoxicillin, if
equally effective for treating severe pneumonia, would address challenges associated with providing parenteral therapy,
including risk of transmission of bloodborne pathogens from contaminated needles, exposure to nosocomial pathogens during
hospitalization, inadequate access to health care facilities, and cost. The recently completed multicenter international trial
of oral amoxicillin versus parenteral penicillin for treatment of severe pneumonia demonstrated the equivalency of these
agents in children with severe pneumonia. This article focuses on the challenges of designing an equivalence study and the
threats to the validity of the trial results, particularly the implications of the bias toward finding equivalence when subjects
are unlikely to respond to either study therapy. These considerations have implications for use of the Amoxicillin Penicillin
Pneumonia International Study (APPIS) results in clinical practice and for potential modification of WHO treatment
guidelines.

In developing countries, acute respiratory Streptococcus pneumoniae and Haemoph- HIV and hepatitis B and C viruses through
infection (i.e., pneumonia, severe pneu- ilus influenzae are still the predominant the use of contaminated needles [710].
monia, and very severe pneumonia) re- causes of severe pneumonia in this age In a recently published trial from Pakistan,
mains a major cause of morbidity and group, the WHOs standard case manage- in which children with pneumonia or se-
mortality in children !5 years of age [1, ment guidelines for children with severe vere pneumonia were randomized to re-
2]. The World Health Organizations pneumonia is directed at treatment of ceive oral amoxicillin or trimethoprim-
(WHOs) standard case management these pathogens [4]. Children receive ben- sulfamethoxazole [11], amoxicillin was
guidelines define severe pneumonia as zyl-penicillin (50,000 units/kg im or iv effective in treating 82% of those with se-
cough or difficult breathing, as well as q6h) for at least 3 days as an inpatient [5], vere pneumonia, which raises the possi-
lower chest wall indrawing [3]. Because and, while improving, they complete a 5- bility that oral amoxicillin may be effective
day course with oral amoxicillin (15 mg/ initial treatment for severe pneumonia.
Received 1 October 2003; accepted 15 April 2004; kg t.i.d.). The prospective, multicenter, randomized,
electronically published 3 August 2004.
a
The challenges of providing the initial controlled clinical trialthe oral amoxi-
Study group members are listed at the end of the text.
Reprints or correspondence: Dr. Patricia L Hibberd, Division 3 days of parenteral therapy to children cillin versus parenteral penicillin for treat-
of Clinical Research Resources, Institute for Clinical Research with severe pneumonia are well recog- ment of severe pneumonia international
and Health Policy Studies, TuftsNew England Medical
Center, 35 Kneeland St., Box 063, Boston, MA 02111
nized [6]. Parenteral therapy is costly to study (APPIS)was designed to evaluate
(phibberd@tufts-nemc.org). administer, requires access to facility- whether these 2 treatments were equiva-
Clinical Infectious Diseases 2004; 39:52631 based health care for hospitalization, and lent. The primary hypothesis was that
 2004 by the Infectious Diseases Society of America. All
rights reserved.
is associated with risks of exposure to nos- treatment failures would be equivalent for
1058-4838/2004/3904-0015$15.00 ocomial pathogens and transmission of children who received oral amoxicillin or

526 CID 2004:39 (15 August) Hibberd et al.

injectable penicillin for 48 h, and the sec- lecting an equivalence margin that is too otics would be inappropriate for children
ondary hypothesis was that the 2 treat- small (a conservative strategy that would who had already taken antibiotics for 148
ments would remain clinically equivalent waste resources) and discussion of the h and continued to have signs of severe
through 5 and 14 days of follow-up. principle that the equivalence margin pneumonia. As a compromise, we elected
should be smaller than a difference that to enroll children who had taken anti-
CHALLENGES IN THE DESIGN would be acceptable for determining su- biotics for !48 h and to evaluate the po-
OF THE TRIAL periority of one treatment over another. tential impact of antibiotics received be-
By means of a consensus process, a dif- fore study enrollment on study outcomes
The challenges and complexities in de-
ference in failure rates of 5% between and conclusions.
signing international studies in children
treatment groups within 48 h after ini- With regard to the second eligibility
are well recognized. To address the antic-
tiating therapy was chosen as the equiv- criterion, specific risks of or lack of re-
ipated challenges, the studys steering
alence margin for the trial. We could not sponse to empirical treatment with
committee, comprised of members from
use a superiority design to assess equiv- amoxicillin or penicillin were recognized
9 international sites, the WHO, and the
alence, because failure to demonstrate su- for 4 groups of childrenthose with
Center for International Health at Boston
periority does not establish equivalence bacterial pneumonia due to pathogens
University (Boston), considered the fol-
and because our study question was not resistant to penicillin or amoxicillin
lowing study design issues during a 7-day
whether penicillin or amoxicillin was (community acquired or nosocomially
protocol development workshop.
superior in the treatment of severe acquired), those with Pneumocystis jiro-
pneumonia. veci pneumonia [17, 18] as a result of
Appropriate Study Design: Equivalence
HIV infection, those with viral pneu-
or Superiority
Threats to the Validity of the Trial monia (primarily due to respiratory syn-
Members of the steering committee
Results and Approach to These Threats cytial virus [RSV]), and those with
unanimously agreed that results from a
Impact of patient selection. Standard- hyperreactive airway disease. Because
randomized clinical trial would be
ized eligibility criteria were established to children who had been hospitalized
needed to support any recommendation
ensure that children enrolled in the study within 2 weeks before screening were at
to modify WHO standard case manage-
met WHO criteria for severe pneumonia. risk of having nosocomial pneumonia
ment guidelines for treatment of children
We considered 2 such criteria that were caused by pathogens not likely to be
with severe pneumonia. Central to this
discussion was whether the study should not part of the WHO definition because treatable by either regimen, they were ex-
be designed to assess the equivalence of of the equivalence design: prior history cluded from the study. We recognized
both or the superiority of 1 of the study of antibiotic use and unlikeliness to re- that it was possible for children to have
drugs. We selected an equivalence design spond to amoxicillin or penicillin in community-acquired S taphylococcus au-
because of its utility in settings in which children. reus pneumonia and even methicillin-re-
the standard therapy (in this case, in- With regard to the first eligibility cri- sistant S. aureus pneumonia. However, at
jectable penicillin) has been shown to be terion, in developing countries, children enrollment, it was not possible to obtain
beneficial but the new treatment (in this with pneumonia frequently receive out- an etiologic diagnosis or to treat accord-
case, oral amoxicillin) is easier to use, has patient antimicrobial therapy before ing to the potential antimicrobial resis-
fewer side effects, or is less costly. How- seeking care at a health care facility [15]. tance pattern of respiratory pathogens,
ever, true equivalence can never be es- Exclusion of children who had taken any because of the lack of rapid diagnostic
tablishedit is necessary to select, a antibiotic before enrollment would limit tests for most pulmonary pathogens and
priori, how large a difference would still the generalizability of the study results, the difficulties of obtaining lung aspirates
be considered equivalent (i.e., the equiv- whereas inclusion of children who had from children [19, 20].
alence margin) [12, 13]. This margin was taken antibiotics would tend to bias the Although nasopharyngeal isolates do
determined by members of the steering results toward equivalence, because the not necessarily predict the etiologic agent
committee [14] on the basis of an antic- study antibiotics might have less influ- of severe pneumonia, we elected to assess
ipated treatment failure rate of 11% ence on study outcome. WHO standard the potential effect of antimicrobial re-
among children treated with benzyl pen- case management guidelines recommend sistance of nasopharyngeal isolates of S.
icillin [6]. Deliberations included com- reassessment at 48 h and advancement to pneumoniae and H. influenzae on study
paring the risks of selecting an equiva- second-line antibiotics if there is evidence outcome [2123]. To minimize the num-
lence margin that is too large (to avoid of treatment failure [16]. The steering ber of children with P. jiroveci pneumo-
a meaningless result) with those of se- committee decided that first-line antibi- nia, we excluded children who had HIV

Design of Pediatric Antibiotic Trials CID 2004:39 (15 August) 527

infection with a clinical category of B or Table 1. Criteria for study eligibility.
C [24]. In the 2 study sites with a high
Inclusion criteria
prevalence of HIV infection (Durban,
Age of 359 months
South Africa, and Ndola, Zambia), HIV
Cough or difficult breathing, as well as lower chest wall indrawing
infection was assessed using a combina- Exclusion criteria
tion of HIV antibody testing (for children Nonsevere pneumonia
15 months of age) and PCR for detec- Very severe pneumonia (danger signs: inability to drink, abnormal sleepiness, central
tion of viral antigen (for children !15 cyanosis, and convulsions)
months of age). Finally, because real-time Hospitalization during previous 2 weeks
RSV testing was not practical at most History of bronchial asthma or 2 prior episodes of wheezing
sites, we evaluated the impact of RSV on Severe malnutrition (either weight for age z score less than or equal to 3 SD or pres-
ence of kwashiorkor)
the study results by bulk testing for the
Measles during previous month
presence of RSV antigen in nasal wash- Known or clinically recognizable chronic conditions (anomalous congenital cardiac or re-
ings collected and frozen during the base- spiratory findings, chronic lung disease, bronchopulmonary dysplasia, neurological im-
line assessment. pairment affecting respiratory function, renal diseases, and malignant or hematologi-
cal diseases)
The steering committee anticipated
Diseases affecting lower chest wall indrawing (rickets, severe pallor, and severe
that hyperreactive airway disease could dehydration)
be confused with severe pneumonia. To Low oxygen saturation (!75% in room air at high-altitude sites [Bogota, Columbia, and
minimize enrollment of children with Mexico City, Mexico] and !80% in room air at other sites)
bronchospasm and without severe pneu- Prior anaphylactic reaction to penicillin or amoxicillin
Antibiotic therapy for 48 h before admission to the hospital
monia, we excluded children with a his-
Inability to tolerate oral medications (3 episodes of vomiting per hour)
tory of bronchial asthma or at least 3
Living outside of the hospitals catchment area
prior episodes of wheezing. Children
Category B or C HIV infection
without this history and with a poten-
tially reversible episode of bronchospasm
were challenged with up to 3 doses of
agnosed comorbid condition, receipt of current WHO-recommended duration of
inhaled salbutamol, and they were con-
another antibiotic, and death. If consent antibiotic treatment) and, for determin-
sidered to be eligible only if other criteria was withdrawn or if the child withdrew ing whether relapse had occurred, treat-
persisted after bronchodilator therapy. from the study against medical advice, ment failure 14 days after enrollment in
Our final eligibility criteria (table 1) re- outcome was also considered to be treat- the study (i.e., 9 days after completion of
flected a compromise between unre- ment failure because it was not known the course of antibiotics).
stricted enrollment and the inclusion of and could not be assumed to be favor- Use of a composite end point is chal-
children who would be less likely to re- able. The conceptual framework was that lenging in any clinical trial [25, 26], but
spond to either treatment, because this the appearance of danger signs, low it is particularly challenging in equiva-
latter group would compromise the oxygen saturation, discontinuation of lence studies. Composite end points re-
equivalence design. study drug by the treating physician, oc- flect real clinical situations and are ap-
Impact of using a composite study currence of new comorbid conditions or propriate when there is no obvious
outcome. Treatment failure 48 h af- complications, and death represented choice of primary outcome [27]. In the
ter initiation of therapya composite failure to respond to antimicrobial ther- APPIS, the goal was to evaluate whether
outcome reflecting clinical deteriora- apy or progressive or persistent disease oral amoxicillin and injectable penicillin
tionwas the primary end point of the due to the presence or development of were equivalent in treating severe pneu-
study. This outcome was defined as oc- an empyema or lung abscess. The timing monia and preventing a range of severe
currence of any of the following signs and of the primary outcome was based on the outcomes in addition to death. The risks
symptoms: danger signs (inability to current WHO recommendation that the of using a composite outcome include
drink, abnormal sleepiness, central cya- initial response to treatment be assessed bias toward equivalence, which could
nosis, or convulsions), low oxygen sat- after 48 h of antimicrobial therapy [3, 6]; be characterized by considering a higher
uration (!80% in room air at sea level this time frame is frequently used to as- number of deaths and lower number of
or !75% in Bogota, Colombia, and Mex- sess initial response in patients with com- severe outcomes in one treatment arm to
ico City, Mexico), persistence of lower munity-acquired pneumonia. Secondary be equivalent to a lower number of
chest indrawing, life-threatening or se- outcomes included treatment failure on deaths but a higher number of severe
rious adverse drug reaction, newly di- the fifth day of antibiotic therapy (the outcomes in the other treatment arm.

528 CID 2004:39 (15 August) Hibberd et al.

Unfortunately, it is often impractical to duct both an intent-to-treat analysis (that effect, and to assess whether baseline
power a study to detect equivalence for included all patients) and a per-protocol characteristics and results were consistent
all components of a composite end point. analysis (that excluded those who were across study sites. A priori baseline co-
For this reason and to prevent misleading lost to follow-up or were !3 months of variates to be included in the model as
conclusions, it is particularly important age) to evaluate the impact on losses to fixed effects were sex, age of !12 months,
that results of all components of the pri- follow-up on study conclusions. To en- breast-feeding at onset of present illness,
mary end point are included in the pub- able us to conduct an intent-to-treat immunization status (current or not),
lished version of the article. The overall analysis, we assigned an outcome of treat- use of antibiotics before admission to
strategy for the APPIS was to obtain and ment failure to children who withdrew the hospital, presence of malnutrition
report unbiased assessments of each from the trial or who left against medical (weight for age z score, less than 2),
component of the composite end point. advice, recognizing that overall treatment fever (temperature, 138C), tachypnea,
Impact of lack of blinding on assess- failure in the trial would represent a and oxygen saturation.
ment of outcome. Although blinding worst-case scenario.
to treatment assignment is an ideal way CHALLENGES AND
of minimizing bias in assessment of out- Monitoring the Safety of an Equivalence SOLUTIONS TO PROBLEMS
comes, administration of placebo injec- Trial DURING STUDY CONDUCT
tions to children !5 years of age was The APPIS was monitored by an inde- Mortality. The DSMB conducted its
considered to be unethical. Similar con- pendent data safety monitoring board first interim analysis after 8 deaths had
clusions have been made by investigators (DSMB) that was charged with evaluat- occurred at 2 study sites in which there
in other randomized clinical trials based ing subject safety throughout the trial. was a high prevalence of HIV infection
in the United States and elsewhere [28 Because the trial outcome was treatment (Zambia and Durban). Seven of the 8 chil-
32]. To reduce the risk of biased assess- failure, the DSMB focused on whether dren who died were !12 months of age
ment of subjective outcomes, we defined there was interim evidence that treatment and were likely to have had P. jiroveci
each outcome to minimize subjectivity failure was occurring more frequently in pneumonia. Even though the results did
and conducted intensive staff training in 1 treatment arm, rather than on whether not approach statistical significance for a
use of study outcomes both on-site and the treatments were equivalent. Because difference between treatment groups, the
via video recordings. To assess adherence sample size calculations for equivalence DSMB recommended that the exclusion
to study and outcome definitions, inde- studies are different from sample size cal- criteria for the study be revised to exclude
pendent monitors (D.M.T., S.Q., and culation for detecting differences be- children !1 year of age who were likely to
O.F.) audited study procedures during tween treatment groups, we calculated be infected with HIV. Children !1 year of
site visits. In addition, we expected that sample size requirements from both per- age with hepatosplenomegaly, oral thrush,
the most subjective outcomeswitching spectives, selecting the larger sample size or known family member(s) with HIV in-
to another antibioticwould occur more for the study. This approach enabled the fection were excluded. Only 1 additional
frequently among children receiving oral overall result to address equivalence and death occurred at the Zambia or Durban
therapy, leading to a bias away from to provide the interim analyses of study sites between May 2000 and the comple-
equivalence. However, we recognize that safety with adequate power to detect sta- tion of the trial.
our inability to blind treatment assign- tistically important differences between DSMB assessment of study power dur-
ment was a limitation of the study. treatment groups. The final sample size ing the study. The target sample size of
Impact of losses to follow-up or ab- for the study was 1722 children (861 per 1722 subjects was based on anticipated
sence of outcome information. Partic- group). We planned to evaluate whether treatment failure rate during or after 48 h
ipants who withdrew from the trial or the 2 treatments were equivalent using of study treatment of 10% in both groups,
left against medical advice tend to dilute the two 1-sided tests procedure and to as described above. After 1034 children
any difference between the groups, which calculate the risk differences and 95% CIs (60%) were enrolled in the study, the pro-
biases the results toward equivalence. To of the primary and secondary outcomes portion of treatment failures was 18.6%
address the effect that absence of infor- [33]. If the 95% CI limits are within the in the amoxicillin group and 19.9% in the
mation has on outcomes for children range of 5% to 5%, the treatments are penicillin group. The DSMB raised the
who were lost to follow-up, we set the considered to be equivalent. We also concern that the equivalence margin for
goal that no more than 1% of the study planned to evaluate predictors of treat- the study would likely be 15% at study
population could have a missing primary ment failure at 48 h using a mixed-effects completion, unless the sample size was in-
end point at 48 h because of losses to model (SAS Institute), with study site as creased, but did not recommend changing
follow-up. We planned, a priori, to con- a random effect and treatment as a fixed the sample size. Staff at the data co-

Design of Pediatric Antibiotic Trials CID 2004:39 (15 August) 529

ordinating center recalculated sample size Yobo, and Kojo Yeboah-Antwi (Ministry oping countries: recent relevant research and
current initiatives. Int J Tuberc Lung Dis
requirements without changing any as- of Health; Kintampo, Ghana); Mumtaz 2000; 4:80726.
sumptions except the proportion of treat- Hassan (Childrens Hospital; Islamabad, 7. Simonsen L, Kane A, Lloyd J, Zaffran M, Kane
ment failures, which was now assumed to Pakistan); Prakash Jeena and Hoosan M. M. Unsafe injections in the developing world
and transmission of bloodborne pathogens: a
be 19.25% in both groups. To retain the Coovadia (University of Natal; Durban,
review. Bull World Health Organ 1999; 77:
5% equivalence margin (16.75% South Africa); Juan M Lozano (Javeriana 789800.
21.75%), a total of 2269 patients would University; Bogota, Colombia); Irene 8. Kane A, Lloyd J, Zaffran M, Simonsen L, Kane
need to be recruited to complete the trial, Maulen (National Institute of Pediatrics; M. Transmission of hepatitis B, hepatitis C and
human immunodeficiency viruses through
but the current sample size was adequate Mexico City, Mexico); George McGillivray unsafe injections in the developing world:
to evaluate equivalence within a margin (University of Cape Town; Cape Town, model-based regional estimates. Bull World
of 5.8% (16.35%22.15%). The alternative South Africa); Archana Patel (Indira Gan- Health Organ 1999; 77:8017.
9. Miller MA, Pisani E. The cost of unsafe in-
was to accept less power to show equiv- dhi Medical College; Nagpur, India); Tom
jections. Bull World Health Organ 1999; 77:
alency. At the original sample size, the Sukwa and Noel Chisaka (Tropical Disease 80811.
power to retain the 5% equivalence mar- Research Centre; Ndola, Zambia); Nguyen 10. Battersby A, Feilden R, Nelson C. Sterilizable
gin (16.75%21.75%) was reduced to ngoc Tuong Vy (Childrens Hospital No. syringes: excessive risk or cost-effective op-
tion? Bull World Health Organ 1999; 77:
81.2%. After discussion between the spon- 1; Ho Chi Minh City, Vietnam); Donald 8129.
sors and the DSMB about the risk/benefit M. Thea, William B. MacLeod and Mat- 11. Straus WL, Qazi SA, Kundi Z, Nomani NK,
ratio of increasing the size of the study thew Fox (Boston University; Boston); Pa- Schwartz B. Antimicrobial resistance and clin-
ical effectiveness of co-trimoxazole versus
sample, the study continued without tricia L Hibberd (TuftsNew England
amoxycillin for pneumonia among children
change to the sample size, with the rec- Medical Center; Boston); and Shamim in Pakistan: randomised controlled trial. Pak-
ognition that the study would evaluate Qazi and Olivier Fontaine (World Health istan Co-trimoxazole Study Group. Lancet
equivalence within the 5.8% margin. The Organization; Geneva, Switzerland). 1998; 352:2704.
12. International conference on harmonisation;
proportion of patients in both groups guidance on statistical principles for clincial
whose condition did not improve or de- trials. Fed Regist 1998; 63:4958398.
teriorated continued to be monitored Acknowledgments 13. Department of Health and Human Services.
International conference on harmonisation;
throughout the study by the DSMB. Financial support. Department of Child and
Adolescent Health and Development, World choice of control group in clinical trials. Fed
Health Organization; and Applied Research in Regist 1999; 64:5176780.
STUDY OUTCOME AND 14. US Food and Drug Administration (FDA).
Child Health Project and US Agency for Inter-
CONCLUSION national Development (grant). Guidance for industry: clinical development
Conflict of interest. All authors: No conflict. for drugs, devices and biologic products for
Injectable penicillin and oral amoxicillin the treatment of rheumatoid arthritis. Wash-
were equivalent in this trial (19% of pa- ington, DC: FDA, 1999.
15. Qazi SA, Rehman GN, Khan MA. Standard
tients in each study group experienced References management of acute respiratory infections in
treatment failure after 48 h of therapy) a childrens hospital in Pakistan: impact on
1. Mulholland K. Magnitude of the problem of
[34]. The components of the composite antibiotic use and case fatality. Bull World
childhood pneumonia. Lancet 1999; 354:
Health Organ 1996; 74:5017.
outcome were almost identical in the 2 5902.
16. World Health Organization (WHO) Pro-
study groups. However, despite the results 2. Mathers CD, Murray CJL, Lopez AD, Stein C.
gramme for the Control of Acute Respiratory
The global burden of disease 2000 project: ob-
and the careful attention to threats to the jectives, methods, data sources and prelimi-
Infections. Technical basis for the WHO rec-
validity of the results, we recognize that ommendations on the management of pneu-
nary results. Evidence and information for
policy (EIP). Geneva: World Health Organi- monia in children at first level health facilities.
our study was limited by inherent biases WHO/ARI/91.20. Geneva: WHO, 1991.
zation, 2001.
toward finding equivalence if subjects are 17. Chintu C, Mudenda V, Lucas S, et al. Lung
3. World Health Organization (WHO). Acute re-
unlikely to respond to study therapy (e.g., spiratory infections in children: case manage- diseases at necropsy in African children dying
because they had nonbacterial pneumo- ment in small hospitals in developing coun- from respiratory illnesses: a descriptive nec-
tries. WHO/ARI/90.5. Geneva: WHO, 1990. ropsy study. Lancet 2002; 360:98590.
nia) and by our inability to conduct a 18. Zar HJ. Pneumocystis carinii pneumonia
4. Shann F. Etiology of severe pneumonia in chil-
blinded study. dren in developing countries. Pediatr Infect (PCP) in HIV-infected African children. SADJ
Dis 1986; 5:24752. 2001; 56:6179.
STUDY GROUP MEMBERS 5. World Health Organization. Cough or difficult 19. Bartlett JG, Dowell SF, Mandell LA, File TM
breathing. 2003. Available at: http://www. Jr, Musher DM, Fine MJ. Practice guidelines
Steering committee of the Oral Amoxi- who.int/child-adolescent-health/publications/ for the management of community-acquired
referral_care/chap3/chap31.htm#3.1. Accessed pneumonia in adults. Infectious Diseases So-
cillin versus Parenteral Penicillin for Treat-
27 July 2004. ciety of America. Clin Infect Dis 2000; 31:
ment of Severe Pneumonia International 6. Rasmussen Z, Pio A, Enarson P. Case man- 34782.
Study: Paul Arthur, Emmanuel Addo- agement of childhood pneumonia in devel- 20. McIntosh K. Community-acquired pneumo-

530 CID 2004:39 (15 August) Hibberd et al.

 nia in children. N Engl J Med 2002; 346: munodeficiency virus infection in children less ulin to reduce nosocomial infections in very-
42937. than 13 years of age. MMWR 1994; 43(RR- low-birth-weight infants. National Institute of
21. Echave P, Bille J, Audet C, Talla I, Vaudaux B, 12):119. Child Health and Human Development Neo-
Gehri M. Percentage, bacterial etiology and 25. Freemantle N, Calvert M, Wood J, Eastaugh natal Research Network. N Engl J Med
antibiotic susceptibility of acute respiratory in- J, Griffin C. Composite outcomes in random- 1994; 330:110713.
fection and pneumonia among children in ru- ized trials: greater precision but with greater 31. van Lingen RA, Simons SH, Anderson BJ, Tib-
ral Senegal. J Trop Pediatr 2003; 49:2832. uncertainty? JAMA 2003; 289:25549. boel D. The effects of analgesia in the vul-
22. Mastro TD, Nomani NK, Ishaq Z, et al. Use 26. Lauer MS, Topol EJ. Clinical trialsmultiple nerable infant during the perinatal period.
of nasopharyngeal isolates of Streptococcus treatments, multiple end points, and multiple Clin Perinatol 2002; 29:51134.
pneumoniae and Haemophilus influenzae from lessons. JAMA 2003; 289:25757. 32. Hrobjartsson A, Gotzsche PC. Is the placebo
children in Pakistan for surveillance for an- 27. ICH harmonised tripartite guideline: statistical powerless? An analysis of clinical trials com-
timicrobial resistance. Pediatr Infect Dis J principles for clinical trials. International Con- paring placebo with no treatment. N Engl J
1993; 12:82430. ference on Harmonisation E9 Expert Working Med 2001; 344:1594602.
23. Forgie IM, ONeill KP, Lloyd-Evans N, et al. Group. Stat Med 1999; 18:190542. 33. Chow SC, Shao J. A note on statistical meth-
Etiology of acute lower respiratory tract in- 28. Emanuel EJ, Miller FG. The ethics of placebo- ods for assessing therapeutic equivalence.
fections in Gambian children. II. Acute lower controlled trialsa middle ground. N Engl J Control Clin Trials 2002; 23:51520.
respiratory tract infection in children ages one Med 2001; 345:9159. 34. Addo-Yobo E, Chisaka N, Hassan M, et al. A
to nine years presenting at the hospital. Pediatr 29. Horng S, Miller FG. Is placebo surgery uneth- randomized multicentre equivalency study of
Infect Dis J 1991; 10:427. ical? N Engl J Med 2002; 347:1379. oral amoxicillin versus injectable penicillin in
24. Centers for Disease Control and Prevention. 30. Fanaroff AA, Korones SB, Wright LL, et al. A children aged 3 to 59 months with severe
Revised classification system for human im- controlled trial of intravenous immune glob- pneumonia. Lancet 2004 (in press).

Design of Pediatric Antibiotic Trials CID 2004:39 (15 August) 531