1057_ET_kollar 7/10/07 7:59 AM Page 1057
Clinical Therapeutics/Volume 29, Number 6, 2007
Meta-Analysis of the Efficacy of a Single Dose of
Phenylephrine 10 mg Compared with Placebo in Adults with
Acute Nasal Congestion Due to the Common Cold
Christine Kollar, MS1; Heinz Schneider, DrMed2; Joel Waksman, PhD3;
and Eva Krusinska, PhD1
1GlaxoSmithKline Consumer Healthcare, Parsippany, New Jersey; 2Consumer Healthcare Products
Association, Washington, DC; and 3Wyeth Consumer Healthcare, Madison, New Jersey
ABSTRACT
Background: Although nonprescription oral phenyl-
ephrine 10 mg has been judged “generally recognized
as safe and effective” by the US Food and Drug Ad-
ministration (FDA), its efficacy as a nasal deconges-
tant has been questioned.
Objective: This study assessed available data on
the efficacy of oral phenylephrine 10 mg as a nasal
decongestant.
Methods: Three sources were used to identify po-
tentially relevant publications—the bibliography of the
phenylephrine section of the 1976 FDA monograph
on over-the-counter cold, cough, allergy, bronchodi-
lator, and antiasthmatic products; a 2004 Cochrane
Review of nasal decongestants for the common cold;
and a search of MEDLINE from 1966 through Janu-
ary 2007 using the term phenylephrine nasal. To
be included in the analyses, studies had to have a
single-dose, randomized, placebo-controlled design; in-
volve an orally administered product in which phenyl-
ephrine 10 mg was the sole active ingredient; enroll
patients with acute nasal congestion due to the com-
mon cold; evaluate nasal airway resistance (NAR) as
the efficacy end point; and have sufficient data in the
study report to allow reanalysis and/or meta-analysis
of phenylephrine 10 mg versus placebo. Reanalysis of
individual studies and fixed-effects and random-
effects meta-analyses were performed. Statistical signifi-
cance at 30 and 60 minutes after dosing (the primary
time points) and a ≥20% reduction in NAR from base-
line were considered indicative of a clinically meaning-
ful difference.
Results: Fifteen potentially relevant studies were
identified, of which 8 met the inclusion criteria. Data
from 7 crossover studies involving a total of 113 sub-
jects were reanalyzed and then pooled for meta-analysis;
June 2007
results from the initial phase of the eighth study, a
parallel-group trial involving 50 subjects, were includ-
ed in the reanalysis of individual studies but not in the
meta-analyses. Significant differences in favor of
phenylephrine were seen in 4 of the 8 studies (P ≤
0.05). Phenylephrine 10 mg was significantly more ef-
fective than placebo at the primary time points and at
90 minutes after dosing in the meta-analyses using
both the fixed-effects and random-effects models (P ≤
0.05). At 45, 120, and 180 minutes after dosing,
phenylephrine 10 mg was also significantly more ef-
fective than placebo in the fixed-effects model (P ≤
0.05). Between 30 and 90 minutes after dosing, per-
cent reductions from baseline in NAR ranged from
6.0 percentage points higher with phenylephrine than
with placebo (at 30 and 45 minutes after dosing) to
16.6 percentage points higher (at 60 minutes after
dosing). From 60 to 180 minutes after dosing, the per-
cent reductions from baseline were generally ≥20%
with phenylephrine.
Conclusion: These meta-analyses of 7 crossover
studies and the reanalysis of a parallel-group study sup-
port the effectiveness of a single oral dose of phenyl-
ephrine 10 mg as a decongestant in adults with acute
nasal congestion associated with the common cold.
(Clin Ther. 2007;29:1057–1070) Copyright © 2007
Excerpta Medica, Inc.
Key words: nonprescription phenylephrine, nasal
airway resistance, nasal decongestant.
Accepted for publication April 12, 2007.
Express Track online publication May 31, 2007.
doi:10.1016/j.clinthera.2007.05.021
0149-2918/$32.00
Printed in the USA. Reproduction in whole or part is not permitted.
Copyright © 2007 Excerpta Medica, Inc.
1057
1057_ET_kollar 7/10/07 7:59 AM Page 1058
Clinical Therapeutics
INTRODUCTION
Phenylephrine is a sympathomimetic drug that has
been used as a nasal decongestant in the United States
and internationally since the 1940s. The Federal Food,
Drug, and Cosmetic Act of 1938, which was in effect
at the time phenylephrine was introduced, required
only proven safety, not effectiveness, for the market-
ing of a new drug. However, in 1972, the US Food and
Drug Administration (FDA) initiated a review process
to determine which over-the-counter (OTC) drugs
could be considered “generally recognized as safe and
effective” (GRASE). After receiving all available safe-
ty and efficacy information from the manufacturers of
OTC drugs and others, the FDA assembled expert ad-
visory panels to review the data and prepare OTC
drug monographs. The monographs were organized
by therapeutic class and specified the active ingredi-
ents that could be contained in nonprescription drug
products. Because phenylephrine was similar to other
active ingredients used in cough and cold medicines,
it was evaluated by the OTC Cold, Cough, Allergy,
Bronchodilator, and Antiasthmatic Drug Products Ad-
visory Review Panel, which concluded that phenyl-
ephrine was GRASE when used as a nasal decongestant
at oral doses of 10 mg.1 In 1994, the Final Mono-
graph for OTC Nasal Decongestant Drug Products
was issued, indicating that phenylephrine 10 mg was
a safe and effective nasal decongestant.2
More recently, societal and legal concerns associated
with the illicit conversion of pseudoephedrine (also a
GRASE decongestant in the OTC Nasal Decongestant
Drug Products monograph) to methamphetamine led
to the replacement of pseudoephedrine with phenyl-
ephrine in many OTC drug products. This substitution
drew renewed attention to phenylephrine’s overall effi-
cacy. The authors of a recent publication questioned
whether the advisory review panel had reached the cor-
rect conclusion when it reviewed phenylephrine in the
1970s, further noting that the review panel had cited
only 4 studies in patients with acute nasal congestion
associated with the common cold that found a signifi-
cant difference in efficacy between phenylephrine 10 mg
and placebo and 7 studies that reported no significant
difference.3 Of these 11 studies, the 4 studies that re-
ported efficacy and 4 of the other studies were company-
sponsored trials. Of the remaining 3 studies that found
no significant difference in efficacy, 1 was a company-
sponsored study in healthy volunteers, 1 was an
investigator-initiated study in patients with a variety of
1058
disorders (including chronic, allergic, and vasomotor
rhinitis), and 1 was an investigator-initiated study in
patients with chronic rhinitis.1
In response to questions about the efficacy of phenyl-
ephrine 10 mg, the Consumer Healthcare Products
Association (CHPA), the national group representing
manufacturers and distributors of OTC medicines and
dietary supplements, asked its member companies to
provide experts in biostatistics and medical science to
form a task group to review existing efficacy studies
of phenylephrine.* The task group’s first action was
to obtain copies of all studies cited in the bibliography
of the phenylephrine section of the 1976 report of the
Cold, Cough, Allergy, Bronchodilator, and Anti-
asthmatic Drug Products Advisory Review Panel.1 In
addition, the group conducted a literature search for
additional studies on the efficacy of phenylephrine.
Review of the data led the task group to conclude
that it would be feasible to conduct both a reanalysis
and a meta-analysis of the available information that
would contribute to the ongoing discussion of phenyl-
ephrine’s efficacy as a decongestant in patients with
acute nasal congestion associated with the common
cold. Reanalyses of the individual studies were con-
ducted to compare single-dose phenylephrine 10 mg
with placebo. Meta-analyses were performed to com-
pare phenylephrine 10 mg and placebo using all avail-
able raw data from the placebo-controlled, single-
dose, crossover studies. This report describes the
methods used in these analyses, and presents and dis-
cusses the results.
METHODS
Literature Searches
Three sources were used to identify relevant studies
of phenylephrine 10 mg in the treatment of acute
nasal congestion associated with the common cold.
First, the task group examined the bibliography of the
phenylephrine section of the 1976 review.1 Second,
the group examined a recent Cochrane Review of nasal
decongestants for the common cold,4 which involved
searches of MEDLINE, EMBASE, and the Cochrane
Central Register of Controlled Trials for randomized,
*The member companies of the CHPA task group were Bayer
HealthCare, GlaxoSmithKline Consumer Healthcare, Pfizer
Consumer Healthcare, McNeil Consumer Healthcare,
Novartis Consumer Health, Perrigo Company, Procter &
Gamble, Schering-Plough HealthCare Products, and Wyeth
Consumer Healthcare.
Volume 29 Number 6
1057_ET_kollar 7/10/07 7:59 AM Page 1059
placebo-controlled trials of nasal decongestants (in-
cluding phenylephrine) in adults and children with the
common cold. Third, the group conducted a search of
MEDLINE from 1966 through January 2007 employ-
ing a broad search strategy using the term phenyl-
ephrine nasal, which was automatically translated by
the PubMed search program to (“phenylephrine”
[MeSH Terms] OR phenylephrine [Text Word]) AND
((“nose” [TIAB] NOT Medline [SB]) OR “nose”
[MeSH Terms] OR nasal [Text Word]).
Study Selection Criteria
To be included in the present analyses, studies had
to have a single-dose, randomized, placebo-controlled
design; involve an orally administered product in which
phenylephrine at a dose of 10 mg was the single active
ingredient; enroll adult patients with acute nasal con-
gestion due to the common cold; have the efficacy end
point of nasal airway resistance (NAR); and contain
sufficient data in the study report (ie, individual data
for each patient and/or means [SEs]) to allow reanaly-
sis and/or meta-analysis of phenylephrine 10 mg ver-
sus placebo. Studies not meeting these criteria were
excluded.
Although pediatric data were considered, the litera-
ture search identified only 1 study conducted in chil-
dren.5 This study was not eligible for the present
analyses, as it used a fixed combination of phenyl-
ephrine, phenylpropanolamine, and brompheniramine
and included no measurement of NAR.
Efficacy Parameters
Two parameters were analyzed in the individual
study analyses and the meta-analyses. The first was
the change from baseline (before dosing) in NAR (cal-
culated as postbaseline NAR at a specific time point –
baseline NAR) at each postbaseline assessment time
(15, 30, 45, 60, 90, 120, 180, and 240 minutes after
dosing). The second was the natural logarithm of the
ratio of the postbaseline value to the baseline value,
expressed as LN-ratio NAR (calculated as LN [post-
baseline NAR] – LN [baseline NAR]) at each post-
baseline assessment time.
According to the original study reports, 5 NAR mea-
surements were taken at baseline and at each post-
baseline time point for each study subject. However,
only the means of the 5 measurements were provided
in the study reports. It can be assumed that the mean
values were rounded for reporting purposes.
June 2007
C. Kollar et al.
The following criteria were used to define a clinical-
ly meaningful result: statistical significance at 30 and
60 minutes after dosing (primary time points) and a
≥20% reduction from baseline in NAR with phenyl-
ephrine. Based on the work of Cohen,6 a ≥20% reduc-
tion from baseline is considered noticeable by patients.
Statistical Methods
The reanalyses and meta-analyses were performed
by the Biostatistics Department, GlaxoSmithKline
Consumer Healthcare, Parsippany, New Jersey. Because
the original reports of the identified crossover studies
did not provide information on the treatment se-
quence to which patients were assigned, treatment se-
quence and period could not be included in the
statistical models, nor could a test for first-order car-
ryover be performed. SAS PROC MIXED version 8.2
(SAS Institute Inc., Cary, North Carolina) was used to
generate the results. All P values for treatment-effect
terms in models 2 and 3 were considered statistically
significant at P ≤ 0.05.
Reanalyses of Individual Studies
The reanalyses of data from the individual studies
employed analysis of covariance (ANCOVA), with ad-
justment for the baseline mean measurement. For
identified crossover studies, the statistical model in-
cluded patient as a random factor; patient was also
considered as a random factor in the model for iden-
tified parallel-group studies, but the patient effect can-
not be statistically separated from the residual error in
this study design. The original studies used analysis of
variance to analyze NAR data, without a covariate to
adjust for baseline.
Meta-analyses
The meta-analyses of the 2 efficacy parameters in-
cluded individual values from the crossover studies.7,8
Both fixed-effects and random-effects analyses were
performed.9 ANCOVA was performed for all analyses,
with the baseline mean measurement as a covariate.
Before using the statistical models to compare
phenylephrine and placebo, an analysis was per-
formed to test heterogeneity at each postdose time
point; that is, to determine whether the difference be-
tween phenylephrine and placebo varied in direction
or magnitude from study to study at any postdose
time point. The purpose of this analysis was to deter-
mine whether phenylephrine differed from placebo in
1059
1057_ET_kollar 7/10/07 7:59 AM Page 1060
Clinical Therapeutics
some studies and not in others, or whether the treat-
ment difference between phenylephrine and placebo
was larger in some studies than in others at specific
time points after dosing. This test for heterogeneity
was a test of the treatment-by-study interaction term
from model 1, described in the following paragraph.
Model 1 was a fixed-effects meta-analytic model
using parametric ANCOVA, with adjustment for base-
line (a covariate) and terms for patient, study (a fixed
factor), treatment (a fixed factor), and treatment-by-
study interaction. Model 1 included 2 assumptions:
model 1a assumed patient as a fixed factor with un-
equal within-subject variance components across
studies, and model 1b assumed patient as a random
factor with unequal within-subject and between-
subject variance components across studies.
Two statistical models (models 2 and 3) were used
to perform ANCOVA comparing the efficacy of
phenylephrine and placebo at each postdose time point.
Model 2 was a fixed-effects meta-analytic model (the
same as model 1, but without the treatment-by-study
interaction term). Study was again assumed to be a
fixed factor. Model 2 included 2 assumptions: model
2a assumed patient as a fixed factor with unequal
within-subject variance components across studies,
and model 2b assumed patient as a random factor
with unequal within-subject and between-subject
variance components across studies. Model 3 was a
random-effects model with terms for baseline, patient,
treatment, study, and treatment-by-study interaction,
but with patient, study, and treatment-by-study inter-
action considered as random. This model assumed un-
equal within-subject and between-subject variance
components across studies.
The assumptions of the parametric statistical mod-
els, namely normality and equality of variance, were
checked by inspection of plots of the residuals versus
the predicted values and box plots of the residuals for
each treatment group. The 2 efficacy parameters ap-
peared to be comparable with regard to how well the
normality and equality of the variance assumptions fit
the data for the treatment factor in the model.
Differences between studies in terms of patient vari-
ability were noted in the original study reports; there-
fore, within- and between-subject variance com-
ponents were allowed to vary in analyses using models
1, 2, and 3.
The results for model 2a were generally compara-
ble to those for model 2b. Thus, in the present analy-
1060
ses, determinations concerning the efficacy of phenyl-
ephrine were based primarily on the results from model
2b and model 3 for the change-from-baseline param-
eter, which is more commonly used than the LN-ratio
for analyzing efficacy data. A sensitivity analysis was
performed using the LN-ratio parameter. Results of
the analyses of the change-from-baseline parameter
and the LN-ratio parameter were generally compara-
ble. Therefore, the results of the change-from-baseline
analyses for model 2b and model 3 are presented here.
RESULTS
The literature review identified 15 potentially relevant
studies of the efficacy of phenylephrine 10 mg as a
nasal decongestant. Of these studies, 8 randomized
placebo-controlled studies, all of which had been in-
cluded in the 1976 FDA review,1 met the criteria for
inclusion in the present analyses (Table I).10–17 The
7 excluded studies are listed in Table II.6,18–23 With
the exception of study 15,6 the excluded studies were
also included in the 1976 FDA review.
Of the 8 studies that met the inclusion criteria,
7 were randomized, double-blind, 2-treatment, 2-period,
2-sequence crossover trials with NAR as the efficacy
end point.10–16 All used the same procedure to deter-
mine NAR, employing a modified Butler-Ivy airflow
device (Sterling-Winthrop Research Institute, Rensselaer,
New York). Data from these 7 studies, which included
a total of 113 subjects, were pooled for the meta-
analyses. The eighth study was a double-blind, parallel-
group trial that was not included in the meta-
analyses.17 This study was conducted in 2 phases: the
first involved 50 patients (25 in each group) and mea-
sured NAR after a single dose of phenylephrine 10 mg
or placebo. An additional 150 patients took part in
the second phase, in which only subjective observa-
tions of the degree of relief from nasal congestion
were obtained. This second phase provided the only
data on repeated dosing: 100 patients each received
either phenylephrine 10 mg or placebo at 4-hour inter-
vals over a 12-hour period. The second phase of study 8
was not included in any of the present analyses be-
cause of its repeat-dose design and lack of objective
efficacy measurements.
In all 7 crossover studies,10–16 the active treatment
and placebo were provided in identical immediate-
release capsules supplied by the same manufacturer.
The single dose consisted of 1 capsule containing
phenylephrine 10 mg or 1 placebo capsule. In the
Volume 29 Number 6
1057_ET_kollar 7/10/07 7:59 AM Page 1061

C. Kollar et al.

Table I. Mean baseline nasal airway resistance (NAR) of subjects in studies included in the analyses.

Mean Baseline NAR*

Study No. (Design) No. of Subjects Phenylephrine Placebo

1 (Crossover)10 16 13.43 13.08
2 (Crossover)11 10 12.98 12.72
3 (Crossover)12 16 22.30 20.61
4 (Crossover)13 15 28.05 26.73
5 (Crossover)14 15 21.15 21.39
6 (Crossover)15 16 24.61 23.85
7 (Crossover)16 25 26.78 28.66
8 (Parallel group)17 50 (25 per group) 5.29 4.99

*In studies 1 through 7, NAR was measured in “units” (not defined); in study 8, NAR was measured in
cm H2O/L/min at 0.5 L/sec flow.

Table II. Studies excluded from the analyses.

Study No. Reason for Exclusion

918 Subjects were healthy volunteers
1019 Investigated phenylephrine at a dose of 25 mg but not 10 mg
1120 Investigated phenylephrine at doses of 5, 15, and 25 mg but not 10 mg
1221 Investigated phenylephrine at doses of 15, 20, and 25 mg but not 10 mg
1322 Patients had nasal obstruction due to a variety of disorders, including coryza, acute or
chronic sinusitis, allergic or vasomotor rhinitis, and hypothyroidism, and no analyses of
subgroups were performed
1423 Participants had chronic rhinitis
156 Lack of individual patient data (only mean treatment estimates were available by time
point)

parallel-group study,17 the active treatment and place-
bo were provided in identical immediate-release tablets;
2 tablets containing phenylephrine 5 mg or 2 placebo
tablets were taken every 4 hours in the repeat-dose
phase of the study.
No demographic information was provided in the
reports of the 7 crossover studies.10–16 The 50 par-
ticipants in the single-dose phase of the parallel-
group study included 22 white men, 27 white wom-
en, and 1 black woman.17 In the phenylephrine
group, the mean age was 54 years and the mean body
weight was 157 pounds. In the placebo group, the
mean age was 47 years and the mean body weight
was 164 pounds.
Reanalyses of Individual Studies
The 7 crossover studies10–16 and the first phase of
the parallel-group study17 were included in the re-
analyses of individual study data. Statistically signifi-
cant differences in the change from baseline in NAR
were found for phenylephrine compared with placebo
in studies 1,10 2,11 3,12 and 8.17 These are indicated in
Table III by negative mean treatment differences whose
95% CIs fall entirely to the left of zero.
No statistically significant differences were found
between phenylephrine and placebo in studies 4, 5, 6,
and 7.13–16 However, nonsignificant treatment differ-
ences favoring phenylephrine were found at some time
points in these studies, as indicated in Table III by nega-

June 2007 1061

1057_ET_kollar 7/10/07 7:59 AM Page 1062
Clinical Therapeutics
tive mean treatment differences whose 95% CIs in-
clude zero. The maximum percent reductions from
baseline with phenylephrine in studies 4, 5, 6, and 7
were 29%, 17%, 17%, and 16%, respectively; the
corresponding maximum percent reductions from
baseline with placebo were 32%, 21%, 22%, and
20% (data not shown).
Meta-analyses
Model 1 provided tests of treatment-by-study inter-
actions. A nominal significance level for this interac-
tion was not prespecified; however, when P ≤ 0.10
was retrospectively considered as indicating statistical
significance, significant treatment-by-study interac-
tions for phenylephrine were seen at all time points
from 15 through 180 minutes after dosing, as was ex-
pected based on the results of the individual studies. In
addition to statistically significant treatment differences
for phenylephrine in 4 of 8 studies analyzed,10–12,17
nonsignificant treatment differences favoring phenyl-
ephrine, as indicated in Table III by negative mean
treatment differences whose 95% CIs include zero, were
seen at 2 to 4 time points in all 8 studies.
For both the fixed-effects model (model 2b) and the
random-effects model (model 3), phenylephrine was
statistically significant compared with placebo at the
primary time points (30 and 60 minutes after dosing)
and at the 90-minute time point (P ≤ 0.05). In the
fixed-effects model (model 2b), phenylephrine was
also statistically significant compared with placebo at
45, 120, and 180 minutes after dosing.
The results of the meta-analyses are shown in
Table IV and Figure 1. The CIs are the 95% CIs for
the treatment difference (phenylephrine – placebo) in
change in NAR from baseline. In the figure, the width
of the CI is directly proportional to the variability (SE)
of the treatment-difference estimate; the longer the
line representing the CI, the greater the variability.
The percent change from baseline and 95% CIs for
the change in NAR from baseline with each treatment
are shown in Figure 2. The CIs shown in the figure are
the lower and upper 95% CIs of the change-from-
baseline–adjusted (least squares) treatment means
(generated from SAS PROC MIXED), expressed as a
percentage of baseline NAR.
It should be noted that the sample size was smaller
at the 180- and 240-minute time points than at earli-
er time points because only 5 studies included data for
these assessment times. Lack of statistical significance
1062
at the 120- and 180-minute time points (model 3) and
the 240-minute time point (models 2b and 3) may re-
flect reduced power, given the greater variance and/or
smaller sample sizes at these assessment times.
Using estimates taken from both model 2b and
model 3, the highest percent reductions from baseline
with phenylephrine were 4%, 9%, 15%, 21%, 21%,
23%, 25%, and 20% at 15, 30, 45, 60, 90, 120, 180,
and 240 minutes after dosing, respectively. Between
30 and 90 minutes after dosing, the percent reductions
from baseline ranged from 6.0 (at 30 and 45 minutes)
to 16.6 percentage points (at 60 minutes) higher for
phenylephrine compared with placebo.
The mean reduction from baseline in NAR was ap-
proximately two thirds to two times greater for phenyl-
ephrine than for placebo between 15 and 90 minutes
after dosing.
DISCUSSION
The effect of phenylephrine 10 mg was analyzed by
examination of its efficacy at all available time points
(from 15 to 120 minutes after dosing in all 8 studies
and up to 240 minutes after dosing in 5 studies). Such
a time-point analysis is in accordance with the FDA
Guidance for Industry on allergic rhinitis.24
Statistically significant differences in favor of
phenylephrine 10 mg over placebo were seen in 4 of
the 8 studies analyzed (P ≤ 0.05).10–12,17 Although the
direction and size of the treatment difference were not
consistently statistically significant at all time points
in all studies, nonsignificant treatment differences fa-
voring phenylephrine 10 mg over placebo were seen
at 2 to 4 time points in all 8 studies. In the 4 studies
that did not find phenylephrine effective,13–16 the maxi-
mum reductions from baseline ranged from 16% to
29% for phenylephrine and from 20% to 32% for
placebo. The findings of the present reanalysis of the
individual studies were identical to those in the origi-
nal study reports.
Both fixed-effects and random-effects meta-analyses
were performed. The results of these meta-analyses
were similar, although the strength of the statistical
evidence favoring phenylephrine over placebo was
stronger in the fixed-effects analysis. The convention-
al wisdom is that the random-effects model is preferred
when there is evidence of a significant treatment-by-
study interaction, as in these studies. However, this
view has been challenged with respect to analyses
whose purpose is to test the null hypothesis that a
Volume 29 Number 6
 Table III. Results of the reanalyses of the individual studies.
1057_ET_kollar

Time After Dosing, min

June 2007
Study No. (Design)/
7/10/07

Statistic 15 30 45 60 90 120 180 240

1 (Crossover)10
Treatment difference –1.26 –3.11 –5.74 –5.44 –4.70 –3.44 NA NA
7:59 AM

(95% CI) (–1.87 to (–3.97 to (–6.60 to (–6.64 to (–6.03 to (–4.91 to

–0.65)* –2.26)* –4.87)* –4.25)* –3.38)* –1.96)*
2 (Crossover)11
Treatment difference –0.05 –1.68 –3.51 –3.82 –2.90 –2.09 –1.17 –0.38
Page 1063

(95% CI) (–0.44 to (–2.33 to (–4.38 to (–4.64 to (–3.65 to (–2.80 to (–1.71 to (–1.05 to
0.35) –1.03)* –2.65)* –3.01)* –2.15)* –1.38)* –0.63)* 0.30)
3 (Crossover)12
Treatment difference –0.17 –2.24 –1.90 –3.14 –4.75 –4.88 –6.81 –6.66
(95% CI) (–1.70 to (–4.36 to (–4.53 to (–7.01 to (–8.90 to (–8.80 to (–11.09 to (–12.38 to
1.36) –0.12)* 0.73) 0.74) –0.59)* –0.95)* –2.52)* –0.94)*
4 (Crossover)13
Treatment difference –0.13 0.31 0.13 –1.81 0.39 1.05 0.63 0.68
(95% CI) (–1.85 to (–1.40 to (–2.47 to (–4.90 to (–2.92 to (–3.22 to (–4.62 to (–5.75 to
1.60) 2.02) 2.74) 1.29) 3.70) 5.31) 5.87) 7.12)
5 (Crossover)14
Treatment difference –0.58 –0.21 –0.07 –0.13 0.15 0.93 NA NA
(95% CI) (–1.93 to (–2.44 to (–2.46 to (–2.75 to (–2.93 to (–2.19 to
0.77) 2.03) 2.31) 2.48) 3.23) 4.05)
6 (Crossover)15
Treatment difference –0.57 –0.06 1.11 1.53 0.17 2.70 0.83 –1.65
(95% CI) (–2.82 to (–3.29 to (–1.22 to (–2.37 to (–3.62 to (–2.45 to (–4.25 to (–9.22 to
1.68) 3.17) 3.43) 5.43) 3.96) 7.84) 5.91) 5.92)
7 (Crossover)16
Treatment difference 0.99 –0.36 2.09 1.44 –0.18 2.89 1.49 1.61
(95% CI) (–0.98 to (–3.61 to (–0.88 to (–2.81 to (–4.00 to (–0.69 to (–1.05 to (–2.82 to
2.95) 2.89) 5.05) 5.70) 3.63) 6.48) 4.02) 6.03)
8 (Parallel group)17
Treatment difference –0.60 –0.67 NA –0.68 NA –0.96 NA NA
(95% CI) (–1.14 to (–1.23 to (–1.28 to (–1.48 to
–0.07)* –0.11)* –0.09)* –0.44)*
NA = not applicable (study design did not include this time point); 95% CI = lower and upper limits of the 95% CI for the treatment difference (phenylephrine – placebo).

1063
C. Kollar et al.

*P ≤ 0.05.
1064
1057_ET_kollar
7/10/07

Clinical Therapeutics
7:59 AM

Table IV. Results of the meta-analyses.

Page 1064

Time After Dosing, min

Model/Statistic 15 30 45 60 90 120 180 240

2b
Treatment difference –0.27 –1.68 –2.71 –3.68 –2.80 –2.02 –1.09 –0.33
(95% CI) (–0.61 to (–2.23 to (–3.57 to (–4.39 to (–3.54 to (–2.67 to (–1.61 to (–1.21 to
0.08) –1.14)* –1.85)* –2.97)* –2.06)* –1.37)* –0.58)* 0.55)
3
Treatment difference –0.41 –1.32 –1.38 –2.30 –2.24 –1.01 –0.95 –0.32
(95% CI) (–1.18 to (–2.56 to (–3.51 to (–4.34 to (–4.17 to (–3.42 to (–4.85 to (–1.21 to
0.36) –0.09)* 0.74) –0.26)* –0.31)* 1.40) 2.96) 0.57)
Model 2b = fixed-effects model, assuming patient as a random factor with unequal within-subject and between-subject variance components across studies; 95% CI =
lower and upper limits of the 95% CI for the treatment difference (phenylephrine – placebo); model 3 = random-effects model with terms for baseline, patient, treat-
ment, study, and treatment-by-study interaction, but with patient, study, and treatment-by-study interaction considered random.
*P ≤ 0.05.

Volume 29 Number 6
1057_ET_kollar 7/10/07 7:59 AM Page 1065

C. Kollar et al.

A
Study 1

Study 2

Study 3

Study 4

Study 5

Study 6

Study 7

Model 2b

Model 3

–9 –8 –7 –6 –5 –4 –3 –2 –1 0 1 2 3 4 5 6 7 8
Treatment Difference (95% CI)

B
Study 1

Study 2

Study 3

Study 4

Study 5

Study 6

Study 7

Model 2b

Model 3

–9 –8 –7 –6 –5 –4 –3 –2 –1 0 1 2 3 4 5 6 7 8
Treatment Difference (95% CI)

Figure 1. Treatment differences (phenylephrine – placebo) and 95% CIs for the change in nasal airway resistance
from baseline by study, assuming patient as a random factor, and in the meta-analyses using the fixed-
effects model (model 2b) and the random-effects model (model 3) at (A) 30 minutes after dosing; and
(B) 45 minutes after dosing. Statistical significance in favor of phenylephrine over placebo is indicated
when the entire CI for the treatment difference lies to the left of zero (P ≤ 0.05).
(continued)

June 2007 1065

1057_ET_kollar 7/10/07 7:59 AM Page 1066

Clinical Therapeutics

C
Study 1

Study 2

Study 3

Study 4

Study 5

Study 6

Study 7

Model 2b

Model 3

–9 –8 –7 –6 –5 –4 –3 –2 –1 0 1 2 3 4 5 6 7 8
Treatment Difference (95% CI)

D
Study 1

Study 2

Study 3

Study 4

Study 5

Study 6

Study 7

Model 2b

Model 3

–9 –8 –7 –6 –5 –4 –3 –2 –1 0 1 2 3 4 5 6 7 8
Treatment Difference (95% CI)

Figure 1 (Continued). Treatment differences (phenylephrine – placebo) and 95% CIs for the change in nasal airway
resistance from baseline by study, assuming patient as a random factor, and in the meta-analyses using
the fixed-effects model (model 2b) and the random-effects model (model 3) at (C) 60 minutes after dos-
ing; and (D) 90 minutes after dosing. Statistical significance in favor of phenylephrine over placebo is in-
dicated when the entire CI for the treatment difference lies to the left of zero (P ≤ 0.05).
(continued)

1066 Volume 29 Number 6

1057_ET_kollar 7/10/07 7:59 AM Page 1067

C. Kollar et al.

E
Study 1

Study 2

Study 3

Study 4

Study 5

Study 6

Study 7

Model 2b

Model 3

–9 –8 –7 –6 –5 –4 –3 –2 –1 0 1 2 3 4 5 6 7 8
Treatment Difference (95% CI)

F
Study 2

Study 3

Study 4

Study 6

Study 7

Model 2b

Model 3

–14 –13 –12 –11 –10 –9 –8 –7 –6 –5 –4 –3 –2 –1 0 1 2 3 4 5 6 7 8

Treatment Difference (95% CI)

Figure 1 (Continued). Treatment differences (phenylephrine – placebo) and 95% CIs for the change in nasal airway
resistance from baseline by study, assuming patient as a random factor, and in the meta-analyses using
the fixed-effects model (model 2b) and the random-effects model (model 3) at (E) 120 minutes after dos-
ing; and (F) 180 minutes after dosing. Statistical significance in favor of phenylephrine over placebo is in-
dicated when the entire CI for the treatment difference lies to the left of zero (P ≤ 0.05).

June 2007 1067

1057_ET_kollar 7/10/07 7:59 AM Page 1068

Clinical Therapeutics

Phenylephrine
Placebo
A

0
% Change from Baseline

–5

–10

–15

–20

–25

–30
0 30 60 90 120 150 180 210 240
Time (min)

–5
% Change from Baseline

–10

–15

–20

–25

–30

–35

–40
0 30 60 90 120 150 180 210 240
Time (min)

Figure 2. Percent change from baseline, with 95% CIs, at specific time points for phenylephrine and placebo in
(A) the fixed-effects model (model 2b); and (B) the random-effects model (model 3).

1068 Volume 29 Number 6

1057_ET_kollar 7/10/07 7:59 AM Page 1069
treatment has no effect.8 Under this hypothesis, the
treatment has no effect in any study, and thus there is
no treatment-by-study interaction. Hence, the fixed-
effects model is more appropriate when testing
whether a treatment has an effect.
The meta-analyses were based on the raw data for
individual patients rather than on the treatment means.
A key advantage of this approach is that it allowed
adjustment of the analyses for the patient’s baseline
NAR (a patient-level covariate in the statistical mod-
els). This adjustment corrected for any baseline imbal-
ances in NAR between phenylephrine and placebo
(which could have confounded the results) and in-
creased the statistical power of the analyses.
In the meta-analyses using both the fixed-effects
and random-effects models, phenylephrine was statis-
tically significantly more effective than placebo at the
primary time points, 30 and 60 minutes after dosing,
and at 90 minutes after dosing (P ≤ 0.05). In addition,
phenylephrine was significantly more effective than
placebo at 45, 120, and 180 minutes after dosing in
the fixed-effects model. Between 30 and 90 minutes
after dosing, the percent reductions from baseline
ranged from 6.0 percentage points (at 30 and 45 min-
utes) to 16.6 percentage points (at 60 minutes) higher
for phenylephrine compared with placebo. The reduc-
tions from baseline with phenylephrine were general-
ly ≥20% from 60 to 180 minutes after dosing (con-
sidered clinically meaningful). In the fixed-effects
analysis, the reductions from baseline in NAR with
phenylephrine were at least 21% from 60 to 180 min-
utes after dosing. In the random-effects analysis, the
reduction was 18% at 60 minutes after dosing and at
least 20% from 90 to 180 minutes after dosing.
The parallel-group study (study 8) was not includ-
ed in the meta-analyses.17 However, the results of this
study were consistent with the outcome of the meta-
analyses in terms of phenylephrine’s efficacy compared
with placebo.
The studies included in the meta-analyses had a
crossover design, with a 1-day washout between treat-
ments. The treatment sequence to which subjects were
randomized was not provided in the original study
reports. The original reports did not mention any
analysis of treatment sequence with regard to a possi-
ble carryover effect, but they gave no indication of a
carryover effect. Phenylephrine’s short half-life ren-
ders a pharmacologic carryover effect in these studies
unlikely.
June 2007
C. Kollar et al.
There is no clear cause for the disparities in esti-
mated treatment differences between phenylephrine
and placebo in the various studies. The original study
reports did not provide information allowing determi-
nation of whether the populations were homogeneous
across studies. Also, no information was provided that
indicated whether differences in baseline NAR could
have been a factor in the differences in results across
studies. (Despite including patients with a relatively
low baseline NAR and, presumably, less severe con-
gestion, study 8 was still able to discriminate between
phenylephrine and placebo.17) Among the 8 studies
included in these analyses, 3 of the 4 studies that re-
ported a statistically significant outcome in favor of
phenylephrine 10 mg had the smallest variances (stud-
ies 1,10 2,11 and 817). This lower variability increased
the sensitivity (power) of the studies.
Heart rate and blood pressure readings were ob-
tained in 5 of the 8 studies. Slight (5% at most), sta-
tistically nonsignificant increases in heart rate and sys-
tolic blood pressure were observed with phenylephrine
10 mg. No adverse events were documented in the
7 crossover studies.10–16 In the parallel-group study
involving repeated administration of phenylephrine
10 mg or placebo,17 the nature and frequency of ad-
verse events were comparable between groups. Eight
of 100 patients in the active-treatment group and 11 of
100 patients in the placebo group reported adverse
events.
The data search used in these analyses was not re-
stricted to studies in which only a single dose of
phenylephrine was administered. However, the second
phase of study 8 was the only study arm identified
in which patients received multiple doses of phenyl-
ephrine or placebo over a 12-hour period.17 In that
phase, in which patients provided subjective evalua-
tions of the degree of relief from nasal congestion, the
phenylephrine group reported significantly greater re-
lief compared with those in the placebo group at all
time points (P ≤ 0.05). Tachyphylactic effects have
been seen with multiple dosing of sympathomimetic
drugs in animals,25 but no studies were identified that
investigated the potential of oral phenylephrine to
produce tachyphylactic effects in humans.
CONCLUSION
These meta-analyses of 7 crossover studies and the re-
analysis of a parallel-group study support the effec-
tiveness of a single oral dose of phenylephrine 10 mg
1069
1057_ET_kollar 7/10/07 7:59 AM Page 1070
Clinical Therapeutics
as a decongestant in adults with acute nasal conges-
tion associated with the common cold.
ACKNOWLEDGMENTS
The authors of the present report were members of
the CHPA task group on phenylephrine 10 mg. The
authors thank Dallas Johnson, PhD, Department of
Statistics, Kansas State University, for contributions to
the resolution of technical matters related to the out-
put generated from SAS PROC MIXED.
REFERENCES
1. Dept of Health and Human Services, Food and Drug
Administration. Establishment of a monograph for OTC
cold, cough, allergy, bronchodilator and anti-asthmatic
products. Fed Regist. 1976;41:38399–38400.
2. Dept of Health and Human Services, Food and Drug
Administration. Cold, cough, allergy, bronchodilator and
anti-asthmatic drug products for over-the-counter human
use; final monograph for OTC nasal decongestant drug
products. Fed Regist. 1994;59:43386–43412.
3. Hendeles L, Hatton RC. Oral phenylephrine: An ineffective
replacement for pseudoephedrine? J Allergy Clin Immunol.
2006;118:279–280.
4. Taverner D, Latte J, Draper M. Nasal decongestants for the
common cold. Cochrane Database Syst Rev. 2004;3:CD001953.
5. Hutton N, Wilson MH, Mellits ED, et al. Effectiveness of
an antihistamine-decongestant combination for young
children with the common cold: A randomized, con-
trolled clinical trial. J Pediatr. 1991;118:125–130.
6. Cohen BM. Clinical and physiological “significance” of
drug-induced changes in nasal flow/resistance. Eur J Clin
Pharmacol. 1972;5:81–86.
7. Whitehead A. Meta-Analysis of Controlled Clinical Trials.
Chichester, UK: John Wiley & Sons; 2002.
8. Senn S. The many modes of meta. Drug Info J. 2000;34:535–549.
9. DerSimonian R, Laird N. Meta-analysis in clinical trials.
Control Clin Trials. 1986;7:177–188.
10. Memo to Wessinger from NA Hulme. Nasal decongestant
study by Elizabeth Biochemical No 2. In: FDA OTC Volume
040298. January 1968.
11. Memo to Blackmore from NA Hulme. Neo-Synephrine—
Elizabeth Biochemical Laboratory Study No 5. In: FDA
OTC Volume 040298. May 1970.
12. Memo to Blackmore from NA Hulme. Oral Neo-
Synephrine—Cintest Labs Study No 1. In: FDA OTC Volume
040298. May 1969.
Address correspondence to: Heinz Schneider, DrMed, Vice President,
Regulatory & Scientific Affairs, Consumer Healthcare Products
Association (CHPA), 900 19th Street NW, Suite 700, Washington, DC
20006. Email: hschneider@chpa-info.org
1070
13. Memo to Blackmore from NA Hulme. Oral Neo-
Synephrine—Cintest Labs Study No 2. In: FDA OTC Volume
040298. January 1970.
14. Memo to Blackmore from NA Hulme. Oral Neo-
Synephrine—Cintest Labs Study No 3. In: FDA OTC
Volume 040298. May 1970.
15. Memo to Blackmore from NA Hulme. Oral Neo-
Synephrine—Huntingdon Research Center Study No 1. In:
FDA OTC Volume 040298. May 1969.
16. Memo to Blackmore from NA Hulme. Oral Neo-
Synephrine—Huntingdon Research Center Study No 2. In:
FDA OTC Volume 040298. January 1970.
17. Cohen BM. Objective and subjective evaluation of
phenylephrine HCl (5 mg) versus placebo tablets. In: FDA
OTC Volume 04088B. June 1975.
18. Memo to Lands from FP Luduena. Comparative study of
the effects of Neo-synephrine HCl and Propadrine HCl on
nasal air resistance (NAR), blood pressure and pulse rate
of volunteers. In: FDA OTC Volume 040298. April 23,
1959.
19. Memo to Suter from NA Hulme. Nasal Decongestant
Study by Elizabeth Biochemical Laboratories No 1. In:
FDA OTC Volume 040298. June 1967.
20. Memo to Blackmore from NA Hulme. Oral Neo-
Synephrine—Elizabeth BioChemical Study No 3. In: FDA
OTC Volume 040298. June 1969.
21. Memo to Blackmore from NA Hulme. Oral Neo-
Synephrine—Elizabeth Biochemical Study No 4. In: FDA
OTC Volume 040298. August 1969.
22. McLaurin JW, Shipman WF, Rosedale R Jr. Oral deconges-
tants. A double blind comparison study of the effective-
ness of four sympathomimetic drugs: Objective and
subjective. Laryngoscope. 1961;71:54–67.
23. Bickerman HA. Physiologic and pharmacologic studies on
nasal airway resistance (RN). In: Current Research
Methodology in the Evaluation of Proprietary Medicines:
Proceedings of a Conference Sponsored by the Scientific Develop-
ment Committee of the Proprietary Association. New York, NY:
The Proprietary Association; 1971:61–72.
24. US Dept of Health and Human Services, Food and Drug
Administration, Center for Drug Evaluation and Research.
Guidance for industry. Allergic rhinitis: Clinical develop-
ment programs for drug products. April 2000. Available
at: http://www.fda.gov/CDER/guidance/2718dft.pdf.
Accessed October 16, 2006.
25. Morishita H, Furukawa T. The vascular changes after
ephedrine tachyphylaxis. J Pharm Pharmacol. 1975;27:574–
579.
Volume 29 Number 6