Journal of Antimicrobial Chemotherapy (2004) 53, 703–707
DOI: 10.1093/jac/dkh154
Advance Access publication 24 March 2004
Clinical significance of antiviral therapy for episodic treatment
of herpes labialis: exploratory analyses of the combined data
from two valaciclovir trials
Spotswood L. Spruance1* and Joanne Hill2
1Division
of Infectious Diseases, RM 4B319, University of Utah School of Medicine, 50 North Medical Drive,
Salt Lake City, UT 84132, USA; 2Department of Statistics, GlaxoSmithKline, Greenford, Middlesex, UK
Received 5 December 2003; accepted 21 January 2004
Valaciclovir (Valtrex) 2 g twice daily for 1 day was recently approved in the United States for treatment of cold
sores. In order to apply more clinically relevant assumptions to the analysis, we examined the effect of different missing data and endpoint assumptions on apparent valaciclovir efficacy. Results of each analysis
demonstrate statistically significant increases in the proportion of subjects whose cold sores were aborted
with valaciclovir compared with placebo, and significant decreases in healing times for subjects with cold
sore lesions who were treated with valaciclovir compared with placebo. These exploratory analyses provide
evidence of the robustness of the results to differing missing data assumptions and show that use of more
clinically relevant endpoint assumptions increases the magnitude of some therapeutic responses. We also
introduce a new measure that combines the two observed drug effects (reduced lesion duration, increased
aborted lesions) into a single endpoint that captures the global benefit of the drug to the patient.
Keywords: cold sores, HSV-1, valaciclovir
Introduction
Recurrent cold sores caused by herpes simplex virus type 1 (HSV-1)
occur in 15–40% of adults worldwide1 but finding an effective
and convenient episodic treatment has been elusive. Valaciclovir, the
L-valine ester prodrug of the acyclic guanosine analogue aciclovir,
was developed to improve aciclovir bioavailability and was recently
approved in the United States for the 1 day treatment of cold sores.
The indication was based on the results of two randomized, placebocontrolled, multi-centre clinical trials of early, high-dose valaciclovir
regimens in subjects (≥12 years of age) with histories of cold sores.2
Subjects initiated treatment upon the first symptoms of a cold sore
with either 2 g of valaciclovir twice daily for 1 day, 2 g of valaciclovir
twice daily for 1 day and then 1 g of valaciclovir twice daily for 1 day,
or matching placebo. In the 1 day treatment group, the median duration
of the episode and time to lesion healing were reduced by approximately 1 day (P < 0.001) and the proportion of subjects in whom cold
sore lesion development was aborted was increased by approximately 7% compared to placebo but did not reach statistical significance. Results were similar in the 1 and 2 day treatment groups and
between studies.
These studies provided evidence supporting a simple, safe and
effective 1 day valaciclovir treatment regimen for cold sores. In order
to apply more clinically relevant assumptions to the analysis, we
combined the data set from the two studies and examined the effect of
different missing data and endpoint assumptions on apparent drug efficacy. The missing data and endpoint assumptions for the primary and
exploratory analyses are summarized in Figure 1.
Exploratory analyses on the combined data set
The primary analysis of the aborted lesion data in these studies2
assumed that all of the subjects for whom information was unknown
were treatment failures, i.e. that vesicular lesions occurred. This is a
conservative assumption regarding the likely outcome for these subjects and almost certainly overestimated the percentage of subjects
with vesicular lesions. Additional analyses were carried out that may
be more clinically relevant whereby patients with missing information on lesion prevention were: (i) excluded, (ii) assumed to be treatment successes, i.e. assumed that vesicle formation was prevented or
(iii) imputed at the placebo rate.
In the previously published assessment of duration of episode and
time to lesion healing, subjects with missing data were assigned
either the maximum observed value plus 1 day, or assigned a value of
15 days, whichever was shorter.2 Additional analyses were carried
..................................................................................................................................................................................................................................................................
*Corresponding author. Tel: +1-801-581-8804; Fax: +1-801-585-6422; E-mail: [email protected]
...................................................................................................................................................................................................................................................................
703
JAC vol.53 no.5 © The British Society for Antimicrobial Chemotherapy 2004; all rights reserved.
Leading article
Results of the exploratory analyses
Prevention (abortion) of lesions
The proportion of subjects in whom cold sore lesion development
was aborted was 37% (223/609) of subjects in the placebo group and
44% (267/609) of subjects in the 1 day group when subjects with
missing data were assumed treatment failures (primary analysis)
(Table1). Cold sore lesion progression data were missing for 23 (4%)
placebo-treated patients and 17 (3%) valaciclovir-treated patients.
Exploratory analyses were carried out with missing data excluded,
missing data assumed as successes (i.e. lesions aborted) or missing
data imputed at the placebo rate and results are summarized in Table 1.
Results of each analysis demonstrate a statistically significant increase
in the proportion of subjects whose cold sores were aborted with
valaciclovir compared with placebo, and provide evidence of the
robustness of the results to differing missing data assumptions.
Episode duration
Figure 1. Missing data and endpoint assumptions in the primary and exploratory
analyses.
out where patients with missing data were: (i) excluded, (ii) assigned
a value of zero days or (iii) assigned the overall median value.
The initial analysis of time to lesion healing included data only for
patients with classical lesions (i.e. vesicle, ulcer and/or crusts developed).2 In order to assess the overall effect of valaciclovir on herpes
labialis lesion healing, we present a post hoc analysis that includes
cases where lesion formation was aborted. In this latter analysis,
aborted lesions are included in the lesion healing time analysis by
assigning them a classical lesion healing time of zero.
Since results for the valaciclovir 1 day and 2 day treatment arms
are comparable, we focus only on the 1 day arm here. Therefore, for
this combined retrospective analysis, the intent-to-treat (ITT) population comprised1218 subjects: 609 in the placebo group and 609 in
the 2 g valaciclovir twice daily for 1 day group. The population was
predominantly female (74%) and white (93%) and the median age was
37.0 years. The median number of years with recurrent cold sore episodes was 20.0 and the median number of episodes of cold sores in
the last 12 months was 5.0–6.0. The primary reason for the missing
data in these studies was premature discontinuation as a result of loss
to follow-up.
In the primary analysis (missing values imputed as 15 days), the duration of cold sore episodes (time to loss of crust for vesicular lesions,
and return to normal skin for non-vesicular lesions) was significantly
decreased (P < 0.001) by approximately 1 day for subjects in the valaciclovir treatment group compared to placebo (6.2 versus 5.2 days
in the placebo and 1 day groups, respectively). Data were missing for
26 (4%) placebo-treated patients and 20 (3%) valaciclovir-treated
patients. All missing data assumptions (values excluded, values
imputed as zero days, or values imputed with the overall median)
show statistically and clinically significant differences for the valaciclovir regimen over placebo and provide consistent estimates for the
treatment differences and percent reduction in episode duration
(Table 2 and Figure 2).
Episode duration is an endpoint of interest to regulatory authorities
and determines the amount of time between the onset of therapy and
loss of crust (for subjects with vesicular lesions) or the time between
the onset of therapy and the return of skin to a completely normal state
— no redness, no residual swelling (for subjects with aborted
lesions). It has the drawback of being difficult to gauge, since the time
point when redness disappears is subjective. Furthermore, persistence of redness is of little clinical significance to patients, who are
more concerned with the duration of blisters, ulcers and eschars,
which are captured by determination of ‘lesion healing time’.
Lesion healing
The mean time to cold sore lesion healing (time to loss of crust for
patients with vesicular lesions) was 6.3 and 4.9 days for the placebo
and 1 day treatment groups, respectively, in the primary analysis
(Table 3). This analysis included only subjects with vesicular lesions.
When all subjects were included in the analysis by assigning aborted
lesions a healing time of zero days, mean healing time was reduced to
4.3 and 3.1 days for the placebo and 1 day treatment groups, and the
effect of 1 day valaciclovir therapy increased from a 21% to 29%
reduction in lesion healing time relative to the primary analysis
(Figure 2). Data were missing for 27 (4%) placebo-treated patients
and 21 (3%) valaciclovir-treated patients. Excluding missing values
or assigning values of zero or the median healing time, rather than
15 days, resulted in a further decrease in healing times and an additional enhancement of the apparent valaciclovir effect. Significant
treatment differences were observed for all of the sensitivity analyses
(Table 3). The percent decrease in healing times for valaciclovir-
704
Leading article
Table 1. Comparison of differencesa in the percentage of aborted lesions using different missing data
assumptions
Aborted lesions (%)
Analyses
Primary: missing
values assumed as
treatment failures
Missing values
excluded
Missing values
imputed at placebo rate
Missing values
assumed as successes
placebo (n = 609);
missing data 4%
valaciclovir; 2 g twice
daily for 1 day (n = 609);
missing data 3%
Differencea (95% CI)
P value
37
44
7.2 (1.6, 12.9)
0.011
38
45
7.0 (1.3, 12.8)
0.015
38
45
6.9 (1.2, 12.6)
0.016
40
47
6.2 (0.5, 12.0)
0.030
P values determined by Mantel–Haenszel test stratified by study.
aDifferences (valaciclovir minus placebo) in percentage of patients with aborted lesions.
has the advantage of capturing the effect of an agent which may
increase the rate of aborted lesions, thereby expressing the drug’s
value to the patient in one global term.
Discussion
Figure 2. Impact of sensitivity analyses on efficacy results for valaciclovir 1 day
regimen. Primary analyses = missing values imputed as 15 days; Excluded =
missing values excluded; 0 days = missing values imputed as 0 days; All = all subjects included; Median = missing values imputed with the median. Percent reduction in number of days refers to the decrease in mean days needed for lesion
healing in valaciclovir-treated subjects compared with placebo.
treated subjects for the different sensitivity analyses are summarized
in Figure 2.
As shown in Table 3 and Figure 2, the effect of treatment on lesion
healing was consistently greater than on episode duration. This is
because episode duration includes aborted lesions, and valaciclovir
did not appear to affect the duration of aborted lesions (data not
shown). Using lesion healing as an endpoint has been criticized
because it excludes part of the data (aborted lesions). We feel aborted
lesions can be included in lesion healing time by assigning them a
value of zero. Since aborted lesions by definition do not contribute to
time in the painful and disfiguring vesicle, ulcer and eschar stages,
they are not ‘lesions’ in the classic sense of cold sores, and assigning
them a healing time of zero is justified. This endpoint redefinition has
the further advantage of adding a qualitative element to time, since
only the most troublesome periods of time are measured. Secondly, it
The different sensitivity analyses carried out on the data were based
on inclusion of aborted lesion data and various missing data assumptions and are exploratory analyses that present the full range of potential clinical outcomes rather than a single analysis based on the most
conservative assumptions. We feel that the assumptions we elected
are more likely to reflect clinical reality. Use of more clinically relevant
endpoint assumptions increased the magnitude of some therapeutic
responses, such as lesion healing (Figure 2). Lesion healing, although
a secondary endpoint in our studies, may actually be more important
to the patient than episode duration since the lesion stages from vesicle
to loss of crust are the most uncomfortable and disfiguring parts of the
episode.
In the primary analysis, patients with aborted lesions were
excluded from the analysis of lesion healing. However, the population of patients with aborted lesions obviously includes some
successfully treated patients, and their exclusion from the lesion healing
analysis may bias the outcome against the drug. These exploratory
analyses included aborted lesions in this measure by assigning them a
lesion healing time value of zero — reflecting the failure of a classical
lesion to form. By doing so, we combined the two observed drug
effects (reduced lesion duration, increased aborted lesions) into a
single endpoint that captures the global benefit of the drug to the
patient. This modified version of the lesion healing endpoint will
likely be useful as a prospective primary endpoint in future herpes
labialis studies. The magnitude of the treatment difference improves
from a 21% reduction in the number of days to healing in the primary
analysis to a 31% reduction for the 1 day valaciclovir group when
missing values were excluded. This clinically meaningful improvement, in association with the convenience of 1 day therapy, should
705
Leading article
Table 2. Comparison of differencesa in duration of episode using different missing data assumptions
Mean duration of episode (days)
placebo (n = 609);
missing data 4%
Analyses
Primary: missing
values imputed as 15 days
Missing values excluded
Missing values imputed
as 0 days
Missing values imputed
with the median (5 days)
valaciclovir; 2 g twice
daily for 1 day (n = 609);
missing data 3%
Differencea (95% CI)
P value
6.2
5.2
–1.0 (–1.4, –0.6)
<0.001
5.8
5.6
4.9
4.7
–1.0 (–1.3, –0.6)
–0.9 (–1.2, –0.5)
<0.001
<0.001
5.8
4.9
–0.9 (–1.2, –0.6)
<0.001
Means are adjusted for study.
aDifferences in mean days are valaciclovir minus placebo.
Table 3. Comparison of differencesa in time to lesion healing using different missing data assumptions
Mean time to lesion healing (days)
Treatment group: primary analysis
Only subjects with
vesicular lesions: missing
values imputed as 15 days
placebo (n = 363);
missing data 1%
valaciclovir; 2 g twice
daily for 1 day (n = 325);
missing data 1%
Differencea (95% CI)
P value
6.3
4.9
–1.3 (–1.7, –0.9)
<0.001
Mean time to lesion healing (days)
Treatment group: all subjectsb
Missing values imputed
as 15 days
Missing values excluded
Missing values imputed
as 0 days
Missing values imputed
with the median (4.8 days)
placebo (n = 609);
missing data 4%
valaciclovir; 2 g twice
daily for 1 day (n = 609);
missing data 3%
Differencea (95% CI)
P value
4.3
3.1
–1.2 (–1.7, –0.8)
<0.001
3.8
3.6
2.6
2.5
–1.2 (–1.5, –0.8)
–1.1 (–1.5, –0.7)
<0.001
<0.001
3.8
2.7
–1.1 (–1.5, –0.8)
<0.001
Means are adjusted for study.
aDifferences in mean days are valaciclovir minus placebo.
bPatients with aborted lesions were assigned a lesion healing time of zero days.
motivate more patients and practitioners to manage cold sores with
episodic antiviral therapy.
Aborted lesions prevent the painful and disfiguring stages of cold
sore episodes and are the most desirable outcomes of cold sore therapy.
Combined analysis of data from these trials also show that 1 day
high-dose valaciclovir therapy significantly increases the frequency
of aborted lesions by 6–7% depending on the data assumptions used
in the analysis. Although an increase in aborted lesions was noted in
one large trial of valaciclovir for treatment of recurrent genital herpes,3
this is the first time that an episodic antiviral treatment alone has
achieved a clinically and statistically significant effect against naturally occurring cold sore lesions.
706
Leading article
Acknowledgements
References
Results of these studies were described in part in an oral presentation
at the 15th International Conference on Antiviral Research, March
19, 2002, Prague, Czech Republic. Funding for this study was provided by GlaxoSmithKline.
1. Embil, J. A., Stephens, R. G. & Manuel, F. R. (1975). Prevalence
of recurrent herpes labialis and aphthous ulcers among young adults on
six continents. Canadian Medical Association Journal 113, 627–30.
2. Spruance, S., Jones, T., Blatter, M. et al. (2003). High-dose, shortduration, early valacyclovir therapy for the episodic treatment of cold
sores: results of two randomized, placebo-controlled, multicenter studies.
Antimicrobial Agents and Chemotherapy 47, 1072–80.
3. Spruance, S. L., Tyring, S. K., DeGregorio, B. et al. (1996). A
large-scale, placebo-controlled, dose-ranging trial of peroral valaciclovir
for episodic treatment of recurrent herpes genitalis. Valaciclovir HSV
Study Group. Archives of Internal Medicine 156, 1729–35.
Transparency declarations
Dr Spruance has been an investigator, speaker and consultant for
GlaxoSmithKline and a consultant for Novartis and Medivir. Ms Hill
is an employee of GlaxoSmithKline and holds stock options.
707