J Periodontol May 2004

679
The Effect of Aspirin Intake on Bleeding
on Probing in Patients With Gingivitis
Daniel Royzman,* Luisa Recio,

Rachel L. Badovinac,

Joseph Fiorellini,

Max Goodson,

Howard Howell,

and Nadeem Karimbux

Background: Bleeding indices are used as a screen for perio-

dontal disease activity, a measure of disease prevalence, and a
measure of effectiveness in clinical trials. Bleeding on probing
(BOP) is widely interpreted as a sign of disease activity whereas
its absence is interpreted as both a sign and predictor of health.
Aspirin use has become increasingly common in the prevention
of cerebrovascular and cardiovascular diseases. Because of its
anti-platelet activity, aspirin is a non-disease factor that has the
potential to affect the appearance of BOP. The hypothesis being
tested is that short-term aspirin use in doses of 81 mg and 325 mg
will increase the number of bleeding sites in a population with
gingivitis.
Methods: Fifty-four subjects were screened initially, those sub-
jects with 20% to 30% whole mouth BOP were randomly assigned
to one of three arms: placebo group, 81 mg aspirin group, or
325 mg aspirin group. Before and after exposure to the respec-
tive regimens, clinical parameters were measured on all the teeth:
the plaque index was recorded at four sites per tooth, and probing
depth and BOP were evaluated at six sites per tooth using an
automated pressure-sensitive probe.
Results: The data obtained in this clinical trial were analyzed
utilizing a linear regression analysis to control for confounding
variables. The primary measure of interest was BOP in patients
clinically demonstrating naturally occurring gingivitis. The results
of this study indicate that while controlling for age, gender, and
plaque, low dose 81 mg and regular dose 325 mg of aspirin
demonstrated a statistically signicant 5.30 (P = 0.001) and 4.13
(P = 0.010) increase from baseline, respectively, in percent BOP.
Conclusion: Failure to consider the effects of aspirin on BOP
could impair proper diagnosis and treatment planning for clini-
cians and introduce a signicant confounding variable in research
situations. J Periodontol 2004;75:679-684.
KEY WORDS
Aspirin/adverse effects; bleeding/prevention and control;
clinical trials; dental probes; gingivitis/diagnosis.
* Private practice, New York, NY.
Private practice, Boston, MA; previously, Department of Oral Medicine, Infection
and Immunity, Harvard School of Dental Medicine, Boston, MA.
Department of Oral Health Policy and Epidemiology, Harvard School of Dental Medicine.
Department of Epidemiology, Harvard School of Public Health, Boston, MA.
Department of Oral Medicine, Infection and Immunity, Harvard School of Dental
Medicine.
The Forsyth Institute, Boston, MA.
A
spirin has been recommended by
the American Heart Association as
a therapeutic agent for cardiovas-
cular disease.
1
In addition, it is commonly
used to treat rheumatoid arthritis, osteo-
arthritis, rheumatic fever, and other inam-
matory joint diseases. Due to aspirins
effect on bleeding, failure to consider its
intake during a routine clinical dental
examination could produce various false-
positive readings, which could result in
an inaccurate patient diagnosis. Aspirin,
a non-steroidal anti-inammatory drug
(NSAID), is commonly distributed in 81 mg
and 325 mg doses. It is absorbed from
the duodenum and converted to salicilate.
Metabolically, it binds irreversibly with
cyclooxygenase and inhibits the release of
prostaglandins, prostocyclins, and throm-
boxanes. Aspirin via the cyclooxygenase
pathway inhibits thromboxanes, which are
responsible for platelet aggregation for a
period of 7 to 10 days.
2
This change in
platelet aggregation could increase the
chances of bleeding and bleeding on prob-
ing for that period of time.
Previous research by Schrodi et al.
3
examined the effect of aspirin on BOP in
patients with a clinically healthy peri-
odontium. Their results demonstrated
that aspirin intake of 325 mg daily for
7 days moderately increased the appear-
ance of BOP in a population that had
greater or equal than twenty percent BOP
sites. Based on the results of the Schrodi
et al. study,
3
the aim of this study is to
determine the relationship between BOP
and a 1-week intake of 81 mg or 325 mg
aspirin in individuals who exhibit 20% to
30% BOP.
3081.qxd 5/10/04 2:29 PM Page 679
680
Aspirin and Gingival Bleeding Volume 75 Number 5
MATERIALS AND METHODS
The clinical study was conducted at the Harvard School
of Dental Medicine and at the Forsyth Institute. The
human subjects review committees of both institutes
approved this protocol, and the study was conducted
in compliance with institutional review board and
informed consent regulations. No monetary compen-
sation was given for participation.
Patient Selection and Clinical Procedures
Subjects with at least 20 teeth excluding third molars,
with probing depths (PD) <5 mm at all sites, and 20%
to 30% whole mouth BOP were entered into the study.
Exclusion criteria were any known contraindication to
aspirin intake, compromising systemic medical condi-
tions, any form of ongoing tobacco use, pregnancy or
lactation, any condition requiring antibiotic premedica-
tion for the prevention of subacute bacterial endocarditis,
a history of systemic or topical use of any type of NSAID
within 2 weeks prior to the study or requiring the use
of any type of NSAID during the study period, previous
diagnosis of chronic periodontitis (ADA type II, type III
or type IV), and periodontal treatment other than a rou-
tine prophylaxis within the previous 12 months.
All of the participating subjects received an exami-
nation of soft and hard tissues and four bitewing radio-
graphs to assess any possible carious lesions and/or
bone loss. Measures of PD, BOP, and plaque index (PI)
were measured and recorded. PD and BOP were mea-
sured at six sites per tooth (mesio-buccal, mid-buccal,
disto-buccal, mesio-lingual, mid-lingual, disto-lingual)
on a maximum of 28 teeth for a maximum total of 168
sites. To reduce error, the PD was recorded using a
controlled force periodontal probe,
#
which has a stan-
dardized force of 0.25 N, a resolution of 0.1 mm for
recording probing depths, and which is linked to a com-
puter database for direct chairside data analysis. PD
for each subject was the average taken over six sites
per tooth. BOP was considered positive if bleeding was
elicited within 30 seconds following the insertion of the
probe to measure the PD. The percentage of sites that
bled was determined for each subject. The hygiene
analysis index (HAI) by Love et al. was used to mea-
sure plaque.
4
The HAI was obtained by measuring the
presence or absence of plaque on four surfaces (mesial,
distal, facial, and lingual) of all qualifying teeth. The HAI
score was expressed as a percentage determined by
adding the number of plaque sites per tooth and divid-
ing by the total number of sites examined. A single
calibrated investigator, who was masked to group mem-
bership, was involved in carrying out all the examina-
tions and data collection.
At day zero, baseline measures of BOP, PD, and PI
were taken. All subjects received instructions and were
provided with seven placebo capsules to be taken for
7 consecutive days. Subjects were instructed to return
at day 7 and bring any remaining capsules to determine
compliance. At day 7 subjects returned for a brief
appointment (no measures taken) and were random-
ized to one of the treatment arms (placebo, 81 mg of
buffered ASA, or 325 mg of buffered ASA). The sub-
jects were issued the appropriate package of tablets,
with the instructions to consume one each day, for 7
consecutive days. At day 14, subjects were re-examined
for outcome measures BOP, PD, and PI. After com-
pletion of data collection, each subject received an
appointment for a prophylaxis.
Statistical Analysis
In all analyses, treatment was entered as three sham
variables, which served to compare change from base-
line to 14 days in the placebo (placebo: day 14 versus
baseline), 81 mg (81 mg: day 14 versus baseline), and
325 mg (325 mg: day 14 versus baseline) groups, res-
pectively, while age, PD, and plaque were entered as
continuous variables.
Bleeding on probing. The primary aim of this study
was to assess the effect of week-long aspirin therapy
on BOP. To accomplish this, linear regression analysis
with robust standard errors was used to assess the uni-
variate relationships between each of the variables and
BOP. Multivariate analysis was performed to determine
the effect of each treatment on change in BOP relative
to baseline, controlling for plaque, age, and gender. It
was decided a priori to enter age, gender, and PI into
the model because of imbalances between the groups
and the inherent potential of these variables to act as
confounding factors. Additionally, we considered adding
baseline PD to the model on the basis of univariate
regression. If the relationship between baseline PD and
BOP were statistically signicant, we would include the
term in the multivariate model. Alpha was set at 0.05
in all cases. Analyses were performed using a statisti-
cal software program.** In both the univariate and mul-
tivariate linear regression analyses, the cluster option
was specied in the program to account for the repeated
measures at baseline and after treatment.
Plaque. Two different types of analyses were com-
pleted to determine the relationships between clinical
and demographic factors and baseline plaque, and the
relationship between the three treatment groups and
change in plaque relative to baseline. A standard
univariate linear regression was used because there
were no repeated measures in the analysis of the rela-
tionship between baseline BOP and PI, age, gender,
and PD, respectively. By contrast, when assessing the
relationship between treatment and change in plaque
relative to baseline, repeated measures were taken into
# Florida Probe Corporation, Gainesville, FL.
** STATA, College Station, TX.
3081.qxd 5/10/04 2:29 PM Page 680
681
J Periodontol May 2004 Royzman, Recio, Badovinac, et al.
account through the specication of the cluster option
in the statistical program.
Probing depth. Two different types of analyses were
similarly completed to determine the relationships
between clinical and demographic factors and baseline
PD and the relationship between the three treatment
groups and change in PD relative to baseline. Standard
linear regression was used in the former aim, while
linear regression with the cluster option specied was
used in the latter.
RESULTS
Study Population
A total of 100 subjects were invited to participate in the
study. Twenty-nine were excluded due to adverse reac-
tions to aspirin, history of smoking, taking birth control
pills, requiring antibiotic prophylaxis, or for personal rea-
sons. Thirteen subjects were excluded due to periodon-
tal bone loss. Fifty-eight subjects were enrolled who met
the inclusion and exclusion criteria described above.
Fifty-four subjects completed the clinical trial. Four
subjects were lost during the 1-week run-in period on
placebo pills: one subject left claiming an adverse reac-
tion to the capsules and three were lost to fol-
low-up failure. Seventeen subjects were
treated with placebo, 17 with low dose
aspirin (81 mg), and 20 with regular dose
(325 mg) of aspirin. The study population
demographics and pre-randomization clinical
measures are presented in Table 1. Despite
the randomized study design, there were
imbalances between the groups. A placebo
group member was more likely to be female,
to be younger, and to have a lower mean PI
score than an experimental group member.
The statistical analyses took these differences
into account.
Bleeding on Probing
The univariate relationships between BOP and
female gender, plaque, PD, and type of treat-
ment arm (placebo, 81 mg or 325 mg aspirin)
were assessed (Table 2). Treatment with both
81 mg and 325 mg aspirin were signicantly
related to BOP (Fig. 1). The coefcient asso-
ciated with plaque was 0.19 (P = 0.007; 95%
confidence interval [CI]: 0.056, 0.33). The
coefficient associated with treatment with
81 mg aspirin relative to BOP change from
baseline was 4.76 (P = 0.008, 95% CI: 1.29,
8.24), and the coefficient associated with
treatment with 325 mg aspirin relative to BOP
change from baseline was 3.46 (P = 0.034,
95% CI: 0.26, 6.65). Treatment with placebo
was not associated with any signicant change
in BOP relative to baseline.
Table 1.
Study Population Demographics
and Pre-Randomization Clinical Measures
Aspirin Aspirin
Characteristic Placebo (81 mg/daily) (325 mg/daily)
Total N 17 17 20
N males (%) 2 (11.8) 7 (41.2) 8 (40.0)
N females (%) 15 (88.2) 10 (58.8) 12 (60.0)
Group mean 25.9 (5.9) 27.6 (5.6) 29.3 (3.9)
age (SD)
Group mean % 24.8 (3.5) 25.4 (2.4) 24.7 (3.3)
BOP (SD)
Group mean PD 1.8 (0.2) 1.9 (0.2) 1.8 (0.3)
in mm (SD)
Group mean PI 15.7 (6.3) 21.7 (10.3) 19.3 (9.2)
score (SD)
Table 2.
Univariate and Multivariate Linear Regressions
Between BOP and Treatment, Gender, Plaque
Index, Age, and PD
Variable Coefcient Robust SE P Value 95% CI
Univariate
Age 0.051 0.16 0.76 0.28, 0.38
Female versus male 0.22 1.23 0.86 2.69, 2.25
Plaque index 0.19 0.069 0.007 0.056, 0.33
PD 0.99 3.07 0.75 5.18, 7.16
Treatment
Placebo: day 14 versus 0.47 1.67 0.78 2.88, 3.82
baseline
81 mg: day 14 versus 4.76 1.73 0.008 1.29, 8.24
baseline
325 mg: day 14 versus 3.46 1.59 0.034 0.26, 6.65
baseline
Multivariate
Age 0.074 0.14 0.61 0.22, 0.36
Female versus male 2.36 1.37 0.091 0.39, 5.11
Plaque 0.27 0.07 0.001 0.12, 0.42
Treatment
Placebo: day 14 versus 1.90 1.74 0.28 1.58, 5.39
baseline
81 mg: day 14 versus 5.30 1.55 0.001 2.20, 8.41
baseline
325 mg: day 14 versus 4.13 1.55 0.010 1.03, 7.24
baseline
3081.qxd 5/10/04 2:29 PM Page 681
682
Aspirin and Gingival Bleeding Volume 75 Number 5
As previously mentioned, we decided a priori to
include variables age, gender, and PI in the multivari-
ate model due to imbalances in the treatment groups
and the potential for confounding. The decision on
whether to include PD was based upon statistical sig-
nicance in univariate analysis. As it did not achieve
statistical signicance, it was excluded from the multi-
variate model. In this context, the coefcient associated
with plaque was 0.27 (P = 0.001, 95% CI: 0.12, 0.42).
Placebo had no signicant effect on BOP relative to
baseline, while both 81 mg and 325 mg did have an
effect on BOP. The coefcient associated with 81 mg
was 5.30 (P = 0.001, 95% CI: 1.29, 8.24), and the coef-
cient associated with 325 mg was 4.13 (P = 0.010,
95% CI: 1.03, 7.24). Based upon the multivariate model,
we would predict that a 28 year-old woman with a PI
of 18.9 would have a percentage of sites with bleeding
on probing of 9.53 at baseline. At day 14, her predicted
percentage of sites with bleeding on probing would be
11.43, 14.84, or 13.67, depending upon whether she
were in the placebo, 81 mg, or 325 mg group.
Plaque
The factors associated with baseline plaque (Table 3)
were analyzed and the treatment with placebo, 81 mg
aspirin, or 325 mg aspirin was determined with respect
to plaque relative to baseline (Table 4). All of these
analyses were performed on a univariate basis. Analy-
sis of the demographic and clinical factors revealed
that only gender had a significant relationship with
baseline plaque. The point estimate revealed that
womens PI scores were 7.97 (P = 0.002, 95% CI: 12.81,
3.12) lower than the mens. Interestingly, placebo was
the only treatment significantly associated with a
change in plaque relative to baseline. The coefcient
associated with placebo was 6.60 (P <0.001, 95% CI:
9.50, 3.70).
Probing Depth
The results of the univariate linear regressions to deter-
mine the associations between age, PI, gender, and BOP,
and PD may be found in Table 5. The results of the
linear regression between treatment and change in PD
from baseline is found in Table 6. None of the clinical
Figure 1.
Change in % BOP from baseline by treatment group. X-axis: treatment
group;Y-axis: mean change from baseline in percentage of sites that
exhibited bleeding on probing, as derived from the univariate linear
regression. Error bars: robust standard errors, derived from the
univariate linear regression.
Table 3.
Univariate Linear Regression Between
Baseline Plaque Index and Age, PD,
and BOP
Variable Coefcient SE P Value 95% CI
Female versus 7.97 2.41 0.002 12.81, 3.12
male
Age 0.05 0.24 0.83 0.53, 0.43
PD 6.89 5.26 0.20 3.66, 17.44
BOP 0.82 0.38 0.036 0.055, 1.59
Table 4.
Univariate Linear Regression Between
Plaque Index and Treatment
Robust
Treatment Coefcient SE P Value 95% CI
Placebo: day 14 6.60 1.45 <0.001 9.50, 3.70
versus baseline
81 mg: day 14 1.15 2.02 0.57 5.21, 2.91
versus baseline
325 mg: day 14 2.22 1.72 0.20 5.67, 1.22
versus baseline
Table 5.
Univariate Linear Regression Between
Baseline PD and Age, Plaque Index,
and BOP
Variable Coefcient SE P Value 95% CI
Female versus 0.022 0.069 0.75 0.12, 0.16
male
Age 0.0041 0.0062 0.51 0.16, 0.0083
Plaque 0.0046 0.0035 0.20 0.0025, 0.012
BOP 0.012 0.010 0.24 0.0084, 0.033
3081.qxd 5/10/04 2:29 PM Page 682
683
J Periodontol May 2004 Royzman, Recio, Badovinac, et al.
or demographic factors were statistically signicantly
associated with baseline PD, but treatment with both
aspirin regimens was signicantly associated with a
change in PD. The coefcient associated with 81 mg
relative to baseline was 0.11 (P = 0.012, 95% CI:
0.19, 0.024), and the coefcient associated with 325
mg was 0.18 (P <0.001, 95% CI: 0.26, 0.091).
DISCUSSION
BOP is used by epidemiologists to measure disease
prevalence and progression and by clinicians to aid in
diagnosis, to measure outcomes of treatment, and to
motivate patients with their home care.
5
Researchers
use BOP as an outcome to measure the effectiveness
of surgical and non-surgical therapies in treating gin-
givitis and periodontitis. A previous study by Heasman
et al. examined the anti-inammatory effects of NSAIDS
on bleeding indices in patients with naturally occur-
ring gingivitis.
6
Schrodi et al.
3
investigated the anti-
thrombolytic effects of aspirin on BOP in patients with
a healthy periodontium. Their results demonstrated
that 81 mg (low dose) or 325 mg (regular dose) aspirin
did not have an effect on BOP in healthy individuals.
However there was a signicant increase in BOP in a
subset of patients who had >28% of whole mouth BOP
taking 325 mg aspirin. Although their sample size was
small, the signicance of their ndings led to the cur-
rent study, which attempted to isolate aspirins anti-
thrombolytic effect on BOP in patients with naturally
occurring gingivitis dened as BOP ranging from 20%
to 30% of a possible 168 sites (28 teeth).
All 54 patients who were placed in randomized
groups completed the trial. The data obtained in this
clinical trial were analyzed utilizing a linear regression
analysis to control for confounding variables. The pri-
mary measure of interest was BOP in patients clinically
demonstrating naturally occurring gingivitis. The results
of this study indicate that by controlling for age, gender,
and plaque, low dose 81 mg and regular dose 325 mg
aspirin demonstrated a statistically signicant 5.30 (P =
Table 6.
Univariate Linear Regression Between PD
and Treatment
Robust
Treatment Coefcient SE P Value 95% CI
Placebo: day 14 0.0026 0.045 0.954 0.093, 0.088
versus baseline
81 mg: day 14 0.11 0.041 0.012 0.19, 0.024
versus baseline
325 mg: day 14 0.18 0.043 <0.001 0.26, 0.091
versus baseline
0.001) and 4.13 (P = 0.010) increase, respectively, in
percent BOP (Table 2). However, even the subjects in
the placebo group experienced a non-signicant increase
in percentage of sites exhibiting BOP (P = 1.90, P =
0.28), revealing a relative increase of 3.40 and 2.23
in the 81 and 325 mg groups, respectively.
Of secondary interest were PD and plaque. The PD
in the population ranged from 1 to 4 mm, as was
expected in a population that was screened to exclude
periodontal bone loss. The overall average PD ranged
from 1.80 to 1.90. It is especially interesting to note
that for both 81 mg and 325 mg treatment groups,
there was a statistically significant increase in BOP
with a decrease in PD. The decrease in PD could be
attributed to the anti-inammatory effects of aspirin,
while the increase in BOP could be attributed to the
anti-thrombolytic effects of aspirin. This hypothesis is
the subject of an additional study. These results could
also be due to the Hawthorne effect, where subjects
altered their regular practices due to their participation
in a study.
The other secondary measure of interest was plaque.
When controlling for age, gender, and treatment modal-
ity, for every one unit increase in plaque there was
0.27 (P = 0.001) increase in percent of BOP. This posi-
tive relationship is supported by an overwhelming
amount of evidence, which suggests that microbial
plaque near the cervical region of teeth causes gin-
givitis.
7,8
Additionally, univariate linear regression
between gender and PI indicates that women began
with a 7.97 lower baseline plaque than men (P =
0.002). This result corroborates previous ndings by
Schrodi et al.
3
and epidemiological studies by Marshall-
Day et al.
9
and Albandar and Kingman,
10
who describe
females as having better oral hygiene, less gingival
bleeding, and less subgingival calculus compared to
males.
Evaluating the effect of treatment on plaque, it is
interesting to point out that the placebo subjects experi-
enced a 6.6 (P 0.001) decrease in PI from baseline,
while subjects in the aspirin groups did not demon-
strate statistically significant decreases in PI. These
results could be attributed to the fact that the placebo
group was mainly composed of female dental hygien-
ists and that, in the context of the study, this group
practiced a better oral hygiene regimen, thus resulting
in lower plaque scores.
In 1998, Bode-Boger et al. reported that 100 mg
enteric-coated aspirin had an effect on platelet aggre-
gation, thromboxane 2 production, and urinary excre-
tion of 2,3-dionr-TXB2 after 7 days of drug exposure.
11
Their results indicated that 100 mg of enteric-coated
aspirin signicantly reduced platelet aggregation after
4 hours of drug intake. The thromboxane 2 level, which
was signicantly reduced after 7 days of the initial drug
uptake, was completely recovered on day 8. Our study
3081.qxd 5/10/04 2:29 PM Page 683
684
Aspirin and Gingival Bleeding Volume 75 Number 5
did not examine blood samples to determine aspirins
effect on platelet function. These additional measure-
ments were not considered necessary considering that
studies demonstrated that the systemic circulation of
aspirin irreversibly binds cyclooxygenase and in effect
inhibits platelet aggregation for a period of 7 to 10 days,
when new platelets will be formed.
1
The results of this study demonstrate widespread
signicance, challenging previously-held notions that
BOP is directly caused by inammation. The ndings
of this study lend credence to the notion that BOP may
be related to an inadequate function of platelets in
patients taking aspirin. Failure of clinicians to take into
account aspirin use in patients undergoing periodon-
tal treatment could lead to false-positive results, which
would lead to an improper diagnosis, treatment choice,
and assessment of disease activity and progression. In
research settings, failure to exclude patients taking
aspirin could signicantly alter the sensitivity and speci-
city if one outcome being measured is BOP.
REFERENCES
1. Hennekens CH, Dykan ML, Fuster V. Aspirin as a thera-
peutic agent in cardiovascular disease; a statement for
health care professionals from the American Heart Asso-
ciation. Circulation 1997;96:2751-2753.
2. Craig CR, Stitzel RE. Modern Pharmacology with Clin-
ical Applications. Philadelphia: Lippincott, Williams &
Wilkins; 1997:456-461.
3. Schrodi J, Recio L, Fiorellini J, Howell H, Goodson M,
Karimbux NY. The effect of aspirin on the periodontal
parameter bleeding on probing. J Periodontol 2002;73:
871-876.
4. Love WD, Ramirez JM, Fultz RP. An oral hygiene mea-
surement system for possible research and clinical use.
J Public Health Dent 1975;35:227-230.
5. Greenstein G. The role of bleeding upon probing in the
diagnosis of periodontal disease. A literature review.
J Periodontol 1984;55:684-688.
6. Heasman PA, Seymor RA, Kelly PJ. The effect of sys-
temically administered urbiprofen as an adjunct to tooth
brushing on the resolution of experimental gingivitis.
J Clin Periodontol 1994;21:166-170.
7. Le H, Theilade E, Jensen SB. Experimental gingivitis
in man. J Periodontol 1965;36:177-187.
8. Page R, Schroeder HE. Pathogenesis of inammatory
periodontal disease. A summary of current work. Lab
Invest 1976;33:235-249.
9. Marshall-Day CD, Stephens RG, Quigley LF. Periodon-
tal disease: Prevalence and incidence. J Periodontol
1955;26:185-203.
10. Albandar JM, Kingman A. Gingival recession, gingival
bleeding, and dental calculus in adults 30 years of age
and older in the United States, 1988-1994. J Periodontol
1999;70:30-43.
11. Bode-Broger SM, Boger RH, Shubert M, Frolich JC.
Effects of very low dose and enteric coated acetylsali-
cyclic acid on prostacyclin and thromboxane formation
on bleeding time in healthy subjects. Eur J Clin Pharm
1998;54;707-714.
Correspondence: Dr. Nadeem Karimbux, Department of Oral
Medicine, Infection and Immunity, Harvard School of Dental
Medicine, 188 Longwood Ave., Boston, MA 02115.
Accepted for publication September 19, 2003.
3081.qxd 5/10/04 2:29 PM Page 684