Can Condom Users Likely to Experience Condom Failure Be Identified?

Abstract

A study based on a convenience sample of 177 couples who each used 11 condoms found that 103 condoms (5.3%) broke before or during intercourse and 67 condoms (3.5%) slipped off during sex. Couples who had not used a condom in the past year were almost twice as likely to experience condom failure as were couples who had used at least one during that period (p < .001). Of the couples who had used a condom in the previous year, the failure rate among those who reported at least one condom break during that period was more than twice the failure rate among those who reported no breaks (p < .001). Among couples who had used condoms in the past year without breaking any, those who did not live with their partner and those who had a high school education or less were at increased risk of condom failure (adjusted odds ratios of 3.2 and 2.7, respectively).

Full text

Can Condom Users Likely to Experience Condom Failure Be Identified?
Author(s): Markus Steiner, Carla Piedrahita, Lucinda Glover, Carol Joanis
Source:
Family Planning Perspectives,
Vol. 25, No. 5 (Sep. - Oct., 1993), pp. 220-223+226
Published by: Guttmacher Institute
Stable URL: http://www.jstor.org/stable/2136075
Accessed: 30/11/2010 11:42
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=agi.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Guttmacher Institute is collaborating with JSTOR to digitize, preserve and extend access to Family Planning
Perspectives.
http://www.jstor.org

Can Condom
Users
Likely
to
Experience
Condom Failure Be
Identified?
By Markus
Steiner,
Carla
Piedrahita,
Lucinda
Glover
and
Carol
Joanis
A
study
based on
a convenience
sample of
177
couples
who each
used 11
condoms
found that
103
condoms
(5.3%1o)
broke
before or
during
intercourse
and 67
condoms
(3.5%1o) slipped off
during
sex.
Couples
who
had not
used a condom
in
the past
year
were almost
twice as likely to
experience condom
failure as were
couples who
had used at least
one
during that
period
(p<.001)
Of
the
couples
who had used
a condom
in
the
previous
year,
the failure
rate
among
those
who
reported
at least
one condom
break
during that
period was
more than
twice the
failure rate
among
those who
reported no
breaks
(p<.001).
Among couples
who had used
condoms
in
the
past
year
without
breaking any,
those who did not
live
with
their
partner
and
those who had a high
school education or
less were
at
increased
risk of condom
failure
(adjusted
odds ratios of
3.2
and
2.7,
respectively). (Family
Planning
Perspectives,
25:220-223 &
226,1993)
The rapid
spread of
the
human
im-
munodeficiency
virus
(HIV) and
other
sexually
transmissible dis-
eases
(STDs) during the
last decade
has
led
to
increased
research on
the
male con-
dom.
Because
condom
breakage
during
intercourse
or
withdrawal
seriously un-
dermines
this
protection,
numerous
stud-
ies
have
examined
the
frequency
of this
problem.'
Data
from these
studies
suggest
that
breakage rates
range from
less than
%/o
to
12% of
condoms used.2
Does this
wide
range in
condom
break-
age
rates
result from
chance,
from
differ-
ences
in
the
material
integrity
of
condoms
or
from
variations
among individual users
in
their
likelihood
of
breaking
condoms?
Anecdotal
evidence
suggests that a rela-
tively small
proportion of
condom
users
are responsible
for
a disproportionate
number
of
breaks.
If
this
is the
case, the
proportion
of
these
"condom
breakers"
in
a study
could
affect the
study's
overall
breakage
rate.
Because so little is known
about the
characteristics of such individ-
uals,
we
cannot
predict accurate condom
breakage
rates for
a given
user.
If
simple
methods of
identifying
condom
breakers
existed,
service
providers
could
maximize
the
impact of their
educational
interven-
tions by
targeting the
cohort
of users
who
experience the
majority
of
breaks.
The
analysis
presented
in
this
article ex-
amines
possible
explanations
for the wide
range of
breakage
rates
presented
in
the
literature
and
attempts
to
provide a
basis
for
identifying
condom users who
are at
increased
risk
of
breaking
condoms.
This
analysis
uses a
subset of
data
from
a
study
conducted
by
Family Health
Internation-
al
(FHI) to
assess
the value of
laboratory
tests
in
predicting
condom
breakage.3
In
the
original
study,
breakage
and
slip-
page data were collected for 20
different
lots of
condoms.
Four
lots
were
new
con-
doms
from
four U. S.
manufacturers;
the
remaining 16
were
recovered
from
ware-
houses
in
Tanzania,
the Dominican
Re-
public,
Egypt,
Mexico,
Kenya,
Jamaica and
Barbados.
All
recovered condoms
were
made
by
a
single U. S.
manufacturer and
were
distributed
by
the
Commodities Pro-
curement and
Support
Division
of
the
U. S.
Agency
for
International
Development.
Three
hundred
couples
were recruited
for the
study
from
professional
organiza-
tions
and
institutions in
the Research Tri-
angle
Park area of North
Carolina
(Raleigh,
Durham
and
Chapel
Hill) via fliers
and
word
of mouth.
Couples who
expressed
interest
in
participating were
sent
a
fact-
sheet
outlining
the
purpose
of the
study
and
a list of
criteria for
participation.
The
study protocol
required
participants
and
their
partners
to
be
in
a
monogamous het-
erosexual
relationship,
at least
18
years
old,
protected
against
pregnancy,
not
practic-
ing
behaviors
that
would
put
them at
risk
of STDs (including
HIV), and free from
known
sensitivities
to latex.
Further, each
participating couple
agreed
to use 20
con-
doms
during the
four-month
study peri-
od and
to
complete
a self-administered
questionnaire.
FHI's
Protection of Human
Subjects Committee
approved the
study
protocol and
informed
consent
forms.
The
20
study
condoms-one from
each
lot-were
equally
divided
into four
pack-
ets,
which
were mailed to
participating
couples
along
with
the
study question-
naire,
a
one-page form on
which
respon-
dents
answered a series of
questions
on
slippage
and
breakage
for
each condom
and
filled
in
an
identifying
code from the
condom
packaging.
The
participants
were
asked to
use the five
condoms
in
each
packet
in
random order.
This
scheme was
devised
to reduce
potential bias from
the
order
in
which
the condoms were
used.
When the
investigators
received
the com-
pleted
questionnaires,
they paid
the par-
ticipating
couples
for each
condom used.
One of the main
conclusions
of
the
orig-
inal
study of condom
breakage
was that
condoms
less than two
years
old are
of
es-
sentially
the same
quality
as new
condoms,
provided
they
have not
been
exposed to
extremely
harsh
conditions
during
storage.
Because
some of
the
16
condom
lots re-
covered
from
overseas
warehouses
had
been
manufactured more than
two
years
earlier and
some
might
have been
stored
under
adverse
conditions,
we considered
both
condom
age
and
laboratory results
when
deciding
whether
to use data for
a
given
condom lot in
the
current
analysis.
Seven
of
those lots and all
four lots
of
new
condoms
passed
the
bursting
and
pressure
standards set
by
the
International
Organi-
zation for
Standardization
(ISO)4
and
were
less than
two
years
old. Data
on
condoms
from the
remaining
nine
lots were
not used
in
this
analysis because the condoms
failed
the ISO standards and
were not
consid-
ered of
acceptable
quality
for distribution.
Although
most
research
has addressed
only
condom
breakage,
we
also calculate
rates
of
condom
slippage and
overall fail-
ure
(slippage
and
breakage
combined).
Overall failure
should not be confused
with
contraceptive
failure,
which
by
def-
inition
is
the
occurrence of a
pregnancy.
Markus Steiner is
a
senior
research
analyst, Carla
Piedrahi-
ta is
a
consultant,
Lucinda Glover
is
a
biostatistician and
Carol
Joanis
is
associate director of
marketing
research,
Family Health
International,
Research
Triangle
Park,
N.
C.
The
study on
which this
article is based
was supported
by funds
from
the United
States
Agency for
Internation-
al
Development.
The views
expressed
in this
article are
those
of
the authors and do
not
necessarily reflect
those
of
the
funding
agency.
Theresa
Burton, Kathy
Hinson,
Kazu Martinez and
Pat
Stewart provided
administrative
assistance and
helped with
data
management.
Rosalie Do-
minik,
Barbara
Janowitz,
Nancy
Williamson,
Douglas
Nichols
and
Robin
Foldesy provided
insightful
comments
on earlier
versions
of
the
manuscript.
220 Family
Planning
Perspectives

Data and
Methods
Sample
Characteristics
Of the 262
couples
who
completed the
original study,
260 used at least one
of
the
11
condoms that
met ISO standards and
were less than
two years old. This analy-
sis is based on 177
couples
who used all
11
of those
condoms. This assures that
rates
will
not be
influenced by the use of
different
combinations of condom lots
by
the participants.
However, to make sure
that we did not
introduce a selection bias
by eliminating
the
experience
of 83 cou-
ples who used
some
but
not
all
of the
11
condoms,
we also report
selected rates
based on
the
larger group
of 260
couples.
The first
packet of condoms sent to the
participants
included a
questionnaire
that
asked for social and
demographic
data
and
information on
previous experience
with
condoms. Female and male
participants
in
our
sample
were similar
in
age,
education
and
ethnic
background.
The median
age
of the
participants
was about
30,
and the
median education
was approximately 15
years.
Caucasians
made
up
the
great
ma-
jority
of the
sample (84%),
followed
by
blacks
(10%),
Asians
(2%), Hispanics (2%)
and those of
other ethnic
backgrounds
(2%).
Most of the
couples (84%)
were either
married or
living together.
About
three-fourths of
participants
re-
ported having
used 10
or more condoms
during
their lifetime
(women, 75%; men,
83%)
and a similar
proportion
had
used
at
least one condom
during
the
year
be-
fore the
study
(women, 80%; men, 80%).
Thirty-nine percent of the women and
41%
of the men
reported
using
10 or more con-
doms
during
that
period. Among
the
par-
ticipants
with
condom
experience
in
the
previous year, 75% of the women and
74%
of the men
reported
no condom
breakage
during
that
time. Three
percent
of men
and
women
(four
couples) reported
five
or
more breaks
in
the
previous year.
Definition of
Slippage
and
Breakage
The self-administered
questionnaire
in-
cluded a series of
specific questions
about
each condom used. The couples were
asked:
1)
if
the condom
slipped
off
during
sex; 2)
if
the condom
broke;
and
3)
if
the
condom did
break,
whether the
break oc-
curred while the package was being
opened,
while the condom was
being
un-
rolled onto
the
penis, during sex,
during
withdrawal, while
the condom was
being
taken
off,
or at an
unknown time.
To avoid double counting and
to calcu-
late accurate
rates for breakage,
slippage
and overall
failure, we used a
hierarchical
convention
similar to one developed
by
James
Trussell and colleagues.5 Unlike
Trussell,
however, we included
in the
breakage rate
calculations condoms
that
broke while the
user was opening the
pack-
age or putting
on the condom. As
in
the
Trussell
convention, those condoms
were
subtracted from
the denominator used to
calculate
the
slippage rate; thus, that rate
is based only on
condoms that were
used
during intercourse. Because the
question
on slippage asked if a condom had
"slipped
off
during sex,"
some
respondents
may have
reported condoms that
slipped
down
(but not
off)
the
penis
or those that
slipped
off
during
withdrawal.
Therefore,
the condom
slippage
rate
presented
in
this
paper may
overstate the
proportion
of con-
doms that
slipped
off
during
intercourse.
We counted condoms that both broke
and
slipped
off as breaks
only (not
as
slips),
because we believe that
in
most
of
these
cases,
the condom
slipped
off
because
it
had broken. We calculated
overall failure
by adding
the number
of condoms that
broke to the number of condoms that
slipped
off
during
intercourse and divid-
ing
the sum
by
the total
number of con-
doms used
by
the
participants.
We divided condom
breaks into two
categories,
based on whether
they
had
clinical
implications.
We classified breaks
that
occurred before intercourse as non-
clinical because
they
would not
expose
the
couple
to the risks of
pregnancy
and STD
transmission. We classified all other breaks
as clinical. We included breaks noticed
during
withdrawal
in
this
category,
be-
cause
many
of these
might
have occurred
during
intercourse.
When
calculating
condom
breakage
rates, some researchers exclude nonclin-
ical breaks because such breaks do not
put
the
couple
at risk.
Although
we
provide
rates for clinical
breakage,
we
based
our
analyses
on all
breakage
because we be-
lieve that both
types
of
breakage
are
im-
portant.
Nonclinical
breakage,
whether
caused
by
misuse or
by
defects
in
the
con-
doms
themselves,
could lead users to dis-
trust condoms. This
in
turn could lead to
increased
user failure
through
nonuse or
inconsistent
use. User failure is as
impor-
tant
as, if not more
important
than,
method failure
in
its
impact
on rates of
pregnancy
and
STD infection
among
con-
dom
users.
Moreover,
if
a
couple
has ac-
cess to
only
one condom and that
condom
breaks before
intercourse,
the
couple
may
engage
in
unprotected
intercourse.
Data Analysis
First, we calculated rates
of
breakage,
slip-
page and overall failure among
the
cou-
ples in
our sample. Next,
we compared
the
expected
binomial distribution and the ob-
served
distribution of
condom failures
in
the
entire sample to
determine
whether
instances of repeated
condom failure
in
this
study population were
independent
events.
(Similarity
in
the two distributions
would
suggest that
condom failure is
the
result of
chance and not
of user behaviors
or
characteristics.)
We used a
chi-square
goodness-of-fit
test to assess whether dif-
ferences
in
the
expected
and observed
fre-
quencies were
statistically significant.
We
then
compared
rates of
breakage,
slippage
and
overall failure
among
cou-
ples
who had used
condoms
in
the previ-
ous
year
with
rates
among
those who
had
not. We
also compared
those rates
among
couples who had experienced
condom
breakage
in
the
year
before the
study
with
rates
among couples
who had not
expe-
rienced
breakage
during
that
period.
These variables were based on the
male
partner's
responses
on the
background
questionnaire.
We used a two-tailed Fish-
er's exact test for both
comparisons.
Finally,
we used
logistic
regression
to
identify social and
demographic
risk fac-
tors for condom failure. The
regression
was
modeled
only
on
couples
who had
used
condoms
during
the
previous
year
and
had not
reported any
breakage,
because
those
who
had previously experienced
condom
failure were
already
identified as
being
at
increased risk. The
logistic
re-
gression
model assessed the
usefulness of
four
sociodemographic
characteristics of
the man-age, race, education and
whether he lived
with
his
partner*
in
pre-
dicting
condom failure.
In
two
instances,
race was missing
from the
man's
ques-
tionnaire,
so we used the woman's re-
sponse.
Collinearity
and interaction of
variables were also evaluated.
Results
Overall Failure Rates
The 177
couples
included
in
our
analysis
used
1,947
condoms. Of these
condoms,
5.3% broke
(1.6%
nonclinical;
3.7%
clini-
cal)
and 3.5%
slipped
off
during sex,
re-
sulting
in
a clinical failure rate of
7/2%
and
an overall
failure rate of
8.7%.
The
larger sample
of 260
couples
who
had used at
least one of the
11
condoms
that met our criteria used a total of
2,519
condoms. Of these
condoms,
6.2%
broke
(2.1/o
nonclinical;
4.2%
clinical)
and
4.1/o
slipped
off
during sex,
for a clinical fail-
*The four variables were
dichotomized for this analysis:
cohabitation
status,
living
with
partner
vs. not
living
with
partner; education, ?12
years vs. >12 years; age, <30 years
vs.
?30 years; race,
Caucasian vs. non-Caucasian.
Volume 25,
Number 5,
September/October 1993 221

Condom
Users
Likely to
Experience
Failure
Table 1.
Binomial
probability of
failure,
and ex-
pected and
observed
number of
couples ex-
periencing
condom
failure, by
number of
fail-
ures per
couple
Failures Binomial No.
of
couples
per
couple probabilityt Expected Observed
0-11 1.0000 177 177
0 .3660 65 110
1 .3852 68 26
2 .1843 33 17
3 .0529 9 8
4 .0101 2 6
5 .0014 0 4
6 .0001 0 4
7 .0000 0 0
8 .0000 0 0
9 .0000 0 2
10 .0000 0 0
11 .0000 0 0
X2=
162.96
df=4
p<.O01
tThe
expected
probability
of one
couple
experiencing
the
given
number of failed
condoms
out of 11
condoms used.
Note:
Cells
4-11 were
combined
to eliminate
any
frequency
of
zero before
conducting
the
goodness-of-fit
x2
test.
ure
rate of
8.3% and an
overall
failure
rate
of
10.2%.
We
also
compared the
failure
rate
of
the
four
new
condom lots
with
the failure
rate
of
the
seven
condom
lots
that
had
been
stored
overseas.
The 177
couples
in our
study
reported
breaking 3.4%
of
the
new
condoms (1.1%/o
nonclinical;
2.3%
clinical)
and
6.4%
of
the
condoms
that
had
been
stored overseas (1.9% nonclinical; 4.5%
clinical);
the
difference
in
overall
break-
age
was significant
(p=.004).
They
said
that 4.7%
of
the new
condoms
had
slipped
off
during
sex,
compared with
2.8%
of
the
condoms
that
had
been
stored
overseas
(p=.038).
Thus, the
overall
failure rate
was
8.1% for the
new condoms and 9.%/o
for
those
that had
been
stored
overseas;
these
rates
were not
significantly different.
The data
for
the
260
couples in
the
larg-
er
sample were similar.
These
couples re-
ported
breaking 4.4% of
the new
condoms
(1.6%
nonclinical; 2.8%
clinical),
compared
with
7.3%
of
the
condoms that
had been
stored overseas (2.2% nonclinical; 5.1%
clinical).
They
said
that 5.2%
of
the new
condoms had
slipped
off
during
sex,
com-
pared
with
3.4%
of
the
condoms stored
overseas.
Thus,
the overall
failure rate
was
9.5% for the
new condoms and 10.6% for
the
condoms
stored
overseas.
Distribution
of
Failure
If
condom
failure
is
truly
independent of
the
user,
the
outcome of
each
subject's
first
condom use
is
independent
of
the
subject's
subsequent
condom use.
The
pattern
of
failures
among
subjects
would
follow
a bi-
nomial
distribution. This
distribution mod-
els
the
number
of
failures
per
subject
where
the
outcome
is
binary
(i.e.,
failure
vs.
non-
failure). We
assumed
that
the
true
condom
failure
rate
was the
observed
rate
of
8.7/o.
Using the
binomial
probability
function,
we
calculated
the
numbers of
subjects
who
would
be
expected
to
experience
each
pos-
sible number
of
condom
failures
(0-11) and
used a
goodness-of-fit
chi-square
statistic
to
compare
the expected
number
with the
actual
number of
subjects
who
experienced
each
of the
possible
numbers
of
condom
failures
(Table
1).
The
expected and
ob-
served
distribution
of
condom
failures
dif-
fered
significantly
(p<.001);
thus,
we
con-
cluded that
the condom
failures
occurring
in
this
study
population
were
not
random
events. In our
sample,
more
couples
expe-
rienced
no
condom
failures,
fewer
couples
experienced
one,
two or
three
failures and
more
couples
experienced four or
more fail-
ures
than
would be
expected
if
condom
failure were
randomly
distributed.
Overall,
170
condoms
failed, and the
small
proportion
of
couples
who
experi-
enced
multiple
condom
failure
account-
ed
for
a
larger
proportion
of
condom fail-
ures
than
expected.
According
to the
binomial
probability
function, two
couples
(1/) would
be
expected to
experience
four
or
more condom
failures
each and
to be re-
sponsible
for
5% of all
condom
failures.
However,
in
this
study
population,
16 cou-
ples
(9%)
experienced
four or
more con-
dom
failures each
(for
a
total
of
86
failures)
and
were
responsible
for
50%
of all con-
dom
failures.
Failure
by
Condom
Use
History
When
we divided
the
couples
in
our
sample
into two
groups
accord-
ing
to
whether
they
re-
ported
using
condoms in
the
year before the
study,
the
condom
breakage,
slippage
and
overall fail-
ure
rates were
signifi-
cantly
higher
for the
group with
no
condom
experience
in
the previ-
ous
year
(Table
2).
The
overall
failure
rate
for
the
group
that had
not
used
condoms
during
the
pre-
vious
year
was nearly
twice
that of
the
group
that
had (13.9%
vs.
75%,
p<.OOl).
We further
divided
the group reporting
condom use in
the
pre-
vious year
by whether
they
reported
breaking
one
or
more
condoms
during
that
year.
The
group
with
reported con-
dom
breakage
in
the
previous
year
had
breakage
and
slippage
rates
more than
twice
as high
as the
group with
no re-
ported
condom
breakage
during
that pe-
riod
(overall
failure
rate,
13.1% vs.
5.6%,
p<.001). These
two
simple
screening
ques-
tions
(whether a
couple had used
condoms
in
the
previous
year
and,
if
so,
whether
they had
experienced a
condom
break dur-
ing
that
period)
identified 70
couples
(40%
of the
sample) who were
at
increased risk
of
condom
failure.
Logistic
Regression
Analysis
We
then
compared
the
observed
and ex-
pected
condom
failures
using
data from
the 107
couples
who
reported
having
used
condoms
in
the
previous
year
without
breaking
any.
Once
again, the
goodness-
of-fit
chi-square
was
significant
(X2=35.81,
p<.0001),
strongly
suggesting that
condom
failures
were
not
independent
events
among
these
couples.
We
conducted
fur-
ther
analyses
to
identify factors
that
dif-
ferentiate risks of
failure.
Table
3 shows the
condom
failure rates
for
this
subgroup
within
categories of
age,
education,
race and
cohabitation
status.
These
results
must be interpreted cau-
tiously
because of the
small
sample
size in
some
of the
cells. The
lowest
failure
rates
were
experienced
by
couples
aged
30 or
older
who had
more than 12
years
of ed-
ucation
(2.5%),
couples
with
more
than
12
Table
2.
Calculation of
condom
failure
rates,
by
type of
failure, ac-
cording to
condom use
experience in
the
year
before
the
study.
Type
of Total Used condom Broke
condom
failure in
previous
year in
previous
year
Yes No Yes No
(n=1
43) (n=34) (n=36) (n=1
07)
Total
failure
No.
used* 1,947 1,573 374 396 1,177
No.
failed 170 118 52 52 66
Rate 8.7 7.5 13.9 13.1 5.6
95% C.1. 7.5-10.1 6.3-8.9 10.6-17.9 10.0-16.9 4.4-7.1
P-value .0002 .0001
Breakage
No.
used* 1,947 1,573 374 396 1,177
No.
brokent 103 71 32 31 40
Rate 5.3 4.5 8.6 7.8 3.4
95% C.I. 4.4-6.4 3.6-5.7 6.0-12.0 5.4-11.0 2.5-4.6
P-value .0029 .0006
Slippage
No.
usedt 1,915 1,553 362 387 1,166
No.
slipped? 67 47 20 21 26
Rate 3.5 3.0 5.5 5.4 2.2
95% C.1. 2.7-4.4 2.2-4.0 3.5-8.5 3.4-8.3 1.5-3.3
P-value .0256 .0030
*AII
condoms
used, including
those
that broke
before
intercourse.
tAll
condoms that
broke,
before,
during
or after
intercourse.
4All
condoms
that did not
break before
they were
put on.
?AII
unbroken
condoms that
slipped.
Notes: Man's
response
was used
to
categorize
couple's
condom use experience.
P-value
determined
by two-tailed Fisher's
exact test.
222 Family
Planning
Perspectives

Table
3. Rates of condom failure
among 107
couples who had not
broken a condom in
the
year
before the
study,
by
selected char-
acteristics of
male partner
Characteristic No. of Failure P-valuet
couples rate*
AGE-GROUP
19-29
10-12 yrs. of education 6 7.6 1.000
13-17 yrs. of education 42 6.9 .091
Not
living
with
partner 14 11.7 .169
Living
with
partner 34 5.1 .002
30-65
10-12 yrs.
of
education 22 7.9 .702
13-17 yrs. of
education 37 2.5 <.001
Not
living
with
partner 1 9.1 1.000
Living with
partner 58 4.4 <.001
EDUCATION
10-12 yrs.
Not
living
with
partner 0 na na
Living
with
partner 28 7.8 .558
Non-Caucasian 3 15.2 .202
Caucasian 25 6.9 .272
13-17 yrs.
Not
living
with
partner 15 11.5 .181
Living
with
partner 64 3.3 <.001
Non-Caucasian 10 4.6 .112
Caucasian 69 4.9 <.001
RACE
Non-Caucasian
Aged
19-29 yrs. 9 4.0 .094
Aged
30-65 yrs. 4 13.6 .269
Not
living
with
partner 2 9.1 1.000
Living
with
partner 11 6.6 .496
Caucasian
Aged
19-29 yrs. 39 7.7 .391
Aged
30-65 yrs. 55 3.8 <.001
Not
living
with
partner 13 11.9 .157
Living
with
partner 81 4.4 <.001
*Each
couple used 11
condoms.
tTwo-tailed
Fisher's exact test used to
determine whether fail-
ure
rate
of
subgroup
differs
significantly
from that of
rest of
population.
Note:
na=not
applicable.
years of
education who were
living to-
gether
(3.3%)
and Caucasian
couples aged
30 or
older
(3.8%).
Couples
with
12
years
or less of
education who were non-Cau-
casian
had
the
highest
failure rate
(15.2%),
but
this
rate is
based on
only three
couples.
Because of
this wide
range of
failure
rates within
subgroups of
our
sample, we
performed
further
analyses
of the data.
We
used
logistic
regression
to assess
the joint
contribution of
age,
education,
race
and
co-
habitation
status to
condom
failure. We
considered all
two-way interaction terms
of the four
predictors.
One term
(cohabi-
tation
status by
education)
had
no
subjects
in a
cell and
was not
included
in
the model.
Another
term
(cohabitation status
by age)
was
removed to
control
collinearity.
Because so few
couples
(0 to 4)
were rep-
resented
in
many of the
strata
defined by
cross-classification of
predictors,
estimates
of interaction
effects could be
extremely
variable.
This
instability
was confirmed
through
use of
regression
diagnostic
pro-
cedures.
Consequently,
we fitted a
model
without
interaction terms.
The rates
in
Table 3
suggest that
the
difference in
risk
of condom
failure at-
tributable
to age may
not be the
same for
both
categories
of race and
that a
model with an in-
teraction term
may
pro-
vide a
better
fit.
Of
the four
factors, co-
habitation
status
and
education were the
most
helpful in identifying
couples
at
risk
of con-
dom failure
(Table
4).
Not living
with
one's
partner
had an
adjusted
odds ratio of
3.2,
while
having
12
years of edu-
cation or less
had
an ad-
justed odds
ratio of
2.7
In
this
population,
race
and
age
were not
signif-
icant
predictors of
con-
dom failure.
Discussion
Because it
is unlikely
that a vaccine against
HIV will
be available
soon,6
the
public
health
community
must
rely
heavily on the
condom
to slow the
spread
of
HIV
among sexually
ac-
tive individuals.
How-
ever,
recent
prospective
condom studies have
found
breakage
rates
varying from less
than
1%
to about
12%.7
Our data
are from a study
population
that
was
protected from
pregnancy by a
method other than
the
condom. Because
the
couples
in
our
sam-
ple were not regular
condom
users, their be-
haviors
and
character-
istics
might differ
from
those of
typical
users.
They
probably
had
less
experience
using
con-
doms and might
not
have been
as
careful
in
their use of
condoms
during the
study
period
because they were
al-
ready using another
form
of
contraception.
However, we believe
that
some
of the
results
presented in
this
article
may apply to typical
users as well.
Our data
support
the
theory
that a small
group of
condom users
is
responsible for
a
disproportionate
number of condom
fail-
ures.
Couples
with
no
condom
experience
in
the year
before the
study and
couples
who had
experienced
condom
breakage
during that
period had
relatively high
rates
of
condom failure. This
suggests that such
couples
have
characteristics or
behave in
ways that increase their risk
of failure. How
useful this
information
would be
in
screen-
ing typical condom users
is unknown.
Couples who
had
used condoms
in
the
year
before the
study
without
experienc-
ing condom
breakage
had a failure
rate of
5.6%.
In
light of the AIDS epidemic
and
the
increasing
prevalence of
STDs, this
failure rate is still
unacceptably high.
Co-
habitation
status and
education were
help-
ful in
identifying couples
in
this
subgroup
who were at
increased risk of condom fail-
ure.
Couples
who were not
living
togeth-
er
had
significantly
higher
failure
rates
than
their
cohabiting
counterparts.
If
the
male
partner
had a
high
school
education
or
less,
the
couples experienced
signifi-
cantly higher
failure rates
than if
he had
more education.
Little
quantitative
research has been
conducted on the
behaviors
that
adversely
affect condom
failure rates.
Qualitative
data
collected
in
the
past
few
years
point
to four main
categories
of
behavior
that
may
contribute to condom
failure: incor-
rect methods
of putting
on
condoms, use
of oil-based
lubricants,
reuse of
condoms,
and
duration
and
intensity
of coitus.8
The
higher condom
failure rates
among
couples
in
less
stable
relationships
and
among
less educated
couples may
be
caused
by
some of these
behaviors. Cou-
(Continued
on
page 226)
Table
4. Adjusted
odds ratios
(and 95%
confidence
intervals)
among
107 couples who
had not
broken a
condom
in
the year be-
fore the
study, by selected
characteristics of the male
partner
Characteristic No. of
condoms Adjusted Adjusted
Used Failed failure odds
ratet ratiot
Cohabitation
Not living with
partner 165 19 8.3 3.2 (1.6-6.5)
Living with
partner 1,012 47 2.6
Education
<12
years 308 24 7.0 2.7 (1.5-5.0)
>12 years 869 42 2.6
Race
Non-Caucasian 528 37 3.3 1.2
(0.6-2.5)
Caucasian 649 29 2.6
Age
<30 143 4 3.8 1.4
(0.8-2.6)
?30 1,034 56 2.6
tThe failure
rate of the
characteristic of
interest when the
other three variables
in the
model
are held
at
"not-at-risk" level.
tThe odds ratio of
the variable of interest
when the other three
variables are included in
the model.
Volume
25, Number
5,
September/October
1993 223

Comparison
of Implant
Adopters and
Pill Users
traception for
women
in
general.
Second, the
pattern
of
implant
adoption
suggests
that
this
method
is
being
chosen
more
often
by women
who are
spacing
their
children or discontinuing child-
bearing
than
by
women who are
post-
poning
childbearing.
This
finding is
sup-
ported
by
the
regression
model,
in
which
having two or
more
children
increased
the
odds
of
adopting
the
implant
and
younger
age
was
not
associated
with
choosing the
implant. If
women
were using
the
implant
to
postpone
childbearing,
we would
ex-
pect
having children to
decrease the odds
of
choosing
the
implant.
An
important
question raised
here is
why, when
the effect of
cost
is
removed,
the
implant
is
not
being adopted
more
fre-
quently by
adolescents
to
postpone child-
bearing. (In the
clinic in
this
study, pro-
viders view
the implant
as
an
appropriate,
but
not
necessarily
preferred, method for
adolescents;
adolescents are
generally not
dissuaded
from
choosing it.)
While
the
availability
of
long-acting
methods
might
be
expected to
reduce
the rate
of
adoles-
cent
pregnancies,
such an
outcome
will
depend
on
adolescents'
early
access to
long-term
methods and a
willingness
to
adopt
them.
A
study
of
adolescents'
per-
ception-s
of the
implant's
attributes,
of
their
understanding
of
the
five-year
time
frame
involved,
and
of
the
desirability
of
early
childbearing
is
needed for this
issue
to
be
examined.
Third,
having
had
an
abortion was
not
associated with
choosing
the
implant in
the
multivariate
analysis,
suggesting
that
the
desire to
avoid
additional
unwanted
pregnancies is not
a major
factor
moti-
vating the
choice
of
the
implant over
the
pill
in
this
clinic
sample.
Further
analysis
will
be needed
to assess whether a re-
duction
in
the rate
of
pregnancy
termina-
tions occurs as a
result of
the
availability
of
long-term hormonal
contraception.
Future
research
should
also follow
wom-
en
who have
adopted
the
implant,
com-
pared
with
women
using other
hormonal
methods,
to ascertain
whether some of
the
variables
identified
here,
including
method
of
payment,
affect how
women
cope
with
method
side effects
and
make
decisions
to
continue
or
discontinue use. In this
group of
early implant
users, nine
women
(or
7% of those
receiving
the
implant)
had
already
had
their
implants
removed with-
in 9-14
months of
insertion.
The
average
duration of use
among
these women
who
discontinued was 38
weeks.
The
reasons
given
for removal
varied
widely,
although
bleeding
problems were mentioned
most
frequently.
All of
the removals were
among
Med-
icaid
patients, for
whom the
cost of
re-
moval was
fully covered. This situation
raises
the question of
whether
Medicaid
reimbursement
might
encourage
early re-
movals
or,
conversely,
whether
having
to
pay
for
removal
out-of-pocket
might
be a
disincentive to
early
removal.
Future re-
search
should
examine how
women per-
ceive
their
options
to
continue
or
discon-
tinue use under different
payment
mechanisms.
An
important
social
policy
question is whether
Medicaid reim-
bursement for
the
implant
would
increase
woman-years
of
protection from unin-
tended
pregnancy,
compared
with
other
effective
contraceptive
methods.
References
1. W.
D. Mosher and W. F.
Pratt,
"Contraceptive
Use in
the
United
States,
1973-88,"
Advance
Datafrom Vital
and
Health
Statistics, No. 182,1990.
2. I.
Sivin,
"Norplant
Clinical
Trials," in S.
E.
Samuels and
M.
D. Smith,
eds.,
Norplant and
Poor
Women,
Henry J.
Kaiser
Family
Foundation,
Menlo
Park;
Calif., 1992.
3. Norplant
Levonorgestrel
Implants:
A Summary of
Scien-
tific
Data,
The
Population
Council, New York,
1990.
4. P. D. Darney et al.,
"Acceptance
and Perceptions of
NORPLANTO Among Users in
San Francisco,
USA,"
Studies
in
Family
Planning,
21:152-160,
1990; and D.
Shoupe et
al.,"The
Significance of
Bleeding
Patterns
in
Norplant
Implant
Users," Obstetrics
and Gynecology,
77:256-260,
1991.
5. D. B.
Pettiti, "Critical
Issues
in
the Evaluation
of
Norplant
in
the United
States,"
paper
presented at
the
Conference
on
Dimensions of New
Contraceptive
Technologies: Nor-
plant and
Low-Income Women,
Henry J.
Kaiser
Family
Foundation,
Menlo
Park, Calif.,
Nov.
20-22,1991.
6. M. L. Frank et
al.,
"Characteristics
and
Attitudes of
Early
Contraceptive
Implant
Acceptors
in
Texas,"
Fam-
ily
Planning
Perspectives,
24:208-213, 1992.
7. J.
D. Forrest
and L.
Kaeser,
"Questions
of
Balance:
Is-
sues
Emerging
from the
Introduction of the
Hormonal
Implant,"
Family
Planning
Perspectives,
25:127-132,1993.
8. M.
L.
Frank
et
al.,
1992, op.
cit.
(see
reference
6).
Condom
Users...
(Continuedfrom
page 223)
ples
who live
together
may
engage in
less
vigorous or
lengthy
intercourse
and
thus
have lower
failure rates.
Couples
likely
to
engage
in
vigorous
or
lengthy
intercourse
could be
supplied
with
extra-strong
con-
doms
and
additional
lubricant.
Less edu-
cated
couples
may
have
difficulty
under-
standing
instructions on
how to
put on
condoms
and
understanding
the
impor-
tance of
not
using
oil-based
lubricants.
In-
structions should
be written at
a
reading
level
comprehensible
to
condom users
of
all
educational
levels.
Although this
study found
relatively
high
slippage,
breakage
and
overall
fail-
ure
rates, our
analysis of the
distribution
of failures indicates that
condoms
are an
appropriate
method for the
prevention
of
pregnancy
and
STDs
(including AIDS) for
the
majority of users. This article
presents
preliminary
methods
by
which users at
high
risk
of failure can
be identified.
Fur-
ther
research is needed
to determine
the
behaviors that
place
these
individuals
at
increased risk. If
these
adverse
behaviors
can be
identified,
it
may
be
possible to de-
velop instructional
materials that
alter
these behaviors and
thus
increase the
de-
gree of
protection
condoms offer
all
users
against
pregnancy
and
STDs.
References
1. A.
Albert, R.
A.
Hatcher and W.
Graves,
"Condom Use
and
Breakage
Among Women
in a Municipal Hospital
Family
Planning Clinic,"
Contraception,
43:167-176,1991;
R.
A. Hatcher et
al.,
Contraceptive
Technology, 1990-1992,
15th ed.,
hrvington Publishers,
New York,
1990; "Can You
Rely on Condoms?" Consumer Reports,
March 1989,
pp.134-142;
M. Free, E. Skiens
and M. Morrow, "Rela-
tionship Between
Condom
Strength and
Failure During
Use,"
Contraception,
22:31-37
1980;
M. Free
et
al.,
"An
As-
sessment
of Burst
Strength
Distribution
Data for
Moni-
toring Quality
of Condom
Stocks
in
Developing Coun-
tries,"
Contraception,
33:285-299,1986;
P.
Gotzsche and
M.
Hording, "Condoms to
Prevent
HIV
Transmission Do
Not
Imply
Truly
Safe
Sex,"
Scandinavian
Journal
of Infec-
tious Diseases,
20:233-234,1988; M. Leeper
and M. Con-
rardy,
"Preliminary Evaluation
of REALITY, A
Condom
for Women to
Wear,"
Advances in
Contraception, 5:229-235,
1989;
L.
Liskin,
C. Wharton and R.
Blackburn,
"Condoms:
Now More
than
Ever," Population
Reports,
Series
H,
No.
8,
Sept. 1990; J.
Richters
et
al.,
"Low Condom
Breakage
Rate in
Commercial
Sex,"
Lancet,
11:1487-1488,
1988; P.
Russell-Brown et al.,
"Comparison of
Condom
Breakage
During
Human Use with
Performance
in
Laboratory
Testing,"
Contraception,
45:429-437 1992;
J.
Trussell, D. L.
Warner and
R. A.
Hatcher,
"Condom
Performance Dur-
ing
Vaginal
Intercourse:
Comparison of
Trojan-Enz' and
Tactylon'T
Condoms,"
Contraception,
45:11-19,1992; and
,Condom Slippage and
Breakage
Rates,"
Family
Planning
Perspectives,
24:20-23,
1992.
2. L. Liskin,
C.
Wharton and
R.
Blackburn, 1990,
op. cit.
(see
reference 1).
3. M.
Steiner
et al.,
"Study to
Determine the
Correlation
Between
Condom
Breakage
in
Human
Use and
Labora-
tory
Test
Results,"
Contraception,
46:279-288, 1992.
4. International
Organization
for
Standardization,
"Rub-
ber
Condoms-Part 6:
Determination
of
Bursting Vol-
ume and
Pressure,"
Geneva, 1990.
5. J.
Trussell,
D. L. Warner and R.
A.
Hatcher,
"Condom
Slippage and
Breakage
Rates,"
Family
Planning
Perspec-
tives,
24:20-23, 1992.
6. R. M.
Anderson
and R.
M.
May,
"Understanding
the
AIDS Pandemic,"
Scientific
American, Vol.
266, No.
5,1992,
pp.
58-66.
7. L.
Liskin,
C. Wharton
and R.
Blackburn, 1990,
op.
cit.
(see reference
1).
8. W. R.
Finger
et
al., "How Human Use
Affects Condom
Breakage,"
Network, Vol.
12,
No.
3,1991,
pp.
10-13.
226 Family
Planning
Perspectives