We defined contraceptive method failure as conception that occurred during a period of
method use. Survey data and research documentation recorded information about periods of
contraceptive use, including method start and stop dates. For IUD and implant users we also
recorded insertion and removal dates. If an IUD user experienced a known expulsion and
became pregnant, the pregnancy was attributed to “no method use.” If the woman was not
aware that the device had been expelled, the pregnancy was considered an IUD failure.
Pregnancies that occurred during periods of “no method use” were not considered a method
failure. Intended pregnancies, defined as a woman who stopped a method with the desire to
conceive, were excluded from this analysis. All pregnancies were reviewed to rule out preinsertion conception instead of a method failure.
Body mass index [BMI: (weight(kg)/height(m
2
))] was calculated using baseline
measurements. We categorized participants as normal weight, overweight, or obese using
NHANES criteria (14). Weight change during use was not considered in this analysis. The
significance level alpha was set at 0.05. Means, standard deviations, frequencies and
percentages were used to describe the baseline demographic characteristics of study
participants. For the comparison between implant and IUD users, Student’s t-test was used
for continuous variables, and χ
2
test was performed for categorical data. Kaplan-Meier
survival curves and logrank tests were constructed to compare implant and IUD
contraceptive failure rates among normal weight, overweight, and obese women. We
recorded segments of contraceptive method use for each participant. If a participant did not
get pregnant, we censored her at the time when she stopped the use of her contraceptive
method. In cases of lost-to-follow-up, we censored participants at the last survey date when
they were reached by telephone. We completed a post-hoc sample size calculation. A sample
Xu et al. Page 3
Obstet Gynecol. Author manuscript; available in PMC 2014 June 03.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
size of 1,168 participants of which 28% are classified as overweight and 35% as obese will
have greater than 80% power to detect a difference in contraceptive failure rates of 0.5% in
the normal weight group, to 0.8% of the overweight group, and 3% in the obese group (with
type I error rate of 0.05) using chi-square test of equal proportions. We used STATA 11
(StataCorp LP, College Station, TX) to perform the statistical analyses.