Journal Club
Long-acting reversible contraception and repeat abortion: Rose et al
George A. Macones, MD, MSCE, Associate Editor
www. [Link]
The article below summarizes a roundtable discussion of a study published in this issue of the Journal in light of its methodology, relevance to practice, and implications for future research. Article discussed: Rose SB, Lawton BA. Impact of long-acting reversible contraception on return for repeat abortion. Am J Obstet Gynecol 2012;206:37.e1-6. The full discussion appears at [Link], pages e10-11.
DISCUSSION QUESTIONS
Why is this question important? What was the overall study design? How would you describe the study population? What analytic methods were used? What were the main study ndings? What were the studys strengths and limitations?
From the Department of Obstetrics and Gynecology, Washington University in St Louis School of Medicine, St Louis, MO: Moderator George A. Macones, MD, MSCE Mitchell and Elaine Yanow Professor and Head Discussants Anthony Odibo, MD Associate Professor Alison Cahill, MD Assistant Professor
0002-9378/free 2012 Mosby, Inc. All rights reserved. doi: 10.1016/[Link].2011.11.008
See related article, page 37
nintended pregnancies and repeat abortion are pressing issues around the world. Recent gures suggest that in the United States, over 1 million unintended pregnancies per year end in abortion. For the patient, abortion can be a source of emotional distress, and the surgical procedure is associated with small risks for bleeding and infection. Overall, unintended pregnancies and abortion represent a major public health issue. Therefore, attention has recently turned to the safety, efcacy, and patient acceptability of long-acting reversible contraceptive (LARC) methods. These methods include injectable, implantable, and intrauterine devices. This month, Journal Club participants discussed a prospective follow-up study from New Zealand that underscores the value of LARC methods in reducing the rate of repeat abortions. In the original 2008 study, which took place over a 10week period, women were offered free long-acting contraception after abortion either the levonorgestrel intrauterine system (LNG-IUS), depot medroxyprogesterone acetate (DMPA), or the copper multiload Cu375 IUD (Cu-IUD). Implantable devices were not offered. Patients also received intense education and patient counseling about these methods. They were then contacted by phone at 6 weeks and 6 months to assess continuation rates. At the 6-month mark, investigators reached 78% of the women and learned that 81% with a LNG-IUS and 74% with a Cu-IUD had retained them; 71% of DMPA users were continuing use. For the new study, the rate of repeat abortion was ascertained at 24 months by matching National Health Index numbers (a unique identier for
each subject) to hospital abortion clinic discharge data.
Time-to-event analyses Study investigators were interested in 2 specic questions. The rst was whether subjects had a repeat abortion within the 24-month period following enrollment in the original study. This question was answered easily by applying common analytical methods to their data; for example, the use of 2 2 tables in the calculation of relative risks and 95% condence intervals. Logistic regression can be performed to control for potential confounding factors. But the authors were also interested in whether the time to repeat abortion differed between LARC users and nonLARC users. A question such as this is best answered with a time-to-event analysis. We dont see these very much in obstetrics and gynecology, except in gynecologic oncology and perhaps in labor curve analysis. Still, it is important to understand the concepts. The unadjusted version of a time-to-event analysis involves the generation of a Kaplan-Meier curve. These survival curves generally graph the cumulative event rate on the y-axis, and time on the x-axis. A log-rank test is then used to test whether the curves (in this case, representing LARC and non-LARC users) differ. If one wishes to adjust for confounding factors in a time-to-event analysis, which is wise for observational studies, then a Cox proportional hazards model is produced. The output from such a model is a hazard ratio, which is the ratio of hazard rates. In this study, the hazard rate was derived as the probability of abortion in a time interval divided by the length of the time inter93
JANUARY 2012 American Journal of Obstetrics & Gynecology
Journal Club
val. A hazard ratio above 1 indicates a raised hazard for the LARC group, below 1 indicates a decreased hazard, and a value equal to 1 means that the hazard of abortion for the LARC group would be equal to that for the non-LARC group. A proportional hazards assumption is inherent in this test. Simply put, this means that the hazard rates of the treatments are proportional over time. If they are not proportional, then more sophisticated analyses are needed.
[Link]
What they learned The major nding was that 2 years after the initial study, the return-forabortion rate was signicantly lower for LARC users: 6.5% vs 14.5% for non-LARC users. Relative to those who chose to use an oral contraceptive after abortion, IUD users had a reduced risk of abortion (hazard ratio, 0.36; 95% condence interval, 0.17 0.77). Comparisons among women using LARCs indicated that those using an intrauterine method had signicantly lower return rates. While use of DMPA reduced the likelihood of return for a repeat abortion, the nding was not signicant. Journal Club members thought this was a very impressive study; one that should spur further research into prevention of unintended pregnancies. In fact, the data merits the attention of those who develop health policy in the United Statesand insurers. f
94
American Journal of Obstetrics & Gynecology JANUARY 2012