21 ly motivated users are likely to be rela- tively low; 7 on the other hand, motivation depends largely on women’s socioeco- nomic and other characteristics. Family planning managers need to know not only the failure rates of different methods, but also the determinants of failure, so they can give proper advice to prospective users, who will then be able to make in- formed choices. Few studies have examined contracep- tive failure in Bangladesh, and some of their findings diverge widely. Estimates of the 12-month failure rate for the pill have ranged from 1% to 26%; for inject- ables, from 2% to 3%; for the IUD, from 2% to 6%; and for the condom, from 1% to 12%. 8 In all of these studies, researchers as- sessed failure on the basis of users’ re- sponses to a question about their reasons for discontinuing a method; all but one of the studies 9 relied on retrospective data. Since high-quality data on contracep- tive failure are not available even for de- veloped countries, 10 and methodological pitfalls may bias studies of contraceptive failure, 11 the reliability of retrospective data on failure in developing countries presumably is limited. Furthermore, re- searchers examining Demographic and Health Survey data on contraceptive fail- ure in 15 developing countries have con- cluded that variations between and with- in countries may be attributable to errors in data; 12 variations in the definition and Radheshyam Bairagi is studies director, Population Stud- ies Centre, and Mizanur Rahman is demographer, MCH- FP Extension Project (rural), both at the International Cen- tre for Diarrhoeal Disease Research, Bangladesh (ICDDR,B). The research on which this article is based was supported by the World Health Organization and the ICDDR,B. The ICDDR,B is supported by countries and agencies that share its concern for health problems of developing countries. The authors are grateful for the comments of Michael A. Koenig and Sarah Salway, and for the computer and statistical assistance of Manoj Kumar Barua and Ahsan Habib. Contraceptive Failure in Matlab, Bangladesh By Radheshyam Bairagi and Mizanur Rahman C ontraceptive failure leads to about 20 million unintended pregnancies in developing countries each year. 1 It has serious consequences for the women involved, the children they bear and fam- ily planning programs. Pregnancies re- sulting from contraceptive failure often end in induced abortion, 2 a procedure that can be provided only clandestinely in many settings and that increases mater- nal morbidity and mortality in develop- ing countries. 3 Furthermore, children whose mother did not intend to become pregnant have well-documented devel- opmental deficits. 4 And family planning programs can have only limited effec- tiveness if contraceptive failure rates are high, regardless of contraceptive preva- lence and continuation rates among the populations they serve. 5 The risk of contraceptive failure may be related to characteristics of the method it- self, but is also associated with character- istics of users. 6 Failure rates among high- perception of failure, as well as response error, also can widen the disparity be- tween findings within a country, such as in the studies of Bangladesh. Therefore, understanding contraceptive failure re- quires analysis based on reliable data and a strong methodology. In the study on which we report in this article, we used a set of prospective, high- quality data to obtain better estimates than have thus far been available of contracep- tive failure among rural Bangladeshi women. We also explore socioeconomic, demographic and programmatic factors that may influence the likelihood of failure. Data and Methods Data Sources Our analysis is based on data from the area of Matlab where the International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR,B) has operated a comprehensive maternal and child health and family planning program since 1977. 13 The area’s population is about 100,000. In 1984, a knowledge, attitude and prac- tice survey was conducted in the inter- vention area. Since every individual in Matlab has been assigned an identification number, information about each woman who participated in the 1984 survey can be linked to other data for the area: the 1982 Socioeconomic Survey; 1977–1989 data from the ICDDR,B record-keeping system, a prospective data collection sys- tem; and the 1987 ICDDR,B Evaluation of Worker Performance. We selected ap- proximately 3,000 currently married women aged 15–49 for interview on the basis of multistage random sampling; after cleaning and linking different files, we had a sample of 2,856 women. The 1982 survey results on women’s ed- ucation and the size of respondents’ dwelling space were used as socioeco- nomic indicators. Size of dwelling space, a good measure of wealth and economic status in the area, was the best indicator among the socioeconomic variables in the survey 14 and was unlikely to have changed appreciably by the time of data collection for this study (1984–1989). Women’s age and number of living chil- dren were measured at the baseline. Contraceptive failure rates and the determinants of failure can be most accurately estimated using prospective data from an area served by a well-established maternal and child health and family planning program. In Matlab, Bangladesh, the cumulative probability of contraceptive fail- ure within one year of method acceptance was 1% for the injectable, 3% for the IUD and 15% for the pill and other temporary methods among 2,856 married women aged 15–49 during the pe- riod 1984–1989. Among women using no method, the 12-month cumulative probability of con- ception was 38%. For the pill, the likelihood of failure was consistently high during the first 12–18 months of use, after which it declined substantially; by contrast, the probability of an IUD failure increased, peaking at 24 months of use. The injectable maintained a low likelihood of failure re- gardless of duration of use, and no pattern was evident for other temporary methods. The qual- ity of community health workers’ performance was associated with the risk of failure of all tem- porary methods except the injectable; women’s background characteristics associated with failure varied by method. Calculations from failure rates suggest that 25% of births in Bangladesh may reflect contraceptive failure. (International Family Planning Perspectives, 22:21–25, 1996) Volume 22, Number 1, March 1996