. RESULTADOS: A média de idade das mulheres de perdas foi mais elevada do que no Controle (32,3±6,3 versus 26,5±6,4 anos; p<0,01). Houve predomínio do consumo de bebidas alcoólicas no grupo de perdas (36,9 versus 22,1%; p=0,01), assim como no estado civil (93,2 versus 66,7%; p<0,01), no qual elas eram casadas ou viviam em união estável, respectivamente. O índice de massa corpórea pré-gestacional foi superior no grupo de perdas (26,9 versus 23,5%; p<0,01). Nos antecedentes obstétricos, 103 mulheres com perdas gestacionais relataram 334 gestações. Destas, 56 tiveram dois ou mais abortos no primeiro trimestre e em 31 delas, duas ou mais gestações evoluíram para abortos tardios/prematuros extremos. CONCLUSÕES: Mulheres com perdas recorrentes incluídas neste estudo tiveram alguns fatores de risco identificados, tais como a idade mais avançada e o índice de massa corpórea superior. As observações são coerentes com as propostas mais recentes a respeito de perdas recorrentes que consideram a inclusão das perdas em várias idades gestacionais.Abstract PURPOSE: To describe the epidemiologic and obstetric characteristics of women with recurrent miscarriages. METHODS: A descriptive and analytical study whose inclusion criterion was every woman that was attended at the clinic for recurrent miscarriage (loss group), between January 2006 and December 2010. Patients that did not live in Salvador, Bahia, Brazil, and those who were not reached by telephone or whose number was not included in the medical record were not included. The Control Group consisted of 204 pregnant women seen at the low-risk prenatal care unit between May 2007 and April 2008. Women who did not accept to be interviewed and those with obstetric risk were excluded from the Control Group. The analyzed variables were: age, education, occupation, marital status, alcohol consumption, body mass index, obstetric history and the gestational age when the losses occurred. The SPSS 18.0 program was used for statistical analysis. Means and standard deviations of continuous variables were compared using the Student's t-test and the frequencies of the nominal variables were compared by the χ 2 test. RESULTS: The mean age of women in the loss group was higher than in the Control Group (32.3±6.3 versus 26.5±6.4 years old, p<0.01).