Firm evidence on the causes of recurrent miscarriage is scant. The true rate is probably artificially heightened by a reproductive compensation effect. The commonest direct cause is probably repeated sporadic chromosome abnormalities, which occur consecutively merely by chance. Congenital and acquired anatomical defects of the uterine fundus and cervix, parental chromosomal rearrangements, gene mutations, antibodies to cardiolipin, and luteal phase defects each make a small contribution. Other causes, such as polycystic ovaries and immune rejection, may play some part but the evidence is not clear. Psychological stress, subclinical infections, thyroid disorders, and diabetes mellitus are probably not relevant. Reassurance and clear statements about prognosis are important and psychological support must be offered throughout investigation and subsequent pregnancy. Much more rigorous scientific studies from which clearer conclusions can be drawn are vital for better understanding of this important clinical problem.

On epidemiological evidence, the definition of recurrent miscarriage should be three or more consecutive pregnancy losses. Data should be collected to 28 weeks' gestation but analysis up to 20-22 weeks' or 500 g fetal weight should also be possible. General practitioners and gynaecologists should do what they feel is suitable for couples whose history does not meet these criteria but a diagnosis of recurrent miscarriage should not be made. Women meeting the definition can be subdivided into primary and secondary groups, respectively consisting of those who have lost all previous pregnancies and those who have had one successful pregnancy followed by consecutive losses.