Testicular cancer is the most common malignancy in young men in many Western countries, and is increasing in incidence. Despite epidemiological investigations over several decades its aetiology remains largely unknown. The peak of incidence of the tumour at around age 30 years and other features of its epidemiology suggest that aetiological factors operate early in life, either prenatally or in childhood and adolescence (Swerdlow, 1997). Investigation of exposures at such young ages is hampered, however, by the difficulty of gaining reliable information about them many years later by questions to adult patients. Case-control studies of malignancy in twins, comparing risk factors in affected twins with those in unaffected co-twins, give an unusual opportunity to investigate aetiological factors in early life, because comparative measures can be used to investigate factors that are usually difficult to recall exactly in order to enable comparison between cases and unrelated controls. We therefore conducted a case-control study of testicular cancer in twins in England and Wales.
MATERIALS AND METHODSThe methods of identification of the subjects and collection of data from them have been described in detail elsewhere (Swerdlow et al, 1996(Swerdlow et al, , 1997. In brief, twin-born men with testicular cancer incident in England and Wales from 1971 to 1989 were identified by clerical linkage between national cancer registration files, birth registers and the National Health Service Central Register. All same-sex twin pairs identified in this way were contacted to administer questionnaires on zygosity (Torgersen, 1979;Mack et al, 1995) and potential risk factors for testicular cancer. These factors included perinatal variables, anthropometric measures at various ages, childhood diet and exercise, puberty, acne, childhood infections, other illnesses and malformations, and diseases and malformations in relatives. The risk factor information analysed was that reported in the questionnaires; no previously recorded information on these variables was available to us. When responses were contradictory or unclear, we attempted to resolve this by recontact with the subjects, usually by telephone.For each risk factor, we examined the responses of the twin pair to questions about each other, in pairs in which both twins had replied, in order to determine whether responses from a single respondent about the two members of the pair would be reliable. When this proved to be so, we analysed data from the case and cotwin when both had replied, as well as from a single twin (about both of the pair) when he had been the sole respondent. When responses from subjects about their co-twins proved to be unreliable, we restricted the analyses to data from subjects about themselves. In analyses of comparative variables, pairs who gave contradictory responses to a question (e.g. each reported that they were the taller at a particular age) were excluded from analysis. In no instance did they constitute more than 5% of pairs. When one twin repor...