Table B-1 and Table B-2 illustrated that the relative risk of breast cancer in women of all ages combined who have ever used oral contraceptives is close to unity, with narrow confidence limits; it was also shown that there has been little overall increase in risk in long-term users (Table B-3 and Table B-4), or in women long after initial exposure (Table B-5 and Table B-6). Three possible interpretations of findings for young women are not incompatible with these overall results. One is that the proportion of cases under age 45 in most of the studies of women of all ages was relatively small, and the increased risk in young women may have contributed too little to the overall estimated value of the relative risk to elevate it appreciably above unity. A second possible interpretation is that women who have used oral contraceptives may tend to have their tumors diagnosed earlier than do nonusers, due to more screening or breast self-examination (screening bias). A third possibility is that oral contraceptives stimulate growth of breast carcinomas, so that women who are destined to develop them tend to do so at an earlier age (growth stimulation). The latter two explanations would eventually result in an observed reduction in risk in women at some time beyond age 45, and no overall alteration of risk.
Table B-23 shows relative risks in long-term users for women in various age groups above and below 45 years. Three of the studies (Paul et al., 1986; McPherson et al., 1987; Vessey et al., 1989) are supportive of either of the latter two possibilities given above, in that relative risks are greater than 1.0 in young women, and under 1.0 in older women. The three other studies shown in the table, however, do not show this variation in relative risks by age.
Other evidence that oral contraceptives are associated with earlier diagnosis or tumor growth stimulation would be a rise, and then a fall, in relative risk with duration of use in current users (as a surrogate for risk in relation to time from initial exposure in women who continue to take the pill). The first three studies shown in Table B-24 reported relative risks by duration of use in women who were last exposed within the previous year. As shown in column 1 of the table, two of these studies (Rosenberg et al., 1984; CASH, 1986) showed some increase and then a fall in relative risk in relation to duration of use in current users, but the WHO (1990) study did not. One might also expect to see a decline in risk with time since last exposure in women in those duration-of-use categories in which an increase in risk in current users was observed. This trend is apparent only in the WHO study.