It is reassuring that Olsen et al.1 in their article confirm our finding of a 10% nonsignificant reduction of breast cancer mortality after the introduction of the Norwegian breast cancer screening program.2
However, we are surprised about the interpretation and conclusion of this study. Despite the result of only a very modest, nonsignificant effect of the screening program (7–11%), the authors claim that the “real effect” of the program would be much larger (about 25%). The study results do not support this conclusion. Further, the authors claim “40% of women used regular mammography prior to the program.” This statement is not based on the results of this study. We wonder how the authors can defend a conclusion that is completely different from the observed results of the study.1
We doubt the scientific grounds for the claimed large amount of nonprogram (opportunistic) screening (“regular mammography”) that the authors cite as explanation for the lack of effect of the screening program. The estimate of 40% is based on a study by the same authors where they did not distinguish screening mammography from clinical mammography.3 The authors seem to interpret all previous mammography as screening.3 However, mammography is not only a screening tool but also a widely used diagnostic tool to examine patients with breast symptoms and for surveillance of patients with breast cancer. Thus, it is not at all surprising that many women have had mammograms at some time in their lives before the introduction of organized screening and cannot serve as an estimate for nonprogram screening. Moreover, the data from the Norwegian Radiation Protection Authority which Olsen et al. cite as reference for preprogram mammography screening activity are, according to the Authority itself, not reliable to use as a source to estimate non-program screening, a fact which the corresponding author (dr. Lynge) is well aware of.
The Norwegian Women and Cancer Study (NOWAC) is the other source of information for non-program screening.3 Several factors could bias the estimate of non-program screening based on the questionnaire from the NOWAC study: First, 92% of women in the NOWAC study (in 2002) reported that they had a mammogram before attending the program, whereas 64% of the women attending the screening program reported the same. Thus, there are 44% more women in the NOWAC study reporting that they have had a mammogram than in the program. This might be the result of a selection bias; how did the authors take account of this? Second, it is not clear if each woman included in the NOWAC study was asked repeatedly about previous mammography activity. Women, who have previously had a mammogram, will always have had a mammogram. As more women are enrolled in the program over time, this might influence the amount (%) of women with previous mammograms (both denominator and nominator get smaller but not necessarily in similar amounts). Third, the question about mammography was: “Do you regularly undergo mammography examination?” The possible answer categories were: “no, ” “yes with an interval of 2 years or less, ” or “yes with an interval of more than 2 years.”3 Nobody could answer just yes without stating a regular pattern. As the questionnaire is built up, everyone who has had a previous mammogram will be counted as having had “regular mammography.” Fourth, in the Olsen et al. study, women born in 1932–1946 (aged 50 to 65 years in 1996) are followed through 2001. Thus, the amount of possible non-program screening registered in 2002 might not influence the mortality estimates.
If there had been a substantial amount of non-program screening, the amount of such screening had to be considerably different in the four study groups to affect the effect of the program. No data are presented to provide such information. Given the amount of women who had “at least one mammography examination in the NOWAC study” in 1996, and 1997–1998 in the screening and nonscreening counties (Table 2 in Ref. 3), the percentage of women aged 50 to 69 years with prior mammograms in the control groups are as follows: The historical control group: 46.6%; the historical regional control group: 40.0%; and the regional control group: 46.7%. The number of women with at least one mammogram based on the NOWAC study is surprisingly similar in all control groups. Thus, it is unlikely that these differences could influence the effect of the screening program. The statement in the conclusion (“40% of women used regular mammography prior to the program”) cannot justify the “estimate” of an expected effect of about 25% associated with the program.
There is reassuring evidence against a high level of non-program screening.2,4 The incidence of ductal carcinoma in situ (DCIS) could be an indicator of the amount of screening; in Denmark, the incidence of DCIS before the start of the screening program was approximately 10–15 per 100,000 person-years in women aged 50–69 years.5 Similarly, in Norway, the incidence of DCIS was 10–12 per 100,000 person-years in women aged 50–69 years before the start of the program.6 Both in Denmark and Norway, the rate of DCIS after screening which was implemented in women aged 50–69 was 40–50 per 100,000 person-years.5,6 Denmark did not have non-program screening,7 and if Norway should have had non-program screening as Olsen et al. claim, the numbers would have been substantially different from the Danish, but they are not. Incidence of invasive cancer among women not offered screening (comparable to the regional control group in the present study) did increase in Norway; however, use of menopausal hormone treatment is a more likely contributor to the increase than non-program screening.4 Olsen and co-workers refer to several sources8, 9 for the amount of such screening.1 However, none of the referred studies have investigated this.8, 9 Even though several studies claim that there have been considerable amount of non-program screening in Norway, and this is repeated several times, it does not necessarily become true.