Controversies in mammography screening: let us not ignore science in this never-ending debate

Abstract

There has been an ongoing debate on mortality reduction from mammography screening since the randomized control trials in the 1970s and 1980s which showed a mortality reduction of about 25%. In addition to the mortality reduction, the debates have been focusing on screening of women below 50 years of age and over-diagnosis. Randomized control trials (RCTs) are the gold standard for investigations of new therapies and diagnostic interventions. The design of a RCT is challenging, and some of the mammographic RCTs had, without doubt, limitations. RCTs are comparing the outcome of women invited to intervention (mammography screening) versus a control group. However, many women invited do not attend (‘nonattenders’) and many women in the ‘control group’ have a mammographic examination outside the program. No correction is made for this bias in the RCTs. Results from the Swedish RCTs have been of utmost importance, and are the scientific rationale for implementing mammographic service screening. The recently published results from the Swedish ‘Two-County Trial’ with a follow-up of three decades has confirmed that a highly significant reduction in breast cancer mortality of 30% in women invited to screening has persisted throughout the long follow-up period (1). The Nordic countries have today nationwide populationbased screening programs. An overview of service screening from 13 Swedish counties reported a substantial reduction in breast cancer mortality between 40% and 45% (2). The Icelandic mammography screening, the oldest populationbased program worldwide, showed a 35–40% reduction in breast cancer mortality based on a case-control study (3). A similar mortality reduction was reported in the Copenhagen screening for women actually participating in the mammography program (4). The Danish screening program has recently become nationwide. The smaller impact on mortality rate reported in the Finnish program has been explained by the younger age group targeted in Finland (5). It should be noted that a remarkably high mortality reduction of more than 70% after correction for selfselection bias has recently been reported (6). Evaluation of the Norwegian Breast Cancer Screening Program (NBCSP) has just started, and the results are expected within one to two years. Another controversy has been mammography screening for women aged 40–49 years. Most of the RCTs were not designed to evaluate the impact on this age group. However, several studies have clearly indicated that screening would be of benefit for women below 50 years of age. A recent publication from Sweden on the effectiveness of population-based service screening for women aged 40– 49 years showed a reduction in breast cancer mortality of 29% in the age group 40–49 years and 37% in the age group 45–49 years for the women who actually attended the mammographic screening program (7). In 2009 the US Preventive Services Task Force (USPSTF) declared that routine screening mammography in women aged 40–49 years should not be recommended (8). This statement has been heavily criticized as it is not supported by science. The age of 50 years is nothing but an arbitrary threshold without biological or scientific importance and has been arbitrarily chosen as a surrogate for menopause (9). Over-diagnosis, defined as a cancer diagnosed at mammography which would never have presented in the women’s lifetime in the absence of screening is another controversy. Over-diagnosis exists, but the question is how big the problem is: 1%, 10%, 25%, or even higher (10–12). The reason is that presently there is no way to determine which cases of DCIS (and perhaps even small low-grade invasive cancers) that will progress to lethal cancers. Spontaneous regression, on the other hand, is obviously a very rare phenomenon (12). Headlines in newspapers and medical journals over the last years on a minor effect of mammographic screening on mortality and a substantial over-diagnosis rate are beginning to disturb our belief. A register-based study on the NBCSP published in New England Journal of Medicine concluded that mammography screening accounted for only one-third of the moderate reduction in breast cancer mortality (13). Another recent publication compared results from Sweden and Norway, and concluded that screening did not play ‘a direct part’ in the reduction of breast cancer mortality (14). Over-diagnosis up to 50% has been postulated (11). If so, the question of course would be whether ‘mammographic screening is justifiable considering its substantial over-diagnosis rate and minor effect on mortality’ (10). The flaws in these estimations have been summarized in an overview on over-diagnosis in mammography screening (12). Regarding the recently reported minor effect on mortality and the high over-diagnosis rate in the NBCSP (10, 11, 13, 14), it is important to be aware that these publications are register-based research without access to individualized data. Register-based studies have several limitations (which is also the case for case-control studies!). For