Twenty five year follow-up for breast cancer incidence and mortality of the Canadian National Breast Screening Study: randomised screening trial

Miller has argued that their hazard ratio comparing death by breast cancer (1.47) between their mammography group and their control group from the first round of screening is expected and Miller cites the Swedish two county trial’s hazard ratio of 1.26 as evidence that their hazard ratio is not abnormally imbalanced against mammographic screening [1]. It should be noted that the 1.26 hazard ratio from the Swedish trial was based on comparing mammography to a control group receiving no additional care, whereas Miller’s hazard ratio of 1.47 is based on comparing mammography to a control group receiving breast examinations. Since breast examinations provide benefit by identifying some malignant tumours, we should expect the CNBSS trial [2] to achieve a lower hazard ratio than the Swedish two county trial’s 1.26, however it is noticeably larger. The risk of death from breast cancer does indeed appear to be inflated for women in the first round of screening in the mammography group of the CNBSS trial [2] and inappropriate randomization is a plausible explanation as identified by Dr. Kopans at Harvard [3].

I have previously mentioned the additional issue that lethal advanced stage tumours caught in the first round of screening contribute to degrading the mortality results of any screening technology being evaluated as the technology was not given the opportunity to detect those tumours at a treatable stage. Rolf Hefti has asked “isn’t this bias relatively irrelevant as it affects both groups”? [4] The answer is no, this bias is not irrelevant because the two techniques being compared (mammography and breast examination) have substantially differing sensitivity to the detection of breast cancer. Miller’s study demonstrated increased tumour yield in the mammography group. The bias from including lethal tumours detected in the first round of screening preferentially degrades the results of any technique that is more sensitive to detecting those lethal tumours, in this case mammography. Mammography was in fact catching more cancers and for every additional lethal cancer mammography catches in the first round of screening, its mortality results degrade. This bias does of course exist in the control group as well (when breast examination catches a lethal tumour in the first round of screening), however, whichever technique detects the most lethal tumours in the first round of screening is negatively affected by this bias the most. Since mammography substantially increases tumour yield in the first round of this trial, this bias is not irrelevant as it preferentially degrades the mammography mortality results. This bias can be avoided in both groups by simply comparing their mortality results from subsequent rounds of screening only. If we look exclusively at the tumours detected after the first round of screening (from both mammography and breast examination) then the results of this study demonstrate a mortality benefit from mammography (hazard ratio 0.9), though admittedly it is smaller than the mortality benefits reported in most major randomized controlled trials. Potential reasons for the underperformance of this trial have been presented previously [3, 5].

Rolf Hefti [4] has also cited a recent publication from Norway [6] and indicated that it supports his anti-mammographic screening position. It should be noted that the Norwegian trial cited [6] is population based and did not include a careful assessment whereby all the women in the experimental group received mammography. The experimental group was mixed including both women who receive mammographic screening and those who do not. Basing conclusions as to the effectiveness of a disease screening technology on a study such as this is fraught with peril, as confounding factors are impossible to disassociate from the analysis. It should also be noted that their control group was young women, whereas their experimental group was substantially older. Given that it is known that older women are more likely to develop breast cancer, this comparison was inappropriate. Studies that carefully compare women who receive mammographic breast cancer screening with those that do not have regularly demonstrated a mortality benefit from the use of mammography, which is well summarized in the Marmot review [7].

Jacob Levman, PhD
Institute of Biomedical Engineering
University of Oxford

[1] Miller, Baines, Re: Twenty five year follow-up for breast cancer incidence and mortality of the Canadian National Breast Screening Study: randomized screening trial, British Medical Journal, 2014;348:g366, March 19, 2014.
[2] Miller et al., Twenty five year follow-up for breast cancer incidence and mortality of the Canadian National Breast Screening Study: randomized screening trial, British Medical Journal, 2014;348:g366.
[3] Kopans, Re: Twenty five year follow-up for breast cancer incidence and mortality of the Canadian National Breast Screening Study: randomized screening trial, British Medical Journal, 2014;348:g366, February 12, 2014.
[4] Hefti, Re: Twenty five year follow-up for breast cancer incidence and mortality of the Canadian National Breast Screening Study: randomized screening trial, British Medical Journal, 2014;348:g366, March 18, 2014.
[5] Tabar, Re: Twenty five year follow-up for breast cancer incidence and mortality of the Canadian National Breast Screening Study: randomized screening trial, British Medical Journal, 2014;348:g366, February 18, 2014.
[6] Lousdal et al., Trends in breast cancer stage distribution before, during and after introduction of a screening programme in Norway, European Journal of Public Health, published online ahead of print edition, March 4, 2014.
[7] Marmot et al., The benefits and harms of breast cancer screening: an independent review, British Journal of Cancer, 2013 108:2205-2240.