Response to Zahl

We read with interest the letter by Zahl commenting on the methods we used in our estimation of the reduction in breast cancer mortality due to screening vs treatment in Norway (1). In the following, we will reply to his points.

“Women in Norway under age 50 years are not invited for screening, but have experienced a considerably larger reduction in breast cancer mortality since 1989 (-35.2%) than women aged 50-69 years (-22.6%) who are invited.” 

The above statement does not go against our assumptions or conclusions. Women younger than 50 years have different types of tumors, more hormonal negative, and fast growing (2, 3), so it is plausible that women younger than 50 years have benefited more from new chemotherapy and targeted treatment than women 50 years and older. Younger and older women are hardly comparable in terms of tumor characteristics and consequently treatment algorithms. We think Zahl’s statement, “Improved treatment and centralized care may thus explain the entire observed reduction in breast cancer mortality also in the screened age groups,” is based on a comparison between 2 very different groups of women.

Our linear assumption is that we have largely discussed this assumption in the discussion section of our paper, with possible implications of the assumption’s violations. The fact that “breast cancer mortality rate was almost constant, and then it started falling from 1995” does not say much about the linear assumption that, in our model, referred only to the treatment effect and not the mortality in general.

We do not agree with the statement, “many women in the invited group were offered trastuzumab in the period 2005-2014.” Trastuzumab was offered to a minority of women diagnosed with breast cancer, 12%-15%, or 300-400 women annually in Norway (4), and even fewer among those with screen-detected breast cancer (3). The effect of trastuzumab on breast cancer mortality might thus be limited in the whole population. Also, as we discussed in our paper, it is hard to understand how the progressive changes in breast cancer treatment from the 1970s might have influenced mortality. It is plausible that the rise of new chemotherapies (eg, taxanes), hormonal therapy (eg, tamoxifen), and use of sentinel node technique in the 1990s has had a much larger impact than trastuzumab in the 2000s. The increased use of trastuzumab Zahl is referring to might be due to use of this medication for treating other types of cancer.

In summary, changes in screening and diagnostic tools, life-style factors (eg, breast awareness and awareness of risk factors, use of hormonal replacement treatment), and improved treatment will always represent a challenge in the estimation of benefits and harms of mammographic screening. However, not even a randomized controlled trial would solve this problem, because of contamination. We agree with Zahl that our observational study might be susceptible to different sources of bias, but at the same time, we think it is reassuring that our results are well in line with other important studies (5–7).

References