Mammography Screening Matters: Response to British Medical Journal

Reading Time: 4 minutes read

I am sure you have been disturbed by the media report questioning the value of mammography screening, based on a “publication” in BMJ, therefore I would like to give you a very simple and easily understandable way of looking at the issue. You are one of the experts in your area, so it is important to be able to argue properly.

A report in the current issue of the British Medical Journal claims that there is no evidence that mammography screening has played a direct role in breast cancer mortality reductions in countries in which screening has been implemented. Lead author Phillipe Autier and his colleagues compared breast cancer mortality trends in three pairs of adjacent countries (Sweden vs. Norway; Northern Ireland vs. Republic of Ireland; and Belgium vs. Netherlands), with each comparison including a country that introduced mammography screening some years earlier than the other. Comparing breast cancer death rates between 1989-2006, the authors observed similar trends in the reduction of breast cancer deaths in each country pair. They conclude that mortality trends are more likely to be influenced by improvements in therapy than the effect of mammography screening.

One should rely on people’s common sense – and try to ask relevant questions:

Mammography screening has been exposed to a massive anti-mammography campaign during the past 10+ years. The “criticism” originates from one-two persons living in Denmark. Interestingly enough, they have not included Denmark in the comparison, although there is a notable difference:

• The country of Denmark has one of the highest breast cancer mortality rates in Europe (!); it is at the same level as in Serbia (WHO 2010 stats).
• Sweden has one of the lowest (!) rates of breast cancer death in Europe.
• The only difference between the two countries in this regard is that Denmark – so far – has not had an organized, nationwide screening, but Sweden started to gradually build up organized mammography screening in the entire country in 1986, following the publication of the results of the Two-County Swedish trial.
• The therapeutic guidelines are identical in the two countries (!). Thus, the main difference is: ongoing organized mammography screening in Sweden for two or so decades, but no organized mammography screening in Denmark. How can one then conclude that it is the therapeutic regimens that make the decline in breast cancer death in Sweden?

One should think about the following as well: the Swedish researchers have recently published the largest 40-49 screening data (Effectiveness of Population-Based Service Screening With Mammography for Women Ages 40 to 49 Years. Evaluation of the Swedish Mammography Screening in Young Women (SCRY) Cohort Barbro Numan Hellquist, Cancer Sept 2010). Half of the country invited women aged 40-49 to screening, half of the country did not (politicians’ random decision). Still, there was a 29% significant decrease in mortality from breast cancer among women who attended screening. Had the therapeutic regimens had the strong effect that the BMJ article assumes and not screening, then there should not have been a difference in breast cancer death between the two halves of Sweden.

Similarly: the Swiss Ministry of Health came out with the stats that the French-speaking part of the country that has been offering screening for a long time has a significantly lower breast cancer death rate than the German speaking part that has not been offering screening.

In addition: fortunately, the Danish Ministry of Health does not seem to care much about the Danish “researchers’ ” steady anti-screening activities, since the authorities have – at last – made a decision about financing nationwide organized screening. The first round of screening has been completed, and a manuscript has already been written about their experience. As expected, the colleagues can pride themselves with very good results.

Further comments: Even if screening is offered for a given population, only about 50% of the cancers will be screen-detected because of the large number of breast cancers that will be diagnosed in the rest of the population – among those who do not attend screening and among those who have never been invited (age <40, <50, >70, >74). Not knowing the “detection mode” of each individual breast cancer case makes any “assessment” of breast cancer death rate in a population a mere speculation. Note that this information is NOT available in the Cancer Registries. The impact of mammography screening should only be measured in women who have been invited to screening, without the “dilution effect” of the cancers diagnosed in the rest of the female population.

Not only does the information about the detection mode of each cancer case have to be collected, but one also has to eliminate all of the cancers diagnosed before the start of screening! Since screening is gradually built up in different regions, this cannot be done using only Register data. Eliminating the cases that received breast cancer diagnosis before the screening epoch is vital because about 52% of those who died from breast cancer during a ten-yr period following screening started received breast cancer diagnosis in the prescreening era! Including these breast cancer cases results in a 52% error, due to “contamination” of the data and underestimates the real impact of screening. The correct way of performing statistical analysis is using the “incidence based mortality” method, i.e. both the date of diagnosis and the date of death should  be included within the examined time-period.

The current BMJ publication thus carries a series of serious professional mistakes.  This paper should not have been accepted for publication. The scientific community has several times pointed out the fatal flaws in the Danish and Norwegian authors’ way of looking at various aspects of population screening with mammography, but, despite many oral and written clarifications, they repeat the erroneous calculations again and again.

Just because two countries seem to share similar geography doesn’t mean that their breast cancer mortality trends are easily compared. For example, compared with Norway, Sweden has about 10% greater breast cancer incidence during the study period, and it was even greater before the study period began. That will influence mortality rates over time, since mortality rates are a function of incidence rates over time and their corresponding survival. The authors did not adjust for differences in the incidence rates between countries.

The publication is full of other mistakes too. Some of them are simply false statements and some of them are deliberate repeat of their mistakes,since their incorrectness have already been demonstrated in peer reviewed journals several times. The sad thing is that some women will stop attending mammography screening and their cancers might grow to an undesirably late stage.

Editor’s Note:  Dr. Tabar recently published a study pointing to the long-term benefits of mammography screening.  You can see these results discussed in a previous Ask Anne post.