DÄ internationalArchive23/2008Benefit and Risk of Mammography Screening – Considerations from an Epidemiological Viewpoint: In Reply

Correspondence

Benefit and Risk of Mammography Screening – Considerations from an Epidemiological Viewpoint: In Reply

Dtsch Arztebl Int 2008; 105(23): 421-2. DOI: 10.3238/arztebl.2008.0422

LNSLNS Our article had three main points:

(1) The effect of mammography screening in increasing treatment effectiveness and prolonging survival should be presented from the viewpoint of the individual woman invited for screening, while the scientific demonstration of its efficacy must be based on the observation of mortality.

(2) Some authors' attempts to present the effect of screening intelligibly have resulted not just in altered formulations, but in substantially changed and inaccurate information.

(3) The screening debate fails to take account of the different definitions of risk which may play a role in the assessment of benefit and harms, which differ as to whether the severity of damage associated with the undesired event is taken into account (as is the case with insurance) or not (as is the case in epidemiology).

Mühlhauser's letter repeats the familiar argument that the effect of screening can only be demonstrated scientifically by means of randomized epidemiological studies with mortality as endpoint, which we address (13).

We agree in general with Beise's comments. We did point out in our article, however, that the effect of screening is, by definition conditional: if an as yet undetected malignancy is present, then it might be advantageous to detect it as early as possible by screening and to treat it (3). Our proposed solution is that the benefit of screening in case such a disease is present should be presented in terms of absolute frequencies, just as Beise suggests. The reference to insurance has nothing to do with conditional probabilities, but rather with the varying definitions of "risk": in epidemiology, risk is the probability of an adverse event, regardless of the magnitude of the associated damage; in insurance, the definition of risk takes this magnitude into account.

Weymayr fails to recognize that our presentation is the computational transformation of the results of the randomized studies with mortality as an outcome into absolute rates of survival. Thus, net effects are presented. Overdiagnosis need not be subtracted, because one does not die from a disease which remains subclinical for the remainder of one's life.

Klemperer's letter from the German Network for Evidence-Based Medicine concurs with our goal of yielding patient information that is as accurate and as comprehensive as possible. We believe it is wrong, however, to take the "magnitude of the damage" in case of a cancer diagnosis limited to a possible lethal outcome, but that the possible implications of treatment and their effects on the patient's quality of life must also be taken into account. The probability of disease should be communicated to the patient as a separate quantity, unaffected by screening.

Abholz extends the discussion of mortality reduction by the early detection of existing disease to individuals who are free of the disease and therefore not at risk of dying from it. This is exactly what we criticized in our article. The probability of getting the disease is relatively low (about 5%); however, given the disease, 31 of 100 will die of it within 10 years without screening, 23 if screening is offered, and 20 if they participated.

Jöckel et al. raise objections to the title of the article and the scaling for the used quantities. The title of the article arose in the course of pre-publication review; originally, we proposed: "Risk communication in mammography screening." Colleagues in epidemiology objected to the term "damage" (or harm, "Schaden") as being not entirely free of emotional connotations; therefore, in the article as printed, we consistently referred to "disadvantageous effects," which we contrasted with benefits. As for the scaling of quantities, due to small numbers we gave incidence figures in the usual epidemiological form, i.e., per 100 000 individuals (5); for more frequent events, we used the scaling usual in that area, i.e., per 1000 individuals (4). The benefit in terms of avoided deaths was transformed into the per 1000 scale in table 1 (bottom) and was contrasted in the text on this scale to the disadvantageous effects "breast cancer surgery with benign findings" and "overdiagnosis." We also pointed out that the most common disadvantageous effect by far is a false positive finding whose frequency was shown in table 2. The authors of the letter present these figures on a larger scale which makes them obviously no more "balanced," but merely larger. As for the effect of false positive findings on the quality of life, we have already dealt with this issue in multiple previous publications (1, 2). The focus in the current article was on the correct interpretation of study results.

Schwartz's first assertion is correct: it was indeed our intention to point out the conditional nature of figures on mortality reduction by screening, and to translate them computationally into lethality figures. Since this was not based on empirical data, lead-time bias, relevant for empirical evaluation, was omitted from consideration (13). Selection bias in relation to disease stage is referred to as length bias and does not affect the outcome of offering screening, as opposed to actual participation. DOI: 10.3238/arztebl.2008.0422


Prof. Dr. rer. nat. Nikolaus Becker
Deutsches Krebsforschungszentrum, Abteilung Krebsepidemiologie
Im Neuenheimer Feld 280, 69120 Heidelberg, Germany
n.becker@dkfz.de

Dr. med. Hans Junkermann
Universitätsfrauenklinik, Sektion Senologische Diagnostik
Voßstr. 9, 69115 Heidelberg, Germany

Conflict of interest statement
The authors of all letters and the reply declare that they have no conflict of interest as defined by the International Committee of Medical Journal Editors.
1.
Becker N: Screening aus epidemiologischer Sicht. Radiologe 2002; 42: 592–600. MEDLINE
2.
Becker N: Die Rolle der epidemiologischen Qualitätsparameter im Mammographie-Screening. Radiologe 2006; 46: 984–992. MEDLINE
3.
Morrison AS: Screening in Chronic Disease. Second edition. Monographs in Epidemiology and Biostatistics, Volume 19. Oxford University Press, New York, Oxford 1992.
4.
Perry N, Broeders M, de Wolf C, Törnberg S, Holland R, von Karsa L, Puthaar E (eds.): European guidelines for quality assurance in breast cancer screening and diagnosis – Fourth Edition. Luxembourg: Office for Official Publications of the European Communities 2006.
5.
Robert-Koch-Institut und die Gesellschaft der epidemiologischen Krebsregister in Deutschland e.V. (GEKID) (Hrsg.): Krebs in Deutschland 2003–2004. Häufigkeiten und Trends. Berlin 2008.
1. Becker N: Screening aus epidemiologischer Sicht. Radiologe 2002; 42: 592–600. MEDLINE
2. Becker N: Die Rolle der epidemiologischen Qualitätsparameter im Mammographie-Screening. Radiologe 2006; 46: 984–992. MEDLINE
3. Morrison AS: Screening in Chronic Disease. Second edition. Monographs in Epidemiology and Biostatistics, Volume 19. Oxford University Press, New York, Oxford 1992.
4. Perry N, Broeders M, de Wolf C, Törnberg S, Holland R, von Karsa L, Puthaar E (eds.): European guidelines for quality assurance in breast cancer screening and diagnosis – Fourth Edition. Luxembourg: Office for Official Publications of the European Communities 2006.
5. Robert-Koch-Institut und die Gesellschaft der epidemiologischen Krebsregister in Deutschland e.V. (GEKID) (Hrsg.): Krebs in Deutschland 2003–2004. Häufigkeiten und Trends. Berlin 2008.