Article Text

Download PDFPDF

Observational study
Mammography screening in Norway caused substantial overdiagnosis and did not reduce late-stage breast cancers
  1. Karsten Juhl Jørgensen
  1. Rigshospitalet, Department 3343, The Nordic Cochrane Centre, Copenhagen, Denmark
  1. Correspondence to:Karsten Juhl Jørgensen
    , Rigshospitalet, Department 3343, The Nordic Cochrane Centre, Blegdamsvej 9, Copenhagen 2100, Denmark, kj{at}cochrane.dk

Statistics from Altmetric.com

Commentary on: Kalager M, Adami HO, Bretthauer M, et al. Overdiagnosis of invasive breast cancer due to mammography screening: results from the Norwegian screening program. Ann Intern Med 2012;156:491–9.

Context

There is no question that breast cancer screening identifies cancers that would never have surfaced in a woman's lifetime in the absence of screening (overdiagnosis). However, high-quality studies are needed to quantify this harm, and this is the main objective of the study by Kalager and colleagues. A review of incidence trends of breast cancers over 20 mm in size (which are by definition considered to be advanced) in seven countries with long-running breast screening found unchanged rates with screening.1 This suggests that the massive increases in early invasive breast cancers and in situ cases seen when breast screening is introduced does not prevent late-stage cancers and is therefore not early detection, but overdiagnosis. It also explains why studies fail to identify a mortality benefit2 ,3 and why mastectomy use increases with screening.4 The main focus of this study is a quantification of overdiagnosis, but it also compares late-stage cancers in screened versus non-screened areas.

Methods

The study included 27 238 invasive breast cancers diagnosed in Norwegian women 50–79 years in the observation period (1986–2005). Of these, 28.6% were diagnosed after screening was introduced.

Breast screening was introduced stepwise in Norway over 9 years. As incidence differed between counties before screening, screened and non-screened regions could not be compared directly. The authors therefore created two historical comparison groups consisting of women in the areas that later either would or would not offer screening. These were then compared to women in the same areas during the screening period using incidence rate ratios (IRR). Overdiagnosis was estimated as the ratio between these IRRs.

To compensate for the increase in incidence due to earlier detection (lead time), the authors used two approaches: (1) Approach 1 included women above the screening age (70–79 years) to take into account any compensatory declines in incidence after screening stopped and (2) Approach 2 excluded the initial prevalence peak and compared incidence in the screened group with that in women 2 and 5 years older in the historical group in the screened areas. To test the robustness to short follow-up in some counties, a separate analysis was done including only areas offering screening for 10 years (40% of the population).

Findings

For Approach 1, overdiagnosis was estimated at 18% (95% CI 11% to 25%) to 25% (19% to 31%) and for Approach 2, 15% (8% to 23%) to 20% (13% to 28%). This was interpreted as 6–10 women overdiagnosed for every 2500 invited to screening.

The graphs show a very modest decline in incidence in previously screened women aged 70–74 years, which may be compensatory. However, a similar difference existed for women 45–49 years, which cannot be due to screening. It is unclear as to whether these declines could be due to chance. In any case, the vast majority of the increase in incidence remains uncompensated. A 24% decline in stages 3 and 4 breast cancers from the prescreening to the postscreening era was seen in both screened and non-screened areas (IRR 0.76 (0.61 to 91)).

Commentary

The estimates of overdiagnosis are conservative. First, Approach 1 compared age ranges where one-third had passed the upper age limit for screening. This dilutes the estimate of overdiagnosis. Including the invited age range only (50–69 years) yields 1.86/1.32×100=41% overdiagnosis (in table 1 of the article). Subtracting the small compensatory decline in women 70–74 years would not change this much. Second, Approach 2 used lead time assumptions, not observations, and excluded the period with highest incidence. The assumed lead times were high, although commonly used. Third, carcinoma in situ lesions were not included. The authors acknowledge that overdiagnosis of in situ lesions contributes to overdiagnosis, but argue that there are ‘fundamental methodological differences’ that prevents including them. However, these are not specified. In situ cases make up 10–25% of screen-detected breast cancers.

A study used the stepwise screening introduction in Denmark to form a contemporary control group and found 33% overdiagnosis,5 but the present study is the first to use the method (stepwise screening to form a contemporary control group) to look at cancer stages. Their finding has major implications for central clinical outcomes. Despite reservations, the methods and data are of high quality. I agree with the authors who say that ‘overdiagnosis and unnecessary treatment of non-fatal cancer creates a substantial ethical and clinical dilemma and may cast doubt on whether mammography screening should exist’.

References

View Abstract

Footnotes

  • Competing interests None.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.