Intended for healthcare professionals

Rapid response to:

Analysis

Breast screening: the facts—or maybe not

BMJ 2009; 338 doi: https://doi.org/10.1136/bmj.b86 (Published 28 January 2009) Cite this as: BMJ 2009;338:b86

Rapid Response:

Estimate of breast screening benefit was 6 times too large

Nicholas J Wald and Malcolm R Law provide a seriously misleading
estimate for the number of breast cancer deaths avoided by screening (1).
They claim that over ten years of screening, 6 women per 2000 will avoid
dying from breast cancer. Their estimate is contradicted by the results
from the randomised screening trials, and there are other problems with
their approach.

The first meta-analysis of the Swedish trials reported a 29% relative
risk reduction in breast cancer mortality for women aged 50 to 69 years
when invited to screening (2). The age range corresponds to that used by
Wald and Law for breast cancer mortality, 55 to 74 years, as it takes some
years for screening to take effect. The Swedes calculated that this effect
corresponds to one breast cancer death being avoided per 1000 over ten
years (2, page 976).

The Cochrane review provides more trials and a more realistic
estimate of the effect, namely a 15% reduction in breast cancer mortality
(3). This estimate agrees both with the systematic review done for the US
Preventive Services Task Force that suggested 16% (4), and with the most
recently updated meta-analysis of the Swedish trials that reported 15%
with the follow-up model (5).

As the effect is only 15%, one breast cancer death is avoided per
2000 over ten years. Wald and Law's estimate can only be correct if one
assumes that screening reduces breast cancer mortality by 90%. They derive
their estimate from a curious and unnecessary detour that involves
erroneous use of observational data. They use the death rate in England
and Wales and adjust this rate upwards to estimate a hypothetical death
rate in the absence of screening, assuming that the effect of screening is
30%, and then calculate that 6 per 2000 will benefit, again assuming that
the effect is 30%. Thus, by using the wrong estimate twice in their
detour, they inflate their estimate twice.

They also overlook that many women who die aged 55 to 74 years die
from slow-growing cancers that were diagnosed before the women became
eligible for screening, as they were too young. This is also a serious
error, somewhat like tallying how many die from myocardial infarction in a
country, and then using the effect of streptokinase to calculate the
number of lives saved. Obviously, many die before they reach hospital, and
others are never treated because too many hours have elapsed since the
start of symptoms.

Finally, Wald and Law multiply their estimate by two and then say
that 12 will benefit after 20 years. This is also misleading. We based our
estimates on 10-year data because there are no trial data spanning 20
years of screening, with an unscreened control group. It is a bias-prone
method to extend estimates far beyond the time period covered by the data.
In addition, one should be aware that although the benefit might increase
with time, so will the major harms, overdiagnosis, overtreatment and false
-positive findings. The balance between benefits and harms would therefore
not be expected to change much.

In the wake of our paper, other gross exaggerations of the benefit of
screening have emerged. In "Behind the headlines", produced by the NHS
Knowledge Service, Professor Julietta Patnick, who is Director of the NHS
Cancer Screening Programmes, is quoted as saying that the ratio, instead
of being one life saved to 10 receiving unnecessary treatment, is much
nearer to a one to one ratio (6). Similarly, Professor Mike Richards, the
National Cancer Director, said to The Times: “We think it’s much nearer
one to one" (7).

We wonder how it can be possible to get the ratio wrong by a factor
of ten. A one to one ratio can be obtained if the effect is assumed to be
90% and the overdiagnosis is 15%. However, in the randomised trials the
effect was 15% and the overdiagnosis was 30% (3), and in countries with
organised screening programmes the overdiagnosis is 50% (8).

These observations support our suggestion that those who are
responsible for cancer screening programmes should not also be those who
provide information about its benefits and harms (9).

1. Wald NJ, Law MR. Response to Gotzsche. BMJ 2009.
http://www.bmj.com/cgi/eletters/338/jan27_2/b86.

2. Nyström L, Rutqvist LE, Wall S, Lindgren A, Lindqvist M, Ryden S,
et al. Breast cancer screening with mammography: overview of Swedish
randomised trials. Lancet 1993;341:973–8.

3.Gøtzsche PC, Nielsen M. Screening for breast cancer with
mammography. Cochrane Database Syst Rev 2006;(4):CD001877.

4. Humphrey LL, Helfand M, Chan BK, Woolf SH. Breast cancer
screening: a summary of the evidence for the U.S. Preventive Services Task
Force. Ann Intern Med 2002;137(5 Part 1):347–60.

5. Nyström L, Andersson I, Bjurstam N, Frisell J, Nordenskjöld B,
Rutqvist LE. Long-term effects of mammography screening: updated overview
of the Swedish randomised trials. Lancet 2002;359:909–19.

6. Breast cancer screening.
http://www.nhs.uk/news/2009/02February/Pages/Breastcancerscreening.aspx.
2009 (accessed 23 Feb 2009).

7. Smyth C. NHS rips up breast cancer leaflet and starts all over
again. The Times 2009; 21 Feb.
http://www.timesonline.co.uk/tol/life_and_style/health/article5776804.ece
(accessed 1 March 2009).

8. Jørgensen KJ, Gøtzsche PC. Overdiagnosis in publicly organised
mammography screening programmes: systematic review of incidence trends.
BMJ 2009 (in press).

9. Gøtzsche P, Hartling OJ, Nielsen M, Brodersen J, Jørgensen KJ.
Breast screening: the facts - or maybe not. BMJ 2009;338:446-8.

Competing interests:
None declared

Competing interests: No competing interests

03 March 2009
Peter C Gøtzsche
Director
Ole J Hartling, Margrethe Nielsen, John Brodersen, and Karsten Juhl Jørgensen.
Nordic Cochrane Centre, Rigshospitalet, 2100 Copenhagen, Denmark