News

Row over breast cancer screening shows that scientists bring "some subjectivity into their work

Susan Mayor, London

The difficulty of getting scientists to reach a consensus on their interpretations of trial data was illustrated last week when two reviewers for the Cochrane Breast Cancer Group published their own review of mammography screening trials after editors from the group had disagreed with some of their conclusions.

The review claimed that there was no reliable evidence to support the value of mammo-graphy screening in reducing deaths from breast cancer and alleged an association with increased rates of breast surgery.

Ole Olson and Peter Gøtsche from the Nordic Cochrane Centre, Righospitalet, Copenhagen, Denmark, reassessedas part of a Cochrane reviewa meta-analysis of seven randomised trials of screening mammography which they had previously carried out. This confirmed their original conclusion, they said, that there was no evidence of a reduction in either total or breast cancer mortality in two of the trials that they considered to be of sufficient quality to analyse.

They added: "We have also confirmed, with additional data, which the editors of the Cochrane Breast Cancer Group have elected to defer from publication until further editorial review has been completed, our earlier finding that screening leads to more aggressive treatment, increasing the number of mastectomies by about 20% and the number of mastectomies and tumourectomies by about 30%" (Lancet 2001;358:1340-2 and www.thelancet.com).

In the Lancet in the same issue (pp 1284-5) the editor, Richard Horton, wrote: "The process of collaboration within the Cochrane Breast Cancer Group has broken down badly in the case of the Gøtzsche and Olsen overview. The resulting tensions among colleagues indicated that even in the best organisations raw evidence alone is sometimes insufficient to influence opinion.

"When the Nordic investigators submitted their systematic review to the editors of the Breast Cancer Group, they found that their conclusions were unwelcome."

Professor Alessandro Liberati, professor of biostatistics at the University of Modena, Italy, and a member of the Cochrane Breast Cancer Group, said: "We were amazed by the tone of the editorial. It implies a plot by the editorial group against the reviewers, and this is clearly not the case."

Professor Liberati, who is also director of the Italian Cochrane Centre, said that the group will answer these allegations in a letter to the Lancet.

"We are in a situation of scientific controversy, with different interpretations of the same dataset," Professor Liberati explained.

The editorial group of theCochrane Breast Cancer Group agreed with the Nordic Centre reviewers about the lack of effect of mammography screening on mortality, and this is already reflected in the Cochrane review that has just been published in the Cochrane Library (Issue 4, 2001, www.cochranelibrary.net).

Professor Liberati said, "But we have substantial disagreement on the interpretation of the data on the impact of mammography screening on surgery and treatment so we have offered to wait for further peer review. In order not to delay the Cochrane review, we suggested that it should go ahead and be updated in subsequent issues of the Cochrane Library after further peer review."

The office of the NHS cancer screening programmes in the United Kingdom also disputed the association between mammography and treatment. In a statement, it said: "It is difficult to evaluate these claims, as they are based on small subgroups of individuals from two studies only, classified as poor quality studies by Gøtzsche and Olsen.

"Furthermore, more `aggressive' treatment of breast cancer might be expected to be beneficial for women with breast cancer. Gøtzche and Olsen did not investigate directly whether or not what they regard as `aggressive' treatment was beneficial."

The statement also noted that the Nordic Centre researchers' conclusion about the lack of benefit of mammo-graphy came from analysis of the two studies they considered to be of medium quality.

"Many researchers would classify all seven studies as of similar quality, and when the results from all seven studies are combined, there is clear evidence of the benefit from mammography," the statement said.

Professor Liberati suggested that the controversy over the Cochrane breast cancer review illustrated several important issues. Firstly, it showed the problems that result from poorly designed trials. "Even after 30 years we don't really have the data we needparticularly for many outcomes that are important to womenin the case of mammography screening." He considered that it would be very difficult, however, to carry out randomised trials with mammography screening now that most countries have national programmes in place.

Secondly, it showed that it was unavoidable that even when scientists tried very hard to be rigorous and methodologically sound they brought some subjectivity into their work. "Despite all the efforts we make even when we undertake rigorous systematic reviews, interpretations may differ. Different people faced with the same raw data will not necessarily come to the same conclusion," Professor Liberati said.

That was why there was a discussion within the Breast Cancer Editorial Group and between it and the authors in order to ensure that both data analysis and interpretations were as rigorous and objective as possible.

Professor Liberati concluded that the most important aim of Cochrane reviews was to help people to make well informed decisions about health care and therefore it was important to convey messages in a way that would help women to make better informed choices.

"We consider that women should be better informed about mammography screening and the results of the Cochrane review should be used for this. It is not unusual that information provided by screening programmes presents results in an overly optimistic fashion. This should be changed to reflect the reality of the situation."

"Even when we undertake rigorous systematic reviews, interpretations may differ"

(Credit: SAMUEL ASHFIELD/SCIENCE PHOTO LIBRARY) Does mammography increase mastectomies by about 20% ? Scientists cannot agree