Mammograms are effective, based on new look at stats

A recent, highly controversial series of papers published by two researchers at the Nordic Cochrane Centre in Copenhagen, Peter Gotzsche and Ole Olsen, concluded that mammography does not save lives and instead exposes women to unnecessary diagnostic and surgical procedures.

Adding a fresh perspective to the debate, experts in statistical methodology James M. Robins (Harvard School of Public Health), David A. Freedman (University of California, Berkeley), and Diana B. Petitti (Kaiser Permanente, Southern California), examined the Gotzsche-Olsen critique and found it to be severely flawed. The statistics team concluded that breast cancer screening does save lives.

What is known about the effectiveness of mammography comes from a small number of clinical trials conducted over the past four decades. Gotzsche and Olsen analyzed the results of these existing trials, but in drawing their conclusions, chose to exclude studies that found screening to be effective – for example, the Health Insurance Plan (HIP) trial in New York and the Two-County trial in Sweden – on the grounds that the positive trials were of poor quality. Gotzsche and Olsen did include in their analysis the Canadian National Breast Screening Study (CNBSS) – which found mammography not to be effective – arguing that CNBSS was of adequate quality.

In their paper “On the Efficacy of Screening for Breast Cancer,” Robins, Freedman, and Petitti find that the Gotzsche-Olsen judgments of trial quality were based on misreadings of the data and the literature. For example, Gotzsche and Olsen claim that HIP’s exclusion of high-risk women from the screening group biased the study. The statistics team show this claim is incorrect. Gotzsche and Olsen say there are inconsistencies in the way the Two-County trial reported its data. But Robins, Freedman, and Petitti show that the “inconsistencies” reflect a misunderstanding of the reporting rules, which were explained in the papers cited by Gotzsche and Olsen.

The three researchers discuss anomalies in the CNBSS data that suggest it is not superior to the two studies excluded by Gotzsche and Olsen. For example, in the CNBSS study, among women aged 40-49, a statistically implausible fraction of those with advanced breast cancer were assigned to the screening group.

The Gotzsche-Olsen paper originally appeared in the Jan. 8, 2000, issue of the British journal Lancet.

In a commentary in the current issue of IJE, Peter Gotzsche defends his prior position.

In another commentary, Anthony Miller (University of Toronto) defends CNBSS, suggesting that the excess of advanced cancers reflects more careful diagnostic work in the screening group.

Donald Berry (University of Texas) concludes there are costs associated with the benefits from screening.

Michael Baum (University College, London) argues against screening, since clinical trials of mammography do not meet contemporary reporting standards that he helped to develop.

The issue of IJE also includes a final rejoinder by Freedman, Petitti, and Robins that addresses the commentaries. Freedman et al. conclude that the data make a compelling case. Addressing the overall controversy, they write: “Clinical trials of mammography have led to substantial advances in understanding breast cancer, and a substantial reduction in mortality from this disease. It is time to move on, although some questions may remain.”