last update of this page: 29 July 2005
click here for abstract of article preceding this additional discussion
screening for breast cancer under the age of fifty
When the preceding article was published in the Journal of the National Cancer Institute, it got accompanied by two editorials and later there was a letter to the editor about the article.(Forrest and Alexander 1995; Smith 1995; Tabr 1996) Though only a minor part of the article was about screening under age 50, all three comments concentrated on that issue.
At the time of publication of the article, the overview of the Swedish randomised trials into breast cancer screening showed a substantially, though not statistically significant, lower mortality reduction for women under age 50 at entry compared to women aged 50-69 at entry.(Nystrm et al. 1993) Apart from that observation there was also the outcome from the Canadian NBSS that was quite unfavourable for screening women in their forties.(Miller et al. 1992)
A problem in the interpretation of the published trial outcomes is that women who are classified as under age 50 at entry to the trial, may have benefited from screening at ages of 50 and higher. Our analysis assumed that that the efficacy of screening suddenly changes at age 50 and that efficacy by age is constant within each of the age ranges of under 50 and over 50. Using these assumptions we showed that the observed screening benefit of women under age 50 at entry to the trial could be attributed for the largest part to screening from age 50 onward. We did not conclude that screening under age 50 was worthless, but that this called for further investigation. The mentioned letter to the editor was a reply to this question and showed that in the Two Counties Study, there was evidence against screenings applied over age 50 being more effective than those applied under age 50 among the age group of women under age 50 at entry.(Tabr 1996) Together with the relatively low observed efficacy of screening under age 50 at entry this implies that the screenings applied over age 50 in that age group were remarkably little effective. But because the numbers involved are so small, it is not prudent to draw any conclusions from that.
In the mean time outcomes of the Swedish trials for a longer follow-up have been published, showing a much smaller difference in mortality reduction by age at entry. But the Canadian NBSS study in women under age 50 still shows an unfavourable effect on mortality from screening. The issue is continuously under vigorous discussion (see for instance (Baines and Miller 1997; Chang et al. 1997; Cox 1997; Feig and Hendrick 1997; Fletcher 1997; Harris 1997; Kopans 1997c)). Some proponents of screening younger women appear to be biased in their arguments. For instance the NBSS trial has been under severe attack, while other trials are also liable to more or less similar criticism.(Baines 1994; Glasziou and Irwig 1997; Gtzsche and Olsen 2000; Kopans 1997a; Kopans 1997b)
current evidence for efficacy of breast cancer screening under age 50
Except for the Canadian NBSS trial, none of the trials were designed to show a significant mortality reduction for women younger than 50 years. They were designed to show an effect in the total trial population. The single fact that they did not show a statistically significant mortality reduction for separate age groups should not be interpreted as a lack of efficacy. However, it is possible that after the start of a trial new ideas develop on questions that need to be answered based on empirical evidence. Clearly, the question whether screening under age 50 is efficacious, is such a question. During the course of the HIP trial it was observed that the mortality reduction was lower among women under age 50 at entry than for the other participants.(Shapiro et al. 1974) This was an important spark off for the discussion, with as fuel, the notion of important biological changes occurring around the age of 50, or rather, the menopause. After longer follow up of the HIP trial, the difference in mortality reduction almost disappeared.(Shapiro et al. 1985) And nobody has substantiated why any particular biological change that occurs around the age of 50 would cause a large difference in efficacy of screening. Dense breasts and faster growing tumours indeed tend to cause a somewhat lower efficacy, but the screening of women under age 50 still detects a very substantial proportion of the cancers that surface in the study and at a practically equally favourable stage as in older women. Therefore, dense breasts and faster growing tumours alone cannot explain a large difference in mortality reduction among women over and under age 50.
The ongoing discussion resulted in two trials that focussing on screening women under age 50. The first is the Canadian NBSS that started in 1980 and shows a (statistically non-significant) higher mortality from breast cancer in the screening group than in the control group.(Miller et al. 1992; Miller et al. 1997) After this result, of course there was sufficient reason for concern about the efficacy of screening under age 50. Therefore a second trial was initiated in the U.K., which is so recent that no results are available yet.
Given the results from the collective randomised trials on breast cancer screening and the discussion about biological influences, it appears that screening under age 50 can reduce breast cancer mortality. But there remains a compelling question why the Canadian trial that was designed to show an effect among women in that age group does not show a reduction in breast cancer mortality.
Criticism on the trial has been convincingly countered.(Baines 1994) What appears to be lacking is a detailed comparison of all intermediate outcomes from this trial with those from other studies. Such a comparison may be facilitated by using a model such as Miscan that can show the influence of local circumstances on the observed results. Insight in the circumstances that determine whether one trial shows a mortality reduction, while another does not, is crucial for any future breast screening. As also chapter 12 shows, our understanding of the natural history of breast cancer in relation to screening is still rather incomplete. A more detailed study of the results from the Canadian and other trials may enhance our understanding of breast cancer screening that in turn may indicate possibilities for improving screening efficacy.
choosing the lower age limit of mass screening for breast cancer
Sufficient evidence for mortality reduction due to screening does not imply that the balance of all favourable and unfavourable effects (including costs) is sufficiently positive. And the best estimate of the balance may be sufficiently positive, but that estimate may also be so uncertain that a decision based on it would not be called 'evidence based'.
Both issues have been under discussion, but the emphasis was thus far on the sufficiency of the evidence. Now that the evidence in favour of screening women at ages under 50 has been increasing over the past several years, there is more need for good estimates on the balance of favourable and unfavourable health effects as well as that of effectiveness and costs. What has been published so far on the issue of balancing can at best be called 'tentative'.(Boer et al. 1995b; Salzmann et al. 1997) Although very different estimates of the balance between favourable and unfavourable effects have been given, no one concluded that there are inevitable unfavourable effects that would preclude screening of women in their forties. Also on cost-effectiveness estimates differed widely, and here using one or the other estimate would make the difference between deciding for or against screening women in their forties. At some points the estimates differ on very fundamental issues. For instance, our research group maintains that in general the law of diminishing returns applies to screening frequency, thus more frequent screening is less cost-effective. Other research groups estimate a better cost-effectiveness for yearly screening than for two-yearly screening.(Chen et al. 1997; Michaelson et al. 1999)
There is clearly a need to gather all available information in order to make a more precise estimate of the balance of effectiveness and costs of breast cancer screening of women under the age of 50, and need for detailed comparison of the different models that have been used to estimate favourable and unfavourable effects, and costs.
There appears to be general agreement only on some issues. Breast cancer is also under the age of 50 an important health problem. Though incidence and mortality from the disease are substantially lower than over the age of 50, the number of life years lost per death from the disease is higher. And due to faster growth of tumours at lower ages, and possibly also lower sensitivity of the screening test, reaching the same percentage of mortality reduction needs more frequent screening. That limited agreement is regretfully not a very good basis for decision making.
furthering the method of estimating improvement of prognosis due to screening
The study described in this chapter only accounted for a limited number of aspects that influence screening efficacy. The model was only adjusted for the aspects of organisation of the screening project and attendance of the invited women. When the study was performed, we assumed that the epidemiology and screening performance in the Swedish trial were similar to those in the Netherlands. Later, a more detailed comparison of the Dutch experience with the Two Counties study showed that there are substantial differences.(Fracheboud et al. 1997) Particularly, the Two Counties study appeared to pick up about twice as many small (up to 1 cm) invasive tumours than the Dutch screening programme. In order for the model to predict the same mortality reduction, but with a kind of screening that finds more, particularly smaller, cancers, the probability for cure due to detection at screening should be lower. Thus, with our current understanding, the estimates presented in the article overestimate the efficacy of screening, therefore the prognoses for effectiveness of screening that were made with this estimate are too optimistic. And more importantly, there is reason to assume that the Dutch breast cancer screening programme can possibly perform better. Currently a study for optimisation of the Dutch programme is on its way.
It would be useful to repeat our study with taking screening performance into account so that a more accurate estimate of improvement of prognosis due to detection at screening will be available.