Is routine screening for breast cancer for asymptomatic women worthwhile?

Question

Relatively recently, the United States Preventive Services Task Force made a small change in what its recommendations for routine mammograms, suggesting that they should not be performed on asymptotic women younger than 50:

But the government panel of doctors and scientists concluded that getting screened for breast cancer so early and so often is harmful, causing too many false alarms and unneeded biopsies without substantially improving women's odds of surviving the disease. "The benefits are less and the harms are greater when screening starts in the 40s," said Dr. Diana Petitti, vice chair of the panel.

The political and public response was not positive (see New York Times report here. The American Cancer Society didn't agree saying (reported here):

This is one screening test I recommend unequivocally, and would recommend to any woman 40 and over

Advice to patients in the UK has also proved controversial (eg this personal view published in the BMJ).

Some people's intuition would (wrongly) suggest that screening is all upside. Some experts seem to think differently having looked at the evidence.

So what does the evidence say? Does screening help save lives? If so, for which groups? Is breast cancer screening worthwhile for any group not known to be at elevated risk?

Answer 1

Yes, they absolutely save lives. Here's one study, relevant quote (apart from the title, "Death from breast cancer occurs predominantly in women not participating in mammographic screening"):

Therefore, 75% of breast cancer deaths occurred in the 20% of women not regularly screened, while 25% of deaths occurred in the 80% regularly screened. If the 13 year mortality rate is 15% in the 192,740 women diagnosed with breast cancer in 2009, regularly screened women will have a 13 year breast cancer mortality rate of 5%, while unscreened women will have 56% mortality rate, similar to years prior to introduction of mammography.

Because of the age addition and the discussion of the Taskforce, here's another interesting facet of the argument that the taskforce didn't take into consideration: younger women who get cancer progress through the stages of the disease much more quickly. In fact, I've covered a lot of my misgivings about the taskforce's recommendations in the question about Evidence-Based Medicine, but here's a less anecdotal approach to the report.

So, again, it's a question about weighing the value of false positives vs false negatives (or no test at all). A false negative almost certainly means that the disease will be further progressed when (and if) it's caught, while a false positive means, physically, the harm of the mammogram (radiation), ultrasound if performed (negligible physical harm as the energies are diagnostic and not therapeutic), and the biopsy itself. Minimally invasive biopsies are drastically less physically damaging than previous techniques (and I can post slides showing the difference if that previous link isn't considered sufficient).

Throw into that calculus that the disease will progress faster in younger women, and while the taskforce is right that it's up to the patient to decide to get treated (and it always is), as a general rule, the more who are screened, the more who can be treated when the disease is in its early stages. The author of this article makes a much more compelling argument than I can, so I'll quote him directly:

Considering that the main harms of screening are the necessary follow-up steps, the main question is whether the benefits of screening women aged 40-49 are really worth the effort. The USPSTF determined that they are not. In doing so however, they do not really explain how they came to this decision. [...] However, treatments for younger women are often more successful than for older women and thus a more appropriate focus would be detection of invasive cancer – for which the difference between the age groups is not as impressively significant.[...] The harms are not greater for 40-49 year old woman than for their older counterparts, they are simply more prevalent, but coupled with the benefit of saving a life with a longer potential life expectancy.

We care about women in this group precisely because they are so young and the disease progresses through them more quickly. (more on this later)

A word about the task force's statistics as well. First, 1 in 8 women will develop breast cancer throughout the course of their lifetimes, and 1 in 233 from ages 30 to 39 and 1 in 69 from ages 40 to 49 (numbers quoted by the task force in their introduction). So there is a discrepancy between the task force's numbers and the numbers reported by SEER for diagnosis rates, and I believe that that discrepancy comes from the task force's selection criteria. Further examination of the code that the task force used is pretty illustrative:

1. They have narrowed the number of studies down to 8 from all the other evidence that exists, using internally-defined criteria of 'goodness.' I don't see a mention of the DMIST trial, which shows the effectiveness of digital mammography as compared to film mammography and showed that younger women benefit from the advances of digital technology. This study is important because it shows how the field of medical imaging is advancing to the point that many of the conclusions of the task force may be obviated by people switching to digital. I believe that they didn't include DMIST because it didn't match their criteria for 'goodness' (acquired by consensus, it seems), or because it wasn't in the MEDLINE database. Which begs the question-- are there other studies that aren't in MEDLINE or that they chose to remove because of this criterion but may be considered valid by the medical community?
2. They have a set of non-justified magic numbers, such as 4 Markov chains, 1000 samples, etc. Perhaps they have a justification for these numbers (ie, they considered them to be sufficiently large to not require any further examination), but I see no particular criteria that suggest that they did that. Not saying the results will change, but that I'd like more justification for why they chose the magic numbers that they did. They used OpenBUGS, which has a prominent warning in the help file Beware: MCMC sampling can be dangerous! Might be good to have used more than just one approach to arrive at their conclusions, or to justify why they chose the particular numbers that they did in their models. Why not a million samples? Why not 500? Why 4 chains and not 8 or 2?
3. One of the studies lists "NA" for the py.int and py.cntl parameters. Why was this study included if one of the parameters that's considered necessary for the analysis isn't there?

I'll have to dig around in the code to find anything else, and if I find something, I'll post as an edit to this question. In my mind, however, I shouldn't be doing their homework and validating their statistical methodology; I think that they should have had more justification up front for this approach, especially when the consequences can be so dire.

And, to be fair, the study I used to show that breast cancer progression was faster in younger patients was done using MCMC analysis as well. Other studies have begun to find cellular explanations for the phenomenon, including the suggestion that

Interestingly, in the BALB-NeuT model, activation of the proteolytic system in breast epithelia was associated with young age and ADH [...] The activation of a specific genetic program early after sex-hormone-induced expression of the transforming oncogene HER-2 may, thus, initiate microinvasion, which can be detected by transmission electron microscopy in breast lesions as early as in the stage of ADH. [...] Evidence for their malignant origin was provided by comparative genomic hybridization.

Boiling that study down-- in rats, at least, younger age and expression of sex hormones induced the spread of malignant cancer. If that finding were to hold true in humans, it would help to explain how younger women would progress through the disease more rapidly.

In response to the recent Canadian study, the ACR has published a rebuttal that basically states that the Canadian study was deeply, deeply flawed. One quotation in particular:

In this latest BMJ paper, only 32 percent of cancers were detected by mammography alone. This extremely low number is consistent with poor quality mammography. At least two-thirds of the cancers should be detected by mammography alone (4). In an accompanying BMJ editorial, Kalager and Adami admit that "The lack of mortality benefit is also biologically plausible because the mean tumour size was 19mm in the screening group and 21mm in the control group... a 2mm difference." The documented poor quality of the NBSS mammography screening alone explains these results and should disqualify the CNBSS as a valid scientific study of modern mammography screening. Yet, the CNBSS trial was even more troubled (5).

Basically, if you take terrible images, you get nothing clinically relevant. Diagnostic quality mammograms do save lives, as shown in the Massachusetts study previously referenced, but if the people taking the images don't know what they're doing, then the screening isn't useful.

Answer 2

No, screening does not save lives even for the higher risk older groups never mind the lower risk 40-50s.

The evidence I'm going to quote below addresses the broader question of whether screening works for any age group (which was my original question). But, since the evidence clearly shows that the risk of having breast cancer grows with age (older women have higher risk) I hope that simple logic will show that this answer will be even more relevant to the younger group the question now addresses.

I also need to make a statistical clarification before quoting what thorough reviews say, as careless language is one of the reason that the stats are often misunderstood. One key problem is the casual confusion of all-cause mortality with breast-cancer mortality. It is easy to assume that mammography (which is designed to screen for breast cancer) must reduce all cause mortality because it catches some cases of breast cancer. But many studies show a reduction in breast-cancer mortality but no reduction in overall mortality. Guess which number is used by advocates of screening?

The latest Cochrane review (published in 2011) is here. The review was designed to assess:

To assess the effect of screening for breast cancer with mammography on mortality and morbidity.

Here is one part of the conclusion relevant to the statistical clarification i made above (I've highlighted the important bits and I quote at length to show the different results from different trials):

Three trials with adequate randomisation did not show a significant reduction in breast cancer mortality at 13 years (relative risk (RR) 0.90, 95% confidence interval (CI) 0.79 to 1.02); four trials with suboptimal randomisation showed a significant reduction in breast cancer mortality with an RR of 0.75 (95% CI 0.67 to 0.83). We found that breast cancer mortality was an unreliable outcome that was biased in favour of screening, mainly because of differential misclassification of cause of death. The trials with adequate randomisation did not find an effect of screening on cancer mortality, including breast cancer, after 10 years (RR 1.02, 95% CI 0.95 to 1.10) or on all-cause mortality after 13 years (RR 0.99, 95% CI 0.95 to 1.03).

The overall conclusion was (my emphasis again):

Screening is likely to reduce breast cancer mortality. As the effect was lowest in the adequately randomised trials, a reasonable estimate is a 15% reduction corresponding to an absolute risk reduction of 0.05%. Screening led to 30% overdiagnosis and overtreatment, or an absolute risk increase of 0.5%. This means that for every 2000 women invited for screening throughout 10 years, one will have her life prolonged and 10 healthy women, who would not have been diagnosed if there had not been screening, will be treated unnecessarily. Furthermore, more than 200 women will experience important psychological distress for many months because of false positive findings. It is thus not clear whether screening does more good than harm. Note that more recent large scale studies are now questioning whether the apparent reduction in cancer mortality is real, weakening even further the argument that screening is useful (see latest update for references).

The review recognizes that it is not up to the medical community to recommend blanket screening as only the informed patient can weigh up the cost against the benefit.

Two key points are worth repeating. Screening isn't costless and the cost of false positives and unnecessary treatment are high. And there is no good evidence that overall mortality is reduced by screening. The intuition expressed by many (especially those lobbyists and politicians who backed away from the original recommendations of the Preventative Services Task Force) is not backed by the hard evidence.

This answers both my original question "does screening reduce mortality" and the more specific question now asked about screening in the low risk 40-50 year olds.

---

Some additional information to clarify some misunderstandings raised in comments

---

What is a Cochrane review?

One thread of comment on the original answer accused me of selectively quoting a result to bolster my preferred point of view. But what I had quoted was not a single result but a Cochrane Review. It appears that not everyone knows what that is and why it is the opposite of a single literature result.

Cochrane Reviews are systematic reviews of the complete literature of studies relevant to a particular claim. They review the known results for quality, bias and the strength of their results and produce a systematic analysis of the weight of evidence designed to summarise as best as possible the known literature. Wikipedia says:

The Collaboration [The Cochrane Collaboration is the independent organisation that conducts Cochrane reviews] aims to provide compiled scientific evidence to aid well informed health care decisions. It conducts Systematic reviews of randomized controlled trials of health care interventions and tries to disseminate the results and conclusions derived from them.

The reviews are described on the organisation's web site as (my emphasis):

Cochrane Reviews are systematic reviews of primary research in human health care and health policy, and are internationally recognised as the highest standard in evidence-based health care. They investigate the effects of interventions for prevention, treatment and rehabilitation. They also assess the accuracy of a diagnostic test for a given condition in a specific patient group and setting. They are published online in The Cochrane Library.

So when I quote results from the Cochrane Library I am not being selective, I am trying to reference the best possible analysis of the whole literature. Moreover, the Cochrane reviews should be less subject to bias than the reviews by government bodies who face political pressure and embarrassment when they change their mind on anything.

---

Why is this a controversial issue?

It is easy to misunderstand the risk-reward pattern of medical screening. Popular understanding tends to assume (in the absence of epidemiological evidence) that screening is cost-free. It's not exactly surgery, is it? This view has been reinforced by the advice produced by governments to encourage patients to participate in screening programmes. This advice has often been criticised for not being honest about the harms of screening (the controversy in the NHS about patient advice is summarised in the BMJ here). The article argues that even the revised NHS leaflet fails in its duty because:

No mention is made of the major harm of screening—that is, unnecessary treatment of harmless lesions that would not have been identified without screening. This harm is well known and acknowledged, even among screening enthusiasts. It is in violation of guidelines and laws for informed consent not to mention this common harm, especially when screening is aimed at healthy people.

A previous analysis of advice to patients from other health systems argued:

The major harm of screening, which is overdiagnosis and subsequent overtreatment of healthy women, was not mentioned in any of 31 invitations.

The early days of screening also happened to coincide with significant improvement in breast cancer mortality which was ascribed to the benefits of screening. In principle, however, mortality can be improved by better treatments which would have been implemented even in the absence of screening. Recent evidence has tipped the balance away from screening (see this BMJ review comparing different European countries).

So there are several reasons why many people will overstate the net benefits of screening. as better evidence emerges it is likely that that evidence will provoke controversy.

---

Update February 2014

Another large long term and properly randomised study (in the BMJ) has concluded that screening doesn't even reduce mortality from breast cancer (at least when good therapies are available). In their words:

Annual mammography in women aged 40-59 does not reduce mortality from breast cancer beyond that of physical examination or usual care when adjuvant therapy for breast cancer is freely available. Overall, 22% (106/484) of screen detected invasive breast cancers were over-diagnosed, representing one over-diagnosed breast cancer for every 424 women who received mammography screening in the trial.

So screening does no good but comes at the cost of significant over treatment (women suffer surgery and other clinical interventions that do harm but create no benefit).

Answer 3

There's a group in the USA who have the job of assessing all of the scientific evidence for breast screening programs, and producing recommendations based on the best available evidence. That group is called the US Preventative Services Task Force which you may remember as the original group mentioned in the question who made the original claim.

Their recommendations, I understand, are not binding on individual doctors.

How they assess evidence

They have a difficult job, and they do it very openly. They have to assess the evidence for the "net benefit" of a treatment - and they publish details about how they do that. My summary: It is difficult to do.

Sometimes, they have insufficient evidence to make a recommendation, but they still get a lot of pressure to make one anyway. They openly describe the dilemma and their procedures to avoid making bad recommendations on poor evidence. My summary: It is difficult to do. They grade the evidence behind their recommendations

Breast Cancer Screening Recommendations

They make a number of number of recommendations about screening for breast cancer including:

• The USPSTF recommends biennial screening mammography for women aged 50 to 74 years.
• The decision to start regular, biennial screening mammography before the age of 50 years should be an individual one and take patient context into account, including the patient's values regarding specific benefits and harms.

They support this with a report that does a review of the evidence: Screening for Breast Cancer, An Update for the U.S. Preventive Services Task Force, Heidi D. Nelson, Kari Tyne, Arpana Naik, Christina Bougatsos, Benjamin K. Chan, and Linda Humphrey, November 2009

In particular, they work out the number needed to treat (or they have called it NNI; number needed to invite) in order to save one life.

Extracting from their Table 1, which pools data from 16 trials.

Age   NNI to Prevent 1 Breast Cancer Death
39-49 1904
50-59 1339
60-69 377

Conclusion

Whether it is worth screening 1904 people to get the opportunity to treat enough people to save one life is an economic and ethical decision - presumably it should be compared to other competing services that could be offered to save more lives for the same price.

Update:

A recent study published in the British Medical Journal apparently updates the results to make mammograms seem even less helpful for 40-60 year old Canadian women, and expect recommendations will eventually be updated based on all recent evidence.