I missed Journal Club today because my house was falling down around me. That’s not a metaphor. Our roof is leaking and chunks of plaster ceiling are falling around me. Times like this, I wish we were renting…

Anyway, I was bummed about missing Journal Club. The topic of the day is the recent mammography guidelines, and I was ready to pick a fight. I have been driven to distraction by the brouhaha over the new USPSTF guidelines. In particular, the politicization of these guidelines is infuriating to me. So since I missed my chance to vent at Journal Club, I am going to vent here. If the post ends mid-sentence, it is likely due to a plaster projectile.

To me, the first question is: why is everybody so surprised by these guidelines? I think that to show surprise about these changes is to announce that you haven’t kept up with the literature even a little bit. When last we left off, at the time of the USPSTF’s 2002 guidelines, the Swedish trials were the talk of the town. Looking specifically at the 40-49 age group, mammography was not beneficial. Largely based on these trials, the International Agency for Research on Cancer (IARC) concluded, in 2002, that there was sufficient evidence for screening women age 50-69 with mammography, but there was limited evidence for the efficacy of screening women age 40-49 by mammography. Based on the Swedish trials, mammography had a rate ratio of 0.81, with a 95% CI of 0.65-1.01. In other words, there was a non-significant reduction in breast cancer mortality of 19%.

Let me break for a bit to discuss statistics. This will be quick. I was an English major for good reason. But I want to be clear about one thing first: when I say “non-significant reduction in breast cancer mortality of 19%,” I am not saying that reducing breast cancer by 19% is not “significant” in the meaning of “important.” I mean it is not significant in the mathematical sense. Note the confidence interval: 0.65-1.01. A confidence interval (CI) are those values that, if we repeated the measurements over and over, we’d expect 95% of the outcomes to encompass. If all of those values favor a particular outcome, then we can be reasonably (but not absolutely) certain that the outcome is real and not a result of chance. In the case of the Swedish trials, the 95% CI includes values that favor no screening over mammography. At best, we can say there was “a trend” toward better outcomes with mammography, but we cannot be reasonably certain by the typical definition of “reasonable” used in medicine–that is, 95%.

Ok, so I probably butchered that explanation. Back to the stuff I understand. Since 2002, there have been 2 articles addressing mammography in patients age 40-49. The first, out of the UK, followed 160,921 women. The design was interesting: women turning 40 were randomized to annual mammography screening for 7 years or to no screening. They then had no screening for 3 years, and then were enrolled in the UK 3-yearly screening program. In this study, the point rate ratio was 0.83, with a 95% CI of 0.66-1.04. Again, a non-significant reduction of, in this case, 17%. Very similar to the Swedish studies (and suggesting that maybe there is something to this statistics stuff). The USPSTF also considered a follow-up to one of the Swedish studies, which found a point rate ratio of 0.79, with a 95% CI of 0.58-1.08. The p-value, for those who prefer it to CI, was 0.14. Not statistically significant.

What does all this look like in meta-analysis form (from which the USPSTF made their recommendations)?

Note that not one trial meets statistical significance. Put them all together and you barely get a statistically significant result. Conclusive? I don’t think so.

The truth is that none of these studies really answers the question. They suggest that maybe mammography might reduce mortality in women age 40-49 by something around 17-20%. But if we are honest, we have to admit that maybe mammography does nothing in this age group. Want to muddy things a little more? Take the Canadian National Breast Screening Study 2 (CNBSS 2). Unlike the trials mentioned above, it did not compare mammography versus no screening; it compared mammography plus regular clinical breast exams versus physical exams alone. 39,405 women ages 50-59 were enrolled. There were 107 deaths in the mammography group versus 105 deaths in the physical exam-only group. Null hypothesis–check. The risk ratio was 1.02 for mammography with a 95% CI of 0.78-1.33. Not surprisingly, mammography had many more false positives and identified many more small (and likely clinically insignificant) tumors.

So does mammography even add benefit to clinical breast exam (CBE) in the 50-59 age group? Maybe not. Does that mean mammography is equivalent to CBE? Not necessarily. As noted in a published response to the CNBSS, if mammography and CBE were equivalent, you would expect some added benefit of doing both. There is none. The conclusion of the commentator is that this suggests they are both ineffective.

Head hurt yet? I’m not done.

Suppose you aren’t willing to concede that mammography is a lousy screening tool. And after all, you say, what’s the harm? Well, overdiagnosis is the harm. First, let’s define the term. Overdiagnosis is the identification of a cancer when that identification does not affect the prognosis. (That may or may not be the “official” definition, but it is the spirit of it anyway.) Suppose cancer is found on a screening test. What does that mean? Well, if it’s a good test, it means you found a cancer that is curable, but would not have been curable if you didn’t do the screening test. But it can also mean:

1) The cancer is detected earlier on screening, but it would have been curable even if detected later clinically (i.e., mammography picks up the tumor earlier than CBE, but it is curable either way)
2) The cancer is detected earlier on screening, but it’s too late and can’t be cured anyway
3) The cancer is detected earlier on screening, but it’s slow growing or destined to regress or what-have-you, and the patient is destined to die of something else before dying of the cancer

You can argue that #1 and #2 fall into the “no harm, no foul” category. I’d argue that they in fact represent a significant and unnecessary expense. But in either case, screening does not help. And think about this: as our treatment options for breast cancer improve, unless our screening tests improves proportionately (it hasn’t), it will become less an less beneficial. If we ever get to a point of being able to cure all breast cancer at any stage, then screening becomes useless. We’re not there, but we are improving quickly. But let’s focus on cancers detected on screening that fall into category #3. These will likely result in some degree of harm. Imagine a tumor that is found on mammography, biopsied, and found to be invasive. Best case scenario, there is a lot of terror, an uncomplicated lumpectomy, lots of further studies, and more terror. Let’s put some numbers on this, and for those numbers, let’s turn to the Cochrane Reviews. They published a review of the risks and benefits of mammography this year (2009). They estimate an overdiagnosis rate of 30%. They also note that screening programs increase the mastectomy rate by 20%. Lumpectomies increased by 71% in the Netherlands when screening was introduced. These 2 facts alone suggest that there are many mastectomies and many more lumpectomies being done that would not be needed otherwise. Do most of them save lives? Almost certainly. But consider a couple things. You could estimate that, if the overdiagnosis rate is 30%, then 30% of those mastectomies are unnecessary. I admit that does not necessarily follow. But surely some of them are unnecessary. And ask this question: if mammography is supposed to pick up cancers earlier, why would mastectomy rates go up? Even though the percentage of cases of carcinoma in situ treated in the US by mastectomy declined from 71% in 1983 to 40% in 1993, the total number of mastectomies increased almost 3-fold. In the UK, after screening programs were started, mastectomies for invasive cancer increased by 36% and for carcinoma in situ they increased by 422% over the period from 1990 to 2001! Now, say that without mammography, some women might be diagnosed at a late stage and no mastectomy is done because they are too far gone. But most are identified clinically, eventually, and would get a mastectomy. How can mammography increase mastectomy rates by such a profound degree unless some (many?) of them are due to overdiagnosis? And of course, overdiagnosis is going to be much more common among women in their 40’s than those in their late 50’s.

We fixate on physical harms, but I’d argue that psychological harms are essential to consider as well. In the same Cochrane review, 1/3 of women with false-positive tests were still undergoing testing and had not been declared disease-free at 6 months. 6 months! That is a long time to think you might have a fatal disease. Women who had an abnormal mammogram but were then declared cancer-free after biopsy or other studies were twice as likely to suffer psychological consequences 3 years later as those who had a normal mammogram. No physical harm, but the psychological harm lasts.

Overdiagnosis is not talked about very much. In journal articles, it is often downplayed, and more so when the author has a stake in screening programs. But it recently hit the mainstream press, and there was a bit of an uproar. “Why don’t doctors talk about this?” became a frequent question on comment threads. Well, kudos for the USPSTF for considering overdiagnosis in their guidelines.

Okay, so there is the data. Do you find the question of mammography for women in their 40’s to be straight forward? Me neither. Now let’s look at the actual words used by the USPSTF:

“The USPSTF recommends against routine screening mammography in women aged 40 to 49 years. The decision to start regular, biennial screening mammography before the age of 50 years should be an individual one and take into account patient context, including the patient’s values regarding specific benefits and harms.” [emphasis mine]

How is this controversial? The data to support universal screening just isn’t there. The test just isn’t that good. You need a certain pre-test probability before it is of benefit, and your 40 year-old non-smoking woman with no family history of breast, prostate, or ovarian cancer just isn’t high enough risk. When the signal is too low, all you’ll get is noise. But should you screen the smoker whose brother has prostate cancer? Of course, and the guidelines agree. To say that the benefits of universal screening of women ages 40-49 clearly outweighs the risks is to ignore the evidence. And when the benefits are likely small, and the risks are real, it is reckless to suggest anything other than making it an individualized decision.