Last Wednesday, right before the four-day Thanksgiving holiday weekend, as I was far more interested in preparing to have family over the next day than in what was going on in the medical news or the blogs, the results of a most fascinating study hit the news. In Medscape, the title of the news report was ; on WebMD, the story ran under the title ; on Medical News Today, by Drs. Per-Henrik Zahl, Jan Maehlen, and H. Gilbert Welch. Not surprisingly, the study found its way out of the medical news and into as well, given how provocative the findings seemingly are. From the on this study:
A mammography study from Norway has come up with the controversial proposal that one fifth of breast cancer detected on screening may spontaneously regress. But there is no easy way to verify whether this is the case, say experts.
The study was published in the November 24 issue of the Archives of Internal Medicine. It found that the cumulative incidence of invasive breast cancer in a cohort of women, aged 50 to 64 years, who received 3 mammograms over 6 years was 22% higher than in a control group of age-matched women who received only 1 mammogram at the end of a 6-year period.
In their review of possible explanations for the difference in the breast cancer rates between the 2 groups — which had similar risk factors for breast cancer — the researchers write that the “natural course for some screen-detected breast cancers may be to spontaneously regress.”
“I anticipate that many clinicians will react negatively to the possibility of spontaneous regression, said coauthor Jan Maehlen, MD, PhD, professor of pathology at the Ulleval University Hospital, in Oslo, Norway, in an interview with Medscape Oncology.
Before I discuss the itself, let me briefly discuss why clinicians may have a bit of a problem with the implications of this study, if they accurately reflect the biology of breast cancer.
The essential problem with this study is not the results per se. Rather, it is that, even if it is true that as many as 22% of breast cancers regress spontaneously, we currently have absolutely no way of distinguishing this minority of tumors from the vast majority that will progress without treatment. Consequently, although this study’s findings are provocative and might even be true (although I do think that Dr. Maehlen oversells and exaggerates his results a bit–more on that later), to the average primary care physician or to the average breast surgeon (like me), this study provides no guidance on what to do. If we see an abnormality on mammography and ultrasound that looks suspicious, we are still obligated to work it up and, if necessary, biopsy it. We are still obligated to excise it if it turns out to be a breast cancer. In other words, because we have no way of distinguishing the “good” cancers from the “bad,” so to speak, we are morally obligated to treat both the same because this study provides zero guidance about which tumors might safely be observed to see if they regress. Worse, it induces doubt in patients over whether or not they truly need a biopsy, which, until we find some way of identifying the tumors that won’t progress, they do. Add to that the fact that failure to diagnose breast cancer consistently ranks either number one or in the top three among causes for malpractice lawsuits, and you can see why primary care physicians might not appreciate this study.
Of course, here’s one area where Dr. Maehlen overstates the negative reaction a bit. I suspect that most surgeons, medical oncologists, and radiation oncologists who actually specialize in breast cancer will be skeptical of the magnitude of the rate of spontaneous regression reported by Dr. Maehlen but unlikely to reject out of hand that such regression may occur more often than has been documented. The reasons for this skepticism are buried within the methodology of the study itself, particularly its underlying assumptions. To some extent, I fear that Dr. Maehlen is confusing overdiagnosis by mammography with spontaneous regression in that he seems to be assuming that all the “overdiagnosed” tumors must have spontaneously regressed, an assumption that is not tenable.
To understand why I’m not particularly surprised that mammography might pick up breast cancers that ultimately spontaneously regress, I refer SBM readers to a series of posts that I wrote a few months back entitled The early detection of cancer and improved survival: More complicated than most people think and Early detection of cancer, part 2: Breast cancer and MRI. In particular, concentrate on the concepts of lead time bias, length bias, and overdiagnosis. Also note that in autopsy series, among women from 40-50 years old, 39% had detectable foci of breast cancer or ductal carcinoma in situ, a rate far higher than the overall percentage of women who will ever be diagnosed with breast cancer somtime during their lifetime. These data alone suggest that there are quite a number of cancers that either regress or never progress beyond a small tumor.
Now, on to .
A study of this sort could only have been accomplished in a nation with a government-run health care system with extensive databases linking information from every hospital in the nation, and that’s true of this study, which was carried out in Norway. Indeed, the numbers of women in each study are staggering: 109,784 in the control group and 119,472 in the “screened” group. So here’s what Drs. Zahl, Maehlen, and his colleagues did.
In essence, they took advantage of a change in the routine screening protocol in Norway. In 1995, the Ministry of Health and Care Services instituted the Norwegian Breast Cancer Screening Program, in which women between the ages of 50 and 64 were invited to undergo a two-view biennial screening program. The women in this study who make up the “screened” group thus underwent a first round of screening mammography in 1996-1997, a second round in 1998-1999, and a third round in 2000-2001. The control group consisted of women (age range 50-69 years) who would have been invited to participate in the screening program between 1992 to 1997, had the Norwegian Breast Cancer Screening Program existed then. (If you’re interested in the details, the is available free online.) These women only underwent one screening mammogram at the end of their six year period, compared to the three mammograms the women in the screened group underwent. One thing that’s impossible for me not to mention here is that this is thus a retrospective study and therefore prone to all the problems to which retrospective studies are prone. However, in the case of this study, the screened and control groups are about as well matched as can be done, but never forget that there could be biases inherent in the retrospective study design that can be very hard to ferret out.
Having settled on the two groups, Dr. Zahl then looked at how many diagnoses of breast cancer were made in each group, and they excluded non-invasive (or preinvasive) breast cancer, more commonly referred to as ductal carcinoma in situ (DCIS). Instead, they concentrated only on invasive carcinoma. The screened group had 22% more diagnoses of invasive carcinoma than the group that only received one screening test at the end of a six year period. The authors make a fairly strong argument that the additional invasive cancers in the screened group are unlikely to be explained by better cancer ascertainment with time (ascertainment was excellent for the entire time period covered by the study); increasing sensitivity of mammography (estimated sensitivity and specificity were not significantly different at the beginning of the study time period compared to the end); an increase in the actual incidence of breast cancer (highly unlikely over a six year period, and an increase in the underlying incidence of breast cancer could only explain at most 4% of the observed difference); or hormone replacement therapy, which does confer an increased risk of breast cancer.
Instead, the authors’ favored explanation is regression of some tumors. This is where, I believe, their argument is least convincing, at least to me: They disregard the possibility of stable tumors that do not progress. True, they do posit what they call an “extreme” case, where 50% of breast tumors are stable and are picked up with 50% sensitivity on each round of screening and show that such a case can at most account for the 22% excess number of breast cancers in the screened group, the implication being that their case is so “extreme” that it makes an explanation of stable cancers that do not progress seem very unlikely. However, what they neglect is the case where far more than 50% of breast tumors are stable, which may be the case, at least if the autopsy series I reference above is correct. In that case, it would be quite easy to postulate an explanation where stable cancers, coupled with increasing quality of mammography, could result in an apparent 22% excess number of cancers diagnosed in the screened population. However, I’m not so convinced that that “extreme” case is all that extreme. Even so, I am also not dismissing Dr. Zahl’s contention that 22% of mammographically-detected breast cancers spontaneously regress, either. What I am arguing is that it is doubtful that all 22% of tumors detected in the screened group must have regressed.
There are also methodological problems with Dr. Zahl’s trial to contend with, as an by Drs. Robert Kaplan and Franz Porzsolt points out:
Because the study by Zahl et al2 was not a randomized clinical trial, methodological concerns may lead to other explanations for these findings. One possibility is that the larger number of mammograms in the multiple screen group could account for the differences. We know, for example, that between 20% and 30% of visible lesions are overlooked. Studies suggest that detection rates are higher if films are reviewed by multiple radiologists. One study showed that among 108 radiologists, there was a range of 40% in the sensitively for detecting breast lesions.12 Women who have had 3 consecutive mammograms may be up to 20% more likely to have a positive result on 1 of the 3 tests. However, the study by Zahl et al2 included 1 additional screen in each group. If the multiple vs single screen explanation is correct, we should have seen a narrowing of the difference between the 2 groups following the additional screen. That did not occur. Furthermore, if we assume that tumors missed in early screens continue to progress, they should have showed up in the tumor registry. They did not. The design of the study has many imperfections, but we should not overlook its strengths. It was population based, it had very high participation, and the outcomes were well documented in an independent tumor registry. Considering the strengths and weaknesses of the methodology, the findings should not be dismissed.
Another reason not to disregard the findings of Zahl et al is that they are consistent with several observations that have troubled investigators for years. For example, randomized clinical trials rarely show the benefits of screening, particularly for women younger than 50 years. The Canadian National Breast Screening Trial hinted of spontaneous regression, and the Wisconsin Breast Cancer Epidemiology Simulation Model required a concept like spontaneous regression to account for observational data. Although the findings of Zahl et al seem counterintuitive, the spontaneous regression hypothesis is difficult to rule out.
Indeed it is, although Drs. Kaplan and Porzsolt overstate the difficulty in finding a survival advantage due to mammographic screening, which in women over 50 is pretty clear, less so in women 40-50 years of age. In any case, the only sure-fire way to rule the hypothesis in or out would be a randomized clinical trial because any such trial designed to answer the question of whether some breast cancers undergo spontaneous regression would require that a significant number of women with diagnosed breast cancer go without treatment. Any such trial would clearly be unethical, because even if this study is correct and roughly one in five breast cancers detected on mammography will regress, that still means that four out of five cancers will progress, meaning that withholding treatment from women with breast cancers detected on mammography would have an 80% chance of resulting in harm in terms of delaying therapy. This means that the real question is not whether some percentage of breast cancer patients didn’t actually require therapy. It’s very likely that some small percentage don’t, although I tend to think that Zahn et al overestimate the level of overdiagnosis, given multiple previous studies suggesting it to be considerably lower. The question is what the risk-benefit ratio is. It also has to do with patient choice and what her tolerance for risk is. After all, if we were to come up with a test that could predict which tumors were likely to regress (or at least not to progress), there would still be a chance of the test being wrong and the tumor’s progressing. Watchful waiting as a management strategy for such tumors may not sit well with patients, many of whom are so terrified of breast cancer that they would rather undergo invasive procedures to biopsy mammographic abnormalities, even when the likelihood of cancer is low. Some will even request bilateral mastectomies for small tumors that could easily be treated with lumpectomy. Indeed, on occasion, I’ve had to try to talk patients out of such overkill. Such situations are where science-based medicine sometimes conflicts with patient desires.
If Dr. Zahn’s results hold up, though, there is one rather interesting implication, and that is with regards to “alternative” medicine modalities that, it is claimed, treat breast cancer. Very early on in the history of this blog, I wrote a post that described how it can appear that “alternative medicine” that is completely ineffective can nonetheless lead to the appearance of efficacy and thus to “testimonials” by some women about how they were “cured” by “alternative” medicine. In essence, many breast cancers are , which can consist of the excisional biopsy that diagnosed the cancer in the first place, if the biopsy completely removed the cancer. The addition of radiation therapy and chemotherapy are “icing on the cake,” so to speak, in that they decrease the risk of recurrence but are not the primary cure. In cases where patients undergo surgical excision of their breast cancers but eschew chemotherapy and/or radiation in favor of “alternative” medicine, they almost always attribute their survival to the alternative medicine rather than the surgery alone. After all, surgery is a nasty and brutally “primitive” and “disfiguring” business, and in the world of alt-med it can’t possibly be of value, can it?
Now add to this a facet of breast cancer biology in which 20% or more of mammographically detected breast cancers spontaneously regress. If true, this observation could explain an additional set of breast cancer “testimonials,” in which the woman giving the testimonial had a needle biopsy but never underwent curative surgery. Indeed, if Dr. Maehlen is correct and 20% of mammographically detected breast cancers spontaneously regress, that would be a major addition to the pool of women who gambled with “alternative” medicine to treat their breast cancer and ended up being one of the lucky one in five. Indeed, if Dr. Mahlen’s results hold up, he may well have done skeptical physicians an unwitting service by providing another potential explanation for the ever-ubiquitous breast cancer “alt-med” testimonial, his exceedingly unfortunate use of the very bad choice of the term “pseudocancers” to describe the additional cancers picked up by mammography notwithstanding.
Finally, if there’s one thing that this study highlights more than anything else, it’s just how little we know about the natural history of small mammographically detected breast cancers, as pointed out in the accompanying editorial:
Perhaps the most important concern raised by the study by Zahl et al is that it highlights how surprisingly little we know about what happens to untreated patients with breast cancer. In addition to not knowing the natural history of breast cancer for younger women, we also know very little about the natural history for older women. We know from autopsy studies that a significant number of women die without knowing that they had breast cancer (including ductal carcinoma in situ). The observation of a historical trend toward improved survival does not necessarily support the benefit of treatment. Improved survival is equally well explained by lead-time bias. Widely cited studies describing the dire consequences of refusing treatment may not be entirely credible. In the best known of these, Verkooijen and colleagues compared women who refused surgery with those who had accepted surgery. In their analysis of 5339 women in the Geneva cancer registry, Verkooijen et al found 1.3% who refused surgery. Among these, approximately half received no treatment at all. Survival among the women who refused treatment was lower than it was for women who accepted treatment. However, the groups were in no way comparable. Those who refused were 10 years older, more likely to be of lower income, single, and at more advanced stages of disease. Larsen and Rose reviewed the literature going back to the first part of the last century and concluded that spontaneous remission of breast cancer occurs but only rarely.
I think that Drs. Kaplan and Porzsolt overstate their case a bit, namely because we do have about the natural history of untreated breast cancer. Granted, these tumors were not mammographically detected, but it’s just not correct to say that we do not have a reasonable idea of what the natural history of untreated breast cancer is likely to be. Drs. Kaplan and Porzsolt would have made a better point if they had said that our knowledge of untreated breast cancer that is not palpable and is detected by mammography alone is less than what we would like it to be. Also, one must keep in mind that this study is definitely an outlier. It’s not as though this issue hasn’t been studied before. Indeed, I discussed the extreme variability of breast cancer growth based on another study out of Norway. Come to think of it, this uses much of the very same data that Dr. Mahlen used, namely women screened in Norway during the period of 1995 to 2001, and it used a far more sophisticated model and design compared to those of Dr. Mahlen’s study, not to mention much larger numbers. I wonder why no one mentioned this study to Dr. Mahler in the context of his trial or why he didn’t cite the study in his manuscript. It’s highly relevant, especially since it looks at a dataset that certainly has significant overlap with the dataset studied by Zahn et al. Surprisingly, the reaction of the American Cancer Society was even more :
The American Cancer Society took issue with the thinking that cancer may regress on its own. “The conclusion that more than 1 in 5 invasive breast cancers is destined to regress without incident if not detected by mammography [the 22% figure cited in the study] is nothing more than an overreaching leap in logic,” Robert A. Smith, PhD, director of cancer screening for the American Cancer Society, says in a prepared statement.
Other studies have found that “overdiagnosis” — not the same as regression — probably occurs in less than 5% of all screen-detected cancer cases if it exists at all, Smith says.
He says the benefits of regular mammograms far outweigh any limitations, such as false-positive results and “possibly a small rate of overdiagnosis.”
I would respectfully disagree that the conclusion of this study is an “overreaching leap in logic.” Although this is only one study and, I suspect, it probably overestimates the rate of overdiagnosis due to mammography, it does make an at least plausible case that spontaneous regression of breast cancer may be more common than than previously thought. Reactions like Dr. Smith’s play right into the hands of who are–shall we say?–less than friendly to science-based medicine who claim that the medical profession views this study as a profound threat.
In any case, the bottom line is that the matter of screening mammography is far more complex an issue than it is often portrayed to be. It’s a matter of balancing the benefits of detecting cancer at an earlier stage versus the risk of overdiagnosis. Moreover, it is quite possible that a significant proportion of breast cancers detected by mammography (although, I hasten to add, not cancers detected as palpable masses) may indeed regress. The problem is that we do not yet know enough to identify which tumors will progress and which may not. Even under the most optimistic scenario, in which Dr. Mahlen’s study is confirmed, four out of five breast cancers will progress and become life threatening. Those are not odds that I would want to mess with, unless someone can develop a predictive test that allows me to identify cancers that it’s safe to subject to watchful waiting to see if they regress. Unfortunately, there ain’t no such beast, but I’m optimistic that there may well be before my career comes to an end.
Finally, it should be remembered that mammographic screening in the U.S. is different than it is in much of Europe. We begin screening at age 40, whereas few European countries start mammographic screening before age 50 except in women with factors that put them at high risk. We tend to screen every year, whereas in Europe, most countries screen every other year, which is the sort of screening regimen institute in Norway in 1995 around which this study was built. Such a screening regimen would be more likely to be able to detect regression were it not for the fact that any suspicious abnormality is in general biopsied and, if malignant, excised. in addition, as mentioned before, doing a randomized study would be unethical, because it would necessitate some women with biopsy-proven breast cancer being left untreated. The complexities involved in determining the efficacy of any mass screening program thus provide an excellent example of just how difficult getting answers in science-based medicine can be. Between the ethical problems that make a gold standard trial impossible, the variable biology of breast cancer, and the imperfections in screening tests, science-based medicine is hard. It’s always risk/benefit analyses and shades of gray. Perhaps that’s why quackery is so attractive. There isn’t all that messy dealing with difficult-to-interpret evidence.
Per-Henrik Zahl, MD, PhD, Jan Mæhlen, MD, PhD, H. Gilbert Welch, MD, MPH (2008). The Natural History of Invasive Breast Cancers Detected by Screening Mammography Archives of Internal Medicine, 168 (21), 2311-2316