Cancer overdiagnosis and overtreatment revisited

I think the message may finally getting through.

That message is that it’s not always the best strategy to treat cancer aggressively. Don’t get me wrong. If I have acute leukemia, I know I’ll need the big guns, every bit of chemotherapy appropriate to the disease that modern oncology can throw at it up to and including a bone marrow transplant. But what about prostate cancer? Breast cancer? Thyroid cancer? It turns out that more treatment isn’t always better for these diseases, depending on the subtype, and in some cases, such as early stage prostate cancer of low aggressiveness, it is quite reasonable to conclude that the treatment is more harmful than the disease. That’s because, if you’re a man and reach your 80s, there’s a 75% chance that you will have small foci of cancer in your prostate that you live with but are not what you ultimate die from. Autopsy series have shown this time and time again.

This concept is counterintuitive and goes against what most people understand about cancer. It even goes against what most physicians understand about cancer to suggest that there are limits to what early detection can do and that the early detection of cancer is more complex than moe people–again, even physicians–think. However, as I’ve described on this blog, study after study over the last few years have shown that detecting cancer ever earlier, finding ever smaller cancerous lesions, even down to small clusters of cells, does not necessarily translate to saving more lives. It also comes with a downside that has frequently been downplayed. As I’ve pointed out, the problem of overdiagnosis–detecting small tumors that would, if left alone, never grow to endanger the life of the patient is huge, but, more importantly, overdiagnosis leads to overtreatment.

The mainstream media is actually starting to pick up on this message, too. Th elatest evidence of this is a story published yesterday by the AP authored by Laura Neergaard entitled Overtreating earliest cancers _ but which ones? As is typical of a news story, Neergaard starts with an anecdote:

D.J. Soviero wanted the least treatment that would beat back her small, early-stage breast cancer, but her first doctor insisted she had only one option: tumor removal followed by radiation and chemotherapy.

Then she found a novel program at the University of California, San Francisco, that gave her an unbiased evaluation of the pros and cons of all treatment options.

“I realized that I didn’t need to use a sledgehammer. It was my choice,” said Soviero, of San Francisco, who went with the lumpectomy and radiation, but refused the chemo.

And it’s true. Again, let me emphasize one thing about this. We are not talking about large, palpable tumors here but rather earlier stage breast cancers, the kind that are typically asymptomatic and detected by screening mammography alone. Symptomatic cancers need to be treated, as we have good evidence for these that regression or failure to progress is rare and that failing to treat these cancers greatly increases a woman’s chances of dying of the cancer. However, as I have taken to blogging about lately, for mammographically-detected small tumors, almost always those detected by screening mammography, it’s not so clear whether all of these need to be treated. Overdiagnosis is being increasingly appreciated as a significant problem, and, indeed, may account for as many as 1 in 3 breast cancers detected by screening mammography (although more common estimates are on the order of 20%). There is even evidence–not bulletproof by any means, but intriguing evidence–that as many as 20% of mammographically detected tumors may actually spontaneously regress. Problems such as this in prostate cancer, for which overdiagnosis due to PSA screening was appreciated as a significant problem long before overdiagnosis was appreciated as a problem in other cancers, and breast cancer have produced new recommendations, such as the USPSTF guidelines on screening mammography that caused such an uproar last fall.

So we know that screening to find every earlier and ever tinier cancers results in overdiagnosis, the finding of cancers that do not require treatment. The problem is that we don’t know which cancers are safe to observe; so we treat them all, as Neergaard reports:

“The message has been, `Early detection, early detection, early detection.’ That’s true for some things but not all things,” said Dr. Laura Esserman, a breast cancer specialist at UCSF. She helped lead a study, reported last week, that found mammography is increasing diagnoses of tumors deemed genetically very low risk.

“It’s not just all about finding any cancer. It’s about being more discriminating when you do find it,” she added.

Today’s cancer screenings can unearth tumors that scientists say never would have threatened the person’s life. The problem is there aren’t surefire ways to tell in advance which tumors won’t be dangerous – .just some clues that doctors use in prescribing treatment.

I blogged about the study by Esserman to which Neergard is referring, by the way.

I feel obligated right here to point out that trying to predict which tumors require aggressive treatment and which do not is nothing new. Oncologists and surgeons have been doing that for decades. Commonly used predictive measures include characteristics such as how a tumor looks under the microscope to a pathologist, some molecular markers, location, and some radiographic characteristics. It’s not even an entirely new concept that some tumors don’t need treatment or don’t need a lot of treatment. Some of the chronic leukemias, for example. What is new is generalizing this concept and learning that it applies to many more cancers than we previously thought, including very common cancers, such as prostate and breast cancer, coupled with an understanding that screening tests tend to detect less aggressive tumors more readily than they detect rapidly growing tumors. I described this problem in detail three years ago. Now, it appears that we may finally be getting up the guts to try to do something about it, compelled by evidence like this on prostate cancer:

Nowhere is the disconnect more obvious than with prostate cancer screening. Most men over 50 have had a PSA blood test to check for it even though major medical groups don’t recommend routine PSAs, worried they may do more harm than good for the average man.

What’s the evidence? A study of 76,000 U.S. men, published last year, concluded annual PSAs didn’t save lives. A separate study estimated two of every five men whose prostate cancer was caught through a PSA test had tumors too slow-growing ever to be a threat.

A European study of 162,000 men screened less aggressively – a PSA every four years versus none – found seven fewer deaths per 10,000 men screened. But 48 men had to be treated to prevent each death, meaning many men who weren’t facing death experienced treatment that can have such side effects as incontinence and impotence.

Neergard also points out a study in JNCI that found that nearly one quarter of cancers diagnosed by mammography may be overdiagnosed. Somehow I apparently missed that study, because I don’t remember it. Maybe I’ll go back and look it up to blog about it, although its findings are in line with the findings of other studies recently, albeit with a lower estimate of overdiagnosis than some studies I’ve seen. Be that as it may, the study also apparently found that much of what mammography diagnoses is ductal carcinoma in situ (DCIS).

DCIS is a difficult problem. Frequently, it’s described as breast cancer that hasn’t broken out of the milk ducts yet. However, after decades of study, it’s still argued whether DCIS is truly a premalignant lesion or a precursor to invasive breast cancer. We do know that death from DCIS (more appropriately, an invasive cancer that we suspect to have developed from DCIS) is uncommon when DCIS is treated but we don’t know what percentage of DCIS lesions would, if left alone, actually progress to frankly invasive breast cancer; so we treat them all with surgical excision and most of them with radiation therapy afterward. Yet estimates of how often DCIS progresses to actual cancer when not treated are hard to come by, ranging from as low as 20% to over 50%. Being able to determine which cases of DCIS require no further treatment after surgical excision or even no treatment at all after biopsy would be a huge boon to women, sparing thousands of them every year from lumpectomy and radiation therapy, often followed by and estrogen blocking drug (Tamoxifen) for five years.

Esserman’s group is taking it even further than that. She has researchers looking at whether drugs like Tamoxifen can allow DCIS patients to avoid surgery altogether. I’ll have to be honest about this study, though. I doubt it will work for the simple reason that a large percentage of DCIS doesn’t even make the estrogen receptor (i.e., is estrogen receptor-negative). Some of it even makes HER2/neu. In fact, I wonder how they got that study through the IRB.

More promising to me are approaches looking at molecular markers described in Neergaard’s article to try to identify which DCIS lesions don’t need treatment. I’m also intrigued by this study:

Beyond DCIS, Esserman is designing a first-of-a-kind study to start by summer’s end at five University of California health centers. Women whose mammograms turn up a specific type of suspicious spot that is unlikely to be aggressive cancer will get the option of skipping today’s usual biopsy and repeating the scan in six months instead. She hopes to learn which early abnormalities are safe to leave alone.

“If you’ve had a normal mammogram and develop a new mass, don’t ignore that. If you have a new symptom, those are things you don’t want to ignore,” Esserman said. “The public also has to understand that it’s complicated and there are some cancers that are very slow-growing.”

A point of clarification is in order here. We actually already do this for certain mammographic lesions. I’m sure there are women out there among my readers who have been called back for the infamous “six month followup” mammogram. Most of those are for a lesion that’s categorized as a BI-RADS category III by the radiologist. (The BIRADS classification is a rating system of the level of suspicion for cancer there is.) In fact, for category III studies, the standard of care is a six month followup mammogram because only 0.5-1.0% of these will turn out to have cancer. Also, early all of those will be DCIS or stage I cancer. Waiting six months doesn’t impact survival in these cases. What it sounds as though Esserman is doing is to study women with category IV studies, for which biopsy is virtually always recemmended and which portend a 25% chance of there being cancer. She’s probably taking a subset of these women at the low end of the risk scale of women with category IV mammograms. While this is a good idea, it’s not going far enough. We need new imaging technology and a deeper knowledge of the biology of DCIS and early breast cancer. The same is true for many other cancers. Until then, we have to assume that every cancer we find is potentially deadly and treat them all. Until then, the key is to be able to explain the uncertainty involved with these issues to patients and help them come to a collaborative decision for care based on science, the level of uncertainty in the science, and their own tolerance for risk.

Neegaard mentions one type of harm that comes from ever-more sensitive imaging technology, namely the problem of the “incidentaloma.” Basically, an incidentaloma is a lesion found on a test done for another reason. Back when I was a resident we used to talk about them in patients who had a CT scan after major trauma in which a mass was found. The classic general surgery question is what to do with an adrenal incidentaloma, because all these abdominal CTs were turning up adrenal incidentaloma, the majority of which were benign. Over time, indicators of malignancy were derived based primarily on size. Below a certain size, and adrenal incidentalomas can be safely watched. Above a certain size, they need to be biopsied and excised because of the high likelihood that they represent a malignancy. If only it were so “easy” for other cancers.

There is another consquence of ever more sensitive imaging technology for cancer patients, and that is overtreatment of a cancer that does actualy need treatment. This is mostly due to what’s known as stage migration (colloquially dubbed the Will Rogers effect 25 years ago), a phenomenon I discussed in detail here. One example in breast cancer is how MRI scans. MRIs are more sensitive than mammography and ultrasound and therefore pick up more lesions cancer patients that, especially because the patient has cancer, surgeons feel obligated to work up fully. Unfortunately, although there was a period where practice changed to offering MRIs to evetry woman with breast cancer, it’s become clear that preoperative MRIs do not improve surgical management by decreasing the percentage of time it’s necessary to go back to take more breast tissue after a lumpectomy. Worse, there is more and more evidence that MRI use is associated with more mastectomies in women who probably didn’t actually need a mastectomy. Over time, surgeons started to realize this, and studies supported the hypothesis that doing MRIs preoperatively on every woman with breast cancer didn’t add much to the treatment plan for most women and are very likely increasing the mastectomy rate.

One of the interesting aspects about blogging is that I can go back and see the evolution of my own thinking on a topic. When I started writing about these issues, beginning back in 2007, I was pretty much of the mindset that screening is good and more screening is better. Well, not exactly, but I didn’t appreciate the phenomena of overdiagnosis and overtreatment anywhere near as much as I do now. My thinking has evolved considerably since then. In fact, I can honestly say that blogging about some of these studies has definitely been a significant contributor to the evolution of my thought on the matter. I’ve now become very much more interested in what approaches we could develop that would improve the specificity of treatment and spare cancer patients who don’t need them the harsh and nasty treatments that we so often subject them to. I can’t say that I’ve figured out yet how to study this problem in a way that hasn’t been well trod before, but I’m definitely working on it.