Making Sense of Cancer Screening Controversies

Q:A near riot of headlines in the medical and laypress has followed in the wake of publication of2 studies from Danish researchers on the value ofscreening mammography.^1,2 Gotzsche and Olsen foundthat the 7 large clinical trials on mammography--onwhich current screening recommendations are based--were too flawed to support those recommendations.Many of my patients--and my physician colleagues--are now confused about the role that such screeningshould play. What is your perspective?A:In challenging the quality of some of the original studies,the Danish authors have raised the importantissue of concealed allocation--that is, whether the participantswere adequately randomized and truly blinded. It appearsthat randomization was inadequate in some of thestudies, so the groups were not properly balanced. Althoughmortality was lower in the women who werescreened, these women may have been healthier for a varietyof reasons (eg, higher socioeconomic status). The authorshave gone to some lengths to try to show that thisbias did in fact occur, and point out that trials with inadequaterandomization exaggerate the estimated interventioneffect by an average of 33% to 41%.Of the 7 studies, only 2 were considered by the authorsto have been of adequate quality for consideration.These 2 studies showed that screening had no effect onbreast cancer mortality. Gotzsche and Olsen therefore concludedthat mammography does not save lives.I, for one, am not convinced that these are fatal flawsin the evidence. First, the fact that there may have been thepotential for selection bias in the randomization processdoes not mean that such a bias definitely occurred and alteredthe results. I believe the authors were overzealous intheir criticism.Second, in their synopsis of the Cochrane review, theauthors point out that there is no evidence that mammographylowers all-cause mortality and use that as an argumentagainst screening.² I do not think this argument isvalid. Many studies show disease-specific rather than allcausemortality reduction, and that has generally been thestandard we use to assess efficacy. It's very hard to showall-cause mortality reduction from a single intervention--such a study would have to involve huge numbers of participants.Even if we reduce the death risk from breast cancer,women are still dying of other cancers, heart disease,and strokes. There are a whole host of reasons why peopledie where we would never expect mammography to makea difference.I would also point out that we have excellent randomizedtrials of colorectal cancer screening that showcolon cancer-specific mortality reduction, not all-causemortality reduction. We've never said that screening wasineffective because the studies failed to show all-causemortality reduction.Q:So, despite any methodologic flaws in thestudies reviewed, is it justifiable to concludethat mammography is of no benefit?A:There is no such thing as a perfect study. But manyof us are uncertain now whether the flaws in randomizationwere sufficiently severe to neutralize previousfindings and increase our skepticism about mammography.However, it's unlikely that there will be another mammographytrial (although, with the emergence of this controversy,there may well be some new momentum in thatdirection).We have to be honest with our patients and let themknow that there is now a basic uncertainty about whethermammography is helpful.Q:Should I still advise my 40-year-old patient tohave a mammogram?A:In the last 5 to 7 years we've debated whether routinescreening mammography is effective in women aged40 to 49 years. We simply don't know whether early detectionof breast cancer leads to improved outcomes in thesewomen. The best evidence we have--assuming we includesome of the studies with concealed allocation--suggeststhat we have to screen close to 2000 women in this agegroup over a 10-year period to save a single life.Because of the uncertainty and because of themedicolegal implications, I do offer mammography to allwomen 40 years and older. But I don't recommend it universallyanymore. I tell a 40-year-old woman that the evidencethat we can save her life with mammography is notcompelling. If she wants a mammogram,I'm happy to order it. However,I will also tell her that if we startscreening at age 40, there's virtually a100% chance that some kind of abnormalitywill show up that will warrant atleast a follow-up mammogram, an ultrasoundscan, or a call from the doctorrecommending a 6-month follow-upexamination. There is also a high likelihoodthat over the course of a lifetime many women willundergo an unnecessary breast biopsy. To allay patients'fears, I tell them that in most cases the finding will turn outnot to be cancer.So the downside of screening is that we create a lot ofworry--much of it needless. Despite this, many womensay they'd rather be worried than run the risk of cancer.Others say that because there is uncertainty aboutwhether the screening is helpful, they'd just as soon forgoit for now.The other risk of screening is finding malignanciesthat don't matter--which, by the way, has always been anissue with prostate cancer screening and may now be anissue with early screening for lung cancer. (Editor's note:For one perspective on early lung cancer screening, please seePulmonary Q&A, page 767.) Mammography is particularlygood at finding ductal carcinoma in situ, which accountsfor nearly 20% of the 230,000 cases of breast cancer eachyear. But we just don't know how dangerous these malignanciesare. Many--if not most--of them might neverspread. But many women are having lumpectomies andmastectomies for these early cancers. It's just not clear yetwhether this is a benefit or not.Q:How do you advise your women patientsolder than 50 years who ask aboutmammography?A:I'm approaching older women more the way I do the40- to 49-year-olds. I say, "I know you've been gettingscreening mammograms every year. However, we're nolonger sure if it's helpful. The evidence is getting more controversial.Is this something you want to continue doing?"And I spend a few minutes discussing it with them.Q:Do we need to change our message from"With early detection, you can be cured" to"Early detection may save some lives"?A:Yes, it's still fair to say that early detection appears tosave some lives. And not to be overlooked is the factthat early detection can also shift the diagnosis of cancer toa stage where treatment is often much less drastic, moreeasily tolerated, and much less expensive.However, we know now that evenif the cancer is detected and treatedearly, we can't always predict the naturalhistory of that cancer. We don'tknow in many cases whether the malignancyis destined to be aggressiveor indolent. So the issue of lead-timebias comes into play. It looks like we'redoing a better job because we've diagnosedthe cancer early. But the reality may be that we'vesimply extended the amount of time the patient is aliveknowing that she has cancer--not necessarily that her lifehas been prolonged. She may live just as long with later diagnosisand treatment.Q:How do you help your patients come to gripswith the implications of what you've just said?A:It's difficult and disconcerting. In this country, mammographyhas been almost the paradigm of thescreening test that prevents tragedy. Unfortunately, this notionmay give a false sense of security.Many women have believed that mammography enabledthem to protect themselves. That belief is based onstudies, some of which were conducted back in the 1970s,when we had less sophisticated screening techniques.Some critics say that we can't base our conclusions aboutmammography on those older screening techniques. Butbetter detection still leaves us uncertain about a lesionwe might find. Is this an important lesion? Will it makea difference if we treat it? We just don't have all the answersyet.Q:Bottom line: Is it fair to say that mammographyis no longer in the category of screening teststhat have unambiguous benefits--such as tests forcolon and cervical cancer? That its benefits areambiguous--like those of prostate and lung cancerscreening?A:These are fair statements. Not only can we now detectlesions that might never harm a patient, but infact this earlier detection may lead him or her to undergoan invasive procedure that entails its own risks. We used tobelieve that with lung cancer, all malignancy matters. Butwith the new screening techniques, we're shifting the detectionpoint back far enough that we're finding lesionswhose natural history we do not yet fully understand. Unfortunately,the morbidity involved in removing that lesioncan be substantial. A procedure such as a transthoracicbiopsy involves potentially significant risks, including pneumothoraxor other major adverse events. Even though thepotential morbidity involved in early intervention in breastand prostate lesions is not inconsequential, the stakes withlung lesions are likely to be even higher.So, as we enter these areas of greater uncertainty, werun the real risk of undercutting the premier rule of medicine,"First, do no harm." There's a higher likelihood nowthat we could be doing harm with unproven benefit.Q:How do you make the time to address theseissues with patients in today's managed-careenvironment?A:It can't be done with every patient at every visit.Screening decisions need to be made based on thecurrent evidence. In my initial talk with a patient, we discussthe issues and the patient states her point of view. If shewants to have annual mammographic screening, I don't gothrough the whole discussion every year. If she doesn'twant to be screened, I will ask her the following year if anythinghas changed.I follow the same strategy with prostate cancer screening.If there are new data, or the patient has had a changeof heart, he can raise the issue at the next visit.Q:You've stated that physicians as well aspatients tend to overstate the risks of cancerand, more important, to exaggerate the risk reductionconferred by screening. Can you elaborate?A:I do think that physicians tend to overstate the benefitsof screening. For example, the benefit conferredby mammography--if there is one--is far from dramatic.The best estimates in the 7 studies cited by Gotzsche andOlsen showed that there was about a 28% risk reduction inthe 50-to-74 age group. What does that mean to the individualwoman? The likelihood of death from breast cancer isbetween 3% and 4% over the course of a lifetime. Assuminga woman has a mammogram every year or two, and assumingall the evidence is correct, the best-case scenario isthat we're lowering the risk from about 4% to about 3%.When we look at it this way, that 28% risk reduction doesn'tsound as dramatic. Clearly, from a population standpoint,the potential mortality reduction is very significant: thenumber of women dying of breast cancer could be reducedby thousands. But the risk reduction at the individual patientlevel is modest.The same is true for colon cancer screening. We tellpatients that there's a 33% risk reduction in colon cancermortality with annual fecal occult blood testing over a periodof 10 to 15 years. But the risk of dying of colon canceris about 2.5% over a lifetime. Thus, we're lowering the riskfrom just over 2 in 100 to just under 2 in 100.Q:What is the risk reduction withcolonoscopy?A:We don't know because there haven't been any randomizedtrials with colonoscopy. Interestingly, withcolon cancer there's almost an inverse correlation betweenthe quality of the screening testand the quality of the evidence that itsaves lives. Even though fecal occultblood testing is considered the leasteffective screening test (it misses a lotof cancers and produces many falsepositives),the results of large, randomizedEuropean and American trialsdemonstrate its efficacy.With sigmoidoscopy, we can visualizeonly 40% to 45% of the colon. Wedon't have any randomized trials forsigmoidoscopy, although we havesome good case-control studies thatshow a colon cancer risk reduction ofabout 40%. Colonoscopy is the goldstandard for screening tests, but we have no randomized trialsor good case-control studies, just case series. These seriesdemonstrate that colonoscopy finds a lot of cancers thatthe other screening tests would have missed. Althoughcolonoscopy has "face validity," we haven't shown that itlowers the risk of colon cancer by more than 33%. And thetest is not without adverse effects, including perforation,bleeding, and cardiorespiratory complications of the sedationused during the procedure.Q:It is ironic that the more uncertain we areof the benefits of a screening test, the more wehave to involve our patients in the decision-makingprocess. What's the most efficient way you've foundto do this?A:Obviously, in areas where we're more certain becausewe have good evidence--such as with colonand cervical cancer screening--we don't need to go into along song and dance. The more evidence we have of benefit,the less patients need to be involved in the decision. Theless evidence we have, the greater our obligation is to sharethe burden of uncertainty.This is more manageable if we can do it at the practicelevel, rather than through individual patient encounters--especially if time is limited. There's plenty of informationavailable in the media and on the Internet. You can provideliterature in the waiting room that outlines the pros andcons of the various screening tests. The more informed patientsare before they see us, the less time we need to spendeducating them in the office and the more efficiently we candiscuss the issues.I'm deeply involved in the issue of colon cancer screening,because there is good evidence that it works. In thissetting, I ask my patients which screening test they wouldchoose. So, for example, a patient might say, "I've readthe pros and cons and I want acolonoscopy. I know my insurancedoesn't cover it, but I'm willing to payout of pocket."Q:A case related to this issuewas recently described inthe lay press. It involved a suitbrought by a patient against hisphysician. The doctor had reviewedthe pros and cons of prostatecancer screening with that patient,who decided against such screening.An aggressive prostate cancersubsequently developed in thepatient. On that basis, the patientsuccessfully sued his doctor. That's a nightmare for allphysicians. Is there a strategy to reduce the risk ofbeing sued in this setting?A:I suspect a case like this is quite rare. It's muchmore common for lawsuits to be won where therewas a cancer that was missed because screening wasn'toffered. But if someone is going to sue us even afterwe've offered screening and he has declined it--that'swhy we have malpractice insurance. It's unfortunate thatthese things happen, but I don't believe we should practice"overly defensive medicine" as a result. What is legalis not necessarily ethical. In this case, if the patient wasinformed of the potential benefits and risks of beingscreened versus not being screened, the physician aloneis not liable for the outcome. I don't know if true informeddecision making took place in this case; it mightbe that the patient wasn't really informed, or that the doctorsaid "I don't think it's a very good test," or "I don'tlike to screen for prostate cancer," or expressed his biasin some way. That's why I strongly recommend documentingthis type of discussion. I note in the chart thatI have discussed the pros and cons of cancer screeningand what the patient's decision was. That's a satisfactorylevel of evidence, in my mind, for demonstrating that theissue was addressed.The bottom line is that evidence for the benefits ofprostate cancer screening is ambiguous. All the national societiesexcept one recommend that patients be involved inthe decision to screen for prostate cancer. If we're followingthose guidelines, we're practicing good medicine--even ifsome of our patients who choose not to be screened ultimatelyget cancer and subsequently die. If we change ourpractice based on some selected bad outcomes from malpracticesuits, then we've failed.Q:Proponents of early screening note a decreasein breast cancer deaths inthe past decade. Do you attributethat to early detection? Or tobetter therapy--for example,tamoxifen?A:Deaths from breast cancer havebeen on a somewhat linear declinesince 1992. We don't know why,although the decline could be attributableto a combination of early detectionand better therapy. Also, tamoxifenis clearly a useful adjunct tosurgery and chemotherapy.Prostate cancer mortality has also been decliningsince about 1992. Advocates of screening attribute this tothe fact that prostate-specific antigen (PSA) testing becamewidely available by 1990. But opponents say that if screeningwere really effective, we wouldn't be seeing the benefitsso early. There would have to be a longer delay than ayear or two before the death rates started dropping. Thepatients who died would have been those who'd been livingwith prostate cancer for 5 to 10 years, not those whosecancer had been diagnosed a year or two earlier with PSAscreening. The trend started earlier than we would havepredicted.We can't explain the decline, just as we can't explainthe rise in incidence until 1990 or so. Prostate cancer mortalityis roughly the same as it was in the late 1970s orearly 1980s--it's no lower. There may be some epidemiologicor environmental factors that we have yet to understand.So, although I believe strongly in preventive medicine,I'm becoming more skeptical about the potential efficacyof some of our interventions. Historically, we'vebeen "burned" a number of times by jumping on thebandwagon before the evidence truly supports wherewe're going.Q:Do you recommend screening for patients witha family history of cancer?A:Those with a family history of colorectal cancer,breast cancer, and prostate cancer are at higher riskfor the disease. Presumably, they will have a higher benefitfrom screening than the general population. However,most cases of breast and colon cancer are sporadic.Q:What's the best health advice we can giveour patients?A:I can't emphasize enough how important it is for peopleto stop smoking. With this one step, hundreds ofthousands of deaths worldwide couldbe prevented every year. And I can'toverstate the importance of preventingand treating obesity. We are largely ignoringthe issues of good nutrition andan active lifestyle.It's ironic that in our society welook for higher-tech, lower-effort interventionsto safeguard our health. Idon't mean to underplay the importanceof appropriate cancer screening,but we're learning that probably it'sthe lower-tech, higher-effort interventionsthat can protect us to a greater extent.

References:

REFERENCES:

1.

Gotzsche PC, Olsen O. Is screening for breast cancer with mammographyjustifiable?

Lancet.

2000;355:129-134.

2.

Olsen O, Gotzsche PC. Cochrane review on screening for breast cancer withmammography.

Lancet.

2001;358:1340-1342.

3.

Pearce KF, Haefner HK, Sarwar SF, Nolan TE. Cytopathologic findings onvaginal Papanicolaou smears after hysterectomy for benign gynecologic disease.

N Engl J Med.

1996;335:1559-1562.

FOR MORE INFORMATION:

• Coley CM, Barry MJ, Fleming C, et al. Early detection of prostate cancer, II:estimating the risks, benefits, and costs. Ann Intern Med. 1997;126:468-479.
• Kerlikowske K, Grady D, Rubin SM, et al. Efficacy of screening mammography:a meta-analysis. JAMA. 1995;273:149-154.
• Ransohoff DF, Sandler RS. Screening for colorectal cancer. N Engl J Med.2002;346:40-44.
• Sawaya GF, Brown AD, Washington AE, Garber AM. Current approaches tocervical cancer screening. N Engl J Med. 2001;344:1603-1607.
• US Preventive Services Task Force. Guide to Clinical Preventive Services. 2nd ed.Baltimore: Williams & Wilkins; 1996. Updates to USPSTF reports may be obtainedat www.ahrq.gov/clinic/uspstfix.htm.