| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Plenary Session |
Department of Radiology, Dartmouth-Hitchcock Medical Center, Lebanon, New Hampshire
 |
Commentary |
|---|
 Top Commentary ReferencesResponseReferences |
|---|
Dr Brant-Zawadzki begins his article with the assertion that "there can be no better evidence for the utility of CT...than the fact that approximately 33 million CT scans were performed in the United States last year." Certainly CT provides exquisite anatomic detail of the human body, and undoubtedly CT improves the outcomes of many patients in particular situations. Nevertheless, acceptance by the medical community does not constitute valid evidence of efficacy. Numerous medical practices have achieved widespread acceptance before well-designed studies demonstrated lack of efficacy or actual harm. For example, about 38% of postmenopausal women in the United States have been using hormone replacement therapy partly because of the widely held belief that it decreased the risk of cardiovascular disease (2). Recently, a large randomized clinical trial showed that the therapy had the opposite effect (3). Imaging tests are even more difficult to evaluate than treatments. As recently pointed out by Jarvik (4), new imaging tests often make referring physicians feel better even if they don’t make patients feel better.
Dr Brant-Zawadzki also refers to 5-year survival statistics in the argument for lung cancer screening. These statistics are always quoted by screening advocates because they always makescreening appear effective, often dramatically so. However, the comparison of 5-year survival rates is inappropriate in the evaluation of screening because of early detection biases: lead time, length, and overdiagnosis bias (5,6). The appropriate metrics for comparison are the rates of morbidity and mortality in those individuals who are eligible for screening, that is, those who do not have any signs or symptoms of the target disease when screening is begun.
The most valid study design for evaluating all medical interventions is the randomized clinical trial because it minimizes the effects of confounding variables. Hormone replacement therapy had achieved widespread acceptance partly because it appeared to reduce cardiac deaths in observation studies (2), which were probably strongly confounded by the subjects’ diet, exercise, and smoking history. Randomized clinical trials are particularly appropriate for the evaluation of screening because they eliminate the early detection biases previously mentioned, as well as the potential for confounding by variables associated with access to screening, such as income and life style.
Brant-Zawadzki raises the important issue of false-positive test results in lung cancer screening but dismisses it too quickly. It is true that the vast majority of patients with false-positive test results can probably be managed with follow-up unenhanced CT alone. However, because the false-positive rate is so high—over 65% in two rounds of screening in the Mayo Clinic experience (7)—a high proportion of screened patients without lung cancer may nevertheless undergo invasive testing or surgery.
Brant-Zawadzki also dismisses too quickly the radiation risk from screening by stating that "there is no scientific proof that cancer induction occurs in this range." Even though the radiation risk is likely to be very small and thus difficult to quantify, any benefits are also likely to be very small because mortality rates in those eligible for screening are generally very low. For example, for adults 50–54 years of age, the mortality rate from cancers of the lung, colon, pancreas, and kidney combined is only 0.8 per 1,000 persons per year (8) in the general population and much lower in the population eligible for screening.
Brant-Zawadzki does not address overdiagnosis, that is, the diagnosis of a condition that would not have become clinically significant had it not been detected by screening (9). Affected individuals receive unnecessary treatment, such as chemotherapy or surgical resection for overdiagnosed cancer. Good patient outcomes are attributed to early diagnosis and treatment, and even bad outcomes are often falsely assumed to be better than the natural course of the disease. Although it was once believed that overdiagnosis of cancer was a rare occurrence, several autopsy studies (10,11) and recent experience with different forms of screening (12–14) suggest that the potential for overdiagnosis is vast, even for lung cancer, which is often assumed to be uniformly lethal. In a recent study of CT screening for lung cancer (12), the rates of screening-detected cancers in smokers versus nonsmokers were almost identical (0.46% vs 0.44%), which strongly suggests that many cases in nonsmokers were not clinically significant. Although overdiagnosis may be the most serious potential harm from CT screening, the public is largely unaware of this side effect (15).
As mentioned, the most valid study design for determining the benefits and harms of screening is a randomized clinical trial. The usual endpoint is the mortality rate from the target disease (16). However, it is also useful to examine the mortality rate from all causes because it eliminates the problem of misclassification in cause of death—which can lead to either an underestimation or overestimation of the disease-specific mortality—and because this metric puts the mortality benefit or harm from screening into proper perspective (17). To date, no randomized clinical trial of cancer screening has demonstrated a statistically significant reduction in all-cause mortality, and there is actually a slight trend toward higher all-cause mortality in the screening arms (17). This latter finding could be due to chance, unrecognized deaths that resulted from screening, or a combination of both.
Because of the potential harms and the widespread confusion over the magnitude of the potential benefits from screening, I think screening with CT should be performed only under two conditions. First, CT screening should be part of a well-defined study protocol (not necessarily a randomized clinical trial) with Institutional Review Board approval that is likely to provide useful information about who should be considered for screening or how screening should be performed. Second, potential participants in any screening study should be given a balanced presentation of the expected benefits and risks. I suspect that eligible individuals would be much more willing to participate in screening randomized clinical trials and much less compelled to pursue the latest screening tests if they were properly informed.
|
References |
|---|
|
TopCommentaryReferencesResponseReferences |
|---|
Department of Radiology, Hoag Memorial Hospital, Newport Beach, California
|
Response |
|---|
|
TopCommentaryReferencesResponseReferences |
|---|
As to the issue of lung cancer screening, I clearly agree with Dr Black that randomized clinical trials are the only path to an answer that will satisfy the statistical proof of efficacy. Nevertheless, observational studies are important. This may be particularly true for evaluating diagnostic as opposed to therapeutic medical interventions. Recent publication of a registry of over 21,000 patients in Japan, including 799 patients who did not undergo surgical treatment for clinical stage I lung cancer, documents that the prognosis of this disease, even at stage I, is dismal, with only 16% of stage IA patients surviving 5 years and 7% surviving 10 years (2). These observational data (admittedly partially) address some of the issues of lead-time bias and "overdiagnosis." Yes, a small minority of stage I lung cancer patients survive 5 years without surgical treatment and seem to harbor indolent tumors, but how are we to tell which tumors are indolent a priori? Given the dismal statistics of lung cancer versus those of the other major cancer killers such as breast and colon cancer, which have seen benefit from screening programs, the argument for at least offering CT screening to individuals who have been informed of the risk-benefit ratio is hard to resist, and offering them surgery for stage I cancer is proper, given the 80% 5-year survival rates with such treatment.
Dr Black’s appropriate discussion of radiation risk versus benefit of cancer detection rates is elucidating. The statistic that the mortality rate for adults 50–54 years of age from lung, colon, pancreas, and kidney cancer combined is "only" 0.8 per 1,000 will make it difficult to quantify benefit from screening CT on a statistical basis; nevertheless, given the considerably larger population at risk beyond these age parameters and the fact that 23% of all individuals die from cancer is sobering. A recent review of cancer statistics from the epidemiologists at the American Cancer Society indicates that (despite dramatic growth of radiation-based diagnostic testing in the past 20 years) cancer incidence and mortality rates have dropped over the past decade, despite an aging population. Yet the authors cautioned that despite such reductions in the rates of cancer death, disparities exist in the rates for the African-American population, which has a higher incidence of cancer. The authors went on to suggest that such disparities reflect diagnosis of cancer at later disease stages and poorer survival rates for each stage—perhaps reflecting such individuals’ lack of access to earlier detection tools (3).
I agree with Dr Black’s statements about the need for evaluating true epidemiologic screening with CT under the conditions of a well-defined study protocol and that patients should be given a balanced presentation of the expected benefits and risks. On the other hand, I find it rare that practicing physicians give patients a balanced presentation of risk versus expected benefits for most treatment options (or that patients truly understand them), and yet these same clinicians use and— certainly from the perspective of experts in efficacy trials—overuse diagnostic imaging and other treatment methods to a large degree. Radiologists faced with physician-referred patients must deal with the same dilemmas as when patients self-refer themselves for diagnostic imaging studies, particularly CT with all of its surrounding controversy. The only difference is that when radiologists perform the tests requested by other physicians they are simply following orders. Indeed, radiologists may be in a better position to provide useful information to patients, including a balanced presentation of expected benefits and risks, if faced with the patient directly. This, however, is pragmatically difficult to do in the practice setting—as is performing diagnostic and therapeutic interventions only when valid evidence of efficacy of such management exists. This does not mean that we should not try to expand that evidence base as much as possible, and Dr Black is certainly one of the capable leaders of such efforts.
|
References |
|---|
|
TopCommentaryReferencesResponseReferences |
|---|
Related Article
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOGRAPHICS | RADIOLOGY | RSNA JOURNALS ONLINE |
