|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Originally published as JCO Early Release 10.1200/JCO.2008.21.0526 on March 2 2009 © 2009 American Society of Clinical Oncology.
Screening for Cancer With High Technology Imaging TestsFrom the Department of Radiology, the University of Virginia, Charlottesville, VA In this month's issue of Journal of Clinical Oncology, Nishizawa et al1 report on an initial experience using high technology imaging modalities to annually screen 1,197 healthy Japanese volunteers for cancer. While the emphasis is on the diagnostic performance of positron emission tomography (PET), the authors also tested the value of a multimodality approach, which included x-ray computed tomography (CT), magnetic resonance imaging (MRI), and an array of nonimaging markers for cancer. The screening protocol identified only one cancer during the two screens after the initial prevalence screen (18 cancers). The sensitivity of PET screening alone was 50% and for all tests together 81.8%. The authors aptly name this article a preliminary report. The sample size is small for a screening study, and only 2.3 years of follow-up were allowed to pass for the last enrollees. The authors also made some unusual choices in whom they screened and how they were screened. The inclusion criteria allowed anyone older than the age of 35 years to participate. The resultant average age of just older than 47 years is younger than might be desirable for efficient cancer screening. Moreover, the investigators chose not to administer contrast material for either the CT or MRI exams, the elimination of which would be expected to decrease both sensitivity and specificity. Altogether then, it is not surprising that the yield of cancers was nearly nil in the incidence screens, nor that the performance of the imaging technologies was disappointing. Given all of the foregoing, I cannot agree with the conclusion of the authors that "...this study supports the possible utility of whole-body cancer screening, including PET."1 Indeed, I would argue just the opposite. The findings of this study illustrate the fundamental problems of employing any high technology imaging modality or combination of such modalities to screen for disease in an undifferentiated healthy population. The yield will be low because the a priori risk of disease is scant. The costs inevitably will be high because of the nature of modern medical imaging. The exquisite spatial and contrast resolution of modalities like CT, MRI, and PET, ensure that there will be findings. However, the vast majority of these findings will not impact the lifespan of the screened individuals. Many will be false-positive detections, anatomic variants, and so-called "incidentalomas" such as complicated renal cysts and adrenal adenomas. As was the case in this study, these tend to be much more common than true-positive findings for malignancy and most often generate additional imaging and occasionally even invasive procedures or unnecessary treatment. Not surprisingly, the authors note that the sample that underwent screening had higher health care costs in the follow-up years than a sample of workers in the same factory who did not participate in the study. None of the additional care related to the false-positive diagnoses improves the health of the affected individuals. Indeed, such individuals suffer harm in the forms of greater anxiety, adverse effects, and complications from inappropriate treatment, and financial contributions toward unnecessary care. In essence, screening erroneously transforms a large number of healthy people into patients. In their concluding sentence, the authors note that, "...our study is limited due to the small number of participants, and a cooperative study with a prospective design will be required."1 I doubt this is a good idea, but let's at least explore the proposal. What the authors appear to be suggesting is embarking on a multicenter continuation of the cohort approach they detail here. But a single-arm cohort investigation will not tell us what we must learn from a screening trial—whether multimodality imaging screening reduces the cancer-related mortality rate, and if so, at what cost. The authors seem to assume that because they found cancers, the participants benefited. However, they present no data that would allow us to test that assumption. Were the cancers they found aggressive malignancies that would have caused death? Were they indolent ones that might never have affected their participants? Something in between? We don't know. Cancers have a broad range of biologic behavior. It may be that detecting the asymptomatic cancers allowed cures of life-threatening disease, but two other possibilities exist. Some of the tumors may have been so aggressive that even finding them before they became symptomatic resulted in no change in when the participants would have died. Others might have been so slow growing that the cancers would never have caused harm. In both of these circumstances, the lesions found on screening would still be worked up, medically treated, and surgically excised, but life would not be extended beyond what it would have been if no screening had occurred. That so many of the detected cancers were identified on the initial prevalence screen (18 of 19), and that so many of these were stage I (12 of 18), raises alarms that much of what was found might have been unimportant to the health of the participants (ie, pseudodisease). Variability in tumor behavior is the basis for the specific biases implicit in observational studies of screening, referred to as lead-time and length biases. The combination of these biases, which respectively make patients appear to live longer with their disease and select for less aggressive tumors in groups of subjects who are screened, make screening tests appear more effective than they really would be in practice. These biases are operant whether there are 1,197, 7,360, or 30,376 subjects, so simply expanding the sample size of a single-arm trial provides no relief from the spurious result. Only a randomized clinical trial—wherein some participants receive the screening and others do not—can reasonably determine whether PET or multimodality imaging screening reduces the death rate from cancer and at what cost. Any other design will only lead to the kind of ambiguity, acerbic public wrangling, and political maneuvering we have witnessed surrounding the notion of CT screening for lung cancer. A bigger question is whether it generally makes any sense to use expensive imaging technologies to perform whole-body screening of low-risk, healthy populations. There is no question but that computerized imaging modalities can find cancer or heart disease or brain abnormalities. However, all of our current imaging technologies are susceptible to the same problems that I have detailed above. They find far more incidentalomas and pseudodisease than abnormalities that can be successfully addressed to improve human health. The considerable downstream costs generated as a result have the potential to overwhelm the resources we have available to address more pressing health care needs. This is not to say that some individuals might not benefit from imaging screening, but rather, it is an argument against any systematic implementation of large scale, society-based imaging screening in the absence of data supporting the cost-effectiveness of a cancer screening program. What we need to consider is how we could make screening more efficient and by doing so take advantage of the impressive qualities computerized imaging offers. One possibility is to migrate from the idea of screening the general population—which has been so popularized in the media over the past decade—to one of using imaging to monitor for life-threatening disease in elements of our society at higher risk. Two sets of advances would be necessary to achieve this end. First, we would require innovations that permit us to assess the risk of disease beyond the relatively gross risk stratification we employ today (eg, annual mammography for women older than 40). As we learn more about the serum and tissue markers that portend higher risk, we can design a more directed approach of periodic imaging monitoring that should result in higher yields of disease for which improved treatments might better achieve the goal of reducing deaths from the target condition. Secondly, as noted earlier, while radiologists are expert in using modern imaging technologies to accurately identify deviations of appearance from normal, the methods to provide insight into whether what is discovered is meaningful to the patient are insufficiently accurate. Effective testing of biologic aggressiveness that could quickly, accurately, noninvasively, and inexpensively help predict the potential importance of imaging findings to patient outcomes would go a long way to improving the efficiency and, therefore, the cost-effectiveness of screening using medical imaging devices. Advances of this nature may not be far off, but for the present, any proposal to apply modern imaging technologies to whole-body screening should be met with skepticism. Only randomized trials can definitively prove the worth of such testing, and these tend to be large, slow, and expensive. A randomized trial of whole-body screening for cancer of the general population using PET or some combination of advanced imaging methods would necessarily be massive because of the low rate of true disease. At the very least, modeling and decision analysis should be employed to evaluate the likelihood of a positive result before further trials are implemented. AUTHOR'S DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST The author(s) indicated no potential conflicts of interest. REFERENCE
1. Nishizawa S, Kojima S, Teramukai S, et al: Prospective evaluation of whole-body cancer screening with multiple modalities including [18F]-fluorodeoxyglucose positron emission tomography in a healthy population: a preliminary report. J Clin Oncol 27:1767–1773, 2009.
Related Article
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2009 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
