The use of modern imaging technology to screen healthy people for signs of incipient disease may be as much a curse as a blessing, making the issue of informed consent of paramount importance.
Hassaun Jones-Bey,Senior Editor
The prospect of early and noninvasive diagnoses of life-threatening conditions gets people excited. Ordinary people with no symptoms of any disease but believing they might be at risk for a life-threatening one read a newspaper article about a new technology with medical diagnostic potential, and they telephone local medical centers, sometimes by the thousands, to sign up for a test. Medical researchers and practitioners get excited about potential life-saving breakthroughs. Medical-insurance companies get excited at the prospect of being inundated with claims. Medical-equipment manufacturers get excited at the prospect of soaring sales revenues. However, some researchers and organizations are beginning to pay more attention to the potential downside of screeningfalse-positive test results and treatments that are ineffective or unnecessary.
There is no simple solution to the problem, particularly as the considerable technical prowess of modern society focuses more and more on health-related issues. For example, physicians concerned with the use of medical imaging technology in screening high-risk individuals for lung cancer are currently trying to determine whether or not the latest breakthrough, spiral computed tomography (CT) scanning, will really provide a health benefit or is simply a case of history repeating itself. The chest x-ray reigned as the highly recommended annual screening test for smokers for decades until a randomized clinical trial commenced by the Mayo Clinic (Rochester, MN) in 1971 indicated that the screening had no effect on lung cancer mortality. In the case of spiral CT, a July 9, 1999, article in The New York Times reported the results of a non-randomized study in which spiral CT scanning detected previously undetectable lung-cancer nodules. The very next day, an article in the same newspaper documented the telephone calls of thousands of smokers and former smokers to medical centers around the country requesting the test. And an article on October 27, 1999, documented efforts by the National Cancer Institute (NCI; Gaithersburg, MD) to rapidly launch a randomized trial, along with the opposite opinion, that the evidence was already so clear that a randomized trial would unnecessarily risk the lives of participants in a control group.
Laser Focus World spoke with William C. Black, M.D., associate professor of radiology at the Dartmouth-Hitchcock Medical Center (Lebanon, NH) and a proponent of randomized clinical trials and patient consent forms for screening procedures, about the concerns that arise whenever a new screening methodology is proposed. Black is also a co-investigator on a research team headed by Denise Aberle, M.D., at the University of California-Los Angeles, that is submitting a proposal to the NCI to perform a randomized clinical trial of spiral CT for lung-cancer screening through the American College of Radiology Imaging Network.
How has diagnostic imaging affected our understanding of disease?
In the past we may have been aware of only the more severe or advanced cases of a particular disease. But as imaging improves we are able to detect more of the milder cases, or at least earlier forms of what may later become serious diseases. So we actually broaden our ability to detect all diseases: cancer, pulmonary emboli, pneumonia, anything you can think of, virtually. The names now that apply to those diseases have actually much broader meaning. This is a general phenomenon that is usually lost in the statistics that we use to describe the burden of disease: the incidence and the prevalence of the disease. Epidemiologists and a lot of other people are aware of it. Scientists are generally aware of the concept. But a lot of them, even though they are aware of the concept, don't recognize how much of a problem it actually is in a particular type of disease. And by and large the general public seems completely unaware of this.
What are the implications for medical imaging as a widespread tool for diagnosis?
As we use more and more imaging, we'll find more disease. Some patients may benefit from earlier detection, but a lot won't. So as we find more disease, we'll be labeling more patients with disease who may or may not have experienced problems without the detection. Of course we'll also be treating more patients, and those treatments obviously are going to have some side effects. Certain populations might benefit from the earlier detection, but in general I would say that there is more potential harm than benefit in treating people unnecessarily.
Can you give examples?
One of the best and most familiar examples is prostate cancer. While microscopic evidence of disease is present in more than half of men over the age of 60 years, only about 4% of men die from prostate cancer. Consequently, as our ability to detect early forms of this disease has increased with prostate-specific antigen (PSA), we have found more cases and treated more individuals, many of whom would not have been affected by their prostate cancer before they died of other causes. To this day, we still don't know if screening with PSA is effective at reducing mortality from this disease. Even lung cancer, which is usually thought of as a rapidly fatal disease, is not infrequently discovered incidentally at autopsy. As we improve our ability to detect lung cancer with spiral CT, we will invariably detect many cases that would not have otherwise become clinically significant.
It sounds as if there is a very large potential for unnecessary treatment?
You have to be very selective in your use of treatment. And there are forces against being selective. Once someone is informed that they might have a serious condition, it's very hard to say, "Okay well just relax and we'll watch it. Then if a couple of years from now you want something done . . ." A person in that situation is often thinking about his or her grandmother or mother who had this condition and had a bad outcome. One of the dictionary definitions of cancer is "a tumor, the natural course of which is fatal." And that's what most people will be thinking. Once you're threatened like that, it's extremely hard to be objective and rational about risks and benefits.
If these factors were broadly understood, might people be less enthusiastic about screening?
Informed consent of screening has been tried. Although it's amazing that so little has been done. Researchers in Virginia examined the effect of informed consent on men's attitudes about prostate cancer screening. The informed consent included objective information about their risks for developing and dying from the disease, about the effectiveness of screening (unknown), and about side effects of the treatment. These men were much less interested in screening than a control group that had not been informed.1
In 1993 you published an article in the New England Journal of Medicine that addressed many of these issues.2 Has understanding of these issues improved since then?
Yes and no. It's like asking, "Are people healthier today?" There are probably more individuals throughout the scientific community who have become aware of this issue. And I know the NCI is much more aware about evaluating the benefits not only of the screening, but also of early treatment. However, if you look at the general medical community and the general population, I don't think they are more aware. There may be more written about the potential downside of early detection, but there's more written about everything. If anything, there is a poorer understanding of the complex screening issues simply because there are more tests and diseases to consider and thus less time to devote to any one of them. And there's still a tremendous amount of misinformation in the popular media.
Furthermore, screening is the politically correct thing to do. A couple of years ago [in 1997], the NIH commissioned a consensus panel on breast cancer screening in women 40 to 49 years of age.3 After reviewing all the scientific information available, the panel basically said that screening was a close call and that each woman in this age group should discuss the benefits and harms with her physician and decide for herself. The NCI and most other medical organizations couldn't deal with that uncertainty. In addition, the US Senate unanimously voted that mammographic screening in this age group does work and is worthwhile. So now screening in this age group is being promoted more aggressively than ever and very little is being done to properly inform these women about the pros and cons of starting at age 40 instead of age 50.
How does the continuous flow of technological improvements impact this problem?
We are always finding earlier and earlier sources of disease, and some of the technology can give very detailed pictures. But people tend to develop too much faith in the technology, and they leave it to the technology to answer the question, "Is this a real cancer or is this a pseudo-cancer?" Someone might argue from the perspective of technological sophistication, "Oh we have advanced ways to look at the genetic makeup of it," or something like that. But more often than not, the real bottom line is, "It's beyond your comprehension. So trust us." And that's what most people do, because they don't know. And they don't have the time to truly make an independent judgment.
Why is it that very few diagnostic tests, mammography being among those few, have undergone randomized clinical trials?
It's very hard to do a randomized trial for a diagnostic test. Randomized trials have been accepted for treatments for quite a long time. Everybody recognizes that they are appropriate for treatment but most people haven't really thought through the implications for diagnostic testing. Most people don't realize that you need the randomized trials, ideally, just as you do for treatment, because if anything it's harder to figure out questions about testing than it is about treatment. More biases can creep in when assessing diagnostic testing, and it becomes extremely difficult to prove anything about diagnostic testing because you need a huge number of patients.
In the case of treatment, if the disease or condition that you are managing is a very serious one in which people have bad outcomes in a short period of timeand if you come up with a new effective treatmentyou'll see a difference in a very short period of time. You only need maybe 20 patients in each limb of the trial. You follow them for a couple of years and you've got your answer. With screening, you're starting with a very healthy population. Only a small percentage of that healthy population will develop the disease that you're looking for. And they'll develop it many years from when you start looking. So that's why it takes tens if not hundreds of thousands of subjects to answer questions about whether or not a screening intervention works.
So it is true that very few randomized trials have been done, and we don't really know what we are accomplishing with a lot of our testing. But, to be fair, it is extremely difficult to perform randomized clinical trials for decisions about testing.
What should be done?
I think we need to develop better tools of getting informed consent from patients before testing. It's been neglected for too long. We have standard consent forms for treatments, but virtually no institutions or organizations have put anything together for informing patients about screening. And there are three good reasons for doing so.
First of all, unlike the treatment scenario where patients are usually suffering, under quite a bit of duress, and are likely to have a hard time making a decision, in the screening scenario, patients are healthy and have the comparative luxury of time to think clearly about what their options are.
Another reason is that the issues are very confusing with screening. Generally we know that the public is misinformed about the true risks of contracting disease and of treatment. They also tend to be misinformed about the effectiveness of treatment. So this needs to be clarified for all types of screening.
Third, when a patient comes to the medical system with a symptomatic disease there is a clear burden on the medical system to do its best to alleviate the patient's problem. In the scenario of screening, however, the person is basically fine and the medical system is trying to bring them in by telling them they are not going to be fine if we don't do something to them. It's the flip side of treatment, and according to the basic principle of medicine, "first do no harm," there should be a higher standard of proof that the screening works, and we should do a much better job of informing people about the true downside.
REFERENCES
- A. M. Wolf et al., "The impact of informed consent on patient interest in prostate-specific antigen screening [see comments]," Archives of Internal Medicine 1996;156(12): 1333-6.
- W. C. Black and H. G. Welch, "Advances in diagnostic imaging and overestimations of disease prevalence and the benefits of therapy," New England Journal of Medicine 1993;328(17):1237-43.
- S. W. Fletcher, "Whither scientific deliberation in health policy recommendations? Alice in the Wonderland of breast-cancer screening," New England Journal of Medicine 1997;336(16):1180-3.
Telepathology surfs the Web
Telemedicinethe use of computer and communications technologies to enhance and enable the delivery of healthcare services to remote locationshas become an increasingly standard application within medical diagnostics, in large part because so many medical imaging techniques are now digitally oriented.
Telepathology is one of the earliest examples of telemedicine, dating back to real-time transmission of black-and-white histological images between Massachusetts General Hospital (Boston, MA) and Logan Airport (Boston) in 1968. By the mid-1980s, dynamic telepathology systems incorporating color video and robotic microscopy enabled real-time transmission of images of permanent sections, cytology specimens, and frozen sections. Today telepathology applications include remote primary diagnosis, digital image analysis, interactive expert consultation, continuing medical education, and distributed case databases.
A typical telepathology system configuration requires a high-end color video camera attached to a confocal microscope, plus transmission equipment and software packages for image manipulation, management, and archiving. The camera captures the specimen as an industry-standard video signal via a video-capture board on a computer workstation; the image can then be stored and transmitted from a high-end or even desktop-based workstation to other computers on a hospital network or to remote sites using standard telecommunications protocols. Complete telepathology systems that incorporate videoconferencing, workstations, cameras, and software cost $35,000-$75,000.
Dynamic telepathology systems that use robotic microscopes allow remote manipulation of pathology slides, enabling a consulting pathologist to render a diagnosis while working with a specially trained technician at the referring end. This application has been enhanced recently with the development of Web-based technologies that enable images to be transmitted and viewed in real time via the Internet. The goal is to bridge the growing gap between where pathology information is generated and where it is used.
For example, pathologists and informatics specialists at the University of Pittsburgh (Pittsburgh, PA) have been working with inexpensive digital image-capture technology on the Web since 1993 and have developed ongoing programs in Web-based distance education, clinical reports, and pathology consultation. This includes building a digital imaging system and Web server directly into their laboratory information system, thereby creating a tool that integrates text and images automatically into clinical reports and disseminates those materials worldwide.
Data compression is key
The transmission and management of electronic medical images requires enormous amounts of bandwidth, which can be costly for healthcare organizations both large and small. In order to reduce transmission times and the need for high-cost telecommunication lines, data-compression technologies have become a key component of image transfer and archiving products since their introduction in the early 1990s.
Generally speaking, there are two types of data compression: lossless and lossy. Lossless compression, which refers to compression ratios of less than 3:1, preserves all of the image data, reconstructing the image bit by bit after transmission. With lossy compression, which exceeds a 3:1 compression ratio, some of the original image data are lost. Physicians were initially uncomfortable with lossy compression, fearing that the loss of too much image data would inhibit their ability to make accurate diagnoses and thus put them at risk of malpractice. Numerous studies over the last few years have, however, laid to rest most of these fears by demonstrating that lossy compression results in little degradation of diagnostic quality, even at ratios of 30:1.
The first compression products to be used with medical images were based on the Joint Photographic Experts Group (JPEG) algorithm. JPEG uses a type of compression known as discrete cosine transformation, which represents the image based on its spatial frequency components. Image areas of very fine detail or sharp edges have higher spatial frequencies than areas of coarse detail. However, while JPEG algorithms can be implemented with a minimum amount of image degradation, compression ratios are generally limited to about 10:1 or less.
The development of wavelet-based compression algorithms has improved the transmission of large image files in the hospital environment. Although wavelet compression is a lossy technique, more data can be compressed with less data loss, primarily because each image is compressed as a whole rather than being broken up into 8 x 8-pixel blocks, as with JPEG. The result is three times as much compression as JPEG, with the same image quality.
Wavelet compression algorithms also have opened the door to image transmission and management via desktop computers and the Internet, which can further reduce costs. In the past year, several products have emerged that use wavelet-based compression to enable large data files to be transmitted on a pixel-by-pixel basis, instead of sending the entire file at one time. These products also feature wavelet-based processing algorithms that generate partial spatial transforms of areas of interest in an image. This "streaming-by" capability makes the most important part of each image available for viewing first. Additional data behind the displayed image are "painted in," allowing users to gain a snapshot of the requested image in three to four seconds while other detail transmits in the background. If more detail is desired, a click of the mouse takes the user to the next level of resolution.
These algorithms do not convert or compress the original image data, nor is a copy of the original image created or physically transmitted to the remote site. Rather, the server does the image processing itself, eliminating the need for an intermediate archive or high-bandwidth communication lines.
Commercial telepathology products that use Web-based technologies in similar ways are now coming onto the market. For example, Dynamic Healthcare Technologies (Lake Mary, FL) earlier this year introduced WebSight for Pathology, an information system designed to enable clinicians to view patients' pathology reports (including images) through direct Web access or have reports delivered to them via e-mail or their personal Internet portal.
Kathy Kincade,Contributing Editor