| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Viewpoint |
1 From the Department of Medicine, Weill Medical College of Cornell University, New York, NY, the Department of Epidemiology and Biostatistics, and the Department of Medicine, McGill University, Montreal, Quebec, Canada (O.S.M.); and the Department of Radiology, New York Presbyterian Hospital-Weill Cornell Medical Center, 525 E 68th St, New York, NY 10021 (C.I.H.). Received October 12, 2000; revision requested October 24; revision received December 29; accepted January 17, 2001. Address correspondence to C.I.H. (e-mail: chensch@med.cornell.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT IntroductionContrasting Beliefs Brought to...Foundations of NihilismCritical Examination of the...Foundation for Inferring...Critical Examination of the...Cautions Extended to CT...Demand for RCTs Extended...REFERENCES |
|---|
Index terms: Computed tomography (CT), utilization • Lung neoplasms, CT, 60.12111 • Lung neoplasms, diagnosis, 60.32 • Lung neoplasms, screening • Lung neoplasms, staging
|
Introduction |
|---|
|
TOPABSTRACTIntroductionContrasting Beliefs Brought to...Foundations of NihilismCritical Examination of the...Foundation for Inferring...Critical Examination of the...Cautions Extended to CT...Demand for RCTs Extended...REFERENCES |
|---|
In his or her mind, however, the practitioner is aware of countervailing ideas. Screening with 1970s-type radiography is said to have been demonstrated not to "save lives," not to have any effect on lung cancer "mortality" (1,2). Moreover, that type of screening already is said to involve a notable problem of overdiagnosis, which is defined as innocent lesions that are not life threatening leading to a diagnosis of cancer and, thereby, to unnecessary resection (1). Further, as outlined later, this problem is said to be even more serious with respect to modern computed tomography (CT)-based screening. The practitioner also knows that official recommendations against screening for lung cancer, adopted in the United States a decade ago (2–4, among others), are still in force.
Considerable screening with the traditional type of radiography has continued after the official recommendations against it, and the recent demonstration of the attainability of much earlier diagnosis with modern CT screening (5,6) has led to considerable demand for, and indeed offering of, CT in the United States. In Japan, radiographic screening for lung cancer has been offered as a matter of public health policy for quite some time already.
Thus, to quite an extent, the heart is dominating over the mind, which suggests that the practitioner does not always find the nihilistic recommendations and cautionary declarations compelling. Nor do we, and in this article we make an attempt to help the practitioner appreciate the principal weaknesses in those ideas.
|
Contrasting Beliefs Brought to Focus |
|---|
|
TOPABSTRACTIntroductionContrasting Beliefs Brought to...Foundations of NihilismCritical Examination of the...Foundation for Inferring...Critical Examination of the...Cautions Extended to CT...Demand for RCTs Extended...REFERENCES |
|---|
There is not nor can there be any question but that radiologic screening, including that before the advent of CT, provides for such early diagnosis of lung cancer. The question is, first, whether earlier diagnosis is desirable. If so, then questions arise about the desirability and justification of the costs of pursuing early diagnosis.
Faced with this case, the radiologist who believes early diagnosis is desirable is pleased that the imaging was performed and joins in recommending early immediate resection. She or he gives no thought—and no serious consideration—to delaying intervention until symptoms and/or signs appear. For he or she is aware that stage I lung cancer is generally known to be resectable and is commonly believed to be curable by means of early resection, in sharp contrast to the outcome after waiting until symptoms and/or signs emerge.
But is this practitioner misguided? If the practitioner believes that early intervention for lung cancer does not serve to improve the dismally high (near 90%) case fatality rate, then she or he should also understand that diagnosis before symptoms or signs, whether prompted by screening or imaging for other reasons, has no virtue relative to diagnosis after the signs and symptoms have appeared. In this case, then, the practitioner should not translate early diagnosis to early intervention but should delay intervention until symptoms and/or signs appear. For belief in the putative uselessness of screening in preventing death due to lung cancer also requires belief that intervention before symptoms or signs is no more effective in preventing a fatal outcome than is intervention after signs and symptoms appear.
Now suppose that a practitioner actually does submit to the nihilistic idea that screening, and thereby early intervention, is not effective and thus ignores the early diagnosis and sets out to wait until symptoms and/or signs appear. Would lawyers not regard this as a blatant case of "missed" lung cancer—which, as all American radiologists know, is one of the principal types of malpractice judgments against them? And if the lawyers do regard it in this way, would a successful defense be accomplished by appealing to official nihilism?
If the practitioner does not take the nihilism seriously, he or she would naturally recommend immediate (early) intervention. But the practitioner also must face the larger question of whether to recommend the pursuit of early diagnosis—screening, that is—in suitably defined persons at high risk and with long life expectancy. For this, the practitioner must have a view about the cost-effectiveness of such screening (7).
|
Foundations of Nihilism |
|---|
|
TOPABSTRACTIntroductionContrasting Beliefs Brought to...Foundations of NihilismCritical Examination of the...Foundation for Inferring...Critical Examination of the...Cautions Extended to CT...Demand for RCTs Extended...REFERENCES |
|---|
Most recently, this nihilism has been founded on experience in the extension of follow-up in the Mayo Lung Project (MLP) to 20 years and more (8). The authors concluded that "extended follow-up of MLP participants did not reveal a mortality reduction for the intervention arm," where "intervention" refers to screening-based early diagnosis and "intervention arm" refers to the intensively screened cohort in a randomized clinical trial (RCT), contrasting such experimentally applied screening with the less intensive screening in actual practice. In accord with the authors’ conclusions, the accompanying editorial (9) stated that "after more than 76,000 person-years of observation in each group, there still was no statistically significant difference in lung cancer mortality." And in the same issue of the Journal of the National Cancer Institute, a News article (10) further underscored the putative importance of this latest finding.
Thus, the key to understanding the foundations of this nihilism is, first, clarity on the meaning of "mortality," notably in the context of RCTs in which experimental screening is contrasted with no screening or with the less intensive screening in usual care. To use that now topical study (8) as an example, in the experimental cohort a total of 337 cases of lung cancer deaths were identified during follow-up from the time of randomization to up to 25 years later. The aggregate follow-up of that cohort’s members, while still alive, was 76,761 person-years. On the basis of these two inputs, the "mortality" rate in this cohort was derived as the ratio of 0.0044 per person-year, or 4.4 deaths per 1,000 person-years. (The essential result was that this was no lower than the corresponding 3.9 deaths per 1,000 person-years in the control cohort.) Thus, the meaning of lung cancer mortality rate here, and indeed elsewhere in RCTs on cancer screening, is that of the average incidence, or the density of those deaths in the cohort’s experience over an arbitrary follow-up period. In that average over that arbitrary period, inherently more contributory is the early part of the follow-up, where larger proportions of the two cohorts are still alive and not yet lost to follow-up. (This is another arbitrary feature of the "mortality.") More on the implications of this mortality concept will be discussed in the next section.
|
Critical Examination of the Foundation of Nihilism |
|---|
|
TOPABSTRACTIntroductionContrasting Beliefs Brought to...Foundations of NihilismCritical Examination of the...Foundation for Inferring...Critical Examination of the...Cautions Extended to CT...Demand for RCTs Extended...REFERENCES |
|---|
To start with natural concerns, the practitioner, even if involved in a community program of screening for lung cancer, is concerned with its justifiability on a case-by-case basis, just as in office practice outside any program. So, as outlined in the context of the preceding scenario, the beginning is the indubitable fact that screening provides for earlier diagnosis and thereby earlier intervention; associated with this is the belief, amply substantiated by evidence (11,12), that lung cancer diagnosed early not only is genuine life-threatening cancer but also is distinctly more curable—less fatal—than the essentially incurable and almost always fatal variety that is diagnosed at the prompting of symptoms and/or signs. Thus, the practitioner is naturally concerned to know, for one, how early lung cancer is diagnosed under a particular regimen of "screening" (of early diagnosis, really). In other words, the first concern is to know the distribution of cases by stage diagnosed under that regimen and, especially within stage I, the distribution by tumor size. Also, the practitioner is concerned to know how much more curable—less fatal—are asymptomatic cancers, especially asymptomatic stage I cancers of various sizes that can be detected at screening, in comparison with what is typical in the absence of screening. For naturally that diagnostic distribution together with that curability pattern jointly imply the curability and fatality rates that pertain to cases diagnosed under that regimen (13,14) in comparison with what is known to prevail in the absence of screening.
Those being the practitioner’s natural concerns, she or he thereby is not concerned to know about the aforementioned mortality rates, which have no meaningful relation to the aimed-for reduction in the case-fatality rate—that is, in the proportion of diagnosed cases (true cases rather than overdiagnosed ones) that, even with appropriate treatment, would eventually result in death due to the disease, given the absence of "competing" causes of death. The manifestation of this fatality rate of interest has a built-in delay consonant with prognostication in practice: When screening an asymptomatic patient, the practitioner is aiming to provide for early intervention for the purpose of preventing a fatal outcome not today or within a year but eventually, years into the future. This is also why the practitioner only screens persons with suitably long life expectancy based on age and other indicators of risk of death due to causes other than lung cancer. On this basis, then, the critically thinking practitioner sees no justification for the inclusion of the earliest deaths from lung cancer in the mortality rates compared in RCTs, because their inclusion only leads to dilution of the "mortality" rate ratio as a reflection of the fatality-rate ratio of relevance to practice.
More to the same effect, the practitioner understands that when screening is discontinued in a particular person, the intended fatality-preventing consequence of its associated early intervention and the assurance against lung cancer fatality inherent in negative screening results bear on potential fatalities for only a limited time after the discontinuation of the screening. And so, the critically thinking practitioner recognizes another dilution in the RCT-based mortality ratio as a reflection of the practice-relevant fatality-rate ratio: With screening discontinued after only 6 years in the MLP, the fatality implications are not at all fully manifest in the mortality ratio that accounts for lung cancer deaths up to 19 years after the discontinuation of screening. While the associated editorial claimed that "the authors provide compelling evidence that a major reduction in the lung cancer mortality was not missed because of insufficient follow-up," the critically thinking practitioner well understands that, apart from the diluting inclusion of early deaths, the manifestation of the aimed-for reduction in the fatality rate was missed because of the excessively long follow-up.
As an illustration of the implications of the necessary restriction of temporal aspects of follow-up for death due to lung cancer in the MLP, it has been shown (14) that had the relevant deaths (potentially preventable by screening-associated early intervention) been confined to the 4–7-year range from the time of assignment to the compared cohorts, the mortality rate ratio, in this interval possibly reflecting the fatality-rate ratio, had a 95% interval estimate ranging down to 0.57, suggesting that the evidence is consistent with a 43% gain in the cure rate. To the extent that this 4–7-year interval represents too early a segment of the follow-up experience, this is in part a consequence of the short duration of the screening: With only 6 years of screening, the fatality reduction from screening-associated earlier interventions may never have been fully manifest in the course of follow-up.
|
Foundation for Inferring Overdiagnosis |
|---|
|
TOPABSTRACTIntroductionContrasting Beliefs Brought to...Foundations of NihilismCritical Examination of the...Foundation for Inferring...Critical Examination of the...Cautions Extended to CT...Demand for RCTs Extended...REFERENCES |
|---|
In the MLP, a substantial proportion of screen-detected cases were probably pseudodisease for three reasons: 1) the mortality rate from all causes is high, about three-fold that in nonsmokers; 2) some squamous cell carcinomas detectable by sputum cytology are very small; and 3) some primary adenocarcinomas detectable by chest radiography grow very slowly.
No explication for this was offered, presumably because it was viewed as unnecessary. The author of the editorial (9) did, however, make the additional point that "overdiagnosis does not reduce disease-specific mortality," and therefore "disease-specific mortality is the most valid for the evaluation of screening effectiveness." The associated News article (10) in the journal was entitled, "Lung Project Update Raises Issue of Overdiagnosing Patients." In this article, the lead author of the MLP article is described as having been "able to look for overdiagnosis" and as having remarked that the MLP conclusion that there was overdiagnosis in its screening arm is "totally heretical" to many advocates of lung cancer screening.
|
Critical Examination of the Foundation of Overdiagnosis Inference |
|---|
|
TOPABSTRACTIntroductionContrasting Beliefs Brought to...Foundations of NihilismCritical Examination of the...Foundation for Inferring...Critical Examination of the...Cautions Extended to CT...Demand for RCTs Extended...REFERENCES |
|---|
The practitioner scarcely accepts the point that better survival among cases diagnosed with the aid of screening indicates possible overdiagnosis. After all, the idea in screening is the pursuit of earlier diagnosis of real cancer, and the attainment of this, even without overdiagnosis, means prolonged survival as of diagnosis. This is well understood to be inherent in the "lead time" that screening is supposed to provide.
To be sure, survival results specific to early-stage resected lesions were also given, focusing on those diagnosed 7–12 years after randomization (so that the inclusions extended beyond the screening period, with resulting dilution). The point estimates of the median periods of survival indeed were quite different between the screening- and symptom-detected cases, 16 versus 5 years, "but the difference was not statistically significant" (P = .16). Again, however, the critically thinking practitioner understands that screening-detected stage I disease—real malignancy—should actually be associated with longer survival than stage I disease diagnosed in the absence of screening: Not only are patients with screening-detected disease asymptomatic and sign free, but the diagnosis of stage I cancer is also established earlier by virtue of smaller lesion sizes in this stage. Longer survival of patients with screening-detected stage I cancer relative to survival in those with symptom-detected cancer, even if statistically significant, thus is no indication of overdiagnosis among the former.
Whereas all of this is, just as is the foundation for nihilism about effectiveness, mere pseudoevidence associated with specious reasoning in the inference, the true manifestation of overdiagnosis occurs in the follow-up of unresected early-stage cases diagnosed at screening. Such follow-up has indeed been documented (11), and the evidence compellingly indicates that screening-detected stage I lung cancer, when left unresected, has a completely malignant course. Part of the experience in that study (11) was derived from the screening experience in the MLP; nonetheless, in the rest the screening was of the MLP type. The evidence clearly indicates that the MLP screening did not lead to overdiagnosis. The result in this study is in full accord with that of the only other similar study (12) of which we are aware.
As the critically thinking practitioner then turns to the editorial author’s (9) aprioristic reasons with regard to why, in the MLP, "a substantial proportion of the screen-detected cases were probably pseudodisease," sheer bafflement may set in. In seeking to resolve this, the practitioner must appreciate that by overdiagnosis the author means detection of mere pseudodisease, which the author defines as a "subclinical condition that would not have produced signs or symptoms before the individual died of other causes." To illustrate this concept, a known case of still latent but genuinely aggressive cancer in a living person cannot yet be said to be a case of real disease, for it may turn out to be a case of mere pseudodisease, and death may turn out to be caused by surgery or car accident, say, before symptoms or signs would have appeared (in the absence of early resection). The author’s aprioristic statement thus has no meaning to the practitioner who shares our concept of real cancer of the lung: It is ultimately fatal in the absence of intervention and "competing" causes of death. In these terms, overdiagnosis is labeling as lung cancer a lesion whose natural history (prospective) does not end in a fatal outcome.
Some of the components in that editorial passage do have screening relevance but outside the genuine topic of overdiagnosis. For one, the prospects for dying of "competing" causes bears on the indication for screening, its beginning and ending. And if the diagnosed cancer can be judged to be very slowly growing, then the indication for early intervention may be judged not to exist, notably if the life expectancy otherwise already is rather short. The potential diagnosis of "very small" cancers by means of sputum cytologic analysis, even if it did lead to true overdiagnosis in such screening, obviously has no relevance for radiologic screening, in which the smaller the merrier, as long as what is at issue is real malignancy in the sense of the substitute definition above.
|
Cautions Extended to CT Screening |
|---|
|
TOPABSTRACTIntroductionContrasting Beliefs Brought to...Foundations of NihilismCritical Examination of the...Foundation for Inferring...Critical Examination of the...Cautions Extended to CT...Demand for RCTs Extended...REFERENCES |
|---|
Practitioners suitably versed with CT screening for lung cancer are well aware that CT not only plays a role in the initial screening test but also plays a critical role in the definitive diagnostics after a positive result from the initial test. In particular, CT affords a rapid determination of whether the lesion is indeed growing and at a rate consistent with malignancy (15). Because this determination is feasible before biopsy (and resection), it serves to substantiate the pathologic diagnosis and thus guard against overdiagnosis. In line with this, the author of the editorial envisions "a mandatory observation period for small nodules."
Against this backdrop, the News article (10) quoted the lead author of the MLP report—an epidemiologist at the National Cancer Institute with no experience with CT—as having opined that "given its excellent resolution, spiral CT will identify lesions at a rate much higher than chest x-ray, but some of them might have absolutely no clinical relevance." No reference was made about the use of CT assessment of the rate of growth, which for the earliest lesions is a prerequisite for biopsy and a basis for confirmation of the positive CT result. Caution indeed is called for, including in relation to officials’ public declarations. For it is not that the rate of detection in multiple repeat screenings is much higher; with respect to these mainly relevant repeat screenings, the superior sensitivity of CT screenings means that those detections are achieved much earlier.
|
Demand for RCTs Extended to CT Screening |
|---|
|
TOPABSTRACTIntroductionContrasting Beliefs Brought to...Foundations of NihilismCritical Examination of the...Foundation for Inferring...Critical Examination of the...Cautions Extended to CT...Demand for RCTs Extended...REFERENCES |
|---|
Yet the MLP is now being used as a paradigm for the evaluation of CT-based screening for lung cancer. The authors of the most recent MLP report (8) ended their article by declaring that "before spiral CT is accepted into medical practice, it is critical to determine whether this promising new screening modality ultimately does more good than harm in a randomized, controlled clinical trial with lung cancer mortality as the endpoint." The author of the editorial (9) in turn put it thus: "randomized clinical trials should be performed, and all causes of mortality should be closely monitored to avoid missing a major benefit or harm from the screening process." And the News article (10) quoted the lead author of the MLP report as having remarked (without much of the reserve that is supposed to characterize the scientific mind) that the MLP is "tremendously important in thinking through what data are necessary to properly assess the usefulness of helical computed tomography for screening and whether such a test—if it is released too soon—could cause harm." So, to follow the paradigm, a 30-year RCT with some half dozen years of actual screening should show reduced "mortality" from lung cancer "before spiral CT is accepted into medical practice," in the 2030s perhaps.
Accepted by whom? Surely, a large number of American radiologists have already accepted CT screening for lung cancer in their medical practices—and still steadfastly refuse to accept the idea that lung cancer diagnosed in an asymptomatic person as a result of such screening safely can and therefore should be delayed until symptoms and/or signs appear.
Ultimately, a call for wisdom is in order: "Tomorrow’s radiologists will need to be critical thinkers, learning how to read books and journals and to listen to ’experts’ more sceptically" (16). As is evident from the foregoing, we agree with this precept, and with respect to today’s radiologists, since we find that heeding it so greatly alleviates the practicing radiologist’s heart-mind conflict in decisions about screening for lung cancer.
|
FOOTNOTES |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTIntroductionContrasting Beliefs Brought to...Foundations of NihilismCritical Examination of the...Foundation for Inferring...Critical Examination of the...Cautions Extended to CT...Demand for RCTs Extended...REFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  J M Reich A critical appraisal of overdiagnosis: estimates of its magnitude and implications for lung cancer screening Thorax, April 1, 2008; 63(4): 377 - 383. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 W. C. Black, D. R. Aberle, C. D. Berg, and For the Executive Committee of the National Lung S Large Field Trial for Lung Cancer Screening: Putting the Wrong Cart before the Horse? Radiology, May 1, 2007; 243(2): 314 - 316. [Full Text] [PDF]
|
|
|
|
 |
|
 C. T. Kolber, G. Zipp, D. Glendinning, and J. J. Mitchell Patient Expectations of Full-Body CT Screening Am. J. Roentgenol., March 1, 2007; 188(3): W297 - W304. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. J. Gordon, L. A. Deters, M. D. Nitz, B. C. Lieberman, B. Y. Yeap, and R. Bueno Differential diagnosis of solitary lung nodules with gene expression ratios J. Thorac. Cardiovasc. Surg., September 1, 2006; 132(3): 621 - 627. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. A. Obuchowski and M. T. Modic Screening with Multisection CT: Unmasking the Benefit Radiology, November 1, 2005; 237(2): 401 - 403. [Full Text] [PDF]
|
|
|
|
|
|
G. S. Gazelle, P. M. McMahon, U. Siebert, and M. T. Beinfeld Cost-effectiveness Analysis in the Assessment of Diagnostic Imaging Technologies Radiology, May 1, 2005; 235(2): 361 - 370. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 F. Bianchi, J. Hu, G. Pelosi, R. Cirincione, M. Ferguson, C. Ratcliffe, P. P. Di Fiore, K. Gatter, F. Pezzella, and U. Pastorino Lung Cancers Detected by Screening with Spiral Computed Tomography Have a Malignant Phenotype when Analyzed by cDNA Microarray Clin. Cancer Res., September 15, 2004; 10(18): 6023 - 6028. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. J. Brenner Radiation Risks Potentially Associated with Low-Dose CT Screening of Adult Smokers for Lung Cancer Radiology, May 1, 2004; 231(2): 440 - 445. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C.I. Henschke, D.F. Yankelevitz, D.I. McCauley, D.M. Libby, M.W. Pasmantier, and J.P. Smith Guidelines for the use of spiral computed tomography in screening for lung cancer Eur. Respir. J., January 1, 2003; 21(39_suppl): 45S - 51s. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. K. Shah, J. H. M. Austin, C. S. White, P. Patel, L. B. Haramati, G. D. N. Pearson, M. C. Shiau, and Y. M. Berkmen Missed Non-Small Cell Lung Cancer: Radiographic Findings of Potentially Resectable Lesions Evident Only in Retrospect Radiology, January 1, 2003; 226(1): 235 - 241. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. G. Armato III, F. Li, M. L. Giger, H. MacMahon, S. Sone, and K. Doi Lung Cancer: Performance of Automated Lung Nodule Detection Applied to Cancers Missed in a CT Screening Program Radiology, December 1, 2002; 225(3): 685 - 692. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Vermont Lung Cancer Screening: Necessary Cartesian Doubtfulness Is Not Nihilism Radiology, December 1, 2002; 225(3): 920 - 920. [Full Text] [PDF]
|
|
|
|
|
|
W. J. Casarella A Patient's Viewpoint on a Current Controversy Radiology, September 1, 2002; 224(3): 927 - 927. [Full Text] [PDF]
|
|
|
|
|
|
F. M. Hall, E. F. Patz Jr, W. C. Black, and P. C. Goodman Screening for Lung Cancer: Been There and Done That * Dr Patz and colleagues respond: Radiology, September 1, 2002; 224(3): 928 - 929. [Full Text] [PDF]
|
|
|
|
|
|
E. F. Patz Jr, W. C. Black, and P. C. Goodman Commentary on Drs Miettinen and Henschke's Viewpoint Radiology, December 1, 2001; 221(3): 597 - 597. [Full Text] [PDF]
|
|
|
|
|
|
O. S. Miettinen and C. I. Henschke Commentary on Drs Patz, Black, and Goodman's Viewpoint Radiology, December 1, 2001; 221(3): 598 - 599. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
