(Chest. 2005;127:1132-1135.)
© 2005
American College of Chest Physicians
Stage of Lung Cancer in Relation to Its Size*
Part 1. Insights
David Yankelevitz, MD;
Juan P. Wisnivesky, MD, MPH and
Claudia I. Henschke, PhD, MD, FCCP
* From the Department of Radiology (Drs. Yankelvitz and Henschke), New York-Presbyterian Hospital-Weill Cornell Medical Center; and Division of General Internal Medicine and Pulmonary, Critical Care, and Sleep Medicine (Dr. Wisnivesky), Mount Sinai School of Medicine, New York, NY.
Correspondence to: Claudia I. Henschke, PhD, MD, FCCP, Department of Radiology, New York Presbyterian Hospital-Weill Cornell Medical Center, 525 East 68th St, New York, NY 10021; e-mail: chensch{at}med.cornell.edu
Key Words: CT • lung cancer • stage • tumor size
 |
Introduction
|
|---|
Renewed interest in screening for lung cancer has been sparked by advances in CT technology that now allow for the detection of smaller tumors than previously was possible with chest radiography (CXR). This interest has been addressed by a group led by Patz and Goodman (P/G group),12 whose fundamental idea has been that1:
The fundamental [presumption] on which low-dose CT screening is based predicts and requires that small 5–10-mm lesions observed on CT have not metastasized, and that these lesions are very different from the 1–3-cm lesions detectable on chest radiographs. This reasoning suggests that smaller lesions should have a very different, earlier stage distribution profile compared with larger lesions, which are presumed to have a more advanced stage distribution (ie, stage shift).
To test this presumption, the P/G group analyzed data from the cancer registry of their hospital. The evidence did not support the stage-shift presumption, and on this basis they cautioned that "the detection of small tumors using screening CT may not result in a shift to an earlier stage distribution." They went on to assert that a "reduction in mortality needs to be demonstrated by the appropriate clinical trials." This study followed an earlier one that focused on stage IA cases (< 3 cm in diameter), essentially from the same registry.2 In that earlier study, the P/G group found no correlation between tumor size and 5-year survival, and this already led them to assert the need for a randomized controlled trial (RCT).
These ideas, studies, and consequent assertions have had a dramatic impact. Most notably, they provided the justification for the National Lung Screening Trial.3 To quote its principal investigator,3 "conventional wisdom suggests that the smaller the tumor when it is found, the more likely the chance of survival, but that remains to be proven"; and according to the National Lung Screening Trial protocol4:
Simply stated, it is not known whether the detection and treatment of a 5 mm diameter T1 cancer will affect a better outcome than the detection of a 20 mm diameter T1 cancer. A recent retrospective study found no survival advantage among T1 cancers based on size ... A major impetus to move from chest radiograph screening to CT screening for lung cancer is the promise of detecting smaller lung cancers. Yet, we do not currently know that outcomes are necessarily better when the cancer is 2 mm as opposed to 20 mm. As purveyors of public policy, we are obligated to avoid the premature endorsement of a screening process before its benefit and liabilities have been reconciled.
Given their impact, these two articles by the P/G group call for careful scrutiny. In this article, we critically review them as to their merits. In part 2 of this article, we present data from the Surveillance, Epidemiology and End Results database, compellingly contradicting the evidence and conclusion presented by the P/G group regarding the relationship of stage to size. In another report,5 we already have given direct evidence that refutes the conjecture of the P/G group that tumor size does not bear on survival with stage I lung cancer.
|
Use of Registry Data
|
|---|
Let us assume that small tumors documented in a cancer registry indeed have the same stage distribution as do larger tumors. Would this imply similar findings for a screening database? The correct answer is "No." In general, cases get to be documented in a registry on diagnosis prompted by symptoms. Lung cancer causes symptoms by two basic mechanisms: by local effects of the tumor (compression or invasion), and by its generally distant metastases. Smaller tumors are less likely to locally cause symptoms and are therefore more likely to be symptomatic because of their distant metastases (Fig 1
). This dilutes the size/stage correlation within a registry database relative to that in a screening database, possibly even inverting a positive correlation to a negative one.
View larger version (13K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1. Size selectivity in registration of lung cancer. Stage I cases (definition operational only) tend to become symptomatic and hence diagnosed if they are large; stage II+ become symptomatic regardless of size.
|
|
This selection bias in registry data were actually acknowledged by the P/G group itself,1 as they remarked that "the data may not accurately reflect those of a screening program, because both symptomatic and asymptomatic patients were included in the study," but they dismissed the relevance of this on the grounds of "the similarity between the disease stage distribution in the current study and that published in the prevalence screen data from the CT screening trials." To us, these databases are seriously incomparable, and any similarity in their stage distributions is merely a matter of chance in the context of how cases got to be documented in the registry of the P/G group (see below) rather than a meaningful scientific finding.
|
The Type of Registry Used
|
|---|
In the size category of 2 to 3 cm, approximately 80% of the cases of the P/G group were of stage I, and approximately 90% of the them (all cases) were resected.1 These proportions are much higher than those in other experiences (including in the Surveillance, Epidemiology and End Results experience addressed in the accompanying second article). This suggests that advanced-stage cancers commonly were not registered, presumably for the reason that the registry was in essence a surgical one. Such a registry does not give a fair picture of the size/stage relationship in a general cancer registry experience.
|
The Evidence Presented
|
|---|
The P/G group did not present evidence indicating that a relationship between tumor size and disease stage does not exist. They merely failed to find a statistically significant relationship between size and stage, notably when limiting the data analysis to tumors < 3 cm in diameter. The size restriction limited their ability to detect whatever size/stage relationship might have existed in the downward-biased experience of their registry. That few of the tumors were < 1 cm in diameter was to the same effect.
|
"Conclusions" and Discussion
|
|---|
From their study on the correlation between size and stage,1 the P/G group presented as the "conclusion" that the "detection of small tumors using screening CT may not result in a shift to an earlier stage distribution [emphasis added]." In the same vein,1 they asserted that "[evidence] suggests that tumors that are going to metastasize may do so before they are visible radiologically or amenable to intervention. Those tumors that are more clinically indolent, without evidence of metastatic disease, may remain so despite their size [emphasis added]." And subsequently again,6 the "assumptions that size correlates with biological behavior and that small lesions are equivalent to early-stage disease have not been confirmed for lung cancer. Tumors may already have demonstrated their potential to remain localized or to metastasize by the time they are visible on CT imaging [emphasis added]."
Technically, that conclusion is not even of the form of a conclusion. Rather, it merely states the inconclusiveness of their study. What is insinuated is this: when latent cancers become large enough to be detected by CT, the ones among them that have not yet metastasized will remain nonmetastatic until they reach 3 cm in size, subsequent to which these too are prone to metastasize. Let us assume that the proportion of early stage cancers among all latent cancers indeed is constant in that range of size and ask, does this imply that cancers are not metastasizing in that range of size? In Figure 2
, we show all the possibilities that can occur as tumors grow. Figure 2 demonstrates that it is entirely possible for the proportion of early stage cancers among all latent cancers to remain constant with increasing cancer size even when metastases do occur in the size range at issue. This results from cancers within that size range no longer remaining latent. Thus, even if the P/G group had demonstrated absence of size correlation of stage in that range of size for latent cancer, they could not have inferred from this finding an absence of metastasizing in that size range.
View larger version (50K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2. Course of cancer as it grows from 1 to 3 cm in diameter. At small size, a relatively large proportion of cancers are latent. As cancers grow over time, some of them advance in stage, some will become patent, and some will cause death. Nevertheless, among the continually latent cancers, the proportion of early stage cancers can remain stable; it can even decrease. L = latent (asymptomatic); P = patent; I = stage I; II+ = anything beyond stage I; D = dead.
|
|
|
Implications and Credibility
|
|---|
Let us go even further and assume that the insinuation stated above is correct (Fig 3
). It would then be appropriate for a patient having a small CT-detected stage I lung cancer to delay treatment until the tumor reaches 3 cm in diameter. For, so long as the tumor is < 3 cm, it is locked into this stage and therefore there would be no advantage to early treatment, while there would be an advantage to delaying life-threatening surgery. We know of nobody recommending a delay in treatment for small stage I lung cancers; it is universally accepted among clinicians that recommending such a delay would be unacceptable. The idea of no stage progression in that range of size is patently devoid of credibility.
View larger version (8K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3. The conjecture of the P/G group of the course of lung cancer. Metastases occur before CT detectability or after detectability by both CT and CXR, but not in the size range of detection by screening of either type, up to 30 mm.
|
|
|
The Putative Requirement
|
|---|
The P/G group posited, as we stated at the onset, that for CT screening to be beneficial, the smaller, CT-detected cancers must have an earlier stage distribution than the larger CXR-detected cancers. Let us presume, however unrealistically, that there is no such stage shift and consider a subject enrolled in a screening trial contrasting CT with CXR in annual screening. If the subject has a cancer below the size necessary for detection, he or she is screened again 1 year later. As the size threshold for detection is significantly smaller for CT than CXR, it is far more likely that a cancer not detected in the CXR study arm grows beyond the 3 cm threshold during the 1-year interval between screenings, relative to a cancer not detected in the CT arm. As an example, let us consider an adenocarcinoma with a volume doubling time of 180 days. It would take a 2-cm cancer < 1 year (315 days) to grow to 3 cm, while for a 0.5-cm cancer it would take > 4 years (1,400 days) to grow to 3 cm. Thus, CT screening would still be beneficial, as a much higher percentage of cancers would be detected by repeat screening before it crossed the 3-cm threshold, after which stage progression occurs even by the admission of the P/G group.1
|
Need for an RCT
|
|---|
In both of their articles,12 the P/G group pronounced the need for an RCT to resolve the tenability of their putative presumptions in thinking that CT screening might be superior to CXR screening, these presumptions being that there must be a correlation between tumor size and stage and/or tumor size and 5-year survival within stage I. As they failed to verify those presumptions in both of their articles, they asserted the need for an RCT contrasting the two types of screening. No rationale was given whatsoever as to how this is implied from their work. The need for an RCT is not a logical consequence of their failure to validate their presumption.
|
The Articles Overall
|
|---|
Even though those two articles12 by the PG group have had great influence on policy thinking regarding CT screening for lung cancer, notably as to research on the subject, we here present insights of erroneous presumptions, irrelevance of evidence, and unjustifiability and untenability of conclusions. While there may be legitimate discussions as to the type of research necessary to evaluate screening CT, in no way do we find that those articles provide rationale to perform an RCT comparing CT vs CXR.
|
Acknowledgements
|
|---|
The authors gratefully acknowledge Professor Olli Mietinen for his thoughtful comments and review of this manuscript.
|
Footnotes
|
|---|
Abbreviations: CXR = chest radiography; P/G group = Patz and Goodman; RCT = randomized controlled trial
Received for publication May 14, 2004.
Accepted for publication November 3, 2004.
|
References
|
|---|
- Heyneman, LE, Herndon, JE, Goodman, PC, et al (2001) Stage distribution in patients with a small (
3 cm) primary nonsmall cell lung carcinoma. Cancer 92,3051-3055[CrossRef][Medline]
- Patz, EF, Jr, Rossi, S, Harpole, DH, Jr, et al Correlation of tumor size and survival in patients with stage IA non-small cell lung cancer. Chest 2000;117,1568-1571[Abstract/Free Full Text]
- Aberle D. Fifty thousand current and former smokers needed for National Lung Screening Trial. Available at: http://www.cancer.gov/newscenter/nlst. Accessed March 11, 2005
- Aberle DR, Black WC, Goldin JG, et al. Contemporary screening for the detection of lung cancer protocol [NLST], 10 May 2002. American College of Radiology Imaging Network (ACRIN sharp6654). Available at: http://acrin.org/current_protocols.html. Accessed March 14, 2005
- Wisnivesky, JP, Yankelevitz, DF, Henschke, CI The effect of tumor size on curability of stage I non-small cell lung cancers. Chest 2004;126,761-765[Abstract/Free Full Text]
- Patz, EF, Jr, Goodman, PC, Bepler, G Screening for lung cancer. N Engl J Med 2000;343,1627-1633[Free Full Text]
This article has been cited by other articles:
|
|
|
|
 
The International Early Lung Cancer Action Program
Computed tomographic screening for lung cancer: the relationship of disease stage to tumor size.
Arch Intern Med,
February 13, 2006;
166(3):
321 - 325.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
Introduction
Use of Registry Data
