Doubling Times and CT Screen–Detected Lung Cancers in the Pittsburgh Lung Screening Study

doi:10.1164/rccm.201107-1223OC

Am J Respir Crit Care Med. 2012 January 1; 185(1): 85–89.

Published online 2012 January 1. doi: 10.1164/rccm.201107-1223OC

PMCID: PMC3262038

Doubling Times and CT Screen–Detected Lung Cancers in the Pittsburgh Lung Screening Study

David O. Wilson,¹ Adam Ryan,² Carl Fuhrman,² Matthew Schuchert,³ Steven Shapiro,¹ Jill M. Siegfried,⁴ and Joel Weissfeld⁵

¹Division of Pulmonary, Allergy and Critical Care Medicine Department of Medicine

²Department of Radiology

³Heart, Lung and Esophageal Surgery Institute; and

⁴Department of Pharmacology and

⁵Department of Epidemiology, University of Pittsburgh, Pittsburgh, Pennsylvania

Corresponding author.

Correspondence and requests for reprints should be addressed to David O. Wilson, M.D., M.P.H., UPMC-Shadyside Place, 580 S. Aiken Ave., Suite 400, Pittsburgh, PA 15232. E-mail: wilsondo/at/upmc.edu

Received July 8, 2011; Accepted September 29, 2011.

See the referenced article on page 8.

This article has been cited by other articles in PMC.

Abstract

Rationale: As computed tomography (CT) screening for lung cancer becomes more widespread, volumetric analyses, including doubling times, of CT-screen detected lung nodules and lung cancers may provide useful information in the follow-up and management of CT-detected lung nodules and cancers.

Objectives: To analyze doubling times in CT screen detected lung cancers and compare prevalent and nonprevalent cancers and different cell types on non small cell lung cancer.

Methods: We performed volumetric and doubling time analysis on 63 non–small cell lung cancers detected as part of the Pittsburgh Lung Screening Study using a commercially available VITREA 2 workstation and VITREA VITAL nodule segmentation software.

Measurements and Main Results: Doubling times (DT) were divided into three groups: rapid (DT < 183 d), typical (DT 183–365 d), and slow (DT > 365 d). Adenocarcinoma/bronchioloalveolar carcinoma comprised 86.7% of the slow DT group compared with 20% of the rapid DT group. Conversely, squamous cell cancer comprised 60% of the rapid DT group compared with 3.3% of the slow DT group. Twenty-eight of 42 (67%) prevalent and 2 of 21 (10%) nonprevalent cancers were in the slow DT group (P < 0.0001; Fisher's exact test). Twenty-four of 32 (75%) prevalent and 1 of 11 (9%) nonprevalent adenocarcinomas were in the slow DT group (P < 0.0002; Fisher's exact test).

Conclusions: Volumetric analysis of CT-detected lung cancers is particularly useful in AC/BAC. Prevalent cancers have a significantly slower DT than nonprevalent cancers and a higher percentage of adenocarcinoma/bronchioloalveolar carcinoma. These results should affect the management of indeterminant lung nodules detected on screening CT scans.

Keywords: lung cancer, doubling times, lung cancer screening

At a Glance Commentary

Scientific Knowledge on the Subject

There is limited information on doubling times and volumetric computed tomography (CT) analysis in CT screening–detected lung cancers.

What This Study Adds to the Field

This study presents information on doubling times and volumetric CT analysis of CT screening–detected non–small cell lung cancer from the Pittsburgh Lung Screening Study and on the management of indeterminant lung nodules detected on screening CT scans.

Computed tomography (CT) screening for lung cancer has been under study for over a decade. With the publication of results from the National Lung Screening Trial (NLST) (1), CT screening for lung cancer may become widespread. We have accumulated over 150 lung cancers as part of the Pittsburgh Lung Screening Study (PLuSS), which started enrolling subjects in 2002 (2). Although there have been several studies of volumetric evaluation of lung nodule doubling times (DTs) for differentiating benign versus malignant pulmonary nodules (3–5), there is less information about DTs in screening-detected lung cancers. We present software volumetric evaluation of DTs in 63 non–small cell lung cancers (NSCLCs) from PLuSS and compare the doubling times in prevalent and nonprevalent cancers.

Methods

PLuSS involved 3,642 subjects and was approved by the Institutional Review Board of the University of Pittsburgh. Written informed consent was obtained from all participants, and the study was compliant with the Health Insurance Portability and Accountability Act. At the time of this study, 148 lung cancers had been diagnosed; 63 were suitable for volumetric nodule analysis.

CT Imaging and Nodule Classification Protocol

CT scans evaluated for volumetric analysis included lung cancer screening studies and clinical studies prescribed for the management of CT-detected abnormalities or intervening symptoms. The lung cancer screening protocol used a single breath-hold, helical, low-dose technique (40–80 mA; 140 kVp) to obtain axial images reconstructed with a high-spatial-frequency lung algorithm at contiguous 2.5-mm intervals. Using standard doses, clinical studies typically recontructed images at 2.5- to 5.0-mm intervals. Volume analysis was performed by a radiology resident (A.R.), a pulmonologist (D.O.W.), and a thoracic radiologist (C.F.). Lung lesions were automatically segmented on a Vitrea 2 workstation (Vitrea Vital Images, Minnetonka, MN) using software supplied with the workstation. The nodule segmentation algorithm is an intensity/density based algorithm that uses tissue density (Hounsfield unit values) to separate nodules from background tissues. To get accurate segmentation of nodules, the algorithm also uses constrained morphological operations to separate nodules from attached vessels and airways. After automatic segmentation, lesion margins were manually corrected, whenever necessary, to include all grossly visible tumor and to exclude adjacent structures such as vessels, chest wall, mediastinum, hilar structures, and atelectasis. Such manual corrections were necessary in approximately 20% of cases. In these cases, the margins were determined based on mild attenuation differences between tumor and air space disease and by excluding linear elements that appeared to represent air space disease. If a distinct nodule could not be manually and reproducibly outlined, the case was excluded. Direct, computer-generated volumes were then automatically displayed.

Two investigators (A.R. and D.O.W.) independently used Vitrea Vital software to complete 126 nodule volume measurements (63 subjects with a tumor nodule measured independently on two CTs separated in time). If a volume measurement differed by more than 5% (n = 15), a third investigator (C.F.) independently measured nodule volumes on both CTs. Data analyses used the geometric mean of the volume measurements for the two investigators in closest agreement.

Cancers visible, possibly in retrospect, as a nodule on an initial or baseline screening CT were considered prevalent. Nonprevalent cancers were not visible on the initial screening CT but were newly visible on a follow-up CT performed for any reason.

Data Analysis

Tumor doubling time estimates used tumor volumes (Vo and Vi) on the first and last available and evaluable pretherapeutic CT, the time interval Ti (in days) between the two exams, and the following formula:

Determinations of the statistical significance of DT differences used standard two-sided, nonparametric rank sum tests (SAS version 9.2) for two-group (Wilcoxon test), nonordered multiple-group (Kruskall-Wallis test), and ordered multiple-group (Jonckheere-Terpstra test) comparisons.

Results

Of the initial 148 lung cancers detected in PLuSS, 63 NSCLCs are included in this analysis (Figure 1). Analyses excluded 85 cases for the following reasons: small cell histology (n = 20), absence of two CT scans more than 90 days apart (n = 41), cancer detected after screening completed (n = 15) with diagnostic CT scans not available for volumetric analysis, and inability to segment the cancer (n = 9). We excluded 15 cases because cancer was detected after screening was completed and diagnostic CT scans were not available for volumetric analysis. We corrected the manuscript accordingly. With respect to the 15 cases in question, eight did not have any nodules detected on any screening CT, five had only tiny (< 0.5 cm) nodules detected on screening CT, and two had indeterminant nodules (the largest was 1.0 cm) detected on screening CT. In the latter two cases, lung cancers were diagnosed at outside institutions 3 and 5 years after the last CT screening. There were three bronchioloalveolar carcinoma (BAC)/ground glass cancers in the combined adenocarcinoma (AC)/BAC group, all of which were amenable to the volumetric nodule analysis. All three cases required hand segmentation and were in the group that all three investigators measured to achieve consistent results. There was no BAC in the excluded group. Only 63 of 128 (49%) NSCLCs were amenable to the volumetric/DT analysis.

Figure 1.

Workflow of included and excluded lung cancers. CT = computed tomography; NSCLC = non–small cell lung cancer; SCLC = small cell lung cancer.

Table 1 compares the 63 evaluable and 65 excluded subjects with NSCLC. The excluded group had fewer adenocarcinomas, a higher proportion of locally advanced stage III and IV cancers (52 vs. 29%), and more deaths.

TABLE 1.

COMPARISON OF EVALUABLE SUBJECTS AND EXCLUDED SUBJECTS

Table 2 summarizes the attributes of the 63 included subjects grouped according to cancer rate of growth, as represented by the nodule DT, rapid growth (DT < 183 d), typical growth (DT 183–365 d), and slow growth (DT > 365 d). There were 10 (15.8%) cancers in the rapid DT group, 23 (36.5%) in the typical DT group, and 30 (47.6%) in the slow DT group. These groups differed according to histology and manner of detection (i.e., primary tumor seen vs. not seen on the prevalent screening CT). AC/BAC cell carcinoma comprised 86.7% of the slow DT group compared with 20% of the rapid DT group. Conversely, squamous cell cancer comprised 60% of the rapid DT group compared with 3.3% of the slow DT group. Of the 42 prevalent cancers, 28 (67%) were in the slow DT group, and only 3 (7%) were in the rapid DT group. Of the 21 nonprevalent cancers, 2 (10%) were in the slow DT group, and 7 (33%) were in the rapid DT group. Expressing results in terms of DT, 7 of 10 (70%) rapid DT tumors were nonprevalent cancer, and 28 of 30 (93.3%) slow DT tumors were prevalent cancer (P < 0.0001: Fisher's exact test) (Table 2). Twenty-four of 32 (75%) prevalent and 1 of 11 (9%) nonprevalent adenocarcinomas were in the slow DT group (P < 0.0002; Fisher's exact test).

TABLE 2.

ATTRIBUTES (PERCENT DISTRIBUTION) OF SUBJECTS GROUPED ACCORDING TO CANCER RATE OF GROWTH, AS REPRESENTED BY THE NODULE DOUBLING TIME

Table 3 shows median DT by manner of detection (prevalent vs. nonprevalent) and histology (AC/BAC vs. squamous cell). Prevalent cancers had a median DT of 514 versus 237 days for nonprevalent cancers (P < 0.0001). AC/BACs had a median DT of 387 versus 160 days for squamous cell cancers (P = 0.0031). Twenty-five percent of AC/BACs had a DT > 711 days, with a maximum of 1,435 days.

TABLE 3.

MEASURES OF THE DISTRIBUTION OF DOUBLING TIME BY MANNER OF DETECTION AND HISTOLOGY

Discussion

We report on 63 subjects with NSCLC and lung tumors that were amenable to segmentation by the VITAL/VITREA 2 algorithm. These cancers represent 42% of the lung cancers detected to date as part of the ongoing PLuSS (Figure 1). The 63 subjects with NSCLC included in the analysis and the 65 subjects with NSCLC excluded from analysis were similar with respect to sex, age, race, smoking history, lung function (GOLD), and a CT measure of emphysema (Table 2). The group excluded from analysis contained more non-AC/BAC, more NSCLC diagnosed at advanced stage (III/IV), and more subjects dying before the end of follow-up. By requiring two CT scans separated by more than 90 days, the DT calculation effectively excluded lung cancer presenting as a large primary tumor, as a central airway abnormality, or as a tumor with lymph node enlargement at screening. This situation, accounting for 29 of 65 excluded cases, appropriately demanded more immediate biopsy intervention and explained the histology, stage, and outcome differences observed between cases excluded and included for DT analysis.

Two factors—histology and manner of detection—differed by DT. AC/BAC histology and prevalent detection characterized cases with slow DT (Table 2). Equivalently, AC/BAC histology and prevalent cases had longer DT (Table 3).

Other studies have looked at DTs in lung cancer (6–10). Several studies have compared DT in squamous cell carcinoma (SCC) and AC, showing significantly shorter DT in SCC than in AC (6–8). Hasegawa and associates reported on DT in early-stage SCC versus AC in a CT screening–detected population from Japan (6). They reported mean DTs for 8 SCC versus 49 AC of 97 versus 533 days. There is one study that showed no significant difference among histologic types of lung cancer and DT; this study included 21 of 149 subjects (14%) that did not grow between the two CT scans (10). Our study is the first to compare prevalent and nonprevalent NSCLC. In light of the recently announced results from the NLST, it is likely that screening CT scans in high-risk individuals will become much more mainstream. If this occurs, many more lung nodules will be found that will require follow-up. The risks of finding indeterminant pulmonary nodules in high-risk individuals include morbidity from overly aggressive diagnostic procedures and radiation exposure from follow-up imaging (2). Tools that allow enhanced prediction of individual lung cancer risk will be critical to the management of CT-detected lung nodules. We and others have published clinical predictive formulas for individual lung cancer risk (11–13). The use of DT data for indeterminate peripheral lung nodules could potentially aid in the risk stratification and follow-up management.

A study by van Klaveren and colleagues has advocated using DT to predict benign versus malignant lung nodules detected on CT screening (14). Focusing on nodules 50 to 500 mm³, these authors report that a DT > 400 days on a follow-up CT scan at 3 months had a point estimate negative predictive value of 99.7%. These data are relevant to indeterminant small pulmonary nodules in a CT screening population. Our data focus on screening-detected lung cancers and show a wide range of DTs (72–4,263 d). Our results are similar to Hasegawa (6) with respect to slower DT for AC. We also found slower DT in prevalent cancers, the majority of which (34/42 [81%]) were ACs (Table 3). The slower DT in prevalent lung cancers is not unexpected. Compared with cancer detected shortly after an initially normal screening, the pool of cases found at an initial screening typically includes a higher proportion of slower-growing prevalent tumors existing in the population for longer and indefinite periods of time. One could argue that the more rapid DT in nonprevalent cancers should be a factor that informs the management of newly detected lung nodules on follow-up CT scans done serially in high-risk patients. It is known that new nodules not previously seen on prior CT scans are more likely to be cancer than indeterminant nodules of unknown duration. Thus, the management of these nodules should be more expectant with shorter-interval follow-up and a lower threshold to biopsy. With respect to using DT data for peripheral indeterminant lung nodules of unknown duration, our data suggest that following nodules detected on prevalence CT scans for 2 years may be inadequate (15). Table 3 shows that 25% of ACs had a DT > 711 days, suggesting that there is a wide range of growth rates for this type of lung cancer. One can hypothesize that in slow-growing prevalent nodules suggestive of adenocarcinoma there is the potential for overdiagnosis and therefore a more conservative diagnostic management approach might be considered in selected patients.

There are reports that CT slice thickness and type of software package can affect the accuracy of volumetric measurements, especially for small (< 10 mm) nodules (16). To minimize this effect, it has been suggested to use the same section thickness and overlap and the same segmentation software for all studies, a protocol we adhered to. In our experience, the VITREA/VITAL nodule segmentation software appears to be robust to slice thickness and other image acquisition parameters. Nodule characteristics (e.g., central location or proximity to vessels or airways), not technical parameters (e.g., slice thickness), explained our inability to segment the nine nodules excluded from volumetric analysis.

We have presented DT data on lung cancer appearing in PLuSS, a study using CT to detect early lung cancer. Using currently available technology, only 63 of 148 (43%) of all lung cancer and 63 of 128 (49%) of NSCLCs were amenable to DT analysis. AC/BAC comprised 46 of 63 (73%) of cancer amenable to DT analysis but 26 of 30 (87%) with slow DT. Squamous cell cancers comprised 8 of 63 (13%) of cancers amenable to DT analysis but 6 of 10 (60%) with rapid DT. Prevalent cancers had significantly slower DT than nonprevalent cancers. These results are most applicable to peripheral indeterminant nodules that are suitable for follow-up with serial CT scan, in contrast to more suspicious nodules that are more appropriate for immediate biopsy. As the field of quantitative CT analysis evolves with improved software and computerized segmentation algorithms incorporating freehand sketching of nodules into automated segmentation, volumetric nodule analysis will be easier and more reproducible. We believe the results presented in this manuscript represent a tip-of-the-iceberg phenomenon with respect to quantitative CT analysis of lung nodules, with future studies incorporating new radiology informatics and imaging biomarkers. If the promise of the NLST is to be fulfilled, the many indeterminant lung nodules detected by CT screening will call for a more standardized approach to their clinical management. These results should affect the management of peripheral indeterminant lung nodules detected on screening CT scans.

Supplementary Material

Disclosures

Click here to view.

Footnotes

Supported by the University of Pittsburgh Lung Cancer SPORE grants NCI P50-CA90440 and 1P50 HL084948, University of Pittsburgh Cancer Institute, and University of Pittsburgh Medical Center.

Author Contributions: Conception and design, D.W., C.F., J.S., J.W.; analysis and interpretation, D.W., A.R., C.F., M.S., S.S., J.W.; drafting manuscript and intellectual content, D.W., S.S., J.S., J.W.

Originally Published in Press as DOI: 10.1164/rccm.201107-1223OC on October 13, 2011

Author disclosures are available with the text of this article at www.atsjournals.org.

References

1. The National Lung Screening Trial Research Team Reduced lung-cancer mortality with low-dose computed tomographic screening. N Engl J Med 2011;365:395–409. [PubMed]

2. Wilson D, Weissfeld J, Fuhrman C, Fisher S, Balogh P, Landreneau R, Luketich J, Siegfried J. The Pittsburgh Lung Cancer Screening Study (PluSS): outcomes within 3 years of prevalence CT scan. Am J Respir Crit Care Med 2008;178:956–961. [PMC free article] [PubMed]

3. Revel MP, Merlin A, Peyrard S, Triki R, Couchon S, Chatellier S, Frija G. Software volumetric evaluation of doubling times for differentiating benign versus malignant pulmonary nodules. Am J Roentgenol 2006;187:135–142. [PubMed]

4. Petrou M, Quint L, Nan B, Baker L. Pulmonary nodule volumetric measurement variability as a function of CT slice thickness and nodule morphology. Am J Roentgenol 2007;188:306–312. [PubMed]

5. Quint L, Cheng J, Schipper M, Chang A, Kalemkerian G. Lung lesion doubling times: values and variability based on method of volume determination. Clin Radiol 2008;63:41–48. [PubMed]

6. Hasegawa M, Sone S, Takashimi S, Yang ZG, Maruyama Y, Watanabe T. Growth rate of small lung cancers detected on mass CT screening. Br J Radiol 2000;73:1252–1259. [PubMed]

7. Lindell R, Hartman T, Swensen S, Jett JR, Midthun DE, Tazelaar HD, Madrekar JN. Five year lung cancer screening experience: CT appearance, growth rate, location and histologic features of 61 lung cancers. Radiology 2007;242:555–562. [PubMed]

8. Winer-Muram H, Jennings S, Tarver R, Aisen AM, Tann M, Conces DJ, Meyer CA. Volumetric growth rate of stage I lung cancer prior to treatment: serial CT scanning. Radiology 2002;223:798–805. [PubMed]

9. Aoki T, Nakata H, Watanabe H, Nakamura K, Kasai T, Hashimoto H, Yasumoto K, Kido M. Evolution of peripheral lung adenocarcinomas; CT findings correlated with histology and tumor doubling time. Am J Roentgenol 2000;174:763–768. [PubMed]

10. Jennings S, Winer-Muram H, Tann M, Ying J, Dowdeswell I. Distribution of stage I lung cancer growth rates determined with serial volumetric CT measurements. Radiology 2006;241:554–563. [PubMed]

11. Wilson DO, Weissfeld JL, Balkan A, Schragin JG, Fuhrman CR, Fisher SN, Wilson J, Leader JK, Siegfried JM, Shapiro SD, et al. Association of radiographic emphysema and airflow obstruction with lung cancer. Am J Respir Crit Care Med 2008;178:738–744. [PMC free article] [PubMed]

12. Cronin KA, Gail MH, Zou Z, Bach PB, Virtamo J, Albanes D. Validation of a model of lung cancer risk prediction among smokers. J Natl Cancer Inst 2006;98:637–640. [PubMed]

13. Spitz MR, Hong WK, Amos CI, Wu X, Shabath MB, Dong Q, Shete S, Etzel CJ. A risk model for prediction of lung cancer. J Natl Cancer Inst 2007;99:715–726. [PubMed]

14. van Klaveren RJ, Oudkerk M, Prokop M, Scholten ET, Nackaerts K, Vernhout R, van Iersel CA, van den Bergh KA, van't Westeinde S, nan der Aalst C, et al. Management of lung nodules detected by volume CT scanning. N Engl J Med 2009;361:2221–2229. [PubMed]

15. Yankelevitz DF, Henschke CI. Does 2 year stability imply that pulmonary nodules are benign? Am J Roentgenol 1997;168:325. [PubMed]

16. Petro M, Quint LE, Nan B, Baker LH. Pulmonary nodule volumetric measurement variability as a function of CT slice thickness and nodule morphology. Am J Roentgenol 2007;188:306–312. [PubMed]

Articles from American Journal of Respiratory and Critical Care Medicine are provided here courtesy of
American Thoracic Society