Early Lung Cancer Action Project†
Initial findings on repeat screening
The Early Lung Cancer Action Project (ELCAP) was designed to evaluate the usefulness of annual computed tomography (CT) screening for lung carcinoma. With the baseline results having been reported previously, the focus of the current study was on the early results of the repeat screenings.
A cohort of 1000 high-risk individuals was recruited for baseline and annual repeat CT screening. At last follow-up, a total of 1184 annual repeat screenings had been performed. A positive result from the screening test was defined as newly detected, one to six noncalcified pulmonary nodules with interim growth. The diagnostic workup of the individuals was guided by recommendations supplied by the ELCAP investigators to the collaborating clinicians.
Of the 1184 repeat CT screenings, the test result was positive in 30 (2.5%). In 2 of these 30 cases, the individual died (of an unrelated cause) before diagnostic workup and the nodule(s) resolved in another 12 individuals. In the remaining 16 individuals, the absence of further growth was documented by repeat CT in 8 individuals and further growth was documented in the remaining 8 individuals. All eight individuals with further nodular growth underwent biopsy and malignancy was diagnosed in seven. Six of these seven malignancies were nonsmall cell carcinomas (five of which were Stage IA and one of which was Stage IIIA) and the one small cell carcinoma was found to be of limited stage. The median size dimension of these malignancies was 8 mm. In another two subjects, symptoms prompted the interim diagnosis of lung carcinoma. Neither of these malignancies was nodule-associated but rather were endobronchial; one was a Stage IIB nonsmall cell carcinoma and the other was a small cell carcinoma of limited stage.
False-positive screening test results are uncommon and usually manageable without biopsy; compared with no screening, such screenings permit diagnosis at substantially earlier and thus more curable stages. Annual repetition of CT screening is sufficient to minimize symptom-prompted interim diagnoses of nodule-associated malignancies. Cancer 2001;92:153–9. © 2001 American Cancer Society.
Inspired by the enhanced potential of computed tomography (CT) in screening for lung carcinoma, we initiated the Early Lung Cancer Action Project (ELCAP) in 1992.1, 2 The purpose of the ELCAP is to assess the usefulness of annual CT screening for lung carcinoma. In this, the principal objective is to assess the extent to which screening serves the diagnostic aim of shifting the distribution of diagnosed cases toward smaller sized tumors and thus toward earlier stages. An added major objective, the interventional one, is to quantify the curability of lung carcinoma as it depends on tumor size and disease stage at diagnosis. Both these objectives may be taken to refer to all lung carcinomas diagnosed under screening, irrespective of whether the diagnosis actually is prompted by screening or by interim symptoms. The diagnostic and interventional components jointly determine the overall rate of curability for cases detected under screening, to be contrasted with the corresponding rate of some 10% in the absence of screening. We believe it has been demonstrated beyond question that resection of screen-detected, early stage lung carcinoma commonly is curative,3–5 and therefore to us the real questions remaining to be answered are whether the diagnostic shift toward smaller and earlier stage lung malignancies and the resulting gain in curability are large enough to provide for cost-effective screening, given suitable specifications of both the screening regimen and its recipients.
We previously reported on our results of baseline screening, confirming the expectation that, relative to traditional chest radiography, CT-based screening enhances the detection of small noncalcified nodules markedly and thus of lung carcinoma at earlier and more curable stages relative to what is known to prevail in the absence of screening.2 On baseline screening, noncalcified nodules were detected 3 times as often (23% vs. 7%), malignancies were detected 4 times as often (2.7% vs. 0.7%), and Stage I malignancies were detected 6 times as often (2.3% vs. 0.4%). Of the CT-detected Stage I malignancies, 83% were not visible on chest radiography. Careful assessment of growth permitted the identification of the malignancies without any patient undergoing lobectomy for benign disease.
In the current study we report the results of annual repeat screening, with a focus on the diagnostic component, principally the distribution of diagnosed malignancies by size and stage. We also report the frequency of false-positive results and the workup associated with them. Investigation of the interventional component (i.e., the curability of malignancies diagnosed during screening) must await the accrual of suitable numbers of cases of lung carcinoma with the requisite follow-up.
Enrollment into the ELCAP was confined to a cohort of 1000 individuals (522 at Cornell University Medical College [CUMC] and 478 at New York University Medical Center [NYUMC]) who were age ≥ 60 years with a history of at least 10 pack-years of cigarette smoking and no history of malignancy (other than nonmelanotic skin cancer) and who were fit to undergo thoracic surgery. Baseline screening, initiated in 1993, was completed in 1998. Consent for both baseline and annual repeat CT screenings was obtained at the time of enrollment into the study cohort (Institutional Review Board: 1296-211 at CUMC and H4954-01 at NYUMC).
Among the 1000 individuals, 841 underwent the first repeat screening 6–18 months after the baseline screening. All individuals were asymptomatic at the time. These 841 individuals also underwent 461 subsequent screenings, 343 of which were performed within 6–18 months of the previous screening. To date there have been a total of 1184 “annual” repeat CT screenings (841 + 343 = 1184), and another 118 screenings with a longer interval.
Of the 841 individuals who underwent at least 1 annual repeat CT screening, 55% were male (and 45% were female); 92% were white, 5% were African-American, 2% were Hispanic, and 1% were classified as being of another racial/ethnic group. The median age at annual repeat screening for these individuals was 67 years, the median number of pack-years of smoking was 44, and the history of asbestos exposure was positive in 14% of the individuals, with no appreciable associations among these variables.
Among the 159 individuals who did not undergo a repeat CT screening 6–18 months after baseline screening, 30 were considered ineligible based on a diagnosis of lung carcinoma resulting from baseline screening, another 5 individuals were known to be alive but were ineligible because of a diagnosis of a malignancy other than lung carcinoma, and another 5 individuals were known to have died of causes other than lung carcinoma before the first repeat screening. Two were considered ineligible because of a symptom-prompted diagnosis of lung carcinoma. Among the remaining 117 individuals, when approached for repeat screening, 111 had a clear reason other than a diagnosis of lung carcinoma for forfeiting the repeat screening: 6 individuals had illnesses other than lung carcinoma, 4 said that their physician did not recommend repeat screening, 77 refused for reasons of their own, and another 24 individuals had moved.
As at the baseline, low-dose CT (LDCT) images were obtained at repeat CT screenings using a HighSpeed Advantage scanner (General Electric, Milwaukee WI) at 140 kilovolt peak (kVp), 40 milliangstrom (mÅ), and 2:1 pitch with a collimation (slice thickness) of 10 mm. The images, encompassing the entire lung region, were acquired in a single breathhold at end-inspiration after hyperventilation, and they were reconstructed with overlapping 5-mm intervals. In instances in which the baseline or a subsequent screen had led to the detection of nodules that at the time of the repeat screening had not yet been proven to be benign, the standard-dose “diagnostic” CT (DXCT) was used in lieu of the LDCT. DXCT images were obtained using 120 kVp, 200 mÅ, and 2:1 pitch with a collimation of 5 mm. These images were supplemented by high-resolution CT (HRCT) of the nodule being monitored using 1-mm collimation with overlapping 0.5-mm reconstructions.
The images were viewed at both lung and mediastinal windows (width 1600 HU, level-650 HU and width 325 HU, level 25 HU, respectively) by a chest radiologist. Comparison with prior images was performed with side-to-side viewing. At CUMC, after August 1998, the images were read on monitors used for all readings in the department (previously read on film), 12 images per film. At NYUMC, all images were read on film, 12 images per film.
Note was taken of each newly detected, noncalcified nodule. The protocol called for inclusive reading, so that discrete “ground-glass opacities” were counted as nodules. A nodule was classified as noncalcified if it did not show “benign calcification.”2, 6 The latter was defined as overall density at least as high as that of the adjacent bony structures in association with size < 20 mm and smooth edges.
For each newly detected, noncalcified nodule, the determination was made whether it was visible, in retrospect, on the previous CT screen. If it was visible and there had been no interim growth, no further diagnostic methods were recommended. For those individuals with any newly detected, noncalcified nodules that had grown (either because it was not visible on the previous CT screen or because, although visible, it was smaller), further workup with DXCT, including HRCT of the nodules, was recommended. Again, on HRCT the nodule was considered noncalcified if no benign calcifications according to the definition given earlier could be identified.
If the DXCT and HRCT confirmed the presence of 1–6 newly detected, noncalcified nodules with interim growth, the result of the screening test was classified as positive; otherwise, the results were considered to be negative.
In all cases of a positive screening result, defined characteristics of the relevant nodules were recorded: size (length and width), location (lobe and distance from the pleura), shape (round or nonround), edge (smooth or nonsmooth), and texture (pure ground glass appearance or other). For the current study, as for the baseline report, the following definitions were used. Size was defined as the average of the length and width, shape was defined as round if the nodule's width-to-length ratio was > 67% (otherwise it was considered nonround), location was defined as peripheral if any part of the nodule was within 2 cm of the costal margin (otherwise it was considered central), and texture was defined as pure ground glass if the nodule did not obscure the lung parenchyma and had no solid component (otherwise it was considered “other”).
Recommended Diagnostic Workup upon Positive Screening Result
The ELCAP protocol has recommendations for the diagnostic workup of positives. However, as with the baseline screening, the validity of the ELCAP for the repeat CT screenings did not require that the recommendations for diagnostic workup be followed, provided the final diagnosis became firmly established. Thus, the decision regarding how to proceed was left to the referring physician, and the actual diagnostic procedures and findings were recorded.
For all instances of a positive screening result, the recommended initial phase of the diagnostic workup was a course of broad-spectrum oral antibiotics, followed by HRCT 1 month after the initial CT screen to assess change(s) in the size(s) of the nodule(s). If the nodule(s) had resolved completely, no further workup was recommended. If the resolution was incomplete and all the nodules initially were ≤ 5 mm in size, additional follow-up HRCT 3 months later was recommended; provided no further growth could be documented, follow-up HRCT scans at 6, 12, and 24 months were recommended. For instances of further growth in these small nodules, biopsy was recommended. The same schedule of follow-up was recommended for pure ground glass opacities provided no solid component appeared; upon such an appearance, biopsy was recommended. For all other instances of incomplete resolution after antibiotic therapy, immediate biopsy was recommended, either by percutaneous CT-guided transthoracic fine-needle aspiration or by video-assisted thoracoscopy procedures. Moreover, all histologic findings from biopsy were documented.2
In each case of diagnosed malignancy, the initial HRCT together with the repeat HRCT (which followed the course of antibiotics, if applied) were used to determine the respective tumor volumes.7, 8 These volumes, together with the time between the HRCT screens, were used to calculate the doubling time of the tumor.7, 8
Documentation of Symptom-Prompted Diagnoses
The screening-induced “diagnostic shift” (to smaller sizes and lower stages at the time of diagnosis) ultimately refers, as was noted, to all diagnoses of lung carcinoma made during screening. It thus was our objective to identify and document all relevant “interim” cases of diagnosed lung carcinoma prompted by symptoms, and to pool these cases with the cases diagnosed based on the prompting of screening.
Cases relevant for this purpose are those in which individuals had previous negative screening results and for whom the diagnosis of lung carcinoma was the reason the next screening was not performed within 18 months. Among the 159 subjects who did not undergo an annual repeat screening, 2 individuals had documented symptom-prompted diagnoses of lung carcinoma. An interim diagnosis of lung carcinoma was excluded based on information provided on all the remaining subjects with the exception of six for whom no contact could be made either directly or indirectly through their relatives and referring physicians.
Frequency and Size Distribution of Newly Detected Nodules
Of the 1184 annual repeat screenings, noncalcified nodules were newly detected on LDCT in 63 cases. In 23 of these 63 cases, reexamination revealed that the nodules were visible (in retrospect) on the previous LDCT screen and that no interim growth had occurred; none of these nodules was > 5 mm in size. In another 10 instances, HRCT rendered the result of the screening test (LDCT plus HRCT) as negative, 5 of which showed no nodule(s) whereas the remaining 5 cases were found to have > 6 nodules. This left 30 cases with a positive test result, representing a rate of approximately 2.5% (30 of 1184 cases). As expected, the rate essentially was the same for the first repeat CT screening and for the subsequent repeat screenings.
Among the 30 cases of positive results of the annual screening test, a solitary nodule of the relevant type was noted in 28 cases, with 2 nodules noted in 1 case and 3 nodules noted in another case. In 9 of these 30 cases, the relevant nodule (the largest nodule if there was > 1) in retrospect was visible on the previous screening. The size of the nodule (the largest nodule if there was > 1) among the 30 cases was 2–5 mm for 16 cases, 6–10 mm for 7 cases, 11–20 mm for 4 cases, and 21–25 mm for 3 cases (Table 1).
Table 1. Diagnostic Workup of 30 Instances of a Positive Result of Annual Repeat Screening by Size of the Largest Nodule
|Positive result of screening||16||7||4||3||30|
| Death prior to workup||—||1||1||—||2|
| Repeat HRCT onlya||7||2||1||1||11|
| Repeat HRCT, then biopsy||3||2||2||1||8|
| Longer-term follow-up||6||2||—||—||8|
| Biopsy and antibioticsb||—||—||—||1||1|
Diagnostic Workup upon Positive Screening Result
In 2 of the 30 cases of a positive result of annual repeat screening, the subject died prior to the scheduled diagnostic workup, with cardiovascular disease the cause of death in both cases. In one of these cases, cardiac failure already was present at the time of the screening test 12 days before, and the other individual incurred a sudden cardiac death 6 weeks after the screening test. In one of these individuals, who had a 13-mm nodule in the right lower lobe, the appearance of the nodule was most consistent with inflammation or atelectasis. In the other subject, who had a pure ground-glass opacity measuring 8 mm in size in the right lower lobe, the nodule in retrospect was noted on the images obtained 1 year earlier and at that time the nodule measured 4 mm in size. No autopsy was performed in either individual; therefore no histologic diagnoses were available.
The 1-month HRCT in the remaining 28 cases showed that the nodule(s) had resolved fully in 12 cases, among them 7 cases in which the recommended course of antibiotics had been used. The original sizes of the nodules in these 12 cases were 2–5 mm in 7 cases, 6–10 mm in 2 cases, 11–20 mm in 1 case, and 21–25 mm in 2 cases. In 1 of these 12 cases involving a 24-mm nodule, the individual immediately underwent percutaneous fine-needle aspiration, which resulted in the diagnosis of pneumonia.
The remaining 16 cases required further diagnostic workup. Biopsy was performed in eight cases and the remaining eight cases were followed by HRCT. Of the 8 cases being followed, 2 involved solid nodules measuring ≤ 5 mm in size; both cases had been followed for < 2 years at the time of last follow-up and neither had demonstrated further growth. The other 6 cases were pure ground glass opacities, 4 of which were visible (in retrospect) and slightly smaller on the previous screening; however, at last follow-up, none of the cases had developed solid components, with 2 of the opacities already having been followed for > 2 years.
To date the diagnostic workup has led to the diagnosis of lung carcinoma in seven cases (Table 2). All seven diagnoses were the result of the eight biopsies that were performed immediately after repeat HRCT failed to detect complete resolution.
Table 2. Diagnosis of Lung Carcinoma Among 28 Instances of a Positive Result of Annual Repeat Screening by Size of the Largest Nodule
| Not yet ruled out||6||2||—||0||8|
| Ruled out||7||3||1||2||13|
The distribution of the seven screen-diagnosed malignancies by size, stage, and other characteristics is shown in Table 3. For the 6 nonsmall cell malignancies, the median size was 7 mm, and all cases but 1 were Stage IA. The exception was a 5-mm adenocarcinoma with mediastinal lymph node metastases (Levels 2 and 4), which thus represented Stage IIIA disease (Case 2 in Table 3). The cytologic diagnosis was one of adenocarcinoma for five of the six cases and pleomorphic carcinoma for the remaining case. The single small cell carcinoma was of limited stage disease. Of the three diagnosed malignancies > 10 mm in size, all were clearly visible (in retrospect) on the previous screening (Table 3), and of the 4 malignancies measuring ≤ 10 mm, only 1 case (or perhaps none) was barely visible on the previous screening. Table 3 also shows the calculated doubling time for each of the screen-detected malignancies. The calculated doubling time was shortest (30 days) for small cell malignancies. For the 6 nonsmall cell carcinoma cases the doubling time ranged from 60 to 170 days with a median of 100 days. All seven screen-diagnosed malignancies were round and solid, four had a nonsmooth edge, four were in the right lung, all but one were in the upper or middle lobes, and five were central.
Table 3. Characteristics of Malignancies Diagnosed During Annual Repeat Screening
In two additional instances the diagnosis of lung carcinoma was prompted by symptoms prior to the scheduled repeat CT screening, with both of these tumors classified as endobronchial (Table 3). In 1 of these tumors, reexamination of the baseline LDCT, taken 6 months earlier, showed no abnormality and the specific diagnosis was small cell carcinoma obstructing the posterior segmental bronchus of the right upper lobe and associated with postobstructive pneumonia. The carcinoma was determined to be of limited stage disease because bone marrow aspiration and biopsy showed no evidence of metastases. In the other malignancy, reexamination of the baseline LDCT, performed 5 months prior to the diagnosis, showed (in retrospect) narrowing and irregularity of the left main bronchus at the site. The specific diagnosis was one of endobronchial squamous cell carcinoma with involvement of a single hilar lymph node (Stage IIB) and associated collapse of the left upper lobe.
All seven nonsmall cell malignancies (six of which were screen-detected and one of which was symptom-detected) were deemed resectable. Two of these malignancies, both of which were screen-detected, were not resected because of medical contraindications, and the other four patients with screen-detected tumors underwent lobectomy. The symptom-detected malignancy required pneumonectomy because of its location (Case 9). All resections were coupled with mediastinal lymph node dissection, which did not reveal contralateral mediastinal lymph node metastases in any of the cases. The surgical specimen confirmed the cytologic diagnosis in all five resected cases.
As expected, our experience with annual repeat CT screening was quite different from that with baseline screening, notably with regard to false-positive results from the screening test. Positive results were, for one, much less common (2.5% vs. 23%). In addition, clinical management with a view to minimizing biopsies on benign nodules was quite successful on the grounds of the ELCAP recommendations for diagnostic workup; in 33% of the positive test cases, the nodules had resolved on follow-up HRCT (half after a course of antibiotics) and in the remainder, the documented absence of further growth in the smallest nodules served to identify the remaining false-positive results.
The nodule-associated malignancies were, as was anticipated, typically of Stage IA and small even at that, but the numbers remained quite sparse and thereby imprecise (Table 3). On the other hand, there is every reason to believe that on repeat screening, the diagnostic distribution is shifted toward earlier stages and smaller sizes, even more so than at baseline screening, in which we found 22 of the 27 nodule-associated malignancies (81%) to be of Stage IA, with a typical size of < 10 mm. In the experience of Sone et al.,9 35 of 40 malignancies diagnosed on annual repeat CT screening (88%) were of Stage IA.
The translation of this diagnostic distribution to its corresponding overall rate of curability under screening requires information concerning the stage-specific and size-specific rates of curability. The 5-year survival rate of Stage IA nonsmall cell malignancies measuring < 20 mm and detected by CT has been reported to be > 90%,10, 11 suggesting a curability rate of these malignancies of > 80%. To our knowledge, curability of the screen-detected small but late stage nonsmall cell and limited stage small-cell malignancies has yet to be quantified.
Because these results and inferences pertain mainly to very small lesions, the question of overestimation on the grounds of potential “overdiagnosis” is prone to arise. For orientation, it is good to appreciate that evidence clearly shows the absence of overdiagnosis in the context of Stage I cases detected during screening by traditional radiography.3–5 However, because the CT-detected lesions are distinctly smaller, the concern remains legitimate, and indeed was a concern in the current study. In an effort to avoid the problem, we naturally have been very careful with the pathologic (cytologic and histologic) criteria for a rule-in diagnosis of malignancy but beyond this we had interim growth in all cases, and this was supplemented by the documentation of further growth before biopsy. As it turned out, all cytologic diagnoses of malignancy (rule-in) were confirmed by the histologic specimens from surgery and furthermore, all the calculated rates of growth were in accord with those of definite carcinomas of the lung.7, 8 Ultimately, once there are a sufficient number of cases that, for various reasons, were not resected within the ELCAP and its “sister” projects, it will be possible to estimate empirically the proportions that were overdiagnosed (specific to size), if any.
We continue to be comfortable with a 1-year interval between screenings. As was noted earlier, the screen-detected, nodule-associated malignancies showed a substantial diagnostic shift toward earlier stages and smaller tumor sizes, and there has not yet been a single instance of a symptom-prompted interim diagnosis of nodule-associated malignancy. That there were two symptom-prompted interim diagnoses of lung carcinoma, neither of which was nodule-associated but were endobronchial instead, serves as a reminder that CT screening for lung carcinoma might well deserve to be supplemented by cytologic screening with a view to the earlier detection of centrally located endobronchial malignancies.12
Even if a given regimen of CT screening for lung carcinoma (perhaps a variant of the one addressed in the ELCAP) does serve to increase the overall rate of curability for lung carcinoma among the screenees, this does not in and of itself justify the use of that regimen of screening. It needs to be applied on indications such that the prospect of early diagnosis and its associated curability translate into a gain in life expectancy sufficient to justify the cost of the “screening,” (i.e., of the screening test together with the result-contingent definitive diagnostics). The issues of cost-effectiveness are somewhat complex, but it is evident that, with suitable specifications of both the screening and its recipients, the cost per life-year saved can be as low as $10,000.13 To our knowledge such a cost per life-year saved is well below that for existing screening programs for breast carcinoma14 or cervical carcinoma15 and is well below the benchmark of $50,000 used in the U.S.
The particulars of potential screening for lung carcinoma currently constitute an actively evolving topic with respect to all its principal elements (the screening test[s], the diagnostic workup vpm of positive screening result, intervention in cases of early malignancy, and the identification of suitable candidates for screening).9 The accruing evidence from the ELCAP and others studies,9 although still insufficient, is continuing to heighten the prospects for cost-effective screening for a malignancy that now is the main cause of cancer deaths in both genders.
The authors gratefully acknowledge the outstanding efforts of their study coordinators, Kimberly Agnello at the Weill Medical College of Cornell University and Linda Michaels at the New York University Medical Center.