Early Lung Cancer Action Project: overall design and findings from baseline screening

Authors Henschke C, McCauley D, Yankelevitz D, Naidich D, McGuinness G, Miettinen O, et al.
Source Lancet 354:99-105. July 10, 1999.
Institutions Cornell and Columbia and NYU medical centers, New York; McGill University, Montreal, Canada.
Support National Institutes of Health, USA; Eastman-Kodak Corporation; General Electric Corporation.

Background

Lung cancer is most frequently detected at a stage when it is too advanced to be curable.  Although 5-year survival for stage I cancer that has been resected can be as high as 70%, the overall 5-year survival for lung cancer is only about 12%.  This implies that screening to detect early disease should be able to improve prognosis.  The main randomized trial which looked at this question, the Mayo Lung Project, did not show a significant benefit for screening by chest x-ray, however.

Because CT scanning should be able to detect lung cancer at an earlier stage than conventional radiography, the current trial was designed to look at the effect of low-dose, non-contrast screening CT scans on the ability to detect small, stage I lung cancers.

The Early Lung Cancer Action Project (ELCAP) was not a randomized study, but was designed as a screening and follow-up trial. 1000 participants at high risk for lung cancer were offered baseline screening, starting in 1992, with annual re-screening.  The endpoints were the detection of non-calcified nodules and the clinical outcomes for those patients with nodules, with a particular emphasis on outcome by size and radiographic appearance of the nodules detected.  This paper reports on the results of baseline screening.

Methods

Patients

Patients were recruited for screening if they were over 60 years of age, had smoked at least 10 pack-years, had no history of cancer, had no symptoms of lung cancer and would be able to undergo thoracotomy.  A total of 1000 volunteers were enrolled (at New York Hospital and NYU Medical Center).  Baseline information recorded included age, sex, race, smoking history and asbestos exposure.

Screening was performed at baseline (the subject of this report) and is to be repeated annually.

Baseline x-ray and CT scan

Standard PA and lateral chest x-rays were obtained.  Helical low-dose CT scanning of the lungs was performed in a single breath-hold (up to 20 secs), at end inspiration after hyperventilation.  The slice thickness was 10 mm; images were reconstructed in overlapping 5 mm increments.

All x-rays were read by two radiologists who then discussed their findings and came to a consensus reading; if no consensus could be reached, a third radiologist was asked to adjudicate.

X-rays and CT scans were analyzed for the presence and absence of nodules.  Each nodule was characterized by its shape (round or not), edge (smooth or not), size (length and width), location (lobe involved and central vs. peripheral) and by the presence or absence of benign calcifications.

Studies were classified as positive if there were between one and six non-calcified nodules (NCN).  Studies were negative if there were no NCN, and were judged to show diffuse disease if there were more than six NCN.

Recommended follow-up

Patients with positive screening CT scans and/or x-rays were recommended to have standard-dose diagnostic CT scans for better delineation of the lesions.  If the non-calcified nodules seen on screening were confirmed and were non-calcified on the diagnostic CT, further work-up was recommended, depending on the size (average of length and width) of the nodule:

  • nodules 5 mm or less: follow up with high resolution CT scans at 3 mos, 6 mos, 12 mos and 24 mos as long as there was no growth.
  • nodules 6-10 mm: if possible, CT guided fine needle aspiration biopsy or thoracoscopic biopsy; if not biopsied, CT follow up as for the smaller nodules.
  • nodules greater than 10 mm: biopsy by fine needle aspiration, bronchoscopy or thoracoscopy.

The actual decisions about diagnostic and therapeutic procedures were left up to the referring physician and patients. Whether or not the recommendations were followed, the results of all diagnostic procedures and any surgery for resectable tumors were recorded.

Analysis

This paper presents the results of baseline screening (prevalence data) and the stage of malignant nodules by nodule size. It is too early in the protocol to present survival data and data on follow-up screening (incidence data).

Results

Participants

Median age was 67;  54% were male; 91% were white. Median number of pack-years smoked was 45. Asbestos exposure was present in14%.

Radiologists agreement

The readers disagreed on the interpretation of the low-dose screening scans in only 31 out of the 1000 scans.  In 28 of these cases, both readers identified abnormalities, but only one reader interpreted them as non-calcified nodules. After adjucation, all were deemed NCN's. None of these nodules turned out to be malignant.

Screening results

On screening CT, a total of 559 nodules were detected, of which 35% had benign calcifications. On standard x-ray, a total of 196 nodules were found, of which 60% had benign calcifications.

CT scanning identified a total of 233 individuals with 1 to 6 non-calcified nodules (23% of the study population). Only 33 of these individuals (14%) also had positive findings on conventional x-rays.  There were 68 individuals with positive results on conventional x-rays; 33 had positive CT scans as mentioned previously, the other 35 had "nodules" that were not confirmed on CT scan (shadow confluence).

Among the 233 individuals with a positive CT screening test, 159 (68%) had one nodule and 43 (18%) had two nodules. The remaining 14% had 3, 4, 5 or 6 nodules. The size of the largest nodule detected in these patients was 2-5 mm in 58%, 6-10 mm in 30%, 11-20 mm in 9% and greater than 20 mm in 2%.  Among the 33 patients whose nodules were also seen on standard x-rays, the sizes were larger overall (2-5 mm - 33%; 6-10 mm - 42%; 11-20 mm - 16% and greater than 20 mm - 9%).

Diagnostic work-up

According to protocol, the 233 participants with positive screening scans should have had high resolution CT scans.  50 participants did not, 14 for "good" reasons (a previous CT scan was available or the nodules were tiny and appeared benign), and 36 because the physician or insurance plan declined or they were lost to follow up.

The remaining 183 participants underwent the high resolution CT scan. In 54 of these, the non-calcified nodules were not confirmed and in 9 the nodules were so benign appearing that only CT follow-up in one year was recommended.

Among the remaining 120 patients requiring follow-up:

  • biopsy was recommended immediately in 16, was performed in 14, and revealed malignant disease in 13 cases.
  • biopsy was recommended after growth was documented by subsequent scans in 14, and revealed malignant disease in all 14.
  • there has been no growth in 90 participants, 36 of these have been followed up for 2 years.

Malignancies detected

A total of 33 patients underwent biopsy, 30 recommended by protocol, as noted above, and three for reasons not indicated by the protocol (all 3 turned out to be benign).

Of the 30 per-protocol biopsies, 27 were malignant. Thus, 27 out of 233 participants (12%) with positive screening tests turned out to have malignant disease.  Among the 68 patients with positive chest x-rays, 7 cases of malignancy were detected (10%).  All of these 7 cases were picked up by CT screening.  Thus CT screening identified 20 (out of 27) cases that would not have been picked up on routine chest x-ray; the yield of CT screening was 4 times greater than that of conventional chest x-ray.

Screening CT also picked up a further 4 malignancies that were not picked up by chest x-ray but were not NCN's (two endobronchial lesions and two mediastinal lesions) and were thus not counted as positive screening results.

Treatment and stage

Diagnosis of malignancy was made by fine needle aspiration in 18 and thoracoscopy in 9; all were judged to be operable. Of the 27 participants with detected malignancies, 24 underwent lobectomy, one underwent mediastinoscopy which revealed stage IIIB disease, one underwent radiosurgery and one declined treatment. 18 of the 27 had adenocarcinoma, the remainder were a mix of adenosquamous, bronchoalveolar, atypical carcinoid and squamous.

Of the 27 malignancies detected, 23 (85%) were stage I.  Of the 7 malignancies detected by conventional x-ray, 4 were stage I.  Thus CT screening detected 6 times as many stage I tumors as conventional x-ray: 23  vs. 4.

Author's discussion

The authors make a number of points about their study:

  • Low-dose CT screening is much more sensitive than conventional x-ray for detecting non-calcified nodules and small lung cancers.  CT detected 3 times as many nodules, 4 times as many malignant tumors and 6 times as many stage I tumors as conventional x-rays.

    26 out of 27 patients with malignant disease were operable.  In the Mayo Lung Project study, which utilized conventional x-rays, only about 50% of tumors detected at baseline were operable.
     
  •  A main objective of this study is to correlate cure rate with nodule size, but follow-up is too short to give any data on this yet.  However, given the small size of tumors that were detected and prior historical data on cure rates according to tumor size, the authors estimate that 5-year survival should be 60%-80% for patients with tumors detected by low-dose CT.  This is much better than the usual 12% 5-year survival rate for lung cancer.
     
  • Contrary to current beliefs, the data presented here demonstrate that the presence of 2-6 non-calcified nodules detected on CT was almost as likely to represent malignant disease (11% were malignant) as the presence of a solitary nodule (12%).
     
  • The finding of many cases of NCN's did not lead to a lot of unnecessary biopsies.  Unnecessary biopsies were avoided by the management recommendations which required an increase in size before small nodules were biopsied.  If one disregards the three biopsies done which were not recommended by the guidelines, only one out of the 30 recommended biopsies was for a benign lesion.
     
  • An argument that can be raised about a study like this one is that it leads to overdiagnosis, i.e. that it leads to the excision of lesions that would not have caused problems in the long run (very slow-growing malignancies).  The authors argue that, using precise measurements of nodule size, the growth rate (doubling time) of the lesions that were excised does not support this hypothesis.
     
  • As data accumulates from this study, it will enable more targetted diagnostic approaches to small nodules (based on the characteristics of those nodules that turn out to be malignant) and the evaluation of novel therapeutic approaches to these nodules.
     
  • Detailed cost-effectiveness analyses will be conducted, but the authors note that screening CT scans should be only slightly more expensive than conventional x-rays and treatment of early lung cancer is cheaper than treatment of more advanced disease.

 

Comment

This study, following 1000 participants at high risk for lung cancer with baseline and annual low-dose CT screening for non-calcified nodules, clearly demonstrates the ability of CT scanning to detect small tumors with much greater sensitivity than conventional radiography.

Will this ability translate into lives saved by early detection? ELCAP was not a randomized trial and thus cannot answer this question directly, but the authors argue that, given historical data linking cure rates to tumor size and given the ability of CT scanning to identify much smaller tumors, it makes perfect sense that screening should improve the cure rates for those cancers detected.

It is important to note that this paper provides no actual data on survival or cure rates. It only provides data on the number of tumors found and their size.  Any speculation on the effect of finding these smaller tumors on cure rates remains -- speculation.  Since further data from this study will be forthcoming, in the future we will have hard data on the cure rates of those tumors detected by CT scanning.

As the authors themselves point out, one problem with this type of non-comparative study is the issue of whether those cancers that were actually detected would have caused problems for the patients in the long run, before other illnesses intervene.  Thus, the early discovery of a small lung cancer is of little significance if the cancer thus detected and removed would have taken years to become clinically significant and the patient dies from a different illness before that time.  The authors argue against this scenario based on the doubling times of the tumors that they detected.  Nevertheless, this is an important point.  Since the patients studied here were 67 years old on average and were smokers, they would be expected to have a relatively high incidence of cardiovascular diseases, so the detection of very small lung cancers would have less impact on survival than might be apparent from the cure-rates alone.

Another issue is lead-time bias.  Lead-time bias relates to the fact that tumors that are detected early will have a proportionately longer survival time, just based on the fact that they were detected early.  Thus, for example, very small tumors that are detected 2 years earlier than those detected after symptoms develop would have a survival time even in the absence of any treatment that is 2 years longer.

The improved detection of small non-calcified nodules comes at the price of a high rate of detection of clinically insignificant nodules, which require further work-up.  Out of 1000 participants, 233 had positive screening tests but only 27 had cancer.  Thus, approximately 200 had false positive test results (assuming that most of those who did not receive appropriate follow-up had benign disease).  Apart from the anxiety that positive screening tests produce, most of these false positive tests will lead to at least one and, for the majority, up to 5 high-resolution CT scans to confirm benign disease.  A false positivity rate of approximately 20% is quite high -- one out of 5 patients who are screened.  The authors do not even comment on this issue.

This high rate of false positive screening examinations will adversely impact on cost and cost-effectiveness. The authors state that low-dose CT screening, which does not use contrast, is relatively cheap.  Indeed, I have been getting letters from local radiologists touting their services and offering screening CT's at $300 each.  However, the performance of five high-resolution scans to rule out malignant disease will cost substantially more, probably in the range of several thousand dollars.

Finally, the population studied here is at particularly high risk for lung cancer (the asbestos exposure rate of 14% is particularly impressive). Even if it turns out that low-dose CT scanning is cost effective in this population, its cost-effectiveness and its positive predictive value will be markedly reduced in any population that is at lower risk for lung cancer.  Given the fear that lung cancer engenders in the population, it is likely that this procedure will be demanded by many patients at much lower risk of the disease.  In this lower-risk population, a false positivity rate of 20% will cause a significant burden of anxiety and cost, without yielding many cases of early lung cancer.

August 29, 1999


References

References related to this article from the NLM's PubMed database. 
 


Reader Comments

October 11, 1999

Letters to the editor about this article, from the October 2, 1999 Lancet.  Topics discussed include the sensitivity and specificity of the test, cost-effectiveness and the risk of overdiagnosis of clinically insignificant lesions.


Date: November 12, 2000
From: rfaeder@aol.com

There were 27 cases of malignant nodules reported. Of the 26 operated on, how many are alive today which is past 5 years?

The authors state that they didn't yet have follow-up data, in terms of survival.  I don't believe this data has been published yet.  It will be important, but, since this was not a randomized trial, the survival data can only be compared to historical controls.  -- mj

 

 

Submit a comment

Journal Club homepage

Site Meter