If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Screening for lung cancer works! Given the high incidence and lethality of lung cancer, an effective screening test for lung cancer has long been desired. The recently published National Lung Screening Study (NLST), a prospective, randomized controlled trial (RCT), has shown that screening with low-dose spiral computed tomography (LDSCT) results in mortality reduction, in other words, fewer deaths from lung cancer.
There is much to be learned about who may benefit from screening computed tomography (CT) and who may not benefit. Despite finding more cancers and more early stage cancers, previous randomized trials of screening with chest radiography (CXR) and sputum cytology did not result in fewer deaths from lung cancer in comparison with no screening or screening at a lesser frequency. Early studies showed CT had promise as a much more sensitive test than CXR in detecting lung cancer. Debate ensued as to whether the apparent survival benefit demonstrated with CT in nonrandomized studies proved effectiveness, because the appropriate measure of screening benefit would be a decrease in cancer mortality. The debate was further intensified by the recognition that CT screening frequently results in the detection of pulmonary nodules that often require follow-up or further evaluation. There will likely be age and quantity-of-smoking criteria for which screening will be recommended and reimbursed. Until those recommendations are available, it is best to discuss the potential benefits and risks of CT screening with each patient at risk who is interested in being screened.
The problem
The unfortunately persistent smoking prevalence rate suggests that the problem of lung cancer will have a secure presence for several decades. The current clinical management of patients with lung cancer too often begins with detection at the time of symptomatic presentation. Symptoms are usually an indication that the lung cancer is in an advanced stage. Early detection is preferred, but until now no test has proven efficacy as a screening test for lung cancer. Prevention is key; however, improving outcomes from lung cancer will result from early detection and more effective treatment. The present 5-year survival of 16% for lung cancer pales in comparison with 5-year survivals of 88% for breast cancer, 65% for colon cancer, and 100% for prostate cancer, respectively.
Part of the reason lung cancer has lagged so far behind in survival is that there has been no established screening test. An effective screening tool has thus been greatly anticipated.
Screening
The results of the NLST are the first confirmation that any screening test can reduce deaths from lung cancer. The most recent edition of the American Cancer Society (ACS) Guidelines for the early detection of cancer state: “At present neither the ACS, nor any other medical/scientific organization, recommends testing for early lung cancer detection in asymptomatic individuals.”
The NLST will likely have game-changing impact on this recommendation. But why hasn’t screening been recommended when it was already known that screening with CXR or CT led to the detection of more cancers, more early-stage cancers, and marked improvement in survival? The answer lies within issues introduced by screening—each of those measures are necessary outcomes but together remain insufficient to show efficacy from screening.
Use of a screening test introduces biases that are inherent in screening. The most significant of these are lead time, length time, and overdiagnosis bias. Lead-time bias occurs when a cancer is detected earlier than it would have been in the absence of screening, yet even with appropriate intervention, the natural history of disease is not changed. As the measure of time between detection and death, apparent survival is lengthened, suggesting benefit. However, true survival is something that cannot be measured—that is, the time from disease onset to eventual death, whether that is due to lung cancer or not. In the situation of lead-time bias, apparent survival is improved by applying a screening test, but mortality remains unchanged. A patient with cancer would appear to live longer simply because the disease was detected earlier, but this is not necessarily an indication of a change in the natural history. An example would be an aggressive cancer that originates from one cell at time zero, results in the onset of symptoms and is detected at year 4, and leads to the patient’s death at 5 years. Survival from diagnosis in this case would be 1 year. Applying a screening tool such as CXR might lead to detection and intervention at year 2. Even if intervention did not change the course of the cancer and death occurred at year 5, survival is now 3 years from diagnosis. Survival time tripled with screening, yet death resulted at the same time it would have without screening. Although lengthening apparent survival is desired, it is not sufficient proof of benefit from screening.
Length-time bias describes an apparent improvement in survival, when that improvement is actually due to selective detection of cancers with a less progressive course while missing cancers that have the most rapidly progressive course. Application of a screening tool at specified intervals would have a higher likelihood of detecting a cancer with a more indolent course than one with a more rapid course that presents with symptoms between screenings. The result is the demonstration of an apparent improvement in survival, though the better outcome reflects the more indolent cancers found with screening. CT screening is most likely to detect peripheral nodular cancers, and these are more likely to be adenocarcinomas with a more favorable outcome than, for example, small cell lung cancers that tend to be central and aggressive, and unlikely to be found by periodic CT screening.
An overdiagnosis cancer is a true cancer but one the patient would have died with rather than died from. Overdiagnosis bias is an extreme form of length-time bias in which indolent lung cancers are detected that would not have altered the expected survival when compared with the normal population. Patients with overdiagnosed lung cancers would have died of a cause other than lung cancer; were it not for screening, these lung cancers may have gone undetected. The natural history of these cancers does not need to be altered with detection and treatment, as they were not destined to do harm.
In summary, because of these biases inherent in screening, survival would be expected to be more favorable even if earlier detection and intervention did not alter the course of disease. For this reason, the apparent improvement in survival demonstrated in observational CT-screening studies was not proof that it saved lives. Mortality reduction, rather than survival improvement, is the ultimate measure of a screening tool’s effectiveness, and needs to be demonstrated by performance of RCTs. The RCT has been accepted as the best scientific method for determining the effectiveness of screening and for other medical-practice interventions.
The National Cancer Institute supported 3 prospective randomized trials evaluating screening for lung cancer in the 1970s at 3 sites: Johns Hopkins, Memorial Sloan-Kettering, and Mayo Clinic.
These studies involved more than 30,000 men, current or former smokers, who were older than 45 years. Screening at Memorial Sloan-Kettering and Johns Hopkins showed no change in mortality rate of lung cancer by adding sputum cytology every 4 months to annual CXR screening, when compared with the use of annual CXR alone. The 5-year survival in these two studies was nearly 35%, considerably above the historical average at the time of 13%, but was simply the result of screen bias.
After prevalence screening to exclude cancer was performed, the Mayo Lung Project (MLP) compared sputum cytology and CXR every 4 months with standard care, which at the time was the recommendation for these tests annually. In the more frequently screened group 206 cancers were detected, of which 46% were resectable, whereas 160 cancers were detected in the control group with only 32% resectable.
However, the anticipated shift from detection of cancers at advanced stage to cancers at earlier stage was not seen. Unresectable lung cancer was identified in 112 participants in the group screened every 4 months compared with 109 in the control group. In the MLP the screened group had a 5-year survival of 35% compared with 15% in the control group, yet the mortality of 3.2 per 1000 person-years in the screened group was not statistically different from the mortality of 3.0 per 1000 person-years in the control group.
This lack of mortality reduction in all 3 of the NCI-sponsored studies essentially failed to prove benefit from screening for lung cancer, and drew the line on screening until now. Moreover, a slight increase in mortality resulting from more frequent screening compared with less frequent screening was subsequently shown by meta-analysis, suggesting that screening actually resulted in more harm than good.
The results of the Prostate, Lung, Colon, and Ovary (PLCO) screening trial that investigated CXR versus no screening for lung cancer are still awaited, but the NLST essentially renders these results irrelevant.
The most sensitive imaging modality for detecting pulmonary nodules is CT scanning. CT was put into clinical use back in the mid 1970s, so why has it taken so long to be used for lung cancer screening? Conventional chest CT required radiation dosages and image-acquisition times that were impractical for screening purposes. The development of LDSCT greatly reduced the radiation dose and the scan time, making LDSCT a feasible screening tool. Conventional CT images are obtained at 140 to 300 mA and are performed over many minutes using multiple breath holds. By contrast, LDSCT images may be completed in 15 seconds during a single breath hold, using at only 20 to 50 mA. LDSCT is comparable in sensitivity and specificity of lung nodule detection with conventional CT mode.
Studies from Japan created excitement in suggesting the viability of LDSCT as a tool for early lung cancer detection. The first report was from Kaneko and colleagues,
who screened 1369 high-risk participants with both LDSCT and chest radiography. CT detected 15 cases of peripheral lung cancer while 11 of these were missed on chest radiography. Of the non–small cell carcinomas identified, an amazing 93% were stage I. Sone and colleagues
authored the second report in the literature, with 3958 participants screened with both LDSCT and CXR. Only 4 lung cancers were detected by CXR whereas 19 were seen on CT; 84% were stage I at resection. In the United States, CT-screening efforts were led by Henschke and colleagues
with the Early Lung Cancer Action Project (ELCAP). This study enrolled 1000 high-risk participants (based on smoking history) and screened with both LDSCT and CXR; initial results were reported in 1999. A total of 27 prevalence lung cancers (ie, present on the baseline scan) were detected by CT; only 7 of those were seen by CXR. At surgery, 23 of the 27 (85%) malignancies were stage IA. Tables 1 and 2 summarize the baseline and incident findings of these and other observational studies using LDSCT and CXR.
Table 1Observational CT-screening studies: prevalence cancers and nodules detected
In each of these early studies, CT detected about 4 times more lung cancers than did CXR. Several additional single-arm observational studies reported similar results with LDSCT in the United States, Germany, Italy, and Japan.
The reported survival results from these studies were promising; however, by design they were insufficient to show that CT screening saved lives. To highlight this point, Bach and colleagues
applied a validated lung cancer prediction model to 3 prospective single-arm observational studies of CT screening combining 3246 participants. CT screening found 3 times the number of expected cancers, resulted in 10 times the expected number of resections (109 of 144 cancers were resectable), and identified more than the expected number of advanced stage cancers. The 4-year actual survival was 94% for participants with surgical stage I cancers, yet CT screening did not result in a predicted reduction in the expected number of deaths from lung cancer. Of note, the 95% confidence intervals within the Bach article
would not have been able to detect a reduction in lung cancer mortality as large as 30%. These results emphasized the need to evaluate the effectiveness of screening by mortality reduction rather than survival improvement.
CT screening: randomized controlled trials
Several RCTs of CT screening are under way or have been completed (Table 3). The NLST is by far the largest of these; it included 53,454 participants who were current or former heavy smokers (the equivalent of a pack per day for 30 years) aged 55 to 74 years. Participants who had quit smoking had to have done so within 15 years of enrollment, and participants could not have had a previous CT scan within 18 months of enrollment. Participants were randomized between LDSCT and CXR at baseline, year 1, and year 2, and then followed for a median of 6.5 years.
Writing Committee, Lung Screening Study Research Group
et al.
Baseline findings of a randomized feasibility trial of lung cancer screening with spiral CT scan vs chest radiograph: the Lung Screening Study of the National Cancer Institute.
Baseline results of the Depiscan study: a French randomized pilot trial of lung cancer screening comparing low dose CT scan (LDCT) and chest X-ray (CXR).
A positive CT-screening result in the NLST was defined as the finding of a noncalcified nodule of at least 4 mm. Twenty-seven percent of participants randomized to LDSCT screening had a positive baseline scan, of which 96% were false positive (ie, not lung cancer). In the CXR screening arm, a positive baseline study was found in 9%, of which 95% were false positives. In the LDSCT arm of the NLST there were 649 cancers that were detected by CT screening, 44 interval cancers (identified during the screening period but not detected by screening), and 367 diagnosed during follow-up post screening.
In the CXR arm there were 279 cancers detected by CXR screening, 137 interval cancers, and 525 diagnosed during follow-up post screening. Among cancers detected by CXR, 47.6% were stage I and 43.2% were stage III or IV. By contrast, within the LDSCT arm, 63% of lung cancers diagnosed from a positive screening test were Stage I, and 29.8% were Stages III or IV. This finding demonstrated that CT resulted in a shift in stage at diagnosis from advanced disease to early stage. By late October 2010, 354 deaths from lung cancer had occurred among those in the LDSCT arm versus 442 deaths among those in the CXR arm, corresponding to a 20.3% mortality reduction with CT screening.
Overall mortality in the CT-screening arm was reduced by 6.7% (P = .02) over the CXR arm, but this was not significant after excluding deaths from lung cancer. The number needed to treat (ie, the number of high-risk participants needed to be screened to save one life from lung cancer) was 320.
The Dutch-Belgian randomized lung cancer screening trial known as the NELSON trial is ongoing. Information was reported for the 7557 participants randomized to receive CT screening.
More than 50% of the participants had one or more nodules identified at baseline. Seventy cancers were diagnosed at the baseline examination, resulting in a prevalence rate of 0.9%; 64% of these cancers were stage I. The second round of CT screening identified an additional 57 cancers in 54 participants; 74% were surgical stage I cancers. No information from the NELSON study has been provided thus far for the control group or regarding mortality.
The ITALUNG study is under way in Italy, wherein 3206 participants have been randomized to LDSCT versus no screening.
The baseline CT was positive (defined as a pulmonary nodule ≥5 mm) in 426 (30.3%) of 1406 subjects. Twenty-one prevalence lung cancers were diagnosed in 20 participants (prevalence 1.5%); 10 (47.6%) were stage I. Control group findings have not yet been reported. The Danish Lung Cancer Screening Trial (DLCST) randomized 4104 participants to CT versus no screening.
Among those who received CT screening, there were 17 cancers found at baseline (prevalence rate of 0.8%); 9 (53%) were stage I, and 8 (47%) were stage III or IV. No information has been provided on the control group.
Two small RCTs of CT screening have also reported data on both the screened and the control group. The Lung Screening Study was the pilot study to determine feasibility of the NLST. A total of 1660 subjects were randomized to the LDSCT arm and 1658 to the CXR arm.
Writing Committee, Lung Screening Study Research Group
et al.
Baseline findings of a randomized feasibility trial of lung cancer screening with spiral CT scan vs chest radiograph: the Lung Screening Study of the National Cancer Institute.
In the CT arm the rate of cancer detection on the baseline scan was 1.9%, and for year 1 was 0.57%. In the CXR arm the rate of cancer detection at baseline was 0.45%, and for year 1 was 0.68%. No mortality information was presented for the study; however, the finding of a nearly twofold higher number of advanced stage cancers in the CT arm suggested there was not likely to be any mortality benefit from CT screening. The Depiscan study performed in France randomized 621 participants between CT and CXR.
Baseline results of the Depiscan study: a French randomized pilot trial of lung cancer screening comparing low dose CT scan (LDCT) and chest X-ray (CXR).
One or more nodules were seen in 152 (45%) of 336 subjects receiving CT with 8 lung cancers identified, whereas only one cancer was detected in the CXR arm.
The DANTE trial is under way in Europe, and enrolled 2472 participants who were randomized between CT versus no CT (following a baseline CXR) with the plan to screen annually for 4 years.
After a mean of 33 months follow-up, lung cancer was detected in 60 participants in the CT arm versus 34 in the control group. Stage I cancers represented 60% of the cancers in the CT-screening group compared with only 15% in the controls. No mortality data have been published from this study. Results for these smaller RCTs may be additive to the NLST and may help define the population(s) for whom CT screening is appropriate, as well as the frequency with which screening should be performed.
Potential problems with CT screening
Now that LDSCT has been shown to reduce mortality, the problems of CT screening are recognized as manageable. A few major concerns raised include the issues of false-positive scans, benign nodule resections, overdiagnosis, and the effect of radiation, as well as cost.
In the Mayo CT study, baseline scanning revealed one or more noncalcified nodules in 51% of the participants; after 4 annual scans 74% of the participants had one or more nodules.
The ELCAP study used a single detector scanner with 10-mm collimation (slice thickness) and increments of 5 mm of overlap. The Mayo study used a multidetector scanner with 5-mm slice width and a 3.75-mm reconstruction interval. In the study by Diederich and colleagues,
noncalcified pulmonary nodules were detected in 43% of the participants; this study also used 5-mm collimation. A Canadian study showed that the number of participants with one or more nodules increased from 36% to 60% when scan slice thickness was reduced from 7 mm to 1.25 mm.
These data indicate that the frequency of nodule detection is a function of slice thickness rather than the geographic location of screening. In response to this high rate of false-positive results, many researchers decide to call a scan “negative” if the largest nodule detected is less than 4 or 5 mm.
Doing this serves to reduce the number of false positives; however, this is at the cost of increasing the number of false negatives. This author would recommend against calling a nodule of any size, even very small, “negative,” as it gives the wrong information; one should be able to convey lack of immediate concern for a tiny nodule without completely ignoring it. The consequence to calling tiny nodules “negative” will be to miss lung cancers and increase the number of false negatives (Fig. 1). Regarding this point, in one study 18% of cancers identified were actually first detected when less than 4 mm.
Fig. 1Progressive growth of a tiny nodule. (A) Baseline screening computed tomography (CT) scan demonstrates a 2-mm nodule in the left upper lobe. (B) Follow-up CT scan at 1 year demonstrates slight growth to 3 mm. (C) Follow-up CT scan at 2 years demonstrates further growth to 6 mm. At resection, the nodule was a stage IA adenocarcinoma. If the nodule had been deemed “negative” on the baseline CT scan, it would have been a false negative.
Overdiagnosis is the detection of a true cancer, but one that grows so slowly as to not compete with comorbidities for the eventual cause of death. Data from the observational CT-screening studies indicate that such tumors, which have long tumor doubling times, contribute to the apparent survival improvement observed with LDSCT screening. A study from Japan reported tumor doubling times ranging from 662 to 1486 days, with a mean of 880 days.
found that 23 of 40 (58%) prevalence cancers in a CT-screened population were found in subjects who were never smokers, suggesting that CT screening may identify a different population of lung cancers than nonscreened patients presenting with clinical symptoms. Volume doubling times were retrospectively calculated in the Mayo CT-screening study; 13 of 48 (27%) screening-detected cancers exhibited slow doubling times that were more than 400 days.
These relatively indolent cancers are often nonsolid radiographically, appearing as either pure ground-glass opacities (GGOs) or as partly solid. GGOs require a different approach than solid nodules.
GGOs may demonstrate stability for several years and then show evidence of change, thus violating the conventional 2-year “no change” rule indicating a solid nodule is benign (Fig. 2). GGOs may not grow in diameter, yet development of invasion may be evident as they become partially solid. Positron emission tomography (PET) is often falsely negative in GGOs and partly solid lesions. Concern for overdiagnosis is diminished, knowing that mortality is reduced with screening—even if nonlethal cancers are detected, lives saved through screening indicate that enough fast-growing cancers are found to make it worthwhile.
Fig. 2Slowly enlarging ground-glass opacity (GGO). (A) Baseline screening chest CT scan shows a 13-mm GGO. (B) Follow-up chest CT scan after 5 years demonstrates slow growth to 20 mm. Limited resection revealed a stage I adenocarcinoma in situ.
Surgery for nodules eventually found to be benign could be considered by some as having been unnecessary, but is part of the physical and financial cost of screening for lung cancer. In the Mayo CT-screening study, 18% of participants who underwent procedures performed were found to have a benign histology.
Similarly, in the study by Diederich and colleagues, the DANTE study, and the NLST study, benign nodules represented 20%, 22%, and 27% of resections, respectively.
Surgical procedures in the DANTE trial, a randomized study of lung cancer early detection with spiral computed tomography: comparative analysis in the screening and control arm.
Overall, CT-screening studies report benign nodule resection rates of between 15% and 30%. In the absence of a perfectly sensitive, noninvasive test, some benign nodules will be removed when showing growth, change in density characteristics, or avidity on PET imaging. However, the desire is clearly to keep benign resections to a minimum. The literature suggests that focused screening programs have done this better than usual practice, which is a concern with the anticipation of a generalized recommendation to screen high-risk patients. Recent surgical series report that benign nodules comprise as many as 50% to 86% of resected nodules.
This rate of benign resections is simply too high. Reducing the removal of benign nodules can be achieved in many circumstances by careful evaluation through review of old images, followed by serial examinations, or evaluating larger nodules with PET or needle biopsy. Published nodule guidelines can assist in these evaluations.
American College of Chest Physicians. Evaluation of patients with pulmonary nodules: when is it lung cancer? ACCP evidence-based clinical practice guidelines (2nd edition).
One other concerning issue is that diagnostic radiation associated with CT screening may actually cause some lung cancers. Radiation absorption is expressed as an effective dose in Sievert (Sv) or millisieverts (mSv) for dose distributions that arenonhomogeneous, as in x-ray radiation (1 mSv = 1 mGy). The average effective dose for a CT of the chest is about 7 mSv.
Committee to assess health risks from exposure to low levels of ionizing radiation. Health Risks from Exposure to low levels of Ionizing Radiation: BEIR-VII.
Age at radiation exposure is also a factor. The number of CT scans of the chest that would be required to cause one radiation-induced cancer is estimated to be 720 for 40-year-old women and 1566 scans for 40-year-old men. Older age reduces the eventual risk; for 60-year-old persons, one cancer would be induced for every 1090 CT scans for women and 2080 CT scans for men.
The NLST investigators estimated that the radiation risk from screening 55-year-old smokers results in 1 to 3 lung cancer deaths per 10,000 screened and 0.3 new breast cancers per 10,000 women screened.
This harm from screening highlights the importance of having proven cancer mortality reduction through an RCT.
Summary
Proof of mortality reduction has been shown with mammography for breast cancer and with fecal occult blood testing for colon cancer, and CT screening for lung cancer can now be added to that list; CT screening saves lives. Results of the nonrandomized trials provided insufficient evidence to move CT screening from the realm of personal choice to a matter of health policy. Due to the frequency and lethality of lung cancer, a 20% reduction in deaths certainly provides some long-awaited good news. The response of the ACS and US Preventive Services Task Force in making recommendations for CT screening in public policy is also anticipated. More information is forthcoming from the NLST regarding quality of life, smoking behavior, health care use, and cost effectiveness of screening.
Further studies such as the UK Lung Screen trial will define the appropriate population to screen. This study is set to open in 2012 and will use a validated method to calculate absolute risk of lung cancer based on age, sex, smoking duration, family history of lung cancer, history of nonpulmonary malignant tumor, history of pneumonia, and occupational exposure to asbestos.
What is a clinician to do in the meantime? The decision to case-find lung cancer with screening requires a discussion between the physician and the patient of the risks and potential benefits. Despite the potential harm in biopsies and surgery performed for benign lesions as well as the potential risks of radiation, the mortality reduction demonstrated in the NLST suggests that screening benefit outweighs the risk. Use of a risk-prediction model may identify individuals at comparable or even higher risk for lung cancer who would benefit from CT screening even though they do not fit the age and pack-year criteria of the NLST. A celebratory response to mortality reduction from screening must not diminish the focus on primary prevention through encouraging smoking cessation, promoting avoidance of smoking initiation, and limiting passive smoke exposure.
References
The National Lung Screening Trial Research Team
Reduced lung-cancer mortality with low-dose computed tomographic screening.
Writing Committee, Lung Screening Study Research Group
et al.
Baseline findings of a randomized feasibility trial of lung cancer screening with spiral CT scan vs chest radiograph: the Lung Screening Study of the National Cancer Institute.
Baseline results of the Depiscan study: a French randomized pilot trial of lung cancer screening comparing low dose CT scan (LDCT) and chest X-ray (CXR).
Surgical procedures in the DANTE trial, a randomized study of lung cancer early detection with spiral computed tomography: comparative analysis in the screening and control arm.
American College of Chest Physicians. Evaluation of patients with pulmonary nodules: when is it lung cancer? ACCP evidence-based clinical practice guidelines (2nd edition).
Committee to assess health risks from exposure to low levels of ionizing radiation. Health Risks from Exposure to low levels of Ionizing Radiation: BEIR-VII.
00090-6&id=gr1.jpg){kind=link}
00090-6&id=gr2.jpg){kind=link}