Benefits and Harms of CT Screening for Lung Cancer:  A Systematic Review FREE

Lung cancer is the leading cause of cancer death in the United States (and worldwide), causing as many deaths as the next 4 most deadly cancers combined (breast, prostate, colon, and pancreas).¹ Despite a slight decline in US lung cancer mortality rates since 1990, lung cancer will account for more than 160 000 deaths in the United States in 2012.²Quiz Ref IDMost patients diagnosed with lung cancer today already have advanced disease (40% are stage IV, 30% are stage III), and the current 5-year survival rate is only 16%.³

Quiz Ref IDEarlier randomized controlled trials (RCTs) involving chest radiographs and sputum cytology for lung cancer screening found that these strategies detected slightly more lung cancers, smaller tumors, and more stage I tumors, but the detection of a larger number of early-stage cancers was not accompanied by a reduction in the number of advanced lung cancers or a reduction in lung cancer deaths.^4- 14 Renewed enthusiasm for lung screening arose with the advent of low-dose computed tomography (LDCT) imaging, which is able to identify smaller nodules than can chest radiographs.

This systematic review focuses on the evidence regarding the benefits and harms of LDCT screening for lung cancer. Other potential screening methods (eg, chest radiographs, sputum cytology or biomarkers, exhaled breath) are not addressed. This review is a collaborative initiative of the American Cancer Society (ACS), the American College of Chest Physicians (ACCP), the American Society of Clinical Oncology (ASCO), and the National Comprehensive Cancer Network (NCCN) and forms the basis for the clinical practice guideline of the ACCP and ASCO. This work will be reassessed when pertinent new data become available, consistent with the Institute of Medicine recommendations for guideline development.¹⁵

The ACS, ACCP, ASCO, and NCCN assembled a panel of experts representing the relevant clinical disciplines and the consumer's perspective. All members cleared all organizations' conflict-of-interest policies for participation in guideline development; none received compensation for participation. The sponsoring organizations donated staff time supported by their general administrative funds. No industry funds were used.

The panel defined a process for selection, data extraction, and outcomes assessment to produce a thorough evaluation of LDCT screening relevant to patient-centered outcomes, including quantifying potential benefits and harms. The target patient population for this initiative is individuals at elevated risk of developing lung cancer because of age and smoking history; and the target audience includes physicians, allied professionals, and policy makers.

The panel was divided into evidence review and writing subcommittees, focusing on 4 key questions. (1) What are the potential benefits of screening individuals at elevated risk of developing lung cancer using LDCT? (2) What are the potential harms of screening individuals at elevated risk of developing lung cancer using LDCT? (3) Which groups are most likely to benefit or not benefit from screening? (4) In what setting is screening likely to be effective?

The literature search was developed and conducted by an experienced systematic reviewer using MEDLINE (Ovid: January 1996 to April 8, 2012), EMBASE (Ovid: January 1996 to April 8, 2012), and the Cochrane Library (April 20, 2012). Additional citations were gleaned from the reference lists of related papers and review articles. The literature search included Medical Subject Headings (MeSH) and Emtree headings and related text and keyword searches in a manner that combined terms related to lung cancer, population screening, and LDCT (eAppendix 1). The search was limited to published data only.

Studies were eligible for inclusion if they involved either an RCT using LDCT screening for lung cancer in one intervention group or a noncomparative cohort study of LDCT screening, provided they reported at least 1 of the following outcomes: lung cancer−specific or all-cause mortality, nodule detection rate, frequency of additional imaging, frequency of invasive diagnostic procedures (eg, needle or bronchoscopic biopsy, surgical biopsy, surgical resection), complications from the evaluation of suspected lung cancer, or the rate of smoking cessation or reinitiation. For lung cancer–specific and all-cause mortality end points, only RCT data were considered eligible for inclusion; for other end points, data from the LDCT group of both RCTs and cohort studies were included.

Exclusion criteria covered studies that only assessed screening among participants with risk factors other than smoking (eg, asbestos), those not published in English, and meta-analysis or case-series reports of outcomes only among patients diagnosed with lung cancer.

The exclusion criteria were determined a priori and guided whether data identified by the systematic literature review were judged to have been reported in a manner appropriate for inclusion. Articles were selected and data were extracted independently by a minimum of 2 reviewers. At the point of abstract review, if 1 of 2 reviewers indicated that a citation may be relevant, the full-text article was retrieved. After full text review, if there was a discrepancy among the 2 reviewers, a third reviewer determined eligibility, and the reviewers came to consensus. In addition, the third reviewer also verified that articles deemed ineligible did not meet eligibility criteria.

Between the 3 reviewers, discrepancies occurred in approximately 12% of cases and were resolved through consensus. Most notably, the small RCT by Garg et al¹⁶ and the smoking cessation study by Schnoll et al¹⁷ were originally excluded, but the decision was reversed on further review. Common reasons for exclusion included the identification of narrative reviews, studies that did not involve high-risk smoking populations, and studies that only followed up patients diagnosed with lung cancer. A full list of the studies excluded from the systematic review and the reasons for exclusion is available from the authors. For included studies, the risk of bias was assessed by a minimum of 2 reviewers using prespecified criteria (eAppendix 2), and discrepancies were resolved through consensus.

The frequency of nodule detection across studies was analyzed both unadjusted and stratified by multiple study design characteristics (eg, CT collimation, minimum smoking exposure criteria for study enrollment, stated threshold for labeling a finding “positive” or “suspicious”).

Eight RCTs (Table 1)^16,18- 27 and 13 cohort studies of LDCT screening (Table 2)^28- 47 were selected from 591 citations identified by the literature search (eAppendix 3). Two smaller RCTs are related to 2 larger RCTs: the Lung Screening Study (LSS) was a pilot study of the National Lung Screening Trial (NLST) and there is a prespecified plan to combine data from the Danish Lung Cancer Screening Trial (DLCST) with the Dutch Belgian Randomised Lung Cancer Screening Trial (NELSON). Several trials were ongoing with only preliminary data available. Two RCTs were excluded because they lacked data on key end points; 1 RCT and several cohort studies were excluded because they involved populations at risk because of factors other than smoking or were for general population screening. For studies reported in multiple publications, all reports were reviewed, but earlier reports superseded by additional data in later reports are not referenced.

The NLST and DLCST had a low risk of bias (eTable 1). Other studies had variable risks of bias, in part because only preliminary reports of ongoing studies are available. The risk of bias in the cohort studies was variable and often high (usually because they lacked justification for the sample size, a definition of the primary end point, or description of funding sources).

Across the RCTs, the minimum smoking history required for enrollment ranged from 15 to 30 pack-years (ie, cigarette packs smoked per day multiplied by years of smoking), with a maximum time since quitting smoking ranging from 10 years to an unlimited number of years (Table 1). The lower age limit ranged from 47 to 60 years, and the upper limit from 69 to 80 years. There was greater variation in entry criteria in the cohort studies (Table 2).

The underlying risk for lung cancer varied substantially between the studies.The NLST,²³ LSS,²⁵ and study by Garg et al¹⁶ generally focused on higher risk; DLCST,¹⁹ ITALUNG,²¹ and DANTE²² on both higher and intermediate risk; and NELSON¹⁸ and Dépiscan²⁷ on a broad range of risk among participants.

Although estimating the average risk of all participants in any of these studies is difficult because of a lack of granular data, the minimum risk level in each study was approximated using established formulas.^48- 49 Over 10 years, the risks of being diagnosed with lung cancer for participants meeting minimum entry criteria of each study, assuming they had quit smoking at time of study entry, were approximately 2% for NLST, 1% for DLSCT, and considerably less than 1% for NELSON. The nodule size deemed large enough to investigate further ranged from “any size” to greater than 5 mm; the size that triggered an invasive intervention (when specified) ranged from 6 to 15 mm.

Effect on Mortality. Three RCTs reported the effect of LDCT screening on lung cancer–specific mortality (Table 3). The NLST found that 3 annual rounds of screening (baseline and 1 and 2 years later) with LDCT resulted in a 20% relative decrease in deaths from lung cancer vs chest radiographs over a median of 6.5 years of follow-up (P = .004).²³ In absolute terms, the chance of dying from lung cancer was 0.33% less over the study period in the LDCT group (87 avoided deaths over 26 722 screened participants), meaning 310 individuals must participate in screening for typically 3 rounds to prevent 1 lung cancer death. Based on a slightly different denominator, the NLST authors reported the number needed to screen with LDCT was 320 to prevent 1 lung cancer death.

The considerably smaller ongoing DANTE and DLCST studies each compared 5 annual rounds of LDCT screening to usual care; after a median of 34 and 58 months of follow-up, respectively, no statistically significant difference in lung cancer mortality was observed in either study (DANTE: relative risk [RR], 0.97; 95% CI, 0.71-1.32; P = .84; DLCST: RR, 1.15; 95% CI, 0.83-1.61; P = .43).^19,22

All 3 studies reported on the risk of death from any cause (Table 4) between study groups and directly or indirectly on the risk of death from any cause other than lung cancer. Only the NLST found a difference in this end point, with fewer deaths overall in the LDCT vs the chest radiograph group (1303 vs 1395 deaths per 100 000 person-years, respectively). Analyses focusing exclusively on deaths not due to lung cancer found no significant differences in any of the 3 studies.²³

Effect on Smoking Behavior. Speculation exists that undergoing LDCT screening may result in justification of continued smoking or, alternatively, may represent an opportunity for smoking cessation interventions. But studies examining the smoking behavior of LDCT-screened individuals have not found evidence that cessation or reinitiation rates are meaningfully altered by participation in screening (eTable 2).^50- 52

Detection of Abnormalities. Low-dose computed tomography identifies both cancerous and benign noncalcified nodules; the latter are often called “false positives.” Although most LDCT screening studies have reported on nodules detected, the categorization and manner of reporting are inconsistent (eg, it is sometimes unclear if newly identified nodules are assigned to that round or to an earlier round if they can be retrospectively seen on an earlier LDCT). Likewise, size thresholds that would trigger an invasive workup are variously and inconsistently reported, as are the potential denominators for estimating false-positive rates, such as per screening round or per person-year.

Based on the study's own size cutoffs, the average nodule detection rate per round of screening was 20% (Table 5, eFigure 1) but varied from 3% to 30% in RCTs and 5% to 51% in cohort studies. Most studies reported that more than 90% of nodules were benign. In general, there is a tendency toward lower nodule detection rates in repeat screening rounds, but the data and reporting are inconsistent (Table 5, eFigure 2). In the NLST, the rate of detection did not decrease until the third round. In that round, the study protocol allowed for ignoring nodules that had been present in the prior rounds. We were unable to find any statistically significant relation between study parameters, such as smoking history of study enrollees, CT scan settings, nodule size cutoffs, and reported nodule detection rates.

Most often, a detected nodule triggered further imaging, but the underlying management protocols were inconsistently reported in the studies. Whether all additional imaging tests were captured in the studies was also uncertain: reported follow-up imaging rates may be underestimated.

The frequency of further CT imaging among screened individuals ranged from 1% in the study by Veronesi et al²⁷ to 44.6% in the study by Sobue et al.³² The frequency of further positron emission tomography (PET) imaging among screened individuals exhibited much less variation, ranging from 2.5% in the study by Bastarrika et al³⁹ to 5.5% in the NLST.²³ The frequency of invasive evaluation of detected nodules was generally low but varied considerably (Table 6, eFigure 3). No patterns were apparent that explained this heterogeneity. In the NLST, 1.2% of patients who were not found to have lung cancer underwent an invasive procedure such as needle biopsy or bronchoscopy, while 0.7% of patients who were not found to have lung cancer had a thoracoscopy, mediastinoscopy, or thoracotomy.²³ In the NELSON study, these numbers were 1.2% and 0.6%, respectively.¹⁸ Invasive nonsurgical procedures in patients with benign lesions were common (eg, 73% in NLST).

Complications of Diagnostic Procedures. The only study reporting on complications resulting from LDCT screening is the NLST. Overall, the frequency of death occurring within 2 months of a diagnostic evaluation of a detected finding was 8 per 10 000 individuals screened by LDCT and 5 per 10 000 individuals screened by chest radiographs. Some of the deaths that occurred after a diagnostic evaluation were presumably unrelated to follow-up procedures, as 1.9 and 1.5 per 10 000 occurred within 2 months when the diagnostic evaluation involved only an imaging study.

Deaths most clearly related to follow-up procedures were those occurring within 2 months when the most recent procedure was a bronchoscopy or needle biopsy (3.4 per 10 000 screened by LDCT and 2.2 per 10 000 screened by chest radiographs). Approximately one-third of the deaths occurred within 2 months of a surgical procedure in both study groups, and the vast majority of these were in the patients with cancer, suggesting perhaps that the surgical procedures in those with cancer were more extensive (ie, resection rather than biopsy; such details were not reported). The 60-day perioperative mortality for patients with lung cancer who underwent a surgical procedure was 1% for the LDCT group and 0.2% for the chest radiographs group.

Overall, the frequency of major complications occurring during a diagnostic evaluation of a detected finding was 33 per 10 000 individuals screened by LDCT and 10 per 10 000 individuals screened by chest radiographs. The rate of (presumably unrelated) complications following imaging alone was similar and low (1.1 and 1.5 per 10 000 screened, respectively); the complication rate after a bronchoscopy or needle biopsy was also low (1.5 and 0.7 per 10 000 for LDCT and chest radiographs, respectively). The vast majority of major complications occurred after surgical procedures and in those patients with lung cancer. The rate of major complications in those patients with lung cancer who underwent surgery was 14%.

Focusing only on patients who had nodules detected by LDCT that were determined to be benign, death occurred within 60 days among 0.06% and major complications occurred among 0.36%. Approximately half of the deaths occurred after imaging alone, whereas the majority of major complications occurred after a surgical procedure (details unknown). Calculating these numbers for an entire screened population, the risks of death and major complications following diagnostic events (including imaging) for nodules that were determined to be benign is 4.1 and 4.5 per 10 000, respectively. These rates are higher than those in the chest radiographs group (1.1 and 1.5 per 10 000 for risks of death and major complications, respectively).

Overdiagnosis. Overdiagnosis refers to histologically confirmed lung cancers identified through screening that would not affect the patient's lifetime if left untreated. This includes patients who are destined to die of another cause (eg, a comorbidity or an unexpected event).⁵⁴ Earlier studies suggested that chest radiographs screening may have an overdiagnosis rate of roughly 25%.^55- 56 The overdiagnosis rate for LDCT screening cannot yet be estimated; NLST data show a persistent gap of about 120 excess lung cancers in the LDCT group vs the chest radiographs group, but further follow-up is needed.

Radiation Exposure. The effective dose of radiation of LDCT is estimated to be 1.5 mSv per examination, but there is substantial variation in actual clinical practice. However, diagnostic chest CT (~8 mSv)⁵⁷ or PET CT (~14 mSv)^57- 59 to further investigate detected lesions rapidly increases the exposure and accounts for most of the radiation exposure in screening studies. We estimate that NLST participants received approximately 8 mSv per participant over 3 years, including both screening and diagnostic examinations (averaged over the entire screened population). Estimates of harms from such radiation come from several official bodies and commissioned studies,^60- 61 based on dose extrapolations from atomic bombings and also many studies of medical imaging.^62- 63 Using the NLST data, these models predict that approximately 1 cancer death may be caused by radiation from imaging per 2500 persons screened.

Quiz Ref IDTherefore, the benefit in preventing lung cancer deaths in NLST is greater than the radiation risk—which only becomes manifest 10 to 20 years later. However, for younger individuals or those with lower risk of developing lung cancer, the trade-off would be less favorable. Preliminary modeling studies suggest that potential risks may vastly outweigh benefits in nonsmokers or those aged 42 years or younger.⁶⁴ Further study is needed, including evaluation of the effects of ongoing annual LDCT beyond 3 successive years.

Quality of Life. The effect of LDCT screening on quality of life is uncertain. We found only 1 study, in which 88% to 99% of 351 participants reported no discomfort, but 46% reported psychological distress while awaiting results.⁶⁵ Although there may be quality-of-life benefits due to lower morbidity from advanced lung cancer, there are also potential detriments due to anxiety, costs, and harms from the evaluation of both false-positive scans and overdiagnosed cancers.

The NLST population is the only one for whom a lung cancer mortality benefit from LDCT has been demonstrated (those aged 55-74 years with ≥30 pack-years of smoking who quit ≤15 years prior to entry [if they have stopped smoking]). Other studies are too small, too preliminary, or too poorly designed to support meaningful conclusions. The value of LDCT screening is likely determined primarily by the risk of lung cancer vs competing causes of death. Little information exists regarding comorbidities, but presumably the NLST participants were deemed healthy. We estimate an average risk of developing lung cancer within 10 years of approximately 10% for the NLST population in the absence of screening (estimated median age 62 years and ~50 pack-years of smoking). However, the calculable risk for individual NLST participants most likely varied by more than 10-fold across the participants, from less than 2% to greater than 20%, and it is unclear which groups experienced benefit.^48- 49 But there is no evidence base for determining how selection criteria for screening should be refined. Incorporating other well-known risk factors has not been studied.

A summary of the setting of the NLST (the only positive study) demonstrates that most (76%) of the NLST sites were National Cancer Institute–designated cancer centers, and 82% were large academic medical centers with more than 400 hospital beds, although screening may have taken place at satellite facilities in some cases (eTable 3). All of these centers are likely to have specialized thoracic radiologists and board-certified thoracic surgeons on staff. The CT scanners used in the NLST underwent ongoing extensive quality control, and the scans were interpreted by chest radiologists who underwent specific training and quality control in the interpretation of images and wording of screening LDCT findings.⁵⁸ A nodule management algorithm was included in the NLST, but adherence or the setting in which nodules were managed was not mandated or tracked by the study.⁵⁸

Most other RCTs and cohort studies of LDCT screening were conducted in facilities similar to the NLST sites: academic medical centers, large hospitals, with the involvement of relevant subspecialist services and a defined nodule management algorithm. The association between the setting of LDCT screening and outcome has not been tested, but variability in rates of false-positive LDCT scans, further imaging, and procedures suggests these may be important.

This report summarizes the systematic review conducted by a multisociety collaborative effort examining the risks and benefits of LDCT screening for lung cancer and forms the basis of the American College of Chest Physicians and the American Society of Clinical Oncology clinical practice guideline (Box and eAppendix 4). The guideline is based on the finding that a reasonable amount of data has been reported regarding the outcomes for LDCT screening for lung cancer and that some conclusions can be drawn regarding its risks and benefits despite many areas of uncertainty.

Recommendation 1

For smokers and former smokers aged 55 to 74 years who have smoked for 30 pack-years or more and either continue to smoke or have quit within the past 15 years, we suggest that annual screening with low-dose computed tomography (LDCT) should be offered over both annual screening with chest radiograph or no screening, but only in settings that can deliver the comprehensive care provided to National Lung Screening Trial (NLST) participants. (Grade of recommendation: 2B.)

Remark 1

Counseling should include a complete description of potential benefits and harms (as outlined in the full guideline text) so the individual can decide whether to undergo LDCT screening.

Remark 2

Screening should be conducted in a center similar to those where the NLST was conducted, with multidisciplinary coordinated care and a comprehensive process for screening, image interpretation, management of findings, and evaluation and treatment of potential cancers.

Remark 3

A number of important questions about screening could be addressed if individuals who are screened for lung cancer are entered into a registry that captures data on follow-up testing, radiation exposure, patient experience, and smoking behavior.

Remark 4

Quality metrics should be developed such as those in use for mammography screening, which could help enhance the benefits and minimize the harms for individuals who undergo screening.

Remark 5

Screening for lung cancer is not a substitute for stopping smoking. The most important thing patients can do to prevent lung cancer is not smoke.

Remark 6

The most effective duration or frequency of screening is not known.

Recommendation 2

For individuals who have accumulated fewer than 30 pack-years of smoking or are either younger than 55 years or older than 74 years, or individuals who quit smoking more than 15 years ago, and for individuals with severe comorbidities that would preclude potentially curative treatment, limit life expectancy, or both, we suggest that CT screening should not be performed. (Grade of recommendation: 2C.)

Full text of the American College of Chest Physicians and the American Society of Clinical Oncology evidence-based practice guideline on the role of CT screening for lung cancer is available in eAppendix 4. This guideline has been endorsed by the American Thoracic Society.

A recent large, high-quality RCT (the NLST) found that annual LDCT screening reduced the relative risk of death from lung cancer by 20% and the absolute risk by 0.33% in a population with a substantially elevated risk for lung cancer. Two smaller RCTs (DANTE and DLCST) comparing LDCT with usual care found no benefit of LDCT screening but are best interpreted as neither confirming nor contradicting the NLST findings. Because a recent large RCT (N = 154 901) demonstrated no lung cancer mortality difference between chest radiographs screening and usual care, the interventions in these 3 studies are reasonably comparable.⁶⁶

Quiz Ref IDThe literature supports the conclusion that LDCT screening can lead to harm. It identifies a relatively high percentage of subjects with nodules (average ~20%), the vast majority of which are benign. The additional imaging that these nodules trigger increases radiation exposure. The rates of surgical biopsy are variable (<1%-4%) as are the percentages of surgical procedures performed for benign disease. The rate of major, and sometimes fatal, complications among persons with benign lesions is low.

The unexplained heterogeneous rates of nodule detection, additional imaging, and invasive procedures that occurred within the structured settings of the controlled trials of LDCT raise concerns about how easily LDCT can be more broadly implemented. There is already substantial variability in the United States in the rates and complications of pulmonary needle biopsy⁶⁷ and outcomes of lung cancer surgery, which are considerably better in dedicated centers (such as those conducting LDCT trials).^68- 69 Furthermore, adherence with screening is consistently lower in cohort studies than in the NLST and could be worse with unstructured implementation, with resulting diminished benefits. Analogous concerns in breast cancer screening led to the Mammography Quality Standards Act.⁷⁰ The position statement by the International Association for the Study of Lung Cancer recommends demonstration projects to evaluate implementation of LDCT screening, establishment of quality metrics, and multiple task forces to address the many critical areas of uncertainty.⁷¹ Given all of these issues, performing an LDCT scan outside of a structured organized process appears to be beyond the current evidence base for LDCT lung cancer screening.

The fear and anxiety that patients can experience once there is even a slight suspicion of lung cancer highlights the need for careful education of LDCT participants and the need for carefully worded scan interpretations. Quiz Ref IDIn addition, even a small negative effect of screening on smoking behavior (either lower quit rates or higher recidivism) could easily offset the potential gains in a population.⁷² Smoking cessation should be considered a valuable component of any screening program.

In the setting of increasing health care costs, the relative cost-effectiveness of LDCT screening compared with other interventions will be a topic of discussion and concern in policy spheres. Medicare is allowed to contemplate a preventive service’s cost-effectiveness before adding it to the package of preventive benefits (Medicare Improvements for Patients and Providers Act of 2008⁷³). Now that an estimate is available of effectiveness, an estimate of cost-effectiveness could be generated.

Some critical elements of such an analysis will include determining what the price of the component services will be and how frequently follow-up procedures will be required.

The ACCP and ASCO recommendations for LDCT screening should be interpreted in light of several limitations. We did not conduct a formal cost-effectiveness evaluation. LDCT would be expected to be less cost-effective when applied to individuals at lower risk of lung cancer, because more individuals will need to be screened to prevent each death from the disease. Making screening available in settings without an organized approach to the evaluation and management of LDCT findings may also lower cost-effectiveness, if the frequency of unnecessary interventions and procedures is higher in these settings.^74- 76

Second, the data on which to base these recommendations are relatively limited. Although LDCT screening appears promising, it is also a clinical intervention in its infancy. Many questions that clinicians and patients might reasonably ask when considering whether or not to pursue screening remain unanswered. How large are the risks from radiation? How does an individual's smoking history interplay with the size of the expected benefit or the risk of harms? How serious a problem is overdiagnosis?

Third, it is not clear whether individuals across the range of risk in the NLST are all sufficiently likely to benefit that all such individuals should be considered for screening, or alternatively if a narrower or broader group should be targeted to achieve an appropriate balance between benefit, costs, and harms. How often to perform screening, or over what period of time it should continue, are also important questions that have not been answered by the available data. It is possible to speculate that benefits of screening could be enhanced if screening were continued for longer periods, but the risks could be amplified as well.

A substantial amount of data on LDCT screening should be reported in the near future, including numerous planned analyses of the NLST data both by its investigators and by the Cancer Intervention and Surveillance Modeling Network (CISNET) investigators. The ongoing RCTs in Europe will also be reporting estimates of both the magnitude of LDCT's mortality benefit and the extent of its harms soon. These data may help inform some of the important questions that still linger regarding LDCT screening.

Screening a population of individuals at a substantially elevated risk of lung cancer most likely could be performed in a manner such that the benefits that accrue to a few individuals outweigh the harms that many will experience. However, there are substantial uncertainties regarding how to translate that conclusion into clinical practice.