Annals of Internal Medicine

Article

Cumulative Incidence of False-Positive Test Results in Lung Cancer Screening

A Randomized Trial
Jennifer M. Croswell, MD, MPH; Stuart G. Baker, ScD; Pamela M. Marcus, PhD; Jonathan D. Clapp, BS; and Barnett S. Kramer, MD, MPH

Background: Direct-to-consumer promotion of lung cancer screening has increased, especially low-dose computed tomography (CT). However, screening exposes healthy persons to potential harms, and cumulative false-positive rates for low-dose CT have never been formally reported. Objective: To quantify the cumulative risk that a person who participated in a 1- or 2-year lung cancer screening examination would receive at least 1 false-positive result, as well as rates of unnecessary diagnostic procedures. Design: Randomized, controlled trial of low-dose CT versus chest radiography. (ClinicalTrials.gov registration number: NCT00006382) Setting: Feasibility study for the ongoing National Lung Screening Trial. Patients: Current or former smokers, aged 55 to 74 years, with a smoking history of 30 pack-years or more and no history of lung cancer (n 3190). Intervention: Random assignment to low-dose CT or chest radiography with baseline and 1 repeated annual screening; 1-year follow-up after the final screening. Randomization was centralized and stratified by age, sex, and study center.

Measurements: False-positive screenings, defined as a positive screening with a completed negative work-up or 12 months or more of follow-up with no lung cancer diagnosis. Results: By using a KaplanMeier analysis, a persons cumulative probability of 1 or more false-positive low-dose CT examinations was 21% (95% CI, 19% to 23%) after 1 screening and 33% (CI, 31% to 35%) after 2. The rates for chest radiography were 9% (CI, 8% to 11%) and 15% (CI, 13% to 16%), respectively. A total of 7% of participants with a false-positive low-dose CT examination and 4% with a false-positive chest radiography had a resulting invasive procedure. Limitations: Screening was limited to 2 rounds. Follow-up after the second screening was limited to 12 months. The false-negative rate is probably an underestimate. Conclusion: Risks for false-positive results on lung cancer screening tests are substantial after only 2 annual examinations, particularly for low-dose CT. Further study of resulting economic, psychosocial, and physical burdens of these methods is warranted. Primary Funding Source: National Cancer Institute.
Ann Intern Med. 2010;152:505-512. For author affiliations, see end of text. www.annals.org

espite the lack of a completed randomized, controlled trial demonstrating the efcacy of low-dose computed tomography (CT) in reducing mortality from lung cancer, its use as a screening tool is gaining increased attention in the past several years. Some hospitals and advocacy organizations have actively promoted CT screening to the public. A 2007 New York Times article quotes the director of surgical oncology at Greenwich Hospital, Greenwich, Connecticut, as predicting that within the next ve years, lung cancer screening will be routine, like mammography and colonoscopy (1). One advocacy group mounted a national Demand a CAT Scan billboard campaign in 2008 (2). The Lung Cancer Mortality Reduction Act of 2009 Senate bill states that signicant and rapid improvements in lung cancer mortality can be expected through greater use and access to lung cancer screening tests (3). Utilization rates of chest radiography or CT screening in the community setting have not been well studied. The DutchBelgian randomized lung cancer (NELSON) (4) screening trial reported that 3.1% of participants received a screening chest radiography or CT examination outside the trial by 24 months after randomization (this may not be representative of U.S. rates). Surveys of U.S. community physicians have demonstrated high enthusiasm for screening. One study found that two thirds of family practitio-

ners, internists, and gynecologists and 82% of general surgeons recommended chest radiography for lung cancer screening every 1 to 2 years (5). However, as is the case with all medical interventions, screening tests may generate both benets and harms. If lung cancer screening becomes national health policy, we must have solid evidence not only about the benets but also the harms of testing. This is particularly important because asymptomatic persons are the target population. Major harms associated with screening include the risk for overdiagnosisthe discovery of indolent lung cancer that would not lead to a persons death or cancer in a person

See also: Print Editors Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 506 Summary for Patients. . . . . . . . . . . . . . . . . . . . . . . I-40 Web-Only Appendix Appendix Tables CME quiz Conversion of graphics into slides
20 April 2010 Annals of Internal Medicine Volume 152 Number 8 505

www.annals.org

Article
Context

False-Positive Test Results in Lung Cancer Screening

The ongoing National Lung Screening Trial aims to define the effectiveness of screening for lung cancer. However, imaging studies to screen for lung cancer are currently marketed to patients.

Contribution
These data from a pilot study for the National Lung Screening Trial show a 33% cumulative incidence of falsepositive results after 2 computed tomography examinations and 15% after 2 chest radiography examinations. Substantial proportions of patients (7% for computed tomography and 4% for chest radiography) with falsepositive results required invasive testing to determine that the screening-detected lesion was not cancer.

All participants signed a consent form approved by the institutional review board before randomization. Eligible participants were aged 55 to 74 years, had a cigarette smoking history of 30 pack-years or more, and were current smokers or had quit in the past 10 years. Exclusion criteria included chest CT within 24 months of enrollment, previous lung cancer, removal of part or all of a lung, current treatment of any cancer except nonmelanoma skin cancer, and ongoing participation in a cancer prevention or screening trial other than a smoking cessation study.
Randomization and Interventions

Implication
Physicians and patients should bear in mind high falsepositive rates when considering screening for lung cancer with computed tomography or chest radiography. The Editors

who would die of a competing cause rst (6, 7)and the risk for false-positive results. False-positive results are important because they may have negative psychological effects (8, 9), affect future adherence to other preventive health measures (10, 11), and generate physical harms and economic costs from surveillance visits and conrmatory procedures. Although the Lung Screening Study has reported the total positivity rate, we have not previously examined cumulative false-positivity rates by using formal statistical methods, and the false-positive component most accurately represents a clinically important burden of screening. We focus on the probability of false-positive test results and resulting diagnostic procedures when chest radiography and CT are used as early detection strategies for lung cancer.

METHODS
Design

Once eligibility was established and consent was obtained by a study center, participants were randomly assigned to a treatment group through a single centralized, secure, Web-based system (which generated random code) operated by the trial coordinating center. This process ensured allocation concealment for study site investigators. Randomization was stratied by age group (in 5-year categories), sex, and study center by using variable block sizes. Once randomization occurred, participants and study investigators were not blinded to the screening method received. Two screening examinations were possible: baseline (T0) and repeated examination (T1) 1 year later. Participants were eligible for the second screening if they did not receive a diagnosis of lung cancer after the rst examination. For inclusion in this analysis of false-positive results, participants had to adhere to at least 1 screening. Low-dose CT scans were obtained with the following technical parameters: 120 to 140 kV peak, 60 ma, scan time of 1 s, 5-mm collimation, pitch of 2 or equivalent, and contiguous reconstructions. Chest radiography consisted of single posteroanterior views and was obtained by using high-kilovolt equipment at a tube-to-receiver distance of 6 to 10 feet. Each study center had 1 or more (range, 1 to 14) board-certied radiologists interpreting the examinations. A second radiologist blinded to the initial interpretation as a quality-control measure reread a small sample of lms (n 20) at each center.
Outcomes and Follow-up

The Lung Screening Study was a 2-year study conducted by 6 centers participating in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial. It was a feasibility study for the ongoing National Lung Screening Trial (12, 13).
Setting and Participants

The Lung Screening Study had a goal of randomly assigning 3000 participants at elevated risk for lung cancer. Enrollment was achieved through mass mailings of recruitment materials, along with public service announcements, posters, and physician recruitment efforts. A total of 3318 persons were randomly assigned from September 2000 to January 2001 to have chest radiography or low-dose CT.
506 20 April 2010 Annals of Internal Medicine Volume 152 Number 8

For CT, the denition of a positive screening result changed slightly between the T0 and T1 scans to better match accumulating prognostic evidence: Noncalcied nodules larger than 3 mm at T0 scan or 4 mm or larger at T1 scan were considered suspicious for cancer. Other abnormalities (including spiculated noncalcied nodules of any size; focal parenchymal opacication; endobronchial lesions; hilar, mediastinal, bony, or pleural masses; and major atelectasis) could also be deemed positive according to the radiologists judgment. For chest radiography, nodules with circular opacity of 3.0 cm or less in diameter, masses greater than 3.0 cm, hilar or mediastinal lymph node enlargement (excepting calcied nodes), major atelectasis, inwww.annals.org

False-Positive Test Results in Lung Cancer Screening

Article

ltrates or consolidation, and pleural masses were considered suggestive of cancer. We dened a false-positive screening result as a positive screening with a completed negative work-up or follow-up of at least 12 months with no diagnosis of lung cancer. Because performing biopsy of all screeningdetected lung abnormalities was impractical and undesirable (because of the potential for harm), we had to choose a denition of a false-positive test result that relied on diagnostic work-ups for suspicious examinations. For persons who did not receive denitive testing, given that lung cancer as a general rule is one of the more aggressive tumors, we felt that a 12-month monitoring period was a reasonable cutoff for a false-positive result. We dened a false-negative examination result as a negative screening associated with a diagnosis of lung cancer within 1 year of the examination. This denition is limited in its ability to discern between types of cancer truly missed by screening and aggressive interval tumors that may develop between tests. Furthermore, because the Lung Screening Study was a feasibility trial, follow-up of negative examination ndings was not done in the same systematic manner as for positive test results. Persons in whom screening was negative at T1 did not continue to have follow-up in the trial; reported falsenegative rates are limited to the period between the T0 and T1 examinations. All positive results were communicated by telephone and mailed to participants and their designated physician within 3 weeks. The Lung Screening Study did not specify a diagnostic algorithm for follow-up of positive results; centers would provide recommendations for diagnostic action if requested. Center personnel abstracted medical records relating to follow-up of positive screening results. This process began after a positive screening result and continued until a conclusive diagnosis was made or 12 months had passed. In addition, study participants completed a study update form at the T1 screening to identify any interval cases of lung cancer. Classications for diagnostic follow-up were divided into categories by author consensus: imaging examinations (noninvasive), minimally invasive procedures (bronchoscopy), moderately invasive procedures (for example, biopsy, thoracentesis, video-assisted thoracoscopy), and major surgical procedures (thoracotomy or lung resections).
Statistical Analysis

false-positive result at baseline or second screening received (14). For the base-case analysis, we considered persons with incomplete follow-up ( 12 months) after a positive screening to have received a false-positive result. As a sensitivity analysis, we assumed that a proportion of persons with insufcient follow-up after a positive examination result did have cancer, in which the proportion was estimated as published positive predictive values (from other trials) of CT (7%) and chest radiography (2%) for screening detection of lung cancer (15, 16). Logistic regression was done through 2 models to identify potential participant characteristics associated with increased odds of a false-positive examination result after the rst screening or the second screening (if the rst screening was negative). Variables included age, current versus former smoking status, and smoking history of 60 pack-years or more versus 30 to 59 pack-years. We adjusted logistic regression analyses for screening center. We calculated odds ratios (ORs) separately for CT and chest radiography. We also examined variance in the falsepositive rate by study center.
Role of the Funding Source

The Lung Screening Study was performed under contracts awarded by the National Cancer Institute, and all data were collected under those contracts. Most of the authors who designed, analyzed, and wrote this study are federal employees.

RESULTS
Demographic Characteristics

Table 1 shows demographic characteristics of the study population. Participants were more likely to be men (59%) and current smokers (57%). As expected, randomization achieved balance in baseline patient characteristics.

Table 1. Demographic Characteristics

Characteristic Low-Dose CT (n 1610), n (%) 516 (32) 1094 (68) Chest Radiography (n 1580), n (%) 502 (32) 1078 (68)

Age 6574 y 5564 y Sex Female Male Smoking status Current Former Smoking history 60 pack-years 3059 pack-years

Our study attempts to answer the question, What is the probability that a person entering a lung cancer screening program involving 1 or 2 screening tests will have at least 1 false-positive CT or chest radiography? A person could contribute to the cumulative risk curve only once, after a rst false-positive test result; this avoided doublecounting of suspicious nodules and articial ination of the curve. KaplanMeier analysis generated cumulative incidence curves based on estimated probability of a rst
www.annals.org

675 (42) 935 (58)

637 (40) 943 (60)

929 (58) 681 (42)

897 (57) 683 (43)

673 (42) 937 (58)

687 (43) 893 (57)

CT

computed tomography.
20 April 2010 Annals of Internal Medicine Volume 152 Number 8 507

Article

False-Positive Test Results in Lung Cancer Screening

Adherence Rates

A total of 1610 participants underwent at least 1 CT, and 1580 participants underwent at least 1 chest radiography (Figure 1). A total of 1374 participants in the CT group (97%) and 1287 participants in the chest radiography group (95.2%) received both examinations. Adherence

was lower in both groups for the second screening than for the baseline test.
False-Positive and False-Negative Results

Figure 1. Study flow diagram.

Mailings (n = 653 417)

Contacted screening center (n = 12 270)

Assessed for eligibility (n = 4828) Excluded (n = 1510) Spiral CT in past 2 y: 148 Did not consent: 1362* Randomly assigned (n = 3318)

A total of 31% (n 506) of participants in the CT group and 14% (n 216) of participants in the chest radiography group received at least 1 false-positive result. In comparison, screening CT was true-positive in 38 instances (2% of participants) and chest radiography was true-positive in 16 instances (1% of participants). Figure 2 illustrates the cumulative risk that a person would receive at least 1 false-positive result in a lung cancer screening program over several years. At baseline screening, the risk for a false-positive result is 21% (95% CI, 19% to 23%) for CT and 9% (CI, 8% to 11%) for chest radiography. These risks increase to 33% (CI, 31% to 35%) and 15% (CI, 13% to 16%), respectively, at second examination. Sensitivity analysis, which assumed that 7% of participants in the CT group and 2% of participants in the chest radiography group with insufcient follow-up after a positive screening had true-positive results, yielded identical rates. Four examinations with false-negative results were reported between the baseline and T1 examinations. All were in the chest radiography group (0.2% of participants in this group).
Diagnostic Follow-up and Invasive Procedures

Allocated to low-dose CT (n = 1660) Received T0 screening: 1586 Did not receive T0 screening: 74 Received T1 screening: 1398 Did not receive T1 screening: 262

Allocated to chest radiography (n = 1658) Received T0 screening: 1550 Did not receive T0 screening: 108 Received T1 screening: 1317 Did not receive T1 screening: 341

T0 screening (n = 1586) Sufficient follow-up: 1580 Insufficient follow-up: 6 T1 screening (n = 1398) Sufficient follow-up: 1369 Insufficient follow-up: 29

T0 screening (n = 1550) Sufficient follow-up: 1545 Insufficient follow-up: 5 T1 screening (n = 1317) Sufficient follow-up: 1303 Insufficient follow-up: 14

Table 2 shows absolute rates of diagnostic follow-up by category. Of persons with at least 1 false-positive CT or chest radiography, 61% (n 308) and 51% (n 110), respectively, received 1 or more secondary imaging tests. The overall percentage of participants who had at least 1 invasive procedure as a result of a false-positive result was 7% for CT and 4% for chest radiography. Bronchoscopies were the most common invasive procedure resulting from false-positive CT (5% of participants with a false-positive result). The proportion of participants who had a moderately invasive procedure was 4% for false-positive CT and 3% for false-positive chest radiography. Rates of participants who had major surgical procedures for benign disease were similar (2%), although absolute numbers differed by a factor of 2 (8 CT recipients and 4 chest radiography recipients).
Participant Characteristics and False-Positive Risk

Analyzed (n = 1610) Excluded from analysis: missed or declined all screenings (n = 50)

Analyzed (n = 1580) Excluded from analysis: missed or declined all screenings (n = 78)

CT computed tomography. * There was a lag between eligibility assessment and randomization at each center. Once the target sample size (n 3000) was reached, persons who were eligible but had not yet consented were not randomly assigned unless they had already been invited to participate. Sufcient follow-up means that a person had a negative screening result, 12 months of documented follow-up after a positive screening result, or a completed diagnostic work-up after a positive screening examination. For the base-case analysis, persons with insufcient follow-up after a positive examination were included in the analysis but were assumed to have received a false-positive test result.
508 20 April 2010 Annals of Internal Medicine Volume 152 Number 8

We did multivariable analyses to investigate potential participant characteristics associated with increased odds of a false-positive result after rst or second screening received (in which the rst screening was negative). Our models included age (65 to 74 years vs. 55 to 64 years), smoking status (current vs. former), smoking history ( 60 pack-years vs. 30 to 59 pack-years), and study center.
First Screening Received

For CT, older versus younger age demonstrated a nonstatistically signicant trend toward increased odds of
www.annals.org

False-Positive Test Results in Lung Cancer Screening

Article

a false-positive screening result (OR, 1.28 [CI, 0.98 to 1.67]). This effect was not seen in the chest radiography group (OR, 1.11 [CI, 0.77 to 1.60]). Current versus former smoking status did not show an association with false-positive results (OR, 1.15 [CI, 0.88 to 1.49] for CT and 1.31 [CI, 0.92 to 1.88] for chest radiography). Greater number of pack-years smoked was not associated with increased odds of a false-positive result (OR, 0.92 [CI, 0.71 to 1.19] for CT and 1.02 [CI, 0.72 to 1.45] for chest radiography). The estimated probability of a false-positive result on the rst screening received varied by center. For the combined categories of age (65 to 74 years), current smoker, and smoking history of 60 pack-years or more, the estimate for CT varied from 10% to 42% across centers. For chest radiography, the estimated false-positive rate for the combined categories of age (65 to 74 years), current smoker, and smoking history of 60 pack-years or more varied from 3% to 19% across centers. The Appendix (available at www .annals.org) provides further details.
Second Screening Received

Figure 2. Cumulative probability (95% CI) of a false-positive result for a person who participated in a lung cancer screening program over several years.
50

Cumulative Probability of a

False-Positive Result, %

45 40 35 30 25 20 15 10 5 0 0

Low-dose CT Chest radiography 21% (19%23%)

33% (31%35%)

9% (8%11%)
1

15% (13%16%)
2

Screening Test Participants at risk, n Low-dose CT Chest radiography 1610 1580 1114 1183

Older versus younger age was not associated with odds of false-positive CT (OR, 1.20 [CI, 0.83 to 1.74]) but was associated with twice the odds of false-positive chest radiography (OR, 2.03 [CI, 1.23 to 3.36]). Current versus former smoking status was not associated with odds of false-positive CT (OR, 1.07 [CI, 0.75 to 1.52]) or chest radiography (OR, 0.85 [CI, 0.51 to 1.40]). More pack-years of smoking was associated with 1.5 times increased odds of a false-positive result for CT (OR, 1.53 [CI, 1.08 to 2.18]), but this was not observed in the chest radiography group (OR, 1.12 [CI, 0.68 to 1.85]).

The cumulative probability is for the rst false-positive result received from a number of tests done. Participants at risk for screening test 1 is the number of participants who received at least 1 screening test (at T0 if both T0 and T1 screenings were taken or at T1 if T0 screening was missed). Participants at risk for screening test 2 is the number of participants who received both tests (at T0 and T1) in whom the result of the rst test received was negative. See the Appendix (available at www .annals.org) for more detailed information. CT computed tomography.

DISCUSSION
To our knowledge, our study is the rst to formally evaluate a persons cumulative risk for receiving at least 1

false-positive test result in a program of low-dose CT screening for lung cancer over several years. The probability of a false-positive result is substantial after 2 annual examinations (33%). Of participants with a false-positive CT scan, 7% had an unnecessary invasive procedure and 2% had major surgery for benign disease. Cumulative false-positive rates associated with screening tests have been infrequently reported; most studies of this nature have focused on mammography (1720). An

Table 2. Rates of False-Positive Results That Prompted Diagnostic Follow-up*

Procedure Type Low-Dose CT (n Participants, n Imaging examinations Minimally invasive procedure Moderately invasive procedure Major surgical procedure Any invasive procedure 308 25 20 8 33 506) False-Positive Results, % 61 5 4 2 7 Chest Radiography (n Participants, n 110 6 7 4 9 216) False-Positive Results, % 51 3 3 2 4

CT computed tomography. * Approximately 3% of all participants with a positive screening result received no diagnostic follow-up. Total number of persons with at least 1 false-positive result in each group. A person could contribute only once to each category, even if he or she had more than 1 false-positive result. Bronchoscopy. Procedures included lung biopsy (any method except open surgery), lymph node biopsy, mediastinoscopy, mediastinotomy, thoracentesis, and video-assisted thoracoscopy. Procedures included lung resection and thoracotomy. Any invasive procedure describes the number and percentage of participants who received at least 1 minimally invasive, moderately invasive, or major surgical procedure. A person could contribute to this group only once, regardless of the number of procedures he or she had. Imaging examinations were not considered invasive procedures.
www.annals.org 20 April 2010 Annals of Internal Medicine Volume 152 Number 8 509

Article

False-Positive Test Results in Lung Cancer Screening

estimate of the cumulative false-positive rate for chest radiography in lung cancer screening as part of a larger evaluation of false-positive rates in screening programs that used multiple testing methods was about 14% after 2 screenings and 22% after 4 screenings, which aligns with our ndings (21). Previous studies have generally reported discovery rates of all noncalcied nodules in persons who had CT screening. These rates have varied widely, depending on screening frequency and other factors; the average range is 25% to 50% of participants (16, 22). Clinical experience suggests that, in general, the larger the nodule, the greater the suspicion that the lesion is or will become cancerous (23). Despite this rule of thumb, no uniform consensus on how best to categorize and manage lesions detected on CT exists. Several studies of CT for lung cancer screening used protocols that called for diagnostic follow-up of varying intensity for all detected nodules (24 29), although most lesions detected by low-dose CT are smaller than 4 mm (30). Fleischner Society guidelines for management of incidental nodules detected on nonscreening CT suggest that lesions smaller than 4 mm pose minimal risk, and as such, generally recommend no further follow-up for low-risk patients and a single repeated screening at 12 months for high-risk patients with no intervention if the lesion is unchanged (31). The denition of a positive test result used in the Lung Screening Study ( 3 mm or 4 mm, depending on year) attempted to rule out lesions that were least likely to be indicative of cancer; the study design sought to minimize the number of false-positive results. The relatively high cumulative risk for a false-positive result after 2 examinations may represent a conservative estimate of rates in community practice, in which all nodules, regardless of size, may be more likely to receive follow-up. Multislice scanners were emerging at the time of the Lung Screening Study, and some, but perhaps not all, of the study sites had 4-row scanners in use. It is not known what the effect of single versus multirow detectors would be on false-positive rates, nor the effect of additional (for example, 16 or 64) slices. Previous studies have reported on surgery rates for benign disease; the proportion of persons with a screeningdetected nonmalignant lesion who had surgery has ranged from about 0.6% to 2.7% for CT (26 28, 3235). This is consistent with our ndings. More than half of participants with false-positive chest radiography or CT had at least 1 additional imaging examinationsome at higher radiation doses than that of the original testwhich exposed these persons to a theoretical risk for radiation-induced carcinogenesis. This is potentially concerning in the target population because current evidence indicates that radiation and smoking damage interact synergistically, and the interaction is near multiplicative (36, 37). Consensus has not been reached on a single diagnostic algorithm for positive screening examinations; however, repeated scans (of varying dose) at 1, 3, 6, 12,
510 20 April 2010 Annals of Internal Medicine Volume 152 Number 8

and/or 24 months have been advocated (2329, 38). Although no long-term longitudinal data on the effects of cumulative radiation exposure are available, estimates of cancer risks have been done by using relevant data from cohort studies of survivors of long-term atomic bomb exposure (39, 40). One estimate found that among female smokers aged 60 years, a series of 3 low-dose chest CTs would generate an excess lung cancer risk for about 1.5 women per 1000 exposed (39). The National Lung Screening Trial is collecting information on cumulative radiation exposure among participants and should help clarify this important issue. Negative psychological consequences of false-positive screening are also of concern. Although data for lung cancer screening are limited, 1 study examining the effect of CT on quality of life found that about half of participants had discomfort and dread while waiting for conrmation of screening results (41). Because positive lung cancer screening results may encompass diagnostic uncertainty for up to 24 months, further investigation of the long-term psychological effects of false-positive test results would be important. A high rate of false-positive test results may place economic burdens on persons and the health care system. One study that investigated medical care expenditures triggered by false-positive test results found that the adjusted mean difference in spending after such a test was an additional $1024 for a woman and $1171 for a man in 1 year (42). Because 31% of participants in the CT group received at least 1 false-positive result in 2 years, the effect these rates could have on the costeffectiveness of CT (if proven effective) is apparent. Our study has several limitations. As with all cancer screening trials, a healthy volunteer effect is likely to be present. This effect has been documented in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial, which includes study centers that participated in the Lung Screening Study (43). The Lung Screening Study was a pilot study and, as such, was limited to 2 rounds of screening and 1 year of follow-up after the last examination. Although it is possible that this study might overestimate the cumulative risk for a false-positive result, it is unlikely that a large proportion of false-positive results would convert to diagnoses of cancer. A small proportion (0.04% for each method) of Lung Screening Study participants was lost to follow-up after positive examination ndings; this studys baseline assumption was that those ndings were false positive. Because this could potentially overestimate the cumulative falsepositive risk, a sensitivity analysis was done; however, the cumulative risk estimates remained unchanged. Because the Lung Screening Study was a feasibility trial, negative screening results were not systematically followed in the same manner as positive screening results. Recording of false-negative results was limited to the period between the T0 and T1 examinations, and the estimated false-negative rate is crude and probably an underestimate.
www.annals.org

False-Positive Test Results in Lung Cancer Screening

Article

The analysis of associations between patient characteristics and false-positive risk should be considered exploratory. The Lung Screening Study had a relatively modest sample size; subgroup populations were small; and, as in any multivariable analysis, multiplicity potentially inated the probability of a type I error. The variation in estimated false-positive rates by center is probably due to the effects of interobserver variability in the assessment of a positive examination among centers. Studies of interobserver agreement on interpretation of chest CT scans have demonstrated notable variations among readers (44 46). Semiautomated volumetric determinations of nodule size or computer-aided detection programs have been proposed as methods that could theoretically reduce observer variability, although the ultimate degree of effect this might have on false-positive rates is unknown. Given the relatively high probability of a false-positive low-dose CT lung cancer screening examination, it is important that providers have careful discussions with patients who request this technology to help them weigh known harms against currently theoretical benets. Further investigation into the physical, psychological, and economic ramications of false-positive low-dose CT screening test results is warranted.
From National Institutes of Health, National Cancer Institute, Bethesda, and Information Management Services, Rockville, Maryland. Portions of this work were presented at the annual meeting of the American Society of Clinical Oncology, Orlando, Florida, 29 May2 June 2009 (Abstract 1502).
Disclaimer: The views expressed in this article are those of the authors and do not necessarily represent the views of the U.S. federal government or the National Institutes of Health. Acknowledgment: The authors thank all participants in the Lung

Requests for Single Reprints: Jennifer M. Croswell, MD, MPH, Ofce of Medical Applications of Research, National Institutes of Health, 6100 Executive Boulevard, Suite 2B-03, Bethesda, MD 20892; e-mail, croswellj@od.nih.gov.

Current author addresses and author contributions are available at www .annals.org.

References
1. Steele MF. Hospitals urge screening in lung cancer detection. New York Times. 4 March 2007. Accessed at www.nytimes.com/2007/03/04/nyregion /nyregionspecial2/04CTtopic.html on August 19, 2008. 2. Bonnie J. Addario Lung Cancer Foundation. Awareness Campaign Launched in Four Major Cities Takes Aim at Lung Cancer. Accessed at www.lungcancerfoundation.org/projects-and-progress/bus-and-subway-campaign/ on 16 July 2009. 3. Lung Cancer Mortality Reduction Act of 2009, S. 332, 111th Cong. (2009). Accessed at www.govtrack.us/congress/billtext.xpd?bill s111-332 on 16 July 2009. 4. Baecke E, de Koning HJ, Otto SJ, van Iersel CA, van Klaveren RJ. Limited contamination in the Dutch-Belgian randomized lung cancer screening trial (NELSON). Lung Cancer. 2009. [PMID: 19801174] 5. Czaja R, McFall SL, Warnecke RB, Ford L, Kaluzny AD. Preferences of community physicians for cancer screening guidelines. Ann Intern Med. 1994; 120:602-8. [PMID: 8117000] 6. Black WC. Overdiagnosis: an underrecognized cause of confusion and harm in cancer screening [Editorial]. J Natl Cancer Inst. 2000;92:1280-2. [PMID: 10944539] 7. Marcus PM, Bergstralh EJ, Zweig MH, Harris A, Offord KP, Fontana RS. Extended lung cancer incidence follow-up in the Mayo Lung Project and overdiagnosis. J Natl Cancer Inst. 2006;98:748-56. [PMID: 16757699] 8. Brewer NT, Salz T, Lillie SE. Systematic review: the long-term effects of false-positive mammograms. Ann Intern Med. 2007;146:502-10. [PMID: 17404352] 9. Gilbert FJ, Cordiner CM, Afeck IR, Hood DB, Mathieson D, Walker LG. Breast screening: the psychological sequelae of false-positive recall in women with and without a family history of breast cancer. Eur J Cancer. 1998;34:2010-4. [PMID: 10070302] 10. Taylor KL, Shelby R, Gelmann E, McGuire C. Quality of life and trial adherence among participants in the prostate, lung, colorectal, and ovarian cancer screening trial. J Natl Cancer Inst. 2004;96:1083-94. [PMID: 15265970] 11. Ford ME, Havstad SL, Flickinger L, Johnson CC. Examining the effects of false positive lung cancer screening results on subsequent lung cancer screening adherence. Cancer Epidemiol Biomarkers Prev. 2003;12:28-33. [PMID: 12540500] 12. Gohagan J, Marcus P, Fagerstrom R, Pinsky P, Kramer B, Prorok P; Writing Committee, Lung Screening Study Research Group. Baseline ndings 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. Chest. 2004;126:114-21. [PMID: 15249451] 13. Gohagan JK, Marcus PM, Fagerstrom RM, Pinsky PF, Kramer BS, Prorok PC, et al; Lung Screening Study Research Group. Final results of the Lung Screening Study, a randomized feasibility study of spiral CT versus chest X-ray screening for lung cancer. Lung Cancer. 2005;47:9-15. [PMID: 15603850] 14. Baker SG, Erwin D, Kramer BS. Estimating the cumulative risk of false positive cancer screenings. BMC Med Res Methodol. 2003;3:11. [PMID: 12841854] 15. Oken MM, Marcus PM, Hu P, Beck TM, Hocking W, Kvale PA, et al; PLCO Project Team. Baseline chest radiograph for lung cancer detection in the randomized Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial. J Natl Cancer Inst. 2005;97:1832-9. [PMID: 16368945] 16. Yau G, Lock M, Rodrigues G. Systematic review of baseline low-dose CT lung cancer screening. Lung Cancer. 2007;58:161-70. [PMID: 17723250] 17. Elmore JG, Barton MB, Moceri VM, Polk S, Arena PJ, Fletcher SW. Ten-year risk of false positive screening mammograms and clinical breast examinations. N Engl J Med. 1998;338:1089-96. [PMID: 9545356] 18. Hofvind S, Thoresen S, Tretli S. The cumulative risk of a false-positive recall in the Norwegian Breast Cancer Screening Program. Cancer. 2004;101:1501-7.
20 April 2010 Annals of Internal Medicine Volume 152 Number 8 511

Screening Study and the centers and health professionals involved in this trial. They also thank Dr. Phil Prorok, Dr. Christine Berg, the National Lung Screening Trial Executive Committee, and Dr. Thomas Louis for their helpful input during study development.
Grant Support: The Lung Screening Study was funded by the National

Cancer Institute; no monies or grants were awarded to any of the authors to perform this study.
Potential Conflicts of Interest: Dr. Croswell: Financial relationships

involving spouse: Husband owns shares in Johnson & Johnson. Dr. Kramer: Money received: Journal of the National Cancer Institute. Disclosures can be viewed at www.acponline.org/authors/icmje /ConictOfInterestForms.do?msNum M09-1802.
Reproducible Research Statement: Study protocol: Available at http: //prevention.cancer.gov/programs-resources/groups/ed/programs/lss/manual. Statistical code: Written in Mathematica 7.0 (Wolfram Research, Champaign, Illinois) and available by written request to Dr. Baker (e-mail, bakers@mail.nih.gov). Data set: Available by written request (to include the readers objectives and the specic data required) to Dr. Baker (email, bakers@mail.nih.gov).
www.annals.org

Article

False-Positive Test Results in Lung Cancer Screening

nissen E, et al. Baseline and second round results from the population-based Dutch-Belgian randomized lung cancer screening trial (NELSON) [Abstract]. J Clin Oncol. 2008;26:1508. 34. Wilson DO, Weissfeld JL, Fuhrman CR, Fisher SN, Balogh P, Landreneau RJ, et al. The Pittsburgh Lung Screening Study (PLuSS): outcomes within 3 years of a rst computed tomography scan. Am J Respir Crit Care Med. 2008; 178:956-61. [PMID: 18635890] 35. Pastorino U, Bellomi M, Landoni C, De Fiori E, Arnaldi P, Picchio M, et al. Early lung-cancer detection with spiral CT and positron emission tomography in heavy smokers: 2-year results. Lancet. 2003;362:593-7. [PMID: 12944057] 36. Tokarskaya ZB, Scott BR, Zhuntova GV, Okladnikova ND, Belyaeva ZD, Khokhryakov VF, et al. Interaction of radiation and smoking in lung cancer induction among workers at the Mayak nuclear enterprise. Health Phys. 2002; 83:833-46. [PMID: 12467291] 37. Pierce DA, Sharp GB, Mabuchi K. Joint effects of radiation and smoking on lung cancer risk among atomic bomb survivors. Radiat Res. 2003;159:511-20. [PMID: 12643796] 38. Henschke CI, Yankelevitz DF, Libby DM, Pasmantier MW, Smith JP, Miettinen OS; International Early Lung Cancer Action Program Investigators. Survival of patients with stage I lung cancer detected on CT screening. N Engl J Med. 2006;355:1763-71. [PMID: 17065637] 39. Brenner DJ. Radiation risks potentially associated with low-dose CT screening of adult smokers for lung cancer. Radiology. 2004;231:440-5. [PMID: 15128988] 40. Brenner DJ, Hall EJ. Computed tomographyan increasing source of radiation exposure. N Engl J Med. 2007;357:2277-84. [PMID: 18046031] 41. van den Bergh KA, Essink-Bot ML, Bunge EM, Scholten ET, Prokop M, van Iersel CA, et al. Impact of computed tomography screening for lung cancer on participants in a randomized controlled trial (NELSON trial). Cancer. 2008; 113:396-404. [PMID: 18484588] 42. Lafata JE, Simpkins J, Lamerato L, Poisson L, Divine G, Johnson CC. The economic impact of false-positive cancer screens. Cancer Epidemiol Biomarkers Prev. 2004;13:2126-32. [PMID: 15598770] 43. Pinsky PF, Miller A, Kramer BS, Church T, Reding D, Prorok P, et al. Evidence of a healthy volunteer effect in the prostate, lung, colorectal, and ovarian cancer screening trial. Am J Epidemiol. 2007;165:874-81. [PMID: 17244633] 44. Erasmus JJ, Gladish GW, Broemeling L, Sabloff BS, Truong MT, Herbst RS, et al. Interobserver and intraobserver variability in measurement of nonsmall-cell carcinoma lung lesions: implications for assessment of tumor response. J Clin Oncol. 2003;21:2574-82. [PMID: 12829678] 45. Bogot NR, Kazerooni EA, Kelly AM, Quint LE, Desjardins B, Nan B. Interobserver and intraobserver variability in the assessment of pulmonary nodule size on CT using lm and computer display methods. Acad Radiol. 2005;12:94856. [PMID: 16087090] 46. Gierada DS, Pilgram TK, Ford M, Fagerstrom RM, Church TR, Nath H, et al. Lung cancer: interobserver agreement on interpretation of pulmonary ndings at low-dose CT screening. Radiology. 2008;246:265-72. [PMID: 18024436]

[PMID: 15378474] 19. Christiansen CL, Wang F, Barton MB, Kreuter W, Elmore JG, Gelfand AE, et al. Predicting the cumulative risk of false-positive mammograms. J Natl Cancer Inst. 2000;92:1657-66. [PMID: 11036111] 20. Njor SH, Olsen AH, Schwartz W, Vejborg I, Lynge E. Predicting the risk of a false-positive test for women following a mammography screening programme. J Med Screen. 2007;14:94-7. [PMID: 17626709] 21. Croswell JM, Kramer BS, Kreimer AR, Prorok PC, Xu JL, Baker SG, et al. Cumulative incidence of false-positive results in repeated, multimodal cancer screening. Ann Fam Med. 2009;7:212-22. [PMID: 19433838] 22. Swensen SJ, Jett JR, Hartman TE, Midthun DE, Mandrekar SJ, Hillman SL, et al. CT screening for lung cancer: ve-year prospective experience. Radiology. 2005;235:259-65. [PMID: 15695622] 23. Jett JR, Midthun DE. Commentary: CT screening for lung cancer caveat emptor. Oncologist. 2008;13:439-44. [PMID: 18448559] 24. Garg K, Keith RL, Byers T, Kelly K, Kerzner AL, Lynch DA, et al. Randomized controlled trial with low-dose spiral CT for lung cancer screening: feasibility study and preliminary results. Radiology. 2002;225:506-10. [PMID: 12409588] 25. Diederich S, Thomas M, Semik M, Lenzen H, Roos N, Weber A, et al. Screening for early lung cancer with low-dose spiral computed tomography: results of annual follow-up examinations in asymptomatic smokers. Eur Radiol. 2004;14:691-702. [PMID: 14727146] 26. Sone S, Li F, Yang ZG, Honda T, Maruyama Y, Takashima S, et al. Results of three-year mass screening programme for lung cancer using mobile low-dose spiral computed tomography scanner. Br J Cancer. 2001;84:25-32. [PMID: 11139308] 27. Swensen SJ, Jett JR, Hartman TE, Midthun DE, Sloan JA, Sykes AM, et al. Lung cancer screening with CT: Mayo Clinic experience. Radiology. 2003;226: 756-61. [PMID: 12601181] 28. MacRedmond R, McVey G, Lee M, Costello RW, Kenny D, Foley C, et al. Screening for lung cancer using low dose CT scanning: results of 2 year follow up. Thorax. 2006;61:54-6. [PMID: 16396954] 29. Chong S, Lee KS, Chung MJ, Kim TS, Kim H, Kwon OJ, et al. Lung cancer screening with low-dose helical CT in Korea: experiences at the Samsung Medical Center. J Korean Med Sci. 2005;20:402-8. [PMID: 15953860] 30. McWilliams A, Mayo J. Computed tomography-detected noncalcied pulmonary nodules: a review of evidence for signicance and management. Proc Am Thorac Soc. 2008;5:900-4. [PMID: 19056713] 31. MacMahon H, Austin JH, Gamsu G, Herold CJ, Jett JR, Naidich DP, et al; Fleischner Society. Guidelines for management of small pulmonary nodules detected on CT scans: a statement from the Fleischner Society [Editorial]. Radiology. 2005;237:395-400. [PMID: 16244247] 32. Blanchon T, Brechot JM, Grenier PA, Ferretti GR, Lemarie E, Milleron B, et al; Depiscan Group. Baseline results of the Depiscan study: a French random ized pilot trial of lung cancer screening comparing low dose CT scan (LDCT) and chest X-ray (CXR). Lung Cancer. 2007;58:50-8. [PMID: 17624475] 33. van Klaveren RV, Oudkerk M, Mali W, Scholten E, Nackaerts K, Thun-

512 20 April 2010 Annals of Internal Medicine Volume 152 Number 8

www.annals.org

Annals of Internal Medicine

Current Author Addresses: Dr. Croswell: Ofce of Medical Applica-

n1b

number at risk for rst FP at T1

x20

x21

x23

tions of Research, National Institutes of Health, 6100 Executive Boulevard, Suite 2B-03, Bethesda, MD 20892. Dr. Baker: Biometry Research Group, Division of Cancer Prevention, National Cancer Institute, EPN 3131, 6130 Executive Boulevard, Bethesda, MD 20892-7354. Dr. Marcus: Division of Cancer Prevention, National Cancer Institute, 6130 Executive Boulevard, Suite 3131, Bethesda, MD 20892-7354. Mr. Clapp: Information Management Services, 6110 Executive Boulevard, Suite 310, Rockville, MD 20852. Dr. Kramer: Ofce of Disease Prevention, National Institutes of Health, 6100 Executive Boulevard, Suite 2B-03, Bethesda, MD 20892.
Author Contributions: Conception and design: J.M. Croswell, B.S. Kramer. Analysis and interpretation of the data: J.M. Croswell, S.G. Baker, B.S. Kramer, J.D. Clapp. Drafting of the article: J.M. Croswell, B.S. Kramer. Critical revision of the article for important intellectual content: J.M. Croswell, S.G. Baker, P.M. Marcus, B.S. Kramer. Final approval of the article: J.M. Croswell, S.G. Baker, P.M. Marcus, B.S. Kramer. Statistical expertise: S.G. Baker. Collection and assembly of data: J.M. Croswell, P.M. Marcus, J.D. Clapp.

We can then write: y1 y1a y1b number with a rst FP n1 n1a n1b number at risk for a rst FP For a second screening at T1 after no FP at T0: y2 number with FP at T1 x01 x03 (1 f ) n2 number at risk for FP at T1 x00 x01 x03 From the above quantities, we computed: h1 hazard for rst FP h1 y1/n1 h2 hazard for FP after FP 0 h2 y2/n2 The estimated probability of no FPs on 2 screenings is (1 h1) (1 h2). Therefore the estimated probability of at least one FP on 2 screenings is P 1 S. From the binomial distribution, var(hj) hj (1 hj)/nj . By using the delta method, var(P) var(S) S2 var(log(S)), in which var(log(S)) var(h1)/(1 h1)2 var(h2)/(1 h2)2, x1/[n1 (n1 x1)] x2/[n2 (n2 x2)], which is the Greenwood formula.

APPENDIX
This appendix describes the statistical calculations in more detail. Summary of the Data The possible outcomes were coded as 0, indicating no falsepositive result (no FP); 1, indicating a false-positive result (FP); 2, indicating that the test was missing (missing); or 3, indicating a positive test result with insufcient follow-up (PWIF) (Appendix Table 1). The count data are summarized in Appendix Tables 1 to 9. Formulas for Computing Estimate False-Positive Rates Let xij denote a number with outcome i at round T0 and outcome j at round T1, for a particular screening method, either chest radiography or low-dose CT. These are the numbers in the cells of Appendix Table 1. Let xiP xi0 xi1 xi2 xi3 denote the row sums. We let f denote the probability that a person who received a positive result on screening with insufcient follow-up had a true-positive result (that is, has disease). Under the base case, we set f 0. Under the sensitivity analysis, we set f 0.02 for chest radiography and f 0.07 for low-dose CT. The estimate of f, which comes from other studies, is the probability of disease given a positive screening test result, which is called the positive predictive value. One minus the positive predictive value is the probability of no disease given a positive screening test result. Our formulas are as follows. For a rst screening at T0: y1a number with a rst FP at T0 x1P x32 (1 f ) n1a number at risk for rst FP at T0 x0P x1P x32 For a rst screening FP at T1 because of missing at T0: y1b number with a rst FP at T1 x21 x23 (1 f )
W-176 20 April 2010 Annals of Internal Medicine Volume 152 Number 8

Computed False-Positive Rates For low-dose CT, using the formulas and data cells described, h1 335/1610 0.21. This is the estimated probability of a false-positive result on the rst screening received. h2 171/114 0.15; this is the estimated probability of a falsepositive result on the second screening received, given that the rst result was negative. The cumulative probability of a falsepositive result is therefore h1 (0.21) for the rst test received and 1 (1 h1)(1 h2) 1 (1 0.21)(1 0.15) 0.33 for the second test received. For chest radiography, h1 147/1580 0.09. This is the estimated probability of a false-positive result on the rst screening received. h2 69/1183 0.06; this is the estimated probability of a false-positive result on the second screening received, given that the rst result was negative. The cumulative probability of a false-positive result is therefore h1 (0.09) for the rst test received and 1 (1 h1)(1 h2) 1 (1 0.09)(1 0.06) 0.15 for the second test received. Logistic Regression Model For each screening method, we t the following logistic regression models: logit[pr(FP on rst screening received)] age I(age) smoker I(smoker) i Ci I(center i) py I(pack-years), logit[pr(FP on second screening received after no FP on rst screening received)] age I(age) smoker I(smoker) i Ci I(center i) py I(pack-years), in which I(smoker) 1 if current smoker, and 0 if former smoker,
www.annals.org

I(age) 1 age is between 65 and 74, and 0 if age is between 55 and 64, I(center i) 1 if from center I, and 0 otherwise, I(pack years) 1 if pack-years is 60 or more, and 0 if less than 60. Recoding Data for Fitting Logistic Regression Models Before tting the logistic regressions we recoded the original data as follows. We assumed the base case in which a PWIF is an FP. We denote the original pair of data by {outcome at T0, outcome at T1}, in which 0 indicated no false-positive result (no FP), 1 indicated a false-positive result (FP), 2 indicated a missing test, and 3 indicated a positive result with insufcient follow-up (PWIF). We recoded the original pair as a new pair, {outcome for rst screening received and outcome for second screening received after no FP on rst screening}, in which 0 indicated no false-positive result (no FP), 1 indicated a false-positive result (FP), and 2 indicated information outside of the risk set. The rules for recoding were as follows. If the outcome at T0 in the original pair was 0 (no FP), then the outcome of the new pair was identical to that for the original pair, except that PWIF was coded as FP under the base case. If original pair {0, 0}, then new pair {0, 0} If original pair {0, 1}, then new pair {0, 1} If original pair {0, 2}, then new pair {0, 2} If original pair {0, 3}, then new pair {0, 1}

If the outcome at T0 in the original pair was 1(FP), then the outcome of the new pair was coded as {1, 2} because the outcome at the second screening is not in the risk set. If original pair {1, 0}, then new pair {1, 2} If original pair {1, 1}, then new pair {1, 2} If original pair {1, 2}, then new pair {1, 2} If original pair {1, 3}, then new pair {1, 2} If the outcome at T0 in the original pair was 2 (missing), then the outcome of the new pair was coded as {outcome at T1, 2} because the screening at T1 is now the rst screening, and the second screening is not in the risk set. Also PWIF was coded as FP under the base case. If original pair {2, 0}, then new pair {0, 2} If original pair {2, 1}, then new pair {1, 2} If original pair {2, 2}, then new pair {2, 2} If original pair {2, 3}, then new pair {1, 2} If the outcome at T0 in the original pair was 3 (PWIF), then the outcome of the new pair was coded as {1, 2} because PWIF is coded as 1(FP) under the base case, and the second screening is not in the risk set. If original pair {3, 0}, then new pair {1, 2} If original pair {3, 1}, then new pair {1, 2} If original pair {3, 2}, then new pair {1, 2} If original pair {3, 3}, then new pair {1, 2}

Appendix Table 1. Count Data for Outcomes of Screening

Outcome of Test at T0 0 (No FP) Low-dose CT 0 (no FP) 1 (FP) 2 (Missing) 3 (PWIF) Chest radiography 0 (no FP) 1 (FP) 2 (Missing) 3 (PWIF) 943 86 17 0 Outcome of Test at T1 1 (FP) 159 157 7 0 2 (Missing) 144 62 50 6 3 (PWIF)

Appendix Table 2. Parameter Estimates (SE) for the Logistic Regression Model for the False-Positive Rate on First Screening Received
Variable Parameter Estimate (SE) Low-Dose CT (n 1160)
age

12 17 0 0

Chest Radiography (n 1580) 0.11 0.27 1.85 2.44 2.86 2.22 4.04 2.42 0.02 0.019 0.18 0.26 0.25 0.30 0.24 0.48 0.24 0.18

smoker

(aged 6574 y) (current smoker)

Center 1

1114 69 28 0

66 24 2 0

222 36 78 5

3 11 0 0

Center 2 Center 3 Center 4 Center 5 Center 6

CT computed tomography; FP false-positive result; missing missing test; no FP no false-positive result; PWIF positive test result with insufcient follow-up.
www.annals.org

py

( 60 pack-years)

0.25 0.14 0.77 2.51 2.25 0.61 1.26 1.94 0.08

0.14 0.13 0.19 0.23 0.23 0.17 0.18 0.19 0.13

CT

computed tomography.
20 April 2010 Annals of Internal Medicine Volume 152 Number 8 W-177

Appendix Table 3. Odds Ratio (95% CI) for the Effect of Age, Smoking Status, and Smoking History on the False-Positive Rate on First Screening Received, Based on Logistic Regression Model*
Variable Odds Ratio (95% CI)* Low-Dose CT (n 1160) Age 6574 y 5564 y Smoking status Current Former Smoking history 60 pack-years 3059 pack-years 1.28 (0.981.67) 1.0 1.11 (0.771.60) 1.0 Chest Radiography (n 1580)

1.15 (0.881.49) 1.0

1.31 (0.921.88) 1.0

0.92 (0.711.19) 1.0

1.02 (0.721.45) 1.0

CT computed tomography. * Estimates are adjusted for center.

Appendix Table 4. Estimated False-Positive Rate on First Screening Received for Low-Dose Computed Tomography, Based on Logistic Regression Model*
Center 1 2 3 4 5 6 0, 0, 0 0.32 0.08 0.10 0.10 0.22 0.13 0, 0, P 0.30 0.07 0.09 0.09 0.21 0.12 0, S, 0 0.35 0.09 0.11 0.11 0.25 0.14 0, S, P 0.33 0.08 0.10 0.10 0.23 0.13 A, 0, 0 0.37 0.09 0.12 0.12 0.27 0.16 A, 0, P 0.35 0.09 0.11 0.11 0.25 0.15 A, S, 0 0.40 0.11 0.13 0.13 0.30 0.17 A, S, P 0.39 0.10 0.12 0.12 0.28 0.16

* A means I(age) examination.

1; S means I(smoker)

1; P means I(pack-years of smoking)

1. Analysis based on 1160 participants who received at least 1 computed tomography

www.annals.org

W-178 20 April 2010 Annals of Internal Medicine Volume 152 Number 8

Appendix Table 5. Estimated False-Positive Rate on First Screening Received for Chest Radiography, Based on Logistic Regression
Model*
Center 1 2 3 4 5 6 0, 0, 0 0.14 0.08 0.05 0.10 0.02 0.08 0, 0, P 0.14 0.08 0.06 0.10 0.02 0.08 0, S, 0 0.17 0.10 0.07 0.13 0.02 0.10 0, S, P 0.17 0.10 0.07 0.13 0.02 0.11 A, 0, 0 0.15 0.09 0.06 0.11 0.02 0.09 A, 0, P 0.15 0.09 0.06 0.11 0.02 0.09 A, S, 0 0.19 0.11 0.08 0.14 0.02 0.11 A, S, P 0.19 0.12 0.08 0.14 0.03 0.12

* A means I(age) examination.

1; S means I(smoker)

1; P means I(pack-years of smoking)

1. Analysis based on 1580 participants who received at least 1 chest radiography

Appendix Table 6. Parameter Estimates (SE) for the Logistic Regression Model for the False-Positive Rate on Second Screening Received*
Variable Parameter Estimate (SE) Low-Dose CT (n 1114)
age

Appendix Table 7. Odds Ratio for the Effect of Age and Smoking Status on False-Positive Rate on Second Screening Received*
Variable Odds Ratio (95% CI) Low-Dose CT (n 1114) Age 6574 y 5564 y Smoking status Current Former Smoking history 60 pack-years 3059 pack-years 1.20 (0.831.74) 1.0 Chest Radiography (n 1183) 2.03 (1.233.36) 1.0

Chest Radiography (n 1183) 0.71 0.17 2.99 3.69 3.29 2.05 3.34 3.30 0.11 0.26 0.26 0.48 0.45 0.42 0.29 0.41 0.38 0.26

smoker

(aged 6574 y) (current smoker)

Center 1 Center 2 Center 3 Center 4 Center 5 Center 6 py

( 60 pack-years)

0.18 0.07 0.47 2.75 3.02 1.38 1.59 2.76 0.43

0.019 0.18 0.25 0.28 0.33 0.25 0.24 0.29 0.18

1.07 (0.751.52) 1.0

0.85 (0.511.40) 1.0

1.53 (1.082.18) 1.0

1.12 (0.681.85) 1.0

CT computed tomography. * When baseline examination was negative.

www.annals.org

CT computed tomography. * When baseline examination was negative. Estimates adjusted for center.
20 April 2010 Annals of Internal Medicine Volume 152 Number 8 W-179

Appendix Table 8. Marginal Counts, Crude Rates (FP Probabilities), and Crude ORs for Each Variable Included in Appendix Table 3
Variable, by FP Result FP Age 6574 y 5564 y Smoking Current Former Smoking history 60 pack-years 3059 pack-years 393 882 Low-Dose CT (n 1610) 0 FP 123* 212 1 Total 516 1094 FP 452 981 Chest Radiography (n 1580) 0 FP 50 97 1 Total 502 1078

724 551

205 130**

929 681

804 629

93 54

897 683

535 740

138 197

673 937

623 810

64 83

687 893

CT computed tomography; FP false-positive; OR odds ratio. * Probability of FP if aged 6574 y 0.24; crude OR 1.30. Probability of FP if aged 6574 y 0.10; crude OR 1.12. Probability of FP if aged 55 64 y 0.19. Probability of FP if aged 55 64 y 0.09. Probability of FP if current smoker 0.22; crude OR 1.20. Probability of FP if current smoker 0.10; crude OR 1.35. ** Probability of FP if former smoker 0.19. Probability of FP if former smoker 0.08. Probability of FP if 60 pack-years 0.21; crude OR 0.97. Probability of FP if 60 pack-years 0.09; crude OR 1.00. Probability of FP if 30 59 pack-years 0.21. Probability of FP if 30 59 pack-years 0.09.

Appendix Table 9. Marginal Counts, Crude Rates (FP Probabilities), and Crude ORs for Each Variable Included in Appendix Table 7
Variable, by FP Result FP Age 6574 y 5564 y Smoking Current Former Smoking history 60 pack-years 3059 pack-years 271 672 Low-Dose CT (n 1114) 0 FP 61* 110 1 Total 332 782 FP 328 786 Chest Radiography (n 1580) 0 FP 31 38 1 Total 359 824

522 421

98 73**

620 494

619 495

34 35

653 530

372 571

83 88

455 659

482 632

32 37

514 669

CT computed tomography; FP false-positive; OR odds ratio. * Probability of FP if aged 6574 y 0.18; crude OR 1.38. Probability of FP if aged 6574 y 0.09; crude OR 1.95. Probability of FP if aged 55 64 y 0.14. Probability of FP if aged 55 64 y 0.05. Probability of FP if current smoker 0.16; crude OR 1.08. Probability of FP if current smoker 0.05; crude OR 0.78. ** Probability of FP if former smoker 0.15. Probability of FP if former smoker 0.07. Probability of FP if 60 pack-years 0.18; crude OR 1.45. Probability of FP if 60 pack-years 0.06; crude OR 1.13. Probability of FP if 30 59 pack-years 0.13. Probability of FP if 30 59 pack-years 0.06.
W-180 20 April 2010 Annals of Internal Medicine Volume 152 Number 8 www.annals.org