This article is a US Government work and, as such, is in the public domain in the United States of America.
Low-radiation dose spiral computed tomography (LDCT) currently is being evaluated as a screening modality for lung carcinoma in a randomized trial. Although several diagnostic algorithms for the workup of positive LDCT screens have been proposed, to the authors' knowledge there is no widely accepted standard to date and there are few nationwide data concerning how such diagnostic workups are actually being performed outside a research protocol setting.
The Lung Screening Study (LSS) was a multicenter feasibility trial that randomized 1660 subjects to undergo LDCT and an equivalent number to undergo chest X-ray. Subjects with positive screens were referred to their own health care providers for diagnostic follow-up; LSS did not specify a diagnostic algorithm. LSS collected and abstracted medical records regarding procedures employed in the diagnostic workup of positive screens.
Of the 522 subjects with a positive LDCT screen at baseline or at Year One, 12% underwent biopsy. Biopsy was less likely to be performed in subjects with 4–9-mm nodules (5%) than in subjects with nodules measuring 10+ mm (25%) or in subjects with no nodules but other suspicious findings (15%). Among 63% of the subjects who underwent chest CT on follow-up, the median time between screening and first follow-up chest CT was 82 days. Only a minority of subjects received diagnostic workups that were consistent with published algorithms.
The data from the current study represent the experience of subjects followed by their health care providers in five different U.S. metropolitan areas and one rural area. As such, they provide some indication of practices in the U.S. with regard to the diagnostic workup of patients with positive spiral CT screens. Cancer 2005. Published 2004 by the American Cancer Society.
Lung carcinoma is the leading cause of cancer-related death in the U.S.1 Currently, to our knowledge no early detection modality has been proven to reduce lung carcinoma mortality, and all can trigger potentially harmful diagnostic testing and therapy. Low-radiation dose spiral computed tomography (LDCT), an advance in CT technology that was introduced during the early 1990s, has been proposed recently as a new modality for lung carcinoma screening.2 LDCT has been shown in direct comparison with chest X-ray to identify more small lesions in the lungs of smokers and former smokers.3
Because the benefit of screening for lung carcinoma with LDCT remains unproven, there are many unresolved issues associated with its use. The primary unresolved issue is whether such screening can reduce the mortality from lung carcinoma. A secondary unresolved issue is the optimal management and diagnostic follow-up of individuals with LDCT screens that are judged suspicious for lung carcinoma. Such follow-up must attempt to minimize the costs and medical harms associated with diagnostic procedures, while still identifying those small tumors for which early detection and treatment may be beneficial. A number of researchers have proposed algorithms for the diagnostic follow-up of small nodules or other suspicious findings observed on CT4–8; however, to our knowledge it is not known to what extent any of these proposed algorithms are being utilized in clinical practice. The Lung Screening Study (LSS) provides a preliminary opportunity to examine diagnostic follow-up in a clinical setting. The LSS was a pilot multicenter study designed to assess the feasibility of performing a definitive, large-scale randomized trial of LDCT versus chest X-ray screening for lung carcinoma. The design of the LSS was such that subjects with positive screens were referred to their own personal health care providers for appropriate follow-up; the LSS did not specify a diagnostic algorithm. Therefore, in this respect, the findings from LSS with regard to patterns of diagnostic follow-up provide some indication of what is currently occurring in clinical practice in the U.S.
In this article, we describe in detail the patterns of diagnostic follow-up occurring after a positive LDCT screen in LSS subjects. We related these patterns to the specific findings at the time of screening, including the type of abnormality (noncalcified nodule or other) and the size of any noncalcified nodules, as well as to the demographic factors of the study subjects.
MATERIALS AND METHODS
The Lung Screening Study (LSS) was a special project conducted within the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial.9 The following six PLCO screening centers participated: Georgetown University in Washington, DC; Henry Ford Health System in Detroit, MI; Washington University in St. Louis, MO; the University of Minnesota in Minneapolis, MN; the Marshfield Clinic in Marshfield, WI; and the University of Alabama in Birmingham, AL. The design and the results of the baseline screening round in LSS are explained in detail elsewhere.9 Briefly, more than 3300 subjects were randomized between September 2000 and January 2001 to undergo either LDCT or a chest X-ray. Current cigarette smokers or former smokers who quit within the last 10 years were eligible for the study if they were ages 55–74 years and had a 30+ pack-year smoking history. Exclusion criteria included a previous history of lung carcinoma, previous removal of any portion of the lungs, current treatment for cancer (except basal cell skin cancer), and a spiral CT examination of chest within the past 2 years. Although the LSS initially was designed as a single-screen study, it subsequently was extended to include an additional screen 1 year after the baseline screen. Subjects were eligible for the second screen provided they were not diagnosed with lung carcinoma after the first screen.
LDCT exams were obtained at a range of 120–140 kilovolt peak, 60 milliamperes (mA), and a 1-second scan time using 5-mm collimation and contiguous reconstructions. Each CT was interpreted by an American Board of Radiology-certified radiologist at the screening center. Radiologists recorded the type (e.g., noncalcified nodule), size, and location of each abnormality. The following findings resulted in a designation of a positive screen: a nodule or mass measuring > 3 mm, spiculated nodules measuring ≤ 3 mm, focal parenchymal opacification, and endobronchial lesions. Subjects with positive screens were referred to their personal health care provider for diagnostic follow-up. The LSS did not specify a diagnostic algorithm for the follow-up of positive screens; however, some of the individual PLCO screening centers provided their own recommended guidelines if requested.
LSS attempted to obtain all medical records pertaining to the diagnostic follow-up of positive screens. Relevant diagnostic procedures occurring within 1 year of a positive screen were abstracted in a standardized manner; the results of these procedures were not captured. CT scans were categorized as follows: CT scan of the chest, CT scan of the chest and upper abdomen, CT scan of the abdomen and pelvis, CT scan of the brain, CT scan of the liver, other CT scans, and thin-section CT. No information regarding the radiation dose of the CT scan was collected.
Multiple logistic regression was used to examine the effect of various factors on the probability of biopsy after a first positive LDCT screen. These factors included patient gender, patient age, maximum nodule size, location of the abnormality, and whether the first positive screen was the subject's first or second LSS screen.
In addition, we also examined the effect of the radiologist reading the screening LDCT. There was considerable variability among radiologists in the current study with regard to screen positivity rates. We hypothesize that there is an inverse correlation between the positivity rate of a radiologist and the rate at which that radiologist's positive screen subjects undergo biopsy because the percentage of a radiologist's positive screens that are in some sense “borderline” may be higher for radiologists with high positivity rates and these subjects with “borderline” positive screens may be less likely to undergo biopsy. A statistical model was used to examine this relation, with the null hypothesis being that there is no correlation between the two rates.10
The model for radiologist effect postulates that each radiologist has an underlying log odds X of calling a screen positive and an underlying log odds Y of his or her positive subjects receiving a certain level of follow-up (e.g., invasive procedures). The model assumes that there is a distribution of (X,Y) among radiologists; specifically, (X,Y) is assumed to have a bivariate normal distribution, with a correlation parameter of rho. Under the null hypothesis, rho = 0. The model was fit by maximum likelihood. Given (X,Y), the likelihood of the observed data for a given radiologist is provided by the product of two binomials, one for “k” positive screens of “n” total screens and the other for “j” subjects with a certain level of follow-up of “k” positive screen subjects. The overall likelihood for each radiologist then is computed by integrating the above conditional probability over the distribution of (X,Y).
All reported P values are two-sided.
Of 1660 eligible subjects randomized to undergo LDCT, 1558 received the baseline (Year 0) screen and 1398 received the Year 1 screen. This analysis reports on the 522 subjects who had at least 1 positive LDCT screen and known diagnostic follow-up (i.e., either a documented record of diagnostic procedures or a physician or self-report that no diagnostic procedures were performed) after the positive screen. Note that an additional 12 subjects had a positive LDCT screen but unknown diagnostic follow-up. Table 1 displays the age, gender, smoking history, and screening examination findings (for the first instance of a positive screen) for the above-mentioned 522 subjects. The mean age, percentage of male subjects, median number of pack-years, and percentage of subjects who were current smokers were found to be similar for this population as for the entire trial population. Approximately 60% of subjects had a positive screen at baseline and the remainder had their first positive screen at Year 1. The most common finding on the (first) positive screen was nodules measuring 4–9 mm (56% of subjects). Nodule sizes of 10–19 mm and 20+ mm were reported in 21% and 9% of subjects, respectively. A total of 37 of the 522 subjects (7%) were diagnosed with lung carcinoma within 1 year of their first positive screen. These subjects did not come predominantly from 1 or 2 screening centers; each center contributed between 9–28% of the positive subjects.
Table 1. Demographics and Screening Findings among Subjects with Positive Screens and Known Follow-Up Status
n = 522
Twenty-eight of these 46 subjects had focal parenchymal opacification.
Nodule was spiculated for 19 of these 29 subjects.
Table 2 displays the highest level of diagnostic procedures undergone by subjects after their first positive screen. Note that all procedures discussed in this section occurred within 1 year of the first positive LSS screen and before the next screen (if any). Procedures were ordered hierarchically as shown in the table, with biopsy being the highest level followed by invasive procedures without biopsy and CT scan of chest, and with clinical examination and no procedure performed being the lowest levels. Overall, 12% of the subjects underwent biopsy, and an additional 55% of subjects underwent a chest CT scan. Approximately 3% of subjects received no follow-up, 4% underwent a clinical examination only, and 12% had comparison with only a prior X-ray or prior CT scan (with or without clinical examination).
Table 2. Highest Level of Follow-Up Procedure after First Positive Screen
Biopsy rates ranged from 5% for subjects with nodules measuring 4–9 mm to 25% for subjects with nodules measuring 10+ mm (Table 2). Multiple logistic regression analysis demonstrated that having a maximum nodule size of 10+ mm (odds ratio [OR] of 6.9; 95% confidence interval [95% CI], 3.6–13.4) and having no nodules but other abnormalities (OR of 3.6; 95% CI, 1.2–10.9) were associated significantly with the biopsy rate compared with having a maximum nodule size of 4–9 mm. In addition, subjects who were negative on their first LSS screen were found to have a decreased rate of biopsy (OR of 0.47; 95% CI, 0.24–0.90) after their positive screen compared with subjects whose first LSS screen was positive. Age, gender, and location of the nodule (upper lobe or middle lobe vs. lower lobe) were not found to be independently associated with the biopsy rate.
The average number of invasive procedures, among those subjects undergoing any, was 1.67 (bronchoscopy with biopsy or thoracotomy with resection were considered as a single invasive procedure). Of the 66 subjects undergoing invasive procedures, 46 (70%) had undergone some prior diagnostic follow-up chest imaging procedure and 34 (52%) had undergone a prior follow-up chest CT scan. A total of 89 biopsies were performed on 62 subjects; 36 (58%) of these subjects (accounting for 55 of the biopsies) were diagnosed with lung carcinoma (an additional case of lung carcinoma was diagnosed through sputum cytology). Of the biopsies, 33% were resections or surgical open biopsies, 24% were fine-needle aspiration biopsies, 11% were endobronchial or transbronchial biopsies, 24% were cytologic brushings, 6% were lymph node biopsies, and 3% were biopsies performed at other sites (nonlung or lymph node biopsies).
A total of 330 subjects (63%) underwent at least 1 follow-up chest CT scan. The rates of chest CT scan were similar among subjects with nodules measuring ≥ 10 mm (61%), subjects with nodules measuring 4–9 mm (61%), and subjects with no nodules but other suspicious findings (63%); approximately 90% of the 29 subjects with nodules measuring 1–3 mm underwent CT scans of the chest. Multiple chest CT scans were received by 122 of the above-mentioned subjects (37%); 89 subjects received 2 CT scans and 33 received ≥ 3 CT scans. The timing of the chest CT scans is shown in Table 3. The median interval from the time of the screen to the first diagnostic chest CT scan was 82 days whereas the median interval between the first and second follow-up CT scans was 119 days. The median intervals between CT scans did not appear to vary significantly by nodule size.
Table 3. Intervals between Chest CTs after First Positive Screen
Although 122 subjects had multiple follow-up chest computed tomography(CT)scans, 22 of these had all their chest CTs taken the same day and therefore were not included in the table.
Fifty-seven subjects (17% of those receiving a chest CT scan) received a total of 68 thin-section CT scans, with 44 of these 68 scans occurring concurrently with a full chest CT scan. Subjects with nodules measuring 4–9 mm underwent thin-section CT scans more frequently than did subjects with nodules measuring 10+ mm (15% vs. 5%; P = 0.003).
Pulmonary function tests were performed significantly less often in subjects with nodules measuring 4–9 mm compared with subjects with nodules measuring 10+mm or subjects with no nodules but other suspicious findings (16% vs. 34% vs. 33%, respectively). Comparison with prior images was significantly more common in subjects with nodules measuring 4–9 mm compared with subjects with nodules measuring 10+ mm or subjects with other suspicious findings (61% vs. 41% vs. 37%, respectively).
Of the 485 subjects not diagnosed with lung carcinoma, 217 (45%) were diagnosed with another (nonlung carcinoma) condition as part of the diagnostic follow-up. The most common diagnoses were as follows: other diseases of the lung, not classified elsewhere (n = 114); chronic airway obstruction, not classified elsewhere (n = 47); postinflammatory pulmonary fibrosis (n = 31); and emphysema (n = 12). One subject was diagnosed with renal carcinoma.
The 522 (initial) positive CT screens were interpreted by 32 different radiologists. Figure 1 displays the observed screen positivity rate for each radiologist and the corresponding rate of biopsy for that radiologist's positive subjects. Visually, there appeared to be an inverse correlation between the two rates. This is borne out by the statistical model for radiologist effect, which demonstrated a statistically significant negative correlation of −0.92 between the underlying (log) odds of a radiologist calling a screen positive and the underlying (log) odds of that radiologist's positive screen subjects undergoing biopsy at the time of follow-up (P = 0.002). According to the model, radiologists with average screen positivity rates of 10%, 20%, and 40% would be expected to have 26%, 14%, and 5%, respectively, of their positive subjects undergo biopsy.
In addition to examining the correlation between positivity rates and the rates of undergoing an invasive procedure at the time of follow-up, we also examined the correlation between the positivity rates of the radiologists and the rate of undergoing a follow-up chest CT scan. We found a nonsignificant correlation of 0.25 (P = 0.9) between the log odds of a radiologist calling a screen positive and the log odds of that radiologist's positive subjects undergoing a follow-up chest CT scan.
A total of 143 of the 522 subjects with positive LSS screens had positive screens at both Year 0 and Year 1 (and known diagnostic follow-up after each screen). Of these 143 subjects, only 3 subjects underwent invasive procedures after the Year 0 (baseline) screen, whereas 110 subjects (77%) underwent a chest CT scan. It is interesting to note that the rate of invasive procedures at baseline was low in this group because subjects in whom tumors were diagnosed at baseline did not undergo a second screen and noncarcinoma subjects whose nodules were biopsied at baseline in general did not have a positive screen a year later. Table 4 shows the highest level of follow-up procedure after the second positive screen based on the second screen finding. Among 97 subjects with nodules measuring 4–9 mm, 2% underwent biopsy, 29% underwent a chest CT scan, and 46% had only comparison with a prior CT scan or X-ray images. For 41 subjects with nodules measuring 10+ mm, 5% underwent had biopsy, 44% underwent chest CT scan, and 20% had only comparison with prior images. One of these 143 subjects was diagnosed with lung carcinoma.
Table 4. Highest Level of Follow-Up Procedure Based on Finding on Second Positive Screen
In the current study, we described the pattern of diagnostic follow-up after the first instance of a positive spiral CT screen in subjects enrolled in the LSS. Because the LSS did not specify a diagnostic algorithm and the subjects were followed at numerous institutions across five metropolitan areas and one rural area, the findings reported herein likely represent a broad section of clinical practice in the U.S. However, because the LSS centers were somewhat clustered geographically (mostly in the Midwest and none in the West), the findings of the current study may not be representative of U.S. practice as a whole. Approximately two-thirds of subjects received a follow-up chest CT scan. The use of chest CT was similar for subjects with nodules measuring 4–9mm as for subjects with nodules measuring 10+ mm (and was slightly higher for subjects with nodules measuring 1–3 mm). A total of 12% of subjects underwent biopsy, with 58% of biopsied subjects subsequently diagnosed with lung carcinoma. Biopsies were more prevalent in subjects with nodules measuring 10+ mm (25%) and in subjects with no nodules but other suspicious findings (15%) compared with subjects with nodules measuring 4–9 mm (5%). Somewhat paradoxically, we found that, controlling for maximum nodule size, subjects with a prior negative LSS screen had a significantly lower rate of biopsy after their first positive screen (OR of 0.47) compared with subjects who were positive on their first LSS screen. However, the rate of cancer diagnosis among those subjects who were biopsied was similar in both groups: 50% in those subjects with a prior negative screen and 60% in those subjects with positive findings on the first screen (58% overall).
To our knowledge there are no hard data published to date based on clinical trials or other studies that demonstrate the superiority of a specific diagnostic algorithm for the follow-up of noncalcified nodules identified on CT screening. Nevertheless, several such diagnostic algorithms have been suggested recently.4–6 Although they differ with regard to some details, these algorithms are structurally similar in that the prescribed follow-up differs according to the size of the largest nodule, and in that they generally recommend an initial thin-section CT followed either by lung biopsy or repeated thin-section CT at some fixed interval depending on whether any growth was observed. For example, Swensen et al. recommend thin-section CT at 3 months after screening for subjects with nodules measuring 4–7 mm, followed by either a repeat thin-section CT scan at an interval of 3 months or biopsy/surgery.5 For nodules measuring 8–20 mm, Swensen et al. recommend thin-section CT as soon as possible, followed by biopsy or further CT scans. For subjects with nodules measuring 4–10 mm, Henschke et al. recommend thin-section CT at 3 months followed by biopsy if growth is detected or additional thin-section CT performed at 6 months, 12 months, and 24 months if no growth is detected.6 For nodules measuring ≥ 11 mm, Henschke et al. recommend either immediate biopsy or thin-section CT followed by biopsy or further CT scans, as discussed earlier. Relatively few subjects in the current study received diagnostic follow-up, which is consistent with the above algorithms. For example, only 11% of subjects with nodules measuring 4–9 mm received a thin-section CT scan within 4 months of the screen and only 24% of subjects with nodules measuring 10+ mm underwent a biopsy or thin-section CT scan within 4 months of the screen.
When we examined the effect of the radiologist reading the CT screens on diagnostic follow-up, we found a strong inverse correlation between the positivity rate of the radiologist and the proportion of that radiologist's positive subjects who underwent biopsy on diagnostic follow-up, but no significant correlation was noted between the positivity rates of the radiologists and the proportion of positive subjects undergoing chest CT scan on follow-up. Therefore, although the initial tests (e.g., chest CT scan) ordered as part of the diagnostic follow-up appear to be independent of the screening radiologist, these initial tests appear to be less likely to trigger biopsy in subjects who are referred by radiologists with high positivity rates than in subjects referred by radiologists with low positivity rates. It is interesting to note that those radiologists with high positivity rates (> 20%) reported a similar percentage of their positive screen subjects as having nodules measuring 10+ mm (29.4%) as did radiologists with low positivity rates (<20%); therefore, the above inverse correlation appears to be largely independent of the reported nodule size. There are additional factors that we did not collect data on, such as the type of health care practitioner, the health care setting, and the insurance status of the subject, which may influence diagnostic follow-up independent of the screening findings. Additional studies are needed to validate and elucidate further the correlation between a radiologist's rate of positive screens and the resulting diagnostic follow-up in his or her subjects.
To our knowledge, the majority of proposed guidelines refer only to the finding of one or more noncalcified nodules. In this study, approximately 9% of positive subjects had no reported noncalcified nodules, but had other suspicious findings (e.g., focal parenchymal opacification). Their follow-up was similar to that observed overall in subjects with noncalcified nodules.
In the current study, we described diagnostic workup patterns occurring in clinical practice after a positive spiral CT screen. These data may be useful in estimating the potential burden and cost of CT screening and in evaluating whether practitioners generally are following reasonable procedures in their workup of subjects with positive CT screens.