Indeterminate Lung Nodules in Cancer Patients: Pretest Probability of Malignancy and the Role of 18F-FDG PET/CT
Abstract
OBJECTIVE. The purpose of this study was to determine likelihood of malignancy for indeterminate lung nodules identified on CT comparing two standardized models with 18F-FDG PET/CT.
MATERIALS AND METHODS. Fifty-nine cancer patients with indeterminate lung nodules (solid tumors; diameter, ≥ 5 mm) on CT had FDG PET/CT for lesion characterization. Mayo Clinic and Veterans Affairs Cooperative Study models of likelihood of malignancy were applied to solitary pulmonary nodules. High probability of malignancy was assigned a priori for multiple nodules. Low (< 5%), intermediate (5–60%), and high (> 60%) pretest malignancy probabilities were analyzed separately. Patients were reclassified with PET/CT. Histopathology or 2-year imaging follow-up established diagnosis. Outcome-based reclassification differences were defined as net reclassification improvement. A null hypothesis of asymptotic test was applied.
RESULTS. Thirty-one patients had histology-proven malignancy. PET/CT was true-positive in 24 and true-negative in 25 cases. Negative predictive value was 78% and positive predictive value was 89%. On the basis of the Mayo Clinic model (n = 31), 18 patients had low, 12 had intermediate, and one had high pretest likelihood; on the basis of the Veterans Affairs model (n = 26), 5 patients had low, 20 had intermediate, and one had high pretest likelihood. Because of multiple lung nodules, 28 patients were classified as having high malignancy risk. PET/CT showed 32 negative and 27 positive scans. Net reclassification improvements respectively were 0.95 and 1.6 for Mayo Clinic and Veterans Affairs models (both p < 0.0001). Fourteen of 31 (45.2%) and 12 of 26 (46.2%) patients with low and intermediate pretest likelihood, respectively, had positive findings on PET/CT for the Mayo Clinic and Veterans Affairs models, respectively. Of 15 patients with high pretest likelihood and negative findings on PET/CT, 13 (86.7%) did not have lung malignancy.
CONCLUSION. PET/CT improves stratification of cancer patients with indeterminate pulmonary nodules. A substantial number of patients considered at low and intermediate pretest likelihood of malignancy with histology-proven lung malignancy showed abnormal PET/CT findings.
A solitary pulmonary nodule (SPN) is a finding that cannot be ignored. The “wait and watch” strategy may be advisable under certain circumstances, but every noncalcified, non–fat-containing nodule must be regarded as potentially malignant until proven otherwise. Opinions differ as to the best approach to the initial management of an SPN. Watchful observation may be justified when the probability of malignancy is low, whereas exploratory surgery is warranted when the probability of cancer is high [1]; biopsy techniques and PET using 18F-FDG can be expected to be of clinical value for the broad spectrum of SPN cases with an intermediate probability of malignancy.
For improvement of the outcomes of the diagnosis and treatment of SPNs, two quantitative prediction models have been developed to facilitate the risk assessment of malignancy in patients with the most troublesome subset of indeterminate pulmonary nodules—namely, those that are radiologically indeterminate. The first is the Mayo Clinic model, which is based on a logistic regression analysis and considers six independent predictors of malignancy: older age, history of smoking, history of an extrathoracic cancer more than 5 years before nodule detection, larger nodule diameter, upper lobe location, and spiculated margins [2]. The second is the Veterans Affairs Cooperative Study model, or VA model, which identified four independent predictors—older age, history of smoking, larger nodule diameter, and shorter time since quitting smoking—using logistic regression analysis [3]. As shown by Schultz et al. [4], both the Mayo Clinic and VA models predict the pretest probability of malignant SPNs with accuracy that is sufficient to make them clinically useful, especially given that neither model is intended to be used as a stand-alone diagnostic test.
Although similar, both models are slightly different. In particular, the VA model included a positive smoking history, older age, and a large nodule diameter together as an independent predictor of a malignant SPN. However, in the VA model, the adjusted odds ratio for a positive smoking history was higher than in the Mayo Clinic model. Conversely, the Mayo Clinic model excluded patients with a history of lung cancer or a history of an extrathoracic cancer within 5 years because these would be strong predictors of malignant SPNs, whereas in the VA model, these factors were not independent predictors of malignancy. Finally, in the VA model, two other variables—specifically, spiculation and a remote history of extrathoracic cancer—were not included.
According to recommendation 9 of Gould et al. [5], in patients with low to moderate pretest probability of malignancy (5–60%) and an indeterminate SPN that measures at least 8–10 mm in diameter, FDG PET should be performed to characterize the nodule (grade of recommendation, 1B). Moreover, in patients with a high (> 60%) pretest probability of malignancy or with a subcentimeter nodule that measures less than 8–10 mm in diameter, FDG PET should not be performed to characterize the nodule (grade of recommendation, 2C). Recommendation 12 states that FDG PET can be useful if an SPN measures 8–10 mm in diameter or when clinical probability of malignancy is either very low (< 5%) or low (< 30–40%), as opposed to observation with serial CT studies (grade of recommendation, 2C).
Based on the aforementioned models and functional and biologic data provided by FDG PET/CT, we ask whether altering the probability of malignant nodules (pretest likelihood) or altering test parameters would change the results of such testing. This question appears relevant for investigating the role of available technologies such as FDG PET/CT for evaluating pulmonary nodules [6]. The majority of published studies have evaluated the probability of malignancy in patients with no history of cancer or with cancer within 5 years before nodule discovery [2, 3, 7]. Therefore, the aim of the current study is to determine the relationship between the likelihood of malignancy for indeterminate lung nodules identified on CT images (based on two standardized models) and FDG PET/CT findings in oncologic patients.
Materials and Methods
Study Population
From a single institutional database, between January 2009 and July 2010, 59 consecutive oncologic patients (mean age ± SD, 67 ± 10 years), with indeterminate solid lung nodules on previous CT images who underwent PET/CT within a few months were selected. Of patients selected, 14% had history of head and neck cancer, 27% of gastrointestinal cancer, 19% of breast cancer, 22% of lung cancer, 7% of melanoma, and 11% of other cancers. The mean time between cancer diagnosis and PET/CT was 38 ± 38 months (range, 1–119 months), and the mean period between last therapy (surgery or chemotherapy or other) and PET/CT was 32 ± 31 months (range, 1–119 months). At the time of diagnosis, 17 patients had cancers at stage I, seven at stage II, five at stage III, and four at stage IV, and 26 had cancers with stage unknown. At the time of PET/CT, all patients were considered to be disease free or to have no progression of disease, according to reviewed charts. All patients had received regular clinical and imaging follow-up evaluations at our center. Twenty-four (41%) patients were smokers, and no patients had a history of pneumonia or other inflammatory lung pathology at the time of CT and PET/CT. At least two serial CT studies were performed before PET/CT for each patient.
All patients were referred to PET/CT for characterization of lung nodules within 3 months of chest CT. Features contributing to the indeterminate nature of the lung lesions were unclear morphologic features, such as margins, contour, attenuation value, and growth pattern. Patients with benign calcification or without a definitive clinical diagnosis were excluded from the study. This study was approved by the local institutional review board for retrospective analysis.
Pretest Probability of Malignancy
In patients with an SPN, we calculated the likelihood of malignancy according to the Mayo Clinic and VA models in a manner similar to that used by Schultz et al. [4]. The following equation was computed for the Mayo Clinic model. The pretest probability of a malignant SPN was ex/(1 + ex) for x = –6.8272 + (0.0391 × age) + (0.7917 × smoke) + (1.3388 × cancer) + (0.1274 × diameter) + 1.0407 × spiculation) + (0.7838 × upper), where e is the base of the natural logarithm, “age” indicates the patient's age in years, “smoke” indicates smoking history in pack-years, “cancer” indicates history of an extrathoracic cancer 5 or more years before nodule identification (0, absent; 1, present in history), “diameter” indicates the measurement in millimeters of the largest nodule, “spiculation” indicates mention of nodule spiculation on any imaging test report (0, absent; 1, present on report), and “upper” indicates location of the nodule within the upper lobe of either lung (0, absent; 1, present in upper lobe) [3].
For the VA model, the respective equation was as follows. The pretest probability of malignancy was ex/(1 + ex) for x = –8.404 + (2.061 × smoke) + [(0.779 × age)/10] + (0.112 × diameter) – [(0.567 × quit)/10], where e is the base of the natural logarithm, “smoke” indicates smoking history in pack-years, “age” indicates the patient's age in years at the time of nodule identification, “diameter” indicates the measurement in millimeters of the largest nodule, and “quit” indicates the number of years since quitting smoking [4]. Because of the limitations of the Veterans Affairs Cooperative Study database, in this model, we did not include study participants who had cancer diagnosed more than 5 years before nodule detection.
Therefore, on the basis of the variables conceived in each model, we calculated the likelihood of malignancy in 31 and 26 patients in accordance with the Mayo Clinic and VA models, respectively. Patients with multiple lung nodules at CT were a priori considered to be at high likelihood of malignancy [8, 9]. According to the guidelines for management of small pulmonary nodules, the pretest probability of malignancy was classified as low when less than 5%, intermediate when 5–60% and high when greater than 60% [3, 9]. Thus, a separate analysis for patients with low (< 5%), intermediate (5–60%), and high (> 60%) pretest probabilities of malignancy was performed. Thereafter, patients were reclassified according to PET/CT results. Furthermore, for elaboration of the broad category of intermediate likelihood of malignancy, a further separate analysis considering the following categories (very low, < 5%; low, < 5–40%; intermediate, > 40–60%; and high, > 60%), as suggested by Gould et al. [5], was computed.
CT Examination and Interpretation
All patients underwent CT of the chest, abdomen, and pelvis to identify metastases. All chest CT examinations were acquired from the lung apices through the lung bases using 8–MDCT (Eclos, Hitachi Medical) or 40–MDCT (Somatom Definition AS, Siemens Healthcare) and using the following parameters: section thickness, 2.5 or 3 mm; reconstruction, 2.5 or 2 mm; gantry rotation time, 0.8 or 0.5 second; pitch, 0.7 or 0.8; tube potential, 120 kV; and current–exposure time product setting adjusted for body weight. All patients received IV contrast medium (Omnipaque 350, GE Healthcare) at a dose of 2 mL/kg injected at a flow rate of 3 mL/s.
CT images were reviewed and interpreted by two radiologists with 3 and 5 years of experience in chest CT interpretation. Both specialists were presented with the patients’ clinical history but were unaware of histologic findings. Pulmonary nodules were assessed for size, number, location (upper, middle, or lower lobe), and laterality (right or left lung). Nodule size was defined as the largest diameter measured with electronic calipers on the CT images. Moreover, the nodules were classified according to their size, their margins (smooth or irregular), attenuation (solid or partly solid vs ground-glass opacity), calcification (absent or present) and growth. For the purpose of the present study, only solid lung nodules were included.
PET/CT Examination and Interpretation
Whole-body FDG PET/CT was performed using a dedicated PET/CT scanner (Biograph 16 HT, Siemens Healthcare). The PET component is a high-resolution scanner with a spatial resolution of 4.7 mm and has no septa, thus allowing 3D-only acquisitions. The CT portion of the scanner is the Somatom Sensation 16-slice CT. Together with the PET system, the CT scanner is used both for attenuation correction of PET data and for localization of FDG uptake. All patients were instructed to fast for at least 6 h before the integrated PET/CT examination. After injection of 300 MBq (8.1 mCi) of FDG per kilogram body weight, patients rested for 60 min in a comfortable chair. Emission images from the proximal femur to the base of the skull were acquired for 3 minutes per bed position. Acquired images were reconstructed using an attenuation-weighted ordered subset expectation maximization iterative reconstruction, with two iterations and eight subsets. Fourier rebinning was used to reduce the 3D dataset to a 2D equivalent dataset, and a 4-mm full width at half-maximum gaussian filter was applied to the image after reconstruction along the axial and transaxial directions. The data were reconstructed over a 128 × 128 matrix with 2-mm pixel size and slice thickness.
Two nuclear medicine physicians with 3 and 5 years of experience of PET interpretation independently evaluated the images. At visual analysis, lung nodules were defined as malignant if they showed hypermetabolic activity of FDG higher than the metabolic activity of the other regions. Conversely, no significant FDG uptake was used for defining a negative scan. In case of discordant imaging findings between the two readers, a third nuclear medicine physician was required for the consensus. Maximum standardized uptake value (SUVmax) of the lung lesion was determined by drawing isovolumetric regions of interest on the attenuation-corrected FDG PET/CT images for each suspected lung nodule. SUVmax was calculated as the ratio of regional radioactivity concentration divided by the administered radioactivity normalized to body weight [10]. In case of multiple nodules, the region of interest was drawn around the largest lesion. A separate analysis was performed using the SUVmax threshold of 4.0 in accordance with Grgic et al. [11].
Net Reclassification Improvement
New methods have recently been proposed to evaluate and compare predictive risk models. These are based primarily on stratification into clinical categories on the basis of risk in an attempt to assess the ability of new models to more accurately reclassify individuals into higher or lower risk strata. Since its first description in 2006, much interest has been generated in reclassification, and although the approach is relatively new, there have been further methodologic developments. Researchers in the fields of breast cancer, diabetes, genetics, and clinical cardiology have published articles on these techniques [12]. The net reclassification improvement assesses risk reclassification and is the difference in proportions moving up and down risk strata among case patients versus control participants—that is, those who did or did not develop the disease during follow-up [13]. Net reclassification improvement is calculated as follows: [Pr(up / cases) – Pr(down / cases)] – [Pr(up / controls) – Pr(down / controls)], where Pr stands for probability; the net reclassification improvement is similar to simple percentage reclassification, but it distinguishes correct direction of changes (up for case patients and down for control participants). Ideally, the predicted probabilities would move higher (up a category) for case patients and lower (down a category) for control participants.
Net reclassification improvement was calculated as a sum of two separate components: individuals with events and individuals without events. An “event” refers to the presence of malignancy (i.e., a true-positive). For events, we assigned 1 for upward reclassification, –1 for downward reclassification, and 0 when people did not change their risk category. The opposite was done for nonevents. The summation of the individual scores and division by numbers of people in each group were used to obtain the value of net reclassification improvement. Assuming independence between event and non-events among individuals and following McNemar logic for significance testing in correlated proportions, we used a simple asymptotic test for the null hypothesis where the net reclassification improvement was 0 (z test) [13].
Final Outcome
Final diagnosis was established by histopathology or by CT follow-up. The diagnosis of malignant lung lesions was made in the following situations: when histopathologic analysis of surgical specimen or histology obtained by needle biopsy was positive for secondary localization or for a primary cancer; and when a significant increase (diameter increase by ≥ 25%) in size or number (or both) of pulmonary nodules was found at imaging studies during follow-up. The nodules that did not change during follow-up or that spontaneously resolved without employing therapy were considered to be benign lesions. The observation follow-up period was at least 2 years. The characterization of lung nodules was verified by histology specimens in 28 patients and imaging studies in the remaining 31 patients. The difference between correct and incorrect reclassification according to patient outcome was defined as the net reclassification improvement.
Statistical Analysis
Continuous variables were expressed as a median (range) and categoric data as a percentage. Comparisons between groups were performed by Mann-Whitney U test, chi-square test with Yates correction, and McNemar test, as appropriate. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and diagnostic accuracy for the detection of the lung lesions were calculated using standard methods. A two-sided p value was used, and p < 0.05 was considered statistically significant (Advanced Models version 15.0, SPSS).
Results
PET/CT Accuracy and Lung Nodules
Thirty-one patients had an SPN, and 28 had multiple lung lesions. The median diameter of the SPNs was 12 mm (range, 5–50 mm), and that of multiple lesions was 10 mm (range, 5–18 mm) (by Mann-Whitney U test, p < 0.05). Moreover, the cumulative mean of multiple lung nodule diameters was 23.2 mm (range, 13–67.2 mm).
Nodules were located in an upper lobe in 29 patients (49%): in 15 of 31 (48%) patients with an SPN and in 14 of 28 (50%) patients with multiple lung lesions. SPNs had spiculated margins in 17 patients, and of these, 12 were located in the upper lobe. Thirty-one patients (52%) had histology-proven (on specimens from bronchoscopy and surgery) or imaging-proven (on serial CT studies) lung malignancy; of these patients, 23 (74%) had metastatic disease, and eight (26%) had a primary lung tumor. Eighteen of 31 (58%) patients with an SPN and 13 of 28 (46%) patients with multiple lung lesions had pulmonary malignancy. PET/CT findings were negative in 32 and positive in 27 patients; in particular, positive findings were present in 45% of patients with an SPN and in 46% of patients with multiple nodules. Table 1 presents the characteristics of lung nodules in accordance with histologic and imaging results. PET/CT findings were true-positive in 24 patients (77%) and true-negative in 25 (89%) patients. The NPV of PET/CT was 78% and the PPV of PET/CT was 89% for all nodule types. Table 2 presents the sensitivities, specificities, PPVs, NPVs, and accuracies of PET/CT for lung nodules. As expected, specificity was higher for SPN than for multiple nodules, whereas a high sensitivity was obtained for multiple lesions at PET/CT.
| Characteristics | Total (n = 59) | No. of Nodules | Histologic Classification | ||||
|---|---|---|---|---|---|---|---|
| Single | Multiple | p | Malignant (n = 31) | Benign (n = 28) | p | ||
| Clinical features | |||||||
| Median age (range), yearsa | 70 (43–86) | 68 (45–86) | 70 (43–86) | NS | 80 (45–86) | 70 (43–77) | NS |
| Male | 32 | 17 | 15 | 16 (52%) | 16 (57%) | NS | |
| Smoking historyb | 32 | 16 | 16 | NS | 15 (61%) | 17 (48%) | NS |
| History of cancerb | NS | ||||||
| ≤ 5 years | 46 | 26 | 20 | 22 (71%) | 24 (86%) | ||
| > 5 years | 13 | 5 | 8 | NS | 9 (29%) | 4 (14%) | |
| Radiologic findings | |||||||
| Upper lobe locationb | 29 | 15 | 14 | NS | 17 (55%) | 12 (43%) | NS |
| Spiculated nodulesb | 27 | 17 | 10 | NS | 16 (52%) | 11 (39%) | NS |
| Median of maximum nodule size (mm)a | 11 (5–50) | 12 (5–50) | 10 (5–18) | < 0.05 | 13.3 (5–50) | 9 (5–22) | < 0.005 |
| Positive PET resultc | 27 | 14 | 13 | NS | 24 (77%) | 3 (11%) | NS |
Note—Except where otherwise indicated, data are given as no. (%). NS = not significant.
a
By Mann-Whitney Utest.
b
By chi-square test.
c
By McNemar test.
| Nodule Type | TP | TN | FP | FN | Sensitivity (95% CI), % | Specificity (95% CI), % | PPV (95% CI), % | NPV (95% CI), % | Accuracy (95% CI), % |
|---|---|---|---|---|---|---|---|---|---|
| All nodules (n = 59) | 24 | 25 | 3 | 7 | 77 (63–92) | 89 (78–100) | 89 (78–100) | 78 (63–93) | 83 (73–93) |
| Solitary pulmonary nodule (n = 31) | 13 | 12 | 1 | 5 | 72 (52–93) | 92 (78–100) | 93 (81–100) | 71 (46–95) | 81 (67–95) |
| Multiple lung nodules (n = 28) | 11 | 13 | 2 | 2 | 85 (65–100) | 87 (69–100) | 85 (65–100) | 87 (69–100) | 86 (73–97) |
Note—TP = no. of true-positives, TN = no. of true-negatives, FP = no. of false-positives, FN = no. of false-negatives, PPV = positive predictive value, NPV = negative predictive value.
Pretest Probability of Malignancy and PET/CT
On the basis of the Mayo Clinic model (n = 31 patients), 18 (58.1%) patients had low, 12 (38.7%) had intermediate, and one (3.2%) had high pretest likelihood of malignancy. According to the PET/CT results, 13 (76.4%) patients had low, three (17.6%) had intermediate, and one (5.9%) had high pretest likelihood of malignancy and had negative findings on PET/CT. Therefore, 28% and 75% of patients with low and intermediate pretest likelihood of malignancy had positive findings on PET/CT, respectively. According to final outcomes, 10 (52.6%) patients with low, nine (47.4%) with intermediate, and none with high pretest likelihood of malignancy had positive histology or convincing enlargement on serial CT studies.
The distribution of pretest likelihood of malignancy, according to the VA model, was low in five (19.2%) patients, intermediate in 20 (76.9%), and high in one (3.8%). Negative scan findings were reported in one (7.1%) patient with low, 12 (85.8%) with intermediate, and one (7.1%) with high pretest likelihood of malignancy; therefore, a high percentage of patients in the low and intermediate risk categories had positive findings on PET/CT (80% and 40%, respectively).
On the basis of the classification by Gould et al. [5] and according to the Mayo Clinic model, 18 (58.1%) patients were had very low, six (19.4%) had low, six (19.4%) had intermediate, and one had high pretest likelihood of malignancy. By contrast, on the basis of the VA model, three (11.5%) patients had very low, 17 (65.4%) had low, five had intermediate, and one had high likelihood of malignancy (Table 3). As illustrated in Table 3, in accordance with the Mayo Clinic model, all participants at intermediate likelihood of malignancy (> 40–60%) had positive findings on PET/CT and a true malignancy at follow-up. On the contrary, the majority of patients at very low (100%) and low probability (41.2%) of malignancy based on the VA model had positive findings on PET/CT and on follow-up. Finally, 13 of 15 (86.7%) patients with high pretest likelihood and negative findings on PET/CT did not have a lung malignancy.
| Classification of Likelihood of Malignancy | Histologically Benign Nodule | Histologically Malignant Nodule | Overall | ||
|---|---|---|---|---|---|
| Negative PET/CT Findings | Positive PET/CT Findings | Negative PET/CT Findings | Positive PET/CT Findings | ||
| Mayo Clinic model | |||||
| Very low (< 5%) | 8 (44.4) | 0 | 5 (27.8) | 5 (27.8) | 18 |
| Low (< 5–40%) | 3 (50) | 1 (16.7) | 0 | 2 (33.3) | 6 |
| Intermediate (> 40–60%) | 0 | 0 | 0 | 6 (100) | 6 |
| High (> 60%) | 1 (100) | 0 | 0 | 0 | 1 |
| Subtotal | 12 | 1 | 5 | 13 | 31 |
| Veterans Affairs Cooperative Study model | |||||
| Very low (< 5%) | 0 | 0 | 0 | 3 (100) | 3 |
| Low (< 5–40%) | 6 (35.3) | 0 | 4 (23.5) | 7 (41.2) | 17 |
| Intermediate (> 40–60%) | 2 (40) | 1 (20) | 1 (20) | 1 (20) | 5 |
| High (> 60%) | 1 (100) | 0 | 0 | 0 | 1 |
| Subtotal | 9 | 1 | 5 | 11 | 26 |
Note—Data are given as no. (%) of patients with described findings.
Overall, in the 59 patients, net reclassification improvement was 0.95 (z = 3.60; p < 0.0001) and 1.6 (z = 4.14; p < 0.0001), respectively, for the Mayo Clinic and VA models. In particular, according to the Mayo Clinic model, of 18 patients with low and intermediate pretest likelihood of malignancy with lung disease, 13 (72.2%) had positive findings on PET/CT. Conversely, of 15 patients with high pretest likelihood of malignancy who had benign findings on follow-up, 86.7% had negative findings on PET/CT. Therefore, the addition of PET/CT yielded a change in classification in 4 of 13 (30.7%) patients without lung malignancy (z = 1.34; p = 0.18) for the Mayo Clinic model and in 9 of 10 (90%) patients (z = 2.52; p < 0.05) for the VA model. Moreover, a change in classification was registered in 13 of 18 (72.2%) patients with lung disease (z = 3.60; p < 0.0001) for the Mayo Clinic model and in 11 of 16 (68.8%) patients (z = 3.31; p < 0.0001) for the VA model. Table 4 presents the correlations among histology, PET/CT, and probability of malignancy based on both the Mayo Clinic and VA models.
| Classification of Likelihood of Malignancy | Mayo Clinic Model | Veterans Affairs Cooperative Study Model | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Histologically Benign Nodule | Histologically Malignant Nodule | Subtotal | Histologically Benign Nodule | Histologically Malignant Nodule | Subtotal | |||||
| Negative PET/CT Findings | Positive PET/CT Findings | Negative PET/CT Findings | Positive PET/CT Findings | Negative PET/CT Findings | Positive PET/CT Findings | Negative PET/CT Findings | Positive PET/CT Findings | |||
| Low | 8 (44) | 0 | 5 (28) | 5 (28) | 18 | 0 | 0 | 1 (20) | 4 (80) | 5 |
| Intermediate | 3 (25) | 1 (8) | 0 | 8 (67) | 12 | 8 (40) | 1 (5) | 4 (20) | 7 (35) | 20 |
| High | 1 (100) | 0 | 0 | 0 | 1 | 1 (100) | 0 | 0 | 0 | 1 |
| Total | 12 | 1 | 5 | 13 | 31 | 9 | 1 | 5 | 11 | 26 |
Note—Data are given as no. (%) of patients with described findings.
On the basis of SUVmax, 19 of 33 (58%) patients with positive findings on PET/CT had a value greater than 4.0. In particular, we found that 30% and 44.4% of patients with intermediate risk had an SUVmax less than 4.0, respectively, for the Mayo Clinic and VA models. Conversely, 10 of 17 (58.5%) patients with high risk had an SUVmax greater than 4.0.
Discussion
Risk stratification with clinical prognostic models may be useful for advising patients and making treatment decisions. The distribution of predicted values for a given model may indicate the number of patients who should be classified into each risk category but not whether this is done correctly [14]. It is difficult to determine which of the two models, the Mayo Clinic or the VA model, is better at classifying pulmonary nodules in individuals or to determine whether individual risk estimates differ between the two models. One may examine comparative distributions through risk reclassification as a novel marker for the incremental value of a test. This represents the primary endpoint of the present study. We show that PET/CT is able to further stratify the presence of pulmonary malignancy in patients with low- to indeterminate-risk lung nodules and history of cancer. In particular, a substantial number of pulmonary nodules considered as having low or intermediate pretest likelihood of malignancy had abnormal findings on PET/CT that were associated with malignancy.
Bayesian analysis involves the use of likelihood ratios for imaging findings and clinical features associated with SPNs to establish probability of malignancy [15, 16]. A general Bayesian model principle is that each characteristic used in the development of the model is conditionally independent of all others. Several previous studies have calculated the likelihood ratios for a range of values for clinical variables in predicting malignancy. Clinical variables that have been correlated with the probability of malignancy include baseline incidence of malignancy, age, smoking history, nodule size, nodule edge characteristics, and presence or absence of occult calcification on CT densitometry. Positive findings on PET/CT appear to represent a significantly better indicator of malignancy compared with findings on CT.
Clinical features—including symptoms, physical examination findings, and laboratory results—are usually nonspecific and may not be able to stratify risk between benign and malignant conditions. Bronchoscopy has limited usefulness in a patient with an SPN. Transthoracic needle aspiration biopsy with fluoroscopic CT or ultrasound guidance is diagnostic in 80–95% of malignant nodules. An indeterminate biopsy report, however, may not exclude malignancy [17]. The objective of PET before exploratory thoracotomy in patients with benign solitary nodules is to determine appropriateness for surgery, as well as to prevent complications that occur in a significant percentage of patients.
We found a prevalence of lung malignancy in 52% of the patients enrolled in the present study. SPNs have been documented to have a 40% malignancy rate [18, 19]. This latter percentage varies with age, smoking history, location, and patient selection [20, 21]. We considered a subset of oncologic patients who represent a high-risk category. In our study, PET/CT showed different diagnostic accuracies based on the type of lung nodules. For example, a similar sensitivity was registered for SPNs and multiple lung lesions (72% vs 78%, respectively), whereas a lower specificity was reported for multiple nodules than for a single nodule (85% vs 92%, respectively). The rate of false-positives can be attributed to abnormal 18F-FDG uptake in benign lesions, including infections such as histoplasmosis and tuberculosis, hamartoma, lung cysts, and vascular abnormalities. Respiratory system toxicity is a common side effect and complication of anticancer therapy. Indeterminate lung nodules in patients with cancer history should be characterized in a minimally invasive manner. The accuracy of PET/CT in our study was high, which is in accordance with previous data [22].
Pulmonary metastases are common. Necropsy studies have reported evidence of pulmonary metastases in 50% or more of patients with sarcoma, melanoma, and cancers of the breast, prostate, thyroid, uterus, and kidney [23, 24]. Our study found a substantial rate of metastatic disease (39%) and primary lung cancer (14%). Nearly half of the patients in the present study who had lung metastases also had breast or gastrointestinal tumors. By contrast, in accordance with Quint et al. [25], primary lung cancer was found to be more frequent in patients with head and neck cancer, although these patients are thought to be at a greater risk for metastatic disease.
We opted to study lung nodules of 5 mm or more in diameter, which includes lesions that challenge the spatial resolution limits of PET/CT and of both the Mayo Clinic and VA models. The VA model considers only those lung nodules with diameter of 7–30 mm, whereas the Mayo Clinic model has been applied to patients with newly discovered 4- to 30-mm radiologically indeterminate SPNs. In the present study, only five patients had an SPN with a diameter of 5 mm or more. Therefore, based on the variables conceived in each model, we calculated the likelihood of malignancy in 31 patients and 26 subjects in accordance with the Mayo Clinic and VA models, respectively. Bryant and Cerfolio [26] showed that FDG PET/CT may be useful for the management of nodules less than 10 mm in diameter. Pomerri et al. [27] concluded that a diameter greater than 5 mm and irregular margins were strong predictors of nodule growth by multiple stepwise regression analysis, whereas calcification was associated with a very low likelihood of progression.
As discussed by Schultz et al. [4], both the Mayo Clinic and VA models appear to be sufficiently accurate to assist in making clinical decisions about the choice and interpretation of subsequent diagnostic tests. The accuracy of both models was similar to that reported in the original articles describing their development [2, 3]. The Mayo Clinic model was slightly more accurate than the VA model, but this difference was not statistically significant. Furthermore, the Mayo Clinic model is probably a better choice in practice settings with a low prevalence of malignant lung nodules; conversely, the VA model may be a better choice in settings with a high prevalence of malignant SPNs. From the present analysis, on the basis of the Mayo Clinic model, it appears that all subjects with intermediate likelihood of malignancy (> 40–60%) had positive findings on PET/CT and a true malignancy at follow-up. By contrast, on the basis of the VA model, the majority of patients with very low (100%) or low (41.2%) probability of malignancy had positive findings both on PET/CT and at follow-up. This is in line with the concept that the VA model is preferred in patients with high probability of disease. We used both models because they are consistently complementary. Nevertheless, the VA model is missing some variables, such as spiculation and remote history of thoracic cancer, that are present in the Mayo Clinic model, and the Mayo Clinic model is missing the variable in the VA model that indicates the number of years since quitting smoking. Moreover, as previously stated, the Mayo Clinic model was conceived for patients with a broader spectrum of nodule diameter than was the VA model.
Physicians may overestimate the probability of a malignant lesion in patients with low risk of malignant disease. According to our results, FDG PET can be considered to be helpful in low-risk patients (i.e., those with a high NPV of normal findings on FDG PET, as reported in Table 4 for the Mayo Clinic model). In patients with intermediate risk, biopsy is still recommended. If biopsy is impractical or yields an equivocal result, FDG PET can be helpful as an additional examination. A positive finding on FDG PET requires further histolopathologic confirmation [28]. In patients with low to moderate pretest probability of malignancy and an intermediate SPN that measures at least 8–10 mm in diameter, the American College of Chest Physicians recommends FDG PET to characterize the SPN [29]. As reported by Herder et al. [7], predictive models tended to underestimate the probability of malignancy, particularly with lower-risk patients. In the present study, we encountered positive findings on PET/CT in 28% and 80% of patients otherwise classified as low risk for the Mayo Clinic and VA models, respectively. In a study by Gould et al. [3], PET was used to refine the management of patients after the pretest probability of cancer was determined. If the pretest probability was low (20%) and findings on PET were negative, then the posttest probability of malignancy was less than 2% and follow-up was suggested. However, when findings on PET are negative and the pretest probability is relatively high (65%), the calculated posttest probability increases to more than 10%, and needle-biopsy or video-assisted thoracic surgery biopsy should still be considered. Therefore, PET can be more useful for refining the management of patients at low risk, rather than those at high risk of malignancy.
Our findings show that the addition of PET/CT results in reclassification of approximately 59% of the sample, with a net reclassification improvement ranging from 0.23 to 1.6. The net reclassification improvement in patients with proven lung malignancy ranged from 0.23 to 1.1, whereas the net reclassification improvement for patients without secondary or primary lung cancer ranged from 0.23 to 0.5. These results suggest that, when applied to a high-risk population with suspicion for primary or secondary lung cancer, an FDG PET/CT–adjusted strategy may effectively identify more individuals who had cancer but it can identify many other individuals at higher risk who do not have malignancy (13 of 15 patients considered to be at high risk by CT were correctly stratified as being at low risk by PET/CT). PET is most useful in the pretest probability range of 5–60% [30, 31], as illustrated in Table 4. Moreover, PET may be helpful in the detection of mediastinal lymph node metastases, even when the nodes are not enlarged on CT [32], and for detection of occult extrathoracic metastasis and of synchronous extrathoracic primary malignancies.
Grgic et al. [11] investigated the prognostic value of SUV in SPN for survival. They found that the highest diagnostic accuracy was achieved with an SUV of more than 4.0 (sensitivity, specificity, and accuracy all of 85%) whereas visual interpretation achieved corresponding values of 94, 70, and 84%, respectively. In the current study, we observed that more than the half of patients at intermediate to high risk of malignancy had an SUVmax greater than 4.0, but approximately 40% of patients at intermediate risk had a low SUVmax. Therefore, SUVmax may not be predictive of malignancy in this setting. In comparison with visual image analysis, a semiquantitative approach has the conceptual advantage of objectiveness. Apostolova et al. [33] reported that correction of motion blur is mandatory for accurate SUV quantification of SPNs. Motion blur causes relevant (≥ 30%) underestimation of actual SUV in most SPNs of up to 30 mm in FDG PET/CT. However, the results of SUV measurements are also prone to heterogeneity owing to prevailing differences in data acquisition and reconstruction methods [34] and are in relation with previous treatments.
The most important limitation of the present study is that histologic diagnosis was made in only 28 of the 59 patients. As previously stated, small nodule size, respiratory misregistration, and inaccurate SUVmax assessment owing to volume averaging may represent additional limitations.
In conclusion, in oncologic patients with low and intermediate likelihood of lung nodule malignancy, FDG PET/CT can be helpful for excluding the presence of primary or metastatic malignant lung lesions. FDG PET/CT is most efficacious in patients with low to intermediate probability of malignancy. A well-structured clinical trial would be important to confirm the utility of FDG PET/CT for the evaluation of radiologically indeterminate lung nodules, in both patients with and those without history of malignancy.
References
1.
Lillington GA. Management of solitary pulmonary nodules: how to decide when resection is required. Postgrad Med 1997; 101:145–150
2.
Swensen SJ, Silverstein MD, Ilstrup DM, et al. The probability of malignancy in solitary pulmonary nodules: application to small radiologically indeterminate nodules. Arch Intern Med 1997; 157:849–855
3.
Gould MK, Ananth L, Barnett PG; Veterans Affairs SNAP Cooperative Study Group. A clinical model to estimate the pretest probability of lung cancer in patients with solitary pulmonary nodules. Chest 2007; 131:383–388
4.
Schultz EM, Sanders GD, Trotter PR, et al. Validation of two models to estimate the probability of malignancy in patients with solitary pulmonary nodules. Thorax 2008; 63:335–341
5.
Gould MK, Fletcher J, Iannettoni MD, et al. Evaluation of patients with pulmonary nodules: when is it lung cancer? ACCP evidence-based clinical practice guidelines (2nd edition). Chest 2007; 132(suppl 3):108S–130S
6.
Dietlein M, Weber K, Gandjour A, et al. Cost-effectiveness of FDG-PET for the management of solitary pulmonary nodules: a decision analysis based on cost reimbursement in Germany. Eur J Nucl Med 2000; 27:1441–1456
7.
Herder GJ, van Tinteren H, Golding RP, et al. Clinical prediction model to characterize pulmonary nodules: validation and added value of 18F-fluorodeoxyglucose positron emission tomography. Chest 2005; 128:2490–2496
8.
Lillington GA, Caskey CI. Evaluation and management of solitary and multiple pulmonary nodules. Clin Chest Med 1993; 14:111–119
9.
Ginsberg MS, Griff SK, Go BD, et al. Pulmonary nodules resected at video-assisted thoracoscopic surgery: etiology in 426 patients. Radiology 1999; 213:277–282
10.
Boellaard R. Standards for PET image acquisition and quantitative data analysis. J Nucl Med 2009; 50:11S–20S
11.
Grgic A, Yüksel Y, Gröschel A, et al. Risk stratification of solitary pulmonary nodules by means of PET using 18F-fluorodeoxyglucose and SUV quantification. Eur J Nucl Med Mol Imaging 2010; 37:1087–1094
12.
Cook NR, Ridker PM. Advances in measuring the effect of individual predictors of cardiovascular risk: the role of reclassification measures. Ann Intern Med 2009; 150:795–802
13.
Pencina MJ, D'Agostino RB Sr, D'Agostino RB Jr, et al. Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med 2008; 27:157–172
14.
Cook NR. Statistical evaluation of prognostic versus diagnostic models: beyond the ROC curve. Clin Chem 2008; 54:17–23
15.
Gurney JW. Determining the likelihood of malignancy in solitary pulmonary nodules with Bayesian analysis. Part I. Theory. Radiology 1993; 186:405–413
16.
Black WC, Armstrong P. Communicating the significance of radiologic test results: the likelihood ratio. AJR 1986; 147:1313–1318
17.
Beigelman-Aubry C, Hill C, Grenier PA. Management of an incidentally discovered pulmonary nodule. Eur Radiol 2007; 17:449–466
18.
Huston J 3rd, Muhm JR. Solitary pulmonary opacities: plain tomography. Radiology 1987; 163:481–485
19.
Lillington GA. Pulmonary nodules: solitary and multiple. Clin Chest Med 1982; 3:361–367
20.
Gupta NC, Maloof J, Gunel E. Probability of malignancy in solitary pulmonary nodules using fluorine-18-FDG and PET. J Nucl Med 1996; 37:943–948
21.
Ost D, Fein AM, Feinsilver SH. The solitary pulmonary nodule. N Engl J Med 2003; 348:2535–2542
22.
Kim SK, Allen-Auerbach M, Goldin J, et al. Accuracy of PET/CT in characterization of solitary pulmonary lesions. J Nucl Med 2007; 48:214–220
23.
Luce JA. Metastatic and malignant tumors. In: Murray JF, Nadel J, eds. Textbook of respiratory medicine, 3rd ed. Philadelphia, PA: Saunders, 2000:1470
24.
Weiss L, Gilbert HA. Pulmonary metastasis. Boston, MA: GK Hall, 1978
25.
Quint LE, Park CH, Iannettoni MD. Solitary pulmonary nodules in patients with extrapulmonary neoplasms. Radiology 2000; 217:257–261
26.
Bryant AS, Cerfolio RJ. The maximum standardized uptake values on integrated FDG-PET/CT is useful in differentiating benign from malignant pulmonary nodules. Ann Thorac Surg 2006; 82:1016–1020
27.
Pomerri F, Pucciarelli S, Maretto I, et al. Significance of pulmonary nodules in patients with colorectal cancer. Eur Radiol 2012; 22:1680–1686
28.
Hellwig D, Baum RP, Kirsch CM. FDG-PET, PET/CT and conventional nuclear medicine procedures in the evaluation of lung cancer: a systematic review. Nuklearmedizin 2009; 48:59–69
29.
Alberts WM. Diagnosis and management of lung cancer executive summary: ACCP evidence-based clinical practice guidelines (2nd edition). Chest 2007; 132(suppl 3):1S–19S
30.
Gould MK, Sanders GD, Barnett PG, et al. Cost-effectiveness of alternative management strategies for patients with solitary pulmonary nodules. Ann Intern Med 2003; 138:724–735
31.
Fischer BM, Mortensen J, Højgaard L. Positron emission tomography in the diagnosis and staging of lung cancer: a systematic, quantitative review. Lancet Oncol 2001; 2:659–666
32.
Goldsmith SJ, Kostakoglu L. Nuclear medicine imaging of lung cancer. Radiol Clin North Am 2000; 38:511–524
33.
Apostolova I, Wiemker R, Paulus T, et al. Combined correction of recovery effect and motion blur for SUV quantification of solitary pulmonary nodules in FDG PET/CT. Eur Radiol 2010; 20:1868–1877
34.
Boellaard R, Buijs F, de Jong HW, et al. Characterization of a single LSO crystal layer high resolution research tomograph. Phys Med Biol 2003; 48:429–448
FOR YOUR INFORMATION
The comprehensive book based on the ARRS 2013 annual meeting categorical course on Body MRI is now available! For more information or to purchase a copy, see www.arrs.org.
Information & Authors
Information
Published In
Copyright
© American Roentgen Ray Society.
History
Submitted: August 15, 2013
Accepted: September 26, 2013
First published: February 20, 2014
Keywords
Authors
Metrics & Citations
Metrics
Citations
Export Citations
To download the citation to this article, select your reference manager software.