ORIGINAL ARTICLE
Indeterminate Pulmonary Nodules in Colorectal Cancer Follow-up Guidelines Based on a Risk Predictive Model Chang Hyun Kim, MD,∗ Jung Wook Huh, MD, PhD,† Hyeong Rok Kim, MD, PhD,∗ and Young Jin Kim, MD, PhD∗ Background: Chest computed tomographic (CT) scans frequently detect indeterminate pulmonary nodules (IPNs) in patients with colorectal cancer. The discovery of such nodules creates a clinical dilemma. Purpose: This study was performed to identify clinical characteristics of IPNs and develop a predictive model to predict the risk of progression to pulmonary metastases in patients with colorectal cancer. Methods: We analyzed data from a prospectively collected database involving 1195 patients with colorectal carcinoma who underwent curative surgery between January 2008 and June 2010. A predictive model was constructed on the basis of the probability risk score and validated in 115 patients collected from a separate treatment period. Results: Of the 1195 patients who underwent a baseline staging chest computed tomography, 326 (27.2%) had IPNs. During a median follow-up of 26.7 months (interquartile range: 18.0–37.2), 74 (28.1%) showed pulmonary metastases. Five variables maintained prognostic significance after multivariate analysis: metachronous nodule, bilateral involvement, positive perineural invasion, increased number of positive lymph nodes, and rectal location of cancer. The 2-year progression-free survival rates for the very low-, low-, intermediate-, and high-risk groups were 96%, 82%, 46%, and 16%, respectively (P < 0.001), with a concordance index of 0.81 (95% confidence interval, 0.75–0.86). This model was validated in a separate patient set (P < 0.001), with a C-index of 0.83 (95% confidence interval, 0.77–0.88). Conclusions: A predictive model for progression of IPNs may be clinically useful in discriminating patients who might benefit from an aggressive surveillance program and early pulmonary metastasectomies. Keywords: colorectal cancer, lung nodule, multidetector computed tomography, prognostic factor, pulmonary metastasis
However, these recommendations are not based on any firm guidelines but on guidelines developed in the context of primary lung cancer–screening programs.10–12 In addition, several studies have investigated the clinical relevance of pulmonary nodule(s) in patients with metastatic tumors, the data from these studies were limited in that they included various sites of metastatic origin, including carcinoma of head and neck, bladder, breast, cervix, esophagus, ovary, prostate, colon, rectum, liver, kidney, and uterus. Accordingly, there is a wide range of clinical characteristics and propensity to undergo pulmonary metastasis, and the extrapolation of these results to the management strategy of patients with CRC may be inappropriate. Subsequently, it was recommended that IPNs be managed according to the clinical situation without any firm practical guidelines.13,14 Furthermore, although some authors have reported the clinical significance of small pulmonary nodules in patients with CRC in various clinical settings,15–23 there is little consensus regarding the definition, incidence, risk factors for progression, and the best follow-up or treatment strategies for IPNs. To date, there is a paucity of evidence addressing the issue of radiologically ambiguous IPNs that are detected on chest computed tomograph in patients with CRC. The aim of this study was to define the characteristics and clinical significance of IPNs in patients after surgery with curative intent for CRC. Our main purpose was to construct a risk score derived from the combination of independent predictors of progression to definitive pulmonary metastases, with the aim of standardizing the appropriate timing of radiologic follow-up and management to optimize outcomes.
(Ann Surg 2015;261:1145–1152)
PATIENTS AND METHODS
T
he lung is the most common extra-abdominal site of metastases after curative resection of colorectal cancer (CRC).1 Although many studies have reported that resection of pulmonary metastases is associated with a favorable 5-year overall survival rate that ranges from 24% to 67.8%,2–6 only 4.1% of patients with synchronous pulmonary metastases and 14.8% of those with metachronous pulmonary metastases are treated with curative surgical intent and benefit from metastasectomy.6,7 To select and expand this patient population at an earlier stage, preoperative evaluation currently includes chest computed tomography (CT) rather than chest radiography.8,9 However, chest CT enables increased detection of indeterminate pulmonary nodules (IPNs). As their name implies, radiologists have difficulty determining the malignant potential of these IPNs and usually recommend a short-term follow-up chest CT. From the ∗ Department of Surgery, Chonnam National University Hwasun Hospital and Medical School, Gwangju, Korea; and †Department of Surgery, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Korea. Disclosure: The authors declare no conflicts of interest. Reprints: Jung Wook Huh, MD, PhD, Department of Surgery, Samsung Medical Center, Sungkyunkwan University School of Medicine, 81 Irwon-ro, Gangnam-gu, Seoul 135-710, Korea. E-mail: [email protected]. C 2014 Wolters Kluwer Health, Inc. All rights reserved. Copyright ISSN: 0003-4932/14/26106-1145 DOI: 10.1097/SLA.0000000000000853
Annals of Surgery r Volume 261, Number 6, June 2015
Between January 2008 and June 2010, data were collected for 1386 patients who underwent CRC surgery at Chonnam National University Hwasun Hospital. This study was approved by the appropriate institutional review board. Exclusion criteria for the present study were (1) unresectable distant metastases, (2) multiple synchronous CRC, (3) familial adenomatous polyposis, (4) recurrent CRC, and (5) combined malignancy. We further excluded 113 patients who did not have a baseline computed tomographic assessment and 35 patients who showed highly malignant features in their initial baseline chest computed tomographic scans. Although IPNs were detected at the time of initial diagnosis, 63 patients with less than 24 months of follow-up were also excluded. Ultimately, a total of 1195 patients were enrolled in this analysis (Fig. 1). All chest computed tomographic scans were evaluated with consensus by 2 radiologists who had at least 5 years of experience reading chest computed tomographic scans with specific regard to identifying and classifying abnormalities of parenchymal pulmonary nodules. IPNs were defined as persistent ground glass nodules smaller than 10 mm in diameter and solid nodules smaller than 20 mm in diameter with nonspiculated contours, no air bronchogram or pseudocavitation, no malignant-type calcification, and no intralesional fat or benign-type calcification.24 Postoperative adjuvant chemotherapy was dependent on the general condition and treatment compliance of the patient and the www.annalsofsurgery.com | 1145
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 261, Number 6, June 2015
Kim et al
synchronous and metachronous nodules and analyzed them separately.
Statistical Analyses
FIGURE 1. Flowchart showing patient recruitment. preference of the treating physician. The standard protocol for adjuvant therapy in patients with histologically proven high-risk stage II or stage III colon cancer included 5-fluorouracil (FU) plus leucovorin (LV) [5-FU/LV; 6 cycles of a monthly bolus of intravenous FU (400–425 mg·m−2 ·d−1 ) on days 1–5 and LV calcium (20 mg/m2 /d) on days 1–5] and FU/LV plus oxaliplatin [FOLFOX; 12 cycles of oxaliplatin (85 mg/m2 ) on day 1, LV calcium (200 mg/m2 ) as a 2hour infusion on day 1, and FU (400 mg/m2 ) as a bolus and as a 600 mg/m2 22-hour infusion on days 1 and 2 bimonthly, respectively]. Of the 123 patients with colon cancer, 59 (47.9%) received chemotherapy (25; 5-FU/LV, 34; FOLFOX). Although preoperative concurrent chemoradiation therapy (CRT) (n = 69) was our standard protocol for patients with locally advanced (radiologic T3/T4 or N±) rectal cancer, postoperative CRT was provided after pathologic confirmation for 16 patients. Radiotherapy consisted of a total of 5040 cGy delivered in 28 fractions. During the first and fifth weeks of radiotherapy, 5-FU/LV was administered identically to the pre- and postoperative CRT groups. FOLFOX chemotherapy was also used to treat tumors located in the upper rectum in 23 patients who had not received any CRT. Follow-up chest computed tomographic scans and abdominal computed tomographic scans were obtained at 3, 6, 9, 12, 18, and 24 months. Of the 263 patients who were eligible for final analysis, 199 (75.7%) completed this follow-up program. All scans were obtained on a 64-MDCT system (Discovery HD 750, GE Healthcare, Milwaukee, WI). After the RECIST criteria requirements, the largest nodule was measured along its longest axis using electronic calipers. During follow-up, progression to metastasis was defined according to RECIST criteria or by a histological diagnosis of metastatic cancer.25 Metachronous IPNs were defined as any nodules that newly appeared in the follow-up computed tomographic scans without growth of any previously identified IPNs. We identified 18 patients who had both 1146 | www.annalsofsurgery.com
Categorical variables are presented as numbers and percentages and continuous variables as medians and interquartile ranges (IQR). The significance of baseline differences was tested by the Mann-Whitney U test, Fisher exact test, or the χ 2 test, as appropriate. To evaluate the independent influence of prognostic factors on progression of IPN to definitive pulmonary metastasis, we used Cox proportional hazards regression to estimate hazard ratios and associated 95% confidence intervals. Variables that were significant at P ≤ 0.1 on univariate analysis were included as covariates in the multivariate model. After excluding variables that had low prevalence or exhibited multicolinearity, 12 variables exhibited prognostic significance by univariate analysis and were entered into the multivariate analysis: tumor location, T-stage, N-stage, perineural invasion, adjuvant chemotherapy, extrapulmonary disease, preoperative carcinoembryonic antigen (CEA), CEA at diagnosis of IPN, number of IPNs, chronicity, bilaterality, and configuration. Both CEA parameters were divided into 2 groups on the basis of a cutoff value of 5 ng/mL. Time to progression was calculated as the time from the baseline computed tomographic scan to the date of confirmation based on follow-up chest CT and/or histological biopsy. Progression-free survival was calculated by the Kaplan-Meier method, and differences in survival between groups were assessed by the log-rank test. The life table method was used to estimate the conditional hazard rate for progression to definite pulmonary metastasis. The conditional hazard rate for progression was defined as the probability of developing definitive pulmonary metastasis in a 3-month interval in patients who were free of pulmonary metastasis at the beginning of each interval. To develop a practical prognostic score, we assigned a risk score on the basis of factors identified by multivariate analysis with weighted points proportional to the β regression coefficient values (rounded to the nearest integer after multiplying by 2). A total progression risk score (PRS) for each patient was then calculated by summing the risk points of each prognostic factor. Based on the total PRS, 4 risk groups were defined using receiver operating characteristic curve analysis. The predictive discrimination of the scoring system was examined by calculating the concordance index (C-index).26 A Harrell’s C-index of 0.5 indicates no predictive discrimination and 1.0 indicates perfect separation of patients. The predictive model was validated in 2 different ways. First, the Cox multivariate models were internally validated using bootstrapping (1000 replications). Second, we selected a total of 115 patients who satisfied aforementioned inclusion criteria between January 2006 and December 2007 for the external validation. Statistical significance was set as P < 0.05 in a 2-tailed test. Statistical analyses were performed using SPSS version 20 (SPSS, Chicago, IL) or R 2.15.2.
RESULTS Patients Demographics Of the 1195 patients who had obtained staging computed tomographic scan, 326 (27.2%) had IPNs. Thirty five patients were considered to have a malignant nodule in the initial chest CT. Among them, 11 patients underwent video-assisted thoracoscopic surgery wedge resection. Two of the 11 cases were diagnosed as benign nodules and the other 9 as pulmonary metastases. Another 24 patients were diagnosed by radiology and showed metastasis in the serial chest CT (Fig. 1). Patient demographics for development set (n = 263) and validation set (n = 115) are listed in Table 1. There was a significantly higher rate of rectal cancer and stage III and stage IV diseases at C 2014 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 261, Number 6, June 2015
diagnosis among the patients with progression to pulmonary metastasis in both groups. Moreover, progression developed more frequently in patients who had perineural invasion (P < 0.001), higher preoperative CEA level (P < 0.001), and extrapulmonary distant metastatic disease (P < 0.001).
Pulmonary Nodule Analysis on CT The relationship between nodule character and progression for development set (n = 263) and validation set (n = 115) is summarized in Table 2. Of the 263 patients in the development set, baseline CT detected 438 synchronous IPNs in 194 patients and 118 metachronous IPNs in 69 patients. The median number of IPNs was 1 (IQR: 1–3) in the synchronous nodule and 1 (IQR: 1–2) in the metachronous nodule. One hundred forty-eight (56.3%) patients had a single nodule, 50 (19%) had 2 nodules, and 65 (24.7%) had more than 2 nodules. One hundred eighty-three patients (69.6%) had nodules, with a maximum size of less than 5 mm. Sixty-nine (26.2%) nodules ranged in size from 0.5 mm to 1 cm, with 11 (4.2%) nodules exceeding 1 cm. The majority of the IPNs (234/263, 89%) had smooth margins and 95.4% were solid nodules. Multiple (P = 0.008), metachronous (P < 0.001), and bilaterally located (P = 0.023) IPNs showed a significantly increased incidence of progression to malignant nodules. In contrast to development set, the size of IPNs showed more significant relationship than number of IPNs in the validation set.
Lung Nodules in Colorectal Cancer
Progression to Pulmonary Metastasis Median follow-up for development set was 26.7 months (IQR: 18.0–37.2). During the study period, a total of 1993 chest CT scans were obtained on 263 patients who had a regular follow-up, with a median of 7 (IQR: 5–9) chest CT scans for each patient and 74 (28.1%) patients progressed to pulmonary metastases. Surgical resections of IPNs were performed in 42 patients. Of these, 34 (81%) were histologically confirmed to have metastatic CRC and the remaining 8 patients had the following benign histologic features: 3 had chronic granuloma, 2 had organizing pneumonia, 2 had benign fibrosis, and 1 had focal emphysema. Another 40 patients showed definitive pulmonary metastasis in the follow-up chest CT. Bilateral multiple pulmonary metastases accompanying the size increase of IPNs and/or radiologic malignant change which include a spiculated contours, air bronchogram or pseudocavitation, and any malignant-type calcification, of index IPNs were observed in 36 patients and in other 4 patients, the size of IPNs increased to larger than 3cm. IPNs with progression were mostly confirmed in the second (IQR: 1–4) follow-up computed tomographic scan; IPNs occurred in 59 (79.7%) patients within 1 year after the baseline CT and in 70 (94.6%) patients within 2 years (Fig. 2). Only in 1.5% (4 of 263) of patients, the progression developed after 2 years since its diagnosis. To evaluate whether chemotherapy can inhibit the growth of metastatic nodules in patients with synchronous IPNs, patients were compared according to the treatment regimen they received (Table 3). Patients who underwent simultaneous resection of IPNs with their primary CRC (n = 7, of which 4 exhibited metastasis)
TABLE 1. Clinicopathologic Factors of Patients With Progression and No Progression Development Set Progression (N = 74) Age Sex Male Female Location of primary tumor Colon Rectum Pathologic TNM stage 0–1 2 3 4 Differentiation Well-differentiated Moderate Poor Lymphovascular invasion Absent Present PNI Absent Present Preoperative CEA Primary tumor size Extrapulmonary disease Absent Present Adjuvant chemotherapy No Yes CEA at diagnosis of IPN, ng/mL
66.0 (57–72)
No Progression (N = 189) 67.0 (59–74)
45 (60.8%) 29 (39.2%)
127 (67.25%) 62 (32.8%)
19 (25.7%) 55 (74.3%)
104 (55.0%) 85 (45.0%)
8 (10.8%) 16 (21.6%) 37 (50.0%) 13 (17.6%)
56 (29.6%) 75 (39.7%) 54 (28.6%) 4 (2.1%)
23 (31.1%) 48 (64.9%) 3 (4.0%)
82 (43.4%) 97 (51.3%) 10 (5.3%)
57 (77.0%) 17 (23.0%)
152 (80.4%) 37 (19.6%)
24 (32.4%) 50 (67.6%) 6.3 (3.2–14.6) 4.0 (2.5–5.5)
138 (73.0%) 51 (27.0%) 3.7 (1.9–6.8) 4.0 (3.0–5.0)
53 (71.6%) 21 (28.4%)
181(95.8%) 8 (4.2%)
20 (27.0%) 54 (63.0%) 3.9 (1.9–12.3)
90 (47.6%) 99 (52.4%) 2.9 (1.7–4.9)
Validation Set P 0.236 0.328
Indeterminate Pulmonary Nodules in Colorectal Cancer Follow-up Guidelines Based on a Risk Predictive Model Chang Hyun Kim, MD,∗ Jung Wook Huh, MD, PhD,† Hyeong Rok Kim, MD, PhD,∗ and Young Jin Kim, MD, PhD∗ Background: Chest computed tomographic (CT) scans frequently detect indeterminate pulmonary nodules (IPNs) in patients with colorectal cancer. The discovery of such nodules creates a clinical dilemma. Purpose: This study was performed to identify clinical characteristics of IPNs and develop a predictive model to predict the risk of progression to pulmonary metastases in patients with colorectal cancer. Methods: We analyzed data from a prospectively collected database involving 1195 patients with colorectal carcinoma who underwent curative surgery between January 2008 and June 2010. A predictive model was constructed on the basis of the probability risk score and validated in 115 patients collected from a separate treatment period. Results: Of the 1195 patients who underwent a baseline staging chest computed tomography, 326 (27.2%) had IPNs. During a median follow-up of 26.7 months (interquartile range: 18.0–37.2), 74 (28.1%) showed pulmonary metastases. Five variables maintained prognostic significance after multivariate analysis: metachronous nodule, bilateral involvement, positive perineural invasion, increased number of positive lymph nodes, and rectal location of cancer. The 2-year progression-free survival rates for the very low-, low-, intermediate-, and high-risk groups were 96%, 82%, 46%, and 16%, respectively (P < 0.001), with a concordance index of 0.81 (95% confidence interval, 0.75–0.86). This model was validated in a separate patient set (P < 0.001), with a C-index of 0.83 (95% confidence interval, 0.77–0.88). Conclusions: A predictive model for progression of IPNs may be clinically useful in discriminating patients who might benefit from an aggressive surveillance program and early pulmonary metastasectomies. Keywords: colorectal cancer, lung nodule, multidetector computed tomography, prognostic factor, pulmonary metastasis
However, these recommendations are not based on any firm guidelines but on guidelines developed in the context of primary lung cancer–screening programs.10–12 In addition, several studies have investigated the clinical relevance of pulmonary nodule(s) in patients with metastatic tumors, the data from these studies were limited in that they included various sites of metastatic origin, including carcinoma of head and neck, bladder, breast, cervix, esophagus, ovary, prostate, colon, rectum, liver, kidney, and uterus. Accordingly, there is a wide range of clinical characteristics and propensity to undergo pulmonary metastasis, and the extrapolation of these results to the management strategy of patients with CRC may be inappropriate. Subsequently, it was recommended that IPNs be managed according to the clinical situation without any firm practical guidelines.13,14 Furthermore, although some authors have reported the clinical significance of small pulmonary nodules in patients with CRC in various clinical settings,15–23 there is little consensus regarding the definition, incidence, risk factors for progression, and the best follow-up or treatment strategies for IPNs. To date, there is a paucity of evidence addressing the issue of radiologically ambiguous IPNs that are detected on chest computed tomograph in patients with CRC. The aim of this study was to define the characteristics and clinical significance of IPNs in patients after surgery with curative intent for CRC. Our main purpose was to construct a risk score derived from the combination of independent predictors of progression to definitive pulmonary metastases, with the aim of standardizing the appropriate timing of radiologic follow-up and management to optimize outcomes.
(Ann Surg 2015;261:1145–1152)
PATIENTS AND METHODS
T
he lung is the most common extra-abdominal site of metastases after curative resection of colorectal cancer (CRC).1 Although many studies have reported that resection of pulmonary metastases is associated with a favorable 5-year overall survival rate that ranges from 24% to 67.8%,2–6 only 4.1% of patients with synchronous pulmonary metastases and 14.8% of those with metachronous pulmonary metastases are treated with curative surgical intent and benefit from metastasectomy.6,7 To select and expand this patient population at an earlier stage, preoperative evaluation currently includes chest computed tomography (CT) rather than chest radiography.8,9 However, chest CT enables increased detection of indeterminate pulmonary nodules (IPNs). As their name implies, radiologists have difficulty determining the malignant potential of these IPNs and usually recommend a short-term follow-up chest CT. From the ∗ Department of Surgery, Chonnam National University Hwasun Hospital and Medical School, Gwangju, Korea; and †Department of Surgery, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Korea. Disclosure: The authors declare no conflicts of interest. Reprints: Jung Wook Huh, MD, PhD, Department of Surgery, Samsung Medical Center, Sungkyunkwan University School of Medicine, 81 Irwon-ro, Gangnam-gu, Seoul 135-710, Korea. E-mail: [email protected]. C 2014 Wolters Kluwer Health, Inc. All rights reserved. Copyright ISSN: 0003-4932/14/26106-1145 DOI: 10.1097/SLA.0000000000000853
Annals of Surgery r Volume 261, Number 6, June 2015
Between January 2008 and June 2010, data were collected for 1386 patients who underwent CRC surgery at Chonnam National University Hwasun Hospital. This study was approved by the appropriate institutional review board. Exclusion criteria for the present study were (1) unresectable distant metastases, (2) multiple synchronous CRC, (3) familial adenomatous polyposis, (4) recurrent CRC, and (5) combined malignancy. We further excluded 113 patients who did not have a baseline computed tomographic assessment and 35 patients who showed highly malignant features in their initial baseline chest computed tomographic scans. Although IPNs were detected at the time of initial diagnosis, 63 patients with less than 24 months of follow-up were also excluded. Ultimately, a total of 1195 patients were enrolled in this analysis (Fig. 1). All chest computed tomographic scans were evaluated with consensus by 2 radiologists who had at least 5 years of experience reading chest computed tomographic scans with specific regard to identifying and classifying abnormalities of parenchymal pulmonary nodules. IPNs were defined as persistent ground glass nodules smaller than 10 mm in diameter and solid nodules smaller than 20 mm in diameter with nonspiculated contours, no air bronchogram or pseudocavitation, no malignant-type calcification, and no intralesional fat or benign-type calcification.24 Postoperative adjuvant chemotherapy was dependent on the general condition and treatment compliance of the patient and the www.annalsofsurgery.com | 1145
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 261, Number 6, June 2015
Kim et al
synchronous and metachronous nodules and analyzed them separately.
Statistical Analyses
FIGURE 1. Flowchart showing patient recruitment. preference of the treating physician. The standard protocol for adjuvant therapy in patients with histologically proven high-risk stage II or stage III colon cancer included 5-fluorouracil (FU) plus leucovorin (LV) [5-FU/LV; 6 cycles of a monthly bolus of intravenous FU (400–425 mg·m−2 ·d−1 ) on days 1–5 and LV calcium (20 mg/m2 /d) on days 1–5] and FU/LV plus oxaliplatin [FOLFOX; 12 cycles of oxaliplatin (85 mg/m2 ) on day 1, LV calcium (200 mg/m2 ) as a 2hour infusion on day 1, and FU (400 mg/m2 ) as a bolus and as a 600 mg/m2 22-hour infusion on days 1 and 2 bimonthly, respectively]. Of the 123 patients with colon cancer, 59 (47.9%) received chemotherapy (25; 5-FU/LV, 34; FOLFOX). Although preoperative concurrent chemoradiation therapy (CRT) (n = 69) was our standard protocol for patients with locally advanced (radiologic T3/T4 or N±) rectal cancer, postoperative CRT was provided after pathologic confirmation for 16 patients. Radiotherapy consisted of a total of 5040 cGy delivered in 28 fractions. During the first and fifth weeks of radiotherapy, 5-FU/LV was administered identically to the pre- and postoperative CRT groups. FOLFOX chemotherapy was also used to treat tumors located in the upper rectum in 23 patients who had not received any CRT. Follow-up chest computed tomographic scans and abdominal computed tomographic scans were obtained at 3, 6, 9, 12, 18, and 24 months. Of the 263 patients who were eligible for final analysis, 199 (75.7%) completed this follow-up program. All scans were obtained on a 64-MDCT system (Discovery HD 750, GE Healthcare, Milwaukee, WI). After the RECIST criteria requirements, the largest nodule was measured along its longest axis using electronic calipers. During follow-up, progression to metastasis was defined according to RECIST criteria or by a histological diagnosis of metastatic cancer.25 Metachronous IPNs were defined as any nodules that newly appeared in the follow-up computed tomographic scans without growth of any previously identified IPNs. We identified 18 patients who had both 1146 | www.annalsofsurgery.com
Categorical variables are presented as numbers and percentages and continuous variables as medians and interquartile ranges (IQR). The significance of baseline differences was tested by the Mann-Whitney U test, Fisher exact test, or the χ 2 test, as appropriate. To evaluate the independent influence of prognostic factors on progression of IPN to definitive pulmonary metastasis, we used Cox proportional hazards regression to estimate hazard ratios and associated 95% confidence intervals. Variables that were significant at P ≤ 0.1 on univariate analysis were included as covariates in the multivariate model. After excluding variables that had low prevalence or exhibited multicolinearity, 12 variables exhibited prognostic significance by univariate analysis and were entered into the multivariate analysis: tumor location, T-stage, N-stage, perineural invasion, adjuvant chemotherapy, extrapulmonary disease, preoperative carcinoembryonic antigen (CEA), CEA at diagnosis of IPN, number of IPNs, chronicity, bilaterality, and configuration. Both CEA parameters were divided into 2 groups on the basis of a cutoff value of 5 ng/mL. Time to progression was calculated as the time from the baseline computed tomographic scan to the date of confirmation based on follow-up chest CT and/or histological biopsy. Progression-free survival was calculated by the Kaplan-Meier method, and differences in survival between groups were assessed by the log-rank test. The life table method was used to estimate the conditional hazard rate for progression to definite pulmonary metastasis. The conditional hazard rate for progression was defined as the probability of developing definitive pulmonary metastasis in a 3-month interval in patients who were free of pulmonary metastasis at the beginning of each interval. To develop a practical prognostic score, we assigned a risk score on the basis of factors identified by multivariate analysis with weighted points proportional to the β regression coefficient values (rounded to the nearest integer after multiplying by 2). A total progression risk score (PRS) for each patient was then calculated by summing the risk points of each prognostic factor. Based on the total PRS, 4 risk groups were defined using receiver operating characteristic curve analysis. The predictive discrimination of the scoring system was examined by calculating the concordance index (C-index).26 A Harrell’s C-index of 0.5 indicates no predictive discrimination and 1.0 indicates perfect separation of patients. The predictive model was validated in 2 different ways. First, the Cox multivariate models were internally validated using bootstrapping (1000 replications). Second, we selected a total of 115 patients who satisfied aforementioned inclusion criteria between January 2006 and December 2007 for the external validation. Statistical significance was set as P < 0.05 in a 2-tailed test. Statistical analyses were performed using SPSS version 20 (SPSS, Chicago, IL) or R 2.15.2.
RESULTS Patients Demographics Of the 1195 patients who had obtained staging computed tomographic scan, 326 (27.2%) had IPNs. Thirty five patients were considered to have a malignant nodule in the initial chest CT. Among them, 11 patients underwent video-assisted thoracoscopic surgery wedge resection. Two of the 11 cases were diagnosed as benign nodules and the other 9 as pulmonary metastases. Another 24 patients were diagnosed by radiology and showed metastasis in the serial chest CT (Fig. 1). Patient demographics for development set (n = 263) and validation set (n = 115) are listed in Table 1. There was a significantly higher rate of rectal cancer and stage III and stage IV diseases at C 2014 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 261, Number 6, June 2015
diagnosis among the patients with progression to pulmonary metastasis in both groups. Moreover, progression developed more frequently in patients who had perineural invasion (P < 0.001), higher preoperative CEA level (P < 0.001), and extrapulmonary distant metastatic disease (P < 0.001).
Pulmonary Nodule Analysis on CT The relationship between nodule character and progression for development set (n = 263) and validation set (n = 115) is summarized in Table 2. Of the 263 patients in the development set, baseline CT detected 438 synchronous IPNs in 194 patients and 118 metachronous IPNs in 69 patients. The median number of IPNs was 1 (IQR: 1–3) in the synchronous nodule and 1 (IQR: 1–2) in the metachronous nodule. One hundred forty-eight (56.3%) patients had a single nodule, 50 (19%) had 2 nodules, and 65 (24.7%) had more than 2 nodules. One hundred eighty-three patients (69.6%) had nodules, with a maximum size of less than 5 mm. Sixty-nine (26.2%) nodules ranged in size from 0.5 mm to 1 cm, with 11 (4.2%) nodules exceeding 1 cm. The majority of the IPNs (234/263, 89%) had smooth margins and 95.4% were solid nodules. Multiple (P = 0.008), metachronous (P < 0.001), and bilaterally located (P = 0.023) IPNs showed a significantly increased incidence of progression to malignant nodules. In contrast to development set, the size of IPNs showed more significant relationship than number of IPNs in the validation set.
Lung Nodules in Colorectal Cancer
Progression to Pulmonary Metastasis Median follow-up for development set was 26.7 months (IQR: 18.0–37.2). During the study period, a total of 1993 chest CT scans were obtained on 263 patients who had a regular follow-up, with a median of 7 (IQR: 5–9) chest CT scans for each patient and 74 (28.1%) patients progressed to pulmonary metastases. Surgical resections of IPNs were performed in 42 patients. Of these, 34 (81%) were histologically confirmed to have metastatic CRC and the remaining 8 patients had the following benign histologic features: 3 had chronic granuloma, 2 had organizing pneumonia, 2 had benign fibrosis, and 1 had focal emphysema. Another 40 patients showed definitive pulmonary metastasis in the follow-up chest CT. Bilateral multiple pulmonary metastases accompanying the size increase of IPNs and/or radiologic malignant change which include a spiculated contours, air bronchogram or pseudocavitation, and any malignant-type calcification, of index IPNs were observed in 36 patients and in other 4 patients, the size of IPNs increased to larger than 3cm. IPNs with progression were mostly confirmed in the second (IQR: 1–4) follow-up computed tomographic scan; IPNs occurred in 59 (79.7%) patients within 1 year after the baseline CT and in 70 (94.6%) patients within 2 years (Fig. 2). Only in 1.5% (4 of 263) of patients, the progression developed after 2 years since its diagnosis. To evaluate whether chemotherapy can inhibit the growth of metastatic nodules in patients with synchronous IPNs, patients were compared according to the treatment regimen they received (Table 3). Patients who underwent simultaneous resection of IPNs with their primary CRC (n = 7, of which 4 exhibited metastasis)
TABLE 1. Clinicopathologic Factors of Patients With Progression and No Progression Development Set Progression (N = 74) Age Sex Male Female Location of primary tumor Colon Rectum Pathologic TNM stage 0–1 2 3 4 Differentiation Well-differentiated Moderate Poor Lymphovascular invasion Absent Present PNI Absent Present Preoperative CEA Primary tumor size Extrapulmonary disease Absent Present Adjuvant chemotherapy No Yes CEA at diagnosis of IPN, ng/mL
66.0 (57–72)
No Progression (N = 189) 67.0 (59–74)
45 (60.8%) 29 (39.2%)
127 (67.25%) 62 (32.8%)
19 (25.7%) 55 (74.3%)
104 (55.0%) 85 (45.0%)
8 (10.8%) 16 (21.6%) 37 (50.0%) 13 (17.6%)
56 (29.6%) 75 (39.7%) 54 (28.6%) 4 (2.1%)
23 (31.1%) 48 (64.9%) 3 (4.0%)
82 (43.4%) 97 (51.3%) 10 (5.3%)
57 (77.0%) 17 (23.0%)
152 (80.4%) 37 (19.6%)
24 (32.4%) 50 (67.6%) 6.3 (3.2–14.6) 4.0 (2.5–5.5)
138 (73.0%) 51 (27.0%) 3.7 (1.9–6.8) 4.0 (3.0–5.0)
53 (71.6%) 21 (28.4%)
181(95.8%) 8 (4.2%)
20 (27.0%) 54 (63.0%) 3.9 (1.9–12.3)
90 (47.6%) 99 (52.4%) 2.9 (1.7–4.9)
Validation Set P 0.236 0.328
