Abstract
OBJECTIVES
Pathological vessel invasion is a well-known prognostic factor in early-stage, non-small cell lung cancer and preoperative predicting vessel invasion may enable us to improve prognosis by additional interventions. We evaluated the importance of vessel invasion as a prognostic factor in clinical stage IA non-small cell lung cancer and predictive performance of simple diameter-based computed tomography image analysis for vessel invasion.
METHODS
The study design was retrospective, and we reviewed 398 patients who underwent surgical resection of clinical stage IA non-small cell lung cancer from 1999 to 2009. The prognostic factors for recurrence-free survival were examined by univariate and multivariate analyses. Additionally, we analyzed preoperative high-resolution computed tomography images of patients with adenocarcinoma. The greatest diameter of the tumor in the lung window and the length of the consolidation part of L in the mediastinal window were measured. Then the ratio (mediastinal window/lung window) was calculated, and the correlation between the ratio (mediastinal window/lung window) and vessel invasion was analyzed by receiver operating characteristic analysis.
RESULTS
Sixty-eight recurrences occurred. Multivariate analysis revealed that vessel invasion, high preoperative serum carcinoembryonic antigen, and history of other malignancy were independent prognostic factors; their hazard ratios were 2.98, 2.45, and 1.98, respectively. The receiver operating characteristic analysis showed that the area under the curve was 0.75. When we set the cut-off value of the ratio (mediastinal window/lung window) at 0.67, the sensitivity and specificity were 75% and 72%, respectively.
CONCLUSIONS
Vessel invasion had the greatest impact on recurrence in clinical stage IA non-small cell lung cancer. Our simple computed tomography image analysis showed good predictive performance for vessel invasion.
Keywords: Lung cancer, Diagnosis, Pathology, Computed tomography
INTRODUCTION
The 5-year survival rate of clinical and pathological stage IA non-small cell lung carcinoma (NSCLC) is 50 and 73%, respectively [1]. The high recurrence rate of NSCLC, even in the early stage, accounts for this poor outcome. As a prognostic factor in stage IA NSCLC, pathological vessel invasion is well recognized [2–4]. If we could predict vessel invasion preoperatively, we may be able to identify the high-risk group for recurrence before surgery. Some studies have shown that preoperative computed tomography (CT) findings reflect vessel invasion [5–11]: tumours with a smaller proportion of ground-glass opacity tend to have a greater degree of vessel invasion. Therefore, in this study, we evaluated the importance of pathological vessel invasion as a prognostic factor for the recurrence in clinical stage IA NSCLC and the predictive performance of a high-resolution CT (HRCT) image analysis for vessel invasion when using a receiver operating characteristic (ROC) analysis. In terms of HRCT image analysis, we emphasized simplicity, objectiveness and consideration of the distortion characteristic of the tumour.
MATERIALS AND METHODS
Three hundred and ninety-eight patients diagnosed as clinical stage IA NSCLC who underwent elective surgery in our hospital during the period between May 1999 and December 2009 were enrolled. Lung tumours diagnosed as metastases from other primary sites by pathological examination (i.e. immunostaining, absence of bronchioloalveolar carcinoma component, etc.) were excluded. The median follow-up time was 30 months (range: 0–129 months). The clinical and pathological staging was based on the 7th edition of the TNM classification of the Union for International Cancer Control. We reviewed the clinical records of all of these patients. Before the study, the Research Review Board at our institution examined and approved the research protocol in accordance with the Declaration of Helsinki. All patients provided their written informed consent for review of their medical charts before surgery.
Resected specimens were examined by haematoxylin-and-eosin and Elastica van Gieson stains and then reviewed by pathologists to evaluate the histological type, the extent of lymph node involvement and the existence of vessel invasion. If there were tumour invasions into lymphatic vessels and/or blood vessels, vessel invasion was considered to be present.
Sub-analysis materials and methods
Patients with an adenocarcinoma were enrolled for sub-analysis. We used thin-slice HRCT images (2.5- or 3.0-mm section thickness) taken just before surgery. HRCT images taken in our hospital were all preserved in Digital Imaging and Communications in Medicine format and reviewed on our hospital computing systems.
The protocol for HRCT images analysis is as follows:
Select a slice that maximizes the tumour diameter in the lung window (window level: −600 Hounsfield Unit (HU) and window width: 1500 HU).
The maximum diameter is defined as ‘lung diameter (L)’.
On the same section, change the window setting from the lung window to mediastinal window (window level: 40 HU and window width: 250 HU).
The length of the consolidation part of L is defined as ‘mediastinal diameter (M)’.
Calculate the ratio of M to L (M/L).
An example of the way to measure M and L is shown in Fig. 1. These measurements were performed by two independent blinded observers (A.T. and M.M.), and a mean value of M/L ratio was used for the ROC curve analysis. In this protocol, we defined consolidation as the area that could be recognized in the mediastinal window and ground-glass opacity as the area that disappeared as the CT window setting was changed from the lung window to mediastinal window.
Figure 1:
Definition of M and L. (a) The dotted line is L, the greatest diameter of tumour in the lung window. (b) M is the length of the consolidation part of the dotted line (L) in the mediastinal window.
Statistical analysis
We defined the endpoint as recurrence or metastasis. The t-test and χ2 test were used to compare the proportions between the groups. In univariate analyses, the recurrence-free survival rate was calculated by the Kaplan–Meier method, and the survival curves were compared between the groups with the log-rank and Wilcoxon tests. The following grouping factors were adopted: age, sex, smoking history, preoperative serum carcinoembryonic antigen (CEA) level, types of surgical procedures undergone, preoperative comorbidities [diabetes mellitus (DM), interstitial pneumonia, autoimmune disease, history of other malignancies, peripheral vessel disease, neurological disease, chronic kidney disease, obstructive pulmonary disorder and restrictive pulmonary disorder] and vessel invasion. Variables that showed a statistically significant difference in the univariate analyses were entered into multivariate analysis. In the multivariate analysis, the Cox proportional hazard regression with the stepwise method was used to identify the independent prognostic factors for recurrence and to estimate their hazard ratios (HZs) with a 95% confidence interval (CI). In the sub-analysis, an inter observer agreement of M/L ratio was quantified as the intra-class coefficient. The predictive performance of M/L ratio for vessel invasion was evaluated by the ROC analysis, and the area under the curve (AUC) was calculated. In all tests, a P-value of <0.05 was considered statistically significant. All statistical analyses were performed with Stata 11 (StataCorp LP, TX, USA).
RESULTS
Clinical and pathological patient characteristics
The clinical and pathological characteristics of all patients are summarized in Table 1. This study included 398 patients with a median age of 68. The preoperative level of CEA was elevated (>5.0 ng/ml) in 127 patients (33.9%). An adenocarcinoma was the most common [262 patients (65.8%)]. The most frequently performed procedure was lobectomy with a systemic node dissection [280 patients (70.4%)]. The most common pathological stage was IA (74.6%). Vessel invasion was seen in 108 patients (27.1%). There were 68 recurrences or metastases (17.1%).
Table 1:
Clinical and pathological characteristics of patients
| Factors | n (%) | Mean ± SD (range) | Median |
|---|---|---|---|
| Total patients | 398 | ||
| Observation time (months) | 41.1 ± 33.0 (0–129) | 30 | |
| Age (years) | 66.6 ± 10.3 (17–85) | 68 | |
| Sex | |||
| Male | 238 (59.8) | ||
| Female | 160 (40.2) | ||
| Smoking history | |||
| + | 248 (62.5) | ||
| − | 149 (37.5) | ||
| Brinkman index | 630.8 ± 747.2 (0–4200) | 400 | |
| Carcinoembryonic antigen (ng/ml) | 5.89 ± 8.78 (0.5–122.8) | 3.7 | |
| ≤5 | 248 (66.1) | ||
| >5 | 127 (33.9) | ||
| Surgical procedure | |||
| Wedge resection | 100 (25.1) | ||
| Segmentectomy | 17 (4.3) | ||
| Lobectomy | 280 (70.4) | ||
| Pneumonectomy | 1 (0.2) | ||
| Preoperative comorbidities | |||
| History of other malignanciesa | 136 (34.2) | ||
| Diabetes mellitus | 51 (14.0) | ||
| Neurological disease | 29 (7.3) | ||
| Autoimmune disease | 18 (4.5) | ||
| Peripheral vessel disease | 18 (4.5) | ||
| Interstitial pneumonia | 15 (3.8) | ||
| Chronic kidney disease | 6 (1.5) | ||
| Obstructive pulmonary disorder | 147 (37.0) | ||
| Restrictive pulmonary disorder | 23 (5.8) | ||
| Recurrence | |||
| + | 68 (17.1) | ||
| − | 330 (82.9) | ||
| Vessel invasion | |||
| + | 108 (27.1) | ||
| − | 290 (72.9) | ||
| Histological type | |||
| Adenocarcinoma | 262 (65.8) | ||
| Squamous cell carcinoma | 53 (13.3) | ||
| Bronchioloalveolar carcinoma | 62 (15.6) | ||
| Others | 21 (5.3) | ||
| Pathological stage | |||
| IA | 297 (74.6) | ||
| IB | 54 (13.6) | ||
| IIA | 8 (2.0) | ||
| IIB | 4 (1.0) | ||
| IIIA | 21 (5.3) | ||
| IIIB | 12 (3.0) | ||
| IV | 2 (0.5) | ||
Diabetes mellitus (DM): HbA1c <6.1%, chronic kidney disease: creatinine >2 mg/dl, obstructive pulmonary disorder: FEV1/FVC <70% (FEV1: forced expiratory volume in 1 s; FVC: forced vital capacity), restrictive pulmonary disorder: %VC <80% (%VC: vital capacity as percent of predicted).
aOther types of malignancies [type (n)]: colon cancer (35), gastric cancer (31), lung cancer (19), hepatic cell carcinoma (19), breast cancer (11), oesophageal cancer (9), prostatic cancer (9), renal cancer (8), bladder cancer (4), ovarian tumour (4), thyroid cancer (3), uterine cervical cancer (3), laryngeal cancer (3), uterine body cancer (2), bile duct cancer (1), duodenum cancer (1), intraductal papillary mucinous tumour (1), malignant lymphoma (1), parotid tumour (1), pharyngeal cancer (1), seminoma (1), spinal tumour (1), thymoma (1) and tongue cancer (1).
Prognostic factor analysis
The results of the univariate analyses are summarized in Table 2. As statistically significant recurrence prognostic factors, the univariate analyses detected sex (log-rank test: P = 0.048), high level of preoperative CEA (P < 0.001), history of other malignancies (P = 0.006), interstitial pneumonia (P = 0.003) and vessel invasion (P < 0.001). Figure 2 shows the recurrence-free survival curves with and without vessel invasion. The 5-year recurrence-free survival rates of patients with and without vessel invasion were 84 and 58%, respectively. We chose the five prognostic factors detected in the univariate analyses as covariates in the multivariate analysis. The multivariate analysis revealed that vessel invasion, high level of preoperative CEA and history of other malignancies were independent prognostic factors; the HR was 2.98, 2.45 and 1.94, respectively (Table 3).
Table 2:
Results of univariate analyses of recurrence-free survival
| Variables |
P-value |
|
|---|---|---|
| Log-rank test | Wilcoxon test | |
| Age (<67 or ≥67) | 0.419 | 0.882 |
| Sex (male/female) | 0.048 | 0.044 |
| Smoking history (+/−) | 0.154 | 0.104 |
| Carcinoembryonic antigen (<5 or ≥5 ng/ml) | <0.001 | <0.001 |
| Surgical procedure | ||
| Lobectomy or others | 0.217 | 0.237 |
| Preoperative comorbidities | ||
| Obstructive pulmonary disorder (+/−) | 0.092 | 0.094 |
| History of other malignancies (+/−) | 0.006 | 0.037 |
| Diabetes mellitus (+/−) | 0.165 | 0.295 |
| Neurological disease (+/−) | 0.125 | 0.105 |
| Restrictive respiratory disorder (+/−) | 0.897 | 0.887 |
| Autoimmune disease (+/−) | 0.435 | 0.725 |
| Peripheral vessel disease (+/−) | 0.346 | 0.543 |
| Interstitial pneumonia (+/−) | 0.003 | 0.018 |
| Chronic kidney disease (+/−) | 0.264 | 0.312 |
| Histological type | ||
| Adenocarcinoma or others | 0.509 | 0.757 |
| BAC or others | 0.054 | 0.054 |
| SCC or others | 0.167 | 0.101 |
| Vessel invasion (+/−) | <0.001 | <0.001 |
BAC: bronchioloalveolar carcinoma; SCC: squamous cell carcinoma.
Figure 2:
Recurrence-free survival curves for all patients. There was a statistically significant difference (P < 0.001) between the groups with and without vessel invasion.
Table 3:
Results of the multivariate Cox proportional hazard analysis of recurrence-free survival
| Variables | Hazard ratio | 95% CI | P-value |
|---|---|---|---|
| History of other malignancies (+) | 1.94 | 1.18–3.18 | 0.009 |
| CEA >5 ng/ml | 2.45 | 1.49–4.03 | <0.001 |
| Vessel invasion (+) | 2.98 | 1.82–4.88 | <0.001 |
CAE: carcinoembryonic antigen.
sub-analysis of the correlation between preoperative high-resolution computed tomography findings and vessel invasion in patients with an adenocarcinoma
In the sub-analysis, 182 patients with an adenocarcinoma were enrolled. Eighty patients with an adenocarcinoma were excluded because their preoperative HRCT scans were performed in other hospitals and there were no appropriate HRCT images that we could analyse in our computing system.
In measurements of M/L ratio, a good inter-observer agreement (intra-class coefficient: 0.82, 95% CI: 0.77–0.86) was obtained. The results of the ROC analysis are shown in Fig. 3. The AUC was 0.75. We adopted 0.67 as the cutoff value for M/L ratio because this point maximized the ‘sensitivity false-positive rate’. At this point, the sensitivity and specificity was 75 and 72%, respectively. Figure 4 shows the recurrence-free survival curves for the groups with M/L ratio 0.67 or more and less than 0.67. There was a significant difference between the two curves (P = 0.004).
Figure 3:
ROC curve of the predictive performance of M/L ratio for vessel invasion. The AUC was 0.75 (95% CI: 0.68–0.83).
Figure 4:
Recurrence-free survival curves of the groups with an M/L ratio of ≥0.67 and <0.67. There was a significant difference (P = 0.004) between the survival curves.
DISCUSSION
Pathological vessel invasion showed the greatest impact on prognosis of recurrence in clinical stage IA NSCLC, as reported previously [2–4]. This study also identified the high level of preoperative CEA and the history of other malignancies as independent prognostic factors for recurrence. Elevated CEA may reflect the invasiveness of the tumour, as shown by some previous studies [10, 12]. In contrast, few reports detected the history of other malignancies as a prognostic factor. It may be because the history of other malignancies was not initially adopted as a variable for the analysis in most of the studies.
In the sub-analysis, we defined the population as patients with an adenocarcinoma, because we thought that if several histological types were involved, the correlation between CT findings and vessel invasion might become vague. Preliminarily performed prognostic factor analyses limited to patients with an adenocarcinoma also confirmed that vessel invasion was the strongest prognostic factor (HR: 2.87, 95% CI: 1.63–5.03).
One-dimensional analysis using diameters as indicators of the size of consolidation and the tumour has a great advantage in terms of simplicity. Austin et al. [13] defined ground-glass opacity as the area showing hazy, increasing attenuation with the preservation of bronchial and vascular margins. Some researchers have pointed out that this visually-based definition is not objective [9]. To obtain objectivity, we changed the CT window to predetermined setting to distinguish between ground-glass opacity and consolidation, as reported previously [7, 9, 10, 14], and the good inter-observer agreement of M/L ratio was obtained in the study. To our knowledge, there are few studies in which variability in the location of consolidation was considered and quantified. Consolidation is not always located in the centre of the tumour, and ground-glass opacity also does not always distribute equally around consolidation. We adopted the definition of M to reflect the difference in the position of consolidation on L. If consolidation lies distant from L, M tends to become smaller.
For one-dimensional analysis, Ohde et al. [8] used Cdmax/Td (Cdmax: the greatest diameter of consolidation of all slices; Td: the greatest diameter of the tumour on a particular slice) to predict vessel invasion (without changing the window setting). With this approach, the AUC was calculated to be ∼0.71 (calculated on the basis of given data). M/L ratio may not be so different from Cdmax/Td in terms of the predictive performance, and Cdmax was possibly approximated with M.
Tateishi et al. [11] performed volumetric analysis and examined the predictive performance of the volume occupied by non-solid components for vessel invasion. The AUC was 0.93 (95% CI: 0.90–0.96). Because simplicity is a great advantage of one-dimensional analysis, we believe that the one-dimensional approach could be acceptable in clinical practice at this moment. If the future innovation in computing technology will provide us with an easier access to three-dimensional analysis, the volumetric approach will become the mainstream in terms of accuracy.
Preoperative tumour biopsy and pathological evaluation are still the gold standard in outlining individualized treatment and management with a specific tumour diagnosis. However, other additional therapy may be necessary to improve the prognosis of patients with vessel invasion. M/L ratio may represent one criterion for determining whether the patients should undergo additional intervention preoperatively.
Limitations of this study
In the sub-analysis, 80 patients without HRCT images were excluded. There was no statistically significant difference between the recurrence-free survival curves between the groups with and without HRCT images (P = 0.759) or the distribution of vessel invasion in both groups (P = 0.196). However, there was a significant difference in the mean follow-up time (37.6 vs. 58.9 months, P < 0.001). In this study population, patients who underwent surgery early in the observation period tend not to undergo HRCT at our hospital; this would affect the difference in the mean follow-up time. There is a possibility that and this difference in the follow-up time may have produced selection bias.
The window settings used in this study are preset in our hospital's computing system. Although these values are not much different from the values used in other studies, the most appropriate window settings remain a matter of research.
We focused on the distortion characteristics of the tumour and defined M as described, but further examination will be required to determine how much distortion characteristics contribute to the predictive performance. In the groups with similar values of M/L ratio, examining the distances between the centre of the tumour and the centre of consolidation as well as analysing the correlation between the distance and vessel invasion may help us to quantify the contributing rate of distortion characteristics. In this analysis, the population of this study was small and a much larger study population is needed for a more statistically accurate analysis.
Conflicts of interest: none declared.
REFERENCES
- 1.Goldstraw P, Crowley J, Chansky K, Giroux DJ, Groome PA, Rami-Porta R, et al. The IASLC Lung Cancer Staging Project: proposals for the revision of the TNM stage groupings in the forthcoming (seventh) edition of the TNM Classification of malignant tumours. J Thorac Oncol. 2007;2:706–14. doi: 10.1097/JTO.0b013e31812f3c1a. doi:10.1097/JTO.0b013e31812f3c1a. [DOI] [PubMed] [Google Scholar]
- 2.Tsuchiya T, Akamine S, Muraoka M, Kamohara R, Tsuji K, Urabe S, et al. Stage IA non-small cell lung cancer: vessel invasion is a poor prognostic factor and a new target of adjuvant chemotherapy. Lung Cancer. 2007;56:341–8. doi: 10.1016/j.lungcan.2007.01.019. doi:10.1016/j.lungcan.2007.01.019. [DOI] [PubMed] [Google Scholar]
- 3.Miyoshi K, Moriyama S, Kunitomo T, Nawa S. Prognostic impact of intratumoral vessel invasion in completely resected pathologic stage I non-small cell lung cancer. J Thorac Cardiovasc Surg. 2009;137:429–34. doi: 10.1016/j.jtcvs.2008.07.007. doi:10.1016/j.jtcvs.2008.07.007. [DOI] [PubMed] [Google Scholar]
- 4.Shoji F, Haro A, Yoshida T, Ito K, Morodomi Y, Yano T, et al. Prognostic significance of intratumoral blood vessel invasion in pathologic stage IA non-small cell lung cancer. Ann Thorac Surg. 2010;89:864–9. doi: 10.1016/j.athoracsur.2009.09.047. doi:10.1016/j.athoracsur.2009.09.047. [DOI] [PubMed] [Google Scholar]
- 5.Aoki T, Tomoda Y, Watanabe H, Nakata H, Kasai T, Hashimoto H, et al. Peripheral lung adenocarcinoma: correlation of thin-section CT findings with histologic prognostic factors and survival. Radiology. 2001;220:803–9. doi: 10.1148/radiol.2203001701. doi:10.1148/radiol.2203001701. [DOI] [PubMed] [Google Scholar]
- 6.Suzuki K, Asamura H, Kusumoto M, Kondo H, Tsuchiya R. ‘Early’ peripheral lung cancer: prognostic significance of ground glass opacity on thin-section computed tomographic scan. Ann Thorac Surg. 2002;74:1365–9. doi: 10.1016/s0003-4975(02)03895-x. [DOI] [PubMed] [Google Scholar]
- 7.Kondo T, Yamada K, Noda K, Nakayama H, Kameda Y. Radiologic-prognostic correlation in patients with small pulmonary adenocarcinomas. Lung Cancer. 2002;36:49–57. doi: 10.1016/s0169-5002(01)00448-2. doi:10.1016/S0169-5002(01)00448-2. [DOI] [PubMed] [Google Scholar]
- 8.Ohde Y, Nagai K, Yoshida J, Nishimura M, Takahashi K, Suzuki K, et al. The proportion of consolidation to ground-glass opacity on high resolution CT is a good predictor for distinguishing the population of non-invasive peripheral adenocarcinoma. Lung Cancer. 2003;42:303–10. doi: 10.1016/j.lungcan.2003.07.001. doi:10.1016/j.lungcan.2003.07.001. [DOI] [PubMed] [Google Scholar]
- 9.Matsuguma H, Nakahara R, Anraku M, Kondo T, Tsuura Y, Kamiyama Y, et al. Objective definition and measurement method of ground-glass opacity for planning limited resection in patients with clinical stage IA adenocarcinoma of the lung. Eur J Cardiothorac Surg. 2004;25:1102–6. doi: 10.1016/j.ejcts.2004.02.004. doi:10.1016/j.ejcts.2004.02.004. [DOI] [PubMed] [Google Scholar]
- 10.Takamochi K, Yoshida J, Nishimura M, Yokose T, Sasaki S, Nishiwaki Y, et al. Prognosis and histologic features of small pulmonary adenocarcinoma based on serum carcinoembryonic antigen level and computed tomographic findings. Eur J Cardiothorac Surg. 2004;25:877–83. doi: 10.1016/j.ejcts.2004.01.049. doi:10.1016/j.ejcts.2004.01.049. [DOI] [PubMed] [Google Scholar]
- 11.Tateishi U, Uno H, Yonemori K, Satake M, Takeuchi M, Arai Y. Prediction of lung adenocarcinoma without vessel invasion: a CT scan volumetric analysis. Chest. 2005;128:3276–83. doi: 10.1378/chest.128.5.3276. doi:10.1378/chest.128.5.3276. [DOI] [PubMed] [Google Scholar]
- 12.Tomita M, Matsuzaki Y, Edagawa M, Shimizu T, Hara M, Onitsuka T. Prognostic significance of preoperative serum carcinoembryonic antigen level in lung adenocarcinoma but not squamous cell carcinoma. Ann Thorac Cardiovasc Surg. 2004;10:76–80. [PubMed] [Google Scholar]
- 13.Austin JH, Müller NL, Friedman PJ, Hansell DM, Naidich DP, Remy-Jardin M, et al. Glossary of terms for CT of the lungs: recommendations of the nonmenclature committee of the Fleischner. Radiology. 1996;200:327–31. doi: 10.1148/radiology.200.2.8685321. [DOI] [PubMed] [Google Scholar]
- 14.Sakao Y, Nakazono T, Tomimitsu S, Takeda Y, Sakuragi T, Natsuaki M, et al. Lung adenocarcinoma can be subtyped according to tumor dimension by computed tomography mediastinal-window setting. Additional size criteria for clinical T1 adenocarcinoma. Eur J Cardiothorac Surg. 2004;26:1211–5. doi: 10.1016/j.ejcts.2004.08.021. doi:10.1016/j.ejcts.2004.08.021. [DOI] [PubMed] [Google Scholar]