Clinical Radiology 69 (2014) 559e566

Contents lists available at ScienceDirect

Clinical Radiology journal homepage: www.clinicalradiologyonline.net

Relationship between CT features and high preoperative serum carcinoembryonic antigen levels in early-stage lung adenocarcinoma M. Yamazaki*, H. Ishikawa, R. Kunii, A. Tasaki, S. Sato, Y. Ikeda, N. Yoshimura, H. Aoyama Department of Radiology, Niigata University Graduate School of Medical and Dental Sciences, Japan

art icl e i nformat ion Article history: Received 8 August 2013 Received in revised form 5 December 2013 Accepted 12 December 2013

AIM: To assess the relationship between thin-section computed tomography (CT) features of primary tumour and high preoperative serum carcinoembryonic antigen (CEA) levels that reportedly suggest poor prognoses in early-stage lung adenocarcinoma. MATERIALS AND METHODS: Two hundred and seventy-five consecutive patients who underwent resection of pathological stage I (T1e2aN0M0) adenocarcinomas with a maximum diameter of 3 cm (144 men, 131 women; mean age 67.8 years) were enrolled. CT features of the primary tumours and clinical characteristics of these patients were statistically evaluated to identify the factors associated with high serum CEA levels (>5 ng/ml). RESULTS: Eighty-one patients (29.5%) had high serum CEA levels. In univariate analysis, lower ground-glass opacity ratio (p < 0.001), lower tumour shadow disappearance rate (TDR: the ratio of tumour area in mediastinal window to that of lung window, p < 0.001), presence of notch (p ¼ 0.015), and coexistence with bullae or honeycomb cysts (p < 0.001) were observed more frequently in the group with high serum CEA levels than that of the group with normal levels. TDR [odds ratio (OR) 0.984; 95% confidence interval (CI): 0.976e0.993; p < 0.001] and coexistence with bullae or honeycomb cysts (OR ¼ 3.08; 95% CI: 1.55e6.12; p ¼ 0.001) remained significant, even after adjusting patients’ age, gender, and smoking status. CONCLUSIONS: Adenocarcinomas with lower TDR and coexisting with bullae or honeycomb cysts are associated with high preoperative serum CEA levels. Although some CEA elevations may be due to benign pulmonary diseases, such tumours are suspected to have poor prognoses, even for early-stage diseases. Ó 2013 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.

Introduction

* Guarantor and correspondent: M. Yamazaki, Department of Radiology, Niigata University Graduate School of Medical and Dental Sciences, 757, Ichiban-cho, Asahimachi-dori, Chuou-ku, Niigata-shi, Niigata-ken 951-8510, Japan. Tel.: þ81 25 227 2315; fax: þ81 25 227 0788. E-mail addresses: [email protected], [email protected] (M. Yamazaki).

Preoperative serum carcinoembryonic antigen (CEA) levels have been reported to be an independent prognostic factor in primary non-small-cell lung cancer (NSCLC),1e6 with high preoperative serum CEA levels even in the early stages indicating poor prognosis. Okada et al.4 analysed 1000 patients with clinical stage I NSCLC and found that the 5-year survival rates were 75.2% and 53.8% for patients with normal and high preoperative serum CEA levels,

0009-9260/$ e see front matter Ó 2013 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.crad.2013.12.009

560

M. Yamazaki et al. / Clinical Radiology 69 (2014) 559e566

respectively. Matsuguma et al.1 analysed 455 patients with pathological stage I (p-stage I) NSCLC, and showed that high preoperative serum CEA level was a poor prognostic factor, independent of pathological status such as blood vessel and visceral pleural invasion.1 It is also generally accepted that thin-section computed tomography (CT) features in primary lung adenocarcinoma are associated with prognosis. Tumours with ground-glass opacity (GGO) ratios greater than 50% indicate favourable prognosis, with a 5-year survival rate of nearly 100%.7e10 Tumour shadow disappearance rate (TDR), defined as the ratio of tumour area of the mediastinal window to that of the lung window, has also been reported to be linked to survival and the risk of metastasis.11,12 In addition, the presence of spiculation,13 notch,14 and absence of air bronchogram15 have also been reported to be poor prognostic CT features. However, previous studies regarding high serum CEA levels as an independent poor prognostic factor did not evaluate these CT features in detail. Therefore, although CT features and preoperative serum CEA levels are both considered to be associated with prognosis, these relationships have not been well examined in the past. In addition, the CT features of early-stage lung adenocarcinoma with high preoperative serum CEA levels remain unclear. Consequently, the aim of the present study was to assess the relationship between thin-section CT features of primary tumour and high preoperative serum CEA levels that reportedly suggest poor prognoses in patients with earlystage lung adenocarcinoma.

Materials and methods Study group This retrospective study was approved by the institutional review board with a waiver of informed consent. Between April 2000 and April 2011, 354 patients underwent pulmonary resection for p-stage I (T1e2aN0M0) lung adenocarcinoma at Niigata University Medical and Dental Hospital. Among them, patients with a history of any other malignancy (n ¼ 9) were excluded, because the aim was to evaluate the elevation of preoperative serum CEA levels due to primary lung lesions. Furthermore, patients who did not undergo preoperative CT at Niigata University Hospital (n ¼ 25) were also excluded. Patients with a maximum tumour diameter (maxD) >3 cm (n ¼ 45) were also excluded because large tumours make it difficult to correctly evaluate CT features owing to the presence of obstructive pneumonia or atelectasis. Consequently, the present study group comprised 275 consecutive patients (144 men, 131 women) with a mean age of 67.8 years (range 39e87 years). Of these, 144 (52.4%) patients had screendetected cancer, 29 (10.5%) had cancers detected according to symptoms, and the cancers of the other 102 (37.1%) were discovered during examination for other diseases. Preoperative histopathological specimens were obtained by

either bronchoscopic biopsy or CT-guided percutaneous biopsy. Patients whose preoperative histopathological diagnoses were not confirmed for various reasons also underwent surgery based on typical CT features that strongly suggested adenocarcinoma.

CT imaging CT was performed using a variety of machines (Light Speed Qxi/4-row-detector, GE; Somatom Sensation/16row-detector, Siemens; Aquilion/64-row-detector, TOSHIBA; Somatom Definition Flash/128  2-row-detector, Siemens). CT images were reconstructed into 1e1.25 mm section thickness using the bone algorithm. Images were displayed with lung (level: 600 HU, width: 1600 HU) and mediastinal (level: 30 HU, width: 400 HU) window settings.

Image analysis The following thin-section CT features of primary lung lesions were retrospectively evaluated by three radiologists with 17, 16, and 8 years of experience in thoracic radiology: (a) maxD on lung window (mm); (b) percentage of GGO component by visual estimation, and lesions were subsequently divided into three groups: 50e100%, 1e49%, and 0%; (c) tumour shadow disappearance rate (TDR) defined by the following formula11: TDR ð%Þ ¼ ½1  maxD  the largest diameter perpendicular to maxD ðperDÞ on mediastinal window=ðmaxD  perD on lung windowÞ  100 (an example measurement is shown in Fig 1); (d) air bronchogram (presence or absence); (e) notch (presence or absence); (f) spiculation (presence or absence); (g) pleural indentation (presence or absence); (h) coexistence with a bulla or honeycomb cyst around the tumour (yes or no). The radiologists independently evaluated these CT features without knowledge of the preoperative serum CEA levels. For continuous variables such as maxD and TDR, a median between the three observers’ measurements was selected. For other CT features, the diagnosis of the majority was selected if any interobserver differences occurred. GGO was defined as hazy and amorphous increased lung attenuation without obscuration of the underlying vascular markings and bronchial walls.16 Spiculation was defined as the presence of linear strands at least 2 mm thick, extending from the tumour margin into the lung parenchyma.17 Pleural indentation was defined as a linear structure originating from the tumour and extending to the pleural surface. Notch was defined as a lobulate margin of the tumour. The tumour was classified as coexisting with a bulla or honeycomb cyst when the tumour margin adjoined the cyst. Bullae and honeycomb cysts were not differentiated because they can occasionally be difficult to differentiate at CT.

M. Yamazaki et al. / Clinical Radiology 69 (2014) 559e566

561

Figure 1 Measurement of the tumour shadow disappearance rate (TDR). The maximum tumour diameter (maxD) and the largest diameter perpendicular to maxD (perD) on lung window (left) is 25 mm and 21 mm, that on mediastinal window (right) is 17 mm and 12 mm. TDR is calculated as following formula: [1  (17  12)/(25  21)]  100 ¼ 61%.

Clinical characteristics of patients Patients’ age, gender, and Brinkman index were recorded because these factors have been considered to influence serum CEA levels.18 The Brinkman index was defined as number of cigarettes smoked per day  number of smoking years.

Statistical analysis All the following univariate and multivariate analyses were performed using Dr. SPSS II software (version 11.01 J; SPSS Japan, Tokyo, Japan). A p-value of 5 ng/ ml were considered to be high.

Univariate analysis To use univariate analysis, continuous variables such as maxD, TDR, age, and Brinkman index were categorized as follows: maxD: 2 cm and >2 cm; TDR: >75%, 25e75%, and 70; and Brinkman index: 400 and >400. Subsequently, three categorical data sets (GGO and TDR) were analysed using the chi-square test and two categorical data sets were analysed using Fisher’s exact test to evaluate the significance of differences in the percentage of patients with high preoperative serum CEA levels.

Multivariate analysis Factors with a p-value of 75%, 25e75%, and 70; p ¼ 0.002), male gender (p ¼ 0.006), and history of heavy smoking (Brinkman index >400; p < 0.001) were observed more frequently in the group with high serum CEA levels than in that with normal serum CEA levels.

Multivariate analysis

Results Among the 275 patients with p-stage I lung adenocarcinoma, 81 (29.5%) exhibited high preoperative serum CEA

MaxD, percentage of GGO, TDR, notch, spiculation, coexistence with a bulla or honeycomb cyst, patient age, gender, and Brinkman index were evaluated using

562

M. Yamazaki et al. / Clinical Radiology 69 (2014) 559e566

Table 1 Summary of computed tomography (CT) features and patients’ clinical characteristics (n ¼ 275). Factors MaxD on lung window (mm) Mean  SD Range Percentage of GGO 50e100% 1e49% 0% TDR >75% 25e75% 5 ng/ml) (n ¼ 81)

MaxD 2 cm 43 >2 cm 38 Percentage of GGO 50e100% 24 1e49% 31 0% 26 TDR >75% 20 25e75% 36 70 47 Gender Male 53 Female 28 Brinkman index 400 34 >400 47

p-Value

Normal (5 ng/ml) (n = 194)

(25%) (37%)

128 (75%) 66 (63%)

0.056

(19%) (33%) (49%)

105 (81%) 62 (67%) 27 (51%)

Relationship between CT features and high preoperative serum carcinoembryonic antigen levels in early-stage lung adenocarcinoma.

To assess the relationship between thin-section computed tomography (CT) features of primary tumour and high preoperative serum carcinoembryonic antig...
812KB Sizes 2 Downloads 3 Views