Patients

Consecutive patients with suspected or pathologically established lung cancer were eligible if they had mediastinal and/or upper retroperitoneal PET hot spots without evidence of further distant metastases, if they had a resectable primary tumour as judged by two thoracic surgeons, and if they were medically operable. All patients gave informed consent and the study was approved by the medical ethics committee of the Groningen University Hospital.

Patients were evaluated by history, physical examination, complete blood cell count, renal and liver function tests, chest radiography, bronchoscopy, CT scan of the chest and upper abdomen, and FDG-PET. All test results were presented in a multidisciplinary session where disease stage, tumour resectability, and medical operability were verified before EUS-FNA was performed.

Imaging tests

CT was performed with contrast enhanced single and multi-slice scanners (Somatom Sensation 16/Somatom AR, Siemens AG, Forchheim, Germany). It was used to assess the resectability of the primary tumour, to evaluate mediastinal lymph nodes, and to exclude distant metastases. Whole body PET was performed with an ECAT 951/31 or ECAT HR+ scanner (Siemens/CTI, Knoxville, TN, USA) 90 minutes after intravenous injection of a median dose of 390 MBq (range 130–670) FDG. Attenuation correction was obtained on the HR+ camera by transmission scanning with a germanium-68/gallium-68 rod source using an interleaved protocol. PET was used for the detection of mediastinal and distant metastases. If mediastinal and/or upper retroperitoneal hot spots were detected, PET was considered positive. Patients with hot spots suggesting metastases outside these areas were excluded from the study. A hot spot was defined as a focal increase in FDG uptake compared with the background not explained by physiological uptake. Both CT and PET were used as guidance for EUS-FNA.

EUS-FNA

EUS was performed with a linear scanning ultrasound endoscope (FG-34UX, Pentax GmbH, Hamburg, Germany) connected to a scanner unit (EUB-525, Hitachi Ultrasound BV, Reeuwijk, the Netherlands) with Doppler flow imaging for the detection of blood vessels. Midazolam 3–5 mg was administered intravenously for conscious sedation with continuous monitoring of pulse rate and peripheral oxygen saturation. The endoscope was introduced into the stomach to scan the left adrenal gland and the area around the celiac trunk. After retraction into the oesophagus, the mediastinum was evaluated by scanning 360° transaxially at 1–2 cm intervals upward up to Naruke level 2. The left adrenal gland was considered enlarged if it had lost its normal elliptical, triangular, or “seagull” shape.¹¹ Lymph nodes were identified as non-pulsating regions with round, ellipsoid, crescent, or triangular shape.¹² They were separately punctured with a 22 gauge needle (Echotip, Wilson-Cook Medical Inc, Winston-Salem, NC, USA) using a 10 ml syringe for suction. A cytotechnician evaluated the cellular contents of the specimens on site with a modified Giemsa method. If a specimen was inadequate, additional punctures were performed. Specimens were categorised as positive (tumour cells), negative (lymphoid but no tumour cells), or inconclusive (poor cellularity or unable to perform adequate puncture). Subsequent mediastinoscopy and/or exploratory thoracotomy with mediastinal dissection was performed in all patients except those with a positive EUS-FNA. The mediastinal lymph node map of the American Joint Committee on Cancer was used to localise abnormalities at PET, EUS-FNA, and mediastinal dissection.²

Because EUS-FNA is a new procedure for lung cancer staging, we were interested in the learning curve of two residents. During each session their ability to identify mediastinal lymph nodes was recorded. One of the residents performed EUS-FNA and an experienced pulmonary physician repeated the procedure. The numbers of observed mediastinal lymph nodes of the residents and the pulmonary physician were compared.

Economic analysis

Economic analysis was performed to assess the cost effectiveness of adding EUS-FNA to the staging algorithm. It was performed from a hospital point of view. All direct medical costs were included until the final lung cancer stage was established. This also included functional tests or additional examinations necessary for the evaluation of the patient’s operability. All cost items of EUS-FNA, mediastinoscopy, and exploratory thoracotomy were estimated by expert opinions of pulmonary physicians, thoracic surgeons, anaesthesiologists, and by observations of two hospital economists. Calculations were based on real costs using the 2002 price level. Costs of hospital admissions were calculated from the actual number of admission days to the general ward and intensive care unit against the Dutch unit price for hospital admission days.¹³ Overhead costs were calculated as a 35% surcharge. Because of the short time horizon of the diagnostic process, discounting for future costs was not applied.

Statistical analysis

The study was designed to determine prospectively the proportion of lung cancer patients with mediastinal and/or upper retroperitoneal PET hot spots in whom EUS-FNA can diagnose malignancy. Since EUS-FNA can reach at least seven of the 14 (right and left) mediastinal lymph node levels and the upper retroperitoneum, we estimated that about 50% of the surgical staging procedures could be avoided. Mediastinal PET has a positive predictive value of 67%.⁶ We therefore expected EUS-FNA to be able to identify 50% of this 67% as pathologically malignant—that is, 34% of all PET positive patients. Eighty patients were needed to estimate this proportion with a precision of 10%.

Statistical analysis was performed with SPSS 10.0 (SPSS Inc, Chicago, IL, USA). Continuous variables are reported as median (range) or mean (SE). Dichotomous variables are reported as percentages with 95% confidence interval (95% CI). The accuracy of EUS-FNA was calculated as the accuracy of the decision to proceed to surgical staging (or not) based on EUS-FNA results. For comparisons between groups of categorical variables, Pearson’s χ² test was used. Interobserver agreement for test results was calculated with Cohen’s κ value. Multiple logistic regression analysis was used to identify ultrasound features predictive for lymph node malignancy. We performed sensitivity analysis of the cost of the staging procedures by varying the major cost elements.^14,15 Reported p values are two sided and p values <0.05 were considered significant.

Patients

Between January 2001 and March 2003, 488 consecutive patients were screened at the outpatient department, 81 of whom fulfilled all the inclusion criteria and were enrolled in the study (table 1⇓). Their median age was 63 years (range 32–80).

Imaging tests

Images from a typical diagnostic procedure are shown in fig 1⇓. Sixty four patients (79%) patients had mediastinal and/or upper retroperitoneal abnormalities on the CT scan. Interobserver agreement (κ value) for the detection of enlarged lymph nodes and left adrenal enlargement on CT scanning was 0.76 (95% CI 0.66 to 0.86). All patients had mediastinal and/or upper retroperitoneal PET hot spots, mostly at Naruke level 4-right, 4-left, and 7 (table 2⇓).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

 Images of a 54 year old man with squamous cell lung carcinoma. (A) Helical contrast enhanced CT scan (transaxial section) showing a left upper lobe tumour (T) and enlarged mediastinal lymph nodes at Naruke 4-left (arrowhead) next to the oesophagus (arrow). (B) PET scan (coronal section) showing FDG uptake in the tumour (T) and at Naruke 4-left (arrowhead). (C) EUS (sagittal section) showing a hypoechoic lymph node at Naruke 4-left (arrowhead) next to the oesophagus (E). The aortopulmonary window is located between the aortic arch (A) and the left pulmonary artery (P).

EUS-FNA

EUS-FNA was positive for malignancy in 50 patients, thereby avoiding subsequent surgical staging procedures in 62% of the patients (95% CI 51 to 73). EUS-FNA was negative for malignancy in six (7%) and inconclusive for malignancy in 25 (31%) patients (fig 2⇓). The accuracy of EUS-FNA was 0.77 (95% CI 0.68 to 0.86). Most patients had EUS abnormalities at Naruke levels 4-left, 5, and 7 (table 2⇑). After PET, EUS detected additional abnormalities at Naruke levels 4-left, 5, 7, 8-right, and 8-left. EUS was not able to visualise accurately Naruke levels 1, 2-right, 2-left, 3, or 4-right.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

 Flowchart of the enrolled patients.

Twenty six of the 31 patients with a negative or inconclusive EUS-FNA underwent a surgical staging procedure: mediastinoscopy in 12 patients, mediastinotomy in one, exploratory thoracotomy with mediastinal dissection in nine, and mediastinoscopy followed by exploratory thoracotomy with mediastinal dissection in four patients. Of the six patients with a negative EUS-FNA, two had positive and three had negative mediastinal lymph nodes at surgery (fig 2⇑). Further staging was not performed in one patient due to rapid progression of the mediastinal lesions. Of the 25 patients with an inconclusive EUS-FNA, 12 were positive and nine were negative for malignancy at surgery (fig 2⇑). Four patients refused further surgical staging and, after follow up of 6 months, all had progressive disease diagnosed with imaging tests.

After all staging procedures including mediastinal dissection, mediastinal or upper retroperitoneal malignancy was not detected in 12 of the 81 patients (15%). A negative or inconclusive EUS-FNA did not reliably prove the absence of mediastinal metastasis: 19 of the 31 patients (68%) had lymph node metastasis at surgery. Of these 19 patients, seven had lymph node metastasis exclusively at the upper right mediastinum, outside the reach of EUS-FNA.

One mediastinal lymph node (which was localised at the PET hot spot) was randomly selected per patient for multiple logistic regression analysis. Lymph node size, shape (flat, round, irregular), boundary (vague, sharp), core structure (hyperechogenic, hypoechogenic), and visual impression (benign, malignant) were used as factors in the analysis. Lymph node size was the only significant predictor of malignancy (p = 0.03). For each additional centimetre the chance of malignancy increased by 0.56 (95% CI 0.04 to 1.00).

The learning process of EUS-FNA of two residents was studied during 54 procedures, 29 by the first resident and 25 by the second. The numbers of lymph nodes detected by the residents failed to reach those identified by the experienced operators. More procedures are therefore required to learn EUS-FNA accurately by those who have not previously performed oesophageal endoscopy.

The median duration of the EUS-FNA procedure (including resident training) was 45 minutes (range 20–60). There was no morbidity or mortality during or after EUS-FNA.

Economic analysis

Costs are summarised in table 3⇓. Hospitalisation was a major cost item for mediastinoscopy and exploratory thoracotomy with a mean (SE) duration of 3.0 (0.3) days for mediastinoscopy and 12.0 (1.9) days for exploratory thoracotomy. Average staging costs were $2101 per patient using a staging algorithm with EUS-FNA. For a staging algorithm without EUS-FNA, average staging costs were $3514 per patient, assuming that the ratio of mediastinoscopies to exploratory thoracotomies remained constant (17:13). Thus, mean cost savings through EUS-FNA were $1413 (40%) per patient. Sensitivity analyses were performed by varying the costs of the staging procedures. Cost savings could be maintained even if the costs of EUS-FNA were 2.5 times the current costs. Likewise, even with a 75% reduction in the costs of mediastinoscopy or a 50% reduction in the costs of exploratory thoracotomy, the outcome still favoured the EUS-FNA algorithm. Increasing the ratio of mediastinoscopies to exploratory thoracotomies from 17:13 to 10:1 in the staging algorithm without EUS-FNA (that is, a reduction in exploratory thoracotomies of 80%) would result in equal costs of both staging algorithms.