Does video-assisted mediastinoscopy have a better lymph node yield and safety profile than conventional mediastinoscopy?

Abstract

A best evidence topic was written according to a structured protocol. The question addressed was whether video-assisted mediastinoscopy (VAM) has a better lymph node yield and safety profile than the conventional mediastinoscopy (CM). A total of 194 papers were found, using the reported searches, of which five represented the best evidence to answer the clinical question. The authors, journal, date and country of publication, patient group studied, study type, relevant outcomes and results of these papers are tabulated. Two studies to date have directly compared CM and VAM with respect to lymph node yield, calculated diagnostics performance and complication rate. In both of these, lymph node yield is shown to be higher using VAM with better sensitivity, negative predictive value and accuracy rates. The favourable figures of lymph node sampling are found to be statistically significant in the single study providing such analysis. Complication rates using VAM are low, however, in the one instance where it is reported as higher than CM, the extensive lymph node dissection used in this technique may be a reasonable explanation for this finding. All studies described here exemplify VAM as a safe and useful tool in mediastinal staging, lymph node dissection and tissue diagnosis of mediastinal diseases given its superior visualization of surrounding structures and advantage of bimanual dissection. The future scope for diagnostic and therapeutic indications of cervical mediastinscopy is anticipated with recent advances and new techniques, such as video-assisted mediastinoscopic lymphadenectomy and virtual mediastinscopy.

INTRODUCTION

A best evidence topic was constructed according to a structured protocol. This protocol is fully described in the ICVTS [1].

CLINICAL SCENARIO

Whilst performing a cervical mediastinoscopy you struggle to get good access to the subcarinal lymph node without risking damage to mediastinal structures. Your registrar suggests that video mediastinoscopy is not only safer but yields higher lymph node harvest. You resolve to check the literature yourself.

THREE-PART QUESTION

In [patients undergoing cervical mediastinoscopy] is [video mediastinoscopy] superior to [conventional mediastinoscopy] in terms of [number of lymph nodes harvested safely]?

SEARCH STRATEGY

Medline was searched from 1948 to March 2011 using the OVIDSP interface with the terms (mediastinoscopy.mp. OR exp Mediastinoscopy/) AND (exp Video-Assisted Surgery/OR video.mp).

SEARCH OUTCOME

One hundred and ninety-four papers were found using the reported search on Medline. From these, five papers were identified that provided the best evidence to answer the question. These are presented in Table 1.

Table 1:

Best evidence papers

Author, date and country, study type (level of evidence)	Patient group	Outcomes	Key results				Comments
Anraku et al. (2010) [2], Ontario, Canada Retrospective non-randomized controlled trial level III evidence	645 patients investigated with mediastinoscopy: CM (May 2004–Oct 2006); VAM (Oct 2006–May 2008) Group 1: Lung cancer staging group (n = 500, CM = 396, VAM = 104) Of 362 patients with NSCLC and N0 or N1 disease on mediastinoscopy, 303 underwent thoracotomy + lung resection + mediastinal reassessment Group 2: Mediastinal mass/ lymphadenopathy Diagnostic group (n = 145, CM = 109, VAM = 36) Included in complication rate analysis	Mean LN yield per case Mean LN stations biopsied per case % cases biopsied per station Post-thoracotomy staging Calculated diagnostics performance (Group 1) Complications		VAM	CM		Statistically significant higher lymph node yield and number of stations biopsied with VAM, except in Subcarinal LN station 7 Higher diagnostic performance in VAM group, though not statistically significant. May be reflection of comparatively small number in VAM group (104 vs. CM 396) Higher complication rate with VAM. Authors feel due to more thorough dissection Limitations Attrition bias: 303 of 362 patients underwent thoractomy, 59 patients deemed inoperable or diagnosed with extrathoracic metastases Cointervention bias, unspecified number of cases received concomitant left anterior mediastinoscopy
				7.0 ± 3.2	5.0 ± 2.8	(P < 0.001)
				3.6 ± 1.1	2.6 ± 1.1	(P < 0.001)
			2R	74.0%	37.1%	(P < 0.001)
			2L	29.8%	10.1%	(P < 0.001)
			4R	96.1%	88.9%	(P < 0.001)
			4L	79.8%	47.2%	(P < 0.001)
			7	82.7%	78.8%	(P < 0.35)
			Agreed with mediastinoscopy in 96.4%
			False negative: 3.6% (VAM: 1, CM: 10)
			Sensitivity:	95%	92.2%
			Specificity:	100%	100%
			PPV:	100%	100%
			NPV:	98.6%	95.7%
			Accuracy:	98.9%	97.2%
			* Not statistically significant
			Overall: 1.2%, VAM: 2.9%, CM 0.8%
			No mortality
			Complications	VAM	CM
			Peri-op bleeding	2*	1
			Wound infection	0	1**
			Left recurrent laryngeal nerve injury	1	0
			Chyle leak	0	1
			Post-op pneumonia	0	1
			Myocardial infarction	0	1
			*1 case required median sternotomy for haemostasis
			**required debridement
Leschber et al. (2007) [3], Berlin, Germany Retrospective non-randomized, controlled trial level III evidence	377 patients investigated CM or VAM June 2003–Dec 2005 Group 1: Diagnostic Group  (n = 366, CM = 132, VAM = 234) Included in complication rate analysis  Group 2: lung cancer staging Group (n = 171, CM = 52, VAM = 119) Of 210 patients with NSCLC and N0 nodal disease on mediastinoscopy, 171 underwent subsequent thoracotomy + lung resection + mediastinal lymphadenectomy	Mean LN yield (range) Mean LN yield station 7 Post-thoracotomy staging Calculated diagnostics performance Complications		VAM	CM		Higher lymph node yield and diagnostics performance in VAM group Comparatively small sample of CM (52) participating in diagnostics performance High false negative rates, majority of these due to metastatic nodes in stations 5 and 6 Complication rate higher in CM group Statistical analysis not included Limitations Statistical significance not included Attrition bias: 10 of 377 patients not included due to incomplete data collection and 39 of 210 patients in Group 2 did not undergo thoracotomy with no comment on exclusion reasons Selective reporting bias: only 87% of Group 2 data collection included information on subcarinal station Potential expectation bias: Lymph nodes counted by performing surgeon rather than pathologist at lymphadenectomy
				8.1 (3–25)	6.0 (3–11)
				2.4 (0–7)	1.5 (0–3)
			False negative	16.8%	19.2%
			NPV	83%	81%
			Accuracy	87.9%	83.8%
			Overall: 4.6%, VAM: 4.3%, CM: 5.3%
			No mortality
			Complications	VAM	CM
			Left recurrent laryngeal nerve palsy > 2 days	5 2.1%	4 3%
			Pleural laceration	1	0
			Main bronchus laceration	1	0
			Post-operative bleeding	2*	3
			Post-operative pneumonia	1	0
			*1 case required mediastinoscopic exploration
Karfis et al. (2007) [4], Greece Retrospective non-randomized uncontrolled study level III evidence	138 patients investigated with VAM 1998-2006 Group 1: Lung cancer staging group (n = 87) 29 patients with NSCLC and N0 or N1 nodal disease on mediastinoscopy underwent thoracotomy + lung resection + mediastinal reassessment Group 2: mediastinal mass/ lymphadenopathy diagnostic Group (n = 51)	Mean LN stations biopsied per case % cases biopsied per station Post-thoracotomy staging Calculated diagnostics performance Complications	1.9 ± 0.8 (range 1–4)				No mention of total LN yield Low percentages of cases biopsied in each station High false negative rate Low complication rate Limitations Retrospective analysis,small sample size
			2R = 50.7%
			2L = 2%
			3 = 35.5%
			4R = 79.7%
			4L = 3%
			7 = 16.6%
			Agreed with VAM in 86.2%
			False negative rate: 13.8%
			Sensitivity = 86.1%
			Specificity = 100%
			Positive predictive value = 100%
			Positive predictive value = 100%
			Accuracy = 88.4%
			1.4%; Azygos vein trauma, mediastinitis (requiring mediastinoscopic drainage)
			No mortality
Venissac et al. (2003) [5], Nice, France Retrospective non-randomized uncontrolled study level III evidence	240 patients with ≥1 enlarged lymph node (>1cm) investigated with VAM (Oct 1998– Dec 2001) Of 67 patients with NSCLC and N0 or N1 disease on mediastinoscopy, 47 underwent subsequent thoracotomy + lung resection + mediastinal lymphadenectomy	Mean LN yield (range) Mean LN stations biopsied per case Post-thoracotomy staging Calculated diagnostics performance Complications	6.0 (range 1–20) 2.4 (range 1–4) Agreed with VAM in 93.6% False negative rate of 6.5% Sensitivity = 97.3% Specificity = 100% Accuracy = 98% 0.83% Pneumothorax, innominate artery tear (requiring manubrial split and suture repair) No mortality				High sensitivity and accuracy Low complication rate. Authors feel increased cost of procedure is counterbalanced by advantages, including low complication rate Limitations Retrospective analysis, small sample size Selection bias: Indication for VAM limited to patients with enlarged LNs, may account for high accuracy rates Co-intervention bias: 26 cases received VAM and video-assisted thoracoscopy Attrition bias: 47 of 67 NSCLC patients underwent thoracotomy
Hammoud et al. (1999) [6], Missouri, USA Retrospective non-randomized uncontrolled study level III evidence	2,137 patients investigated with CM (Jan 1988–Sept 1998) Of 1,323 patients with lung cancer and N0 or N1 disease on mediastinoscopy, 1,216 underwent subsequent thoracotomy + exploration Of 422 patients with N2 or N3 disease on mediastinoscopy, 28 underwent subsequent thoracotomy + lung resection + mediastinal lymphadenectomy indicated by patient participation in a neoadjuvant protocol and in cases where disease was limited to one station	Post-thoracotomy staging Calculated diagnostics performance Complications	Agreed with CM in 92% False negative rate of 8% Sensitivity 85.2% 0.6% Arrhythmia, pulmonary artery laceration*, oesophageal perforation*, excessive bleeding, IV fluid extravasation, pneumothorax, hypotension * (repaired at immediate thoracotomy) 4 peri-operative deaths (0.2%), 1 (0.05%) attributable to CM due to aortic laceration in presence of tumour infiltration				Large sample size No mention of LN yield or station biopsy Majority of false negative rate due to metastatic nodes in subaortic station 5, posterior subcarinal, pulmonary ligament nodes Low complication rate Limitations Retrospective analysis limiting reported outcome measures Attrition bias: 107 of 1,323 patients excluded from thoracotomy due to co-morbidities Co-intervention bias: 68 cases received anterior mediastinoscopy and 113 received combined anterior and cervical mediastinoscopy

LN, lymph node; NSCLC, non-small cell lung cancer; PPV, Positive predictive value; NPV, Negative predictive value.

RESULTS

Anraku et al. [2] conducted a retrospective non-randomized controlled study of 645 patients comparing the lymph node yield and safety profile of video-assisted mediastinoscopy (VAM) to conventional mediastinoscopy (CM), demonstrating statistically significant higher yield in the number of individual lymph nodes (7.0 ± 3.2 vs. 5.0 ± 2.8) and lymph node stations biopsied (3.6 ± 1.1 vs. 2.6 ± 1.1) with VAM. Though not statistically significant, the calculated sensitivity (95 vs. 92.2%), negative predictive value (98.6 vs. 95.7%) and accuracy (98.9 vs. 97.2%) at post-thoracotomy staging are consistently higher with VAM. It should be noted that an unspecified number of cases in this study received concomitant left anterior mediastinotomy to access stations 5 and 6, introducing a co-intervention bias that may improve the calculated diagnostics performance reported here. The complication rate was also higher in the VAM group (2.9 vs. 0.8%), which the authors postulated to be a likely reflection of the more thorough lymph node dissection demonstrated with this technique by the higher lymph node yield.

Leschber et al. [3] conducted a retrospective non-randomized controlled study directly comparing CM and VAM. Though no statistical analysis is included, this study demonstrates higher mean lymph node yield (6.0 vs. 8.1) and higher yield at subcarinal station 7 (1.5 vs. 2.4) with VAM. The calculated negative predictive value (83 vs. 81%) and accuracy (87.9 vs. 83.8%) were also higher with VAM, though the comparatively smaller patient group of 52 undergoing CM as opposed to 119 in the VAM group should be noted. The lower negative predictive values seen in this study in comparison with Anraku et al. [2] are likely attributable to the co-intervention bias mentioned previously. Despite the smaller sample size receiving CM, the complication rate was higher in this group in comparison to VAM (5.3 vs. 4.3%).

Karfis et al. [4] performed a retrospective non-randomized study concentrating on their seven year experience of using VAM at a single centre. With no mention of lymph node yield, it focuses on accessibility of lymph node stations (mean number of stations biopsied 1.9 ± 0.8) and percentage of cases biopsied in each lymph node station. These percentages are consistently lower when compared to those seen in Anraku et al's study. The difference in these results can be explained in part by the increased frequency and experience of the procedure, illustrated by the difference in sample size. The complication rates of VAM in this study are low at 1.4%. The authors highlight the advantages of better visualization, improved safety and greater opportunity for skills development at the expense of increased costs associated with the procedure.

Venissac et al. [5] conducted a retrospective non-randomized uncontrolled study on 240 patients with one or more enlarged mediastinal lymph nodes (>1 cm) on computerized tomography (CT) scan undergoing VAM. This study demonstrated an adequate mean lymph node yield (6.0) and a high number of lymph node stations biopsied (2.4). The high sensitivity (97.3%) and accuracy (98%) may be influenced by the selection bias established by limiting the indication for VAM to patients with one or more enlarged lymph nodes, where all other studies described performed VAM for lung cancer staging regardless of CT lymph node appearance. The complication rate for VAM is low at 0.83%. The authors emphasize the ease of lymph node dissection in both normal and pathological mediastinal structures, resulting in a safer procedure with lower complication rates to counterbalance the increased costs.

Hammoud et al. [6] conducted a large retrospective non-randomized uncontrolled study of 2137 patients undergoing CM for diagnosis of a mediastinal mass or staging of lung cancer. There is no mention of lymph node yield or number of stations biopsied in the procedures performed. Thoracotomy and mediastinal exploration was performed on 1216 of the 1323 patients with lung cancer found to have N0 or N1 nodal disease on CM, revealing a false negative rate of 8% and a low sensitivity of 85.2%. Despite the large sample size in this study, there is a paucity of information on the calculated diagnostics performance of CM. The complication rate was low (0.6%) and mortality (0.2%) was mostly ascribed to pre-operative co-morbidity with a single case (0.05%) attributable to the procedure.

CLINICAL BOTTOM LINE

VAM demonstrates higher lymph node yields and number of lymph node stations biopsied in both of the only two studies to date [2, 3] that directly compare the procedures. Though Anraku et al's large study [2] shows this difference in lymph node yield to be statistically significant and is currently the only study providing such analysis, this interpretation must be employed with caution due to the co-intervention bias introduced by performing concomitant left anterior mediastinoscopy to sample lymph node stations 5 and 6. As might be expected, the sensitivity of VAM increases to 97.3% in Venissac's study when used to investigate patients with enlarged lymph nodes (LNs) on CT scan. The comparative complication rate in these studies is inconclusive. One study reports higher complications with VAM, reasonably explained by the authors as caused by the extensive lymph node dissections performed.

Authors discussed here unanimously comment on the improved visualisation of mediastinal structures and benefit of bimanual dissection in lymph node sampling with VAM, as well as its value as a teaching and training tool. Technical progress with VAM has led to procedures such as video-assisted mediastinoscopic lymphadenectomy [7], an en bloc resection of the right and central compartments with dissection of the left, currently being assessed for its role in Stage III lung cancer. Improved accuracy and safety of cervical mediastinoscopy may be the future with concurrent use of virtual mediastinoscopy, conjuring realistic images of mediastinal anatomy [8].

Conflict of interest: none declared.