Journal of Obstetrics and Gynaecology, May 2015; 35: 377–381 © 2015 Informa UK, Ltd. ISSN 0144-3615 print/ISSN 1364-6893 online DOI: 10.3109/01443615.2014.958448
GYNAECOLOGY
Is quantification of lymphovascular space invasion useful in stage 1B2 cervical carcinomas? J. Scurry1,2, N. F. Hacker3,4, E. Barlow3, M. Friedlander3,5 & M. Jackson6,7 1Hunter Area Pathology Service, John Hunter Hospital, New Lambton Heights, NSW Australia, 2Faculty of Health Sciences, University of
J Obstet Gynaecol Downloaded from informahealthcare.com by Deakin University on 08/10/15 For personal use only.
Newcastle, Newcastle, NSW Australia, 3Gynaecological Cancer Centre, The Royal Hospital for Women, Randwick, NSW, Australia, 4School of Women’s and Children’s Health, University of New South Wales, Randwick, NSW, Australia, 5Department of Medicine, Clinical School, University of New South Wales, Randwick, NSW, Australia, 6Department of Radiation Oncology, Prince of Wales Hospital, Randwick, NSW, Australia, and 7University of New South Wales, Randwick, NSW, Australia
The objective was to determine whether quantification of lymphovascular space invasion (LVSI) by simple techniques adds prognostic information above its mere identification in stage 1B2 cervical cancer. The method was to quantify LVSI by extent, density and distance from the advancing front in 88 consecutive stage 1B2 cervical cancers treated by radical hysterectomy and pelvic lymphadenectomy and to compare them with pelvic lymph node status and local and distant recurrence. The results were that LVSI involved more tumour blocks, was denser and extended a further distance in those with positive nodes. However, effective adjuvant therapy confounded the association between quantification of LVSI and local recurrence. Furthermore, pelvic lymph node status was a stronger predictor of distant recurrence than any degree of LVSI. In conclusion, quantifying LVSI in stage 1B2 cervical cancer is a good predictor of lymph node metastasis, but is not useful where the lymph node status is known. Keywords: Lymphovascular space invasion, quantification, stage 1B2 cervical cancer
Introduction Early stage cervical cancer may be treated equally effectively with radical hysterectomy and pelvic lymphadenectomy or radiotherapy (Landoni et al. 1997). An advantage of surgery is the availability of a specimen for histopathology. However, noninvasive techniques of positive emission tomography (PET) and magnetic resonance imaging (MRI) can provide some of the data that has traditionally been the province of histopathology. Thus, lymph node metastases 5 mm can be identified by PET and the size, depth of invasion and parametrial spread can be assessed by MRI. By contrast, a prognostic feature that relies totally on histology for identification is lymphovascular space invasion (LVSI). The Gynecologic Oncology Group (GOG) prospective clinicopathological study of adjuvant pelvic radiation therapy in women with local high risk factors, identified three independent risk factors: tumour size, the relative depth of cervical stromal invasion and LVSI (Delgado et al. 1990). Other studies, however, did not find LVSI to be an independent prognostic factor in women with early cervical cancer (Gauthier et al. 1985; Smiley et al. 1991). One
possible explanation for these divergent results is that the quantity of LVSI has not been taken into account. For the purposes of routine reporting and in most studies, LVSI is assessed simply as being seen or not seen, even though there appears to be a great variation of the number of examples per case. We aimed to determine whether there are any simple quantifying measures that better predict pelvic lymph node metastasis and recurrence than LVSI status and whether quantification is worthwhile once pelvic lymph node status is known. We studied the extent, density and distance from the invasive front of LVSI in 88 unselected, consecutive stage 1B2 cervical carcinomas.
Materials and methods After obtaining approval of the South Eastern Sydney Area Health Service Human Research Ethics Committee, all patients with FIGO stage 1B2 as assessed by clinical palpation cervical cancer, treated at the Royal Hospital for Women, Sydney, from 1988 to 2008, were identified from our database. There were 106 patients identified. Medical records of all the women were reviewed and relevant clinical data extracted. A clinicopathological review of this cohort, excluding quantification of LVSI has been published (Hacker et al. 2013). A total of 13 women (12%) underwent primary radiation therapy because of medical comorbidities or personal preferences. Adjuvant radiation was given to high-risk patients, defined as those with a GOG score of 120. The GOG score is a calculation involving the multiplication of the three independent risk factors: clinical tumour diameter, depth of invasion and status of LVSI. Of the 93 women who underwent primary surgical management, 88 had a complete set of histological slides on file. The slides of the radical hysterectomies of these women were reviewed by a gynaecological pathologist (JS) without reference to the lymph node status or follow-up. The tumours were typed as squamous cell of adenocarcinomas. Adenosquamous carcinomas were placed in the adenocarcinoma group. Special types of tumours with a different prognosis were identified. Tumours were graded empirically as well-differentiated (grade 1); intermediate (grade 2); poorly-differentiated (grade 3). Tumours were divided into those with pushing and infiltrative histological growth patterns. The growth pattern of the tumour was assessed empirically at the advancing front of the tumour.
Correspondence: J. Scurry, Anatomical Pathologist, Hunter Area Pathology Service, and Conjoint Associate Professor, Faculty of Health Sciences, University of Newcastle, Newcastle, NSW 2310, Australia. E-mail: [email protected]
J Obstet Gynaecol Downloaded from informahealthcare.com by Deakin University on 08/10/15 For personal use only.
378
J. Scurry et al.
Pushing tumours had smooth, broad advancing edges and infiltrative tumours had an irregular edge with numerous small tumour nests and tentacles. Cases that appeared intermediate between these extremes were classified as ‘intermediate’ growth pattern. The amount of fibroblastic stromal reaction was assessed empirically as moderate to marked (stroma-rich) or none to mild (stroma-poor). Carcinomas were assessed empirically according to their degree of lymphocytic reaction as lymphocyte-rich (moderate to marked lymphocytic reaction) or lymphocyte-poor (none to mild). The depth of invasion was measured as the perpendicular distance between the basement membrane of the surface epithelium of non-involved cervix to a horizontal line drawn from the deepest invasive cell. In circumferential exophytic tumours, the original cervical wall thickness was assumed to be 15 mm and the depth of invasion estimated by a subtraction method (15 mm – distance from the deepest tumour cell to outermost cervix). Carcinomas were divided into those showing inner-third, middle-third and outer-third invasion of the cervical wall. Parametrial invasion was defined as direct invasion, LVSI, in transit metastasis or lymph node involvement of parametrial fat. Direct involvement of the corpus was defined as tumour histologically extending proximal to the first endocervical cell of the surface or glandular field epithelium. LVSI was identified when a tumour embolus lay entirely within an endothelial-lined space beyond the advancing front of the tumour. It was the laboratory practice to receive fresh specimens from the operating theatre, immediately open them and fix them in abundant formalin. At the time of gross dissection, the entire cervix was embedded in cases of cervical cancer. For this study, blocks with tumour were identified by holding slides up to the light and looking for irregular and/or blue (haematoxylinophilic) areas of tumour. For practicality, when 20 blocks contained tumour, only 20 were examined. Only blocks with the invasive front of tumour were examined, including the blocks with the deepest and largest areas of tumour. When more than one focus of LVSI was seen, no attempt was made to determine whether this was due to multiple tumour emboli in multiple vessels or a single tortuous vessel being involved with one elongated tumour embolus. LVSI was quantified as follows. The fraction of tumour blocks involved with LVSI (extent), maximum number of foci in the worst affected slide density and the maximum distance between an LVSI focus and the advancing front of the tumour (distance), as measured by an eyepiece graticule, were recorded. In those women with LVSI, these forms of quantified LVSI were compared with pelvic lymph node metastases, local and distant recurrence and adjuvant therapy using the t-test and the Wilcoxon unpaired test. For the outcome ‘distant recurrence’, several explanatory variables were tested to see if any was stronger than the others. Logistic regression models were fit using the three LVSI quantification measures and pelvic lymph node metastases. Each model contained only one of these variables and the AICc (Aikake Information Criterion with correction for small samples) was used to test if any variable performed differently.
Figure 1. Two foci of tumour are seen in the myometrium beyond the advancing front. On the left side, the tumour lies within a clearly defined endothelial-lined space and this is counted as a focus of LVSI. On the right side, inflammatory cells obscure any endothelium that might be present and, while LVSI is likely, this focus does not fulfil the criteria for LVSI (Haematoxylin and eosin 200).
adenocarcinomas. Two special types of squamous and four of adenocarcinoma were found. The special types of squamous cell carcinoma comprised one nasopharyngeal and one papillary. The special types of adenocarcinoma comprised one adenoid basal carcinoma combined with carcinosarcoma; one adenoid cystic carcinoma and two glassy cell carcinomas. No small cell carcinomas were found. A total of 65 women had at least one focus of LVSI identified. We encountered two difficult scenarios where it was difficult to determine whether a focus of inflammation should be counted as LVSI or not. The most troublesome were cases which showed foci of inflammation potentially obscuring tumour beyond the advancing edge of the tumour. It was difficult to determine on routine histology whether enlarged cells seen in these foci were reactive endothelial, inflammatory or tumour cells (Figure 1). Such examples were not counted as LVSI unless tumour cells were definitely identified. The other problem was the finding of separate tumour foci beyond the advancing edge of the tumour not explainable by direct growth of tumour or perineural invasion (Figure 2). While it is difficult to conceive how this scenario occurred except by LVSI, such cases were not counted as LVSI
Results The mean age of the 88 women was 46 years. All women were primarily treated by radical hysterectomy and, at least, bilateral pelvic lymphadenectomy. A total of 73 women (83%) received some form of adjuvant therapy (radiation plus or minus chemotherapy). The carcinomas were 59 squamous cell and 29
Figure 2. A tumour focus lies next to a small artery, a typical location of a lymphatic vessel. While LVSI is the most likely explanation, the tumour is not surrounded by an endothelial-lined space meaning that it does not fulfil the criteria for LVSI (Haematoxylin and eosin 100).
Is quantification of lymphovascular space invasion useful in stage 1B2 cervical carcinomas? 379
J Obstet Gynaecol Downloaded from informahealthcare.com by Deakin University on 08/10/15 For personal use only.
Table I. Pathological features vs presence of at least one focus of LVSI. n (n 88)
n with LVSI (n 65)
59 29
42 23
0.4
7 44 37
6 32 27
0.3
20 15 53
10 10 45
0.009
38 50
28 37
1.0
36 52
28 37
0.49
3 12 77
0 4 61
0.001
29 59
24 41
0.3
Type SCC Adeno Grade Grade 1 Grade 2 Grade 3 Growth pattern Pushing Intermediate Infiltrative Stromal reaction Stroma-poor Stroma-rich Lymphocytic reaction Lymphocyte-poor Lymphocyte-rich Depth of cervical wall involvement Inner-third Middle-third Outer-third Corpus involvement Yes No
p value
unless a surrounding endothelium could be seen. By contrast, the well-known mimics of LVSI, translocation and retraction artefact, were rare and easily distinguished from LVSI. The relationship of tumour characteristics to the presence of LVSI is shown in Table I. The presence of LVSI correlated with greater depth of cervical wall invasion and an infiltrative growth pattern. However, occasional examples of deeply invasive tumours without LVSI and pushing growth pattern tumours with much LVSI were encountered. No relationship of LVSI to tumour type, grade, stromal or lymphocytic reaction or corpus involvement was found. A total of 40 women (45%) had positive pelvic lymph nodes: 37 of these 40 (93%) had LVSI and three did not. Of the 65 women with LVSI, LVSI involved a greater fraction of blocks, had a higher mean maximum density in the worst affected slide and was a further distance from the advancing edge of the tumour in those with positive compared with negative pelvic nodes (Table II). A graph of the chances of nodal metastasis compared with the fraction of LVSI blocks/total tumour blocks is shown in Figure 3. Follow-up ranged from 6 to 283 (mean 96) months. Five (6%) women had a local recurrence. Of the 73 women who received adjuvant therapy, 59 had LVSI and 14 did not. Of the 14 who did not receive adjuvant therapy, three of the six with LVSI (3/6) developed local recurrence compared with 1/9 without LVSI (p 0.09). The relationship between local recurrence, LVSI, pelvic lymph node metastases and adjuvant therapy is shown in
Figure 3. Relationship between fraction of tumour blocks involved with LVSI and chance of pelvic lymph node metastases.
Table III. We found that four out of five of the women with local recurrence had not been given adjuvant therapy. A total of 13 (15%) women had distant recurrence. The relationship between distant recurrence, LVSI, pelvic lymph node metastases and adjuvant therapy is shown in Table IV. The usual profile of the women with distant recurrence, seen in 11/13 (85%), was LVSI-positive, lymph node-positive and given adjuvant therapy. Using the AICc, pelvic lymph node metastasis was found to be a stronger explanatory variable than any of the three quantified LVSI measures, i.e. (1) fraction of tumour blocks with LVSI/ total tumour blocks; (2) maximum density of LVSI in worst affected slide and (3) longest distance of LVSI from advancing tumour front. The AICc values were 67.8, 72.3, 72.4 and 74.5, respectively (lower values better).
Discussion Only a few studies have quantified the amount of LVSI. Roman et al. (1998) found that the quantity of LVSI, rather than its mere presence, was a significant risk factor for pelvic lymph node metastases. In their study, the entire cervix was blocked for histological assessment and the amount of LVSI quantified in four ways: (1) the percentage of cervical histopathological sections containing LVSI; (2) the percentage of sections with tumour containing LVSI; (3) the total number of foci with LVSI and (4) the maximal number of foci of LVSI in 10 high powered fields. Only the percentage of cervical sections containing LVSI reached significance. Logistic regression found that LVSI in 45% of all cervical histopathological sections to be a significant predictor of nodal metastases. However, once the pelvic lymph node status is known, is quantifying LVSI any advantage? Chernofsky et al. (2006), using the same cohort and method of LVSI quantitation as Roman and colleagues, answered this question in the affirmative by finding
Table II. Quantitation of LVSI vs pelvic lymph node metastases, local recurrence and distant recurrence in those women who have LVSI present.
Extent Density Distance
PLN (n 37)
PLN (n 28)
p value
LR (n 4)
LR (n 61)
p value
DR (n 12)
DR (n 53)
p value
47 5.9 3.5
29 4 1.4
0.01 0.001 0.001
62 3.3 1.7
40 8.5 2.6
0.4 5.2 4.4
50 10 4.3
39 7.8 2.1
0.02 0.008 0.02
Extent: mean % blocks with LVSI/total blocks of cancer examined. Density: mean number of foci of LVSI per worst affected slide. Distance: mean distance (mm) of furthest LVSI from advancing front of tumour. PLN, pelvic lymph node; LR, local recurrence; DR, distal recurrence.
380
J. Scurry et al.
Table III. Relationship between local recurrence, LVSI, pelvic lymph node status and adjuvant therapy.
J Obstet Gynaecol Downloaded from informahealthcare.com by Deakin University on 08/10/15 For personal use only.
LVSI Pelvic lymph node metastases Adjuvant therapy
Local recurrence (n 5)
No local recurrence (n 83)
4 1 1
61 39 72
time to recurrence was significantly reduced in women with the combination of 29% of sections with LVSI and a total number of foci with LVSI 5. Roman and colleagues’ method (1998), while providing proof of principle, is time-consuming and requires the entire cervix of cancer specimens to be blocked; a practice which is not performed routinely in many laboratories. Sykes et al. (2003) in a study involving one of us (JS), used a much simpler method. They counted the number of foci of LVSI in the most affected slide in node-negative stage 1B1 cervical carcinomas and compared this with recurrence. While there was a trend, their study produced a negative result. Possible reasons why their study failed to reach significance are: (1) their cut-off for mild (1–2 foci of LVSI), moderate (3–9 foci) and severe LVSI ( 9 foci) were arbitrarily set at the beginning rather than being identified statistically at the end of the study and (2) the wrong parameter, density (number of foci per slide) was chosen rather than the extent (number of blocks involved). We found that only two pathological features of the primary tumour correlated with the presence of LVSI: the degree of cervical wall invasion and an infiltrative growth pattern of the advancing front of the tumour. The association of depth with cervical wall invasion is presumably a volume effect. The greater the volume of tumour, the more chances of tumour cells and vessels will approximate and LVSI occur. The correlation of LVSI with an infiltrative advancing front is consistent with higher biological aggressiveness, as has been observed in SCC of other sites. In another study involving one of us (NFH), an infiltrative growth pattern was been found to be associated with an increased risk of recurrence in SCC of the vulva (Heaps et al. 1990). The presence of LVSI strongly correlated with pelvic lymph node metastasis. However, the absence of LVSI is not a reliable guide that pelvic lymph nodes will be negative. Three of the 40 women with lymph node metastasis did not have LVSI. Possible explanations for this small group are: (1) that some tumours may metastasise to nodes via transport of single or small groups of cells gaining access to lymphatics within the tumour without forming the usual intravascular tumour columns visible beyond the confines of the tumour; (2) some patterns of tumour growth beyond the advancing edge, which are true LVSI, are not recognised as such using the standard criterion of a tumour mass completely surrounded by endothelium; and (3) sampling error, as even if all the cervix is submitted, usually only one section per 3–4 mm slice is examined. Table IV. Relationship between distant recurrence, LVSI, pelvic lymph node status and adjuvant therapy. Distant recurrence No distant recurrence (n 13) (n 75) LVSI Pelvic lymph node metastases Adjuvant therapy
12 11 11
53 29 62
When LVSI was quantified, women with a higher fraction of blocks with LVSI (extent), a higher maximum number of foci per block (density) and a further distance of LVSI from the invasive front (distance) were more likely to have positive pelvic nodes. Similarly, Roman et al. (1998) found that a higher fraction of tumour blocks with LVSI predicts an increased likelihood of pelvic lymph node metastases and Memarzadeh et al. (2003) found parametrial as opposed to cervical stromal LVSI carried a higher risk of nodal metastasis. While our study showed quantification of LVSI is very helpful in predicting pelvic lymph node status, this information is of no practical use if the lymph node status is known, as it is in women treated by the surgical standard of radical hysterectomy and pelvic lymphadenectomy. We therefore looked at whether quantification of LVSI gave a better prediction of local and distant recurrence than a yes/no assessment of LVSI. Adjuvant therapy was clearly effective in reducing local recurrence. While the proportion of local recurrences in women not given adjuvant therapy was similar to those not given radiotherapy in the GOG study (27% vs 28%), if adjuvant therapy was given, our local recurrence rate dropped to 1.5% (Sedlis et al. 1999). Quantification of LVSI was of no value in predicting local recurrence. The explanation of this finding is that high-count LVSI women were likely to be node positive and to be given adjuvant therapy, which protected against local recurrence. Distal recurrence correlated with both quantification of LVSI and pelvic lymph node metastases. However, logistic regression using AICc, showed that pelvic lymph node metastasis was a stronger predictor for any form of quantification of LVSI. The importance of positive pelvic nodes as a prognostic factor is well illustrated by Chen et al.’s 2013 study of 588 cancer from stages IA2 to IIA. The 5-year overall survival rate of those with node positive cancer was found to be much lower than those without (22.4% vs 84.4%). The 1B2s in that study were fairly similar to our own cohort in that they contained a high number of cases with multiple node metastases, accounting for a bigger drop in survival than in previous literature. Our results support a stepwise progression of cervical cancer from primary tumour → LVSI → regional pelvic lymph node metastases → distant metastases (Saharinen et al. 2004). This ‘halsteadian’ hypothesis provides the rationale behind current standard surgical and adjuvant therapy protocols. While a jump from primary tumour to distant metastases is possible by (1) inefficient lymph nodes allowing tumour cells to pass through, (2) lymphovenous anastomosis or (3) systemic spread of cancer cells entering the peripheral blood within the tumour, distant spread without pelvic lymph node metastasis is unusual in cervical cancer. While our study has shown no benefit in quantifying LVSI is 1B2 cervical cancers, it may be helpful in smaller cancers, as these are less likely to have lymph node metastases and less likely to have been given adjuvant therapy. Sykes et al. (2003) showed a trend for more LVSI to predict recurrence in the node-negative stage. Further studies are needed to determine the significance or otherwise of their findings. Another potential use of quantifying LVSI is in endometrial cancer as lymphadenectomy is not standard surgery in that cancer. Hachisuga et al. (1999) found that grading of LVSI correlated with lymph node metastasis in endometrial cancer. They graded LVSI on a very simple system of none, one focus and multifocal. A more detailed method of grading LVSI, as used in our study, could be applied to endometrial cancer in order to see whether this could improve the accuracy of nodal metastasis prediction. In summary, we have shown that using relatively simple methods of quantification of LVSI of extent, density and distance
Is quantification of lymphovascular space invasion useful in stage 1B2 cervical carcinomas? 381
J Obstet Gynaecol Downloaded from informahealthcare.com by Deakin University on 08/10/15 For personal use only.
from the tumour front, there is a strong correlation between increased LVSI on all three measures and pelvic lymph node metastases in stage 1B2 cervical cancers. However, since the pelvic lymph node status is known in women treated by standard surgical therapy, quantifying LVSI to predict nodal status is of no benefit. When we attempted to predict local and distant recurrence by quantifying LVSI, we found that adjuvant therapy confounded the prediction of local recurrence and that distant recurrence was better predicted by nodal status. We conclude that there is no benefit in quantifying LVSI in 1B2 cervical cancers when the nodal status is known. Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper. Ellen Barlow was supported by a grant from the GO Research Fund, an account within the Royal Hospital for Women Foundation.
References Chen Y, Zhang L, Tian J, Fu X, Ren X, Hao Q. 2013. Significance of the absolute number and ratio of metastatic lymph nodes in predicting postoperative survival for the International Federation of Gynecology and Obstetrics Stage IA2 to IIA Cervical Cancer. International Journal of Gynecological Cancer 23:157–163. Chernofsky MR, Felix JC, Muderspach LI, Morrow CP, Ye W, Groshen SG et al. 2006. Influence of quantity of lymph vascular space invasion on time to recurrence in women with early-stage squamous cancer of the cervix. Gynecologic Oncology 100:288–293. Delgado G, Bundy B, Zaino R, Sevin BU, Creasman WT, Major F. 1990. Prospective surgical-pathological study of disease-free interval in patients with stage IB squamous cell carcinoma of the cervix: a Gynecologic Oncology Group study. Gynecologic Oncology 38:352–357.
Gauthier P, Gore I, Shingleton HM, Soong SJ, Orr JW Jr, Hatch KD. 1985. Identification of histopathologic risk groups in stage IB squamous cell carcinoma of the cervix. Obstetrics and Gynecology 66:569–574. Hachisuga T, Kaku T, Fukuda K, Eguchi F, Emoto M, Kamura T et al. 1999. The grading of lymphovascular space invasion in endometrial carcinoma. Cancer 86:2090–2097. Hacker NF, Barlow E, Scurry J, Gebski V, Farrell R, Robertson G et al. 2013. Primary surgical management with tailored adjuvant radiation for stage IB2 cervical cancer. Obstetrics and Gynecology 121:765–772. Heaps JM, Fu YF, Montz FJ, Hacker NF, Berek JS. 1990. Surgical-pathologic variables predictive of local recurrence in squamous cell carcinoma of the vulva. Gynecologic Oncology 38:309–314. Landoni F, Maneo A, Colombo A, Placa F, Milani R, Perego P et al. 1997. Randomised study of radical surgery versus radiotherapy for stage IB-IIA cervical cancer. Lancet 350:535–540. Memarzadeh S, Natarajan S, Dandade DP, Ostrzega N, Saber PA, Busuttil A et al. 2003. Lymphovascular and perineural invasion in the parametria: a prognostic factor for early-stage cervical cancer. Obstetrics and Gynecology 102:612–619. Roman LD, Felix JC, Muderspach LI, Varkey T, Burnett AF, Qian D et al. 1998. Influence of quantity of lymph-vascular space invasion on the risk of nodal metastases in women with early-stage squamous cancer of the cervix. Gynecologic Oncology 68:220–225. Saharinen P, Tammela T, Karkkainen MJ, Alitalo K. 2004. Lymphatic vasculature: development, molecular regulation and role in tumor metastasis and inflammation. Trends in Immunology 25:387–395. Sedlis A, Bundy BN, Rotman MZ, Lentz SS, Muderspach LI, Zaino RJ. 1999. A randomized trial of pelvic radiation therapy versus no further therapy in selected patients with stage IB carcinoma of the cervix after radical hysterectomy and pelvic lymphadenectomy. A Gynecologic Oncology Group Study. Gynecologic Oncology 73:177–183. Smiley LM, Burke TW, Silva EG, Morris M, Gershenson DM, Wharton JT. 1991. Prognostic factors in stage IB squamous cervical cancer patients with low risk for recurrence. Obstetrics and Gynecology 77:271–275. Sykes P, Allen D, Cohen C, Scurry J, Yeo D. 2003. Does the density of lymphatic vascular space invasion affect the prognosis of stage Ib and IIA node negative carcinoma of the cervix? International Journal of Gynecological Cancer 13:313–316.
GYNAECOLOGY
Is quantification of lymphovascular space invasion useful in stage 1B2 cervical carcinomas? J. Scurry1,2, N. F. Hacker3,4, E. Barlow3, M. Friedlander3,5 & M. Jackson6,7 1Hunter Area Pathology Service, John Hunter Hospital, New Lambton Heights, NSW Australia, 2Faculty of Health Sciences, University of
J Obstet Gynaecol Downloaded from informahealthcare.com by Deakin University on 08/10/15 For personal use only.
Newcastle, Newcastle, NSW Australia, 3Gynaecological Cancer Centre, The Royal Hospital for Women, Randwick, NSW, Australia, 4School of Women’s and Children’s Health, University of New South Wales, Randwick, NSW, Australia, 5Department of Medicine, Clinical School, University of New South Wales, Randwick, NSW, Australia, 6Department of Radiation Oncology, Prince of Wales Hospital, Randwick, NSW, Australia, and 7University of New South Wales, Randwick, NSW, Australia
The objective was to determine whether quantification of lymphovascular space invasion (LVSI) by simple techniques adds prognostic information above its mere identification in stage 1B2 cervical cancer. The method was to quantify LVSI by extent, density and distance from the advancing front in 88 consecutive stage 1B2 cervical cancers treated by radical hysterectomy and pelvic lymphadenectomy and to compare them with pelvic lymph node status and local and distant recurrence. The results were that LVSI involved more tumour blocks, was denser and extended a further distance in those with positive nodes. However, effective adjuvant therapy confounded the association between quantification of LVSI and local recurrence. Furthermore, pelvic lymph node status was a stronger predictor of distant recurrence than any degree of LVSI. In conclusion, quantifying LVSI in stage 1B2 cervical cancer is a good predictor of lymph node metastasis, but is not useful where the lymph node status is known. Keywords: Lymphovascular space invasion, quantification, stage 1B2 cervical cancer
Introduction Early stage cervical cancer may be treated equally effectively with radical hysterectomy and pelvic lymphadenectomy or radiotherapy (Landoni et al. 1997). An advantage of surgery is the availability of a specimen for histopathology. However, noninvasive techniques of positive emission tomography (PET) and magnetic resonance imaging (MRI) can provide some of the data that has traditionally been the province of histopathology. Thus, lymph node metastases 5 mm can be identified by PET and the size, depth of invasion and parametrial spread can be assessed by MRI. By contrast, a prognostic feature that relies totally on histology for identification is lymphovascular space invasion (LVSI). The Gynecologic Oncology Group (GOG) prospective clinicopathological study of adjuvant pelvic radiation therapy in women with local high risk factors, identified three independent risk factors: tumour size, the relative depth of cervical stromal invasion and LVSI (Delgado et al. 1990). Other studies, however, did not find LVSI to be an independent prognostic factor in women with early cervical cancer (Gauthier et al. 1985; Smiley et al. 1991). One
possible explanation for these divergent results is that the quantity of LVSI has not been taken into account. For the purposes of routine reporting and in most studies, LVSI is assessed simply as being seen or not seen, even though there appears to be a great variation of the number of examples per case. We aimed to determine whether there are any simple quantifying measures that better predict pelvic lymph node metastasis and recurrence than LVSI status and whether quantification is worthwhile once pelvic lymph node status is known. We studied the extent, density and distance from the invasive front of LVSI in 88 unselected, consecutive stage 1B2 cervical carcinomas.
Materials and methods After obtaining approval of the South Eastern Sydney Area Health Service Human Research Ethics Committee, all patients with FIGO stage 1B2 as assessed by clinical palpation cervical cancer, treated at the Royal Hospital for Women, Sydney, from 1988 to 2008, were identified from our database. There were 106 patients identified. Medical records of all the women were reviewed and relevant clinical data extracted. A clinicopathological review of this cohort, excluding quantification of LVSI has been published (Hacker et al. 2013). A total of 13 women (12%) underwent primary radiation therapy because of medical comorbidities or personal preferences. Adjuvant radiation was given to high-risk patients, defined as those with a GOG score of 120. The GOG score is a calculation involving the multiplication of the three independent risk factors: clinical tumour diameter, depth of invasion and status of LVSI. Of the 93 women who underwent primary surgical management, 88 had a complete set of histological slides on file. The slides of the radical hysterectomies of these women were reviewed by a gynaecological pathologist (JS) without reference to the lymph node status or follow-up. The tumours were typed as squamous cell of adenocarcinomas. Adenosquamous carcinomas were placed in the adenocarcinoma group. Special types of tumours with a different prognosis were identified. Tumours were graded empirically as well-differentiated (grade 1); intermediate (grade 2); poorly-differentiated (grade 3). Tumours were divided into those with pushing and infiltrative histological growth patterns. The growth pattern of the tumour was assessed empirically at the advancing front of the tumour.
Correspondence: J. Scurry, Anatomical Pathologist, Hunter Area Pathology Service, and Conjoint Associate Professor, Faculty of Health Sciences, University of Newcastle, Newcastle, NSW 2310, Australia. E-mail: [email protected]
J Obstet Gynaecol Downloaded from informahealthcare.com by Deakin University on 08/10/15 For personal use only.
378
J. Scurry et al.
Pushing tumours had smooth, broad advancing edges and infiltrative tumours had an irregular edge with numerous small tumour nests and tentacles. Cases that appeared intermediate between these extremes were classified as ‘intermediate’ growth pattern. The amount of fibroblastic stromal reaction was assessed empirically as moderate to marked (stroma-rich) or none to mild (stroma-poor). Carcinomas were assessed empirically according to their degree of lymphocytic reaction as lymphocyte-rich (moderate to marked lymphocytic reaction) or lymphocyte-poor (none to mild). The depth of invasion was measured as the perpendicular distance between the basement membrane of the surface epithelium of non-involved cervix to a horizontal line drawn from the deepest invasive cell. In circumferential exophytic tumours, the original cervical wall thickness was assumed to be 15 mm and the depth of invasion estimated by a subtraction method (15 mm – distance from the deepest tumour cell to outermost cervix). Carcinomas were divided into those showing inner-third, middle-third and outer-third invasion of the cervical wall. Parametrial invasion was defined as direct invasion, LVSI, in transit metastasis or lymph node involvement of parametrial fat. Direct involvement of the corpus was defined as tumour histologically extending proximal to the first endocervical cell of the surface or glandular field epithelium. LVSI was identified when a tumour embolus lay entirely within an endothelial-lined space beyond the advancing front of the tumour. It was the laboratory practice to receive fresh specimens from the operating theatre, immediately open them and fix them in abundant formalin. At the time of gross dissection, the entire cervix was embedded in cases of cervical cancer. For this study, blocks with tumour were identified by holding slides up to the light and looking for irregular and/or blue (haematoxylinophilic) areas of tumour. For practicality, when 20 blocks contained tumour, only 20 were examined. Only blocks with the invasive front of tumour were examined, including the blocks with the deepest and largest areas of tumour. When more than one focus of LVSI was seen, no attempt was made to determine whether this was due to multiple tumour emboli in multiple vessels or a single tortuous vessel being involved with one elongated tumour embolus. LVSI was quantified as follows. The fraction of tumour blocks involved with LVSI (extent), maximum number of foci in the worst affected slide density and the maximum distance between an LVSI focus and the advancing front of the tumour (distance), as measured by an eyepiece graticule, were recorded. In those women with LVSI, these forms of quantified LVSI were compared with pelvic lymph node metastases, local and distant recurrence and adjuvant therapy using the t-test and the Wilcoxon unpaired test. For the outcome ‘distant recurrence’, several explanatory variables were tested to see if any was stronger than the others. Logistic regression models were fit using the three LVSI quantification measures and pelvic lymph node metastases. Each model contained only one of these variables and the AICc (Aikake Information Criterion with correction for small samples) was used to test if any variable performed differently.
Figure 1. Two foci of tumour are seen in the myometrium beyond the advancing front. On the left side, the tumour lies within a clearly defined endothelial-lined space and this is counted as a focus of LVSI. On the right side, inflammatory cells obscure any endothelium that might be present and, while LVSI is likely, this focus does not fulfil the criteria for LVSI (Haematoxylin and eosin 200).
adenocarcinomas. Two special types of squamous and four of adenocarcinoma were found. The special types of squamous cell carcinoma comprised one nasopharyngeal and one papillary. The special types of adenocarcinoma comprised one adenoid basal carcinoma combined with carcinosarcoma; one adenoid cystic carcinoma and two glassy cell carcinomas. No small cell carcinomas were found. A total of 65 women had at least one focus of LVSI identified. We encountered two difficult scenarios where it was difficult to determine whether a focus of inflammation should be counted as LVSI or not. The most troublesome were cases which showed foci of inflammation potentially obscuring tumour beyond the advancing edge of the tumour. It was difficult to determine on routine histology whether enlarged cells seen in these foci were reactive endothelial, inflammatory or tumour cells (Figure 1). Such examples were not counted as LVSI unless tumour cells were definitely identified. The other problem was the finding of separate tumour foci beyond the advancing edge of the tumour not explainable by direct growth of tumour or perineural invasion (Figure 2). While it is difficult to conceive how this scenario occurred except by LVSI, such cases were not counted as LVSI
Results The mean age of the 88 women was 46 years. All women were primarily treated by radical hysterectomy and, at least, bilateral pelvic lymphadenectomy. A total of 73 women (83%) received some form of adjuvant therapy (radiation plus or minus chemotherapy). The carcinomas were 59 squamous cell and 29
Figure 2. A tumour focus lies next to a small artery, a typical location of a lymphatic vessel. While LVSI is the most likely explanation, the tumour is not surrounded by an endothelial-lined space meaning that it does not fulfil the criteria for LVSI (Haematoxylin and eosin 100).
Is quantification of lymphovascular space invasion useful in stage 1B2 cervical carcinomas? 379
J Obstet Gynaecol Downloaded from informahealthcare.com by Deakin University on 08/10/15 For personal use only.
Table I. Pathological features vs presence of at least one focus of LVSI. n (n 88)
n with LVSI (n 65)
59 29
42 23
0.4
7 44 37
6 32 27
0.3
20 15 53
10 10 45
0.009
38 50
28 37
1.0
36 52
28 37
0.49
3 12 77
0 4 61
0.001
29 59
24 41
0.3
Type SCC Adeno Grade Grade 1 Grade 2 Grade 3 Growth pattern Pushing Intermediate Infiltrative Stromal reaction Stroma-poor Stroma-rich Lymphocytic reaction Lymphocyte-poor Lymphocyte-rich Depth of cervical wall involvement Inner-third Middle-third Outer-third Corpus involvement Yes No
p value
unless a surrounding endothelium could be seen. By contrast, the well-known mimics of LVSI, translocation and retraction artefact, were rare and easily distinguished from LVSI. The relationship of tumour characteristics to the presence of LVSI is shown in Table I. The presence of LVSI correlated with greater depth of cervical wall invasion and an infiltrative growth pattern. However, occasional examples of deeply invasive tumours without LVSI and pushing growth pattern tumours with much LVSI were encountered. No relationship of LVSI to tumour type, grade, stromal or lymphocytic reaction or corpus involvement was found. A total of 40 women (45%) had positive pelvic lymph nodes: 37 of these 40 (93%) had LVSI and three did not. Of the 65 women with LVSI, LVSI involved a greater fraction of blocks, had a higher mean maximum density in the worst affected slide and was a further distance from the advancing edge of the tumour in those with positive compared with negative pelvic nodes (Table II). A graph of the chances of nodal metastasis compared with the fraction of LVSI blocks/total tumour blocks is shown in Figure 3. Follow-up ranged from 6 to 283 (mean 96) months. Five (6%) women had a local recurrence. Of the 73 women who received adjuvant therapy, 59 had LVSI and 14 did not. Of the 14 who did not receive adjuvant therapy, three of the six with LVSI (3/6) developed local recurrence compared with 1/9 without LVSI (p 0.09). The relationship between local recurrence, LVSI, pelvic lymph node metastases and adjuvant therapy is shown in
Figure 3. Relationship between fraction of tumour blocks involved with LVSI and chance of pelvic lymph node metastases.
Table III. We found that four out of five of the women with local recurrence had not been given adjuvant therapy. A total of 13 (15%) women had distant recurrence. The relationship between distant recurrence, LVSI, pelvic lymph node metastases and adjuvant therapy is shown in Table IV. The usual profile of the women with distant recurrence, seen in 11/13 (85%), was LVSI-positive, lymph node-positive and given adjuvant therapy. Using the AICc, pelvic lymph node metastasis was found to be a stronger explanatory variable than any of the three quantified LVSI measures, i.e. (1) fraction of tumour blocks with LVSI/ total tumour blocks; (2) maximum density of LVSI in worst affected slide and (3) longest distance of LVSI from advancing tumour front. The AICc values were 67.8, 72.3, 72.4 and 74.5, respectively (lower values better).
Discussion Only a few studies have quantified the amount of LVSI. Roman et al. (1998) found that the quantity of LVSI, rather than its mere presence, was a significant risk factor for pelvic lymph node metastases. In their study, the entire cervix was blocked for histological assessment and the amount of LVSI quantified in four ways: (1) the percentage of cervical histopathological sections containing LVSI; (2) the percentage of sections with tumour containing LVSI; (3) the total number of foci with LVSI and (4) the maximal number of foci of LVSI in 10 high powered fields. Only the percentage of cervical sections containing LVSI reached significance. Logistic regression found that LVSI in 45% of all cervical histopathological sections to be a significant predictor of nodal metastases. However, once the pelvic lymph node status is known, is quantifying LVSI any advantage? Chernofsky et al. (2006), using the same cohort and method of LVSI quantitation as Roman and colleagues, answered this question in the affirmative by finding
Table II. Quantitation of LVSI vs pelvic lymph node metastases, local recurrence and distant recurrence in those women who have LVSI present.
Extent Density Distance
PLN (n 37)
PLN (n 28)
p value
LR (n 4)
LR (n 61)
p value
DR (n 12)
DR (n 53)
p value
47 5.9 3.5
29 4 1.4
0.01 0.001 0.001
62 3.3 1.7
40 8.5 2.6
0.4 5.2 4.4
50 10 4.3
39 7.8 2.1
0.02 0.008 0.02
Extent: mean % blocks with LVSI/total blocks of cancer examined. Density: mean number of foci of LVSI per worst affected slide. Distance: mean distance (mm) of furthest LVSI from advancing front of tumour. PLN, pelvic lymph node; LR, local recurrence; DR, distal recurrence.
380
J. Scurry et al.
Table III. Relationship between local recurrence, LVSI, pelvic lymph node status and adjuvant therapy.
J Obstet Gynaecol Downloaded from informahealthcare.com by Deakin University on 08/10/15 For personal use only.
LVSI Pelvic lymph node metastases Adjuvant therapy
Local recurrence (n 5)
No local recurrence (n 83)
4 1 1
61 39 72
time to recurrence was significantly reduced in women with the combination of 29% of sections with LVSI and a total number of foci with LVSI 5. Roman and colleagues’ method (1998), while providing proof of principle, is time-consuming and requires the entire cervix of cancer specimens to be blocked; a practice which is not performed routinely in many laboratories. Sykes et al. (2003) in a study involving one of us (JS), used a much simpler method. They counted the number of foci of LVSI in the most affected slide in node-negative stage 1B1 cervical carcinomas and compared this with recurrence. While there was a trend, their study produced a negative result. Possible reasons why their study failed to reach significance are: (1) their cut-off for mild (1–2 foci of LVSI), moderate (3–9 foci) and severe LVSI ( 9 foci) were arbitrarily set at the beginning rather than being identified statistically at the end of the study and (2) the wrong parameter, density (number of foci per slide) was chosen rather than the extent (number of blocks involved). We found that only two pathological features of the primary tumour correlated with the presence of LVSI: the degree of cervical wall invasion and an infiltrative growth pattern of the advancing front of the tumour. The association of depth with cervical wall invasion is presumably a volume effect. The greater the volume of tumour, the more chances of tumour cells and vessels will approximate and LVSI occur. The correlation of LVSI with an infiltrative advancing front is consistent with higher biological aggressiveness, as has been observed in SCC of other sites. In another study involving one of us (NFH), an infiltrative growth pattern was been found to be associated with an increased risk of recurrence in SCC of the vulva (Heaps et al. 1990). The presence of LVSI strongly correlated with pelvic lymph node metastasis. However, the absence of LVSI is not a reliable guide that pelvic lymph nodes will be negative. Three of the 40 women with lymph node metastasis did not have LVSI. Possible explanations for this small group are: (1) that some tumours may metastasise to nodes via transport of single or small groups of cells gaining access to lymphatics within the tumour without forming the usual intravascular tumour columns visible beyond the confines of the tumour; (2) some patterns of tumour growth beyond the advancing edge, which are true LVSI, are not recognised as such using the standard criterion of a tumour mass completely surrounded by endothelium; and (3) sampling error, as even if all the cervix is submitted, usually only one section per 3–4 mm slice is examined. Table IV. Relationship between distant recurrence, LVSI, pelvic lymph node status and adjuvant therapy. Distant recurrence No distant recurrence (n 13) (n 75) LVSI Pelvic lymph node metastases Adjuvant therapy
12 11 11
53 29 62
When LVSI was quantified, women with a higher fraction of blocks with LVSI (extent), a higher maximum number of foci per block (density) and a further distance of LVSI from the invasive front (distance) were more likely to have positive pelvic nodes. Similarly, Roman et al. (1998) found that a higher fraction of tumour blocks with LVSI predicts an increased likelihood of pelvic lymph node metastases and Memarzadeh et al. (2003) found parametrial as opposed to cervical stromal LVSI carried a higher risk of nodal metastasis. While our study showed quantification of LVSI is very helpful in predicting pelvic lymph node status, this information is of no practical use if the lymph node status is known, as it is in women treated by the surgical standard of radical hysterectomy and pelvic lymphadenectomy. We therefore looked at whether quantification of LVSI gave a better prediction of local and distant recurrence than a yes/no assessment of LVSI. Adjuvant therapy was clearly effective in reducing local recurrence. While the proportion of local recurrences in women not given adjuvant therapy was similar to those not given radiotherapy in the GOG study (27% vs 28%), if adjuvant therapy was given, our local recurrence rate dropped to 1.5% (Sedlis et al. 1999). Quantification of LVSI was of no value in predicting local recurrence. The explanation of this finding is that high-count LVSI women were likely to be node positive and to be given adjuvant therapy, which protected against local recurrence. Distal recurrence correlated with both quantification of LVSI and pelvic lymph node metastases. However, logistic regression using AICc, showed that pelvic lymph node metastasis was a stronger predictor for any form of quantification of LVSI. The importance of positive pelvic nodes as a prognostic factor is well illustrated by Chen et al.’s 2013 study of 588 cancer from stages IA2 to IIA. The 5-year overall survival rate of those with node positive cancer was found to be much lower than those without (22.4% vs 84.4%). The 1B2s in that study were fairly similar to our own cohort in that they contained a high number of cases with multiple node metastases, accounting for a bigger drop in survival than in previous literature. Our results support a stepwise progression of cervical cancer from primary tumour → LVSI → regional pelvic lymph node metastases → distant metastases (Saharinen et al. 2004). This ‘halsteadian’ hypothesis provides the rationale behind current standard surgical and adjuvant therapy protocols. While a jump from primary tumour to distant metastases is possible by (1) inefficient lymph nodes allowing tumour cells to pass through, (2) lymphovenous anastomosis or (3) systemic spread of cancer cells entering the peripheral blood within the tumour, distant spread without pelvic lymph node metastasis is unusual in cervical cancer. While our study has shown no benefit in quantifying LVSI is 1B2 cervical cancers, it may be helpful in smaller cancers, as these are less likely to have lymph node metastases and less likely to have been given adjuvant therapy. Sykes et al. (2003) showed a trend for more LVSI to predict recurrence in the node-negative stage. Further studies are needed to determine the significance or otherwise of their findings. Another potential use of quantifying LVSI is in endometrial cancer as lymphadenectomy is not standard surgery in that cancer. Hachisuga et al. (1999) found that grading of LVSI correlated with lymph node metastasis in endometrial cancer. They graded LVSI on a very simple system of none, one focus and multifocal. A more detailed method of grading LVSI, as used in our study, could be applied to endometrial cancer in order to see whether this could improve the accuracy of nodal metastasis prediction. In summary, we have shown that using relatively simple methods of quantification of LVSI of extent, density and distance
Is quantification of lymphovascular space invasion useful in stage 1B2 cervical carcinomas? 381
J Obstet Gynaecol Downloaded from informahealthcare.com by Deakin University on 08/10/15 For personal use only.
from the tumour front, there is a strong correlation between increased LVSI on all three measures and pelvic lymph node metastases in stage 1B2 cervical cancers. However, since the pelvic lymph node status is known in women treated by standard surgical therapy, quantifying LVSI to predict nodal status is of no benefit. When we attempted to predict local and distant recurrence by quantifying LVSI, we found that adjuvant therapy confounded the prediction of local recurrence and that distant recurrence was better predicted by nodal status. We conclude that there is no benefit in quantifying LVSI in 1B2 cervical cancers when the nodal status is known. Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper. Ellen Barlow was supported by a grant from the GO Research Fund, an account within the Royal Hospital for Women Foundation.
References Chen Y, Zhang L, Tian J, Fu X, Ren X, Hao Q. 2013. Significance of the absolute number and ratio of metastatic lymph nodes in predicting postoperative survival for the International Federation of Gynecology and Obstetrics Stage IA2 to IIA Cervical Cancer. International Journal of Gynecological Cancer 23:157–163. Chernofsky MR, Felix JC, Muderspach LI, Morrow CP, Ye W, Groshen SG et al. 2006. Influence of quantity of lymph vascular space invasion on time to recurrence in women with early-stage squamous cancer of the cervix. Gynecologic Oncology 100:288–293. Delgado G, Bundy B, Zaino R, Sevin BU, Creasman WT, Major F. 1990. Prospective surgical-pathological study of disease-free interval in patients with stage IB squamous cell carcinoma of the cervix: a Gynecologic Oncology Group study. Gynecologic Oncology 38:352–357.
Gauthier P, Gore I, Shingleton HM, Soong SJ, Orr JW Jr, Hatch KD. 1985. Identification of histopathologic risk groups in stage IB squamous cell carcinoma of the cervix. Obstetrics and Gynecology 66:569–574. Hachisuga T, Kaku T, Fukuda K, Eguchi F, Emoto M, Kamura T et al. 1999. The grading of lymphovascular space invasion in endometrial carcinoma. Cancer 86:2090–2097. Hacker NF, Barlow E, Scurry J, Gebski V, Farrell R, Robertson G et al. 2013. Primary surgical management with tailored adjuvant radiation for stage IB2 cervical cancer. Obstetrics and Gynecology 121:765–772. Heaps JM, Fu YF, Montz FJ, Hacker NF, Berek JS. 1990. Surgical-pathologic variables predictive of local recurrence in squamous cell carcinoma of the vulva. Gynecologic Oncology 38:309–314. Landoni F, Maneo A, Colombo A, Placa F, Milani R, Perego P et al. 1997. Randomised study of radical surgery versus radiotherapy for stage IB-IIA cervical cancer. Lancet 350:535–540. Memarzadeh S, Natarajan S, Dandade DP, Ostrzega N, Saber PA, Busuttil A et al. 2003. Lymphovascular and perineural invasion in the parametria: a prognostic factor for early-stage cervical cancer. Obstetrics and Gynecology 102:612–619. Roman LD, Felix JC, Muderspach LI, Varkey T, Burnett AF, Qian D et al. 1998. Influence of quantity of lymph-vascular space invasion on the risk of nodal metastases in women with early-stage squamous cancer of the cervix. Gynecologic Oncology 68:220–225. Saharinen P, Tammela T, Karkkainen MJ, Alitalo K. 2004. Lymphatic vasculature: development, molecular regulation and role in tumor metastasis and inflammation. Trends in Immunology 25:387–395. Sedlis A, Bundy BN, Rotman MZ, Lentz SS, Muderspach LI, Zaino RJ. 1999. A randomized trial of pelvic radiation therapy versus no further therapy in selected patients with stage IB carcinoma of the cervix after radical hysterectomy and pelvic lymphadenectomy. A Gynecologic Oncology Group Study. Gynecologic Oncology 73:177–183. Smiley LM, Burke TW, Silva EG, Morris M, Gershenson DM, Wharton JT. 1991. Prognostic factors in stage IB squamous cervical cancer patients with low risk for recurrence. Obstetrics and Gynecology 77:271–275. Sykes P, Allen D, Cohen C, Scurry J, Yeo D. 2003. Does the density of lymphatic vascular space invasion affect the prognosis of stage Ib and IIA node negative carcinoma of the cervix? International Journal of Gynecological Cancer 13:313–316.
