Magnetic resonance imaging predicts survival and occult metastasis in oral cancer: a dual-centre, retrospective study

Magnetic resonance imaging predicts survival and occult metastasis in oral cancer: a dual-centre, retrospective study

Available online at www.sciencedirect.com British Journal of Oral and Maxillofacial Surgery 51 (2013) 696–701 Magnetic resonance imaging predicts su...

425KB Sizes 0 Downloads 32 Views

Available online at www.sciencedirect.com

British Journal of Oral and Maxillofacial Surgery 51 (2013) 696–701

Magnetic resonance imaging predicts survival and occult metastasis in oral cancer: a dual-centre, retrospective study Paul W. Boland a,∗ , Steve R. Watt-Smith b , Colin Hopper c , Stephen J. Golding a a b c

Radiology Group, Nuffield Department of Surgery, University of Oxford, John Radcliffe Hospital, Headley Way, Headington, Oxford OX3 9DU, UK Department of Oral & Maxillofacial Surgery, John Radcliffe Hospital, Headley Way, Headington, Oxford OX3 9DU, UK Unit of Oral and Maxillofacial Surgery, UCL Eastman Dental Institute, 265 Grays Inn Road, London WC1X 8LD, UK

Accepted 3 September 2013 Available online 27 September 2013

Abstract The purpose of this study was to investigate the effectiveness of tumour variables measured on magnetic resonance imaging (MRI) to predict 2-year disease-related survival and occult cervical lymph node metastasis in oral carcinoma. In this retrospective, dual-centre study the volume and thickness of tumours were measured using archived MRI staging scans of 199 patients who had curative primary resection for histologically confirmed oral carcinoma. Tumour volume predicted survival when grouped using the median (3.0 cm3 , HR 3.41, p 0.005) and first and third quartiles (0.5 cm3 , HR 8.22, p 0.04; 8.0 cm3 , HR 18.6, p 0.005). Tumour thickness predicted survival using a median of 11.0 mm (HR 2.65, p 0.02). Volume predicted occult cervical lymph node metastasis using a median of 3.0 cm3 (HR 5.02, p < 0.001) and quartiles of 0.5 cm3 (HR 6.92, p = 0.01) and 8.0 cm3 (HR 11.3, p 0.005); thickness predicted it using a median of 11.0 mm (HR 4.39, p 0.002) and quartiles of 4.0 mm (HR 4.33, p 0.06) and 16 mm (HR 11.9, p 0.003). The thickness and volume of tumour measured on MRI may predict 2-year disease-related survival and occult cervical lymph node metastasis in oral cancer. © 2013 The British Association of Oral and Maxillofacial Surgeons. Published by Elsevier Ltd. All rights reserved. Keywords: Oral cancer; Magnetic resonance imaging; Staging; Maxillofacial surgery; Survival; Metastasis

Introduction The TNM staging of oral cancer based on surface diameter has been reported as being inadequate.1–4 Diameter is also not necessarily related to prognosis. Large superficial tumours can be managed well, but some small, deep tumours evade locoregional control. Although most oral cancers are classified as T2, it is unlikely that such a varied group of tumours has a similar prognosis.1 Tumours larger than 4 cm usually invade surrounding structures and are designated T4, which leaves a paucity of T3 tumours.1–5 There is a need to look for variables such as the volume and thickness of tumour to replace the existing TNM staging system.



Corresponding author. E-mail address: [email protected] (P.W. Boland).

Occult cervical lymph node metastasis is found in 30–40% of patients who have no sign of metastasis on clinical and imaging examination.6–9 As nodal metastasis is associated with a 50% reduction in 5-year survival, current practice is to do elective neck dissection in those with no sign of metastasis in the neck.10 Accordingly, a large number of patients who have this operation have no pathological evidence of regional metastases. There is a need to identify those with a low risk of occult cervical lymph node metastases to spare them this procedure. Tumour depth has been reported to predict occult cervical metastasis and survival with low-risk cut-points that vary from 1.5 to 12 mm.7–9,11,12 Most of these studies use histological measurement of the depth of the tumour after resection, but an alternative is to use imaging. Practically, depth, which is defined as the distance from the reconstructed mucosal surface to the point of deepest invasion, cannot be measured on MRI, so thickness (the distance from the surface to the point

0266-4356/$ – see front matter © 2013 The British Association of Oral and Maxillofacial Surgeons. Published by Elsevier Ltd. All rights reserved.

http://dx.doi.org/10.1016/j.bjoms.2013.09.001

P.W. Boland et al. / British Journal of Oral and Maxillofacial Surgery 51 (2013) 696–701

of deepest invasion) is used as a surrogate, and the ability of MRI to evaluate it has been validated.13–15 Preliminary work by our group has shown that the volume of a tumour measured on MRI predicts occult cervical lymph node metastasis and survival in lingual carcinoma using cut-points of 3.0 cm3 and 8.0 cm3 , respectively.16 The purpose of this study was two-fold. First, we aimed to investigate whether tumour volume and thickness measured on MRI could be used to predict 2-year disease-related survival, and whether it could be used in a refined TNM staging system. Secondly, we aimed to use these data to predict occult cervical lymph node metastasis and to propose a new decision rule for elective neck dissection.

Method The sample was gathered from two tertiary centres: The John Radcliffe Hospital, Oxford, and University College London Hospital. We retrospectively studied 397 consecutive patients diagnosed with histologically confirmed squamous cell carcinoma (SCC) of the oral cavity from 1998 to 2008 at the John Radcliffe, and from 2006 to 2008 at University College. Patients were identified from tumour registries. All patients with tumours staged using MRI and treated with curative resection were included. No patient had preoperative chemotherapy or radiotherapy. Patients whose MRI studies could not be recovered (n = 70, 18%) or were inadequate (n = 17, 4%), those who had had biopsy examinations of the tumour less than 7 days before MRI (n = 17, 4%), those with previous or synchronous occult cervical lymph node metastases (n = 5, 1%), and those whose records were missing or inadequate (n = 58, 15%), were excluded. Patients with primary SCC of the lip (n = 31, 8%) were also excluded as these lesions are rarely imaged. A total of 199 patients were included (mean age 61.8 years, range 22–92). Patients’ details are summarised in Table 1. All measurements were done in duplicate by a single observer and averaged. A consultant radiologist who specialises in the imaging of occult cervical lymph node metastases reviewed selected studies. The images were anonymised and read in random order, and at least one week was allowed to pass between readings to minimise recall. In most cases T2-weighted spin-echo and fast spin-echo pulse sequences were used to measure volume and thickness. However, where T2-weighted studies were not available, short tau inversion recovery unenhanced and enhanced T1-weighted studies were used. Variables of MRI such as the size of the pixel matrix, field-of-view, and thickness of the slices, varied from patient to patient. Scans were deemed inadequate if imaging artefacts obscured the region of interest, or if the field-of-view did not include the entire tumour. All measurements used ImageJ software (V. 1.41o, National Institutes of Health, USA). Volume was calculated using manual segmentation. Measurements of thickness were taken perpendicular to the

697

Table 1 Patients’ details. Missing cases are shown in brackets in column 1. No. (%) (n = 199) Hospital John Radcliffe University College Sex Male Female Smoking (5) Yes No Alcohol (6) Yes No Site Oral tongue Floor Lower alveolus Upper alveolus Hard palate RMT Buccal Neck dissection (1) None Selective (I–III) Modified radical Bilateral Tumour cT 1 2 3 4 pT (1) 1 2 3 4 cN 0 1 2 pN (1) 0 1 2

177(89) 22 (11) 102 (51) 97 (49) 104 (54) 90 (46) 86 (45) 107 (55) 97 (49) 31 (16) 26 (13) 14 (7) 2 (1) 18 (9) 11 (6) 27 (14) 63 (32) 81 (41) 27 (14)

64 (32) 65 (33) 9 (5) 61 (31) 79 (40) 46 (23) 11 (6) 62 (31) 136 (68) 60 (30) 3 (2) 129 (65) 20 (10) 49 (25)

surface of the tumour to the deepest extent of invasion. In cases where histologically confirmed occult cervical metastasis was not visible, a volume and thickness of 0 cm3 and 0 mm were assigned. Examples of measurements are shown in Fig. 1. Depth of disease and staging were obtained from archived medical records. Occult cervical lymph node metastasis was defined as metastatic SCC of the oral cavity in a patient classified as N0 after clinical examination and staging with MRI. A total of 144 patients had ipsilateral elective neck dissection and 27 had bilateral neck dissection. In the remaining 27, the absence of cervical metastases was inferred from the lack of regional recurrence after 2 years’ follow-up. Of 136 patients initially

698

P.W. Boland et al. / British Journal of Oral and Maxillofacial Surgery 51 (2013) 696–701 Table 2 Disease-related survival. Volume and thickness are categorised using the mean measurement and their first and third quartiles. Values could not be computed for thickness or pathological stage as no events took place in the low-risk group. HR: hazard ratio; CI: confidence interval. Event

Fig. 1. Example of the measurement of tumour volume and thickness on MRI. The image depicts the manual segmentation of a lingual SCC from a single MRI slice using a yellow outline. An example of the measurement of thickness is shown in red.

classified as cN0, 25 (18%) were found to have occult cervical lymph node metastases on pathological examination. All periods of survival were measured from the date of primary resection. Patients were censored if they were lost to follow-up or if they had no documented event at their last outpatient visit. A disease-related survival event was defined as any death in the 2-year follow-up period, which was confirmed on review of the chart to be a result of the disease, with the exception of perioperative deaths related to surgical complications. Eight (4%) patients were excluded from the survival analysis because they died in the perioperative period, and one was excluded because he was lost to follow-up 19 days postoperatively. In total, there were 29 disease-related deaths. Mortality data were obtained from patients’ charts. R-project (V. 2.9.1, The R Foundation) was used for all statistical computation. Prediction of occult cervical lymph node metastasis was tested using logistic regression. Survival was tested using Cox proportional hazards (hazard ratios (HR), and 95% confidence intervals (95% CI)). A 95% CI greater than 1 was deemed significant. Correlation between ordinal variables was evaluated using the Spearman correlation test. Probabilities of 0.05 were considered significant. Ethical approval for this chart audit, which had no direct impact on patient care, was not required.

Results There was a strong correlation between tumour thickness measured on MRI (median (IQR) 11.2 (4.4–16.4) mm) and depth measured histologically (median (IQR) 10.0 (4.8–15.0) mm) (r = 0.79, p < 0.001). Disease-related survival was significantly predicted by tumour volume when grouped using the 3.0 cm3 cut-point. The results were also significant when divided using the

Censored

Volume (2 groups) (cm3 ) ≤3.0 7 90 >3.0 22 71 Volume (3 groups) ≤0.5 1 50 >0.5, ≤8.0 14 80 >8.0 14 31 Thickness (2 groups) (mm) ≤11.0 8 86 >11.0 21 75 Thickness (3 groups) ≤4.0 0 43 >4.0, ≤16.0 12 77 >16.0 17 41 cT class 1 5 58 2 9 54 3 2 7 4 13 42 cN class 0 8 124 1 19 36 2 2 1 pT class 1 5 72 2 5 41 3 2 8 4 17 39 pN class 0 4 122 1 8 11 2 17 26 Pathological stage 1 1 64 2 0 27 3 4 14 4 24 54

HR – 3.41 – 8.22 18.55 – 2.65 – – –

95% CI

p-Value

– 1.46–7.98

– 0.005

– 1.08–62.53 2.44–141.17

– 0.04 0.005

– 1.18–6

– 0.02

– – –

– – –

– 1.85 2.54 3.06

– 0.62–5.53 0.49–13.10 1.09–8.60

– 0.27 0.27 0.03

– 6.63 39.45

– 2.90–15.14 7.77–200.33

– <0.001 <0.001

– 1.80 3.09 5.31

– 0.52–6.23 0.60–15.94 1.96–14.39

– 0.35 0.18 0.001

– 15.77 17.22

– 4.74–52.45 5.78–51.30

– <0.001 <0.001

– – 14.91 24.17

– – 1.67–133.39 3.27–178.75

– – 0.02 0.002

0.5 cm3 and 8.0 cm3 cut-points. Tumour thickness was a significant predictor of disease-related survival using the 11.0 mm cut-point. Statistics could not be computed when the thickness was grouped by the 4.0 mm and 16.0 mm cutpoints as there were no events in the group with tumours less than 4.0 mm thick. Only those tumours classified as cT4 and pT4 predicted disease-related survival using conventional T-classification. Clinical N-classification was strongly significant for cN1 and cN2 tumours. Pathological N-classification was also significant for pN1 and pN2 tumours (Table 2 and Fig. 2). Tumour volume was a significant predictor of occult cervical lymph node metastasis when grouped using the 3.0 cm3 cut-point. When grouped into 3 categories using the 0.5 cm3 and 8.0 cm3 cut-points, the relation remained significant. The negative predictive value for volumes of less than 3.0 cm3 was 90% (70/78 cases) and for those of less than 0.05 cm3 it was

P.W. Boland et al. / British Journal of Oral and Maxillofacial Surgery 51 (2013) 696–701

699

Fig. 2. Kaplan–Meier curves depicting disease-related survival and the volume and thickness of tumour derived from MRI measurements. Results are for median and first and third quartiles.

Table 3 Occult cervical lymph node metastasis. Volume and thickness are categorised using the mean measurement and their first and third quartiles. OR: odds ratio. CI: confidence interval. Volume (2 groups) (cm3 ) ≤3.0 >3.0 Volume (3 groups) ≤0.5 >0.5, ≤8.0 >8.0 Thickness (2 groups) (mm) ≤11.0 >11.0 Thickness (3 groups) ≤4.0 >4.0, ≤16.0 >16.0 cT class 1 2 3 4 pT class 1 2 3 4

No metastasis

Metastasis

78 33

8 17

45 52 14

OR

95% CI

p-Value

– 5.02

– 1.97–12.78

– <0.001

2 16 7

– 6.92 11.25

– 1.51–31.78 2.09–60.49

– 0.01 0.005

70 41

7 18

– 4.39

– 1.69–11.40

– 0.002

38 57 16

2 13 10

– 4.33 11.88

– 0.93–20.30 2.33–60.41

– 0.06 0.003

50 40 3 18

7 8 1 9

– 1.43 2.38 3.57

– 0.48–4.28 0.22–26.18 1.16–11

– 0.52 0.48 0.03

63 24 4 19

7 7 2 9

– 2.62 4.50 4.26

– 0.83–8.28 0.69–29.15 1.4–12.97

– 0.10 0.12 0.01

700

P.W. Boland et al. / British Journal of Oral and Maxillofacial Surgery 51 (2013) 696–701

96% (43/45). Tumour thickness divided using the 11.0 mm cut-point was significant. When grouped into 3 categories, it was significant for those thicker than 16.0 mm while the intermediate group showed a trend towards significance. The negative predictive value for tumours less than 11.0 mm thick was 90% (63/70) while for those less than 4.0 mm it was 94% (36/38). Regarding conventional T-classification, only those tumours classified as cT4 or pT4 were significant predictors of occult cervical lymph node metastasis. Age, sex, institution, smoking, and alcohol were not significant. Results are shown in Table 3.

Discussion We have shown that MRI-derived tumour volume is a predictor of 2-year disease-related survival when volume is divided into categories using a median of 3.0 cm3 and quartiles of 0.5 and 8.0 cm3 . Grouping the cases into quartiles provides low, intermediate, and high-risk groups with respective 2year mortalities of 2%, 15%, and 31%. For thickness quartiles of 4.0 and 16.0 mm, there was a similar stratification with mortalities of 0%, 14%, and 29%. Survival in the T2 and T3 tumour groups was no different from that in the T1 group. Even within the cT1 and pT1 groups, 9%and 7%, respectively, had disease-related survival events, substantially higher than the low risk volume and thickness groups, which suggested that MRI-based measurement may be superior at identifying low risk patients. Pathological tumour stage (pS) predicted survival with one event in 65 patients (2%) in the pS1 group, compared with 24 in 78 patients (31%) in the pS4 group. This was driven by the high prognostic value of the nodal status. However, the accuracy of thickness and volume measured on MRI compared with the pathological result suggests that an MRI-derived staging system may be superior to TNM staging. Further study is warranted. The volume and thickness of the tumour are significant predictors of occult cervical lymph node metastasis. When volume was grouped using quartiles 0.5 and 8.0 cm3 it resulted in distinct low, intermediate, and high-risk groups. While the intermediate risk group (thickness of 4.0 or 16.0 mm) was not significantly different from the low risk group, the result approached significance. Tumours with a volume or thickness that was less than the first quartile had a 4.3% and 5% chance of metastasis, respectively. These patients may be spared elective neck dissection. The risk of occult cervical lymph node metastasis was less defined using the T-classification. The risk of metastases in T2 and T3 tumours was no higher than that in T1 tumours. The risk of occult cervical lymph node metastasis remained substantial in the cT1 and pT1 groups at 12.3% and 10%, respectively, which suggests that the volume and thickness of tumour may define low risk patients accurately. Despite the benefit of using depth and thickness as robust predictors of survival and occult cervical lymph node metastasis, the American Joint Committee on Cancer (AJCC) has

been slow to adopt them, probably because of serious disagreement among authors about the appropriate values with which to define groups stratified by risk. Suggested cut-points range from 1.5 to 12 mm.12 Tumour thickness and pathological depth were strongly correlated, which suggests that MRI is a valid surrogate for pathological staging. Thickness was slightly greater than pathological depth, but this is expected given that specimens shrink with fixation, and different techniques are used for measurement. We have previously shown that MRI-derived tumour volume predicts occult cervical lymph node metastasis and survival in lingual carcinomas for cut-points of 3.0 cm3 and 8.0 cm3 , respectively.16 In a retrospective study of 17 patients, Chew et al. showed that MRI-derived tumour volume predicts survival using a cut-point of 13.0 cm3 .17 However, Yuen et al. found that postoperative histopathological analysis gave no prognostic value.18 Pathological measurement however, is subject to an estimated 30% shrinkage of tumour volumes in formalin. Preda et al. also found no prognostic value when they calculated tumour volume on MRI based on the assumption of an elliptical shape.13 The manual segmentation used in the present study accounted for irregularities in contour that were not addressed by the assumption of an ellipse, and it is a more accurate method of measuring volume. All these studies were limited by small samples. Our sample of 199 is the largest used to study MRI-based measurement of dimensions of oral cancer. Our findings that tumour volume and thickness measured on MRI are predictors of occult cervical lymph node metastasis agree partly with our previously published study, which showed that volume, but not thickness, was a predictor. This may reflect the increased sample size, and it also agrees with Iwai et al., who showed that measurement of tumour thickness with MRI was significant. Measurements of volume and thickness on MRI have the potential to provide valuable information that will guide preoperative decisions. They are also superior to the current TNM staging system. Classification using the cut-points of 0.5 and 8.0 cm3 for volume and 4 and 16 mm for thickness provides more accurate risk stratification than the T-classification. Also, patients who are in low risk groups for tumour volume and thickness and who have no sign of metastases in the neck, have a lower risk of occult cervical lymph node metastasis than those with T1 tumours. Measurement of tumour volume and thickness may identify those who can be spared neck dissection in favour of a watch-and-wait approach. Our study has several shortcomings. Measurements were relatively subjective, but this reflects day-to-day practice in the imaging department. Manual measurement of volume usually takes minutes while measurement of thickness takes seconds, and it is likely to be beyond the scope of routine radiology practice. However, automated segmentation would improve the speed of tumour volumetry making the process more practical. In the present study volume and thickness are equally effective in the prediction of occult cervical lymph

P.W. Boland et al. / British Journal of Oral and Maxillofacial Surgery 51 (2013) 696–701

node metastasis and survival. Given that thickness can be measured simply and relatively fast it seems to be the metric of choice. Even though the sample size was 199 it remains small. The oral cavity is composed of several subsites that differ in natural history. However, study of these subsites in isolation would result in sample sizes of insufficient statistical power. Exclusion of patients with missing or inadequate MRI images or records introduces bias to this consecutive group as it favours more recent cases. Measurement bias could have been improved by the use of 2 or more independent investigators. The scan variables we used for MRI were heterogeneous and took place over 10 years. The 2-year follow-up period was short and limited the ability of the study to suggest changes to current practice. However, most disease progression and deaths are expected early in cases of occult cervical lymph node metastasis. Stell found that the median time to recurrence of oral cancer was 24 weeks (95% CI 19–28).19

Conclusion We have shown that tumour volume and thickness measured on MRI is significantly related to 2-year disease-related survival and occult cervical lymph node metastasis. Tumour volume and thickness, when grouped using the quartiles 0.5 and 8.0 cm3 , and 4.0 and 16.0 mm, may indicate low, intermediate and high risk groups for survival and predict occult cervical lymph node metastasis. As these groupings may stratify risk more accurately than conventional TNM classification, tumour volume and thickness show promise as novel staging variables. In our study the incidence of occult cervical lymph node metastasis was less than 5% in patients who had no sign of metastases in the neck and tumour volume and thickness in low-risk categories. These patients may be suitable for a watch-and-wait approach, and elective neck dissection could be avoided. Shortcomings in methodology limit clinical application however, and further prospective work is warranted.

Conflict of interest statement None declared.

701

References 1. Sobin LH, Gospodarowicz M, Wittekind C, editors. TNM classification of malignant tumours. 7th ed. Chichester: John Wiley and Sons; 2009. 2. Hibbert J, Marks NJ, Winter PJ, et al. Prognostic factors in oral squamous carcinoma and their relation to clinical staging. Clin Otolaryngol Allied Sci 1983;8:197–203. 3. Krause CJ, Lee JG, McCabe BF. Carcinoma of the oral cavity. Arch Otolaryngol 1973;97:354–8. 4. Lee JG, Krause CJ. Radical neck dissection: elective, therapeutic, and secondary. Arch Otolaryngol 1975;101:656–9. 5. Moore C, Flynn MB, Greenberg RA. Evaluation of size in prognosis of oral cancer. Cancer 1986;58:158–62. 6. Boyle JO, Strong EW, Shah JP. Oral cavity cancer. In: Shah JP, editor. Cancer of the head and neck. Hamilton: BC Decker; 2001. p. 100–25. 7. Asakage T, Yokose T, Mukai K, et al. Tumor thickness predicts cervical metastasis in patients with stage I/II carcinoma of the tongue. Cancer 1998;82:1443–8. 8. O-charoenrat P, Pillai G, Patel S, et al. Tumour thickness predicts cervical nodal metastases and survival in early oral tongue cancer. Oral Oncol 2003;39:386–90. 9. Hayashi T, Ito J, Taira S, et al. The relationship of primary tumor thickness in carcinoma of the tongue to subsequent lymph node metastasis. Dentomaxillofac Radiol 2001;30:242–5. 10. Chong V. Oral cavity cancer. Cancer Imaging 2005;5:S49–52. 11. Po Wing Yuen A, Lam KY, Lam LK, et al. Prognostic factors of clinically stage I and II oral tongue carcinoma – a comparative study of stage, thickness, shape, growth pattern, invasive front malignancy grading, Martinez-Gimeno score, and pathologic features. Head Neck 2002;24:513–20. 12. Pentenero M, Gandolfo S, Carrozzo M. Importance of tumor thickness and depth of invasion in nodal involvement and prognosis of oral squamous cell carcinoma: a review of the literature. Head Neck 2005;27:1080–91. 13. Preda L, Chiesa F, Calabrese L, et al. Relationship between histologic thickness of tongue carcinoma and thickness estimated from preoperative MRI. Eur Radiol 2006;16:2242–8. 14. Iwai H, Kyomoto R, Ha-Kawa SK, et al. Magnetic resonance determination of tumor thickness as predictive factor of cervical metastasis in oral tongue carcinoma. Laryngoscope 2002;112:457–61. 15. Lam P, Au-Yeung KM, Cheng PW, et al. Correlating MRI and histologic tumor thickness in the assessment of oral tongue cancer. Am J Roentgenol 2004;182:803–8. 16. Boland PW, Watt-Smith SR, Pataridis K, et al. Evaluating lingual carcinoma for surgical management: what does volumetric measurement with MRI offer. Br J Radiol 2010;83:927–33. 17. Chew MH, Khoo JB, Chong VF, et al. Significance of tumour volume measurements in tongue cancer: a novel role in staging. ANZ J Surg 2007;77:632–7. 18. Yuen AP, Lam KY, Wei WI, et al. A comparison of the prognostic significance of tumor diameter, length, width, thickness, area, volume, and clinicopathological features of oral tongue carcinoma. Am J Surg 2000;180:139–43. 19. Stell PM. Time to recurrence of squamous cell carcinoma of the head and neck. Head Neck 1991;13:277–81.