Objective  To evaluate pretreatment tumour volume as a predictor of survival in patients with cervical cancer using both endovaginal and external coil magnetic resonance imaging in order to achieve high spatial resolution and delineate small volume disease.

Design  A retrosfection case series.

Setting  A tertiary referral centre for gynaecological oncology.

Population/Sample  One hundred and six consecutive women with invasive carcinoma of the cervix underwent endovaginal and external coil magnetic resonance imaging on a 0.5-T or 1.5-T scanner.

Methods  T2-W FSE images, sagittal and transverse to the cervix, were obtained and tumour volume was calculated on the sagittal images by the standard technique of multiplying the sum of the areas by the slice thickness. Patients were treated in accordance with normal clinical practice and their subsequent outcome was recorded. The relationships between clinical or imaging parameters and survival were assessed with Cox's proportional hazard method.

Main outcome measures  Disease-free survival.

Results  In 89 of these women, the tumour was Stage I and 88 of the 106 were treated principally by surgery. The median tumour volume was 4.75 cm3 (upper and lower quartiles 22 and 0.6). The median length of follow up of surviving patients was 223 weeks (quartiles 158 and 274 weeks). Stage, treatment type, lymphovascular space involvement, invasion of the parametrium, closeness of the excision margin, lymph node metastases, and magnetic resonance imaging measurements of tumour volume, parametrial invasion and lymph node disease were all significantly associated with survival in univariate analysis. Only magnetic resonance imaging measurement of tumour volume remained consistently and strongly associated with survival after multivariate analysis of parameters available prior to treatment (P= 0.001, Wald statistic 10.74). A receiver operating characteristic curve of tumour volume and disease-free survival confirmed the utility of this investigation and suggested that a cutoff around 13.0 cm3 would predict survival with a positive predictive value of 0.93 and a negative predictive value of 0.75.

Conclusion  Magnetic resonance imaging assessment of tumour volume using both an endovaginal and an external coil approach provides an accurate prediction of prognosis in cervical cancer and defines a population of women at high risk of recurrence and death. The predictive value of this investigation is superior to the clinical and histological parameters previously used. Use of this technique permits a more accurate choice of treatment options. These results suggest that it is the size of tumour burden that determines the outcome rather than invasion beyond the anatomical margins of the uterus.