Accuracy of magnetic resonance imaging in predicting residual disease in patients treated for stage IB2/II cervical carcinoma with chemoradiation therapy

Correlation of Radiologic Findings With Surgicopathologic Results

Authors


  • We thank Lorna Saint Ange for editing.

Abstract

BACKGROUND

The evaluation of residual disease (RD) after chemoradiation therapy (CRT) in stage IB2/II cervical carcinoma conventionally is based on a clinical examination and magnetic resonance imaging (MRI) performed 3 to 8 weeks after the end of treatment. Very few studies have correlated MRI and histologic findings specifically in cervix cancer. This was the objective of the current study.

METHODS

A retrospective review was undertaken of patients who fulfilled the following inclusion criteria: 1) stage IB2/II cervical cancer, 2) external radiotherapy (45 grays [Gy]) was given with concomitant chemotherapy followed by uterovaginal brachytherapy (15 Gy), 3) MRI studies were obtained between 3 weeks and 8 weeks after brachytherapy, and 4) completion surgery included at least a hysterectomy. Postsurgical histologic findings and MRI results were compared.

RESULTS

Forty-four patients who were treated between 2003 and 2006 fulfilled all inclusion criteria. Twelve patients (27%) had achieved a radiologic complete response, 16 patients (36.5%) had unclassified lesions (RD or ‘fibrosis’), and 16 patients (36.5%) had radiologic residual tumor. According to the histologic results, 19 patients (43%) had no RD, 10 patients (23%) had inframillimeter RD, 2 patients (5%) had RD that measured <1 cm, and 13 patients (29%) had RD that measured >1 cm. The sensitivity and specificity of MRI in evaluating RD were 80% and 55%, respectively.

CONCLUSIONS

The current results indicated that the evaluation of RD 3 to 8 weeks after CRT with MRI is difficult, and the risk of false-positive results is high. Another radiologic procedure or a more technologically advanced MRI technique, such as diffusion-weighted MRI, should be evaluated. Cancer 2008. © 2008 American Cancer Society.

Two management options are proposed for the treatment of cervical cancers that measure >4 cm: chemoradiation therapy (CRT) or neoadjuvant chemotherapy. A European randomized trial comparing CRT and neoadjuvant chemotherapy in cervical carcinoma is ongoing (European Organization for Research and Treatment of Cancer Trial 55994). Nevertheless, CRT is considered the standard treatment for bulky cervical cancer (stage IB2 or greater according to the International Federation of Gynecology and Obstetrics [FIGO] classification) by many North American and Western European teams.1

The place of completion surgery after CRT still is under debate.2 Several teams, particularly in Europe, have evaluated response at the end of CRT so that completion surgery can be considered for the subgroup of patients who have ‘residual cervical disease.’3–6 This evaluation is based on clinical examination and imaging studies, especially magnetic resonance imaging (MRI). However, to date, very few investigators have correlated the MRI appearance and histologic findings after CRT in patients who have residual disease. This was the objective of the current study.

MATERIALS AND METHODS

Data concerning women who were treated with CRT followed by pelvic surgery for locally advanced cervical cancer between February 2003 and July 2006 were reviewed retrospectively. Patients who fulfilled the following inclusion criteria were included in the current study: 1) stage IB2 or II cervical carcinoma according to the FIGO classification with no suspicious para-aortic lymph nodes on initial abdominopelvic MRI or computed tomography (CT) scans; 2) histologic subtype either squamous cell, adenocarcinoma, or adenosquamous; 3) patients had received external beam radiation therapy (45 grays [Gy]) delivered to the pelvic cavity and platinum-based, concomitant chemotherapy (if cisplatin was used, then the dose had to be 40 mg/m2 per week); 4) patients had received brachytherapy at a dose of 15 Gy according to International Commission on Radiation Units recommendations7 after CRT (some patients with stage II disease and/or bulky pelvic lymph nodes on initial imaging studies had received a lateral pelvic boost of 15 Gy); 5) patients underwent pelvic (or abdominopelvic) MRI at least 3 weeks after the end of brachytherapy; and 6) patients underwent completion surgery, including at least a hysterectomy (or cervicectomy in patients who had a previous history of subtotal hysterectomy) with or without a pelvic and/or para-aortic lymphadenectomy. The timing between MRI and surgery was not to exceed 80 days. The radicality of the hysterectomy was dependent on the presence of residual disease.8

Image Analysis

MRI studies were analyzed blindly and independently by 2 senior radiologists with experience in the imaging of gynecologic malignancies. They were unaware of the clinical, histologic, and imaging findings of the patients. MRI studies were obtained performed on a 1.5 T Magnet (Signa Excite; GEMS, Milwaukee, Wis) with a phased-array coil and included axial, sagittal, and coronal T2-weighted images of the pelvis and the abdomen. Three-dimensional, T1-weighted images with fat suppression and dynamic sequences of the pelvis were acquired after the injection of gadolinium chelate.

Results of this radiologic analysis were interpreted with a consensus between the 2 senior radiologists. Pretreatment MRI studies were consulted if available by both readers.

Imaging findings were classified into 3 groups according to the appearance of the cervix:

  • Normal in patients who achieved total restoration of the normal radioanatomy of the cervix (1 criterion for a normal cervix was restoration of the hypointense signal of the cervix with no early contrast uptake);

  • Uncertain in patients who had unclassified modifications during imaging, such as radiation therapy-induced sequels, fibrosis, and suspicious (but uncertain) images suggestive of residual disease (a heterogeneous, hyperintense signal on the cervix or moderate uptake after injection); and

  • Certain in patients who had lesions that were considered ‘residual disease’ by radiologists (a lesion was considered certain if a hyperintensity mass on T2-weighted images was visible on the cervix, parametria, or vagina or if rapid contrast uptake was observed in the cervix.9–11)

MRI studies before and after treatment were compared to facilitate the analysis and the conspicuity of residual disease.

Clinical Gynecologic Examination

A clincial gynecologic examination was evaluated in the same week (few days before or after) that a posttreatment MRI study was obtained. The results of this examination were classified according to the same criteria that were used for the radiologic examination:

  • Normal if the gynecologic examination showed no change;

  • Uncertain in patients who exhibited an unclassified change during macroscopic examination of the cervix and/or pelvic touches; and

  • Certain in patients who had unambiguous residual disease in the cervix during the clinical examination.

In some patients, cytologic and biopsy samplings were obtained. Clinical categorization of the gynecologic examination was based on the retrospective data described on the report of the clinical examination at the end of treatment.

Histologic Analysis

During the period of this study, patients underwent completion surgery regardless of the results of their clinical and radiologic examinations (except for patients who were included in a randomized trial that opened during the study period, in whom patients who had complete clinical, radiologic, and cytologic responses after CRT were randomized between a ‘systematic extrafascial hysterectomy’ group and a ‘no hysterectomy’ group). Therefore, except for the patients who were included in that randomized trial, biopsy or cytology results did not modify the indication for subsequent surgery.

All specimens that were removed during the surgical procedure were submitted for histologic analysis. Four groups of patients were defined according to the histological findings in the cervix (and uterus if present):

  • 1)No residual disease in patients who exhibited a complete histologic response after CRT;
  • 2)Patients with limited residual disease in the form of isolated cells or residual disease that measured ≤1 mm;
  • 3)Patients who had residual disease that measured <1 cm; and
  • 4)Patients who had residual disease that measured ≥1 cm.

Statistical Analysis

The sensitivity, specificity, and positive and negative predictive values of MRI were determined by using the histologic results as the reference standard. The histologic results were divided into 2 groups for these calculations: patients with histologic residual disease <1 cm and patients with histologic residual disease >1 cm. The others patients (without histologic residual disease or with residual disease in the form of isolated cells) were classified either as free of residual disease or with inframillimeter residual disease.

The accuracy of MRI could not be determined based on the inclusion of all patients, because the results of this examination were not classified into 2 groups (‘positive’ or ‘negative’). However, because the MRI results were classified into 3 groups, MRI accuracy was evaluated including the cases in which the conclusion of the reviewed MRI was contributive (the MRI was considered ‘normal’ or showed ‘residual disease’). Cases in which the reviewed MRI was ‘uncertain’ were not included in the determination of the accuracy of MRI. Thus, 28 patients (12 with ‘normal’ MRI studies and 16 with an MRI appearance synonymous with radiologic residual tumor) were included in the analysis to determine MRI accuracy. The accuracy of the clinical examination (compared with histologic results) was determined using the same methodology: Patients who had an ‘uncertain’ clinical examination (13 patients) were excluded from the calculation.

RESULTS

During the study period, 29 patients who fulfilled the 5 inclusion criteria described above were included in a randomized trial comparing ‘hysterectomy’ versus ‘no hysterectomy’ after CRT. Thus, 14 of these patients did not undergo hysterectomy (inclusion criteria 6) and were not included in the current study, because no histologic analysis of the uterus specimen was available. Except for these patients, all others patients who were treated during the study period and fulfilled all inclusion criteria were studied.

Forty-four patients fulfilled the inclusion criteria. The median age of patients was 47 years (range, 25-70 years). Patient characteristics are detailed in Table 1. All patients had received chemoradiation therapy (45 Gy of pelvic external radiation therapy) with concomitant platinum-based chemotherapy (cisplatin 40 mg/m2 per week for all patients except for 1 patient who received carboplatin). Details on radiation therapy modalities and surgery are provided in Table 2. The median interval between brachytherapy and MRI evaluation was 35 days (range, 21-64 days). The median interval between MRI evaluation and surgery was 25 days (range, 6-80 days).

Table 1. Patient Characteristics
CharacteristicNo. of Patients%
  • *

    Stage was determined according to the 1995 International Federation of Gynecology and Obstetrics classification.

Median age (range), y47 (25-70) 
Stage*  
 IB21739
 II2761
  IIA37
  IIB2454
Histologic subtype  
 Squamous cell carcinoma3580
 Adenocarcinoma614
 Glassy cell carcinoma24
 Adenosquamous12
Tumor differentiation  
 Well differentiated1945
 Moderately differentiated2253
 Poorly differentiated12
 Unknown2
Median tumor size (range), mm  
 Clinical47.5 (4-70) 
 Radiologic46 (15-80) 
Table 2. Treatment Delivered in 44 Patients
TreatmentNo. of Patients%
  1. Gy indicates grays.

External radiation therapy  
 45 Gy44100
 Median no. of fractions (range)25 (20-28) 
 Median no. of d (range)36 (29-60) 
 Concomitant chemotherapy44100
  Cisplatin: 40 mg/m2/wk4398
  Carboplatin12
 Uterovaginal brachytherapy: 15 Gy44100
 Lateral boost of external radiation therapy: Pelvic lymph nodes or parametria1022
Surgical approach  
 Laparoscopy1943
 Laparotomy2557
Pelvic surgery  
 Extrafascial hysterectomy3375
 Radical hysterectomy1023
 Radical cervicectomy12
Lymph node surgery  
 Pelvic lymphadenectomy or selective adenectomy716
 Para-aortic lymphadenectomy4091

Radiologic Results

Pretreatment MRI studies were available and, thus, were consulted by both readers (if they were available) in 43 patients (98%). Posttreatment MRI studies initially were analyzed by 20 different radiologists, but MRI studies from 30 patients (68%) were analyzed mainly by 2 radiologists. Correlation between the initial posttreatment MRI results and the consensual results of our 2 radiologists are given in Table 3.

Table 3. Correlation Between the Initial Magnetic Resonance Image Interpretation and Consensual Results From 2 Radiologists
Initial ResultResult of Consensual Review of 2 Senior Radiologists
NormalUncertainCertain
Normal1065
Uncertain131
Certain1710

During the MRI examination, 12 patients had what were considered normal imaging studies, 16 patients had studies that were considered ‘uncertain’, and 16 patients had studies with ‘certain’ residual disease in the cervix and/or uterus. Correlation between these 3 groups and data from the clinical examination performed at the time of MRI are given in Tables 4, 5, and 6.

Table 4. Clinical Evaluation and Histologic Results From the Group of 12 Patients With ‘Normal’ Posttreatment Magnetic Resonance Imaging Studies
Clinical Evaluation of RDNo. of Patients 
 Histologic RD 
No RDIsolated Cells<1 cm≥1 cmTotal
  1. RD indicates residual disease.

No RD32106
Uncertain12003
Certain20013
Total641112
Table 5. Clinical Evaluation and Histologic Results From the Group of 16 Patients With ‘Uncertain’ Posttreatment Magnetic Resonance Imaging Studies
 No. of Patients
 Histologic RD
Clinical Evaluation of RDNo RDIsolated Cells<1 cm≥1 cmTotal
  1. RD indicates residual disease.

No RD610310
Uncertain10023
Certain21003
Total920516
Table 6. Clinical Evaluation and Histologic Results From the Group of 16 Patients With ‘Certain’ Residual Disease on Post-treatment Magnetic Resonance Imaging Studies
 No. of Patients
 Histologic RD
Clinical Evaluation of RDNo RDIsolated Cells<1 cm≥1 cmTotal
  1. RD indicates residual disease.

No RD30104
Uncertain12047
Certain02035
Total441716

Histologic Results

Nineteen patients (43%) had no residual disease on the cervix or uterus, 10 patients (23%) had minimal residual disease in the form of isolated cells, 2 patients (5%) had residual disease that measured <1 cm, and 13 patients (29%) had residual disease that measured >1 cm. Five patients had extracervical residual disease (3 in the parametria, 1 in the vagina, and 1 in the septum between bladder and vagina). Four of those 5 patients with residual disease located at an extracervical site also had residual disease in the cervix that measured >1 cm.

Among 9 patients with adenocarcinoma (including 1 patient with an adenosquamous lesion and 2 patients with glassy cell carcinoma), 3 patients had residual disease that measured ≥1 cm (2 had ‘certain’ residual disease, and 1 had ‘uncertain’ residual disease on MRI), 3 patients had residual disease in the form of isolated cells (2 had ‘uncertain’ residual disease, and 1 had an absence of residual disease on MRI), and 3 patients had an absence of residual disease (2 had ‘uncertain’ residual disease, and 1 had an absence of residual disease on MRI).

In the subgroup of patients with no histologically proven residual disease or residual disease in the form of isolated cells (29 patients), the clinical examination was considered ‘normal’ in 15 patients (52%), ‘uncertain’ in 7 patients (24%), and ‘certain’ (that clinical residual disease was present) in 7 patients (24%). In the subgroup of patients with histologically proven residual disease that measured <1 cm or ≥1 cm, the clinical examination was ‘normal’ in 5 patients (33%), ‘uncertain’ in 6 (patients 40%), and ‘certain’ (that clinical residual disease was present) in 4 patients (27%).

Accuracy of MRI Imaging

The sensitivity, specificity, and positive and negative predictive values of MRI were 80% (8/8 + 2), 55% (10/10 + 8), 50% (8/8 + 8) and 83% (10/10 + 2), respectively (Table 7). The false-positive and false-negative rates were 50% and 17% respectively.

Table 7. Summary of the Numbers of Patients in Each Group Used to Determine the Accuracy of Magnetic Resonance Imaging
 No. of Patients
MRI RDAbsence of Histologic RD or Isolated CellsHistologic RD <1 cm or ≥1 cm
  1. MRI indicates magnetic resonance imaging; RD, residual disease.

No102
Certain88

In the subgroup of 8 patients who had false-positive results, the interval between MRI and surgery ranged between 8 days and 51 days, but only 3 patients had an interval >30 days. Two patients who had false-positive results were observed, but only the patient with residual disease >1 cm had an imaging study that was considered ‘normal’ but a clinical examination that was considered abnormal. The intervals between brachytherapy and MRI and between MRI and surgery were 36 days and 14 days, respectively. A laparoscopic extrafacial hysterectomy was performed (with a para-aortic lymphadenectomy). Bulky residual cervical disease measuring 17 mm (without lymph node spread) was identified with free margins.

DISCUSSION

MRI is the standard imaging technique used for the initial evaluation of local spread (tumor size, spread to parametria, vagina, bladder, and rectum) and lymph node status in patients with cervical cancer.12–18 However, its accuracy in predicting response after CRT still is under debate. This is a crucial question, because all teams use at least MRI to evaluate response after CRT. Most of the time, patients with suspected residual lesions on MRI are considered candidates for completion surgery based on (radical) hysterectomy.2, 5, 6

Very few studies have analyzed the accuracy of MRI after CRT. Nam et al recently published an article on a series of 38 patients who were treated with CRT (external radiotherapy at a dose of 45 Gy followed by high-dose-rate brachytherapy with concomitant cisplatin chemotherapy) and had response evaluated by MRI.19 A radiologic response was obtained in 35 patients (92%). The local control rate in those 35 patients was 100%.19 However, the accuracy of MRI in this interesting study was based on the follow-up of patients, but there was no histologic verification of the cervix.

The true accuracy of MRI in cervical cancer should be correlated with the histologic results, which are the most reliable. Eight years ago, Hatano et al reported on a correlation between MRI and cytology and/or punch biopsy findings in this context. Those procedures were performed 1 month, 3 months, and 6 months after the end of radiation therapy in 42 patients who were treated for stage I through IV cervical cancer.18 The sensitivity, specificity, and negative and positive predictive values of MRI were 100%, 78%, 43%, and 100%, respectively.18 A strong correlation was observed between MRI findings and histology in procedures that were performed 3 months and 6 months after therapy (no false-negative or false-positive results). Histologic verification in that study was based on superficial (<1 cm in depth) punch biopsy specimens. Under such conditions, MRI cannot be truly correlated with the histology of the cervix, because cervical and uterine specimens were not examined completely like they were in our study. Furthermore, that study was performed when external radiation therapy was considered the standard of care (chemoradiation became the standard management for advanced-stage cervical cancer in 19991).

Therefore, to our knowledge, the current study is the first to correlate MRI with histology of the entire cervix and uterus in patients who receive CRT for stage IB2/II disease. Furthermore, the inclusion criteria in our series were very strict: All patients had received the same treatment modalities and were studied over a short period. Therefore, even if the current series was retrospective, the conclusions are likely to be more reliable given the homogeneity of our patients. One of the areas that could be discussed in our methodology is combining patients with no histologic residual disease and patients with very small residual disease (<1 mm) under the form of isolated cells. This was performed for 3 reasons: 1) the prognosis for patients who undergo completion surgery with very small, microscopic residual disease or the absence of residual disease was similar in 3 series.3, 4, 8 The management of these patients, in terms of discussing the place of completion surgery, should be similar. 2) The radiosensitivity of these patients is very close; and, if patients with very small residual disease had undergone surgery later, then most of them probably would have had a complete response. 3) ‘Conventional’ MRI is unable to discriminate residual disease constituted of isolated cells <1 mm. For these 3 reasons both histologic subtypes were mixed.

The accuracy of MRI was lower in our study than in previous studies discussed herein.18, 19 In 3 of 4 patients who exhibited ‘residual disease’ on MRI, histologically proven residual disease was identified, but 50% of the patients had minimal residual disease (only 7 of 16 patients had residual disease >1 cm), and 25% of patients had no residual disease. Consequently, systematic radical surgery in patients with suspicious, radiologically identified residual disease was pointless in 50% of them. The morbidity of such surgery in this context has been evaluated well.2

Conversely, 8% of patients with no residual disease on MRI did have histologically proven disease that measured >1 cm. This is the most disturbing result of this study, because these patients would have been under treated if they had not undergone completion surgery (even if completion surgery in patients with bulky residual disease has not yielded a gain in terms of survival).5, 6 Furthermore, this suboptimal MRI accuracy increased the rate of positive margins (as observed in the current study), because the radicality of surgery based on radiologic findings was not sufficiently effective to guarantee clear margins.

How can we explain the difference between the studies? The first point concerns MRI interpretation difficulties. In this study, all imaging studies were reviewed by 2 senior radiologists with experience in imaging gynecologic malignancies. Even with such experience, distinguishing between normal and suspicious cases was impossible because of posttherapy changes in 16 of 44 patients. This is why only MRI studies with a contributive conclusion (‘normal’ or ‘certain’ residual disease) were included in the accuracy analysis in the current series.

The second point concerns the intervals between brachytherapy and imaging and between MRI and surgery. Concerning the first interval, if MRI is performed too ‘early’ after the end of brachytherapy, then the rate of patients with radiologic residual disease will increase. The optimal interval between brachytherapy and MRI is unknown, but most teams perform this procedure between 4 to 6 weeks after brachytherapy and sometimes even later (12 weeks). In a series of patients who received external radiation therapy without surgery, Hatano et al suggested that the accuracy of MRI was increased when this procedure was performed 3 months after radiation therapy.18 If MRI is performed at this interval posttreatment, then ‘fibrosis’ would increase, and, thus, perioperative morbidity (in patients undergoing completion surgery) also would increase. In our series, 86% of patients had such an interval (between 26 days and 47 days); therefore, our series is relatively ‘homogeneous’ in this respect. This interval may be too short and probably explains the high rate of patients with ‘residual disease’ on imaging in our series.

The second important interval is between MRI and surgery. If it is too ‘long’ (>4 to 6 weeks), then it could increase the rate of false-positive results, because some patients who have radiologic findings suggestive of residual disease ultimately would exhibit no histologic residual disease, because the effect of radiation therapy persists during the interval between MRI and surgery. In our series, among the 8 patients who had radiologic residual disease defined as ‘certain’ and had no (or minimal) histologic residual disease, only 3 patients had an interval >30 days between imaging and surgery.

Finally, using MRI to predict response after CRT in stage IB2/II cervical cancer is not the optimal technique for predicting response after CRT because of the risk of false-positive results, as observed in the current series. It is more difficult to interpret MRI studies after the treatment of cervical cancer than at the time of disease staging. Radiologists and surgeons should be aware of these difficulties, and pretreatment images should be available for a more accurate appraisal of posttreatment images. Adequate sequences also are required, including T2-weighted images and dynamic sequences after injection of contrast media.

New procedures or a combination of techniques should be evaluated for the evaluation of response to treatment in patients with cervical cancer. Tumor volume regression may be a better response criterion during treatment.19, 20 Perhaps imaging also should be correlated with biopsy specimens, as suggested by Nijhuis et al21: or perhaps the accuracy of such procedures would be improved if they were associated with the monitoring of serum squamous cell carcinoma antigen levels.22–24 Different imaging procedures, such as diffusion-weighted MRI or positron emission tomography, may be promising in this context.25–27 Furthers studies should be performed to improve the accuracy of the evaluation of response to CRT, which currently is based on clinical evaluation and MRI studies.

Ancillary