Can radicality of surgery be safely modulated on the basis of MRI and PET/CT imaging in locally advanced cervical cancer patients administered preoperative treatment?

Authors


Abstract

BACKGROUND:

The goal of this study was to prospectively analyze the diagnostic performances of magnetic resonance imaging (MRI) and positron emission tomography (PET)/computed tomography (CT) in predicting pathologically assessed residual disease in a large, single-institution series of locally advanced cervical cancer (LACC) patients triaged to neoadjuvant treatments followed by radical surgery.

METHODS:

Between April 2007 and March 2010, 96 patients with histologically documented cervical cancer (any histology) and FIGO stage IB2-IVA were enrolled. MRI and PET/CT were recommended to be performed within 4-6 weeks from the end of treatment, and histology was the reference standard. Sensitivity, specificity, and accuracy were compared using the McNemar test.

RESULTS:

For residual disease in the cervix, sensitivity was higher for MRI than for PET/CT (86.1% vs 63.1%; P = .002), while specificity was significantly higher for PET/CT compared with MRI (P = .002). There was no difference in accuracy values between the 2 imaging modalities. For MRI analysis of lymph node groups, sensitivity, specificity, and accuracy were 35.7%, 95.9%, and 88.0%, respectively. Conversely, sensitivity, specificity, and accuracy for PET/CT were 28.6%, 97.8%, and 88.7%, respectively. Absence of follicular structures replaced by prevalent sclerosis and/or sinus histiocytosis was the most frequently documented morphological pattern in false-positive cases.

CONCLUSION:

Neither MRI nor PET/CT accurately detected residual disease in LACC patients triaged to radical surgery after neoadjuvant treatment, disallowing the option of avoiding or modulating completion surgery. Cancer 2011;. © 2011 American Cancer Society.

The role of magnetic resonance imaging (MRI) and [18F]-fluoro-2-deoxyglucose (FDG) positron emission tomography (FDG/PET) or PET/computed tomography (PET/CT) in cervical cancer staging has been widely recognized,1, 2 such that these imaging modalities are more and more often used in the routine setting and have been recently included by the National Comprehensive Cancer Network among the procedures for pretreatment work-up of stage ≥IB2 cervical cancer (www.nccn.org/professionals /physician_gls). FDG/PET has also been actively investigated for prognostic characterization, prediction of response to treatment, and surveillance,1, 3-5 and recently, the diagnostic accuracy of PET/CT, which combines metabolic information with anatomic details6, has been reported to be even higher than FDG/PET, MRI, or CT techniques in cervical cancer.7, 8 In particular, MRI sensitivity in the preoperative evaluation of lymph node disease in untreated patients has been reported to be 30%-73%,7, 9, 10 with an overall accuracy of 70%-90%. On the other hand, the diagnostic performance of FDG/PET and PET/CT was shown to be higher, with sensitivity values up to 90%, and an overall accuracy of 78%-100%.1, 9 Conversely, few data on anecdotal cases or small, retrospective series are available regarding the diagnostic performances of MRI and FDG/PET in detecting pathologically assessed residual disease after multimodal treatment of locally advanced cervical cancer (LACC) patients.11-16 This issue is potentially of interest considering that the direct effects of treatment on tumor cells as well as treatment-induced necrosis and inflammation are likely to produce a derangement of tissue architecture and boundaries, such that the performance of any imaging approach might intuitively be lower in posttreatment cases compared with untreated ones. In this context, it must be noted that while exclusive chemoradiation (CT/radiotherapy [RT]) has been accepted since 1999 as the gold standard for treatment of LACC patients,17 investigational approaches including neoadjuvant chemotherapy and CT/RT followed by radical surgery have also been attempted, with encouraging results.18-20 Because multimodal approaches are associated with a higher rate and/or severity of intraoperative and postoperative complications,21, 22 the availability of imaging techniques that can accurately define the extent of residual tumor, thus allowing to select patients who can be spared or at least be offered a more tailored surgery, would be clinically relevant.

The aim of the study was to prospectively analyze the diagnostic performances of MRI and PET/CT in predicting pathologically assessed residual disease in a large, single-institution series of LACC patients triaged to neoadjuvant treatment followed by radical surgery.

MATERIALS AND METHODS

Work-up and Treatment

Between April 2007 and March 2010, patients with histologically documented cervical cancer (any histology) and stage IB2-IVA disease according to FIGO classification were prospectively enrolled at the Gynecologic Oncology Unit of the Catholic University in Campobasso and Rome, Italy. Other inclusion criteria included age 18-75 years and Eastern Cooperative Oncology Group performance status of 0-1. Institutional Review Board approval was obtained, and all patients provided written informed consent to participate in the study. Pretreatment work-up included clinical examination, pelvic and recto-vaginal examination in anesthesia, chest radiography, complete blood count, and measurement of liver and renal function. MRI and PET/CT were also performed for staging, but surgeons were blinded to the results, hence they were not considered in the evaluation of posttreatment imaging. Cystoscopy and proctoscopy were performed if there was a clinical suspicion of organ involvement. Patients received preoperative CT/RT or neoadjuvant chemotherapy in case of refusal of radiation treatment. CT/RT was administered to the pelvic region (39.6-50.3 Gy) according to specific protocols, and concomitant chemotherapy included cisplatin and 5-fluorouracil.20 Cases administered neoadjuvant chemotherapy received a drug-based regimen of paclitaxel, cisplatin, and ifosfamide.23 Patients were evaluated for response according to RECIST criteria to plan surgery. MRI and PET/CT were recommended to be performed within 4-6 weeks of the end of treatment. With the exception of patients who progress during treatment, patients should be considered for radical hysterectomy according to Piver et al.24 plus systematic bilateral pelvic lymphadenectomy. Aortic lymphadenectomy up to the level of the inferior mesenteric artery was performed in case of bulky lymph nodes at surgery or in case of pelvic lymph nodes that are positive for metastasis at frozen section. Surgery was recommended to be performed within 4 weeks of imaging.

MRI and PET/CT Imaging

MRI examinations were performed as described14 with some modifications. Briefly, a 1.5-T superconducting magnet (Excite HD; GE Medical Systems, Milwaukee, WI) was used with the 8-channel pelvic phased-array surface coil. Pelvic panoramic T1-weighted rapid acquisition with relaxation enhancement (fast-spin echo) axial views were obtained using the following imaging parameters: repetition time/echo time, 600/7 msec; echo train length (ETL), 5; slice thickness, 8 mm; interslice gap, 1 mm; matrix, 384 × 256; and acquisition time, 1.12 minutes. Panoramic axial T2-weighted fast recovery fast-spin echo (FRFSE) images were obtained using the following parameters: TR/TE, 4320/108 msec; ETL, 27; slice thickness, 8 mm; interslice gap, 1 mm; matrix, 416 × 256; and acquisition time, 1.48 minutes. Sagittal T2-weighted FRFSE images were obtained using the following parameters: TR/TE, 3500-4000/105 msec; ETL, 25; slice thickness, 3 mm; interslice gap, 1 mm; matrix, 384 × 224; and acquisition time, 3.48 minutes. For oblique and short axis T2-weighted FRFSE images of the uterus, acquisition times were between 3.59 and 4.49 minutes. For oblique coronal and short axis T2-weighted FRFSE images of the uterus, the acquisition time was between 35.9 and 4.49 minutes. Residual disease in the cervix was evaluated on T2-weighted images considering tumor size, infiltration of cervical stroma and vaginal fornices, and relationship between residual tumor and internal os of the endocervical canal. Lymph nodes with short axis diameter >1 cm were considered disease-involved.

PET/CT imaging was performed as described25 using PET/CT combined with a hybrid, commercially available FDG–PET/CT scanner (Gemini Dual or Gemini GXL; Philips, Cleveland, OH) that consists of a combination of a dual-detector row spiral CT and a high-resolution PET scanner with 5-mm spatial resolution and a 3-dimensional image acquisition algorithm. A workstation (Sun Blade 2000, Syntegra) was employed for image display, fusion, and analysis. Initially, low-dose CT was performed with the patient breathing normally and included an area from the external auditory meatus to below the symphysis pubis. The CT parameters were as follows: section thickness, 6.5 mm; time per table rotation, 0.75 seconds; pitch, 1.5; tube voltage, 140 kVp; tube current, 30 mA; and field of view, 60 cm. This procedure was followed by the acquisition of PET images. Eight acquisition beds were performed, and the scan-bed acquisition time was 3 minutes. PET images were acquired in a 3-dimensional mode. Matched CT and PET images were reconstructed with a field of view of 50 cm. Iterative reconstruction and CT-based attenuation correction were used, and attenuation-corrected PET images were retrieved in transverse, sagittal, and coronal planes. PET data were also displayed in a rotating maximum-intensity projection. The patients fasted for 6 hours prior to image acquisition. Saline solution (500 mL) and activities of 5-10 mCi (185-370 MBq) of FDG were administered intravenously according to body mass index, and the images were obtained after a 60-minute uptake period during which the patient rested. Patients were asked to void just before the start of acquisition. No oral or intravenous contrast agents were administered to any patient. Neither urinary bladder catheterization nor additional provocative maneuvers were used. For image evaluation, visual criteria for both primary tumor and lymph nodes were applied, and every region with an uptake above the background was read as positive. MRI and PET/CT images were reviewed respectively by radiologists and nuclear physicians with >5 years experience in gynecologic oncology imaging who were unaware of patients' clinical features or results from pretreatment imaging modalities. Interpretation discrepancies were resolved by consensus.

Histopathologic Evaluation

Both MRI and PET/CT performances were evaluated using histology as the reference standard.

Histopathological evaluation corresponded to the standard procedures performed in our Institution: the cervix was sectioned clockwise in at least 12 blocks, and entirely embedded in paraffin. From each block, 2 slides, 3- to 4-μm thick, were cut at different levels and stained with hematoxylin and eosin. When appropriate, additional sections were prepared for specific immunohistochemistry tests.26 For lymph node evaluation, slices at 2-μm intervals perpendicular to the greatest dimension were cut. Thin sections were examined by 2 independent pathologists who had >10 years experience in gynecologic oncology pathology. Residual disease at any site was expressed in millimeters, and cases were grouped as no residual disease, microscopic (1-3 mm) residual disease, and macroscopic (>3 mm) residual disease.26 Considering the clinical objective of the study, absence of residual disease was chosen as the cut-off to divide histologically negative cases from positive ones. Additional cut-off values (up to 3 mm, up to 5 mm residual disease) were also investigated.

Statistical Analysis

Sample size was calculated on the basis of Simon design in order to detect a 15% difference in the accuracy between the 2 imaging modalities; considering the MRI diagnostic accuracy = 75% (p0), an α type 1 error = 0.01, and a β type II error = 0.1 (power = 90%), a total of 86 patients was required. Taking into account a drop out rate around 10%, a sample size of at least 95 patients was planned. Sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV), and accuracy were calculated using standard formulas. The pretest probability and posttest probability values for MRI and PET/CT were also calculated. The 95% confidence intervals were determined for each parameter. Sensitivity, specificity, and accuracy were compared using the McNemar test. The diagnostic performances of MRI and PET/CT were calculated on a per-patient as well as on a per-lesion basis: the lesion-based analysis was conducted considering residual disease in the cervix (n = 96), lymph node groups (n = 425), and in the whole series (n = 521). Lymph node groups included aortic and pelvic lymph nodes represented by lower and upper pelvic lymph nodes, bilaterally.27

A 2-sided P < .05 was considered statistically significant. Statistical analyses were performed with SPSS-Windows, version 16 (SPSS Inc. Chicago, IL).

RESULTS

Table 1 shows the clinico-pathological characteristics of cases examined: 65 cases (67.7%) had FIGO stage IIB disease; the vast majority of patients (81.2%) showed response to treatment, and 18 patients only achieved stable disease. All patients underwent surgery. The mean ± SD time interval from completion of neoadjuvant treatment and imaging was 27.0 ± 8.1 days; 60 (62.5%) patients underwent both imaging procedures within 4-6 weeks of the end of treatment. The mean ± SD time interval between imaging and surgery was 22.3 ± 12.6, with 78 (81.2%) patients undergoing surgery within the predefined time limits. Reasons for not performing surgery were incomplete recovery from neoadjuvant treatment (n = 13) and logistic problems (n = 5).

Table 1. Clinico-pathological Characteristics, Treatment Details, and Pathological Findings of the Study Population
CharacteristicsValues
  • Abbreviations: CT, computed tomography; LN, lymph nodes; NACT, neoadjuvant chemotherapy.

  • a

    Out of 40 cases.

  • b

    For definitions of lymph nod''rials and Methods.

No. of patients96
Age, y, median (range)50.5 (28-75)
FIGO stage, no. (%) 
 IB/IIA bulky12 (12.5)
 IIB65 (67.7)
 III/IVA19 (19.8)
Histotype, no. (%) 
 Squamous81 (84.4)
 Adenocarcinoma/adenosquamous15 (15.6)
Grade, no. (%) 
 1-226 (27.1)
 340 (41.7)
 Not available30 (31.2)
Neoadjuvant treatment, no. (%) 
 CT/radiotherapy68 (70.8)
 NACT28 (29.2)
Clinical response, no. (%) 
 Complete15 (15.6)
 Partial63 (65.6)
 No change18 (18.8)
Type of radical hysterectomy, no. (%) 
 Piver II11 (11.5)
 Piver III/IV85 (88.5)
Type of lymphadenectomy, no. (%) 
 Pelvic96 (100)
 Pelvic plus aortic40 (41.7)
Pelvic LN removed, median (range)25 (7-65)
Aortic LN removed, median (range)12 (1-31)
Residual tumor in the cervix, no. (%) 
 None31 (32.3)
 1-3 mm22 (22.9)
 >3 mm43 (44.8)
Metastatic pelvic LN, no. (%)25 (26.0)
Metastatic aortic LN, no. (%)11 (27.5)a
Residual tumor in lymph node groupsb (n = 425), no. (%) 
 None369 (86.8)
 1-3 mm14 (3.3)
 >3 mm42 (9.9)

At pathological examination, residual tumor in the cervix was absent in 31 (32.3%) patients; in the remaining patients, it ranged between 1 and 30 mm (median, 16 mm). Overall, metastatic involvement of pelvic lymph nodes was documented in 25 (26.0%) patients, and aortic lymph node involvement was documented in 11 (27.5%) patients. Fifty-six out of 425 lymph node groups (13.2%) were metastatic.

Diagnostic Performances

In the whole series, sensitivity was higher for MRI than for PET/CT (P = .0001), whereas specificity was higher for PET/CT (P = .0001). No difference was found with accuracy (Table 2).

Table 2. MRI and PET/CT Diagnostic Performances in the Whole Series on a Per-Lesion Basis
 Histopathology of Residual Disease at MRIMedian (95% CI)Histopathology of Residual Disease at PET/CTMedian (95% CI)Pa
PresentAbsentPresentAbsent
  • Abbreviations: CI, confidence interval; CT, computed tomography; MRI, magnetic resonance imaging; NPV, negative predictive value; PET, positron emission tomography; PPV, positive predictive value.

  • a

    McNemar test; statistically significant values are indicated in bold.

Present76355714
Absent4536564386
Sensitivity, %62.8 (54.2-71.4)47.1 (38.2-56.0).0001
Specificity, %91.2 (88.5-94.0)96.5 (94.7-98.3).0001
Accuracy, %84.6 (81.5-87.7)85.0 (81.9-88.1).9
NPV, %89.0 (86.0-92.0)85.8 (82.5-89.0)
PPV, %68.5 (59.8-77.1)80.3 (71.0-89.5)

We then analyzed MRI and PET/CT diagnostic performances separately for residual disease in the cervix (Table 3). MRI was found to correctly identify as positive 56 out of 65 cases with pathological evidence of disease, and was negative in 11/31 cases with absence of disease after treatment. Of 9 false-negative cases, 7 were shown to have residual tumor of 1-3 mm maximum diameter, and 2 had residual tumor >10 mm.

Table 3. MRI and PET/CT Diagnostic Performances in the Cervix
 Histopathology of Residual Disease at MRIMedian (95% CI)Histopathology of Residual Disease at PET/CTMedian (95% CI)Pa
PresentAbsentPresentAbsent
  • Abbreviations: CT, computed tomography; MRI, magnetic resonance imaging; NPV, negative predictive value; PET, positron emission tomography; PPV, positive predictive value.

  • a

    McNemar test; statistically significant values are indicated in bold.

Present5620416
Absent9112425
Sensitivity, %86.1 (77.8-94.5)63.1 (51.3-74.8).002
Specificity, %35.5 (18.6-52.3)80.6 (66.7-94.5).002
Accuracy, %69.8 (60.6-78.9)68.7 (59.5-78.0).9
NPV, %55.0 (33.2-76.8)51.0 (37.0-65.0)
PPV, %73.7 (63.8-83.6)87.2 (77.7-96.8)

PET/CT results were positive in 41/65 cases with pathologically documented persistence of disease, and was negative in 25/31 cases with absence of tumor foci. Of 24 false-negatives, 12 had residual tumor of 1-3 mm maximum size, 4 had residual tumor between 4-10 mm, and 8 had residual tumor >10 mm.

Sensitivity was shown to be higher for MRI compared with PET/CT (P = .002), whereas specificity was significantly higher for PET/CT compared with MRI (P = .002). There was no difference in accuracy values between the 2 imaging modalities.

As shown in Table 4, there was no difference in sensitivity, specificity, or accuracy between the 2 imaging approaches in analyzing lymph node status on a per-lesion or per-patient basis.

Table 4. MRI and PET/CT Diagnostic Performances for LFN Status
 Histopathology of Residual Disease at MRIMedian (95% CI)Histopathology of Residual Disease at PET/CTMedian (95% CI)Pa
PresentAbsentPresentAbsent
  • Abbreviations: CT, computed tomography; MRI, magnetic resonance imaging; NPV, negative predictive value; PET, positron emission tomography; PPV, positive predictive value.

  • a

    McNemar test.

Per-lesion basis       
 Present2015168
 Absent3635440361
 Sensitivity, %35.7 (23.2, 48.3)28.6 (16.7, 40.4).28
 Specificity, %95.9 (93.9, 97.9)97.8 (96.3, 99.3).14
 Accuracy, %88.0 (84.9, 91.1)88.7 (85.7, 91.7).68
 NPV, %90.8 (87.9, 93.6)90.0 (87.1, 92.9)
 PPV, %57.1 (40.7, 73.5)66.7 (47.8, 85.5)
Per-patient basis       
 Present116168
 Absent3635440361
 Sensitivity, %39.3 (21.2, 57.4)25.0 (8.9, 41.0).22
 Specificity, %91.2 (84.4, 97.9)94.1 (88.5, 99.7).68
 Accuracy, %76.0 (67.5, 84.6)73.9 (65.2, 82.7).70
 NPV, %78.5 (69.4, 87.5)75.3 (66.1, 84.5)
 PPV, %64.7 (42.0, 87.4)63.6 (35.2, 92.1)

Table 5 shows the detailed correlation between imaging and pathological evaluation of lymph node groups of false-negative cases: of 36 false-negative lymph node stations at MRI, 23 (63.9%) were shown to have residual tumor below the threshold of imaging detection (≤10 mm), whereas up to 70.0% (28/40) of false-negatives at PET/CT had residual disease ≤10 mm. Figure 1 shows 2 representative examples of false-negative cases.

Figure 1.

Representative examples of 2 false-negative cases are shown. In case 1, posttreatment axial T2-weighted MRI (A) and PET/CT (B) documented the absence of any apparent residual disease at the lymph node level. At low-power magnification, lymph node structure was characterized by the presence of sclerosis (C). At higher magnification (D), the presence of cancer cells is clearly shown. Note the signs of response to treatment (cell gigantism, vacuolization, cytoplasmic droplets). Bars indicate 1 mm (C) and 0.1 mm (D). In case 2, posttreatment and axial T2-weighted MRI and PET/CT documented the absence of abnormal lymph nodes (E) or radiotracer uptake (F). At low-power magnification, microcystic, unevenly distributed spaces in the tissue are shown (G). At higher magnification (H), the presence of cancer cells is clearly shown. Bars indicate 1 mm (G) and 0.1 mm (H).

Table 5. Correlation Between Imaging and Pathologically Assessed Residual Disease in Lymph Node Groups of False-Negative Cases
Patient No.Lymph Node GroupResidual Disease at MRIResidual Disease at PET/CTPathologically Assessed Residual Disease, mm
  1. Abbreviations: CT, computed tomography; LPN, lower pelvic node; MRI, magnetic resonance imaging; PET, positron emission tomography; UPN, upper pelvic node.

1Left LPN1
2Right LPN1
3Left UPN1
4Left LPN+1
5Right LPN+1
6Left LPN2
7Left LPN2
8Right LPN2
9Right UPN2
10Aortic2
11Aortic2
12Aortic2
13Left LPN+2
14Left LPN4
15Right LPN4
16Right LPN+4
17Right LPN5
18Aortic5
19Aortic+5
20Right LPN6
21Right LPN6
22Left LPN+6
23Aortic8
24Left LPN9
25Aortic9
26Left LPN10
27Left LPN10
28Right LPN+10
29Aortic10
30Right LPN11
31Right UPN11
32Right LPN12
33Right UPN12
34Left LPN13
35Left UPN13
36Left LPN18
37Left UPN18
38Aortic19
39Left LPN20
40Right LPN20
41Right UPN20
42Aortic+25

Table 6 reports the macroscopic and microscopic pathologic findings of the cervix for each false-positive case: in 13 out of 22 cases showing residual disease at imaging, there was no macroscopic evidence of disease at pathology, whereas erosion or ulceration of cervical tissue was documented in 4 and 3 cases, respectively. Finally, 2 cases showed only tissue fibrosis. Twenty-one cases showed microscopic evidence of chronic cervicitis, which was also associated with multinuclear giant cells (n = 7), histiocytic cells (n = 2), or fibrosis (n = 1). One case only showed endocervical glandular hyperplasia with multinuclear giant cells.

Table 6. Correlation Between Imaging and Pathological Findings in the Cervix in False-Positive Cases
Patient No.Residual Disease at MRIResidual Disease at PET/CTMacroscopic Pathological FindingsMicroscopic Pathological Findings
  1. Abbreviations: CT, computed tomography; MRI, magnetic resonance imaging; PET, positron emission tomography.

1++No suspicious lesionChronic cervicitis
2++Cervical erosion (1.2 cm)Chronic cervicitis with multinuclear giant cells
3++Cervical erosion (2.0 cm)Chronic cervicitis
4++Cervical ulceration (1.2 cm)Chronic cervicitis
5+Cervical ulceration (1.1 cm)Chronic cervicitis
6+Cervical ulceration (1.2 cm)Endocervical glandular hyperplasia with multinuclear giant cells
7+Cervical tissue fibrosisChronic cervicitis with multiple fibrotic foci
8+No suspicious lesionChronic cervicitis with multinuclear giant cells
9+No suspicious lesionChronic cervicitis
10+Cervical tissue fibrosisChronic cervicitis with multinuclear giant cells
11+No suspicious lesionChronic cervicitis with multinuclear giant cells
12+Multiple erosive areasChronic cervicitis with histiocytary cells
13+Multiple erosive areasChronic cervicitis with histiocytary cells
14+No suspicious lesionChronic cervicitis
15+No suspicious lesionChronic cervicitis
16+No suspicious lesionChronic cervicitis
17+No suspicious lesionChronic cervicitis
18+No suspicious lesionChronic cervicitis with multinuclear giant cells
19+No suspicious lesionChronic cervicitis
20+No suspicious lesionChronic cervicitis with multinuclear giant cells
21+No suspicious lesionChronic cervicitis with multinuclear giant cells
22+No suspicious lesionChronic cervicitis

As shown in Table 7, which summarizes the correlation between imaging and pathological assessment of lymph node groups of false-positive cases, there were 15 false-positives at MRI and 8 false-positives at PET/CT. The pathologic macroscopic evaluation failed to show any suspicious lesion in 15 lymph node groups, while revealing only 5 enlarged lymph node groups. Microscopically, most lymph node groups (n = 13) showed chronic lymphadenitis, often associated with granulomatous inflammation, sclerosis, and/or sinus histiocytosis (Figure 2). Regarding the analysis of only aortic lymph node stations, MRI was found to correctly identify as positive 2 out of 11 cases with pathological evidence of disease, and was negative in all cases (n = 30) exhibiting no residual disease after treatment. Conversely, PET/CT results were positive in 2/3 cases with pathologically documented aortic lymph node involvement and were negative in 28/37 cases with absence of residual disease. There was no statistically significant difference in sensitivity, specificity, or accuracy between the 2 imaging techniques, although these results have to be interpreted with caution considering the small sample size.

Figure 2.

Representative examples of 2 false-positive cases are shown. In case 1, posttreatment axial T2-weighted MRI (A) documented the presence of an enlarged lymph node (short axis = 14 mm) in the right external iliac lymph node group (arrow), and posttreatment PET/CT (B) documented a focus of abnormal radiotracer uptake (standardized uptake value = 4.3) in the same area. Low-power magnification of a representative lymph node from that area revealed the absence of follicles and prevalence of regularly dilated sinuses circumscribed by sclerosis (C). A representative area surrounded by the black box is shown at higher magnification (D) and documents the absence of cancer cells. Bars indicate 1 mm (C) and 0.1 mm (D). In case 2, posttreatment axial T2-weighted MRI (E) documented an enlarged lymph node (short axis = 13 mm) in the left obturatory group, and posttreatment PET/CT (F) documented an abnormal radiotracer uptake (standardized uptake value = 5.8) in the same area. Low-power magnification revealed the absence of follicles and a more solid pattern (G). At higher magnification (H), a slight prevalence of sinus histiocytosis over sclerosis is shown. No neoplastic cell is present. Bars indicate 1 mm (G) and 0.125 mm (H).

Table 7. Correlation Between Imaging and Pathological Findings in Lymph Node Groups in False-Positive Cases
Patient No.Lymph Node GroupResidual Disease at MRIResidual Disease at PET/CTMacroscopic Pathological FindingsMicroscopic Pathological Findings
  1. Abbreviations: CT, computed tomography; LPN, lower pelvic node; MRI, magnetic resonance imaging; PET, positron emission tomography; UPN, upper pelvic node.

1Right LPN++Normal sizeReactive lymphadenitis
2Right LPN++Enlarged size (1.5 cm)Reactive lymphadenitis
3Left LPN++Normal sizeReactive lymphadenitis
4Left LPN+Normal sizeChronic lymphadenitis
5Left LPN+Normal sizeChronic lymphadenitis
6Left UPN+Normal sizeChronic lymphadenitis
7Left UPN+Normal sizeChronic lymphadenitis
8Aortic+Normal sizeReactive lymphadenitis
9Right LPN+Enlarged size (2.0 cm)Chronic granulomatous inflammation
10Right LPN+Normal sizeChronic granulomatous inflammation
11Left LPN+Enlarged size (2.0 cm)Chronic granulomatous inflammation
12Left LPN+Normal sizeReactive lymphadenitis
13Left LPN+Normal sizeChronic granulomatous inflammation with sclerosis and histiocytosis
14Left LPN+Normal sizeReactive lymphadenitis
15Right UPN+Enlarged size (2.0 cm)Chronic granulomatous inflammation
16Right UPN+Normal sizeReactive lymphadenitis
17Left UPN+Enlarged size (2.0 cm)Chronic granulomatous inflammation
18Left UPN+Normal sizeReactive lymphadenitis
19Left UPN+Normal sizeChronic lymphadenitis
20Left UPN+Normal sizeChronic lymphadenitis

The pretest probability of having residual tumor in the cervix after treatment was 67.7%, but the posttest probability was 73.7% for positive MRI and 87.2% for positive PET/CT. The pretest probability of presenting metastatic lymph node involvement after treatment was 29.2%; the posttest probability rose up to 64.7% for positive MRI, and to 63.6% for positive PET/CT (data not shown).

There was no improvement in the diagnostic performance of the 2 approaches when using different cut-off values (3 mm or 5 mm) to define histologically negative cases in any analysis performed (data not shown).

DISCUSSION

To our knowledge, this is the first prospective, observational study assessing and comparing the performances of MRI and PET/CT in the evaluation of pathologically assessed residual disease after neoadjuvant treatment in a large series of LACC patients.

In the whole population, we failed to find any difference in the overall accuracy of the 2 imaging procedures; however, a separate analysis of cervix versus lymph nodes was supposed to provide more reliable and clinically relevant information on size and specific site of residual tumor, hopefully helping surgeons to decide whether to perform completion surgery, and to what extent for hysterectomy and/or lymphadenectomy. Indeed, both techniques showed higher sensitivity and lower specificity for cervix evaluation than lymph nodes, thus emphasizing the need to consider primary tumor and lymph node disease as separate entities for testing the performances of any imaging tool. Moreover, whereas MRI and PET/CT performances diverged in the evaluation of residual disease in the cervix (higher sensitivity and lower specificity for MRI compared with PET/CT), they were similar in the evaluation of lymph nodes. These findings, together with the documentation that accuracy values are not statistically different between the 2 techniques at the primary and lymph node setting, lead to retain the quality of the 2 diagnostic procedures as equivalent.

Conversely, the NPVs for both MRI and PET/CT in the evaluation of the cervix were quite disappointing, because almost half of cases with negative findings at imaging showed the presence of residual tumor, which was even larger than 10 mm in 22.2% of false-negatives for MRI and 33.3% of false-negatives for PET/CT. These data suggess that planning a less radical hysterectomy, or even deciding not to perform it at all, on the basis of negative imaging would lead to a risky underestimation of residual tumor in the cervix with deleterious consequences in terms of local control and survival.

Regarding per-lesion–based analysis of lymph node status, the NPVs for both MRI and PET/CT appeared quite higher (≈90%) considering that the proportion of false-negative cases (63.9% and 70.0% at MRI and PET/CT scan, respectively) that had residual tumor below the threshold of imaging detection (≤10 mm) is not negligible. However, if we focus on what really matters from a clinical point of view (ie, the likelihood that patients defined as lymph node–negative at imaging are truly free of disease at pathology), NPVs for MRI and PET/CT drop to 78% and 75%, respectively, which means that approximately 20%-25% of patients considered as lymph node–negative would be erroneously triaged to a less radical, if any, lymphadenectomy and possibly left with residual tumor. One could argue that the choice of “no residual tumor” as the reference value for negative histology would intuitively lead to high values of false-negative cases. Indeed, the performances of MRI and PET/CT were shown not to improve when using other cut-off values. Moreover, although some investigators combined cases with no residual tumor and cases with only microscopic tumor based on the equivalence of these 2 groups in terms of clinical outcome,19, 20 our recent observations suggest that with an appropriately longer follow-up period, a statistically significant difference in terms of survival is in fact evident between these 2 groups,28 thus clinically supporting the rationale for the cut-off we have chosen.

As far as PPVs are concerned, PET/CT showed a more favorable performance than MRI for residual tumor in the cervix, although the proportion of false positive cases was not negligible for either procedure (26.3% for MRI, 12.7% for PET/CT). Moreover, the proportion of false-positives was even higher for the evaluation of lymph node status in both lesion-based and patient-based analysis. Indeed, consequences of low PPVs are mainly represented by overtreatment and surgery-related complications, whose clinical impact, although considered lower compared with that associated with the risk of leaving residual tumor in situ, should be not underestimated. We acknowledge that the relatively low PPVs in our study could be ascribed to the fact that MRI and PET/CT were not performed within the predefined time limits in 37.5% of cases. On the other hand, the interval between imaging and surgery seems acceptable, because in most cases it did not exceed 4 weeks. It is conceivable that the low PPVs are due to treatment-induced tissue modifications in which depleted follicular structures have been replaced by sclerosis, sinus histiocytosis, or, as already reported, multinuclear giant cells.26, 29 Although our data suggest that the performances of both techniques do not differ according to type of neoadjuvant treatment, this issue remains to be established in a larger series.

In conclusion, we showed that both MRI and PET/CT, which play a major role in preoperative staging, prognostic characterization, and monitoring of response to treatment, are not accurate enough in the detection of residual disease in LACC patients triaged to radical surgery after neoadjuvant treatment, thus disallowing to the option of avoiding or modulating completion surgery. The acknowledgement of the detection limits of MRI and PET/ CT in this specific subset of patients stimulates the improvement of conventional approaches and the development of novel ones: apart from functional techniques such as contrast-enhanced or diffusion-weighted MRI, which are actively investigated as response biomarkers, fused MRI/PET has also been reported to provide additional value compared with PET/CT for detecting lymph node metastases in cervical cancer.1, 30

FUNDING SOURCES

No specific funding was disclosed.

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

Ancillary