PATIENTS AND METHODS

Patient selection

Ten patients (five males, five females), median age 68 years (range 54–87), with 13 resectable primary or recurrent colorectal cancers, were recruited at diagnosis from local surgical clinics. Demographics are given in table 1. Written informed consent was obtained following approval from the Local Research Ethics Committee and the Administration of Radioactive Substances Advisory Committee. All patients underwent whole body ¹⁸FDG scanning followed by an ¹⁸FLT scan (median intervening period 5 days (range 1–19)).

Imaging

¹⁸FDG was prepared using a modification of the technique described by Hamacher and colleagues²⁷ while ¹⁸FLT was prepared according to the method of Cleij and colleagues.²⁸ Radiochemical purity was >99% and >95% for ¹⁸FDG and ¹⁸FLT, respectively. All patients were asked to fast for six hours prior to scanning, and blood glucose levels were checked using a glucometer (Glucometer Elite; Bayer Pharmaceuticals, Newbury, Berkshire, UK) prior to injection of the tracers. All patients were normoglycaemic with a mean blood glucose concentration of 5.4 mmol/l (SD 0.5) for ¹⁸FDG and 4.6 mmol/l (SD 0.8) for ¹⁸FLT. Mean time of data acquisition post injection was 57 minutes (SD 10) and 62 minutes (SD 12) for ¹⁸FDG and ¹⁸FLT, respectively. The timing of ¹⁸FLT scanning in patients with this disease was determined by previous work carried out by our group using time-activity curves.²⁹ Mean injected activity was 371 MBq (SD 24) and 351 MBq (SD 52), respectively. All data were acquired using the Discovery LS PET/computed tomography (CT) combined scanner (GE medical systems, Milwaukee, USA). PET emission datasets were attenuation corrected utilising the CT capability of the hybrid scanner.³⁰ Spiral CT scans were acquired using a speed of rotation and couch movement of 0.8 s and 22.5 mm per rotation, respectively, with a slice thickness of 5 mm. CT images were subsequently converted to maps of PET attenuation coefficients using a bilinear transformation. Transaxial emission images of 4.3×4.3×4.25 mm³ (matrix size 128×128×35) were reconstructed using ordered subsets expectation maximisation with two iterations and 28 subsets.

For each patient, regions of interest (ROI) were drawn over the tumour and standardised uptake values (SUVs) were obtained for both tracers. SUVs were calculated using the average tumour activity concentration (A) in kBq/ml, the patient’s body weight (W) in kg, and the injected activity (A_inj) in kBq:

The average SUV of five consecutive slices was obtained. The slices selected comprised the one with the maximum count density over each ROI and the four immediately adjacent slices.

Immunohistochemistry

Routine histological examination was performed on 4 μm thick haematoxylin and eosin (H&E) stained sections. Each specimen was sampled in a routine manner for diagnostic purposes in accordance with established Royal College of Pathology clinical guidelines. Immunohistochemistry for Ki-67 using the MIB-1 antibody was then performed on routine diagnostic blocks using 3 μm thick sections cut from paraffin blocks and dried overnight at 60°C. Sections were taken from xylene (two changes), through graded alcohols (100% and 70%) to water. Antigen retrieval was performed by pressure cooking for two minutes in a conventional 15 lb pressure cooker. Slides were then rinsed in 0.05% Tween 20 in Tris buffered saline (TBS/Tween). Endogenous peroxidase activity was blocked for 10 minutes using a commercially available peroxidase blocking solution (Dako, Ely, Cambridgeshire, UK). Primary mouse monoclonal antibody (MIB-1 (Dako) diluted 1/50 in TBS, pH 7.4) was applied for 60 minutes at room temperature.

Sections were rinsed in TBS/Tween and secondary antibody (goat antirabbit/mouse; Dako) was applied for 30 minutes. Sections were then rinsed in TBS/Tween and streptavidin-horseradish peroxidase (Dako) was applied for 30 minutes. Sections were again rinsed in TBS/Tween and diaminobenzidene used as a chromogen. A haematoxylin counter stain was applied.

Sections were dehydrated through graded alcohols (70% and 100%), cleared in xylene (two changes), and coverslipped with DPX mountant.

Examination of sections

Sections of tumour were identified on the H&E slides and the corresponding MIB-1 sections were examined independently by two experienced observers, A and B, blinded to the results of the scans.

Three MIB-1 stained sections of tumour were used for each lesion to account for tumour heterogeneity. Thirty high power fields (×40 objectile lens) were examined and 50 tumour cells were assessed per field using a graticule, making a total of 1500 cells per case. A labelling index (LI) was calculated as the percentage of positive staining nuclei out of all nuclei examined. “Positive” nuclear staining was regarded as staining of the entire nucleus. Nuclei with no staining or nucleolar staining only was regarded as negative.

Statistical analysis

Interobserver reliability was assessed using Pearson’s correlation coefficient, and linear regression analysis using the SPSS statistical package was employed to correlate uptake of both ¹⁸FLT and ¹⁸FDG with MIB-1.

RESULTS

Histopathology of all lesions (n=13, five primary, five liver metastases, and three peritoneal) confirmed adenocarcinoma in 12 of 13 cases. The remaining case was thought to be a malignant peritoneal recurrence by preoperative CT, ¹⁸FDG PET imaging, and at intraoperative evaluation but on microscopic histological examination was found to be an inflammatory lesion secondary to fat necrosis.

There was excellent interrater reliability between the two observers responsible for quantifying proliferation using MIB-1 immunohistochemistry (Pearson’s correlation coefficient 0.9) (fig 1).

Quantification of tracer uptake in addition to the corresponding LIs can be seen in table 1. All five primary tumours in addition to both malignant peritoneal lesions were visualised with increased focal uptake using both ¹⁸FLT and ¹⁸FDG. Three of five liver metastases were clearly seen to be avid for ¹⁸FLT, and all showed a corresponding high proliferation LI.

The three lesions (peritoneal inflammatory lesion and two liver metastases) which did not show increased uptake of ¹⁸FLT and were therefore not visualised using this tracer had corresponding low proliferation rates, as determined by the LIs, ranging from 19 to 26.

SUVs for those lesions seen with ¹⁸FDG ranged between 3.7 and 11 compared with a range of 1.4–9.1 for the 10 lesions visualised using ¹⁸FLT.

There was a statistically significant correlation (r =0.8, p<0.01) between SUVs of the 10 tumours visualised with ¹⁸FLT and the corresponding LIs (fig 2). No such correlation (r =0.4) was demonstrated in the case of ¹⁸FDG avid lesions (fig 3).

Differences in tracer uptake compared with macroscopic histology and MIB-1 immunohistochemistry are highlighted in fig 4.

DISCUSSION

Our results suggest that ¹⁸FLT reflects proliferative activity in CRC, as determined by MIB-1 staining. Lack of correlation between LIs and ¹⁸FDG supports previous evidence that ¹⁸FDG is not a specific marker of proliferation.^21,22 The lesions successfully imaged with ¹⁸FLT demonstrated high SUVs in addition to high LIs. On the other hand, those lesions not visualised by ¹⁸FLT were shown to have correspondingly low LIs.

It is known that glucuronidation of ¹⁸FLT within the liver leads to high hepatic background activity which may inhibit its diagnostic utility.³¹ This was exemplified by failure of ¹⁸FLT PET to show two of the five liver metastases against this high liver background. However, both of these lesions displayed low proliferation labelling indices (19% and 26%), suggesting relatively slow growth compared with the remaining three liver metastases (labelling indices of 87%, 87%, and 73%). The fact that certain liver lesions are not seen with ¹⁸FLT may have prognostic implications in terms of the aggressiveness of these tumours.

The well documented problems of specificity due to uptake of ¹⁸FDG by activated macrophages^32–34 is also highlighted in this study by the peritoneal lesion thought to be malignant because of a high uptake of ¹⁸FDG. Microscopic examination of the entire lesion showed a florid inflammatory response and fat necrosis secondary to previous surgery, but no tumour. In contrast, ¹⁸FLT showed no increased uptake within this lesion, suggesting that this new tracer may have a role to play in improving the diagnostic specificity of PET.

¹⁸FLT PET has recently been found in the investigation of solitary pulmonary nodules to be specific for malignant lesions. In addition, ¹⁸FLT correlates with the Ki-67 labelling index and estimation of prognosis.³⁵ This hypothesis has been further validated by Vesselle and colleagues³⁶ who assessed proliferation rates in 11 histologically proven non-small cell lung cancers using ¹⁸FLT PET and again demonstrated significant correlation with Ki-67.

The very nature of PET allows metabolic assessment of the whole tumour in its heterogeneic entirety. This coupled with the unique ability to quantitate the uptake of the tracers used, non-invasively and in vivo, has opened new avenues in the assessment of malignancy by functional imaging.

A system for grading the aggressiveness of CRC was first developed by Cuthbert Duke in 1932,³⁷ and it was not until 1995 that a classification of prognostic markers was first suggested by the College of American Pathologists.³⁸ Seventy years on, Dukes and TNM, along with the histological grade of the tumour, still remain the only reliable methods to dictate whether patients with CRC will receive adjuvant therapy. Unfortunately, under current guidelines, which have been set up in light of these prognostic indicators, there is still a cohort of patients who unnecessarily receive toxic chemotherapy while others who could potentially benefit from it are not recommended to receive it. In view of this, the prognostic value of Ki-67 in colorectal cancer has recently been revisited by the National Surgical Adjuvant Breast and Bowel Project Collaborative Study group in an attempt to define a marker on which therapeutic decisions could be made with greater precision.³⁹

Ki-67 was selected for investigation because of its potential value in identifying populations of individuals who may or may not benefit from adjuvant chemotherapy. Significant prognostic value was found for patients with Dukes’ B and C disease although it was unable to differentiate between groups who might benefit from adjuvant chemotherapy. Prior to this study, there have been contradictory reports regarding the role of Ki-67 role as a prognostic indicator in CRC. Palmqvist and colleagues⁴⁰ showed that those tumours with low Ki-67 expression at the invasive margin had a poor prognosis while Kimura and colleagues⁴¹ concluded that the tumours with high expression at the site of deepest invasion had a worse prognosis. In contrast, both Kubota and colleagues⁴² who analysed 100 cases of CRC using Ki-67, and Kyzer and colleagues⁴³ who examined 30 resected specimens of CRC, concluded that Ki-67 could not be used as a prognostic determinant in CRC.

Our study has demonstrated that ¹⁸FLT PET may be used as a non-invasive method of quantitating cellular proliferation in CRC. This may offer the opportunity for CRC to be imaged at a pretreatment stage for those undergoing surgery and perhaps define the population of patients who might benefit from neoadjuvant therapy. In addition, it may also help in identifying a group of patients with recurrent disease who would not benefit from firstline chemotherapy as a result of the low proliferation rate of the tumour, as demonstrated by the low SUV on ¹⁸FLT PET imaging.

This would offer the clinician a further prognostic marker with which to assess the aggressiveness of the disease and perhaps allow a more precise method for therapeutic decisions to be made. In real terms, a decision regarding response to chemotherapy in CRC is usually judged six weeks into a 12 week treatment cycle. The potential of avoiding six weeks of unnecessary toxic chemotherapy by differentiating the non-responders from the responders at a pretreatment stage is a prospect which cannot be ignored.

Future work will help in validating ¹⁸FLT PET as a potential tool for the non-invasive in vivo assessment of CRC and may enable more accurate evaluation of the antitumoral effects of new as well as established anticancer drug treatments aimed at competing with DNA synthesis.