FLT-PET for early response evaluation of colorectal cancer patients with liver metastases: a prospective study

Response evaluation during cancer treatment is normally based on the response evaluation criteria in solid tumours (RECIST) looking for tumour shrinkage [1]. However, molecular changes precede tumour shrinkage, and therefore, there is an increased interest in functional imaging in the search for early response evaluation attributes [2, 3].

[¹⁸F]-3′-Deoxy-3′-fluorothymidine (FLT) is a modified thymidine analogue developed as a positron emission tomography (PET) tracer to reflect proliferation. FLT crosses the cell membrane and becomes phosphorylated by thymidine kinase 1 (TK-1), which is the first enzyme in the salvage pathway of thymidine synthesis as part of the DNA synthesis. FLT is not incorporated into the DNA, but becomes dephosphorylated at a slow rate. Phosphorylation causes FLT to be trapped inside the cell for a relatively long period, making it suitable for tracer imaging [4]. Several small studies have shown a correlation between the uptake of FLT and the histopathological marker of proliferation Ki-67 [5]. It has also been shown that FLT accumulates proportionally with TK-1 [6]. Furthermore, we have previously shown in preclinical studies that FLT is a sensitive and early predictor of morphological therapy response in various cancers [7‐10].

Colorectal cancer (CRC) is the third most common type of cancer worldwide [11]. Approximately 20% of patients with newly diagnosed CRC have distant metastases at the time of diagnosis, the most common site being the liver [11]. Moreover, another 50% of CRC patients will develop liver metastases at some time during their disease. The only definitive treatment for metastatic CRC is surgical resection if possible [12]. A revolution in the treatment of metastatic CRC has taken place in the last decade—especially in the surgical field—resulting in more patients being candidates for resection, which has been followed by a significant improvement in the 5-year survival rate of patients resected for liver metastases [12]. It is therefore crucial to identify patients who are not responding as early as possible as an ineffective treatment could jeopardise their chance of being candidates for resection. Furthermore, the benefit of saving patients from unnecessary side effects of a treatment that does not work is also worthwhile in a palliative setting.

Preclinical studies with cell lines and xenografts [4, 13‐16] as well as a few clinical studies [17, 18] have suggested the possibility that FLT-PET could be used for the early assessment of treatment response in CRC. To the best of our knowledge, to date, the potential of FLT-PET for the early response evaluation of CRC during chemotherapy has been investigated in only three clinical studies with 2, 12 and 18 patients, respectively [17‐19]. Just one of these studies used liver metastases as measurable outcome lesions [17], while none of them used the currently preferred treatment regimen, i.e. addition of a targeting agent to the chemotherapy.

The present study therefore aimed to assess the potential of FLT-PET/computed tomography (CT) in the early response evaluation of CRC based on measurable liver metastases. The early changes in FLT uptake were compared to RECIST after three cycles of treatment. It is acknowledged that RECIST, especially after the introduction of bevacizumab, have limitations in response evaluation in liver metastases since increasing necrosis does not necessarily lead to tumour shrinkage [20, 21]. Hence, it has been suggested that, in combination with volumetric changes, the morphologic changes of liver metastases such as, sharpness of the rim and the heterogeneity of contrast enhancement seen on CT or magnetic resonance (MR) may also be useful for evaluating response. Although morphology is not included in the clinical guidelines, a study by Chun and co-workers found that morphological criteria based on attenuation changes on CT correlated with the pathologic response and was associated to overall survival (OS) [20]. Therefore, in this study, changes in FLT uptake shortly after the initiation of treatment were compared not only to the outcome of RECIST but also to the changes in attenuation of liver lesions on CT after three treatment cycles.

Imaging reflecting early cellular changes predicting the response to a given treatment is a non-invasive approach to personalised medicine. The current clinical means of measuring the clinical response by imaging is performed by evaluation using RECIST which is strictly based on changes in tumour size [1]. Since cellular and molecular changes may precede changes in tumour volume, new methods are under investigation [24]. As [18 F]-fluoro-deoxy-glucose (FDG) not only accumulates in tumour tissue but is also seen in inflammatory tissue, new tracers more specific for malignancy are therefore warranted. FLT has been assessed for use in early response evaluation in a few small studies and has been found to be superior to FDG in differentiating tumour from inflammation [25, 26].

To the best of our knowledge, this study is the largest study evaluating FLT-PET/CT for early response assessment of chemotherapy-treated CRC patients with liver metastases. Only one other study has previously evaluated liver metastases as measurable lesions. The study included a mixture of patients with colorectal and breast cancer, and a change in FLT could separate responders from non-responders with a sensitivity of 83% and a specificity of 78% [17]. Other studies have only evaluated extrahepatic lesions. A reason may be that 18-F accumulation is high in normal liver tissue because of FLT glucuronidation in the hepatocytes [18, 19]. In spite of this obstacle, we found it important to investigate the FLT methodology in liver metastases because they very often are the main target in treatment of CRC. Furthermore, liver metastases are more suitable for accurate repeated measurements in contrast to the primary tumour which may be mobile and therefore difficult to capture in the same position from scan to scan. A significant decrease in FLT uptake was seen in the group of responders compared to non-responders; however, this must be interpreted cautiously due to the small sample size in the non-responders’ group. In this patient cohort, looking at the patients individually, we could not separate responders from non-responders based on the changes in FLT uptake.

It can be argued that it is always meaningful to characterise SD according to RECIST as non-responders because SD may represent lesions decreasing up to 30% [1]. For the same reason, instead of forcing data to fit four bins, it has become more and more common to study treatment outcome in waterfall plots since changes in tumour size occur as a continuum [27]. In common clinical practice, SD suggests that the current treatment regimen is continued since this policy has shown superior survival data in metastatic CRC [28, 29]. In this study, all patients with SD had a decrease in lesion size, meaning that the treatment had a beneficial effect. The key point of early response evaluation is to characterise true non-responders so that the current treatment can be discontinued. This is especially important for those patients with potentially resectable disease where an aggressive approach is necessary to achieve resectability. Hence, those patients with “true” PD need to be identified. In this study, the one patient with PD stood out as having the highest increase in FLT uptake; however, no conclusion can be drawn from a single patient with PD.

The majority of the patients in this study received a molecular targeted treatment (70%), in most cases bevacizumab. Bevacizumab is an anti-angiogenic drug that causes morphological changes with a decrease in vascularity followed by a decrease in attenuation and enhancement at imaging even though no changes in tumour size are necessarily seen. It can therefore be debated whether it is reasonable to compare FLT/PET to CT using RECIST because these criteria are often found to be inadequate when bevacizumab is part of the treatment [20, 21]. The fact that many liver metastases resected after neoadjuvant chemotherapy may exhibit a considerable degree of necrosis even though no shrinkage has been proven prior to resection indicates that changes in tumour size do not always reflect the pathologic response [30]. A study by Chun et al. [20] evaluated the treatment response in CRC liver metastases treated with chemotherapy, including bevacizumab. They found that both the morphological changes and RECIST outcome correlated with the histological evaluation of resected metastases, focusing on vital tumour cells, but prediction of minor pathological response was significantly better by morphological changes compared to RECIST. They also demonstrated a correlation between the morphological changes after three cycles of treatment and survival; in contrast, no correlation was seen with RECIST. There are no previous studies on FLT-PET for response assessment during treatment with bevacizumab in CRC. However, a preclinical study investigated the effect of bevacizumab on orthotopic glioblastoma xenografts and found no changes in FLT when bevacizumab was administered as monotherapy [31]. Four of the patients included in the present study were characterised as having SD but showed a good morphologic response, reflecting the inadequacy of RECIST for response evaluation of targeted treatments.

Even though this is the largest FLT/PET study performed on humans with CRC liver metastases, the sample size is still relatively small with too few patients with PD. Another limitation is the variation of treatment regimens even though this reflects the everyday clinical situation. It can be debated whether correlation of the early changes in proliferation measured by changes in FLT uptake with a RECIST assessment of tumour shrinkage or a CT morphologic response is best. Our analysis of changes in FLT uptake and survival was hampered by the fact that a majority of the patients subsequently had a liver resection and thereby were potentially cured. In the ideal setting, all patients should have a liver resection after three cycles of treatment, thus making it possible to correlate changes in FLT uptake to the histological tumour regression, or if no patients were liver resected, it would be possible to correlate FLT uptake changes to the entire cohort’s progression-free survival or OS. In our study, a rather large fraction of the patients (16 out of 27) did undergo a liver resection, thereby making it impossible to correlate early changes in FLT uptake to progression-free survival. It is difficult to predict who will become resectable for their liver metastases following neoadjuvant treatment.

Earlier studies have raised concern about the use of FLT for early response assessment due to the impact of the drugs given on the de novo and salvage pathway of thymidine synthesis. 5-FU inhibits thymidylate synthase and consequently inhibits the de novo pathway following a depletion of the thymidine pool necessary for DNA synthesis. It is known that 5-FU gives rise to an increase in FLT uptake shortly after the drug is given, termed the FLT flare. This flare has been explained by a redistribution of the transporter from the intracellular compartments to the cell surface with no changes in the level of ATP and TK-1 [32], meaning that FLT reaches the cell but is not phosphorylated and therefore not trapped inside the cell. Previous studies have shown that changes in FLT uptake seen after this flare reflect the changes in tumour shrinkage when treated with 5-FU. In our study, all patients were treated with 5-FU either as an infusion or as the prodrug capecitabine. The evaluation FLT-PET scans in this study were performed considerably later than when the flare is reported to occur [6, 18, 32, 33]; therefore, we do not have any concerns regarding potential influence from the flare.

In conclusion, we were not able to demonstrate correlation between early changes on FLT-PET following the first cycle of treatment and response on subsequent CT scan, neither when evaluated according to RECIST or morphology criteria (Chun). Thus, FLT-PET does not seem to be an applicable methodology for early response evaluation of mCRC treated with 5-FU based chemotherapy.