Clinical value of 3'-deoxy-3'-[18F]fluorothymidine-positron emission tomography for diagnosis, staging and assessing therapy response in lung cancer

Lung cancer has the highest mortality rate of any tumour type. The main driver of lung tumour growth and development is uncontrolled cellular proliferation. Poor patient outcomes are partly the result of the limited range of effective anti-cancer therapies available and partly due to the limited accuracy of biomarkers to report on cell proliferation rates in patients. Accordingly, accurate methods of diagnosing, staging and assessing response to therapy are crucial to improve patient outcomes. One effective way of assessing cell proliferation is to employ non-invasive evaluation using 3'-deoxy-3'-[18F]fluorothymidine ([18F]FLT) positron emission tomography [18F]FLT-PET. [18F]FLT, unlike the most commonly used PET tracer [18F]fluorodeoxyglucose ([18F]FDG), can specifically report on cell proliferation and does not accumulate in inflammatory cells. Therefore, this radiotracer could exhibit higher specificity in diagnosis and staging, along with more accurate monitoring of therapy response at early stages in the treatment cycle. This review summarises and evaluates published studies on the clinical use of [18F]FLT to diagnose, stage and assess response to therapy in lung cancer.


Background
Lung cancer (LC) remains the leading cause of cancerrelated mortality worldwide, accounting for about 18.4 % of all cancer cases in 2018. It ranks first in mortality and incidence in men. In women it has the second highest mortality and the third highest incidence [1]. LC is subcategorised histologically into small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC), of which adenocarcinoma (ADC), squamous cell carcinoma (SqCC) and large cell lung carcinoma (LCC) represents almost 80 % of diagnosed cases [2]. These subcategories show different growth patterns and might be associated with different prognoses [3].
NSCLC can be stratified according to the tumour size, nodal involvement and metastases (TNM). SCLC is usually divided into limited or extensive disease, although the TNM system has been adopted by some clinical committees. These models, such as those provided by the Veterans Administration Lung Study Group and later modified by the International Association for the Study of Lung Cancer (IASLC), have proven useful to guide prognosis and patient stratification for therapy [4].
Uncontrolled cell proliferation is a fundamental hallmark of malignant tumour growth. The biomarker Ki-67 is considered the gold standard for assessment of cell proliferation due to the strong correlation between the cell proliferation rate and Ki67 expression in cells [5,6]. However, it exhibits several potential drawbacks in the clinic, including the requirement for invasive collection of biopsies and possible sampling bias, due to poorly representative biopsies collected from heterogeneous tumours [7]. Over the past few decades, noninvasive positron emission tomography (PET) imaging has played an increasingly important role in LC management. [ 18 F]Fluorodeoxyglucose ([ 18 F]FDG) is the most widely used radiotracer in PET. It is a glucose analogue which monitors glucose metabolism, based on the concept that tumour cells take up significantly higher levels of glucose than normal tissues. It has been a beneficial adjunct in characterisation of intermediate solitary lung nodules and pre-treatment detection and staging of distant metastases [8]. However, [ 18 F]FDG uptake is modulated by multiple signalling pathways, so is not selective enough to specifically assess changes in the cell proliferation rate. Moreover, a meta-analysis of [ 18 F]FDG in LC showed its extreme heterogeneity in LC in pulmonary areas with inflammations [9].
The first PET radiotracer introduced for in vivo proliferative imaging was 11 C-labelled thymidine ([ 11 C]thymidine); however, its rapid degradation with a half-life time of only 20 min is a major limitation [10]. 3'-deoxy-3'-[ 18 F]fluorothymidine ([ 18 F]FLT) was therefore subsequently employed, as this 18 F-labelled thymidine analogue has a half-life of about 110 min. [ 18 F]FLT is initially phosphorylated by cytosolic thymidine kinase 1 (TK1) into FLT-monophosphate and then subsequently further phosphorylated to make the diphosphate and triphosphate nucleotide (Fig. 1). TK1 is the principal enzyme that controls the rate of nucleotide recycling via the salvage pathway of DNA synthesis. After [ 18 F] FLT is phosphorylated to the triphosphate analogue it cannot be further metabolised or incorporated into the DNA molecule. It is therefore trapped intracellularly as its high hydrophilicity means that it cannot readily cross the cell membrane. This cell sequestration of [ 18 F]FLT is due to the substitution of the hydroxyl group at the 5'-end of thymidine, which is essential for ligation of DNA, with a fluorine-18 radionuclide. TK1 expression levels increase dramatically in proliferating cells, and there is close correlation between expression of the enzyme and the cell proliferation rate [11,12]. Close correlation has also been observed between Ki-67 scoring and [ 18 F]FLT uptake in many tumour types [5], including LC [6]. Therefore, [ 18 F]FLT is a very plausible alternative to Ki-67 for assessing the proliferation rate in tumours. It also offers the advantages of avoiding the need to collect invasive biopsies and allows evaluation of proliferation heterogeneity across the entire tumour, minimising both patient discomfort and sampling errors. Moreover, [ 18 F]FLT-PET is able to produce repeated 3D images for multiple cancer sites simultaneously, which is a distinct advantage for accurately assessing response to therapy. Alternative proliferation radiotracers have been also developed as direct biomarkers for DNA synthesis as they can be incorporated into DNA. This includes 2'-deoxy-2'-fluoro- . However, these radiotracers are poor substrates for TK1 and show low uptake in highly proliferating tissue compared with [ 18 F]FLT. This probably reflects their phosphorylation by mitochondrial thymidine kinase 2 (TK2) rather than TK1 [13], making [ 18 F]FLT potentially superior to other proliferation radiotracers.
Relatively similar imaging protocols are followed in most [ 18 F]FLT-PET scans. No specific preparations are required by patients although it is advantageous if they are instructed to drink up to 1 L water before imaging to stimulate tracer excretion from the renal calyces [14,15]. The tracer dose administered ranges from 130 to 550 MBq. Most protocols employ static scans, although some more involved studies use dynamic scans to obtain more complex datasets. Subtly different scan timings and imaging parameters such as field of view and reconstruction techniques are also performed. For PET/CT, the patient is usually positioned supine, with arms raised to minimise beam hardening. Image evaluation is usually performed by two or more nuclear medicine physicians blinded to patient's data.
The aim of this literature review paper is to review the clinical value of [ 18 F]FLT-PET proliferative imaging for diagnosis, staging and assessing response to therapy in LC. A secondary objective is to analyse the uptake values of [ 18

Search strategy
To identify all relevant publications, a systematic search of Scopus/PubMed databases was implemented from inception to 1st December 2020 using combinations of the following keywords: "positron emission tomography", "PET", "lung cancer", "fluorothymidine" and "[ 18 F]FLT".

Selection process
All potentially relevant publications were screened for eligibility. Initially, titles and abstracts were screened and if necessary, the full texts were scrutinised as well. Studies were included if they met the following criteria:

Search results
The literature search identified a total of 319 studies: 210 in Scopus and 109 in PubMed. After removing the studies that were found in both databases, 221 studies remained. Figure 2 presents the flow chart of the search and selection procedure. Out of 221, only 31 studies were included as eligible, and out of these 31 studies, 17 studies were included for quantitative synthesis. A summary of [ 18 F]FLT-PET LC clinical studies and their findings is listed in Table 1 and 2.

Diagnosis studies
The past two decades have witnessed a growing attention to [  Uptake patterns with both tracers in NSCLC sub-types are consistent with the proliferative patterns demonstrated by Ki-67 in a previous study in the different subtypes of NSCLC histology [26].

Staging studies
Staging of LC is another clinical application that has been evaluated in several studies. In general, [ 18 [29]. Using [ 18 F]FDG, anatomic responders showed significantly lower uptake than anatomic non-responders 2-3 weeks after treatment, while no significant difference was found between these subgroups using [ 18 F]FLT.

Radiotherapy
Preliminary data of studies that evaluated response to radiotherapy is more promising. Vera et al. [31] demonstrated significant changes in SUVmax in 5 patients treated with 46 Gy radiotherapy. This decline of [ 18 F]FLT uptake was supported by another study (n = 20) which tested correlation with clinical results after treatment with carbon ion radiotherapy [32]. Baseline SUVmax was predictive of response as patients who died or developed recurrence had significantly more [ 18 F]FLT uptake than those did not (p = 0.007 and p = 0.008, respectively). Furthermore, primary tumours showed decreased SUVmean and SUVmax despite the absence of morphological change [33].

Chemoradiotherapy
Three studies have investigated the potential of [ 18 F]FLT for assessing response to chemoradiotherapy. Response to chemoradiotherapy was firstly assessed in a small study (n = 5), which observed decreased [ 18 F]FLT uptake in primary tumours after therapy [34]. A subsequent study (n = 20) found that [ 18 F]FLT was more sensitive than [ 18 F]FDG in assessing response to radical chemoradiation [35]. Median SUVmax were 14 and 6 at baseline and 10 and 3 two weeks post-therapy for [ 18 F]FDG and [ 18 F]FLT, respectively. The same group also studied correlation with clinical outcome in a larger cohort (n = 60) treated with 60 Gy radiotherapy combined with carboplatin and paclitaxel or cisplatin (32). Stable disease, as assessed using RECIST criteria, was associated with longer OS and progression-free survival (PFS) than patients with complete or partial response. This paradox could be ascribed to the weakening of tumouricidal effect of chemotherapy by suppressed proliferation [36].

Targeted therapy
Multiple studies have found high potential of [ 18 F]FLT to predict LC response to targeted therapy. Yang et al. analysed the prediction of response to anti-angiogenic  [38,39]. In the static PET study, SUVmax was decreased −11% in cycle 1 treatment with the anti-angiogenic x-82. After administration of cycle 2 x-82 combined with the chemotherapy docetaxel on day 21, SUVmax was greatly decreased to −44% (34). Decline of cell tumour proliferation and vasculature were also exhibited 2 weeks after treatment with axitinib followed by an increase in washout period during the week of treatment break [39]. Assessment of epidermal growth factor receptor-tyrosine kinase inhibitors (EGFR-TKI) response with [ 18 F]FLT is also an active area of research. Typical scans of [ 18 F] FLT uptake before and after EGFR-TKI therapy in a LC patient responding to therapy are shown in Fig. 4. The value of [ 18 F]FLT in this role is due to well-established clinical benefits of EGFR-TKI, as they are being used as first-line therapy in select NSCLC patients [47]. Three studies investigated the performance of [ 18 F]FLT in assessing tumour progression in correlation with tumour size measured by computed tomography. Sohn et al. utilised [ 18 F]FLT to predict response to gefitinib in 31 adenocarcinoma patients [40]. A significant difference (p < 0.001) was shown between responders and non-responders. Mileshkin [42]. Moreover, further clinical assessment of [ 18 F]FLT role for early monitoring of response to newer generations of EGFR-TKI agents is still required in order to better understand their associated mechanisms of responsiveness and resistance [48].
The predictive role of clinical outcome using [ 18 F]FLT uptake was also studied in EGFR-TKI studies and showed positive results. Sohn [44]. Moreover, apart from one study with a small patient cohort (n = 15) [45], [ 18 F]FLT uptake was correlated with PFS in multiple independent studies with larger cohorts [41][42][43]. However, in most studies OS did not correlate with [ 18 F]FLT uptake [42,45].  [50]. Furthermore, highly proliferating cells such as bone marrow act as a sink, reducing the amount of [ 18 F]FLT available in the body to be taken up by tumours [51]. In contrast, cancer cells tend to take up relatively high levels of [ 18 F]FDG. This is due to multiple factors. Firstly, they express specialised receptors such as GLUT1, which facilitate the active transport of glucose or [ 18 F]FDG across the cell membrane, and partly due to the Warburg effect, which results in relatively high glucose metabolism in tumour cells. Secondly, whilst most cancer cells are highly metabolically active only a proportion of them are proliferating at any time. Finally, inflammatory cells are usually present in malignant tumours, which will further increase [ 18 F]FDG uptake in regions of interest.

Discussion
One interesting suggestion is that using [ 18 F]FLT and [ 18 F]FDG-PET in combination could potentially provide better diagnostic performance than the individual tracers. Indeed, this strategy showed higher sensitivity and specificity in diagnosing pulmonary lesions [21]. However, this benefit is offset by the higher costs associated with performing two PET scans and the increased radiation burden imposed by this dual tracer approach. There is an effective dose equivalent of 0.031 mSv/MBq and 0.029mSv/MBq for [ 18 (Fig. 3). This result was expected given the defined uptake mechanism for [ 18 F]FDG in tumour cells combined with the Warburg effect, which results in tumour cells utilising high amounts of glucose.
A comparison of [ 18 F]FDG-SUVmax detected no significant difference between the 4 sub-types of LC (p > 0.05). However, the statistical power of this analysis is limited by the number of SCLC studies performed to date and the relative high variation in uptake values between studies in the other tumour types. Publication of additional LC datasets is required to confirm if there is any clear differences in [ 18 F]FDG uptake between LC sub-types. In contrast, a comparison of [ 18 F]FLT-SUVmax between the different histotypes observed that [ 18 F]FLT-SUVmax was significantly lower in SCLC than in SqCC (p < 0.05). This is a new observation which suggests that there may be something subtly different between these LC sub-types. There are 3 potential straightforward explanations for this observation. The first possibility is that SCLC proliferates at a slower rate. This seems unlikely given that SCLC is known to be an aggressive sub-type that grows quickly. The second possibility is that there is less cellular [ 18 F]FLT uptake into SCLC, either because these tumours have limited access to [ 18 F]FLT in the blood or because less radiotracer can cross the plasma-membrane, perhaps suggesting that the cells may express fewer hENT transporters. The third possibility is that SCLC could rely more than SqCC on the de novo pathway to synthesise nucleotides rather than using the salvage pathway [53], resulting in less [ 18 F]FLT uptake into cells via hENT transporters. To our knowledge neither of the latter possibilities have been investigated to date.
Evaluating the potential of [ 18 F]FLT-PET for assessing early response to therapy is a logical step, given the clear utility of [ 18 F]FDG-PET in this role in LC. The rationale for using [ 18 F]FLT-PET to assess therapies that selectively affect proliferation is especially compelling, as these treatments may inhibit proliferation without affecting either metabolic rate or causing tumour shrinkage [54]. Only 4 out of the 18 LC response assessment studies identified in this review did not observe some utility for [ 18 F]FLT in this role [28,29,36,42]. One of these studies [28] assessed a chemotherapeutic agent that is antagonist to thymidylate synthase. This may impact TK1, resulting in poor correlation between [ 18 F]FLT uptake and clinical outcomes [53]. Whilst another study [29], which studied platinum-based chemotherapy, showed no correlation with CT response. The reason for this discrepancy is unknown. These findings suggest that the complex effect of cytotoxic therapeutics on cell signalling pathways needs to be fully considered to understand how each agent may affect [ 18 F]FLT uptake into cells. Indeed, a previous systematic review that evaluated the role of [ 18 F]FLT as a measure of therapy response in different tumours also concluded that [ 18 F]FLT-PET is not as useful as [ 18 F]FDG-PET for assessing chemotherapeutic response in NSCLC [55]. However, studies evaluating these types of therapies are still scarce and findings from more studies with larger patient cohorts might be more conclusive. A chemoradiotherapy study which employed [ 18 F]FLT-PET to evaluate therapy response observed that stable disease measurements compared with complete or partial response based on CT was associated with longer OS and progression-free survival (PFS) [36]. This paradox could be ascribed to the weakening of tumouricidal effect of chemotherapy by suppressed proliferation [36]. Finally, a study evaluating [ 18 F]FLT-PET to assess EGFR-TKI response found no correlation with CT measurements [42]. However, this observation may be due to utilisation of size measurements to assess EGFR-TKI; change in tumour size may be delayed or may not occur at all with this type of therapy, creating a discrepancy between these two different metrics [54].
The most promising application for [ 18 F]FLT is assessing response to targeted agents that selectively inhibit cell proliferation. In this role [ 18 F]FLT is utilised as a precise tool for detecting the effects of antiproliferative therapies. Apart from one study with a small patient cohort (n = 15) [45], [ 18 F]FLT uptake correlated with PFS in multiple independent studies with larger cohorts [41][42][43]. Interestingly, in most studies OS did not correlate with [ 18 F]FLT uptake [42,45]. This may be due to the subtly different metrics captured by PFS and OS. The OS statistics may be somewhat compromised by the limited number of patients recruited in most of these studies and/or the limited follow up time. Indeed, the study with the largest patient group (n = 51) and follow up time observed correlation between [ 18 F]FLT uptake and OS 8 weeks after treatment; the longest follow up time in other studies was only 6 weeks [41,42]. Although OS is a gold standard measure to demonstrate clinical efficacy, larger patient numbers and longer follow up are required to establish [ 18 F]FLT as a reliable measure of this endpoint [56]. PFS is more advantageous than OS in that it assesses both stable disease and responsive disease. This renders it a more reasonable clinical measure for targeted therapies, which often benefit patients mainly through prolonged stable disease rather than tumour shrinkage [57]. The assessment of pharmacodynamic endpoints is another extremely important function for [ 18 F]FLT. Scarpelli et al. [38,39] demonstrated that the increase of [ 18 F] FLT uptake is apparently caused by the washout period during treatment. This type of study, which evaluates the pharmacodynamic endpoints, is of great importance to develop better understanding of drug resistance mechanisms, as it helps clinicians to tailor more effective treatments and to minimise systematic toxicity caused by ineffective drugs [38]. Together, these preliminary findings suggest that [ 18 F]FLT is a useful tool for evaluating response to targeted therapies with anticipated cytostatic effects, but further studies with different targeted agents, larger patient cohorts and longer follow ups are warranted to fully understand its potential in this role.
The studies mentioned above have multiple limitations. One limitation is that most studies are singlecentre observational studies with small patient cohorts. This makes it difficult to know how readily the study findings would translate to other research centres or if similar results would be obtained using larger patient cohorts. While most studies used a relatively similar imaging protocol, there were no standard instructions for patient preparation. Some studies encouraged patients to drink water before imaging, which would help to reduce background radiation, whereas others did not. In addition, there were differences between studies in tracer dose administered, which is likely to affect the signal to background ratio. Most studies utilised static scans, with similar times between tracer injection and imaging. However, the use of different reconstruction techniques and metrics for assessing uptake (e.g. FLTmax vs FLTmean) would result in subtly different values being calculated from the same dataset. It is important that methods are optimised and standardised, so that the results obtained in one study can be directly compared with those from another, and so that multi-centre studies can generate equivalent data at all participating centres. This issue has been recognised within the PET community and has resulted in core labs being set up to ensure quality control for PET clinical trials [58]. The limited number of studies evaluating LC response to various types of treatment affects the conclusiveness of study findings. Moreover, size-based measurements were used as a standard reference in studies evaluating responses to targeted therapy although tumour size would not be affected with this type of therapy. OS was also used as the gold standard for assessing clinical outcome in studies evaluating responses to targeted therapy although it showed limitations compared with surrogate metrics such as PFS. Therefore, extra attention is also needed to ensure that the most suitable metrics of clinical outcome are used to study such therapies with [ 18 F]FLT-PET.

Conclusions
1. Overall, [ 18 F]FLT seems to have better specificity in diagnosis and staging but lower sensitivity with significantly lower uptake than [ 18 F]FDG. This significant difference was also observed after comparing uptake values of both tracers in the main histotypes of lung cancer that showed consistence with established histological patterns of cell proliferation in