Update on Quantitative Imaging for Predicting and Assessing Response in Oncology

Abstract

Molecular imaging has revolutionized clinical oncology by imaging specific facets of cancer biology. Through non-invasive measurements of tumor physiology, targeted radiotracers can serve as biomarkers for disease characterization, prognosis, response assessment, and predicting long-term response / survival. In turn, these imaging biomarkers can be utilized to tailor therapeutic regimens to tumor biology. In this article, we review biomarker applications for response assessment and predicting long-term outcomes. F-fluorodeoxyglucose (FDG), a measure of cellular glucose metabolism, is discussed in the context of lymphoma and breast and lung cancer. FDG has gained widespread clinical acceptance and has been integrated into the routine clinical care of several malignancies, most notably lymphoma. The novel radiotracers 16α−¹⁸F-fluoro-17β-estradiol (FES), and ¹⁸F-fluorothymidine (FLT) are reviewed in application to the early prediction of response assessment of breast cancer. Through illustrative examples, we explore current and future applications of molecular imaging biomarkers in the advancement of precision medicine.

Introduction

New oncologic therapies targeted to specific aspects of tumor biology have revolutionized clinical oncology. By pairing such therapies with diagnostic tests that predict and monitor response, oncologic care can be optimized. This paradigm—deemed precision medicine—can increase the probability of a favorable response (1). Toward this goal, cancer biomarkers have been developed and integrated into research and clinical settings. A biological marker, or biomarker, is a measure of a biologic characteristic that can be used to characterize a normal or pathologic process, or measure a pharmacologic response to an intervention. In characterizing disease, biomarkers have been utilized for diagnosis, staging, prognosis, and prediction/monitoring of response (2). In oncology, both tissue and imaging biomarkers have been employed successfully.

Tissue-based biomarkers have been translated into the clinic with direct impact on patient care. For example, the Oncotype DX breast recurrence score from tissue specimens has been utilized to direct adjuvant therapy of breast cancer. Adjuvant chemotherapy is recommended for patients with a high recurrence score, but not for patients with a low recurrence score (3). However, inherent limitations in tissue sampling restrict the use of tissue-based biomarkers. The invasive nature of tissue acquisition limits the frequency of evaluation, of particular importance in studies monitoring the effect of treatment where serial measurements may prove useful. With limited sites of disease often sampled, heterogeneity of a patient’s disease burden typically cannot be assessed simultaneously, potentially failing to capture sites of disease that will ultimately drive progression. Tissue sampling is also prone to sampling error (4). Non-invasive imaging biomarkers, such as those acquired with positron emission tomography (PET) imaging, overcome many of these limitations and capture a more comprehensive view of a patient’s disease status. Imaging biomarkers complement tissue biomarkers, enabling more personalized oncologic care.

As a non-invasive measure of specific molecular processes, PET imaging biomarkers have been utilized in many aspects of clinical medicine and research, including prediction/prognosis, early and late response assessment, and as a surrogate response endpoint (Figure 1). Indeed, imaging with the glucose analog FDG has found widespread clinical acceptance in staging and responses assessment of several malignancies, perhaps most notably lymphoma. The development of combined PET/CT imaging further accelerated clinical utilization of FDG-PET, both improving accuracy of the test as well as providing anatomic correlation (5, 6).

Figure 1: Types of Oncologic Response Assessments.

This chart demonstrates the points at which molecular imaging evaluation can impact patient care during the oncologic treatment timeline. This includes prognostic evaluation at the time of diagnosis, predictive response evaluation prior to treatment selection, early response evaluation after initiating treatment, and late or end-of-treatment response evaluation to predict survival or risk of relapse.

In this review, we summarize the current approach to standardization of PET radiotracer uptake analysis. We discuss the most utilized quantitative and qualitative criteria for treatment response assessment. We also discuss notable PET imaging biomarkers, including selected novel imaging agents for response assessment in oncology, with examples from both clinical and research settings. In particular, we will focus on early response assessment. It is at this time point, whereby molecular imaging can demonstrate an early pharmacodynamic effect of treatment before an anatomic change, that we believe PET molecular imaging holds particular promise in response assessment.

Standardized Approach to Uptake Analysis

Given its ability to measure absolute radionuclide concentration, PET is an inherently quantitative imaging modality. Indeed, quantitative measures of uptake—most notably SUV_max —have become widely utilized for staging and response assessment in the clinic. Accurate quantification of radiotracer uptake, and by extension accurate assessment of response, requires standardized study protocols and patient preparation.

The uptake of radiotracer in tumors is a dynamic time-dependent process. After injection into the bloodstream, the radiotracer must exit the vasculature and interact with its intended target. To characterize these complex interactions, full kinetic modeling of dynamic images can be performed to estimate biologically relevant rate constants (e.g. the delivery of FDG and the rate of FDG metabolism) that are independent of time. Such imaging, though, requires prolonged imaging, a measure of tracer availability (e.g. an input function that may require blood sampling), and additional analysis (7–9). To simplify imaging protocols for clinical trials and the clinic, static imaging approaches are utilized. For FDG-PET, imaging at one hour after FDG injection has been widely accepted. FDG uptake at one hour, though, is not constant; as long as FDG uptake is available for the tumor from the blood, radiotracer uptake will increase over time. As such, a constant imaging time after radiotracer injection is strongly recommended for static imaging (8, 10).

The dynamic interaction of radiotracer with tissue also necessitates standard patient preparation. For imaging with FDG, a glucose analog, plasma glucose levels should be measured prior to imaging to avoid potential competition with FDG. A goal of 150–200 mg/dL or less is desired. Although complex correction factors have been proposed to account for native glucose concentrations, corrections are not routinely used if a patient has an acceptable blood glucose level, and patients are generally rescheduled for elevated blood glucose levels. In addition, patients should fast for at least four hours prior to injection so as not to elicit an insulin response to food, which would drive glucose into specific tissues and alter the biodistribution. Similarly, strenuous exercise is also discouraged. Patient weight should also be measured at each imaging session, noting that SUV is a semi-quantitative measure of radionuclide concentration normalized to dose and patient weight. For patients undergoing active treatment, weight changes may fluctuate, underscoring the importance of a timely, accurate weight (8, 10). These factors are known to impact the qualitative biodistribution of FDG, but may also impact quantitative tumor uptake (7).

Ideally, patients should be imaged on the same PET scanner to control for inherent differences in scanner characteristics, including calibration factors and reconstruction. A 2019 study of qualified PET scanners from four institutions demonstrated an average difference in lesion SUV_max of 18% between scanners at different sites. A difference of 8% was seen between scans performed at the same institution, either the on the same scanner or a different scanner of the same model. These differences were not attributable to patient factors described above (11). In an effort to standardize quantitative imaging approaches for clinical trials, two organizations have been formed: the Quantitative Imaging Biomarkers Alliance (QIBA) (12) and Quantitative Imaging Network of the National Cancer Institute (QIN) (13).

As discussed above, the semi-quantitative SUV obtained on static imaging has gained widespread acceptance in both clinical trials and research. To calculate an SUV, radionuclide concentration in a region of interest is normalized to the injected activity and a measure of radiotracer distribution volume. While body weight is often utilized as a proxy for distribution volume, alternative options, such as body surface area and lean body mass, may better account for tracer distribution (8). The most commonly used clinical uptake measurement is SUV_max, which is a measure of uptake in the voxel with the greatest uptake in a lesion. SUV_peak (average uptake over 1 cm³ in a lesion centered at the area of higher uptake in a tumor) is more commonly used for research purposes. Whereas SUV_max is the uptake in a single voxel in a region of interest, SUV_peak is the mean uptake in a small volume, and as such, thought to be less susceptible to statistical variation. SUV_mean over a larger volume, such as over the liver or mediastinal blood pool, is often used to assess background scan uptake for a general assessment of image quality (14). For FDG, these measures are highly reproducible quantitative values of glucose metabolism, and are important for developing standardized tools for response assessment (11, 15). Additional metabolic indices include metabolic tumor volume (MTV, the volume of a tumor above a defined threshold of signal intensity), and total lesion glycolysis (TLG, or the product of SUV_mean and MTV). These measures have been investigated as a biomarker of prognosis in several diseases (16, 17), but have not yet achieved widespread clinical acceptance.

Determination of Tumor Response by PET

Traditional response assessment in oncology is based on changes in tumor size as assessed by anatomic imaging such as CT or MRI. A favorable response to therapy should result in cellular death and, consequently, a decrease in anatomic size; the converse also holds. This concept forms the basis for the Response Evaluation Criteria in Solid Tumors (RECIST) criteria, most recently updated in 2009 (RECIST 1.1) (18). RECIST 1.1 is based on standardized anatomic measurements of up to five lesions to categorize treatment response into one of four categories: complete response, partial response, stable disease, and progressive disease (18). These criteria are widely utilized in clinical trials of novel therapeutics for solid tumor malignancies and by oncologic regulatory agencies worldwide (14, 19). However, RECIST 1.1 has some notable limitations, including the loss of valuable information on tumor response and predicting survival that occurs from reducing continuous data on tumor size to the four defined categories of response (14). FDG-PET/CT information can be used to inform the assessment of progression in RECIST 1.1, but only in a qualitative fashion and only in conjunction with anatomic findings. In addition, RECIST 1.1 limits response assessment for some sites of disease. For example, assessing response of breast cancer bone metastases by anatomic imaging is challenging, and RESIST 1.1 excludes bone metastases as a measurable site for response assessment. Patients with bone dominant metastatic disease are therefore often excluded from response-based clinical trials (18, 20).

Molecular imaging offers tumor-specific in vivo metabolic and biochemical information, overcoming some of the limitations of anatomic response assessment. Anatomic changes in tumor size seen on CT or MRI typically lag behind the treatment effects on a biochemical and molecular level. That is, anatomic changes secondary to cell death are downstream of molecular changes, such as those that can be detected with PET. As such, PET measures obtained after treatment initiation, or at the end of treatment, may inform clinicians to adjust therapy as well as serve as a surrogate endpoint and predict survival. With this goal, PET-specific response criteria have been developed, such as the PET Response Criteria in Solid Tumors (PERCIST) (14) and Lugano Criteria for lymphoma (21). These criteria are particularly helpful when evaluating response to cytostatic chemotherapies, which do not result in anatomic tumor reduction, but may demonstrate molecular changes, such as decreased metabolism on FDG-PET (22).

PERCIST was introduced in 2009 to overcome some of the limitations of anatomic response assessment (14). Under the updated criteria, PERCIST 1.0, SUV is corrected for lean body mass (SUL), and SUL is determined for up to five tumor sites with the most intense FDG uptake. Response is assessed as a continuous variable and reported as the percentage change in SUL_peak between the pre- and post-treatment studies (14). Preliminary studies demonstrate value of PERCIST in predicting treatment outcomes, including in patients with bone metastases. A study of bone dominant metastatic breast cancer patients imaged with FDG-PET prior to initiating new therapy demonstrated that both time-to-disease-progression and time-to-skeletal-related-events were both longer in responders compared to non-responders by PERCIST criteria (23).

Given that RECIST 1.1 and PERSIST measure two different aspects of malignancy, it is expected that the two methods may lead to different assessments of response, although few studies to date have directly compared the two. In a study of 20 patients with metastatic breast cancer, PERCIST demonstrated significantly higher accuracy in predicting progression free survival (PFS) compared to RECIST 1.1 (24). Additional studies also demonstrate superior predictive performance of PERCIST compared to RECIST 1.0 or 1.1 in non-small-cell lung cancer (25, 26), esophageal cancer (27), and Ewing sarcoma (28).

Compared to PERCIST, a more qualitative approach to tumor response is used in lymphoma treatment response. This is based on the Lugano Criteria, which utilizes the Deauville score, a 5-point visual qualitative scale comparing lesional FDG uptake to that of background liver parenchyma or blood pool (29). A comparison between this standardized qualitative approach and quantitative treatment response analyses will be described in detail later in this review.

Prognostic Biomarkers versus Predictive Response to Therapy

Prognostic biomarkers quantify the aggressiveness of a patient’s disease and the likelihood of mortality secondary to disease. In theory, these measures are independent of therapy, indicating an intrinsic characteristic of the disease (4). Ideally, studies of prognostic biomarkers should prospectively follow patients over the course of their disease. In practice, though, these studies are often retrospective and specialized statistical methods are frequently needed to account for varied time of follow-up and potential confounding factors (30).

Predictive biomarkers measure a biologic characteristic that predicts therapeutic efficacy of a certain treatment. This represents a slightly different paradigm from prognostic biomarkers, where only aggressiveness of the disease is selected, but not the actual treatment, although biomarkers often have both capabilities (4). For instance, a PET measure of drug-target expression may serve as a predictive biomarker for treatment efficacy with the paired targeted agent. Such predictive biomarkers can then be utilized to select patients for targeted treatments, increasing the probability of response. For example, FES-PET has been utilized as a non-invasive assay of estrogen receptor expression and as a predictor of response to targeted therapy. FES-PET has been validated as an accurate quantitative measure of in vitro ER expression (31–34), with a high specificity of 98% for ER-positive lesions (35). FES-PET accurately assesses ER expression in multiple sites of disease, including axillary lymph nodes as well as distant metastases (31, 32). In predictive biomarker studies, FES uptake predicts response to endocrine therapy, including selective estrogen receptor modulators and aromatase inhibitors, as both first line and salvage therapy (34, 36–39). Likewise, low or absent FES uptake, like immunohistochemistry (IHC), identifies patients with breast cancer lacking ER expression who are unlikely to respond to endocrine therapy (36, 38, 40). FES-PET has recently been approved for routine clinical use and is discussed in greater detail as an early response biomarker.

Early Pharmacodynamic Response

Measuring early (pharmacodynamic) response after treatment initiation offers early insight into treatment effect, allowing early prediction of eventual response. Such early measures may guide further treatment, possibly allowing discontinuation of a futile treatment or continuation of a promising treatment. Metabolic changes in tumor biology in response to treatment precede anatomic changes allowing for earlier assessment of response with PET than with anatomic imaging such as computed tomography (CT) (30). In this section, we will discuss early response biomarkers with a focus on breast cancer.

Breast Cancer

Breast cancer is a markedly heterogeneous disease, encompassing a number of biologically distinct tumor subtypes with unique pathologic features and prognoses. Molecular profiling has classified breast cancer based on expression of three receptors: the estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor 2 receptor (HER2). The presence or absence of expression of these three receptors has both prognostic and treatment implications, and determination of receptor status has become standard of care. The development of numerous targeted therapies in recent years has substantially improved the prognosis of breast cancer, particularly among the subtypes that express the ER, HER2, or both (41–43). However, even within subtypes, treatment response remains variable, and both intrinsic and acquired resistance to targeted breast cancer therapy occurs frequently (44, 45). Consequently, a clinical need exists for an early response biomarker. [¹⁸F]-fluorodeoxyglucose (FDG), 16α−¹⁸F-fluoro-17β-estradiol (FES), and [¹⁸F]-fluorothymidine (FLT) have each been studied in this context, summarized below.

FDG

The glucose analog FDG is the most extensively utilized radiotracer in oncologic molecular imaging, underscoring the ubiquity of dysregulation of glycolysis in malignancy. FDG-PET has become a key modality in cancer staging and restaging for many cancer types (9, 46–48). Currently, the National Comprehensive Cancer Network recommends that FDG-PET/CT be considered in the workup of Stage III and higher breast cancer (49) for staging and restaging. FDG-PET/CT offers high sensitivity for the detection of both extra-axillary nodal and distant metastases (50–52), and is among the most accurate modalities for staging recurrent breast cancer, particularly for identifying disease outside of the breast, with both high sensitivity and specificity (53–55). FDG-PET/CT performs favorably in the detection of bone metastases, particularly lytic lesions, compared to anatomic imaging and bone scan, as FDG images actual tumor as opposed to osseous reaction (56–58).

The role of FDG-PET continues to evolve, with emerging applications beyond staging/restaging. Because glucose metabolism is downstream of numerous therapeutic targets, FGD-PET may provide an early indicator of treatment response. In this context, FDG has been studied in the setting of neoadjuvant chemotherapy, where patients with large or locally advanced breast cancers are treated medically to reduce tumor volume and enable breast conservation surgery. FDG-PET has demonstrated high sensitivity for predicting tumor response with neoadjuvant chemotherapy (40, 53, 59–64). Several early prospective clinical trials of patients with Stage II or III breast cancer, comprised of multiple histologic tumor subtypes and neoadjuvant regimens, found that relative declines in tumor SUV of 45 to 50% after the first cycle of chemotherapy could accurately predict pathologic response (65, 66). A meta-analysis also demonstrated that FDG-PET/CT predicts early response to neoadjuvant therapy, with a sensitivity of 81% and specificity of 79%, and with a trend toward higher sensitivity after the second course compared to the first (60).

Despite early evidence suggesting the utility of FDG-PET in the early assessment of neoadjuvant response in breast cancer, FDG-PET has not yet gained widespread acceptance for this application. This is perhaps in part due to a lack of consensus that early imaging evaluation of treatment response could be utilized to effectively alter therapy and improve outcomes (61, 64, 67). Additionally, initial trials in this space were conducted in populations of mixed breast cancer subtypes, limiting the ability to define and quantify histologic subtype, and treatment-specific early response (61, 68). Given that baseline FDG uptake and kinetics vary by histologic subtype (69, 70), and that serial changes in FDG uptake can vary by subtype (68, 71), subtype-specific trials are necessary to direct FDG-PET assessment of early neoadjuvant response. Several such trials have been undertaken.

Image-based neoadjuvant response assessment has been well studied in triple negative breast cancer (TNBC), a subtype with poor prognosis and higher rates of metastasis and relapse (72, 73). Defined by the lack of expression of ER, PR, and HER2, these tumors account for 10–20% of invasive breast cancer, lack targeted treatment options, and are often resistant to the available cytotoxic chemotherapies (74, 75). TNBC demonstrates relatively high baseline SUV uptake on FDG-PET/CT, in keeping with inherent aggressiveness (76, 77), and suggesting potential for FDG as a biomarker of early response to guide therapy. The largest study to date investigating FDG in neoadjuvant response to breast cancer included 50 stage II and III TNBC patients, and found that the change in SUV_max between the baseline study and the follow-up study (performed after the second cycle of neoadjuvant therapy) was more strongly associated with pathologic outcome than was the absolute value of SUV_max on either exam. The mean decrease in SUV_max was 72% in patients who achieved pathologic complete response (PCR), compared to 38% in patients who did not (78). A cut-off of 50% decrease in SUV_max offered the highest accuracy (80%) for predicting pathologic complete response, with a positive predictive value of 67%, and negative predictive value of 96% (78). A similar study of mixed breast cancer subtypes included a cohort of TNBC, and also found that the decrease in FDG uptake from baseline to follow-up scan (performed at 6 to 8 weeks post initiation of neoadjuvant therapy) was predictive of PCR (79). However, another study of mixed subtypes found that the change in SUV from the baseline FDG-PET/CT to the follow-up scan (performed after the first course neoadjuvant therapy) was not predictive of PCR in the small cohort of TNBC (77). The discrepancy in results may be related to variation in the timing of the follow-up scans, treatment heterogeneity, or relatively small patient numbers. Larger trials in this space are indicated.

FDG-PET also offers a valuable role in the early assessment of HER2-targeted neoadjuvant breast cancer treatment. HER2-targeted therapy leads to a rapid decrease in FDG uptake in these tumors, which is dependent on the particular treatment regimen (61, 64, 80). In a prospective study of 57 women with HER2-positive breast cancer, treated with the anti-HER2 drug trastuzumab plus taxane-based neoadjuvant therapy, Humbert et al. found that low residual FDG uptake (defined as SUVmax < 2.1) after the first cycle of neoadjuvant therapy was predictive of pathologic complete response (81). A similar study by Groheux et al., with different timing of the post-treatment PET (after the second cycle), also found that low SUV_max post-treatment was most predictive of PCR, with an accuracy of 90% (82). In the prospective Neo-ALTTO (Neoadjuvant Lapatinib and/or Trastuzumab Treatment Optimization) trial of HER2-positive subjects, a subset of the patients underwent FDG-PET/CT at baseline and again at two weeks and six weeks post initiation of neoadjuvant therapy. Trial results showed that metabolic response was already apparent in the primary tumors after two weeks of therapy, and was highly correlated with metabolic responses at six weeks. In addition, pathologic complete response was associated with greater decrease in FDG uptake at both two and six weeks (80).

FDG-PET has also been studied prospectively to guide neoadjuvant therapy in HER2-positive breast cancer patients in the multicenter randomized phase 2 AVATAXHER trial. The trial aimed to assess the ability of FDG-PET to predict pathologic complete response early in the course of the standard neoadjuvant regimen of docetaxel plus trastuzumab, and to determine if the addition of the angiogenesis inhibitor bevacizumab could improve response in patients deemed unlikely to respond to the initial treatment regimen. Of the 142 enrolled subjects, 69 were predicted to be responders to the standard regimen based on the change in FGD uptake after the first cycle of therapy. 37 of the 69 subjects (53.6%) demonstrated pathologic complete response. The 73 subjects deemed non-responders by FDG-PET were randomized to continue the standard regimen or to receive bevacizumab in addition to the standard regimen. Pathologic complete response was seen in 21 of 48 (43.8%) of the non-responders who received the addition of bevacizumab, compared to 6 of 25 (24%) of the non-responders who remained on the standard regimen (83).

Although several studies of mixed breast cancer subtypes have suggested a lack of utility of FDG-PET for predicting treatment response in HER2-postive disease (79, 84), the low accuracy could be related to the small number of subjects in the HER2-positive cohorts, the timing of the scan following the initiation of treatment, or the specific quantitative parameters used to assess PET response (61, 64, 79, 84). The preponderance of evidence suggests some utility of measuring early response with FDG in HER2-positive patients, and larger multicenter trials are necessary to better optimize the integration of FDG-PET into treatment algorithms.

In both clinical breast oncology and research, the semi-quantitative SUV values obtained from static FDG-PET imaging have become standard for analysis of treatment response. However, full kinetic modeling from dynamic FDG-PET imaging may offer a more robust quantification of response assessment. Early studies suggested that kinetic analysis of dynamic FDG-PET may provide greater accuracy for measuring early response assessment and predicting outcomes than static SUV measurements (85, 86). A larger study by Dunnwald et al. sought to compare the accuracy of kinetic parameters, including K₁ (FDG blood-to-tissue delivery) and K_i (FDG flux constant) versus SUV as measures of treatment response and predictors of outcomes in patients with locally advanced breast cancer on neoadjuvant therapy. Patients underwent dynamic FDG-PET at baseline and again while on neoadjuvant therapy. Results showed that changes in kinetic measurements more accurately predicted pathologic response than did SUV (87). The authors postulate that kinetic parameters are more accurate in assessing therapy response and predicting outcomes for two reasons. First, evidence suggests that low baseline SUV values are associated with lower maximum detectable percent changes from baseline to mid-therapy PET scans, limiting the sensitivity of SUV for the evaluation of response in tumors that have low baseline uptake (85, 87). However, no correlation was found between pretherapy FDG uptake and percent change while on therapy for the kinetic measures. Thus kinetic analyses, unlike SUV, are independent of the baseline level of FDG uptake (87). Second, the improved accuracy of kinetic analyses compared to SUV is likely related to its ability to quantify multiple properties of the tumor biology (87). While clinically impractical due to the length of time needed for dynamic imaging, these example serve to illustrate that time domain data made add clinical value to FDG-PET, and may be included in some fashion in the current paradigm of delayed imaging after injection.

FES

Targeted therapy to the estrogen receptor has revolutionized care of patients with ER-positive tumors (43). Tumor biopsy specimens are now routinely assessed for ER expression via IHC. However, sampling error, invasiveness, and the inability to sample all sites of disease hamper the utility of tissue biopsy. ER expression is often heterogeneous across the disease burden, with expression in the primary breast tumor or a single metastasis not necessarily the same as other metastases (36, 88, 89). PET imaging of the estrogen receptor circumvents these challenges. FES, the most promising ER radiopharmaceutical to date, has recently gained FDA approval for PET imaging of breast cancer imaging (Figure 2).

Figure2. FES-PET/CT demonstrating a sternal metastasis and left axillary lymph node metastases.

62-year-old woman with chronic lymphocytic leukemia and left breast invasive ductal carcinoma. Sagittal PET, CT, and fused PET/CT images (A) demonstrate FES uptake within a biopsy-proven sternal breast cancer metastasis. Axial PET, CT, and fused PET/CT images (B) demonstrate FES uptake within left axillary lymph node breast cancer metastases; non-avid right axillary nodes reflect the patients known chronic lymphocytic leukemia.

Several studies with FES have investigated its ability to offer an early assessment of treatment response. An early study by Mortimer et al. demonstrated that an early decrease in FES uptake following initiation of endocrine therapy corresponds to successful ER blockade and is associated with treatment response (37). In this study of 40 ER-positive breast cancer patients treated with tamoxifen, SUV decreased on FES-PET performed after 7–10 days of tamoxifen treatment compared to the baseline FES-PET, indicative of binding of tamoxifen to the ER. Furthermore, the percentage decrease in SUV on tamoxifen was greater in patients who ultimately had a clinical response compared to patients who did not have disease response (37). A later retrospective study by Linden et al. utilized FES uptake to assess and compare early regional ER blockade in patients treated with tamoxifen versus fulvestrant. Thirty patients with metastatic breast cancer undergoing salvage endocrine therapy underwent a pre-therapy FES-PET and a second FES-PET performed early (median of 6 weeks) in the course of endocrine therapy. The results demonstrated that both tamoxifen and fulvestrant effectively decreased FES binding, as measured by percent SUV decrease. However, FES demonstrated notable differences in the degree of blockade between the two agents; while tamoxifen resulted in complete blockade of the ER (measured by an SUV of 1.5 or less), fulvestrant showed incomplete blockade (90). The relatively lower ER blockade by fulvestrant likely explains its suboptimal clinical performance compared to tamoxifen, which could have been the result of insufficient dosing at the time of the study (90). A phase III trial demonstrating increased efficacy at higher fulvestrant doses supports this hypothesis (91). Importantly, the study by Linden at al highlights the ability of FES-PET to evaluate the pharmacodynamics of endocrine therapy and guide treatment and dosing (90). Additionally, a recent multicenter Phase I clinical trial successfully utilized FES-PET/CT to evaluate ER occupancy during treatment of metastatic breast cancer patients with a novel ER-targeted therapy, and to help guide dosing for upcoming Phase II studies (92).

Research to date validates FES-PET as a quantitative measure of ER expression in breast cancer across the disease burden, and as a biomarker of response to endocrine therapy. With the recent FDA approval of FES for PET in breast cancer, as well as ongoing multicenter clinical trials to better define FES-PET as a marker of disease response, the utilization of FES-PET in the clinic and for research will likely continue to grow.

FLT

Measurements of cell proliferation, in conjunction with tumor size, grade, receptor status, and nodal status, offer valuable prognostic information, and may be used to guide therapy (93). A decrease in proliferation occurs early after initiation of successful cancer therapy with both cytotoxic and cytostatic therapies (94). Indeed, changes in cell proliferation after induction chemotherapy or endocrine therapy are predictive of treatment outcome (94–97).

While multiple laboratory assays measure cellular proliferation, including mitotic index and S-phase fraction, the most validated and widely used assay is total cellular proliferation with the nuclear protein Ki-67 via (93). Numerous studies have established that higher grade cancers have higher levels of Ki-67 (93, 98), and that Ki-67 index is correlated with other markers of proliferation (93). However, like all tissue-based assays, Ki-67 has several limitations (99) that could be circumvented with an imaging probe.

The most successful approach to imaging proliferation is to image the salvage pathway of thymidine incorporation into DNA. Early studies utilized ¹¹C-thmyidine-PET to successfully image tumor proliferation as well as treatment response changes (100). However, the short half-life of ¹¹C and its complex metabolism preclude its routine clinical use (101), and the fluorinated thymidine analog,¹⁸F-fluorothymidine (FLT), provides a more practical measure of cellular proliferation (102). Cell uptake of FLT is dependent on thymidine kinase-1, which is relatively active in proliferating cells, including malignant cells, but low or absent in quiescent cells (103, 104). Once intracellular, FLT is phosphorylated by TK-1 and trapped within proliferating cells; it is thus a marker of sites of proliferation (93). FLT uptake in breast cancer patients correlates with Ki-67 expression (103, 105).

Initial studies of FLT-PET in breast cancer demonstrated early changes in FLT uptake following initiation of chemotherapy (97, 106–108) (Figure 3). In a study of 12 breast cancer patients, Pio et al. found that the mean change in FLT uptake in both primary and metastatic tumors after the first course of chemotherapy significantly correlated with late changes in tumor marker levels. Early change in FLT uptake also predicted late changes in tumor sizes as measured by CT (106). Kenny et al. found that FLT-PET can detect proliferation changes at one week following chemotherapy, with a significant difference in uptake between responders and non-responders (97). Contractor et al. found that decreased FLT uptake two weeks after initiation of the first or second cycle of docetaxel predicts anatomic tumor response at mid-therapy (after three cycles) with high sensitivity (107). Another small study found that change in FLT uptake within two weeks of treatment is correlated with decreased circulating tumor cells (108).

Figure 3. Early Post-Treatment Response of Breast Cancer on ¹⁸F-FLT PET/CT.

¹⁸F-FLT-PET/CT axial pre-treatment images (A) demonstrate increased ¹⁸F-FLT uptake in the right breast cancer (arrow) and satellite lesion (arrow head). Early 7-day post-treatment imaging (B) demonstrates decreased radiotracer uptake in the dominant primary breast cancer (arrow) and resolution of radiotracer uptake in the satellite lesion. This illustrates the ability of proliferation imaging to measure response at very early time points after therapy.

Recently, a multicenter Phase II study was performed as part of ACRIN 6688 to assess whether early changes on FLT-PET/CT could predict pathologic complete response among primary breast cancer patients on neoadjuvant therapy. Patients with invasive ductal carcinoma underwent FLT-PET/CT scans prior to the initiation of neoadjuvant therapy, after the first cycle of therapy (first post-therapy), and after therapy completion. FLT uptake on the post-treatment PET correlated with Ki-67 on the surgical specimens. There was a marginal difference in the percentage change in SUV_max from the pre-therapy PET to first post-therapy PET between patients with pathologic complete response versus those without pathologic response (109). While this study does suggest some efficacy of FLT-PET as an early indicator of therapeutic response, the neoadjuvant regimen was not specified by the study protocol, confounding the analysis and demonstrating the need for additional clinical trials.

PET Biomarkers to Predict Survival

Late or end-of-treatment response has been utilized as a biologic surrogate for therapeutic outcomes such as progression-free survival or overall survival. Whereas survival data may take months to year to acquire, depending on the disease, late or end-of-treatment response analysis may be acquired soon after therapy. In clinical trials, such biomarkers could decrease trial duration and costs, facilitating a more rapid clinical translation (4, 110). In clinical practice, end-of-treatment response has been used to guide additional therapy, most notably in lymphoma.

Lymphoma

The role of FDG-PET or PET/CT as an integrated biomarker during and after therapy has been extensively studied in the treatment of lymphoma. FDG-PET/CT has gained widespread clinical acceptance for staging and end-of-treatment response; the Deauville score and Lugano criteria have been widely embraced for these assessments (21, 111). The Deauville criteria visual assessment categorizes lesion uptake as follows: 1) No uptake; 2) Uptake less than or equal to that of the mediastinal blood pool; 3) Uptake more than the mediastinum but less than or equal to that of liver, 4) Uptake moderately more than the liver, and 5) Uptake markedly more than liver parenchyma or new sites of disease (111). This 5-point scale is applied as the input for the Lugano Criteria response assessment (21)

Using the Deauville criteria, a complete metabolic response is defined as a Deauville score of 1–3, whereas a Deauville score of 4 or 5 signifies either a partial response or progression of disease. Based on these criteria, end of treatment FDG-PET/CT demonstrates high positive predictive value and negative predictive value in both Hodgkin Lymphoma (>90% and 95–100%, respectively) and Non-Hodgkin Lymphoma (50–100% and 80–100%, respectively) (21).

The prognostic value of FDG-PET has been leveraged as an early response indicator after initiating chemotherapy in both Hodgkin and Non-Hodgkin lymphoma. In a response-adapted treatment paradigm, an interim FDG-PET is performed after two cycles of chemotherapy. A partial response (PR) or progression of disease (PD) on interim PET imaging may result in intensification of therapy and possible consolidative radiation therapy. In contrast, a complete metabolic response (Figure 4). would result in continuation of the current treatment plan. In patients with Hodgkin lymphoma, interim FDG-PET testing predicts PFS, treatment failure, and overall survival better than the initial stage of disease and staging prognostic risk scores (112, 113). The EORTC/LYSA/FIL H10 randomized trial published in 2017 investigated early response assessment with FDG-PET to improve selection of previously untreated stage I and II Hodgkin lymphoma patients for reduced versus more intensive chemotherapeutic therapy. FDG-PET was performed after two cycles of doxorubicin, bleomycin, vinblastine, and dacarbazine (ABVD). In the standard arm, all patients continued ABVD followed by involved-node radiotherapy (INRT), regardless of PET results. In the experimental arm, PET-positive patients had escalated drug therapy plus INRT; PET-negative patients received ABVD only. Among PET-positive patients, five-year progression-free survival was significantly greater for the escalated therapy patients compared to the patients who remained on standard therapy (114). These results demonstrate the potential value of FDG-PET in improving the balance between treatment efficacy and toxicity (114). Similarly, the phase 3 response-adapted therapy for advanced Hodgkin lymphoma (RATHL) trial demonstrated de-escalating therapy for patients with CMR on interim PET-CT after 2 cycles of ABVD chemotherapy to AVD (omitting bleomycin), reduced pulmonary toxicity without significantly lowering treatment efficacy (115).

Figure 4. ¹⁸F-FDG PET/CT for lymphoma demonstrating a complete metabolic response on interim and post-therapy studies.

31-year-old man with Burkitt lymphoma with FDG-avid lymphadenopathy above and below the diaphragm on pre-treatment FDG-PET/CT. (A) Whole body MIP, with axial PET and CT images demonstrate large juxtacrural/posterior mediastinal lymphadenopathy (arrows), SUV_max 27.3. Interim-treatment imaging (B) after 4 cycles of R-EPOCH demonstrates a complete metabolic response with resolution of the large hypermetabolic mediastinal mass, SUV_max below blood pool (Deauville 1), Lugano classification: Complete metabolic response. Note is made of FDG-avidity throughout the bone marrow and spleen related to bone marrow stimulation. Post-treatment imaging (C) demonstrates a continued complete metabolic response to therapy.

Quantitative vs. Qualitative Response Assessment in Lymphoma

As described above, the Lugano Criteria are based on the Deauville criteria, a 5-point visual qualitative scale which compares lesional FDG-uptake to that of background liver parenchyma or blood pool. Given the potential for interobserver variability in interpretation of this visual assessment, a more quantitative measure of tracer uptake, such as ΔSUV_max, may offer a more reproducible assessment (116–118). Itti et al. studied 114 patients with newly diagnosed diffuse large B-cell lymphoma, and evaluated interobserver variability and progression free survival based on Deauville visual assessments versus ΔSUV_max (using a ΔSUV_max threshold of 66%) after two cycles of chemotherapy compared to baseline. While both the Deauville criteria and ΔSUV_max were predictive of outcome, ΔSUV_max demonstrated better reproducibility, with very high interobserver agreement, as well as higher performance for predicting outcome (116).

The CALGB 50303 Clinical Trial in diffuse large B-cell lymphoma performed a subset analysis to assess the 5-point visual assessment compared to a quantitative cut-off of ΔSUV_max of 66% as biomarkers of response to therapy using interim and end-of-treatment FDG/PET-CT. Although estimates of PFS and overall survival (OS) were higher in patients with a negative interim PET using the visual assessment, neither difference reached statistical significance. Similarly, there was no significant association between visual end-of-treatment PET findings and patient outcomes. However, the ΔSUV_max 66% cut-off demonstrated a significant association with OS and a positive trend for PFS (119). Given these findings, the use of a more quantitative response assessment may be of value in lymphoma clinical trials; however, more research is needed to assess clinical efficacy.

Non-Small Cell Lung Cancer:

FDG-PET has a well-established role in evaluating early response to therapy in non-small cell lung cancer, with a sensitivity and specificity of 100% and 92%, respectively (120). However, there is also an important and evolving role for FDG-PET to predict survival or tumor recurrence. In an early heterogeneous study of 113 NSCLC patients treated with chemotherapy, surgery, or radiation, Patz et al. found a statistically significant increased survival among patients with a negative post-therapy PET. 85% of PET-negative patients were alive at the completion of the study, with a median survival of 34.2 months, compared to a median survival of 12 months in the PET-positive patients (121).

Higher post-treatment tumor FDG avidity is associated with worse survival outcomes, risk of local recurrence and distant progression. In the ACRIN 6668/RTOG clinical trial of stage III NSCLC patients, Machtay et al. sought to determine if SUV on post-treatment FDG-PET correlated with survival. While the pre-specified binary SUV value of 3.5 was not associated with survival, the results demonstrated worse survival outcomes with higher SUV_peak when studied as a continuous variable, and an increase of 1.0 point in post-treatment SUV_peak translated to a 9% increase in the risk of death (122).

More recent studies indicate that an interval change in FDG avidity on interim PET during the course of chemotherapy may predict treatment response and survival. For example, Kim et al. retrospectively studied 42 stage IIIA-N2 NSCLC patients who underwent FDG-PET/CT after receiving two to four cycles of induction chemotherapy with or without radiation. The median follow-up period was 28 months. Subjects who demonstrated a complete response on PET/CT following had a significantly longer relapse-free survival time compared to those categorized as incomplete responders. (28 versus 9 months). However, this particular study found no significant difference in overall survival time between complete and incomplete PET responders. (123). In a prospective multicenter clinical trial of interim treatment PET imaging by Vera et al., higher SUV_max on interim PET predicted death and tumor progression at one year (124). More recently, the utilization of interim PET imaging for NSCLC has been evaluated by Cremonesi et al in a large systematic review. Although there was appreciable heterogeneity in the 21 studies reviewed, early identification of metabolic tumor response was found to be a promising predictor of response and prognosis in NSCLC patients (125).

Conclusion

The development of novel targeted oncologic therapies has transformed clinical oncology. Molecular imaging offers great value as a noninvasive assessment of treatment response across the entire disease burden, and numerous PET imaging biomarkers have been investigated for this purpose. This review highlights examples of quantitative PET biomarkers for both early response assessment and for predicting long-term response / survival.

FDG remains the most utilized imaging biomarker for response assessment. While FDG-PET has an established role in the clinical staging of numerous malignancies, its role in treatment response continues to evolve. Clinical studies demonstrate the value of FDG in breast cancer for early response assessment to neoadjuvant therapy (65, 66), including HER2-targeted therapy (80, 83). The utilization of FDG-PET has been well validated as a predictive marker for survival in lymphoma, and while qualitative response assessment has traditionally been utilized, more quantitative response measures may more accurately predict outcomes (116, 119). In addition, there is an evolving role for utilization of FDG-PET in predicting survival among non-small cell lung cancer patients (123, 124).

Although FDG dominates clinical PET imaging, research to date supports promising clinical applications of novel PET biomarkers. We review several novel biomarkers of response assessment that are primed for utilization in clinical oncology or are ready to undergo further investigation in advanced clinical trials, including FLT and FES. Recent studies suggest value of FLT as an early predictor of treatment response in a number of malignancies, including breast cancer (107, 109), and demonstrate the need for further study in clinical trials. Targeted therapy of the estrogen receptor has markedly improved prognosis of ER-positive breast cancer, and with the recent FDA approval of FES for breast cancer, FES-PET is poised to enter clinical oncology as both a predictor and early response assessment of endocrine therapy. These novel imaging biomarkers have the potential to profoundly impact response assessment in precision medicine.