Using positron emission tomography to assess tumor proliferation in non-Hodgkin lymphoma

Positron emission tomography (PET) allows quantitative three-dimensional imaging of various physiologic processes following administration of a positron-emitting radiopharmaceutical. In combination with computed tomography (PET/CT), both function and anatomy can be characterized contemporaneously providing a powerful non-invasive tool assisting diagnosis, staging, and restaging of lymphoma. The radiotracer used most commonly in clinical practice, fluorodeoxyglucose (FDG), is a glucose analog allowing whole body and regional quantification of glucose metabolism. While this is not a direct measure of tumor proliferation, there will clearly be a relationship between the two, as tumor cells typically rely on glycolysis to fuel proliferation [1] and many of the key signaling pathways involved in malignant transformation are intimately linked to the regulation of glycolytic metabolism [2]. While visual analysis of images is usually sufficient, uptake can also be assessed semiquantitatively using the standardized uptake value (SUV). The SUV is calculated by knowing the dose of tracer injected, time from injection to imaging, and normalizing for patient weight. More accurate quantitative analysis of the intact tracer being delivered to the tumor is also possible but this requires more labor-intensive protocols and arterial sampling which is not practical for routine clinical practice.

Cell proliferation rate is an important factor for grading human cancers. Ki-67 is a protein which undergoes phosphorylation and dephosphorylation during the active part of the cell cycle changing its localization pattern within the nucleus [3]. Using antibodies raised against antigens of this protein which are expressed in all phases of the cell cycle except G0, cell proliferation can be quantified immunohistochemically by dividing the number of cells that stain positively for Ki-67 with the total number of cells in the sample [4]. Since its first description in 1983, there has been a rapid rise in interest with multiple studies investigating the role of Ki-67 proliferation index (PI) and clinical outcomes in a variety of malignancies [3].

In this issue of the journal, Watanabe et al. report a significant correlation (r = 0.69) between Ki-67 PI and the SUV at the biopsy site in a study of 36 untreated patients with non-Hodgkin lymphoma (NHL) who underwent FDG PET imaging [5]. The population studied was heterogeneous including patients with indolent and aggressive lymphoma. The correlation was maintained at both nodal and extra-nodal sites, and in a sub-analysis of the 16 patients with diffuse large B cell lymphoma (DLBCL). These results are concordant with previous smaller studies correlating Ki-67 with FDG uptake in lymphoma [6,7].

In several other studies that did not involve direct correlation with proliferative indices such as Ki-67, lower FDG uptake has been demonstrated in indolent compared to aggressive lymphoma but there is considerable overlap in the range of SUV values for each sub-group [7–9]. This limits the clinical utility of this measure except in patients with either very high or low SUVs. In one of the largest studies, Schoder et al. [8] found a sensitivity and specificity of 71% and 81% using a SUV cut-off of 10. Broyde et al. reported similar results using Ki-67 PI in a study of 260 patients, with a Ki-67 PI of 45% providing a sensitivity and specificity of 82% and 77% to differentiate indolent from aggressive lymphoma [9]. Both studies are limited by post hoc statistical analysis to identify an optimal cut-off level which introduces significant bias and limits the generalizability of the results. Indeed, other studies find different cut-off values in their post hoc analyses [10,11]. Furthermore, the disparate pathological features of these disease processes make this differentiation of limited clinical relevance except in patients suspected of transformed lymphoma.

Accordingly it is probable within any given histopathological subtype that parameters assessing proliferation may have prognostic value. To assess the potential clinical utility of PET as a non-invasive imaging technique, one must first demonstrate that other more established measures, such as Ki-67, have independent prognostic value compared to known clinical, laboratory and staging parameters, particularly those included in prognostic scores such as the International Prognostic Index (IPI). In a retrospective study of 171 patients with indolent and aggressive NHL, Gerdes et al. [12] demonstrated a better survival in patients with lower Ki-67; paradoxically, in the subset of patients with DLBCL, there was longer survival in patients with higher Ki-67. Hall et al. [13] describe similar results with a poorer prognosis for patients with low-grade lymphoma and high Ki-67 (>5%) but a trend for patients with high-grade lymphoma and very high Ki-67 (>80%) to have better survival. More recent studies, however, have not confirmed these findings and have found poorer prognosis in DLBCL with high Ki-67 PI [14–17]. In DLBCL, Grogan et al. found Ki-67 to be an independent predictor of survival [16], whereas other studies demonstrated inferiority of Ki-67 compared with the IPI [9,18]. Interestingly, Broyde et al. [9] reported that in a subgroup of patients with bulky disease (>10 cm), Ki-67 provided marked survival stratification suggesting a good prognosis for tumors that reach bulkiness with a low proliferative index. Thus, the prognostic relevance of tumor growth fraction appears to differ in NHL subtypes and may result from factors including the type of treatment delivered. For example, it may be that proliferation is simultaneously a predictive and a prognostic biomarker wherein a high Ki-67 may predict for high responsiveness to cytotoxic therapy but in the presence of refractory disease be associated with an inferior prognosis.

While of significant value in providing further insights into the linkage between FDG uptake in lymphoma and biological factors such as proliferation, definitive conclusions from the data of Watanabe et al. presented in this issue [5] are somewhat limited by the heterogeneous population evaluated. While Ki-67 and SUV may play a role in specific subpopulations of NHL, further research is needed before both are incorporated routinely into management planning. In mantle cell lymphoma (MCL), Ki-67 expression has evolved as independent prognostic factor in multivariate analysis [19]. This is due to key genomic changes that are intimately linked to cell cycling [20]. Further to this, Karam et al. demonstrated survival stratification utilizing FDG SUVmax [21]. In MCL, the 5-year overall survival was 87.7% with SUV <5 compared to 34% with SUV >5. However, the relationship between SUV and survival is not linear, with no additional stratification in patients with SUV >5; similar results are found with Ki-67. This suggests the tumor proliferation indices may be identifying two different MCL-like phenotypes with different biological behaviors.

One of the strengths of PET imaging is the facilitation of rapid whole body imaging. Biopsy is limited to specific sites, which is usually guided by ease of access and clinical feasibility. This may, in part, explain some of the variability of Ki-67 prognostic significance between studies. PET also allows simple identification of sites with different rates of tumor proliferation that cannot be differentiated structurally on anatomic imaging with CT or MRI. Thus, in patients with follicular lymphoma, FDG PET has been demonstrated to be useful in the identification of high-grade transformation [22,23].

The relationship between SUV and tumor proliferation, particularly with a tracer like FDG that measures glucose metabolism rather than cellular proliferation is complex. Glucose transporter expression and tissue perfusion will also influence FDG uptake. There is interest in other tracers such as fluorothymidine (FLT) that more specifically measures proliferation [24] but there is limited evidence to suggest a specific role in lymphoma over FDG. Another area of interest is whether PET can be used to differentiate Grade I/II from Grade III follicular lymphoma. Tang et al. [25] examined the correlation between Ki-67 and FDG SUV in such patients and found a moderate correlation (r = 0.4); Ki-67 clearly separated the various grades of follicular lymphoma, while there was significant overlap of FDG SUV. Higher correlation (r = 0.84) has been reported with FLT-PET [26] suggesting a role for more specific radiotracers in this patient population.

Regardless of the radioactive tracer used, many factors have the potential to influence SUV measurements including uptake phase duration, patient weight/body habitus, partial volume effects, and scanner calibration factors. Nevertheless, a very high correlation between semi-quantitative SUV and quantitative regional metabolic rate (r = 0.95) in patients with lymphoma has been demonstrated [7]. SUV is frequently considered as an independent variable from stage but the two are closely linked. If the disease process has higher FDG avidity, this makes identification of additional sites of disease, particularly those associated with smaller lesions, more likely to upstage patients. Thus, apparent PET stage and SUV may be partially co-dependent variables. Measures of metabolic burden which incorporate both PET uptake and the volume of PET abnormality may prove more useful for risk stratification and response assessment [27].

The prognostic and predictive power of early PET restaging is promising with good evidence to suggest that it provides more powerful risk stratification than conventional means [28]. The utility of PET in this regard is likely to dwarf the power of proliferative markers such as Ki-67 or surrogates such as SUV prior to treatment. Several prospective trials are underway to examine if early change in treatment after interim PET improves outcome; analysis of baseline SUV in these studies would be of interest. Further research is still needed within specific histological subgroups to elucidate if the addition of proliferative indices or SUV to existing clinicopathological parameters can be used to alter patient management.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the article.