PET Scans for Staging and Restaging in Diffuse Large B-Cell and Follicular Lymphomas

PET-CT is regarded as mandatory for staging patients with DLBCL because (i) CT alone understages patients by missing extranodal disease and (ii) a staging scan is required to identify initial disease sites when reporting end-of-treatment scans [2••]. Changes occur during treatment that need to be distinguished from lymphomatous involvement, which is evident on the staging scan. Changes may be FDG-avid but represent inflammation, infection and/or stimulation of normal marrow rather than lymphoma [4]. Recent publications have examined the effect of the more accurate staging with PET-CT on the prognostic ability of IPI [5, 6]. There is also growing interest in the significance of disease features on baseline PET, particularly in relation to the extent of extranodal involvement and total disease burden, both of which show promising prognostic value [7, 8, 9•].

There are three prognostic indices available to risk-stratify patients: the IPI [10], the revised IPI (R-IPI) [11] introduced to account for improved prognosis after rituximab and the most recent National Comprehensive Cancer Network IPI (NCCN-IPI) [12] with further refinement based on grading some of the prognostic factors.

All use clinical factors - age, performance status (PS) and lactate dehydrogenase (LDH) level and imaging defined factors - Ann Arbor stage and extranodal involvement. These indexes were formulated using CT imaging. PET, now recommended for staging, upstages a significant proportion of patients, specifically detecting more extranodal disease, in particular with respect to bone marrow involvement [2••].

Recent studies examined the effect of PET upstaging on the performance of prognostic indices. El-Galaly and colleagues [5] evaluated stage, extranodal involvement and prognostic indices in 443 newly diagnosed patients with DLBCL staged with PET-CT, treated with R-CHOP or R-CHOP-‘like’ chemotherapy. PET-based IPI, R-IPI and NCCN-IPI were all predictive of outcome. Patients with very good (R-IPI) or low risk (NCCN-IPI) had excellent outcomes: 3y-PFS was 100 % (95 % confidence interval 88, 100) [n = 50] for very good risk and 3y-PFS and OS were 100 % (88, 100) [n = 54] for low-risk patients. However, the IPI and R-IPI failed to identify the group who are unlikely to be cured with RCHOP, with 3y-PFS >50 % for the poor risk group, but the NCCN-IPI was better at identifying this group (3y-PFS 33 % and OS 40 %).

Another emerging theme in the study by El-Galaly et al. is the strong prognostic value of the extent of extranodal involvement. Two thirds of patients had extranodal disease on PET-CT, and significantly more treatment failures occurred with an increasing number of involved extranodal sites. By example, 3y-PFS was 25 % (7, 53) in patients with four or more extranodal sites compared to 79 % (71, 87) in patients with nodal disease only. Corresponding 3y-OS was 36 % (16, 56) vs 82 % (66, 98), respectively. Patients with extranodal involvement were older and had worse PS, more B symptoms and a higher prevalence of raised LDH. The extranodal sites associated with worse PFS and OS were bone marrow, pleura and gynaecological organs [13].

Bulk is regarded as an adverse prognostic indicator in DLBCL, at least in patients with otherwise good prognosis [14]. It is commonly measured as the largest dimension of the largest mass. Metabolic tumour volume (MTV) is a more sophisticated, albeit time-consuming measurement of total disease burden based on volumetric measurement and metabolic activity. MTV is calculated by adding volumes of tumour, selected using a standardised uptake value (SUV) typically ≥2.5 or a percentage of the maximum SUV in areas of tumour [15, 16]. Recent publications report that MTV predicts prognosis [17, 18] better than bulk [9•]. An inherent problem, as with all measurements with a continuous distribution, is where to draw the cutoff between ‘high’ and ‘low’ MTV. A range of thresholds has been reported from 220 to 550 cm³, derived from receiver operating characteristic curves [9•, 17, 18]. The cutoff is influenced by the characteristics of the study population, with patients with earlier stage nodal disease only [17] having lower cutoffs than patients with high-risk more advanced disease [18]. Patients with low MTV tend to have better outcomes, with 3y-PFS 77–92 %, compared with patients with high MTV who have worse outcomes, 3y-PFS 48–56 % [9•, 17, 18], but again not sufficiently poor to consider treatment escalation. Mikhaeel et al. [9•] combined baseline MTV with early response assessment using Deauville scores at 2 cycles in 147 consecutive unselected patients treated for DLBCL with R-CHOP at a single institution. Patients could be separated into three distinct prognostic groups—good risk (low MTV regardless of PET-2 response), intermediate risk (high MTV, with CMR at 2 cycles) and poor risk (high MTV, no CMR at 2 cycles, Deauville scores 4, 5). 5y-PFS was >90 %, 58 % and 30 %, respectively (median FU = 3.8 years). The poor risk group contained 31 % of patients who experienced 58 % of the study events. These results demonstrate the previously unexplored interaction between pre-treatment prognosis and early response which enables better prognostication at an individual patient level. Validation of these results in larger prospective studies is warranted, as such an approach could improve on the prognostic power of interim PET (Fig. 1).

Post-treatment remission assessment is important, as cure is the goal of treatment in DLBCL. PET-CT is the standard method in international guidelines [3••] and is widely used in routine practice. Nevertheless, several points are worth addressing:

Several studies have shown that residual activity in sites of previous disease on PET-CT, normally defined as Deauville score 4–5, is strongly predictive of residual disease. In R-CHOP-treated patients, progression-free survival (PFS) ranges from 24 to 35 % for positive end-of-treatment PET [28••, 29••]. This may be considered high enough to consider further treatment without biopsy confirmation. However, we would recommend that biopsy should be considered whenever possible to exclude the less common false-positive cases usually histologically reported as xanthomatous granulomatosis [2••]. Ultimately, the decision depends on the balance between the type of biopsy required (e.g. imaging-guided versus open surgery) and the intensity of treatment being considered (e.g. consolidation radiotherapy versus autologous stem cell transplant (ASCT)). Consideration of pre-treatment prognosis and response on interim imaging may also help the decision.

The negative predictive value of end-of-treatment PET is high, and most patients enjoy a long-term remission or cure. However, recent evidence also suggests that high-risk patients who achieve CMR still have a considerable risk of relapse. In a population-based study of 223 consecutive patients treated with R-CHOP-like immunochemotherapy [34], NCCN-IPI and R-IPI remained predictive of relapse irrespective of CT- or PET-defined remission status. Patients with NCCN-IPI 6–8 had a dismal outcome and higher risk of CNS relapse, whether or not they achieved CMR. The median age in this group, however, was 75 years. None responded to salvage treatment. In this elderly high-risk group, the authors suggested that alternative therapeutic approaches, e.g. novel agents, may be the only prospect for cure.

Early response assessment using interim PET is a more controversial area in DLBCL due to variable results reported in studies [30‐37] and the debate on what action, if any, should be taken. We first examine the prognostic significance of interim PET negative and positive results and subsequently what action might be appropriate.

Studies demonstrate that 60–80 % of patients achieve CMR after 1–4 cycles of systemic therapy and tend to have an excellent PFS, usually in excess of 75–80 % [28••, 29••, 38]. More recent evidence also shows that achievement of early CMR predicts final CMR with very low risk of conversion to PET positivity (i.e. progression) after treatment [28••, 29••, 39].

Mamot el al. [28••] recently reported prospective results in 138 patients with DLBCL treated with R-CHOP-14 with response assessed using PET-CT. This important study employed standardised methods for PET with a common imaging protocol and strict observance of timing of scans in relation to chemotherapy. Treatment was not adapted according to interim PET, although patients with progression went off-study, suggesting that clinicians considered that it was unacceptable to continue with standard treatment in the presence of clear evidence of progression on interim PET, in line with recent international recommendations [2••]. There was a higher proportion of events than anticipated, likely because radiotherapy was included as an event, but the authors found similar results when using PFS as the end point. Analysis during the study used older International Harmonization Project reporting criteria [40], but a post hoc expert central review was performed using the five-point scale or Deauville criteria (DC) with scores 1–3 regarded as CMR. 2y-event-free survival (EFS) according to interim PET at 2 cycles was 75.9 % (63.7, 84.5) vs 41.4 % (28.7, 53.6) for patients with CMR and Deauville scores 4–5, respectively, p < 0.001. According to end-of-treatment PET, 2y-EFS was 71.5 % (61.2, 79.5) for CMR and 24.0 % (9.8, 41.7) for Deauville scores 4–5. Of note is that all patients with CMR on interim PET (60 %) remained in CMR at end of treatment and the EFS for a negative PET was very similar at interim and end of treatment.

Similar findings were reported in an earlier study [41] and a more recent study by Huntington et al. [39] where 79 % of patients had CMR at interim, all of whom had CMR at end of treatment. This confirms that interim PET shows CMR early in a significant proportion of patients and supports the premise that end-of-treatment scans are not necessary in patients who have CMR at interim.

Another interesting study was reported by Carr et al. [29••] combining interim (2–3 cycles) and end-of-treatment PET assessment from an international cohort of patients with DLBCL from disparate healthcare systems, sponsored by the International Atomic Energy Agency. They set a priori criteria which were similar, but not identical to DC. There was no significant difference in outcomes by country according to PET, suggesting that the technology is equally well applied in low- and high-income countries. In 327 patients, interim PET was negative in 64 % and positive in 36 %. The authors stratified 312 patients into four groups using the results of both interim PET and end-of-treatment PET. The best outcome was in the largest group (62 %) with interim and end-of-treatment CMR; 2y-EFS was 97 % (92, 98). Over half the patients who did not achieve CMR at interim had CMR at end of treatment (19 %), and these ‘slow’ responders, in whom bulky disease was more common, also had good outcomes, with 2y-EFS of 86 % (73, 93), although the hazard ratio (HR) for relapse compared to the previous group was 2.56 (1.08–6.11). These results imply that treatment escalation according to a positive interim PET could significantly over-treat many patients. Patients who had both positive interim PET and end-of-treatment PET (16 %) had 2y-EFS of 35 % (22, 48). Finally, 13 (4 %) patients progressed on treatment, with negative interim but positive end-of-treatment scans, 11 with biopsy-proven disease.

On the basis of these studies, it is clear that an early CMR on interim PET identifies a group with excellent prognosis, with the advantage of reassuring patients of the expected good outcome early. On the other hand, approximately half of patients with a positive interim PET will enter remission and change or escalation of treatment will over-treat an unacceptable proportion.

But more importantly is the fact that, to date, there has been no alternative treatment that proved to be superior to R-CHOP. In fact, studies which examined change or escalation on the basis of interim PET have not shown any benefit. Five studies have been reported either fully [42‐44] or in abstract form [45, 46] using different PET criteria, timing and escalation strategies, summarised in Table 1. These studies demonstrated two findings: (1) the prognosis of interim PET positive is significantly worse, and (2) there was no improvement in this prognosis with changing therapy. The hope is that future research may produce better regimens which improve the outcome of patients not responding well to RCHOP. Until then, there is no justification for an early change in therapy unless there is definite evidence of progression.

So where does this leave the role of interim PET in DLBCL? We assert that it is an excellent negative test, showing remission early and enabling reassurance of patients whilst still having treatment. In addition, it may show lack of any response or early progression in a small proportion of patients. Is it essential? The answer is no, but if interim imaging is performed, we would recommend PET-CT in preference to CT for the above reasons and recommend that if the result shows CMR then an end-of-treatment-PET is not required, which saves resources and inconvenience. On the other hand, if the result is positive, there is no indication to change treatment, but close monitoring of these patients during treatment may be warranted, as a proportion of them may progress. In our practice, if the initial disease was poor risk (e.g. high IPI) and PET-2 is positive, we prefer to monitor the patient with repeat PET after 4 cycles of treatment [9•].

PET-CT detects more nodal and extranodal sites than CT in patients with FL undergoing induction chemotherapy [52•]. The impact on staging is higher in patients with apparently limited stage on CT in whom PET is performed (i) to determine if local RT treatment is appropriate (ii) to plan RT fields. In a recent phase III prospective study which randomised patients to one of three R-chemotherapy treatments, PET-CT altered the Follicular Lymphoma International Prognostic Index (FLIPI) score in 35/142 patients, increasing the score in 18 % of patients overall and in 62 % of patients (15/24) with limited disease on CT [52•], confirming earlier reports [51, 53, 54]. Although outcomes from patients selected for RT using PET-CT have not been published, this would appear to be a sensible indication for PET-CT in patients with FL. Bone marrow biopsy detected bone marrow involvement in 46/108 (43 %) of patients without marrow abnormalities on PET [52•], indicating that BMB is required for diagnostic workup. The ability of PET to select sites for biopsy in cases with suspected transformation has also been previously described [58, 59] and is recommended in current guidelines [2••].

PET-CT was initially supported for response assessment in the research setting [60] but has now become the standard imaging modality [2••, 3••]. This was based on three multicentre studies where PET-CT was used to assess response in patients with high tumour burden symptomatic FL or advanced disease, which included 122 [61], 112 [62] and 205 [63] patients, respectively. All reported end-of-treatment PET-CT to be predictive of PFS, independent of the FLIPI and superior to CT-based response. Interim PET (cycle 4) was predictive of PFS, but not as strongly predictive as the end-of-treatment scan in the prospective PET-Folliculaire study [62].

A pooled analysis of PET-CT response in the three studies, using central scan review and DC, was published in 2014 [64••]. The analysis included 246 patients with PET-CT scans available for review. Median FU was 54.8 m. Seventy-three percent of patients were treated with R-CHOP, 15 % with R-CVP and 12 % with R-FM. Eighty-three percent of patients had a negative scan (Deauville scores 1–3). The study revealed three important findings. Firstly, there was a significant number of patients who had their response re-classified with PET compared to CT-based International Working Group (IWC) criteria. Of 128 patients with CR/CRu, 17 (13 %) had Deauville scores 4–5, and of 72 with PR/SD/PD, 50 (69 %) were re-classified as CMR (Deauville scores 1–3). Secondly, the PET-based response was more predictive of PFS and OS than IWC, in the whole group, in the RCHOP-treated patients and in the responding patients according to IWC. In the whole group, 4y-PFS was 63.4 % (55.9, 70.0) for patients with CMR, compared with 23.2 % (11.1, 37.9) for patients without CMR [HR 3.9 (2.5,5.9) p < 0.0001]. The difference in median PFS was very large: 74 and 16.9 months for patients with scans showing CMR and no CMR, respectively. Finally, PET-based response was an independent and stronger predictor of PFS than FLIPI and IWC response.

These data establish PET-CT as the imaging modality of choice for remission assessment in FL, but more importantly, it shows that PET response using DC can identify the small group of patients who are likely to have an early progression (median PFS 16.9 m). As discussed in the “Introduction” section, it has been shown that patients who progress within 2 years have worse OS (5y-OS 50 % compared to 90 % for longer remission) [56] who may be candidates for close monitoring and testing different treatment approaches if they relapse early. Patients with CMR can be confidently reassured about the prospect of long PFS. The discriminatory ability of PET response to first-line therapy in FL lends itself to being used in studies testing response-adapted therapy, further refining the management of this disease with varied natural history.

Lastly, salvage treatment and ASCT is an option for patients with refractory or relapsed disease who are sufficiently fit. An initial report from the Lymphoma Studies Association in 59 patients with relapsed/refractory disease after first-line R-CHOP suggested that PET may be able to predict response following high-dose chemotherapy prior to ASCT [65], as previously reported in HL and DLBCL.