Midtreatment ¹⁸F-FDG PET/CT Scan for Early Response Assessment of SMILE Therapy in Natural Killer/T-Cell Lymphoma: A Prospective Study from a Single Center

DISCUSSION

We found an early midtreatment ¹⁸F-FDG PET scan after 2–3 cycles of chemotherapy to be useful in the prediction of therapy response in ENKTL patients. With visual assessment, the DS was found to be the only significant independent predictor of OS and PFS and thus more powerful than IPI, conventionally the most important prognostic scoring system (16). Unlike the 2-point positive and negative score of the International Harmonisation Project (IHP) criteria (17), the DS offers a 5-point scale that provides more continuous criteria to define tumor response, which is pertinent especially for the midtreatment scans, because the DS better reflects the therapy changes in tumors that are ongoing and in continuum. It has been recommended that interim PET scans should be assessed using DS in clinical practice and in trials (18). Treatment response assessment using DS was found to be accurate and reproducible, with good interobserver agreement, ranging from 0.69 to 0.84 measured with the Cohen κ (19). In our cohort, DS using a cutoff of 3 was able to predict OS and PFS with an accuracy of 83%, and this improved at the end-treatment scan to 94%. Both NPV and PPV at midtreatment scan for OS and PFS were moderately good at 83%. The 2-y estimate for OS of the entire population, patients with a DS of 1–3 and 4–5 at midtreatment scan, was 65%, 81%, and 17%, respectively, and the 2-y estimate for PFS was 52%, 62%, and 17%, respectively. At the end-treatment scan, the DS cutoff of 2 or 3 would have yielded the same results because there were no patients with a DS of 3. Using a DS cutoff of 2 to define complete metabolic response in the negative scans would essentially be equivalent to using the IHP criteria and may be more appropriate at end-treatment to increase the NPV of PET. In a study from Southern China of mature aggressive T-cell and ENKTL, the subset of ENKTL patients with interim PET scans (n = 34) or end-treatment PET scans (n = 17) was evaluated regardless of whether pretreatment baseline scans were obtained. Li et al. found interim and end-treatment PET scans assessed using the visual criteria of the IHP to be predictive of PFS (P = 0.013) but not of OS (12).

We did not find ΔSUV_max to be predictive of OS or PFS, which might be due to limitations with the accuracy of quantification of residual small lesions after treatment, lesions becoming no longer visible, and the effect of partial-volume averaging. Moreover, ΔSUV_max sometimes does not accurately reflect the impact of new disease because the new lesion may be less ¹⁸F-FDG–avid than the index lesion, and this was the case in 7 scans with new lesions (DS, 5) in our study. Nevertheless, ΔSUV_max was found to have a higher predictive value than visual analysis in several studies evaluating therapy response in DLBCL patients (20–22). Lin et al. found ΔSUV_max using a cutoff of 66% to be more predictive than visual analysis in predicting outcome after 2 cycles of therapy in DLBCL patients, especially in improving PPV, but a follow-up study by Itti et al. did not show this advantage when patients were assessed after 4 cycles of therapy (20,21). Subsequently, these results were validated by Casasnovas et al. who showed ΔSUV_max to be feasible in a cohort of high-risk DLBCL patients after both 2 cycles and 4 cycles of treatment and better than visual analysis using both IHP and Deauville criteria (22). Finally, in a recent study by Pregno et al., interim PET using ΔSUV_max was not predictive of outcome in a cohort of DLBCL patients treated homogeneously with R-CHOP (rituximab-cyclophosphamide, vincristine, doxorubicin, prednisolone) (23). Thus, this semiquantitative parameter requires further validation in prospective studies (20–23).

We found absolute SUV_max of the hottest lesion in the mid- and end-treatment scans to be predictive of OS by univariate analysis, using a cutoff value of 3 for SUV_max at midtreatment and 2 at end-treatment scans. However, although semiquantitative analysis using SUV_max is a more objective assessment, it did not provide any advantage over visual analysis, because it provided a lower accuracy and did not impact independently on OS.

In our cohort, pretreatment SUV_max of the hottest lesion was not predictive of OS or PFS. Suh et al. in a study of ENKTL limited to the head and neck region found SUV_max of the primary tumor (nasal cavity, oral cavity, and nasopharynx) to be predictive of disease-specific survival and OS. The cohort differed in the location of the primary tumor, and patients were mostly of stage I and II disease (10). Hence, our results are not directly comparable.

Our study has several advantages over the few series of PET for therapy response assessment in ENKTL reported so far. First, patients were treated with currently the most active regimen (SMILE) for this lymphoma (13). Second, patients were managed uniformly with the same protocol. These 2 points are important, because prognosis changes with different regimens, and prognostic indicators established with less-effective treatment protocols might not be applicable when more active regimens are used. Furthermore, prognostic indicators could be meaningfully established only when a uniform protocol is used. Finally, our study was prospective, so that PET scans were obtained at predetermined time points during treatment. Results of the interim scan did not affect the planned treatment, so that results of the end-treatment scan were fully interpretable. Although the number of patients involved was relatively small and thus a limitation of the study, this is a rare tumor, even in countries in which it is considered to be more prevalent.

As previously noted, SMILE appears to date to be the most active regimen for treatment of ENKTL (1). However, there is still a paucity of data for risk factors that may predict disease relapse after treatment with SMILE. Prognostication is especially important for early-stage ENKTL, because some patients may be curable with radiotherapy and conventional chemotherapy, whereas others may require additional or innovative therapy. Although there is no evidence to date that altering therapy based on interim PET findings will improve patient outcome, our findings support the use of ¹⁸F-FDG PET as a tool for early midtreatment response assessment in clinical trials of ENKTL to evaluate its usefulness in prediction of response and future relapses when other novel regimens are tested.