To the Editor:

We read with great interest the report by Moskowitz et al¹ describing the remarkably low positive predictive value of interim fluorodeoxyglucose–positron emission tomography (FDG-PET) in predicting outcome after four cycles of risk-adapted dose-dense immunochemotherapy for advanced diffuse large B-cell lymphoma (DLBCL). Although we essentially agree with the conclusion of the authors that biopsy confirmation of an abnormal interim PET should be performed before changing therapy because of the ongoing definition of interim PET interpretation criteria,² we would like to note that the biopsy approach itself suffered from relatively low yields for detection of residual disease and prediction of relapse.¹ Five (13%) of 38 patients for whom interim PET was positive who underwent biopsy were biopsy positive, and of these patients, two died as a result of their disease. However, 10 DLBCL-related events, including five disease-related deaths, were observed in these 38 patients. Thus, biopsy predicted DLBCL-related events or deaths in only five (50%) of 10 and two (40%) of five patients, respectively. This fact, combined with the relatively low overall yield of biopsy of only 13%, leads to the question of whether biopsy should indeed be performed in all patients for whom interim PET is positive or reserved for those who are more likely than the average patient to have a positive biopsy.

As Moskowitz et al¹ point out, unlike in other studies, no meaningful standardized uptake value (SUV) or δ SUV cutoff could be found that would predict outcome of treated patients with reasonable accuracy.^1–3 This finding is probably related to the significant overlap between FDG uptake associated with active disease and that associated with inflammatory post-therapy changes, which in our opinion has been augmented by the dose-dense therapy and probably high-disease bulk at diagnosis in most patients included in the study.^1,4–6 Nevertheless, it may be reasonable to attempt to investigate SUV cutoffs, which could be used to identify those patients more likely to have a negative rather than positive biopsy, sparing them this invasive procedure.

In the patient data reported in Table 2 of the report by Moskowitz et al,¹ we observed that a post-therapy SUV cutoff of 3.5 or less was associated with low likelihood of a positive biopsy; post-therapy SUV values were available in 36 of 38 patients who underwent biopsy. Only one (5.0%) of 20 patients (95% CI, 0.1% to 24.9%) with an SUV of 3.5 or less had a positive biopsy. In contrast, four (25%) of 16 patients (95% CI, 7.3% to 52.4%) with an SUV of 3.5 or more had a positive biopsy. Although this difference is not statistically significant (P = .149 by Fisher's exact test), it is conceivable that the lack of significance is a result of the small sample size (N = 36) in this study and that a higher sample size may have increased statistical power. Unfortunately, the δ SUV computation, which potentially could have reduced false-positive interpretations, could not be reliably investigated in this study. Indeed, the pretherapy SUV data provided in Table 2 were computed retrospectively at the initial site of positive PET post-therapy mass and not at the site of the highest tumor uptake pretherapy; the two have been shown to differ from each other in 18% of patients with DLBCL.⁷

Of course, there is always the possibility that the difference we note is simply a result of random chance; yet biologically, it seems logical that post-therapy inflammatory changes in general exhibit relatively low metabolic activity (ie, generally lower than what is seen at the site of gross residual disease). For example, in the study by Moskowitz et al,¹ only one (20%) of five patients with residual disease by biopsy had an SUV of 3.5 or less at biopsy site, whereas this was the case in 19 (61%) of 31 patients in whom an SUV was computed, without evidence of gross residual disease by biopsy (ie, those in whom only or predominantly post-therapy inflammatory changes resulted in a positive FDG-PET finding). Moreover, 25 (81%) of 31 patients with a negative biopsy had an SUV of 5.0 or less at biopsy site versus two (40%) of five patients with a positive biopsy. Using the cutoff value of 3.5 to determine the need for biopsy would have, in this particular study, spared more than half the patients (20 of 36 patients) for whom interim PET was positive a biopsy with low yield of finding disease (one [5%] of 20 patients).

In summary, we believe that the SUV-based approach may have some merit in determining which patients are unlikely to benefit from biopsy because of the high likelihood of a negative result. We encourage Moskowitz et al¹ to investigate this issue using larger sample sizes of patients to determine the validity of our supposition.

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The author(s) indicated no potential conflicts of interest.