Higher survival rates outside US
Both PFS and OS were higher in the international versus U.S. cohort. At a median follow-up of 45 months, the PFS rate in the United States was 65%, and the OS rate was 70%.
In the international cohort, after a median follow-up of 52 months, the PFS rate was 75%, and the OS rate was 76%, the investigators found.
Reasons for the differences may be because of differences in treatment regimens, socioeconomic and racial disparities in the United States versus other countries, or to decentralized Burkitt lymphoma therapy in the United States, Dr. Olszewski said.
In univariate analysis, factors significantly predictive of worse PFS included age 40 years or older, ECOG performance status 2 or greater, stage 3 or 4 disease, marrow involvement, CNS involvement, LDH more than three times the upper limit of normal, and hemoglobin <11.5 g/dL (P < .001 for all preceding), as well as albumin <3.5 g/dL (P = .001).
“However, the multivariable analysis was more complicated, because many of these factors were overlapping, and most patients with high LDH also had advanced disease, and this group also encompassed patients who had bone marrow and CNS involvement,” he said.
Using the two types of regression analysis mentioned before, investigators identified ECOG performance status 2 or greater (P = .001), age 40 and older (P = .005), LDH greater than three times the upper limit of normal (P < .001) and CNS involvement (P = .002) as significant predictors for worse outcomes in multivariable analysis, and were included in the final model.
“We initially had five groups according to the number of these factors, but we observed that the survival curves for patients with two, three, or four factors were overlapping, and not significantly different, so ultimately we had three risk groups. In the derivation (U.S.) cohort, patients in the low-risk group, with no risk factors, a 3-year PFS of 92%, compared with 72% for patients with one risk factor (intermediate risk), and 53% for patients with two to four risk factors (high risk).
Respective hazard ratios for worse PFS in the low-, intermediate-, and high-risk groups were 1 (reference), 4.15 (95% confidence interval, 1.99-8.68), and 8.83 (95% CI, 4.32-18.03).
Respective HR for worse OS was 1, 7.06 (95% CI, 2.55-19.53), and 15.12 (95% CI, 5.58-40.99).
There were no significant differences in either PFS or OS when either LDH or stage was added into the model.
The BL-IPI was prognostic for PFS and OS in all subgroups, including HIV-positive or -negative patients, those with MYC rearrangements, stage 1 or 2 versus stage 3 or 4, or those treated with rituximab versus those who were not.
As noted before, 3-year PFS rates in the validation cohort for low, intermediate, high-risk groups were 96%, 82%, and 63% respectively, and 3-year OS rates were 99%, 85%, and 64%.
Why the CNS discrepancy?
In the question and answer session following the presentation, comoderator Christopher J. Melani, MD, from the Lymphoid Malignancies Branch at the National Cancer Institute in Bethesda, Md., said that “it was interesting to see the difference between CNS involvement in both the U.S. and the international cohort,” and asked whether Dr. Olszweski could elaborate on whether baseline CNS involvement was assessed by contrast-enhanced MRI of flow cytometry studies of cerebrospinal fluid.
“Could some of these differences between the U.S. and the international cohort be from the baseline assessment differing between the two?” he asked.
Dr. Olszewski replied that the retrospective nature of the data precluded capturing those data, but added that “I do suspect there may be some differences in the way that central nervous system is staged in different countries. In the United States the use of flow cytometry is more commonly employed, but we don’t know how it is used internationally. We do not know how often this is staged radiographically.”
Asked by others who viewed the presentation whether extranodal disease or peripheral blood involvement were prognostic in the final model, Dr. Olszewski replied that “one has to understand that, when one constructs a prognostic index, there is a balance between trying to input as much information as possible and to create something that is useful, clinically meaningful, and accurate.”
He said that, despite trying different models with different factors, “we couldn’t get the discrimination to be much better than the basic model that we ultimately created, so we favored using a more parsimonious model.”
No study funding source was reported. Dr. Olszewski reported research funding from Spectrum Pharmaceuticals, Genentech, TG Therapeutics, and Adaptive Biotechnologies. Dr. Melani reported having no relevant conflicts of interest.
SOURCE: Olszewski AJ et al. ASH 2020, Abstract 705.