Page 106 - 2020_11-Haematologica-web

A. Conconi et al.
ed for proportions. The c2 test or Fisher exact test was used as appropriate for comparing proportions. Hazard ratios (HR) and their 95% confidence intervals (95% CI) were estimated using a Cox proportional hazard model. Multivariable analysis of clinical prognostic factors (including the international prognostic scores, IPI12 and MALT-IPI7) for OS was performed by Cox regression13 with backward stepwise selection. To identify factors associated with early POD, logistic regression was also performed with back- wards stepwise selection. P-values <0.05 (two-sided test) were considered statistically significant.
Results
Test set
The analyzed population consisted of 401 patients enrolled in the IELSG-19 study, 131 treated with chloram- bucil, 132 with chlorambucil and rituximab and 138 with rituximab; their main clinical features are summarized in Table 1. Estimated hazard curves showed that the peak risk of progression occurred within the first 24 months after diagnosis (Figure 1A). Among these 401 patients, 69 (17%) had early POD, relapsing within 24 months of start- ing treatment. Of the remaining 332 patients, 315 (79%) had no relapse or death during the first 24 months and form the reference group. Relapses were observed later in 64 (20%) patients in the reference group. Nine patients were lost to follow-up and eight patients died without POD within 24 months of starting treatment (Figure 2, left panel).
The median age of the 69 patients with early POD was 62 years (range: 31 to 81 years), 32 (46%) patients were male and 26 patients (38%) had a primary gastric localiza- tion (Table 2).
Early POD was most frequent in patients with Eastern Cooperative Oncology Group performance status >1 (P=0.042) and elevated serum LDH (P=0.002). Patients with early POD were more likely to have high-risk MALT- IPI scores (P=0.005) and high-risk IPI scores (P=0.013) than
the reference group. In contrast, elevated serum β2- microglobulin level, advanced disease stage (III-IV vs. I-II), multiple extranodal sites of involvement, primary site of disease localization (gastric vs. extra-gastric), age at diagno- sis (with either 60 or 70 years cut-off) were not associated with early POD. An unbalanced distribution of patients with early POD was evident across the treatment arms, with early POD occurring more frequently (34/132, 26%) in the single agent rituximab arm and less frequently in the combination treatment arm (13/125, 10%) when com- pared with the standard arm of single agent chlorambucil (22/127, 17%) (c2 test, P=0.006) (Table 2).
In a stepwise logistic regression (including the above- mentioned individual factors predicting early POD at uni- variate analysis: treatment arm, LDH concentration, per- formance status, high-risk IPI score, high-risk MALT-IPI score), only high-risk MALT-IPI score retained statistical significance (P=0.006; odds ratio: 2.39; 95% CI: 1.29-4.45).
The proportion of subjects with early POD was also higher among patients achieving partial remission after first-line therapy than among complete responders (P<0.0001) and, notably, transformation into aggressive histology was detected more frequently in patients with early POD than in the reference group (7/69 vs. 3/315; P<0.0001).
With a median follow-up time of 7.4 years, 18 of 69 patients with early POD died and the OS rates at 5 and 10 years after the risk-defining event were 80% (95% CI: 69- 88%) and 64% (95% CI: 45-78%), respectively, in the early POD group versus 91% (95% CI: 87-94%) and 85% (95% CI: 79-90%), respectively, in the reference group (HR=2.42; 95% CI: 1.35-4.35; log-rank P=0.002) (Figure 3A).
Early POD maintained its predictive power with regards to OS (after a risk-defining event) together with a high-risk MALT-IPI score and age (as a continuous variable) in a step- wise Cox model after controlling for treatment arm, LDH concentration, performance status, disease stage, age, B- symptoms, multiple extranodal sites and high-risk IPI groups (Table 3).
Table 2. The distribution of patients’ characteristics in the test and validation sets according to whether they had early progression of disease or not.
Number of patients
Early POD subset
69
62 years (40-78) 32/37
36 (52%)
3 (4%)
14 (20%)
36 (16%)
26 (38%)
20 (29%)
19 (28%)
22 (32%) 13 (19%) 34 (49%)
6.0 years (5.2-9.4)
Test set (IELSG-19) Reference subset
315
60 years (28-80) 153/162
131 (42%)
2 (1%)
25 (8%)
8 (16%) 136 (43%) 51 (16%)
43 (14%)
105 (33%) 112 (36%) 98 (31%)
8.0 years (5.9-9.9)
P-value 0.667
0.741 0.108 0.042 0.002 0.906 0.403 0.013 0.005
0.006
0.024
Early POD subset
64
64 years (31-87) 24/40
41 (64%)
5 (8%)
18 (31%)
15 (41%)
27 (42%)
30 (48%)
25 (43%)
42 (66%) 11 (17%) 11 (17%)
7.0 years (3.5-10.2)
Validation set Reference subset
160
60 years (23-92) 67/93
68 (43%)
6 (4%)
16 (11%)
39 (41%)
71 (44%)
40 (26%)
28 (19%)
85 (53%) 42 (26%) 33 (21%)
6.9 years (4.3-10.9)
P-value 0.410
0.547 0.004 0.300 <0.001 0.957 0.766 0.001 0.001
0.210 0.569
Median age at diagnosis (IQR) Male/female ratio
Stage III-IV, n (%)
Performance status >1, n (%) Lactate dehydrogenase >UNL, n (%) β2 microglobulin >UNL, n (%) Primary gastric lymphoma, n (%)
IPI, high-intermediate/high risk, n (%) MALT-IPI, high risk, n (%)
First-line treatment, n (%) Chemotherapy Immuno-chemotherapy Immunotherapy
Median follow-up (IQR)
IELSG-19: International Extranodal Lymphoma Study Group-19 study; POD: progression of disease; IQR: interquartile range; LDH: serum lactate dehydrogenase; UNL: upper normal limit;IPI:InternationalPrognosticIndex;MALT-IPI,Mucosa-AssociatedLymphoidTissuelymphomaInternationalPrognosticIndex.P-valuesrefertothecomparisonofproportions in early POD versus reference subsets by a c2 or Fisher exact test, as appropriate.
2594
haematologica | 2020; 105(11)