Stem cell transplantation for follicular lymphoma relapsed/refractory after prior rituximab

A comprehensive analysis from the NCCN lymphoma outcomes project




Stem cell transplant (SCT)-related outcomes and prognostication for relapsed/refractory follicular lymphoma (FL) are not well-defined in the post-rituximab era.


Through the National Comprehensive Cancer Network (NCCN) lymphoma outcomes study, 184 patients with relapsed/refractory FL who underwent autologous SCT (autoSCT) or allogenic SCT (alloSCT) following disease relapse after prior rituximab-based therapy were examined.


Patients who underwent autoSCT (N = 136) were older compared with patients who underwent alloSCT (N = 48) (54 versus 51 years, respectively, P = .01) and more frequently had grade 3 FL (35% versus 8%, respectively, P = .006). Patients who underwent alloSCT received more prior therapies (4 versus 3, respectively, P < .0001) and more often had resistant disease at SCT (19% versus 6%, respectively, P = .008). Cumulative 100-day nonrelapse mortality (NRM) for autoSCT and alloSCT were 1% and 6%, respectively (P < .0001), whereas 3-year NRM rates were 3% versus 24%, respectively (P < .0001). For autoSCT and alloSCT, cumulative rates of relapse, progression, and/or transformation were 32% versus 16%, respectively (P = .03), whereas 3-year overall survival rates were 87% versus 61% (P < .0001); there were no differences in failure-free survival. AlloSCT was associated with increased risk of death on multivariate analysis (hazard ratio = 2.77, 95% confidence interval = 1.46-5.26, P = .002). This finding persisted on propensity scoring/matching. Multivariate analysis for autoSCT patients identified age > 60 years and > 3 prior therapies as adverse factors. Furthermore, a survival model was created for the autoSCT cohort based on number of factors present (0, 1, 2); 3-year failure-free survival was 72%, 47%, and 20%, respectively (P = .0003), and 3-year overall survival was 96%, 82%, and 62%, respectively (P < .0001).


AutoSCT remains an effective therapy for patients with FL. For alloSCT, continued strategies to reduce NRM are needed. Cancer 2013;119:3662–3671. © 2013 American Cancer Society.


Autologous stem cell transplantation (autoSCT) and allogeneic SCT (alloSCT) have been shown to be effective therapeutic modalities for relapsed or refractory follicular lymphoma (FL).[1-5] However, most patients in these analyses did not have disease that was relapsed or refractory after prior rituximab-based therapy prior to SCT. Recent data from diffuse large B-cell lymphoma showed that outcomes with autoSCT for relapsed/refractory disease were poor, especially among patients who relapsed after prior rituximab-containing therapy.[6] The impact of prior treatment with rituximab, if any, on the outcomes and prognostication of patients with relapsed/refractory FL who undergo subsequent SCT is not clear.

There has been an increasing use of alloSCT for relapsed/refractory FL, in part due to the use of reduced-intensity conditioning, which allows older and/or less physically fit patients to undergo alloSCT. Early nonrelapse mortality (NRM) is reduced with reduced-intensity alloSCT strategies, although there remains a lack of data studying longer-term outcomes, especially late toxicity and mortality. In addition, there is a paucity of data in the contemporary era comparing autoSCT versus alloSCT in relapsed/refractory FL. There are no available randomized trials comparing these modalities. Moreover, a recent clinical trial attempting to examine this question closed prematurely due to poor accrual.[7]

The National Comprehensive Cancer Network (NCCN) Non-Hodgkin Lymphoma (NHL) Outcomes Database is a prospective cohort study examining comprehensive clinical, treatment, and outcomes data for patients with NHL at 7 NCCN centers. This data set is a unique resource collecting prospective data with relatively large numbers of patients over long periods. The aim of the current analysis was to examine and compare the outcomes of patients with FL with relapsed or refractory disease after prior rituximab-based therapy who underwent either an autoSCT or alloSCT. Furthermore, prognostic factors predictive of survival within each SCT cohort, and across all patients, were investigated. To our knowledge, this is the largest report of SCT for FL relapsed/refractory after prior rituximab treatment.



We identified all cases in the NCCN NHL database of relapsed or refractory FL who had been treated with prior (and relapsed from) rituximab and who received subsequent autoSCT or alloSCT. Patients who received > 1 SCT were excluded. The NCCN database prospectively collects detailed NHL-specific data including patient and disease characteristics, treatment, and outcomes.[8, 9] The intent of the NCCN cohort study is to capture all patients with NHL who were treated at these participating centers. Among all NHL subjects in the database (N = 5395), 1670 had FL, of whom 240 underwent one SCT during the study observation period of January 1, 2001, through December 31, 2009 (follow-up through May 11, 2012). All SCTs were performed at an NCCN center. Quality assurance review was performed for all cases; after quality assurance review, we identified cases that did not meet the predefined study criteria: did not have FL at transplant (ie, transformed lymphoma, N = 22), did not have relapsed/refractory FL (ie, SCT done in first response/remission, N = 13), did not receive prior rituximab (N = 11), patient had undergone > 1 SCT (N = 9), or disease was not FL (N = 1). In total, 184 patients were included in the final analysis. Table 1 depicts the disease status at presentation to the respective NCCN center. Given the observational nature of this cohort study, treatment was not uniform or predetermined; therapy was given at the discretion of the patients' individual treating physicians.

Table 1. Disease Status at Presentation to NCCN Center
Disease StatusType of SCT
at PresentationAutologousAllogeneicTotal
  1. a

    Diagnosed at NCCN center and all treatment (including SCT) received at NCCN center.

  2. b

    Diagnosed at outside center, no prior therapy; all therapy received at NCCN center.

  3. c

    Initial treatment received at outside center; therapy for relapsed disease (including SCT) received at NCCN center.

  4. Abbreviations: SCT, stem cell transplantation; NCCN, National Comprehensive Cancer Network.

Newly diagnoseda (percent)35 (19)14 (8)49 (27)
Previously diagnosed, not treatedb (percent)2 (1)1 (1)3 (2)
Previously diagnosed and treated, SCT at NCCNc (percent)98 (53)34 (18)132 (71)
Totals135 (73)49 (27)184 (100)

Statistical Analysis

Definitions in the database included scoring of comorbidities according to the Charlson comorbidity index, whereas resistant disease (prior to SCT) included patients without partial or complete response to salvage therapy. Relapse after SCT was defined as progressive disease (PD) from remission. Progression was defined as PD after SCT without entering remission. Failure-free survival (FFS) was defined as time from SCT to relapse, transformation, PD, or death. Overall survival (OS) was defined as time in years from SCT to death. Baseline demographic and clinical characteristics were compared between autoSCT and alloSCT cohorts, using chi-square test for categorical variables and Wilcoxon rank-sum for medians. Based on cause of death, cumulative incidence curves were generated to compare treatment-related mortality and nonrelapse mortality between type of SCT. P values were calculated using Gray's test for equality.[10-12]

Univariate Cox proportional hazards regression was used to assess associations of prognostic factors within each SCT cohort for survival. Variables with P < .20 were entered into a multivariate Cox model within each type of SCT. In a similar fashion, Cox regression was used to assess FFS and OS comparing autoSCT versus alloSCT cohorts. To further examine differences in OS, a matched cohort survival analysis using propensity scores was performed.[13] Logistic regression was used to calculate each patient's propensity score, estimating the likelihood of receiving an alloSCT conditioned upon the observed covariates. AlloSCT patients were exactly matched (without replacement) to an autoSCT patient to the 10th decimal place of the propensity score on a 1:1 basis. Cox proportional hazards regression was performed with stratification on the propensity score (rounded to the 10th place) to account for the matching nature of the data.[14] Finally, based on the number of significant prognostic factors identified in the multivariate Cox model for the autoSCT cohort, a survival model was created using classification and regression tree (CART) analysis. The presence of each prognostic factor was given a value of 1 point, and the sum of all factors present was taken as the patient's final score. Kaplan-Meier survival plots were generated by stratification on the final scores. All analyses were performed using SAS version 9.2 software.


Patient Characteristics

There were several differences in patient and disease characteristics between the autoSCT and alloSCT cohorts (Table 2). Patients who received autoSCT were significantly older at time of SCT compared with the alloSCT cohort, whereas alloSCT patients had received a higher number of therapies prior to SCT compared with autoSCT patients. However, the overall use of rituximab prior to SCT appeared similar. For patients who had autoSCT and alloSCT, the mean number of rituximab/chemotherapy regimens were 1.41 (0-3) and 1.47 (0-4), respectively, and the mean number in whom rituximab was used alone were 0.59 (0-2) and 0.90 (0-3), respectively. Immediately prior to SCT, rituximab was a part of salvage therapy in 27% and 45% of autoSCT and alloSCT patients, respectively (P = .03). There were no differences between SCT groups identified for race, sex, insurance, comorbidities, participation in a clinical trial, or performance status. Further, the median time from original FL diagnosis to SCT was 26 months (range, 6-207 months) for patients who underwent autoSCT compared with 36 months (range, 8-178 months) for those who underwent alloSCT (P = .13).

Table 2. Demographics and Clinical Characteristics
 Type of SCT  
 AutologousAllogeneicAll Subjects 
CharacteristicN = 135N = 49N = 184Pa
  1. a

    P value comparing autoSCT vs alloSCT, reported from chi-square analysis on categorical variables and Wilcoxon rank-sum on medians.

  2. b

    Comorbidity scoring via Charlson index/method at initial diagnosis of FL.

  3. c

    Related to the SCT.

  4. Abbreviations: A/PI, Asian/Pacific Islander; ECOG, Eastern Cooperative Oncology Group; FL, follicular lymphoma; NOS, not otherwise specified; PS, performance status; SCT, stem cell transplantation.

Age at SCTMedian, y (range)55 (29-70)50 (27-64)54 (27-70).005
Sex .76
MaleN (%)82 (61)31 (63)113 (61)
FemaleN (%)53 (39)18 (37)71 (39)
Race .65
Caucasian non-HispanicN (%)109 (81)42 (86)151 (82)
HispanicN (%)18 (13)3 (6)21 (12)
BlackN (%)1 (1)1 (2)2 (1)
A/PIN (%)4 (3)2 (4)6 (3)
UnknownN (%)3 (2)1 (2)4 (2)
0N (%)107 (79)37 (76)144 (78)
1N (%)20 (15)10 (20)30 (17)
2N (%)3 (2)1 (2)4 (2)
UnknownN (%)5 (4)1 (2)6 (3)
Comorbidityb .69
0N (%)120 (89)45(92)165 (90)
1N (%)11 (8)4(8)15 (8)
2N (%)3 (2)03 (<2)
3N (%)1 (1)01 (<1)
Enrollment in a clinical trialc    .74
NoN (%)93 (69)35 (71)128 (70)
YesN (%)42 (31)14 (29)56 (30)
FL histology .005
Grade 1/2N (%)78 (58)40 (82)118 (64)
Grade 3N (%)47 (35)4 (8)51 (28)
NOSN (%)10 (7)5 (10)15 (8)
Stage at relapse (prior to SCT)    .29
IN (%)5 (4)3 (6)8 (4)
IIN (%)9 (7)2 (4)11 (6)
IIIN (%)62 (46)16 (33)78 (42)
IVN (%)59 (44)28 (57)87 (48)
Number therapies prior to SCTMedian (range)3 (2-8)4 (2-9)3 (1-9)<.0001
Disease status at SCT .01
SensitiveN (%)124 (92)37 (76)161 (88)
ResistantN (%)8 (6)9 (18)17 (9)
UnknownN (%)3 (2)3 (6)6 (3)

Disease Status

AutoSCT patients were more likely to have grade 3 FL compared with the alloSCT cohort. Analyzed as a continuous variable and grouped (ie, stage I/II versus stage III/IV), there were no differences in disease stage at relapse (prior to SCT) between the 2 cohorts (Table 2). Subjects who underwent alloSCT, however, more commonly had resistant disease at the time of SCT compared with patients who underwent autoSCT.

Transplant Information

A total of 96% of patients had peripheral stem cell collection, whereas 3% had bone marrow harvest, and 1% had combined peripheral blood with marrow. Among patients who underwent alloSCT, 63% had a matched-sibling donor, whereas 37% had a matched-unrelated donor. Eighty-six percent had peripheral blood stem cell collection, 8% were from umbilical cord, and 6% were from bone marrow. Further details regarding donor status (eg, human leukocyte antigen match/mismatch, cytomegalovirus status, and the like) were not available. The conditioning regimens for patients who underwent autoSCT were cyclophosphamide, bis-chloro-ethylnitrosourea (BCNU), and etoposide (CBV)-based in 52%, BCNU, etoposide, cytarabine, and melphalan (BEAM) for 28%, total body irradiation (TBI)-based in 18% (cyclophosphamide/etoposide with TBI most common), and 2% having other regimens. For alloSCT patients, conditioning regimens were fludarabine/melphalan in 31%, TBI-based for 29% (fludarabine-TBI and cyclophosphamide-TBI being most common), busulfan/fludarabine for 23%, and 17% having other regimens. Dosing of conditioning therapy or graft-versus-host disease prophylaxis was not available.


Median follow-up among surviving patients from the date of SCT to last known vital status was 4.0 years with a range of < 1 to 10 years. Table 3 shows outcomes according to type of SCT. The proportion of patients whose first event experienced during follow-up was progression, relapse, and/or transformation was 32% for the autoSCT cohort compared with 17% for patients who underwent alloSCT (P = .03). The only difference among patients who experienced disease progression versus those who did not, besides type of SCT, was older age (ie, median age 56.4 versus 52.4 years, respectively, P = .01). There was no appreciable difference in FFS (Fig. 1A); however, the 3-year OS was superior for the autoSCT cohort compared with alloSCT patients (Fig. 1B).

Table 3. Stem Cell Transplantation (SCT)-Related Outcomes
 Type of SCTAll Subjects
OutcomeN = 135N = 49N = 184
  1. a

    Cell frequencies are not mutually exclusive (ie, a patient may have experienced more than one event and thus contributed to >1 event category).

  2. b

    One autologous SCT subject died from unknown causes and was censored.

Relapse/transformation post-SCTa 93 (69)44 (90)137 (75)
NoN (%)
RelapseN (%)37 (27)4 (8)41 (22)
TransformationN (%)5 (4)1 (2)6 (3)
Progression post-SCTa 124 (92)44 (90)168 (91)
NoN (%)
YesN (%)11 (8)5 (10)16 (9)
Death post-SCT 109 (81)28 (57)137 (75)
NoN (%)
YesN (%)26 (19)21 (43)47 (25)
Death while in remissionb
NoN (%)127 (95)36 (73)157 (79)
YesN (%)7 (5)13 (27)27 (11)
Figure 1.

Nonrelapse mortality and survival are shown. Graph (A) shows the cumulative nonrelapse mortality rates for autologous stem cell transplantation (autoSCT) and allogeneic SCT (alloSCT), and (B) shows the unadjusted Kaplan-Meier failure-free survival (FFS) for autologous SCT versus allogeneic SCT (N = 184). The 3-year FFS for autologous SCT patients was 57% (95% CI = 47%-66%) versus 52% (95% CI = 36%-67%) for allogeneic SCT (FFS log-rank P = .14). (C) shows the unadjusted overall survival (OS) for autologous SCT versus allogeneic SCT (N = 184). The 3-year OS for autologous SCT was 87% (95% CI = 81%-94%) versus 61% (95% CI = 47%-76%) for allogeneic SCT patients (OS log-rank P < .0001).

Causes of Death

Nineteen percent of autoSCT and 43% of alloSCT subjects have died. Among the 26 deaths in the autoSCT cohort, 18 (69%) were due to PD compared with 8 due to PD of 21 (38%) deaths in alloSCT subjects. Interestingly, the 100-day NRM rates for autoSCT and alloSCT patients were 1% (95% confidence interval [CI] = 0.1%-4%) and 6% (95% CI = 2%-16%), respectively (overall P < .0001 Gray's test for equality), whereas the 3-year NRM rates were 3% (95% CI = 1%-8%) for autoSCT versus for 24% (95% CI = 12%-39%) alloSCT (P < .0001 Gray's test for equality) (Fig. 1C). The causes of death by type of SCT and time are detailed in Table 4. Among all patients, there were 3 cases of myelodysplastic syndrome/acute myelogenous leukemia, all which occurred in patients who received autoSCT (incidence 2.2%).

Table 4. Cause and Time to Death.
SCTCause of DeathTime to Death, Days (Month)
  1. a

    Counting from day 0 of SCT to date of death.

  2. Abbreviations: SCT, stem cell transplantation; FL, follicular lymphoma; GVHD, graft versus host disease; GI, gastrointestinal; COPD, chronic obstructive pulmonary disease; MDS, myelodysplastic syndrome.

AllogeneicMultiple organ failure31 (1)
AllogeneicProgressive disease36 (1)
AllogeneicAcute GVHD45 (2)
AllogeneicAcute GVHD/acute renal failure62 (2)
AllogeneicGI hemorrhage153 (5)
AllogeneicProgressive disease167 (6)
AllogeneicChronic GVHD181 (6)
AllogeneicChronic GVHD198 (7)
AllogeneicInterstitial pneumonia202 (7)
AllogeneicProgressive disease226 (8)
AllogeneicSecond malignancy (sarcoma)251 (8)
AllogeneicProgressive disease286 (10)
AllogeneicProgressive disease421 (14)
AllogeneicChronic GVHD458 (15)
AllogeneicProgressive disease483 (16)
AllogeneicProgressive disease525 (18)
AllogeneicPulmonary failure978 (33)
AllogeneicChronic GVHD1726 (58)
AllogeneicPneumonia2232 (74)
AllogeneicSecond malignancy (cervical)2244 (75)
AllogeneicProgressive disease2782 (93)
AutologousPulmonary failure76 (3)
AutologousProgressive disease83 (3)
AutologousProgressive disease88 (3)
AutologousProgressive disease100 (3)
AutologousProgressive disease207 (7)
AutologousProgressive disease212 (7)
AutologousProgressive disease392 (13)
AutologousProgressive disease636 (21)
AutologousProgressive disease657 (22)
AutologousProgressive disease762 (25)
AutologousCOPD939 (31)
AutologousSecond malignancy (MDS)982 (33)
AutologousProgressive disease1037 (35)
AutologousProgressive disease1093 (36)
AutologousEncephalitis1126 (38)
AutologousProgressive disease1238 (41)
AutologousProgressive disease1269 (42)
AutologousProgressive disease1458 (49)
AutologousUnknown1470 (49)
AutologousProgressive disease1779 (59)
AutologousProgressive disease2047 (68)
AutologousSecond malignancy (esophageal)2059 (69)
AutologousSecond malignancy (leukemia)2164 (72)
AutologousProgressive disease2372 (79)
AutologousSecond malignancy (leukemia)2715 (91)
AutologousProgressive disease2953 (98)

Prognostic Factors Within Respective SCT Cohorts

Factors that predicted survival on univariate analysis for patients with relapsed/refractory FL who underwent autoSCT were ages > 60 years and > 3 prior therapies (Table 5). Figure 2 shows survival rates within the autoSCT cohort based on the presence of a single prognostic factor (identified on multivariate analysis). Furthermore, a CART survival model was formed for the autoSCT cohort using these prognostic covariates (Fig. 3A,B). The impact of each adverse risk factor appeared to be approximately equal based on the similar OS and FFS univariate hazard ratios. An increasing number of these 2 independent variables (ie, age ≥ 60 years and > 3 prior therapies) was associated with divergent survival rates as noted in Figure 3A,B.

Table 5. Univariate Analysis for Prognostic Factors Within the Autologous SCT Cohort.
  1. a

    Abbreviations: FFS, failure-free survival; OS, overall survival; N, number; HR, hazard ratio; CI, confidence interval; SCT, stem cell transplantation; FL, follicular lymphoma; NOS, not otherwise specified; BEAM, BCNU, etopoide, ara-c, melphalan; CBV, cyclophosphamide, BCNU, etoposide/VP16; TBI, total body irradiation.

VariableStratumNHR95% CIPHR95% CIP
Age at SCTContinuous, Years1351.091.04,1.150.0011.061.02, 1.100.002
Age at SCT<=6099baseline  baseline  
 >60363.151.40, 7.070.0052.011.16, 3.470.01
GenderFemale53baseline  baseline  
 Male820.610.28, 1.330.210.760.45, 1.300.32
RaceOther26baseline  baseline  
 Caucasian1091.260.43, 3.650.680.580.32, 1.050.07
InsuranceMedicare/Medicaid24baseline  baseline  
 Managed Care1020.650.25, 1.660.360.560.30, 1.040.07
 Other (excluded)9    
FL Histology1/278baseline  baseline  
 3470.690.30, 1.600.390.790.45, 1.390.41
 NOS (excluded)10    
Performance Status0107baseline  baseline  
 1/2231.130.39, 3.320.820.930.44, 1.980.85
 Unknown (excluded)5    
Stage at SCTI/II/III76baseline  baseline  
 IV590.880.40, 1.930.751.040.62, 1.770.88
Number therapies prior to SCTContinuous (1-9)1351.321.05,1.660.021.241.05, 1.460.01
Number therapies prior to SCT1-390baseline  baseline  
 >3453.411.54, 7.560.0032.341.38, 3.960.002
Disease Status at SCTSensitive124baseline  baseline  
 Resistant80.770.18, 3.300.730.840.30, 2.330.74
 Other (excluded)3    
Conditioning RegimenBEAM38baseline  baseline  
 CBV711.930.56, 6.700.371.140.59, 2.180.78
 TBI-based241.090.25, 4.72 0.900.40, 2.02 
 Other (excluded)3    
Figure 2.

Survival is shown within the autologous stem cell transplantation (SCT) cohort based on presence of single prognostic factors. Kaplan-Meier failure-free survival (FFS) and overall survival (OS) were calculated for autologous SCT patients (N = 135) based on age and number of prior therapies. (A) shows the 3-year FFS rates for autologous SCT patients age < 60 years and ≤ 3 prior therapies of 72% (95% confidence interval [CI] = 58%-82%), ≥ 60 years and ≤ 3 prior therapies of 47% (95% CI = 24%-68%), < 60 years and > 3 prior therapies of 45% (95% CI = 25%-63%), and ≥ 60 years and > 3 prior therapies of 20% (95% CI = 3%-47%), FFS log-rank P = .001. (B) shows the 3-year OS rates among SCT patients age < 60 years and ≤ 3 prior therapies of 96% (95% CI = 86%-99%), ≥ 60 years and ≤ 3 prior therapies of 86% (95% CI = 63%-95%), < 60 years and > 3 prior therapies of 80% (95% CI = 58%-91%), and ≥ 60 years and > 3 prior therapies of 62% (95% CI = 27%-84%), OS log-rank P = .0003.

Figure 3.

Survival models are presented. Panels (A) and (B) show the CART survival model formed for autologous stem cell transplantation (SCT) patients using the 2 significant factors identified in multivariate analysis in this cohort: 3-year failure-free survival (FFS) and overall survival (OS) for patients who underwent autologous SCT based on the number of the adverse prognostic factors present (ie, age > 60 years and > 3 prior therapies). An increasing number of adverse factors portended an increasingly poor outcome: 3-year FFS with 0, 1, or 2 factors: 72% (95% confidence interval [CI] = 60%-84%), 47% (95% CI = 32%-61%), and 20% (95% CI = 4%-45%), respectively (FFS log-rank P = .0003); and the associated 3-year OS rates were 96% (95% CI = 91%-101%), 82% (95% CI = 71%-94%), and 62% (95% CI = 32%-92%), respectively (OS log-rank P < .0001). (C) shows the Kaplan-Meier OS figure for the exact matched cohort (on propensity matching analysis) comparing autologous SCT with allogeneic SCT. The 3-year OS on this matched analysis was 85% (95% CI = 73%-97%) for autologous SCT versus 63% (95% CI = 47%-79%) for allogeneic SCT (P = .011).

Within the alloSCT cohort, there were no significant prognostic factors identified on univariate analysis. On multivariate Cox regression analysis, age and number of prior therapies remained significant for FFS and OS for autoSCT patients, whereas age > 50 years and resistant disease emerged as prognostic factors for the alloSCT cohort (Table 6). Time from original FL diagnosis to SCT was not predictive of FFS or OS (data not shown).

Table 6. Multivariate Cox Regression Analysis for Prognostic Factors Within SCT Cohortsa
    FFS model P = .0008 OS model P = .0004
Variable Stratum N HR 95% CI P HR 95% CI P
  1. a

    Final models shown based on model building process of variable elimination and model fit assessment.

  2. Abbreviations: CI, confidence interval; FFS, failure-free survival; HR, hazard ratio; OS, overall survival; SCT, stem cell transplantation.

Autologous SCT (N = 135)
Age at SCT≤60 y99baseline .03baseline .01
>60 y361.881.08, 3.262.861.27, 6.43
Number therapies prior to SCT1-390baseline .003baseline .005
>3452.241.32, 3.803.171.43, 7.05
    FFS model P = 0.02 OS model P = 0.05
Variable Stratum N HR 95% CI P HR 95% CI P
Allogeneic SCT (N = 49)
Age at SCT≤50 y24baseline .02baseline .08
>50 y253.071.16, 8.132.430.89, 6.68
Disease status at SCTSensitive37baseline .08baseline .04
Resistant92.370.90, 6.272.901.05, 7.95
No data3   

Outcomes Comparing AutoSCT Versus AlloSCT

Cox regression multivariate analysis was performed across all patients (N = 184) in order to compare survival for autoSCT versus alloSCT cohorts (Table 7). After controlling for significant competing factors including increasing age (analyzed as continuous and dichotomous variable), resistant disease at time of SCT, and number of prior therapies, alloSCT remained associated with a significantly increased risk of death compared with autoSCT (Table 5). The risk of death remained increased with alloSCT when FL grade 3 patients were excluded (unadjusted hazard ratio [HR] = 2.66, 95% CI = 1.37-5.17, P = .004), which persisted on multivariate analysis (data not shown). It is important to note, however, that this group of patients was small and there were no data delineating grade 3a versus 3b. Patients were also matched exactly with propensity scoring on age, number of prior therapies, and disease resistance at time of SCT. The risk of death with alloSCT remained more than twice that for autoSCT. Figure 3C shows the Kaplan-Meier OS plot for the matched cohort (on propensity analysis) comparing autoSCT with alloSCT. Among the 49 alloSCT patients, 3 had unknown disease status at SCT and were excluded from the matching process, whereas an additional 3 alloSCT patients were unable to have a matching autologous SCT patient found, leaving 86 patients in total (ie, 43 matched pairs) for the final matched propensity analysis.

Table 7. Multivariable Cox Regression for Overall Survival Comparing by Type of SCT (N = 184)a
VariableAll Subjects (N = 184)
StratumHR (OS)95% CI P
  1. a

    Final models shown based on model building process of variable elimination and model fit assessment.

  2. Abbreviations: CI, confidence interval; FFS, failure-free survival; HR, hazard ratio; OS, overall survival; SCT, stem cell transplantation.

Age at SCTContinuous, y1.061.01, 1.10.009
Number of therapies prior to SCT1-3baseline .03
>32.121.09, 4.10
Disease status at SCTSensitivebaseline .19
Resistant1.680.78, 3.67
Type of SCTAutologousbaseline .003
Allogeneic2.701.40, 5.23
 OS model P < .0001


This analysis reports the outcomes and associated prognostic factors among a large cohort of relapsed/refractory FL subjects who underwent autoSCT or alloSCT after relapse from prior rituximab. To our knowledge, this is the largest report of SCT for relapsed/refractory FL in the post-rituximab era. Comparing autoSCT and alloSCT, we noted no difference in FFS; however, we found that the 3-year OS was superior for autoSCT versus alloSCT patients. Moreover, the OS advantage for autoSCT persisted on multivariate Cox regression and propensity scoring analyses. In addition, we identified key prognostic factors within each SCT cohort. In interpreting these observations, several factors should be considered.

Outcomes with autoSCT for relapsed/refractory FL have been reported.[1-3, 5, 15-20] However, these analyses consisted primarily of patients treated in single-institution series, and most were in the pre-rituximab era. Because rituximab is now included as a component of frontline therapy in the vast majority of patients with FL, it is of major interest to examine SCT-related outcomes for relapsed/refractory patients who have progressed after treatment with (failure from) prior rituximab-based therapy. There is currently minimal available data toward this end. A study by Le Gouill et al examined FL subjects in first relapse following treatment on the front-line FL2000 study.[21] They demonstrated excellent outcomes among relapsed FL subjects who underwent autoSCT, with 3-year OS of 92% versus 63% for patients who did not undergo autoSCT; the OS advantage persisted on multivariate analysis. However, of the 42 subjects who underwent autoSCT, only 13 had relapsed from/after rituximab-based therapy. In a retrospective series, Kang et al compared 35 patients with relapsed/refractory FL who were treated and relapsed after prior rituximab with 71 rituximab-naive relapsed patients; both groups received autoSCT.[22] Relapse-free survival and OS were similar. We identified here a 3-year FFS of 56% and 3-year OS of 86% for 135 relapsed/refractory FL patients who received autoSCT, all of whom had relapsed after treatment with prior rituximab. These outcomes appear comparable with prior autoSCT reports, including ones with rituximab-naive populations.

Among recent alloSCT reports for relapsed/refractory FL, the 4- to 5-year OS rates were 57% to 85%, whereas the NRM rates ranged from 15% to 35%[23-25]; 35% to 74% of patients in these series had received rituximab prior to SCT. The 3-year OS for the alloSCT cohort in the current report was 63%. Moreover, a surprising finding in the current data was the superior OS documented for relapsed/refractory FL subjects who received autoSCT compared with those who received alloSCT. The reason for this difference is not known, but several explanations are possible.

Many alloSCT procedures today are based on a reduced intensity platform, which allows older and/or less physically fit patients to undergo this therapy. The conditioning regimens for alloSCT in this study were heterogenous, although most appeared consistent with reduced-intensity conditioning. Reduced-intensity conditioning regimens are typically associated with lower rates of early TRM (ie, 100-day); however, as our data show, there was a later and continual risk of death while in remission for patients who underwent alloSCT. In addition, it may not be possible for nonmyeloablative conditioning regimens to overcome refractory/resistant disease in select cases. Interestingly, resistant disease did not predict FFS or OS within the autoSCT cohort; however, that subset represented a minority of subjects and thus may not have been powered to detect a difference. The higher risk features of alloSCT patients in the current analyses (eg, more prior therapy and resistant disease) likely contributed to the poorer outcome. Yet, the inferior survival for alloSCT persisted after adjustment for these competing risk factors (ie, resistant disease, number of prior therapies, and age) on multivariate Cox regression analysis as well as on exact matching through propensity scoring analysis.

The NCCN Outcomes database is a multicenter research effort prospectively collecting data from 7 participating NCCN centers. These centers are National Cancer Institute–designated Comprehensive Cancer Centers, each with expertise in SCT. Further, a significant strength of this data is all patients diagnosed and treated with NHL at these centers were included in the database regardless of patient/disease characteristics or therapy administered. Despite the shortcomings of this report, such as lack of treatment details and nonuniform therapy, it is able to overcome inherent limitations associated with single-center SCT reports. Furthermore, subjects are tracked prospectively for potential late effects, including nonlymphoma events. We identified several late nonrelapse deaths among alloSCT subjects, including deaths more than 5 years after SCT. It is not clear if the impact of prior therapies received, including rituximab therapy/failure, may adversely affect outcomes associated with alloSCT. Longer term follow-up will be essential in this data set in confirming the OS advantage identified for the autoSCT cohort and whether a plateau will persist on either of the survival curves. It is also important to acknowledge that despite the prospective cohort data collection and the detailed statistical methodology used here, there may still be unintended biases that affected these data.

The data presented here do not address the role of whether SCT should be recommended in relapsed/refractory FL, which is an important consideration. However, when SCT is performed, we identified several critical prognostic factors that predicted survival within the autoSCT and alloSCT cohorts. Furthermore, we created a novel prognostic survival model for patients with relapsed/refractory FL in the post-rituximab era. We also found that autoSCT and alloSCT were associated with comparable FFS. However, including adjustment for competing prognostic factors on multivariate analysis and propensity score analysis, OS was superior among relapsed/refractory FL patients who underwent autoSCT compared with alloSCT. Based on these data, we conclude that autoSCT remains an effective therapy for relapsed/refractory FL in the post-rituximab era. For high-risk patients with FL who are considered for alloSCT, continued strategies to reduce NRM are needed.


No specific funding was disclosed.


Dr. Millenson has received support from a National Comprehensive Cancer Network (NCCN) grant. Dr. Czuczman has received a grant from NCCN Data Outcomes. Dr. Rodriguez has received grants for research funding from Pfizer, GlaxoSmithKline, and Amgen. Dr. Niland has received support from an NCCN contract with her institution for the Data Coordinating Center for the support of responsibility for data analysis for this article. Dr. Zelenetz has acted as a consultant for Genentech and Roche and has received grants from both to support unrelated trials. Dr. Friedberg has acted a member of the Trubion Data Safety Monitoring Committee, the Bayer Data Safety Monitoring Committee, and the Sanofi Advisory Board.