Second Primary Malignancies in Diffuse Large B-cell Lymphoma Survivors with 40 Years of Follow Up: Influence of Chemotherapy and Radiation Therapy

Abstract

Purpose

Previous studies have shown an increased risk of second primary malignancies (SPMs) in survivors of diffuse large B-cell lymphoma (DLBCL). Survivors live longer due to the intensification of and improvements in therapy; thus, we aimed to characterize SPM patterns in patients with DLBCL by treatment modality.

Methods and Materials

Standardized incidence ratio and absolute excess risk of SPMs were assessed in patients with primary DLBCL from 1975 to 2016 in the National Cancer Institute's Surveillance, Epidemiology, and End Results Program. A subgroup analyses based on, sex, race, age at the time of diagnosis, latency, and treatment modality were performed. Propensity score-adjusted cumulative incidence curves were generated, stratified by treatment and accounting for death as a competing risk.

Results

In total, 45,946 patients with DLBCL were identified with a mean follow up of 70 months. Overall, 9.2% of patients developed an SPM with a standardized incidence ratio of 1.23 (95% confidence interval, 1.20-1.27). There was no difference in SPM risk between men and women or Black and White patients. Patients age <25 years were particularly susceptible to the development of SPMs, with a risk 2.5 times greater than patients aged 50 to 74 years. Temporal patterns showed increasing risk of solid malignancies and decreasing risk of hematologic malignancies over time, with bladder cancer posing the greatest absolute excess risk of any cancer type after 15 years. Patients treated with radiation therapy (RT), chemotherapy (CT), and chemoradiation therapy (CRT) all had an increased risk of SPM development compared with the general population. The cumulative incidence of SPMs was the lowest in patients treated with RT and the highest when treated with CRT. In the modern treatment era, the cumulative incidence of SPM for patients treated with CT versus CRT was not significantly different.

Conclusions

In this large population-based study, we demonstrate unique SPM risk patterns based on age, latency, and treatment modality that have important implications for the treatment and screening of patients diagnosed with DLBCL.

Introduction

An estimated 80,470 people will be diagnosed with non-Hodgkin lymphoma (NHL) in the United States in the year 2022.¹ Survivors of NHL are at an increased risk of developing second cancers, known as second primary malignancies (SPMs), and may be at a greater risk of developing SPMs than survivors of other cancer types.2, 3, 4 The long-term management of these patients can be challenging, because surveillance strategies are not well defined.

Survivors of diffuse large B-cell lymphoma (DLBCL), the most common subtype of NHL, are susceptible to this increased risk of SPMs.^5,6 A variety of risk factors have been implicated, including radiation therapy (RT), chemotherapy (CT), the use of anti-CD20 monoclonal antibodies (ie, rituximab), immunosuppression, demographics, and genetics.7, 8, 9, 10 With modern treatment advances, survivorship among patients with DLBCL has greatly improved. The 5-year disease-specific survival rate has increased from 43% to 63% since 1995.¹¹ Given these improvements in survival, understanding patient and disease characteristics that predict the increased risk of SPMs has significant clinical implications. Current guidelines and recommendations for SPM screening are limited for DLBCL survivors.¹² Understanding the patterns of SPM development is crucial to guide cancer long-term follow up and surveillance.

To update the SPM risk with modern DLBCL treatments and survival outcomes, we used the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) program's database with updated diagnosis years from 1975 to 2016.

Methods and Materials

The SEER 9 Regs Custom Data for Standardized Incidence Ratio (SIR) was used to identify patients diagnosed with a first primary DLBCL (lymphoma subtype recode/World Health Organization 2008: 2(a)2.3 Diffuse large B-cell lymphoma) between 1975 to 2016. The SEER*Stat Multiple Primary-SIR tool (version 8.3.8) was used to calculate the SIR and absolute excess risk (AER). The SIR compares the number of SPMs in DLBCL survivors with the incidence rates of cancers for a matched U.S. population, and is expressed as the ratio of observed (O) to expected (E) events (O/E). The AER gives the absolute number of excess events that occurred per 10,000 person years among DLBCL survivors compared with a matched U.S population, taking into account age, sex, race, patient years at risk, and the year of DLBCL diagnosis. SPM was defined as a second malignancy occurring at least 6 months after the first primary DLBCL diagnosis. Only invasive malignances were included. Nonmelanoma skin cancers were excluded. Patients who were recorded to have developed DLBCL as a second malignancy were excluded.

The risk of SPMs was evaluated in DLBCL survivors as a whole, as well as stratified by sex, race, stage (including early [I-II] and advanced [III-IV] stage) and SPM subtype. A subgroup analysis was performed to evaluate the relationships between SPM risk and age at the time of DLBCL diagnosis, time since DLBCL diagnosis in 5-year increments (latency), and treatment type, including RT alone, CT alone, and chemoradiation therapy (CRT). We compared subgroups to test for a heterogenous effect (eg, whether each subgroup has a similar increased risk relative to their respective endemic populations). Statistical tests and 95% confidence intervals (CIs) were based on the assumption that the observed number of SPMs was distributed as a Poisson variable. The O/E ratio was deemed statistically significant when the 95% CI did not cross 1. When comparing 2 subgroups, statistical significance was determined on the basis of 95% CIs overlapping or not, which is a conservative assessment of significance when comparing intervals from 2 different populations.^13,14 Information regarding radiation dose, sites to which RT was directed, and chemotherapy regimens are not available currently in the SEER database, and could not be included in the analysis.

An additional analysis was performed to compare incidence of SPMs between treatment cohorts after propensity score adjustment. For years with SEER data on Ann Arbor staging (1983-2015), we estimated the propensity score among treatment arms using a multinomial logistic regression and calculated matching weights.¹⁵ Balancing covariates included age at the time of diagnosis, race, year of diagnosis, sex, Ann Arbor staging, and health service area. Time from DLBCL diagnosis to the earliest of secondary malignancy, death (competing risk), or end of follow up (censoring) was determined. Cumulative incidence curves of secondary malignancies were generated using the Aalen–Johansen estimator.¹⁶

Differences between arms were tested by bootstrap sampling, recreating the propensity score, and calculating the difference in mean time of secondary malignancies over 30 years. This analytical strategy was used over the Fine-and-Gray model, because the latter requires strong proportional hazard and correct mean model specification assumptions and the sum of cumulative incidence rates can exceed 100% of the data, especially when there is a long follow up.^17,18 Additional stratification in the modern treatment era, in which rituximab and intensity modulated RT have become widespread, was performed. The year 2001 was chosen to delineate between the prerituximab and modern treatment era, as previously used.^5,7

The analyses were performed in R, using the VGAM package to calculate multigroup matching weights (per appendix¹⁵), and the survival package to calculate the 30-year mean time of secondary malignancies.^15,19

Results

Patient characteristics

A total of 45,946 patients with DLBCL diagnosed between 1975 and 2016 were identified, representing 268,493 patient years at risk. Within our cohort, the mean age at the time of diagnosis was 62 years (Table 1), with slightly more male (54%) than female patients and most of White ethnicity (85%). A slight majority of patients had early stage disease (53%). The majority of patients were treated with CT (66%), 26% with CRT, and 8% with RT.

Table 1.

Patient characteristics

	All patients, n (%)	Chemotherapy, n (%)	Chemoradiation therapy, n (%)	Radiation therapy, n (%)
Total patients	45,946	25,997	10,017	3194
Patient years at risk	268,493	140,060	71,295	17,911
Patients with second primary malignancy	4247 (9.2)	2275 (8.8)	1071 (10.7)	305 (9.5)
Mean age, DLBCL diagnosis, y	62	61.6	58.2	67.9
Mean age, second primary malignancy diagnosis, y	70	70	68.8	71.4
Mean follow up, mo	70.1	64.7	85.4	67.3
Sex
Male	24,876 (54.1)	14,385 (55.3)	5540 (55.3)	1535 (48.0)
Female	21,070 (45.9)	11,612 (44.7)	4477 (44.7)	1659 (52.0)
Race
White	38,806 (84.5)	21,833 (84.0)	8451 (84.4)	2767 (86.6)
Black	3169 (6.9)	1944 (7.5)	578 (5.8)	183 (5.7)
Other	3808 (8.3)	2129 (8.2)	961 (9.6)	239 (7.5)
Unknown	163 (.4)	91 (.4)	27 (.3)	5 (.2)
Stage
Early stage*	20,990 (53.3)	9685 (42.2)	6182 (69.6)	1692 (78.4)
Advanced stage†	18,386 (46.7)	13,264 (57.8)	2703 (30.4)	465 (21.6)

DLBCL, diffuse large B-cell lymphoma.

^⁎Stage I and II DLBCL.

^†Stage III and IV DLBCL.

Incidence of SPM

Overall, 4247 patients developed an SPM, which accounted for 9.2% of the study cohort. Some patients developed >1 SPM and, in total, 4896 SPMs were observed. SPM development was significantly higher in the DLBCL cohort compared with the general U.S. population (O/E: 1.23; 95% CI, 1.20-1.27; AER: 32.05). DLBCL survivors were at an increased risk of developing head and neck, stomach, colon, anal, hepatobiliary, lung, bone and soft tissue, bladder, kidney, and thyroid cancers, as well as Hodgkin lymphoma, NHL (excluding DLBCL), acute lymphocytic leukemia, acute myeloid leukemia (AML), chronic myeloid leukemia, and Kaposi sarcoma (Table 2). Of these, Kaposi sarcoma (O/E: 10.83; 95% CI, 8.01-14.31), Hodgkin lymphoma (O/E: 7.34; 95% CI, 5.78-9.19), AML (O/E: 4.66; 95% CI, 3.95-5.46), and acute lymphocytic leukemia (O/E: 3.33; 95% CI, 1.66-5.95) had the highest O/E ratio. NHL (AER: 6.75), AML (AER: 4.20), lung (AER: 3.23), head and neck (AER: 3.14), Hodgkin lymphoma (AER: 2.29), and bladder cancers (AER: 2.04) demonstrated the highest population-level risk. There were decreased rates of prostate cancer and chronic lymphocytic leukemia.

Table 2.

Observed-to-expected ratio and AER for second primary malignancy subtypes in patients with diffuse large B-cell lymphoma

				95% confidence interval
Malignancy/tumor site	Observed cases, n	AER*	Observed-to-expected ratio	Lower	Upper
All sites	4896	32.05	1.23†	1.20	1.27
Head and neck	222	3.14	1.68†	1.47	1.92
Esophagus	53	0.28	1.18	0.89	1.55
Stomach	91	0.67	1.27†	1.02	1.56
Colon, excluding rectum	372	01.59	1.14†	1.03	1.26
Rectum	96	–0.65	0.84	0.68	1.02
Anus and anal canal	32	0.71	2.77†	1.90	3.92
Hepatobiliary	113	0.92	1.30†	1.07	1.57
Pancreas	119	0.00	1.00	0.83	1.20
Lung	681	3.23	1.16†	1.07	1.25
Sarcoma	42	0.60	1.70†	1.22	2.29
Melanoma	186	0.85	1.15	0.99	1.33
Breast	439	–0.49	0.97	0.88	1.06
Gynecologic	180	0.09	1.01	0.87	1.17
Prostate	632	–1.98	0.92†	0.85	0.99
Bladder	291	2.04	1.25†	1.11	1.40
Kidney and renal pelvis	173	2.03	1.51†	1.29	1.75
Brain	41	0.04	1.03	0.74	1.40
Thyroid	111	2.30	2.46†	2.02	2.96
Hodgkin lymphoma	76	2.29	7.34†	5.78	9.19
Non-Hodgkin lymphoma	363	6.75	2.14†	1.92	2.37
Acute lymphocytic leukemia	11	0.27	3.33†	1.66	5.95
Chronic lymphocytic leukemia	25	–0.97	0.47†	0.31	0.70
Acute myeloid leukemia	153	4.20	4.66†	3.95	5.46
Acute monocytic leukemia	15	0.46	8.40†	4.70	13.86
Chronic myeloid leukemia	31	0.55	2.03†	1.38	2.89
Kaposi sarcoma	49	1.55	10.83†	8.01	14.31

AER, absolute excess risk.

^⁎Number of cases per 10,000 person – years.

^†P < .05.

Risk factors for SPM development

Comparison of baseline demographics showed no difference in SPM development between male (O/E: 1.23; 95% CI, 1.19-1.28) and female (O/E: 1.23; 95% CI, 1.18-1.28) patients or between White (O/E: 1.22; 95% CI, 1.19-1.26) and Black (O/E: 1.19; 95% CI, 1.04-1.35) patients. However, patients with their race labeled as other were more likely than White patients to develop SPMs (O/E: 1.44; 95% CI, 1.27-1.40).

We also evaluated SPM risk by DLBCL stage at the time of diagnosis. Patients with advanced-stage disease (stage III and IV) had an increased risk of SPM (O/E: 1.33; 95% CI, 1.27-1.40) compared with those with early stage disease (stage I and II; O/E: 1.18; 95% CI, 1.13-1.22). Age at the time of diagnosis was also a significant predictor for increased SPM risk. Patients age <25 years at the time of DLBCL diagnosis were 3 times more likely to develop an SPM than the general population (O/E: 2.99; 95% CI, 2.29-3.84). The increased risk remained statistically significant for patients age 25 to 49 years (O/E: 1.76; 95% CI, 1.64-1.88) and 50 to 74 years (O/E: 1.18; 95% CI, 1.14-1.23) compared with the general population. The increased SPM risk in patients age <25 years was particularly pronounced compared with their older counterparts for the development of AML, Hodgkin lymphoma, sarcomas, and breast, lung, hepatobiliary, stomach, and thyroid cancer SPMs (Fig. 1). O/E ratios and AERs for all SPM types by age at the time of diagnosis are shown in Supplemental Table 1.

Fig. 1.

Observed-to-expected ratios for various secondary primary malignancies based on age at time of diffuse large B-cell lymphoma diagnosis.

SPM risk by treatment modality

We compared SPM risk by treatment modality, and included patients treated with RT alone, CT alone, and CRT. Regardless of treatment type, DLBCL survivors were significantly more likely to develop SPMs compared with their matched counterparts in the general population. This was true for patients treated with RT (O/E: 1.15; 95% CI, 1.03-1.28), CT (O/E: 1.25; 95% CI, 1.20-1.30), and CRT (O/E: 1.30; 95% CI, 1.23-1.38). In all treatment groups, patients were at an increased risk of multiple SPM subtypes (Supplemental Table 2).

A few notable differences of SPM development between treatment modality were found. There was a significantly increased risk of breast cancer in patients treated with CRT versus CT (O/E: 1.32; 95% CI, 1.12-1.56 vs O/E: 0.89; 95% CI, 0.77-1.01, respectively). Patients treated with CT had a higher risk of developing any leukemia than those who did not receive CT (O/E: 2.33; 95% CI, 2.03-2.67 vs O/E: 1.40; 95% CI, 0.98-1.93, respectively). Patients treated with CT alone were less likely than their population-matched counterparts to develop prostate cancer (O/E: 0.83; 95% CI, 0.74-0.93). O/E ratios for malignancies where screening may be performed, stratified by treatment with CT, RT, and CRT, are shown in Figure E1.

The propensity score-adjusted cumulative incidence of SPMs showed a significantly higher incidence of SPMs among patients receiving CRT than patients receiving CT (P = .001), which, in turn, was higher than patients receiving RT (P < .001; Fig. 2). Measured baseline covariates were well balanced between the arms (standardized mean difference ≤0.12 for all covariates). Associated cumulative incidence of death is shown in Figure E2. The cumulative incidence of SPMs in the prerituximab era (1983-2000) showed a significantly higher incidence in patients receiving CRT than patients receiving CT (P < .001) or RT (P < .001), but no difference between CT and RT (P = .42; Fig. 3).

Fig. 2.

Propensity score-adjusted cumulative incidence of second primary malignancies, stratified by treatment modality. Sample size is reweighted from propensity score-matched weights.

Fig. 3.

Propensity score-adjusted cumulative incidence of second primary malignancies, stratified by treatment modality in the A, prerituximab era (1983-2000) and B, modern treatment era (2001-2015).

In the modern treatment era (2001-2015), there was no longer a difference in the cumulative incidence of SPMs between patients treated with CRT versus CT (P = .355); however, there was a significant difference between CRT and RT (P = .015) and a trend toward significance between CT and RT (P = .059). In the modern treatment era, the cumulative incidence of SPMs was numerically higher in all 3 treatment cohorts compared with the prerituximab era.

SPM risk by latency

Temporal patterns of SPM development were analyzed for both solid and hematologic malignancies. The mean interval from DLBCL diagnosis to development of an SPM was 8 years. There was a significantly elevated risk of solid tumor development at all timepoints (0-4, 5-9, 10-14, 15-19, ≥20 years); however, the increased risk was the greatest after the first decade. For leukemia, there was a significantly increased risk during the first decade, which decreased to nonstatistically significant levels thereafter.

Solid and hematologic malignancies with notably increased O/E ratios and AERs over time are shown in Figure 4. Sarcomas, head and neck cancer, and bladder cancer had the highest O/E ratios of any solid malignancies and an upward trend over time. Breast cancer had a statistically decreased O/E ratio during the first 5 years, which subsequently increased over time. Within the first 15 years from diagnosis, lung cancer had the highest AER of any solid malignancy, but bladder cancer had the highest AER after 15 years. When evaluating hematologic malignancies, Hodgkin lymphoma had the highest O/E ratio of any SPM, followed by AML at all timepoints. O/E ratios and AERs for all SPMs by latency are shown in Supplemental Table 3.

Fig. 4.

Observed-to-expected ratios and absolute excess risks for various A, B, solid malignancies and C, D, hematologic malignancies. Circles indicate ratios in 5-year intervals. Solid circles indicate statistically significant values with P < .05 compared with matched general population. Non-Hodgkin lymphoma is excluding diffuse large B-cell lymphoma.

Discussion

This is the largest population-based study evaluating SPM risk in DLBCL survivors with the longest follow up to date. We identified unique SPM risk patterns based on age at the time of diagnosis, latency, and treatment modality. We demonstrated the first-ever cumulative incidence curves stratified by treatment modality in both the prerituximab and modern treatment eras. These findings have important implications for both the treatment and surveillance of these patients.

Age at the time of diagnosis significantly affected the overall risk of SPM development, as well as the types of SPMs for which these patients were at risk. Patients diagnosed age <25 years demonstrated the highest risk of developing an SPM, followed by patients age 26 to 49 years and then patients 50 to 74 years. Prior studies, including the childhood cancer survivor study, have shown that survivors of childhood cancer are at a 2- to 6-fold increased risk of developing SPMs.20, 21, 22, 23, 24, 25 Women treated as children and young adults with more historic thoracic RT techniques have been shown previously to have a cumulative incidence of breast cancer between 13% to 20% by age 40 years, which later led to the discovery that early breast magnetic resonance imaging screening improves survival in these patients.^12,26

Furthermore, young patients treated for NHL specifically have been shown to be at a significantly higher risk of hematologic cancers after treatment.^20,27 This may be related to the fact that children and young adults with cancer are more likely to have genetic predispositions that may be present in up to 10% of pediatric patients based on recent reports using genome-scale germline sequencing of pediatric cancer cohorts.28, 29, 30, 31 Additionally, environmental factors, such as smoking and obesity, may compound risks from treatment.^32,33 Regardless of the etiology, patients diagnosed with DLBCL at a younger age clearly have an increased SPM risk compared with their older counterparts.

The last few decades have seen a DLBCL transition toward omission of RT, with increasing frequency of CT-only treatment approaches.^7,34 In some instances, the rationale for this shift in practice patterns is the concern for RT-induced SPM. A previous study using the California Cancer Registry showed an increased risk of SPM in patients with DLBCL with the use of RT; however, this risk did not persist when only evaluating patients treated since 2001.⁷ Studies using the SEER database have been mixed in patient with DLBCL, with one study showing no increased risk of SPM development in patients treated with RT and another showing an increased risk of specific cancer types in patients treated with RT.^5,6 In our evaluation, treatment with RT, CT, and CRT were all associated with an increased risk of SPM development compared with the matched general population.

Our propensity score-matched analysis comparing treatment modalities demonstrated that the cumulative incidence of SPMs was the lowest in patients treated with RT alone, followed by CT and then CRT. Among patients treated in the modern era, the cumulative incidence of SPMs remained the lowest in the RT cohort with no statistically significant difference between the CRT and CT cohorts. Possibly, advances in radiation oncology and the adoption of involved-site RT with associated improvements in conformality, smaller treatment fields, and lower doses have contributed to a relative decrease in the risk of SPMs observed in patients who receive RT.

These results may help alleviate concerns regarding SPM risk in patients receiving chemotherapy who could otherwise benefit from the addition of RT. The notable exception, which has been demonstrated in our data as well as multiple other studies, is the increased risk of breast cancer in patients treated with radiation.^3,5,35 The numerically higher cumulative incidence of SPMs among all treatment modalities in the modern treatment era compared with the prerituximab era is likely multifactorial, and may be due to changes in treatment practice over time and the addition of rituximab. However, given the relative increase among patients in all treatment cohorts, the most likely explanation is increased detection through improvements in imaging and cancer screening protocols.

Temporal patterns of SPM development varied between solid and hematologic malignancies. SPMs of multiple solid malignancy types, including head and neck, lung, sarcoma, bladder, and breast cancer, increased over time, and is consistent with the previous literature evaluating radiation-induced SPM.^36,37 Chemotherapy has also been shown to increase the risk of solid malignancies, which is typically seen >10 years after exposure.³⁶ Hematologic malignancies tend to occur a few to several years after treatment, with leukemia peaking 5 to 7 years after treatment,³⁸ which was also supported by our data. These temporal patterns can help guide various screening recommendations for different malignancy types.

One of the most striking findings in our data was the drastic and significantly delayed risk of development of bladder cancer for DLBCL survivors. Cyclophosphamide has been well established to be a significant risk factor for the development of bladder cancer in a dose-dependent manner.^39,40 A recent systematic review evaluated 285 patients from 121 studies who had received prior cyclophosphamide and later developed urothelial carcinoma. Among these patients, a median latency of 10 years from initiation of cyclophosphamide to the development of bladder cancer was shown, but a significant number of patients developed bladder cancer with a latency >15 years. ⁴⁰ In this study, bladder cancer had the highest AER of any cancer type beyond 15 years, and accounted for more cases than both breast and lung cancers combined.

This highlights the significant risk cyclophosphamide poses in DLBCL survivors, and should be a consideration in patients with limited-stage DLBCL, because radiation can be used to decrease the number of cycles of rituximab, cyclophosphamide, doxorubicin hydrochloride (hydroxydaunomycin), vincristine sulfate (oncovin), and prednisone combination therapy given. Additionally, there are no screening recommendations for patients with DLBCL or those exposed to cyclophosphamide. Yearly screening via urinalysis and a lower threshold for imaging for patients who develop hematuria and irritative urinary symptoms have been suggested.⁴⁰ Although data are extremely limited regarding the utility and cost effectiveness of these measures, our data highlight the importance of future studies in bladder cancer screening among patients with DLBCL.

This study has inherent limitations due to the retrospective nature of a database analysis. SEER does not collect or include information on comorbid conditions, specific chemotherapy regimens, RT dose, RT technique, location of treatment, or other cancer risk factors (eg, smoking). With regard to radiation specifics, both dose and radiation techniques have been associated with a higher risk of SPM development.³⁶ We would expect this same association to exist in DLBCL; however, we were unable to explore these associations, because this information is unavailable in the SEER database. In addition, although the SEER database has been shown to have an excellent positive predictive value for patients coded as having received RT or CT, the database also has imperfect sensitivity for patients coded as not having received CT or RT. This suggests that some treatment fails to be captured.⁴¹ Finally, we defined statistical significance as P < .05, allowing for a false discovery rate >5%. Focused, affirmative, follow-up studies would strengthen the conclusion of each finding.

Despite the aforementioned shortcomings, the strength of this study lies in the analysis of a very large cohort of patients with sufficient size and statistical power to identify patterns of SPM risk, as well the long follow up necessary to fully evaluate SPM risk over time.

Conclusions

This is the largest study to examine SPM risk in patients treated for DLBCL with >40 years of follow up. Patients age <25 years were found to be particularly susceptible to the development of SPMs, with a risk 2.5 times greater than patients age 50 to 74 years. Temporal patterns showed an increasing risk of solid malignancies and decreasing risk of hematologic malignancies over time, with bladder cancer posing the greatest absolute excess risk of any cancer type after 15 years. Patients treated with RT, CT, and CRT all had an increased risk of SPM development compared with the general population. The cumulative incidence of SPMs was the lowest in patients treated with RT and the highest in those treated with CRT. In the modern treatment era, the cumulative incidence of SPMs for patients treated with CT versus CRT were not significantly different. Understanding these risks is essential in the management of patients with DLBCL, and should inform both treatment and future survivorship guidelines for patients with DLBCL.